How does the recombinant human interferon
beta induce antibodies in immune tolerant
mice?
Grzegorz Kijanka
Printing of this thesis was financially suported by:
UIPS, Utrecht Institut of Pharmaceutical Sciences.
How does the recombinant human interferon beta induce antibodies in immune tolerant
mice?
PhD thesis with a summary in Dutch
Department of Pharmaceutics, Utrecht University, The Netherlands
ISBN: 978-90-393-6045-3
Chapter 2: Copyright ©2013. From Pharmaceutical Formulation Development of Peptides
and Proteins by Lars Hovgaard, Sven Frokjaer, Marco van de Weert. Reproduced by
permission of Taylor and Francis Group, LLC, a division of Informa plc.
Chapter 1, 3-8: Copyright © 2013 GM Kijanka.
Design of the cover and illustrations for chapters’ title pages:
Magdalena Kijanka
Printed by CPI Wöhrmann Print Service
How does the recombinant human interferon beta
induce antibodies in immune tolerant mice?
Hoe induceert recombinant humaan interferon beta een
antilichaamrespons in immuuntolerante muizen?
(met een samenvatting in het Nederlands)
Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag
van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolge het besluit
van het college voor promoties in het openbaar te verdedigen op maandag 16
december 2013 des middags te 2.30 uur
door
Grzegorz Maciej Kijanka
geboren op 27 februari 1984, te Chrzanow, Polen
Promotor: Prof. dr. H. Schellekens
Co-promotor: Dr. V. Brinks
Studies described in this thesis have been partially funded by F. Hoffmann-La Roche Ltd.,
Basel.
Table of contents.
CHAPTER 1
General introduction.
9
CHAPTER 2
Immunogenicity of therapeutic proteins.
17
CHAPTER 3
Immunogenicity of recombinant human interferon
45
beta in immune tolerant mice is characterized by early
involvement of CD4+ T cells, formation of germinal
centers, but no apparent immunological memory.
CHAPTER 4
The impact of adjuvants and immune suppressants on the
61
production of anti rh interferon beta antibodies in immune
tolerant mice.
CHAPTER 5
Recombinant human interferon beta forms complexes with
79
bacteria - potential risk factor for immunogenicity?
CHAPTER 6
Effect of treatment regimen on the immunogenicity of
97
human interferon beta in immune tolerant mice.
CHAPTER 7
Influence of aggregation and route of injection on the
111
biodistribution of mouse serum albumin.
CHAPTER 8
Summary.
129
APPENDICES
Nederlandse samenvatting.
139
Affiliations of corresponding authors.
149
Curriculum Vitae.
151
List of publications.
153
Acknowledgements.
157
CHAPTER 1.
General introduction.
Chapter 1.
10
Introduction.
1. Interferon beta and multiple sclerosis treatment.
Interferon beta (IFNβ) is an anti-inflammatory cytokine produced in response to viral
infections. It was the first compound found to be effective in MS therapy. The mechanisms
underlying the therapeutic effects of IFNβ in MS patients remains unclear. It was suggested
that IFNβ increases the expression of anti-inflammatory agents and simultaneously down
regulates pro-inflammatory cytokines 1. Moreover, it appears to limit the trafficking of
inflammatory cells through blood-brain barrier 1. Overall IFNβ reduces the number and
strength of relapses, slows the progression of MS and improves patients’ wellbeing 2–8.
In the early 1990s three originator recombinant human IFNβ (rhIFNβ) drugs were introduced
onto the market: Betaferon®/Betaseron® (Schering AG, Germany and Berlex, USA),
Avonex® (Biogen Indec, USA) and Rebif® (Merck Serono, USA) (Table 1). Two types of
rhIFNβ are formulated in those products. RhIFNβ-1b, the drug substance in Betaferon®,
is produced in modified E.coli and therefore lacks posttranslational modifications such
as glycosylation. Moreover, two amino acids of its sequence have been modified when
compared to the endogenously expressed counterpart. The Cys-17 is replaced by Ser-17
and it lacks N-terminal methionine. RhIFNβ-1a, the drug substance of Avonex® and Rebif®,
is produced in Chinese hamster ovarian cells (CHO). It is glycosylated and its sequence is
identical to the endogenously produced IFNβ protein. Due to the mutations and the lack of
glycosylation, rhIFNβ-1b is less stable in solution than rhIFNβ-1a and it is more prone to
aggregation. As a consequence, it is approximately 10 times less potent than rhIFNβ-1a 9.
2. Immunogenicity of rhIFNβ.
RhIFNβ products, as all therapeutic proteins, induce the development of anti drug
antibodies (ADA) in patients. Betaferon® is the most immunogenic product among the
originator products, and induces ADA in up to 90% of treated patients 10–13. The second
Table 1. Differences in formulation, administration regime and immunogenicity of rhIFNβ originator
drugs. Adopted from Harmelling et all 16 and van Beers at all 14.
Drug
Form
Betaseron®/
Betaferon®
Lyophilized
powder
Avonex®
Lyophilized
powder
Rebif®
Ready-touse syringe
Excipients
HSA, di- and monobasic
sodium phosphate,
sodium chloride, final
pH 7.2
HSA, di- and monobasic
sodium phosphate,
sodium chloride, final
pH 7.2
Arginine HCl,
polysorbate 20, sodium
acetate, glacial acetic
acid, final pH 4.8
Mannitol, HSA, sodium
acetate, acetic acid,
sodium hydroxide, final
pH 3.8
Dose
[μg]
Route
Frequency
Binding
ADAs
[%]
Neutralizing
ADAs [%]
250
SC
Every other
day
81-97
28-47
30
IM
Once
weekly
25-89
5-38
30
IM
Once
weekly
10-45
2-14
22 or
44
SC
Three
times per
week
11
Chapter 1.
most immunogenic originator product is Rebif® with 20-40% of ADA positive patients,
and the least immunogenic is Avonex® inducing ADAs in less than 20% of patients 10–13.
Many factors have been suggested to contribute to the high immunogenicity of Betaferon®
including the lack of posttranslational modifications, sequence modifications and the
administration regime (including the highest dose and the highest frequency of injections
among all rhIFNβ products) 12. However, the most potent trigger of ADA responses seems to
be high content of rhIFNβ aggregates in Betaferon® 14. In fact, most of rhIFNβ in Betaferon®
was found to be aggregated, whereas the majority of rhIFNβ in both Rebif® and Avonex® is
in its monomeric form 9,15.
2. Human IFNβ immune tolerant mice.
Transgenic mouse models are powerful tools to study the mechanisms underlying
immunogenicity of rhIFNβ and are an essential in the experiments described in this thesis. The
mice used in this thesis were originally developed by Suzanne Harmeling 16. DNA construct
of hIFNβ gene under control of a mouse interferon beta promoter was microinjected into
fertilized C57Bl/6 ova. The offspring was analyzed for the presence of hIFNβ gene in their
chromosomal DNA by polymerase chain reaction (PCR). Transgenic (tg) mice were shown
to express the hIFNβ protein and to be tolerant for hIFNBβ-1a 16. However, as rhIFNβ-1a
did not elicit an immune response in the non-transgenic (non-Tg) control mice, the model
was modified by cross-breeding with FvB/N mice 17. As a result, hybrid C57Bl/6:FvB/N
heterogeneous transgenic mice were obtained, which showed increased sensitivity to form
ADA against rhIFNβ products. These hybrid transgenic mice therefore allow studying a
variety of product related and patient related factors contributing to immunogenicity of
rhIFNβ, and are suitable to study the immunological processes occurring during the ADA
response.
Tg mice display a number of advantages over conventional mouse models. As they
express human rhIFNβ, the immune system builds up tolerance towards this protein during
a process known as negative selection. Like in patients, the hIFNβ is considered a self
protein, and it is believed that the mechanism and factors underlying immunogenicity in
patients are similar to the mechanisms and factors in the tg mice. As the tg mice have a
fully functional immune system, the myriad of immune cells and processes (e.g. B cells, T
cells, complement etc.) potentially related to immunogenicity of rhIFNβ products can be
studied. An important feature of the tg model is the lack of cross-reactivity between hIFNβ
and mouse IFNβ 18,19. hIFNβ is therefore not biologically active in mice, making it possible
to study immunogenicity of IFNβ independent from its possible interference on the immune
system. As the IFNβ possesses anti-inflammatory activity, injection of hIFNβ might have a
suppressing effect on ADA formation. Moreover, the immune response towards rhIFNβ in
tg mice will not cause neutralization of the endogenous protein, therefore impacting with
the endogenous IFNβ function.
Although tg mice seem to be (one of) the best models for studying immunogenicity,
12
Introduction.
they are not free of drawbacks. The biggest disadvantage of the tg mouse model is the
evolutionary distance between the immune systems of humans and mice. This may result in
misinterpretation of results as the same components of the human and mouse system might
display different activities. For example, mouse immunoglobulins (Ig), the main readout
for immunogenicity in the tg mouse model, are not direct homologues of human Ig 20.
Moreover, as the mouse repertoire of major histocompatibility complex (MHC) molecules
differs from human, it is likely that different epitopes of the hIFNβ sequence will be
recognized and presented on antigen presenting cells 21. Moreover, lack of hIFNβ activity
might be also considered as disadvantage of our model. It has been suggested that activity of
protein might be crucial in formation of ADA. Lack of hIFNβ and mIFNβ cross-reactivity
might results in altered ADA production, e.g. lower pressure to produce of neutralizing
ADA. Therefore translation of results from mouse to human should be done with caution.
Nevertheless the basic mechanisms and immune components involved in immunogenicity
of rhIFNβ in humans and tg mice are likely the same.
3. Aim and the outline of the thesis.
During recent years many improvements have been implemented in the design and
production processes of protein drugs. These improvements increased the quality and safety
of therapeutic proteins. However, despite all those efforts therapeutic drugs, including
rhIFNβ, still induce the production of ADA in patients decreasing treatment efficiency
and increasing therapy costs. The immune processes underlying immunogenicity remain
poorly understood. The main goal of the project described in this thesis was to improve our
understanding of the immune mechanisms leading to ADA formation. For this I have used
the immune tolerant tg mouse model as described above. Simultaneously, further efforts
were implemented to get new insight in factors, e.g. interaction of rhIFNβ with bacteria, that
might contribute to the development of ADAs’ production.
Chapter 2 summarizes the current knowledge about immunogenicity of therapeutic
proteins. It reviews the available literature regarding factors influencing immunogenicity,
possible immunological mechanisms leading to the formation of ADA and strategies used
to prevent immunogenicity.
Several studies designed to improve our knowledge on the immune processes involved in
the ADA response against rhIFNβ are described in Chapter 3 to 7. Chapter 3 first describes
an experiment in which the effect of CD4+T cell removal at different time points during
rhIFNβ treatment in tg mice has been evaluated. Second, the formation of germinal centers
and the presence of an immunological memory response after rhIFNβ administration have
been studied.
The impact of immune modulators on ADA production against rhIFNβ in tg mice is
described in Chapter 4. Agents potentially enhancing (adjuvants) and inhibiting (immune
suppressants) the ADA production have been studied.
13
Chapter 1.
In recent years a number of reports have suggested a correlation between bacterial infections
and the development and progression of MS. Experiments to study the hypothesis that
bacterial infections might be affecting immunogenicity of rhIFNβ is presented in Chapter
5. Here, the interaction between several species of gram positive and negative bacteria with
(aggregated) rhIFNβ has been evaluated.
Although treatment related factors have been commonly acknowledged to alter the risk of
ADA production, most conclusions are drawn from vaccination studies or clinical reports
in which different products with distinct rhIFNβ proteins, treatment routes and doses are
compared for immunogenicity. Chapter 6 presents the results of a study focusing on the
influence of the dosage regime, administration frequency and route of injection using 1
rhIFNβ product.
Due to the low bioavailability of therapeutic proteins after oral administration the parental
administration remains the primary route of delivery. However, almost no information is
present on how (aggregated) rhIFNβ will distribute in the body after administration via
different routes. Chapter 7 describes studies in which the differences in in vivo and ex vivo
biodistribution between monomeric and aggregated proteins have been investigated after
injection via different routes.
In Chapter 8 a summary of the major findings and conclusions described in this thesis
are given. Also several perspectives for future research on the immunological processes
underlying immunogenicity are presented.
14
References.
Introduction.
1.
Kieseier BC. 2011. The mechanism of action of interferon-β in relapsing multiple sclerosis.
CNS drugs 25:491–502
2.
Coppola G, Lanzillo R, Florio C, Orefice G, Vivo P, Ascione S, Schiavone V, Pagano A,
Vacca G, De Michele G, Morra VB. 2006. Long-term clinical experience with weekly
interferon beta-1a in relapsing multiple sclerosis. Eur J Neurol 13:1014–1021
3.
Arbizu T, Alvarez-Cermeño JC, Decap G, Fernández O, Uría DF, García Merino A,
Izquierdo G, Montalbán X. 2000. Interferon beta-1b treatment in patients with relapsing-remitting multiple sclerosis under a standardized protocol in Spain. Acta Neurol Scand
102:209–217
4.
Trojano M, Liguori M, Paolicelli D, Zimatore GB, De Robertis F, Avolio C, Giuliani F,
Fuiani A, Livrea P; Southern Italy MS Group. 2003. Interferon beta in relapsing–remitting
multiple sclerosis: an independent postmarketing study in southern Italy. Mult Scler 9:451–
457
5.
1993. Interferon beta-1b is effective in relapsing-remitting multiple sclerosis. I. Clinical
results of a multicenter, randomized, double-blind, placebo-controlled trial. The IFNB
Multiple Sclerosis Study Group. Neurology 43:655–661
6.
Paty DW, Li DK. 1993. 1993. Interferon beta-1b is effective in relapsing-remitting multiple
sclerosis: II. MRI analysis results of a multicenter, randomized, double-blind, placebocontrolled trial. Neurology 43:662–667
7.
Trojano M, Pellegrini F, Fuiani A, Paolicelli D, Zipoli V, Zimatore GB, Di Monte E,
Portaccio E, Lepore V, Livrea P, Amato MP. 2007. New natural history of interferon-betatreated relapsing multiple sclerosis. Ann Neurol 61:300–306
8.
Milanese C, La Mantia L, Palumbo R, Martinelli V, Murialdo A, Zaffaroni M, Caputo D,
Capra R, Bergamaschi R; North Italy Multiple Sclerosis Group. 2003.A post-marketing study
on interferon beta 1b and 1a treatment in relapsing-remitting multiple sclerosis: different
response in drop-outs and treated patients. J Neurol Neurosurg Psychiatry 74:1689–1692
9.
Runkel L, Meier W, Pepinsky RB, Karpusas M, Whitty A, Kimball K, Brickelmaier M,
Muldowney C, Jones W, Goelz SE. 1998. Structural and functional differences between
glycosylated and non-glycosylated forms of human interferon-beta (IFN-beta). Pharm Res
15:641-649
10.
Sominanda A, Rot U, Suoniemi M, Deisenhammer F, Hillert J, Fogdell-Hahn A. 2007.
Interferon beta preparations for the treatment of multiple sclerosis patients differ in
neutralizing antibody seroprevalence and immunogenicity. Mult Scler 13:208–214
11.
Perini P, Facchinetti A, Bulian P, Massaro AR, DePascalis D, Bertolotto A, Biasi G,
Gallo P. 2001. Interferon-beta (INF-b ) antibodies in multiple sclerosis patients. Prevalence,
kinetics, cross-reactivity, and factors enhancing interferon-b immunogenicity in. Eur
Cytokine Netw 12:56-61
12.
Bertolotto A, Deisenhammer F, Gallo P, Sölberg Sørensen P. 2004. Immunogenicity of
interferon beta: differences among products. J Neurol 251:II15–II24
13.
Boz C, Oger J, Gibbs E, Grossberg SE; Neurologists of the UBC MS Clinic. 2007. Reduced
effectiveness of long-term interferon-beta treatment on relapses in neutralizing antibodypositive multiple sclerosis patients: a Canadian multiple sclerosis clinic-based study. Mult
Scler 13:1127–1137
14.
Van Beers MM, Jiskoot W, Schellekens H. 2010. On the role of aggregates in the
immunogenicity of recombinant human interferon beta in patients with multiple sclerosis. J
Interferon Cytokine Res 30:767–775
15.
Barnard JG, Babcock K, Carpenter JF. 2013. Characterization and Quantitation of Aggregates
15
Chapter 1.
and Particles in Interferon-β Products : Potential Links Between Product Quality Attributes
and Immunogenicity. J Pharm Sci 102:915–928
16.
Hermeling S, Jiskoot W, Crommelin, D, Bornaes C, Schellekens H. 2005. Development of
a transgenic mouse model immune tolerant for human interferon Beta. Pharm Res 22:847–
851
17.
van Beers MM, Sauerborn M, Gilli F, Hermeling S, Brinks V, Schellekens H, Jiskoot W.
2010. Hybrid transgenic immune tolerant mouse model for assessing the breaking of B cell
tolerance by human interferon beta. J Imunol Methods 352:32–37
18.
Ruotsalainen J, Martikainen M, Niittykoski M, Huhtala T, Aaltonen T, Heikkilä J, Bell J,
Vähä-Koskela M, Hinkkanen A. 2012. Interferon-β sensitivity of tumor cells correlates with
poor response to VA7 virotherapy in mouse glioma models. Mol Ther 20:1529-1539
19.
Qin XQ, Beckham C, Brown JL, Lukashev M, Barsoum J. 2001. Human and mouse IFNbeta gene therapy exhibits different anti-tumor mechanisms in mouse models. Mol Ther
4:356–364
20.
Mestas J, Hughes CC. 2004. Of mice and not men: differences between mouse and human
immunology. J Immunol 172:2731–2738
21.
De Groot AS, McMurry J, Moise L. 2008. Prediction of immunogenicity: in silico paradigms,
ex vivo and in vivo correlates. Curr Opin Pharmacol 8:620–626
16
CHAPTER 2.
Immunogenicity of therapeutic proteins.
Grzegorz Kijanka
Wim Jiskoot
Melody Sauerborn
Huub Schellekens
Vera Brinks
Pharmaceutical Formulation Development of Peptides and Proteins,
CRC Press, 2012: pp 297-322
17
Chapter 2.
Abstract.
Therapeutic proteins are one of the most rapidly growing groups of drugs in respect to
the number of newly approved drugs and market value. However, the positive effect of
protein drug therapy is often abrogated by the production of anti drug antibodies (ADA).
These ADA might decrease or abolish treatment efficiency. However, in rare cases, the
ADA may neutralize endogenous proteins leading to sever side reactions. Many factors
have been suggested to increase the risk of ADA development. Among them the presence of
protein aggregates in the formulated drug seems to be the most potent factor in triggering
ADA production. Depending on the origin of the protein drug ADA can be produced via
the classical immune response mechanisms or via the not yet fully understood pathway
referred to as “breaking of tolerance”. A number of in silico, in vitro and in vivo tools have
been developed to predict the immunogenicity of therapeutic proteins. However, due to
insufficient knowledge on the mechanisms underlying immunogenicity they are not reliable.
This manuscript summarizes current knowledge on immunogenicity of therapeutic proteins.
18
1. Introduction.
Immunogencity of therapeutic proteins.
The use of therapeutic proteins originates in the 19th century when polyclonal horse antitetanus and anti-diphtheria immunoglobulins were used for the first time. Then in the 1920s
bovine and porcine insulin were introduced for the treatment of diabetes. Since then insulin
and many more protein drugs, including growth factor, erythropoietin and a whole range of
monoclonal antibodies became major therapies. Currently, a substantial part of the newly
approved drugs are therapeutic proteins. A common feature of all therapeutic proteins is their
immunogenicity, the potential to evoke an adverse immune response. In the field of protein
drugs, the presence of anti-drug antibodies (ADA) is often referred to as seroprevalence,
whereas the term immunogenicity is referred to as a quantitative measurement of ADA
titers 1. However this distinction is not a general rule. Therefore in this chapter the term
immunogenicity will be used in a general manner, describing the potential of the therapeutic
protein to evoke an adverse immune response.
The immunogenicity of the first therapeutic proteins, such as insulin, has been explained by
their non-human origin, resulting in a classical immune response (see section 3.2). Protein
drugs of the next generation, e.g. plasma-derived clotting factors 2 and growth hormone
extracted from the pituitary glands of human cadavers 3, were deprived of this drawback
and therefore application of these human-derived proteins had been believed to solve the
problem of immunogenicity. Surprisingly it had not. Moreover, their potential to spread
infections like HIV and Creutzfeld-Jacob resulted in restriction of application of those
drugs. The real breakthrough in protein drug development took place in the 1980s when
DNA recombination methods were introduced in the production of therapeutic proteins. This
drastically opened the possibility of large-scale production of highly purified, therapeutic
proteins homologous to endogenous proteins. Although assumed to be non-immunogenic,
we are now still far away from the elimination of unwanted immune responses and
understanding the exact mechanisms underlying immunogenicity of therapeutic proteins
that closely mimic the structure of endogenous proteins. Many hypotheses trying to explain
immunogenicity are based on assumptions rather than facts, which additionally blur our
insight into immunogenicity.
In this chapter the current knowledge in understanding immunogenicity of therapeutic
proteins is summarized. Insight into potential risk factors, mechanisms underlying unwanted
immune responses as well as the available tools and assays to determine immunogenic
potential of (new) therapeutic proteins are described.
2. Immunogenicity as a clinical problem.
Almost all patients treated with a non-human therapeutic protein developed an antibody
response against the therapeutic protein 4. The development of recombinant human
proteins, together with the improvement of production processes allowed a reduction of
adverse reactions associated with protein drug therapy. However, immunogenicity is still
an important issue, which has to be included in the assessment of a protein drug before its
19
Chapter 2.
Figure 1. Schematic representation of the difference in binding between binding (BAbs) and
neutralizing antibodies (NAbs). Binding antibodies recognize a part of the protein which is not
involved in enzymatic activity or receptor recognition (active site). Therefore BAbs don’t block
the activity of the protein. In contrast, neutralizing antibodies recognize the active site or an epitope
nearby, directly or indirectly blocking the activity of the protein.
introduction onto the market. Moreover, the development of new, more sensitive ADAdetecting assays revealed that protein drugs induce an antibody response in a much higher
percentage of patients than was initially believed 5.
2.1. Clinical significance of unwanted immune responses.
Immunogenicity of therapeutic proteins can have a number of clinical consequences
which differ in their impact on therapy outcome or patient health, varying from no obvious
clinical consequences to life-threatening conditions. In the development of an immune
response usually binding anti-drug antibodies (binding ADA, BAbs) are observed first 6.
Such antibodies recognize drug epitopes that are not involved in receptor recognition or
substrate processing, and often lead to rapid clearance of the therapeutic protein-antibody
complexes from the blood stream. However, some drugs in complex with antibodies show
a prolonged half-life 7. BAbs do not always lead to lower efficiency of therapy. In contrast
to BAbs, neutralizing antibodies (neutralizing ADA, NAbs) recognize the active site of
the therapeutic protein with high affinity and inhibit substrate processing or its receptor
interaction 6. Figure 1 shows the comparison between BAbs and NAbs. How NAbs affect
20
Immunogencity of therapeutic proteins.
treatment strongly depends on their titers. When NAbs are produced in low or moderate
titers, their influence on therapy can be similar to that of BAbs, whereas in patients with
high NAb titers, the effect of treatment can be completely inhibited 1,8.
Since most of the protein drugs are recombinant human proteins, it is possible that ADA
interact with corresponding endogenous proteins. Such an interaction would correlate
with a serious risk of autoimmune disease, e.g. pure red cell aplasia (PRCA) in the case
of erythropoietin 9, as described later (section 2.2). Therefore, if there is suspicion of
ADA production, e.g. significant reduction of treatment efficiency or increased disease
symptomatology, the levels of ADA including neutralizing activity should be tested. If the
risk of severe side effects is serious, the treatment has to be terminated and an alternative
therapy applied. Physicians are recommended to consider switching the treatment of
multiple sclerosis (MS) patients from interferon beta (IFNβ) to alternatives (e.g. glatiramer
acetate or natalizumab) as soon as NAbs are detected 10. However, before a decision about
the termination of treatment is made, the balance between therapy outcome and possible
risk of antibody formation has to be considered. For example, in patients with MS antiIFNβ NAb titers, especially low or moderate ones, have a tendency to decrease over time.
Deciding too early that the treatment should be terminated can bereave the patient of the
positive effect of IFNβ, and might even lead to a faster disease progression 10.
2.2. Examples of immunogenicity.
Introducing protein drugs into the clinic has been a real breakthrough in the treatment of a
number of diseases like: rheumatoid arthritis (monoclonal antibody 11), anemia (erythropoietin
12
), diabetes (insulin 13) and hepatitis (interferon alpha 14). However, the beneficial effect
of therapeutic proteins is often significantly reduced by the unwanted immune response
against the drug. In this section four examples of protein drugs with clinical consequences
of an unwanted immune response will be described.
The most well-known therapeutic protein with severe immunogenicity issues is
erythropoietin. Immunogenicity of this therapeutic protein shook the belief that therapeutic
proteins are very safe, and began the discussion about the significance of immunogenicity
in the clinic. Antibody formation against erythropoietin can induce pure red cell aplasia,
when antibodies against the drug cross-neutralize endogenous erythropoietin. Although
PRCA incidence has been reported since erythropoietin therapy was introduced for patients
suffering from renal failure, PRCA was extremely rare and a connection between the
disease and treatment was not clear. In 1998, Johnson & Johnson introduced reformulated
erythropoietin named Eprex® for the therapy of patients with chronic renal failure. Following
EMA guidance, human serum albumin (HSA) was replaced by polysorbate 80 and glycine
as stabilizers (outside US). Between 1998 and 2004, the occurrence of PRCA significantly
increased and, in the majority of patients, was related to Eprex® treatment 9,15. A number of
studies correlated immunogenicity to the use of non-coated rubber stoppers of ready-touse syringes with Eprex®. According to the manufacturer, leachates containing phenolic
21
Chapter 2.
derivates, extracted by polysorbate 80 from uncoated rubber, would have acted like an
adjuvant inducing an immune response 16. Since thousands of patients undergo Eprex®
therapy and PRCA developed only in a small percentage of them, many scientists question
the importance of adjuvant properties of leachates as the main immunogenic factor 9,17.
More probably the usage of polysorbate 80 led to lower stability of the protein, which could
have induced immunogenic impurities e.g. as a result of improper handling. However, the
exact mechanism of erythropoietin associated PRCA remains unknown.
IFNβ is the major drug used as first line therapy of MS. Although it does not cure the
disease, it reduces the number and strength of relapses of MS, significantly improving
patients’ quality of life. The first available formulation of IFNβ was Betaseron®, produced in
bacteria (Escherichia coli) and introduced on the market in 1993. The first reports showing
its immunogenicity were published already in the same year. Betaseron® and biosimilar
Extavia® (which has been recently marked), are formulations of the IFNβ-1b type, which
differs from the endogenous protein in three points; (i) it lacks N-terminal methionine, (ii)
the cysteine from residue 17 is replaced by serine and (iii) it is not glycosylated. In contrast,
IFNβ-1a, the active compound in Avonex® (Biogen) and Rebif® (Merck Serono), is produced
in Chinese hamster ovary (CHO) cells and does not differ in amino-acid sequence from
endogenous IFNβ. Although Avonex® and Rebif® show lower seroprevalence (percent of
ADAs’ positive patients) than Betaseron®, Rebif® has been shown to induce the highest titers
of NAbs among those drugs 1,18,19. The difference in immunogenicity between the products
is probably not only related to protein structure, but also to their activity, formulation, route
and frequency of administration 20. Betaseron® contains human serum albumin (HSA) as
a stabilizer, whereas Avonex® and Rebif® are available in a formulation with and without
HSA. Betaseron® and Avonex® with HSA are sold in lyophilized form, which has to be redissolved directly before injection. HSA-free Avonex® and Rebif® are supplied as liquids in
ready-to-use syringes. Betaseron® and Rebif® are administered subcutaneously (SC) every
other day and three times a week, respectively, whereas Avonex® is injected intramuscularly
(IM) every week. Overall, the immune response against IFNβ develops within 3-12 months
of treatment. After 24 months, if treatment is not terminated, antibody titers have a tendency
to decrease 8,21. However, if high titers of NAbs are developed the therapy outcome is
significantly altered 8. Therefore, a major item in the society of MS experts is the question
whether treatment should be continued or terminated when NAbs are detected 10,21,22.
Monoclonal antibodies are the fastest growing class of biopharmaceuticals. Due to their
nature (possession of highly variable antigen-binding regions), they can be designed
to selectively bind almost any target. The first monoclonal antibodies used in the clinic
were of murine origin, very immunogenic and often induced severe anaphylactic or
anaphylactoid reactions 23. Thanks to the recombinant technology developed in the 1980s,
most monoclonal antibodies nowadays present in the clinic are humanized or fully human
proteins 24. In general, the more human the monoclonal antibody is, the less immunogenicity
it evokes. However, also a human antibody can show a significant immunogenicity. After a
22
Immunogencity of therapeutic proteins.
few months of treatment with infliximab, a chimeric antibody against tumor necrosis factor
alpha (TNF-α), antibodies are found in 20% to 60% of the patients (depending on study) 25.
Adalimumab, a fully human antibody against the same therapeutic target, has been shown
to evoke antibody production in a similar percentage of patients 5,26.
The last group of therapeutic proteins described in this section are bone morphogenic
proteins (BMPs) belonging to the family of growth factors. There are two types of BMPs
available on the market, BMP-2 and BMP-7. They are used to induce spinal fusion or are
given to treat long non-union fractures. In contrast to the majority of other protein drugs,
which are used mainly in the therapy of chronic diseases, BMPs are used as single treatment,
by surgical implantation. Depending on the study, BMPs were shown to evoke unwanted
immune responses in 40%-60% percent of the patients. Although severe side effects are very
rare, development of anti-BMP antibodies significantly reduces the success of treatment 27.
3. Immunological mechanism underlying immunogenicity of therapeutic proteins.
The immune system is complex, and evolved to protect organisms from all kinds of pathogens
such as viruses and bacteria. The wide range of infectious agents forced the development
of a very sophisticated defense system, which includes the cooperation of many proteins
and cell types. Thanks to many years of intensive studies our knowledge about the immune
responses against many antigens has drastically improved, although new findings still bring
more insight into the complexity of the immune system.
3.1. Innate and adaptive immune system.
The first line of defense against pathogens is the innate immune system composed of physical
barriers and phagocytic cells, which are responsible for preventing invasion of pathogens
and for the removal of pathogens from the organism, respectively. A typical feature of the
cellular component of the innate immune system is its rapid, but rather unspecific manner of
antigen removal. The efficiency of the innate system strongly depends on the identification
of invaders, which is based on the recognition of foreign compounds on the surface of
pathogens by mannose receptors, scavenger receptors and Toll-like receptors (TLRs) 28.
When the innate system cannot sufficiently fight the infection, the adaptive system comes
into action. In contrast to the innate system, the adaptive system is capable of mounting
highly specific responses. The main players of adaptive immune response are T and B
lymphocytes.
A very important feature of the adaptive immune system is tolerance for self antigens.
During development, immature lymphocytes undergo selection based on their affinity to
self epitopes in the thymus (T lymphocytes) and in the bone marrow (B lymphocytes) 29,30.
This mechanism of T and B cell selection is known as central tolerance. However, since it is
virtually impossible to present all possible self epitopes in the thymus or bone morrow, some
auto reactive cells may escape the elimination. Those self reactive T and B lymphocytes are
controlled by a mechanism known as peripheral tolerance. Peripheral tolerance is maintained
mainly due to the lack of the “danger” signal, usually derived from other immune cells,
23
Chapter 2.
needed for T or B cell activation 30. In addition, specialized regulatory cells (both B and
T type) are capable of eliminating self reacting immune cells from periphery, creating an
additional safety mechanism.
3.2. Immune reaction against therapeutic proteins.
There is no universal immune mechanism underlying the immunogenicity of therapeutic
proteins. Mostly, the immune response against protein drugs is explained as a classical
immune response against foreign epitopes or as a breakage of tolerance (Table 1). A classical
immune response is usually triggered by external antigens and involves the T cell dependent
(TD) activation of B cells (Figure 2). This type of immune mechanism is found when
treating patients with therapeutic proteins having a non-human origin. However, most of
the recombinant proteins are human analogues. They likely lead only to a classical immune
response in patients who have a genetic disorder and thus lack tolerance for a certain protein
(e.g. haemophiliacs with a factor VIII gene defect 2).
In general, the classical immune response involves a fast and robust immune reaction leading
to the production of NAbs as well as memory formation. For example, Muromonab-CD3,
a murine antibody which was the first monoclonal antibody approved for usage in humans,
induces production of NAbs in all patients within two weeks after administration 31.
While the classical immune response is quite well understood, the breaking of tolerance
by human derived or recombinant proteins remains a mystery. Tolerance for self-derived
epitopes has to be at least temporally disturbed to develop a reaction against those types of
protein drugs, e.g. due to presence of aggregates which might activate B cells via a T cell
independent (TI) pathway. Development of an immune response against recombinant or
human proteins usually takes a few months and leads to the production of NAbs only in a
part of all ADAs’ positive patients. Clinical and experimental data show also that during the
proposed breaking of tolerance IgG is produced, while memory B cells are not formed 32.
Moreover, the immune response against majority of protein drugs does not lead to severe
side effects.
3.3. T cell independent immune response.
The TI response is triggered by molecules or structures that are typical for lower organisms
like bacteria or viruses. The T cell independent activation of B cells can be divided into
two types: polyclonal, induced by bacterial molecules like lipopolysaccharide (LPS) –
Type I, and not polyclonal, induced by repetitive structures – Type II (Figure 3). As in
the TD immune response, a strong “danger” signal has to be provided to activate B cells
and therefore induce antibody production. In a TI type I response, binding of antigens like
LPS or bacterial DNA to TLRs directly leads to B cell activation. A TI type II response
is elicited by an antigen-induced change of B cell receptor (BCR) reorganization on the
naïve B lymphocyte, leading to crosslinking or dissociation of BCRs 33. To induce such a
24
Immunogencity of therapeutic proteins.
Figure 2. Schematic representation of the B cell activation pathway in T cell dependent manner.
Antigen is caught and digested by an antigen presenting cells (APC). The APC then presents short
peptides, created during digestion, to T cells via II MCH molecules. The interaction between MCH II
and TCR is the first activation signal for T cells. Upon additional stimulation, i.a. co-stimulation via
CD4 and CD80/86 interaction with CD28, T cells become activated. Simultaneously B cells interact
with an antigen via their BCRs. However, without additional stimulation from direct contact with T
cells and T cell derived cytokines, B cell remains in a state of anergy. Activated B cells undergo a
maturation process and differentiate into antibody secreting plasma cells or memory cells.
25
Chapter 2.
Figure 3. Schematic representation of the B cell activation pathway in a T cell independent manner.
The first step in the TI activation of a B cell is the interaction of an antigen with the BCR. In order to
fully activate a B cell, an additional “danger” signal is needed. TI type I antigens bind to a receptor
designed to recognize bacterial structures e.g. toll like receptor (TLR). In contrast, TI type II antigens
bind to several BCRs changing their distribution on the membrane and leading to BCR cross-reaction.
To bind more than one BCR antigen, repetitive epitopes have to be present. These repetitive epitopes
can be part of the polymeric structure of an antigen (e.g. sugar moieties of polysaccharides) or can
be displayed by protein aggregates.
reorganization of BCRs, a special type of antigen is required. The typical TI type II antigen
is a polymer or an aggregate, with a molecular weight above 100 kDa, possessing on its
surface repetitive epitopes, which form a two-dimensional network with a spacing of about
50-100Å (5-10 nm) between epitopes 34. The TI pathway leads to direct B lymphocyte
activation and antibody production. However, in contrast with the TD pathway, the B
lymphocytes do not differentiate into memory B cells and maturation of the antibody
response does not occur or is very limited. Therefore a TI immune response is characterized
26
Immunogencity of therapeutic proteins.
Table 1. Comparison of the suggested immune mechanisms leading to ADA production. While the
TD and TI pathways are well known, the characteristics of the breaking of tolerance pathway are not.
The data presented on the breaking of tolerance is a combination of both patient-and animal data.
Immune mechanism
TD
TI
Bacterial/Viral compounds
e.g. LPS
5-7 days
Breaking of tolerance
Type of antigen
Time of antibodies
arising
Formation of
germinal centers
Foreign protein
Yes
APCs,
No
Cells required
B lymphocytes,
B lymphocytes
B lymphocytes,
T lymphocytes.
Yes
No
T lymphocytes (?)
Yes
Yes
No
?
Yes
Yes
Yes
No
Antibody isotype
switching
Affinity maturation
of produced
antibodies
Production of
neutralizing
antibodies
Formation of
memory B cells
5-7 days
Recombinant human protein
After repeated exposures, usually
> 1-2 months
?
APCs (?)
Yes/No
(Strongly depends on protein and
its formulation)
No
by the formation of IgM as the major antibody class 33.
It is very important to keep in mind that the TD and TI response are not mutually exclusive.
In fact, most immune responses are a combination of a TD and a TI mechanism. It is
known that the TI immune response can be modulated by cytokines produced by T cells
35
. Reorganization of BCRs, induced by TI type II antigens, also leads to a much faster
development of a TD immune response 36. Moreover, a few years ago the production of
memory B cells via a TI response was observed 37. The insight to date on therapeutic protein
immunogenicity will be discussed in the next section.
4. Why are therapeutic proteins immunogenic?
Explanations why therapeutic proteins evoke an antibody response have been changing
over time. For the first therapeutic proteins, their foreign origin was believed to underlie the
immunogenicity, e.g. a classical immune response against bovine insulin 4. As recombinant
human proteins and human antibodies also turned out to be immunogenic, we now realize
that the mechanisms of antibody formation are much more complex than a distinction
between ‘self’ and ‘non-self’. Since our knowledge on the mechanisms underlying
immunogenicity of current therapeutic proteins is limited, factors potentially affecting
immunogenicity should be studied for each (class of) therapeutic protein. In general, there
are three categories of factors that can affect immunogenicity; patient, product, ADA
detecting assays and treatment related (Table 2.).
27
Chapter 2.
4.1. Patient related factors.
The significance of patient related factors influencing immunogenicity of a protein drug
is often underestimated during drug development. However, clinical data show that most
therapeutic proteins induce an unwanted immune response only in some patients, suggesting
that patient characteristics may be crucial for immunogenicity. For example, the genetic
background of patients can have a significant impact on the treatment outcome.
Some diseases, for example hemophilia A, are correlated with various mutations of the gene
encoding the endogenous protein (in case of hemophilia A, factor VIII) 38. Approximately
20% of patients with severe hemophilia A develop antibodies during treatment with factor
VIII. Those patients present the most severe form of hemophilia A, which is a direct result of
large deletions or nonsense mutations in the factor VIII gene. As a consequence patients are
not tolerant to functional factor VIII. In contrast, patients with mild or moderate hemophilia
A, who do produce functional factor VIII, are much less prone to the development of
antibodies against factor VIII 38. Moreover, a significant part of protein therapies have been
designed to replace or to supplement endogenous protein deficiency, mainly enzymes.
Such therapies will always be associated with a higher risk of immunogenicity compared
to therapies for patients that do not have any (genetic) abnormalities in endogenous protein
expression 39,40.
Not only lack of functional protein expression can influence the immunogenicity of
therapeutic proteins but also patients’ human leukocyte antigen (HLA) haplotype. Diabetic
patients with the human leukocyte antigen D-related 7 (HLA-DR7) have been reported to
produce higher titers of insulin antibodies than other diabetic patients, whereas patients who
are HLA-DR3 homozygous produce lower titers 41. Also in MS patients treated with IFNβ an
association between MHC class II haplotype and the level of anti-IFNβ antibody production
was found. In vitro studies showed that the IFNβ molecule contains T-cell epitopes which
can be recognized and presented on B cells expressing DRB1*0701 MHC class II allele
42
. Clinical data confirm that patients whose antigen presenting cells (APCs) express
DRB1*0701 MHC II allele are more prone to develop an unwanted immune response than
Table 2. Factors influencing the immunogenicity of therapeutic proteins.
Product related
1. Source of protein
2. Posttranslational
modifications
3. Excipients
4. Impurities (including
degradation products such
as aggregates)
5. Pegylation
28
Factors influencing immunogenicity
Therapy related
ADA detecting assay
1. Dose
2. Dosing schedule
3. Duration of therapy
4. Co-medication
1. Sensitivity of assay
2. Specificity of assay
Patient related
1. Genetic background
(e.g. non-sense
mutations, HLA
haplotype)
2. Age
3. Type of disease
patients who lack this MHC II allele [33].
Immunogencity of therapeutic proteins.
ADA production can also be influenced by the type of disease. Patients who suffer from
chronic infectious diseases and whose immune systems are in an activated state, are more
likely to produce antibodies against interferon alpha (INFα) rather than patients suffering
from cancer 43. In contrast, patients with impaired immune system may be less prone to the
development of ADA than patients with an intact immune system 44.
4.2. Product related factors.
The production of a protein is a complicated, multi-step process. During product
development many factors that may influence the immunogenicity of the final product have
to be taken into consideration. In this section the most important product related factors will
be described.
One of the most important product related factors, which can enhance the probability of
an immune response against the protein drug, is its origin 4. The immune mechanisms
underlying immunogenicity of non-human proteins are usually explained on the basis of
the sequence differences between the product and the endogenous protein. However, one
has to keep in mind that other factors, e.g. protein aggregation, can be equally important
for triggering the immune system. Non-human proteins are recognized as foreign antigens
and evoke a response via the TD pathway. This type of immune response is responsible for
the immunogenicity associated with the first, animal derived therapeutic proteins as well as
with the immune reaction against plant or bacterial proteins.
Although sequence variation between endogenous and exogenous proteins seems to be
an important risk factor for immunogenicity, 100% sequence homology does not lead to
absence of immunogenicity. Usage of human tissue derived proteins like growth hormone
7
or human interferon beta 45 is accompanied by clinically significant immune responses.
All recombinant therapeutic proteins, despite the similarity of their sequences to natural
proteins, evoke an immune response. In fact, some of the recombinant “human” proteins
have been found to be more immunogenic than their human tissue derived counterparts;
for example, whereas natural interleukin 2 (IL-2) shows only minimal immunogenicity, the
recombinant IL-2 produced in Escherichia coli cultures induces an immune response in over
50% of patients 46. This phenomenon can be explained by differences in posttranslational
modifications between bacteria and mammals, such as glycosylation (not present in bacteria),
methylation or acetylation of protein. Differences in posttranslational modifications can
result in the exposure of epitopes on endogenous proteins, which are normally covered
by polysaccharides. These different modifications themselves may also be recognized as
foreign and thereby trigger an immune response. Lack of glycosylation is one of the possible
reasons explaining higher immunogenicity of IFNβ-1b compared to IFNβ-1a formulations.
One of the crucial steps in product development is formulation development. In some
formulations human serum albumin is used (HSA). HSA is generally considered as a non-
29
Chapter 2.
immunogenic excipient with very good stabilizing characteristics. However, the addition
of HSA was correlated with formation of aggregates and increased immunogenicity of
IFN-alpha2a formulations 47. When HSA and IFN-α2a are mixed, aggregation between the
therapeutic protein and HSA may occur. Moreover, addition of bovine serum albumin (BSA)
to IFN- α2b in molar ratio equal or higher than 5:1 (BSA:IFN- α2b) significantly disrupted
the secondary structure of IFN- α2b 48. HSA may have a similar impact on protein in
formulations. Despite the possible immunological consequences of adding HSA as
an excipient, its removal has to be carefully considered as it may lead to a less robust
formulation (cf. PRCA case with Eprex®, see section 2.2).
During recent years, several papers have been published that show the potential
immunogenicity of protein aggregates. Formation of aggregates can occur due to many
physical or chemical processes like oxidation, protein cleavage, or conformational changes
49,50
. Oxidation can change the epitope structure and induce faster degradation of the protein.
Misfolding, degradation or aggregation may be also induced by leachates from containers,
silica droplets and contact with glass 16. Deglycosylation of the protein backbone can unmask
hydrophobic epitopes, which are hidden under the polysaccharide residues 28.
A high content of aggregates has been correlated with increased immunogenicity of
different IFNβ formulations: Betaferon® (high aggregate content) is more immunogenic
than Avonex® (low aggregate content) 19,20. The presence of aggregates has been shown to
correlate with increased immunogenic potential of human-derived growth factor. Patients
treated with highly aggregated hGH (50-70% of aggregates) developed persistent antibody
production, which remained after switching to non-aggregated hGH 7. On the other hand,
patients treated with a formulation of hGH containing low amounts of aggregates (less than
5%), stopped producing antibodies when they received a non-aggregated product 7.
Chemical conjugation of proteins with polyethylene glycol (PEG) might affect their
immunogenicity. Pegylation of the protein chain usually does not lead to conformational
changes, while it does prolong the protein’s circulation time. In addition, it may shield
immunogenic sites, like glycosylation, and therefore reduce the risk of an immune
reaction. For example, pegylation successfully decreased immunogenicity of therapeutic
enzymes such as asparaginase 51 or arginase 52, without affecting their enzymatic properties.
However, studies with pegylated rhIFNα2a have shown that although its immunogenicity
was decreased and serum half life increased, activity of the drug in vitro was lowered by
93% 52. Moreover, it has been claimed that pegylation might lead to the production of antiPEG antibodies, but the influence of PEG on immunogenicity is not proven.
4.3. ADA detection - assay related factors.
The identification of ADA responsive patients, and thus the observed immunogenicity of
the therapeutic proteins, is significantly influenced by the assays used to detect ADAs. Lack
of standardized assays and protocols results in high variability in ADA detection between
studies. The reported percentage of patients developing ADAs against Betaseron® varied
30
from 30% up to 97% 19.
Immunogencity of therapeutic proteins.
The seroprevalence can be underestimated when ADA are measured in serum in the
presence of drug, especially when ADA titers are relatively low and drug circulation time
is long. Recently a report showing the impact of assay improvement on seroprevalence of
patients treated with Adalimumab® has been published. Acidic treatment used to dissociate
ADA from the drug resulted in an increase of ADA positive samples from 5% to almost
90% 5. This example reveals the problem of lack of ADA assay standardization. Therefore
both FDA and EMA prepared guidelines in which recommendations for sample collection,
preparation and ADA assay formats are enclosed 53,54.
4.4. Therapy related factors.
In addition to patient and product characteristics, design of therapy can significantly
influence the immunogenicity of protein drugs. By changing the dose, frequency, route of
administration it is possible to efficiently decrease the drug’s immunogenicity. Moreover,
additional drugs often used to increase the treatment efficiency may have impact on the
immunogenicity of therapeutic protein. For instance, co-medication with immunosuppressant
drugs can result in lower or less frequent immune responses against therapeutic proteins 44.
Proteins can be administrated by many different routes which may enhance or decrease
their immunogenic potential. In general the probability of unwanted immune response is
the highest when protein is administered SC, followed by the IM and IV route 4,18,31. The
immune response against both IFNβ-1a (Rebif®) and IFNβ-1b (Betaseron®) was higher after
SC injection than after IM administration 18,55. Clinical data suggest a correlation between
increased immunogenicity and increased frequency of drug administration or its dose 31.
However, clinical data are not always consistent for all therapeutic drugs. Whereas a higher
dose was correlated with higher immunogenicity in patients treated SC with rhIFNβ 18, the
administration of a higher dose of Adalimumab led to a decreased percentage of patients
with measurable ADA 56.
Duration of treatment is also correlated with immunogenicity of protein drugs. Long
lasting, chronic therapy is usually accompanied with higher risk of an unwanted immune
reaction than acute treatment 4,31. However some drugs, like murine monoclonal antibody
Muromonab-CD3 or human recombinant BMPs, can induce significant unwanted immune
response after only single administration27,31.
5. Pre-clinical assessment of the immunogenicity of therapeutic proteins.
Protein drugs are the leading group of medicines in respect to their increasing market value.
More and more companies are increasing their efforts in developing new protein drugs as
well as generating biosimilars. Simultaneously, both the regulatory bodies and the industry
put a higher emphasis on the understanding and prediction of their immunogenicity. In
general, three types of approaches are being used to predict potential immunogenicity of
protein drugs: in silico, in vitro and in vivo methods (Table 3). All methods suffer from
31
Chapter 2.
Table 3. Comparison of the different approaches for immunogenicity prediction.
Tools
Assays
Used for
In silico
1. T cell epitope prediction,
2. B cell epitope prediction,
3. Prediction of Aggregateprone sequences
1. Primary screening in drug
discovery phase
In vitro
1. MHC II binding assays
2. T cell activation assays
3. B cell activation assays
1. Validation of epitopes
found in silico,
2. Screening of formulated
products
In vivo
1. Non-transgenic animals
2. Non-human primates
3. Transgenic animals
1. Evaluation of in silico and
in vitro screening approaches
2. Evaluation of formulated
product
Features
1. Time efficient
2. Relatively cheap
3. Outcome depends on input
data that are at best partly
known
1. Relatively easy
2. Time consuming
3. Performed in artificial
environment
1. Expensive
2. Time consuming
3. Encompasses the entire
immune system
4. Translation to
immunogenicity in patients
questionable
lack of data about the exact mechanism underlying immunogenicity and their presumed
predictive value is largely based on assumptions derived from the mechanisms described by
“classical immunology”. For example, almost all in silico algorithms consider the unwanted
immune response against therapeutic proteins as vaccine-like. As such, in silico prediction
of immunogenic T cell epitopes is based on the TD pathway. Whereas this may be useful in
antigen selection for vaccine development, in case of therapeutic proteins it can lead to misor overinterpretation of experimental data. Moreover, in silico methods are purely based on
(mainly primary) structure of the protein and do not take into account how product quality
(e.g. presence of impurities such as aggregates and host cell proteins) and treatment regime
(e.g. route of administration, concomitant medication, etc.) affect the risk of an immune
response. The aim of this section is to present existing tools for immunogenicity prediction
and their usefulness in the process of protein drug development.
5.1. In silico tools.
The first in silico algorithm for the determination of proteins’ immunogenic properties,
was presented in the 1980s 57. This algorithm calculates the immunogenic properties of
proteins on the basis of their relative hydrophobicity, based on their amino acid structure.
Since then, new tools for immunogenicity prediction in silico have been developed and
improved. Nowadays a number of algorithms designed for immunogenicity prediction are
available. The majority of algorithms were developed for the prediction of T cell rather than
B cell epitopes. This is explained by the fact that T cells recognize short, linear sequences
of amino acid chain presented on MHC II molecules, whereas most epitopes recognized
by B cells have three-dimensional structures and/or are discontinuous in nature. Moreover,
new structural B cell epitopes can be created if structure of a protein is disrupted e.g. due to
oxidation or aggregation.
Years of intensive studies revealed the mechanism of peptide:MHC II interaction. The most
32
Immunogencity of therapeutic proteins.
important amino acids in the peptide sequence, which determine binding to the MHC II
pocket, are called anchoring residues. The remaining amino acids are recognized by the TCR
and therefore create T cell epitopes. At the moment there are three types of computational
models that are used for the prediction of peptide:MHC II complex formation. Most of the in
silico statistical methods calculate the probability of peptide:MHC II complex formation on
the basis of known interactions between different amino acid sequences and HLA allotypes.
Less sophisticated methods like the Syfpeithi 58 database rely only on sequence similarities
between known binders and potentially immunogenic, studied antigens. Other developed
tools create a scoring matrix that not only is based on the similarities between binding
peptides, but also scores each amino acid depending on its position and character. Such an
approach is applied in tools like Tepitope 59 or Rankprep 60. More recent algorithms like
ANNprep and Comprep 61, which are based on artificial neural networks (ANNs), have been
developed. The superiority of ANN-based tools over score matrices is due to their ability to
adjust the identification of immunogenic epitopes under impact of newly acquired data on
immunogenic epitopes. Therefore ANNs can be used not only for epitope prediction, but
also to search for unknown binding patterns. The last and the most recent in silico tools are
structure based methods, which aim to predict the possibility of peptide binding on the basis
of the three dimensional structure of the binding grooves of MHC II complexes derived
from crystallographic studies. Therefore such tools, e.g. Epibased 62, are less dependent on
the experimental data input than other in silico tools.
Whereas T cells recognize only linear, outstretched sequences, 90% of the epitopes
recognized by B cells are three dimensional structures 63. Although most of the in silico tools
focus on the prediction of T cell epitopes, there are some models attempting to predict B cell
epitopes, for example Preditop 64 and Bepitop 65. They use propensity scales, corresponding
to hydrophilicity, accessibility, flexibility, or secondary structure propensities, to assess
which amino acids on the surface of an antigen are most likely involved in the epitope.
Generally, prediction of B cell epitopes lacks power since even a slight change in tertiary or
quaternary structure of the protein can lead to formation of neoepitopes.
In general in silico methods for T and B cell epitope prediction are a promising field, which
will definitely allow to significantly reduce the number of potential drugs for further in
vitro and in vivo validation in the future. Nevertheless, even the most sophisticated in silico
tools will not eliminate the necessity of in vitro and in vivo methods for immunogenicity
screening of therapeutic proteins. So far, in silico methods have a tendency to over-estimate
the number of potential immunogenic peptides. For example, they do not take into account
that some epitopes can be hidden inside the core of the protein or covered by polysaccharides.
Such an epitope can give a high immunogenic score in silico, but this will not be confirmed
in vivo. Moreover, the influence of other cells, e.g. regulatory cells, is omitted.
Although the prediction of T- or B cell epitopes are the most often referred tools for
immunogenicity prediction and reduction, the increasing understanding of the immunogenic
33
Chapter 2.
potency of aggregates has led to the incorporation of algorithms designed to predict
aggregate prone sequences (APRs) of proteins into the panel of immunogenicity prediction
tools. Two main approaches are used to predict the ability of the protein to form aggregates.
The first approach, e.g. Aggrescan 66, TANGO 67 and PAGE 68, base the prediction only on
the amino acid sequence of the protein. Less sophisticated algorithms (e.g. Aggrescan),
base their prediction on sequences that are known to form β-cross aggregates, i.e. they
score the susceptibility to form aggregates by assessing the possibility of forming cross
reactive β-sheets. More sophisticated tools belonging to the first approach include also the
character of every amino acids residue (e.g. possession of aromatic ring, charge, solubility
and hydrophobicity) which builds the linear peptide of a minimum length of 5-9 amino
acids. The second approach creates algorithms based on molecular simulation techniques.
They are designed to identify the hydrophobic patches on the surface of the protein, e.g.
SAP 69 and a method described by Pechmann 70. In contrast to prediction tools from the
first group (e.g. Aggrescan or Tango), SAP and Pechmanns’ algorithm can predict potential
hydrophobic patches also in form of the three dimensional structures (e.g. on the surface of
protein molecule) and not only as the linear sequence.
5.2. In vitro tools.
The majority of in vitro assays, similar to the computational methods, are based on
assumptions that immunogenicity against therapeutic proteins is T cell dependent and,
hence, that protein drugs have to possess T cells epitopes. A number of different approaches
have been designed to study T cell activation or antigen binding to class II MHC class II in
vitro.
In order to validate in silico predictions, two types of MHC II-based in vitro assays can
be used. The first is to assess the possibility of binding of the potentially immunogenic
peptides to MHC class II molecules, which can be performed by incubation of peptides
with a lymphoblastoid cell line of B cells 71. Since those cells express different MHC
class II alleles, it is possible to predict which sequence fragment of the protein is most
immunogenic as well as which haplotype is correlated with increased risk of an unwanted
immune response. In the second in vitro assay, which is a competition epitope binding
assay, recombinant MHC class II is immobilized on a plate and then incubated with class
II MHC:epitope complexes 71,72. If the affinity of an epitope is higher for the immobilized
allele compared to the soluble one, the peptide will be detected on the plate after washing.
That will provide information on which MHC class II haplotype might be correlated
with a higher risk of developing an immune response against the epitope. Although the
information derived from both assays could be very valuable in predicting and validating
T cell epitopes, both approaches have important drawbacks. They are time and material
consuming, covering of all MHC class II haplotypes seems to be unlikely, and, especially
in the case of the competition assays, they are performed in non physiological conditions
which can influence the outcome. Moreover, like in silico assays, these tests only focus on
one aspect of product-related factors contributing to immunogenicity.
34
Immunogencity of therapeutic proteins.
In contrast to the assays described in the previous paragraph, T-cell based in vitro assays are
less complex and can give wider prediction potential since not only separate peptides but
also intact proteins, including formulated ones, can be used.
In T-cell based in vitro assays, T cells are isolated from the blood of a naive or drug exposed
subject and are used alone or mixed with other peripheral blood mononuclear cells (PBMCs)
71,73
. While cells from a naive subject can give information about immunogenic properties
of a protein, cells from a previously exposed patient can be used to detect and expand
memory T cells. Then, depending on the approach, protein or peptide-induced proliferation
of T cells can be measured, for example by measuring the uptake of labeled substrates
(e.g. radio-labeled thymidine) incorporated into newly formed cells. In addition, protein or
peptide induced production of activation markers such as IL-2 or IFNγ can be measured.
While the proliferation assay gives only quantitative information on the proliferation of
T-cells and thus which proteins or peptides can activate T-cells, measuring which activation
markers (e.g. cytokines) are produced gives additional data about the nature of the immune
response. Irrespective of the method used, the T cell response to the protein of interest has
to be compared to the basic activity of non-stimulated cells obtained from the same subject
(since T cell activation state can differ between individuals) 71,73.
Despite many advantages of the present T-cell assays, some drawbacks have to be pointed
out. Similarly to the MHC class II assays, full coverage of MHC II haplotypes is difficult
to obtain. Also, since in general T cells are less susceptible for development of response in
vitro than in vivo, depletion of suppressing T cells is often necessary to unmask the response
of other kinds of T cells.
A very interesting strategy for the discovery of T cell epitopes, MHC-Associated
Peptide Proteomics (MAPPs), was established by Röhn et al 74. In this technique, class
II MHC:epitope complexes are isolated from dendritic cells (DCs) that were previously
stimulated with a protein in vitro or in vivo. The peptides are then dissociated from the
MHCII molecule and analysed in a system composed of a two-dimensional capillary liquid
chromatography and tandem mass spectrometry system (LC-MS/MS). The most powerful
features of this technique are: (i) it provides the researcher with the immunodominant
epitopes, (ii) it takes into account the post-translational modifications of MCH II molecule,
which may affect the peptide recognition and/or selection, (iii) it includes those MHC II
alleles which (still) are not covered by in silico tools, and (iv) it can be performed on fully
formulated protein 74,75. The reliability of MAPPs technique requires trained personnel as
well as complex calibration and validation of analytic system. Moreover, due to patients’
peripheral tolerance, MAPPs always generate few percent of false positives. MAPPs can
also give false negatives because of mass spectrometry detection system limitations, e.g.
2-3% of peptides cannot be detected by ion trap mass spectrometry 75.
In the previous paragraphs, the in vitro methods for T-cell epitope prediction are described.
The identification of B cell epitopes in cell-based approaches is also possible, but it is
35
Chapter 2.
burdened with many disadvantages. The most important one, which practically eliminates
B cell assays from the set of in vitro prediction tools, is the nature of B cell activation and
the fact that in vitro responses of B cells are much weaker than in vivo responses. Only
potent TI antigens like LPS or TNP-Ficoll are potent enough to measurably activate naïve B
cells in vitro 76. In case of weaker TI antigens, a costimulation from other agent/cell type is
needed in order to obtain measurable activation of B-cells. It was shown that the presence of
T cells increases the speed of TI B cell activation, however addition of T cells into a B-cell
assay makes it impossible to distinguish between direct activation of B cells by the protein
(aggregates) or activation by MHC II derived or structural epitopes. Therefore, B cell based
assays are not commonly used as prediction tools.
5.3. In vivo tools.
Immunogenicity prediction might also be performed in vivo. In general, animal models,
especially non-human primates and transgenic rodents, allow the most accurate prediction
of immunogenic properties of therapeutic drugs compared to in silico and in vitro tools. In
silico and in vitro models simplify the immune system, and therefore may lead to a wrong
estimation of immunogenic potential 73. Moreover, the physico-chemical characteristics of a
drug can change after administration. This is not detected in in silico and in vitro models. The
number of unknown variables related to drug administration make it impossible to mimic
this process in silico or in vitro with a 100% accuracy. For example, the micro environment
of the injection spot can have a tremendous impact on immunogenicity; different routes of
administration can be compared in vivo but cannot be recreated in vitro 18. Therefore animal
models still remain an important part of immunogenicity prediction.
Mouse and rat models have been widely used to assess immunogenic properties of protein
drugs. However, the prediction power of those models is rather weak. Since human proteins
are exogenous antigens, animals will generally develop a classical TD immune response.
Therefore, direct extrapolation of immunogenicity of human-derived or recombinant
proteins from those models to humans is difficult and unreliable. In contrast, non-human
primates were found to be useful in the prediction of the relative immunogenicity of some
human proteins like growth hormone or thrombophoietin 77. However, the reliability of nonhuman primates is limited to proteins with conserved structure.
Better models to study immunogenicity are transgenic animals. Two different transgenic
models are nowadays commonly used in pre-clinical studies: human HLA transgenic mouse
models and animals transgenic for a particular protein of interest. The HLA transgenic
models are animals with a deficient murine MHC class II that do express human HLA
alleles. These mice are routinely used for the evaluation of an immune response against
vaccines 78,79. They show a clear correlation to effects in humans in terms of a vaccine
induced T dependent immune response. This correlation suggests that these animals might
also have predictive value when assessing immunogenicity of therapeutic drugs of both
human and non-human origin. The existence of several of transgenic strains expressing the
36
Immunogencity of therapeutic proteins.
most common HLA alleles allows us to correlate HLA haplotype with a potential risk of
treatment. However, one has to keep in mind that all human therapeutic proteins are external
antigens for HLA transgenic animals, therefore direct extrapolation of the immunogenicity
from those animals to patients cannot be done.
The other group of transgenic animals models in immunogenicity research includes mice
designed to be immune tolerant to a native human protein of interest. These mice have the
human gene of the protein of interest encoded in their genome and therefore express this
human protein. However, by treating these transgenic mice with the protein drug of interest,
their tolerance against the protein can be broken and antibody production is initiated.
Importantly, the mechanism of the antibody response in transgenic mice is likely similar to
that in humans.
Several immune tolerant transgenic animal models have been used to predict
immunogenicity. In addition, they are used to evaluate critical product-related parameters
leading to the immune response against protein drugs. The first animal models implemented
in immunogenicity research were transgenic mice producing human plasminogen activator
(tPA) 80. Although these animals were tolerant to the natural tPA, they were capable of
producing antibodies against a modified form of tPA (single amino acid substitution). Since
then several other transgenic mouse models were developed and validated. For example,
studies using transgenic animals tolerant for human IFNα 81, IFNβ 82, and insulin 83 have
been and are being carried out. Several of those studies revealed that presence of aggregates
is an important risk factor for unwanted immune responses. Both IFNα and IFNβ, when
present in monomer form, are not able to induce an immune response in transgenic mice.
In contrast, some, but not all, formulations containing aggregated forms of those proteins
elicit antibodies.
Although transgenic animals expressing human proteins of interest are valuable models
for immunogenicity studies, they still have some important disadvantages. In contrast to
HLA transgenic animals they express murine MHC II alleles, therefore even though they
are tolerant for a human protein, during the immune response against aggregated or altered
protein a different pattern of epitopes can be recognized compared to patients. Moreover,
many available therapeutic drugs contain HSA as stabilizer. During the assessment of
immunogenic properties of HSA-containing products caution is needed as the impact of
HSA (a foreign protein for the transgenics, but obviously not for human patients) on the
immune response against the protein drug is not known.
6. How to decrease/eliminate the immunogenicity of therapeutic proteins?
The existence of even the most sophisticated and accurate tools to predict immunogenicity
would be worthless without the availability of techniques that allow elimination of
immunogenic sites. Before the age of recombinant technology, the immunogenic
potential of protein drugs could only be reduced by optimizing the production process,
including formulation, and assuring proper storage and administration conditions. In fact,
37
Chapter 2.
immunogenicity of the first protein drugs was significantly reduced after improving their
purification and storage conditions.
The development of recombinant technology opened a new possibility of rational protein
modification in order to achieve higher stability and reduced immunogenicity. In combination
with solving the sequence of proteins (at both DNA and amino acid level), recombinant
technology allows direct elimination of the most immunogenic parts of the protein sequence
which are thought to be responsible for triggering the immune response.
6.1. Elimination of T cell epitopes.
The elimination of T cell epitopes is one of the most potent ways of limiting immunogenicity
of therapeutic proteins e.g. used for reducing immunogenicity of factor VIII 84,85. Although
this method has been mainly applied for non-human protein drugs, in principle it can be
used for all types of protein drugs. While non-human proteins contain a number of potential
MHC II binding peptide sequences, usually only few of them are dominant. A small change
in these potential epitopes, such as replacing one amino-acid residue, can significantly
reduce its recognition by immune cells. For example, a single amino acid substitution
(glutamic acid to serine or alanine) in the T cell epitope of staphylokinase lowered the
titers of produced antibodies by 35% 86. Also immunogenicity of the human derived or the
recombinant human proteins can be lowered by applying directed mutagenesis to remove
potentially immunogenic epitopes. A good example is human derived IFNβ, in which the
possibility of T cell activation in vitro was inhibited by a point mutation 87. Moreover,
immune response of BALB/cByJ mice against mutated human IFNβ was significantly
lower than against non-mutated form 80. However, lower immunogenicity of this mutated
protein has not been confirmed in patients.
During T cell epitope removal caution should be applied since every mutation in the protein
might change the conformation, the activity, or increase susceptibility to aggregation.
In case of monoclonal antibodies another approach has been undertaken to lower
immunogenicity. Because many monoclonal antibodies were (partially) foreign, these
proteins are subjected to humanization, a process during which almost the whole murine
sequence, except complementarity-determining regions (CDRs), is replaced with its human
counterpart 84,88. Monoclonal antibodies are especially suitable for that modification due
to their characteristic, conserved structure including the possession of Fab and Fc regions.
Since the specificity and the affinity of an antibody are dependent only on the sequence and
conformation of Fab region, (especially) the rest of residues can be modified in order to
reduce its immunogenicity or increase its stability 89. Modification of amino acid residues
not involved in epitope binding is a common approach of increasing stabilization and safety
of monoclonal antibodies. Humanization of murine antibodies by complete exchange of
the murine Fc region by its human counterpart was the first successful attempt in reducing
immunogenicity of the first therapeutic monoclonal antibodies. The humanization of these
chimeric monoclonal antibodies was pushed even further by the development of humanized
38
Immunogencity of therapeutic proteins.
antibodies, with only the CDRs being non-human, and fully human antibodies. However,
one has to keep in mind that even a fully human antibody contains “foreign” CDRs, which
cannot be removed without loss of the antibody’s activity 89.
6.2. Elimination of aggregation-prone sequences.
The basic approach underlying elimination of aggregation-prone sequences is similar as in
T cell epitope removal. Directed mutations allow changing the nature of aggregate-prone
sequences and thereby limit the possibility to form aggregates. Elimination of aggregateprone sequences suffers from the same drawbacks as elimination of T cell epitopes.
Any modification of the protein structure can influence the specificity and activity, and
conformation of the protein. A change in protein conformation can lead to the formation of
neo-epitopes, which can be more efficiently recognized by immune cells. Elimination of
aggregate-prone sites is especially complicated in monoclonal antibodies as they are often
localized in the CDRs 90.
7. Conclusions.
In the near future therapeutic proteins will dominate the drug market. However, to
fully benefit from their therapeutic potential a deeper knowledge about all aspects of
immunogenicity is needed. During the past few years great progress in understanding
the importance of the immunogenicity and accompanied risk factors has been made. The
development of new, more sensitive assays for immunogenicity development revealed the
commonness of unwanted immune responses even for drugs which until recently were
considered hardly immunogenic. Despite this progress the exact immunological mechanism
of immunogenicity of especially recombinant proteins, still remains poorly understood and
further intensive studies are needed in order to understand mechanism of immunogenicity
and rationally design less immunogenic protein therapeutics.
39
Chapter 2.
References.
1.
Sominanda A, Rot U, Suoniemi M, Deisenhammer F, Hillert J, Fogdell-Hahn A. 2007.
Interferon beta preparations for the treatment of multiple sclerosis patients differ in
neutralizing antibody seroprevalence and immunogenicity. Mult Scler 13:208-214
2.
Jacquemin MG, Saint-Remy JM. 1998. Factor VIII immunogenicity. Haemophilia 4:552557
3.
Milner RD. 1985. Growth hormone Brit Med J 291:1593-1594
4.
Schellekens H. 2003. Immunogenicity of therpeutic proteins Nephrol Dial Transpl 18:
1257-1259
5.
van Schouwenburg PA, Bartelds GM, Hart MH, Aarden L, Wolbink GJ, Wouters D. 2010. A
novel method for the detection of antibodies to adalimumab in the presence of drug reveals
“hidden” immunogenicity in rheumatoid arthritis patients. J Immunol Methods 362:82-88
6.
Chamberlain P. 2002. Immunogenicity of therapeutic proteins Part 1: Causes and clinical
manifestations of immunogenicity. Regulatory Rev 5:4-9
7.
Moore WV, Leppert P. 1980. Role of aggregated human growth hormone (hGH) in
development of antibodies to hGH. J Clin Endocrinol Metab 51:691-697
8.
van der Voort LF, Gilli F, Bertolotto A, Knol DL, Uitdehaag BM, Polman CH, Killestein
J. 2010 Clinical effect of neutralizing antibodies to interferon beta that persist long after
cessation of therapy for multiple sclerosis. Arch Neurol 67:402-407
9.
Pollock C, Johnson DW, Hörl WH, Rossert J, Casadevall N, Schellekens H, Delage R, De
Francisco A, Macdougall I, Thorpe R, Toffelmire E. 2008. Pure red cell aplasia induced by
erythropoiesis-stimulating agents. Clin J Am Soc Nephrol 3:193-199
10.
Polman CH, Bertolotto A, Deisenhammer F, Giovannoni G, Hartung HP, Hemmer
B, Killestein J, McFarland HF, Oger J, Pachner AR, Petkau J, Reder AT, Reingold SC,
Schellekens H, Sørensen PS. 2010. Recommendations for clinical use of data on neutralising
antibodies to interferon-beta therapy in multiple sclerosis. Lancet Neurol 9:740-750
11.
Markham A, Lamb HM. 2000. Infliximab: a review of its use in the management of
rheumatoid arthritis. Drugs 59:1341-1359
12.
Ludwig H, Fritz E, Leitgeb C, Krainer M, Kührer I, Sagaster P, Umek H. 1992 Erythropoietin
treatment for chronic anemia of selected hematological malignancies and solid tumors. Ann
oncol 4:161-167
13.
Macleod JJ. 1922. Insulin and diabetes: A general statment of the physiological and
therapeutic effects of insulin. Br Med J 2:833-835
14.
Smith CI, Kitchen LW, Scullard GH, Robinson WS, Gregory PB, Merigan TC. 1982.
Vidarabine monophosphate and human leukocyte interferon in chronic hepatitis B infection.
JAMA 247:2261-2265
15.
Casadevall N, Nataf J, Viron B, Kolta A, Kiladjian JJ, Martin-Dupont P, Michaud P, Papo T,
Ugo V, Teyssandier I, Varet B, Mayeux P. 2002. Pure red-cell aplasia and antierythropoietin
antibodies in patients treated with recombinant erythropoietin. N Engl J Med 346:469-475
16.
Sharma B. 2007. Immunogenicity of therapeutic proteins. Part 2: Impact of container
closures. Biotechnol Adv 25:318-324
17.
Schellekens H. 2007. Lesson learned from Eprex-associated pure red cell aplasia. Kidney
Blood Press Res 30:9-12
18.
Ross C, Clemmesen KM, Svenson M, Sørensen PS, Koch-Henriksen N, Skovgaard GL,
Bendtzen K. 2000. Immunogenicity of interferon-beta in multiple sclerosis patients:
influence of preparation, dosage, dose frequency, and route of administration. Danish
40
Immunogencity of therapeutic proteins.
Multiple Sclerosis Study Group. Ann Neurol 48:706-712
19.
Bertolotto A, Deisenhammer F, Gallo P, Sölberg Sørensen P. 2004. Immunogenicity of
interferon beta: differences among products. J Neurol 251:II/15-II/24
20.
van Beers MM, Jiskoot W, Schellekens H. 2010. On the role of aggregates in the
immunogenicity of recombinant human interferon beta in patients with multiple sclerosis. J
Interferon Cytokine Res 30:767-75
21.
Hartung HP, Polman C, Bertolotto A, Deisenhammer F, Giovannoni G, Havrdova E,
Hemmer B, Hillert J, Kappos L, Kieseier B, Killestein J, Malcus C, Comabella M, Pachner
A, Schellekens H, Sellebjerg F, Selmaj K, Sorensen PS. Neutralising antibodies to interferon
beta in multiple sclerosis: expert panel report. J Neurol 254:827-837
22.
Herndon RM, Rudick RA, Munschauer FE 3rd, Mass MK, Salazar AM, Coats ME, Labutta
R, Richert JR, Cohan SL, Genain C, Goodkin D, Toal M, Riester K. 2005. Eight-year
immunogenicity and safety of interferon beta-1a-Avonex treatment in patients with multiple
sclerosis. Mult Scler 11:409-419
23.
Dillman RO. 1999. Infusion reactions associated with the therapeutic use of monoclonal
antibodies in the treatment of malignancy. Cancer Metastasis Rev 18:465-471
24.
Kang SP, Saif MW. 2007. Infusion-related and hypersensitivity reactions of monoclonal
antibodies used to treat colorectal cancer--identification, prevention, and management. J
Support Oncol 5:451-457
25.
Cassinotti A, Travis S. 2009. Incidence and clinical significance of immunogenicity to
infliximab in Crohn’s disease: a critical systematic review. Inflamm Bowel Dis 15:12641275
26.
West RL, Zelinkova Z, Wolbink GJ, Kuipers EJ, Stokkers PC, van der Woude CJ. 2008.
Immunogenicity negatively influences the outcome of adalimumab treatment in Crohn’s
disease. Aliment Pharmacol Ther 28:1122-1126
27.
Hwang CJ, Vaccaro AR, Lawrence JP, Hong J, Schellekens H, Alaoui-Ismaili MH, Falb D.
2009. Immunogenicity of bone morphogenetic proteins. J Neurosurg Spine 10: 443-451
28.
Seong SY, Matzinger P. 2004. Hydrophobicity: an ancient damage-associated molecular
pattern that initiates innate immune responses. Nat Rev Immunol 4:469-478
29.
Melchers F, Rolink AR. 2006. B cell tolerance-how to make it and how to break it. Curr Top
Microbiol Immunol 305:1-23
30.
Li L, Boussiotis VA. 2006. Physiologic regulation of central and peripheral T cell tolerance:
lessons for therapeutic applications. J Mol Med 84:887-99
31.
Singh SK. 2009. Impact of product-related factors on the immunogenicity of biotherapeuticals.
J Pharm Sci 354–387
32.
van Beers MM, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
33.
Goldsby, R., Kindt, T., Osborne, R., Kuby, J. 2006 Immunology, ed. 6th: W.H. Freeman &
Co.
34.
Vos Q, Lees A, Wu ZQ, Snapper CM, Mond JJ. 2000. B-cell activation by T-cell-independent
type 2 antigens as an integral part of the humoral immune response to pathogenic
microorganisms. Immunol Rev 176:154-170
35.
Vos Q, Snapper CM, Mond JJ. 2000. Th1 versus Th2 cytokine profile determines the
modulation of in vitro T cell-independent type 2 responses by IL-4. Int. Immunol 12: 13371345
41
Chapter 2.
36.
Snapper CM, Kehry MR, Castle BE, Mond JJ. 1995. Multivalent, but not divalent, antigen
receptor cross-linkers synergize with CD40 ligand for induction of Ig synthesis and class
switching in normal murine B cells. A redefinition of the TI-2 vs T cell-dependent antigen
dichotomy. J Immunol 154:1177-1187
37.
Obukhanych TV, Nussenzweig MC. 2006. T-independent type II immune responses generate
memory B cells. J Exp Med 203:305–310
38.
Fakharzadeh SS, Kazazian HH Jr. 2000. Correlation between factor VIII genotype and
inhibitor development in hemophilia A. Semin Thromb Hemost 26:167-171
39.
Richards SM, 2002. Immunologic considerations for enzyme replacement therapy in the
treatment of lysosomal storage disorders. Appl Immunol Rev 2:241-253
40.
Lidove O, West ML, Pintos-Morell G, Reisin R, Nicholls K, Figuera LE, Parini R, Carvalho
LR, Kampmann C, Pastores GM, Mehta A. 2010. Effects of enzyme replacement therapy in
Fabry disease--a comprehensive review of the medical literature. Genet Med 12:668-679
41.
Reeves WG, Gelsthorpe K, Van der Minne P, Torensma R, Tattersall RB. 1984. HLA
phenotype and insulin antibody production. Clin Exp Immunol 57:443-448
42.
Barbosa MD, Vielmetter J, Chu S, Smith DD, Jacinto J. 2006. Clinical link between MHC
class II haplotype and interferon-beta (IFN-B) immunogenicity. Clin Immunol 118:42-50
43.
Antonelli G, Dianzani F. 1999. Development of antibodies to interferon beta in patients:
technical and biological aspects. Eur Cytokine Netw 10:413-422
44.
Schellekens H. 2002. Immunogenicity of therapeutic proteins: clinical implications and
future prospects. Clin Ther 24:1720-1740
45.
Fierlbeck G, Schreiner T, Schaber B, Walser A, Rassner G. 1994. Neutralizing interferon
beta antibodies in melanoma patients treated with recombinant and natural interferon beta.
Cancer Immunol Immunother 39:263-269
46.
Prtimmer O. 1997. Treatment-induced antibodies to interleukin-2. Biotherapy 10:15-24
47.
Palleroni AV, Aglione A, Labow M, Brunda MJ, Pestka S, Sinigaglia F, Garotta G, Alsenz J,
Braun A. 1997. Interferon immunogenicity: preclinical evaluation of interferon-alpha 2a. J
Interferon Cytokine Res 17:s23-s27
48.
Johnston MJ, Nemr K, Hefford MA. 2010. Influence of bovine serum albumin on the
secondary structure of interferon alpha 2b as determined by far UV circular dichroism
spectropolarimetry. Biologicals 38:314-320
49.
Philo JS, Arakawa T. 2009. Mechanisms of protein aggregation. Curr Pharm Biotechnol
10:348-35.
50.
Mahler HC, Friess W, Grauschopf U, Kiese S. S. 2009. Protein aggregation: pathways,
induction factors and analysis. J Pharm Sci 98:2909-2934
51.
Pisal DS, Kosloski MP, Balu-Iyer SV. 2010. Delivery of Therapeutic Proteins. J Pharm Sci
99: 2557-2575
52.
Chirino AJ, Ary ML, Marshall SA. 2004. Minimizing the immunogenicity of protein
therapeutics. Drug Discov Today 9:82-90
53.
European Medicines Agency, 2007. Guideline on immunogenicity of biotechnology derived
therapeutic proteins.
54.
Food and Drug Administration, 2009. Guidance for industry assay development for
immunogenicity testing of therapeutic proteins.
55.
Perini P, Facchinetti A, Bulian P, Massaro AR, Pascalis DD, Bertolotto A, Biasi G, Gallo
P. 2001. Interferon-beta (INF-beta) antibodies in interferon-beta1a- and interferon-beta1btreated multiple sclerosis patients. Prevalence, kinetics, cross-reactivity, and factors
42
Immunogencity of therapeutic proteins.
enhancing interferon-beta immunogenicity in vivo. Eur Cytokine Netw 12:56-61
56.
Miyasaka N, CHANGE Investigators. 2008. Clinical investigation in highly diseaseaffected rheumatoid arthritis patients in Japan with adalimumab applying standard and
general evaluation: the CHANGE study. Mod Rheumatol 18:252-262
57.
De Groot AS, McMurry J, Moise L. 2008. Prediction of immunogenicity: in silico paradigms,
ex vivo and in vivo correlates. Curr Opin Pharmacol 8:620-626
58.
Rammensee H, Bachmann J, Emmerich NP, Bachor OA, Stevanović S. 1999. SYFPEITHI:
database for MHC ligands and peptide motifs. Immunogenetics 50:213-219
59.
Bian H, Hammer J. 2004. Discovery of promiscuous HLA-II-restricted T cell epitopes with
TEPITOPE. Methods 34:468-475
60.
Reche PA, Glutting JP, Zhang H, Reinherz EL. 2004. Enhancement to the RANKPEP resource
for the prediction of peptide binding to MHC molecules using profiles. Immunogenetics
56:405-419
61.
Lata S, Bhasin M, Raghava GP. 2007. Application of machine learning techniques in
predicting MHC binders. Methods Mol Biol 409:201-251
62.
Desmet J, Meersseman G, Boutonnet N, Pletinckx J, De Clercq K, Debulpaep M, Braeckman
T, Lasters I. 2005. Anchor profiles of HLA-specific peptides: analysis by a novel affinity
scoring method and experimental validation. Proteins 58:53-69
63.
Ansari HR, Raghava GP.2010. Identification of conformational B-cell Epitopes in an antigen
from its primary sequence. Immunome Res 6
64.
Pellequer JL, Westhof E. 1993. PREDITOP: a program for antigenicity prediction. J Mol
Graph 11:204-10
65.
Odorico M, Pellequer JL. 2003. BEPITOPE: predicting the location of continuous epitopes
and patterns in proteins. J Mol Recognit 16:20-22
66.
Conchillo-Solé O, de Groot NS, Avilés FX, Vendrell J, Daura X, Ventura S. 2007.
AGGRESCAN: a server for the prediction and evaluation of “hot spots” of aggregation in
polypeptides. BMC Bioinformatics 8:65
67.
Fernandez-Escamilla AM, Rousseau F, Schymkowitz J, Serrano L. 2004. Prediction of
sequence-dependent and mutational effects on the aggregation of peptides and proteins. Nat
Biotechnol 22:1302-1306
68.
Tartaglia GG, Cavalli A, Pellarin R, Caflisch A. 2005. Prediction of aggregation rate and
aggregation-prone segments in polypeptide sequences. Protein Sci 14:2723-2734
69.
Chennamsetty N, Voynov V, Kayser V, Helk B, Trout BL. 2009. Design of therapeutic
proteins with enhanced stability. Proc Natl Acad Sci USA 106:11937-11942
70.
Pechmann S, Vendruscolo M. 2010. Derivation of a solubility condition for proteins from
an analysis of the competition between folding and aggregation. Mol Biosyst 6: 2490-2497
71.
De Groot AS, Moise L. 2007. Prediction of immunogenicity for therapeutic proteins: state
of the art. Curr Opin Drug Discov Devel 10:332-340
72.
McMurry JA, Gregory SH, Moise L, Rivera D, Buus S, De Groot AS. 2007. Diversity of
Francisella tularensis Schu4 antigens recognized by T lymphocytes after natural infections
in humans: Identification of candidate epitopes for inclusion in a rationally designed
tularemia vaccine. Vaccine 25:3179-3191
73.
Hobeika AC, Morse MA, Osada T, Ghanayem M, Niedzwiecki D, Barrier R, Lyerly
HK, Clay TM. 2005. Enumerating antigen-specific T-cell responses in peripheral blood:
a comparison of peptide MHC Tetramer, ELISpot, and intracellular cytokine analysis. J
Immunother 28:63-72
43
Chapter 2.
74.
Röhn TA, Reitz A, Paschen A, Nguyen XD, Schadendorf D, Vogt AB, Kropshofer H. 2005.
A novel strategy for the discovery of MHC class II-restricted tumor antigens: identification
of a melanotransferrin helper T-cell epitope. Cancer Res 65:10068-10078
75.
Kropshofer H, Singer T. 2006. Overview of cell-based tools for pre-clinical assessment of
immunogenicity of biotherapeutics. J Immunotoxicol 3:131-136
76.
Mond JJ, Vos Q, Lees A, Snapper CM. 1995. T cell independent antigens. Curr. Opin.
Immunol 7:349-354
Wierda D, Smith HW, Zwickl CM. 2001. Immunogenicity of biopharmaceuticals in
laboratory animals. Toxicology 158:71-74
77.
78.
Man S, Newberg MH, Crotzer VL, Luckey CJ, Williams NS, Chen Y, Huczko EL, Ridge JP,
Engelhard VH. 1995 Definition of a human T cell epitope from influenza A non-structural
protein 1 using HLA-A2.1 transgenic mice. Int Immunol 7:597-605
79.
Depil S, Angyalosi G, Moralès O, Delacre M, Delhem N, François V, Georges B, Hammer J,
Maillère B, Auriault C, Pancré V. 2006. Peptide-binding assays and HLA II transgenic Abeta
degrees mice are consistent and complementary tools for identifying HLA II-restricted
peptides. Vaccine 24:2225-2229
80.
Stewart TA, Hollingshead PG, Pitts SL, Chang R, Martin LE, Oakley H. 1989. Transgenic
mice as a model to test the immunogenicity of proteins altered by site-specific mutagenesis.
Mol Biol Med 6:275-281
81.
Hermeling S, Schellekens H, Maas C, Gebbink MF, Crommelin DJ, Jiskoot W. 2006.
Antibody response to aggregated human interferon alpha2b in wild-type and transgenic
immune tolerant mice depends on type and level of aggregation. J Pharm Sci 95:1084-1096
82.
van Beers MM, Sauerborn M, Gilli F, Hermeling S, Brinks V, Schellekens H, Jiskoot W.
2010. Hybrid transgenic immune tolerant mouse model for assessing the breaking of B cell
tolerance by human interferon beta. J Immunol Methods 352:32-37
83.
Ottesen JL, Nilsson P, Jami J, Weilguny D, Dührkop M, Bucchini D, Havelund S, Fogh JM.
1994. The potential immunogenicity of human insulin and insulin analogues evaluated in a
transgenic mouse model. Diabetologia 12:1178-1185
Jones TD, Crompton LJ, Carr FJ, Baker MP. 2009. Deimmunization of monoclonal
antibodies. Methods Mol Biol 525:405-423
84.
85.
Song C, Martin W, DeGroot A, Scott D. 2010. Deimmunization of HLA-DR epitopes in the
C2 domain of human FVIII. J Immunol 184
86.
He J, Xu R, Chen X, Jia K, Zhou X, Zhu K. 2010. Simultaneous elimination of T-cell and
B-cell epitope by structure-based mutagenesis of single Glu80 resiude within recombinant
staphylokinase. Acta Biochim Biophys Sin 42:209-215
87.
Yeung PV, Chang J, Miller J, Barnett C, Stickler M, Harding FA. 2004. Elimination of
an immunodominant CD4+ T cell epitope in human IFN-B does not result in an in vivo
response directed at the subdominant epitope. J Immunol 172:6658-6665
Roque-Navarro L, Mateo C, Lombardero J, Mustelier G, Fernández A, Sosa K, Morrison
SL, Pérez R. 2003. Humanization of predicted T-cell epitopes reduces the immunogenicity
of chimeric antibodies: new evidence supporting a simple method. Hybrid Hybridomics
22:245-257
88.
89.
Harding FA, Stickler MM, Razo J, DuBridge RB. 2010. The immunogenicity of humanized
and fully human antibodies: residual immunogenicity resides in the CDR regions. MAbs
2:256-265
90.
Wang X, Singh SK, Kumar S. 2010. Potential aggregation-prone regions in complementaritydetermining regions of antibodies and their contribution towards antigen recognition: a
computational analysis. Pharm Res 27:1512-1529
44
CHAPTER 3.
Immunogenicity of recombinant human
interferon beta in immune tolerant mice is
characterized by early involvement of CD4+
T cells, formation of germinal centers, but no
apparent immunological memory.
Grzegorz Kijanka
Melody Sauerborn
Louis Boon
Huub Schellekens
Vera Brinks
Submitted for publication.
45
Chapter 3.
Abstract.
The mechanisms underlying immunogenicity of recombinant human proteins remain
largely unknown. Animal and clinical studies revealed that immunogenicity shares both T
cell dependent (requirement of CD4+ T cells and isotype switching) and T cell independent
(involvement of Marginal Zone B cells and apparent lack of memory) characteristics.
We used transgenic immune tolerant mice to study the mechanisms underlying
immunogenicity of recombinant human interferon beta (rhIFNβ) in more detail and to
follow up our previous results. By removal of CD4+ T cells at different time points during
rhIFNβ treatment we studied when these cells are important for antidrug antibody formation.
We also investigated whether an immunological memory response could be induced by
increased exposure to antigen. Finally, we studied if germinal centers were formed as part
of the immune response against rhIFNβ.
We found that CD4+ T cells are important very early during antibody responses against
rhIFNβ, similar to a classical CD4+ T cell dependent immune response. In addition germinal
centers are formed as part of the response. Nonetheless, despite these two major features
of a classical immune response being present, no clear immunological memory response
could be induced at different antigen doses. Moreover, aluminum hydroxide was inefficient
in enhancing immune response towards rhIFNβ.
Our findings show that early response events of immunogenicity reaction share similarities
with CD4+ T cell dependent mechanism. However, lack of apparent memory after increased
exposure to rhIFNβ and inability of adjuvant to enhance response indicate deviations from
CD4+ T cell dependent process of B cell maturation.
46
1. Introduction.
CD4+ T cells, Germinal Centers and Memory.
Therapeutic proteins are a rapidly growing group of drugs, increasing both in number of
approved drugs as well as in their market value. Unfortunately the efficacy of protein drugs
is often limited by the formation of the anti-drug antibodies (ADA), which can alter the drug
half-life, decrease efficacy and sometimes lead to severe, life threatening side effects 1. ADA
response against therapeutic proteins is therefore a health risk and needs to be minimized.
ADA formation can only be reduced on the condition the underlying immune mechanism
is known. Two distinct mechanisms may lead to the production of antibodies. The first
mechanism is the classical immune response to foreign proteins. Such proteins can trigger
B cells to produce antibodies (Abs) via a CD4+ T cell dependent (TD) mechanism which is
characterized by isotype class switching and affinity maturation of secreted Abs, formation
of germinal centers (GCs) in lymph organs where maturation and differentiation of B cells
occurs and memory B cells are formed 2–5.
The second mechanism, T cell independent (TI), results in the direct activation of B
lymphocytes without necessity of CD4+ T cells involvement. A TI mechanism is usually
triggered by bacterial molecules such as CpG, LPS or polysaccharides. In contrast to the TD
mechanism, a Tind response leads to lack or reduced GC formation, is mainly restricted to
IgM and usually lacks the formation of immunological memory 6,7.
In previous studies we used immune tolerant mice to study the immune mechanisms
underlying immunogenicity of recombinant human interferon beta (rhIFNβ). Depending on
the product, rhIFNβ induces ADA production in up to 90% of patients 8–11 These antibodies
interfere with rhIFNβ’s efficacy, but it is unclear why and how these products cause ADA
formation. Both clinical studies as well as our studies in immune tolerant mice revealed
an immune response with both TD and TI characteristics. For example, both clinical and
mouse studies identified IgGs as the main ADA isotype 12–14. In immune tolerant mice CD4+
T cells were shown to be needed for production of ADA, suggesting a TD mechanism
underlying immunogenicity of rhIFNβ 14. However, other clinical and mouse data showed
features typical for a TI response such as an apparent lack of immunological memory and
involvement of marginal zone B cells 13–15.
In this report we describe a series of experiments designed to improve our understanding
of the mechanisms underlying immunogenicity of rhIFNβ using immune tolerant mice.
Because immune system of these mice is tolerant for hIFNβ, the development of antibody
response upon rhIFNβ treatment is supposed to mimic the immunogenicity reaction found
in patients 16. Moreover, as hIFNβ has been shown not to be active in mice, the immune
response is not biased by its activity which was shown to be different from murine IFNβ 17–
20
. We aimed to determine at which time point during ADA formation CD4+ T cells are most
crucial and whether an immunological memory formation could be induced by increasing
antigen exposure. Finally, we wanted to assess the formation of germinal centers to explain
47
Chapter 3.
the apparent lack of immunological memory found in our previous studies.
2. Material and Methods.
2.1. Animals.
Heterozygoous transgenic (tg) C57Bl/6 mice carrying the human interferon beta (hIFNβ)
gene originally developed in our institute were bred at the Central Laboratory Animal
Institute (Utrecht University, the Netherlands) 21. Tg C57Bl/6 males were crossbred with
wild-type FVB/N mice (Janvier, BioServices, Uden, The Netherlands) 22. The hIFNβ
genotype of the offspring (F1) was determined using PCR s in chromosomal DNA isolated
from ear tissue. Both PCR positive (tg) and PCR negative (non-tg) littermates of 6-10 weeks
of age were used for the experiments. Mice were housed in standard perspex cages and given
access to food (Hope Farms, Worden, The Netherlands) and water (acidified) ad libitum. All
experiments were performed according to Institutional Ethical Committee Regulations of
Utrecht University, the Netherlands.
2.2. Recombinant human interferon Beta.
Betaferon® (recombinant human IFN β-1b, rhIFNβ-1b drug product) was obtained from
Schering (Berlin, Germany). Lyophilised powder containing 300 μg of rhIFNβ, 15 mg of
human serum albumin (HSA) and 15 mg of mannitol was reconstituted in 1 ml of 10 mM
sodium phosphate pH 7.4, 137 mM sodium chloride (PBS). Before injection, Betaferon® was
further diluted to desired concentration in PBS. RhIFNβ-1a drug substance was supplied by
Biogen Idec (Cambridge, MA, US) and was formulated to a concentration of 270μg/ml in
100 mM sodium phosphate pH 7.2, 200 mM sodium chloride.
2.3.1. Experiment 1: Timing of CD4+ T-cell involvement in the ADA response against
Betaferon®.
Both tg and non-tg animals received intraperitoneal (IP) injections of 5 μg of Betaferon® in
100 μl of PBS for 3 weeks (on days 1-5, 8-12 and 15-19).
To study the involvement of CD4+ T cells, mice received IP injections of 100μg of anti
CD4 antibody GK1.5 (Bioceros, the Netherlands) in 100 μl of PBS on three consecutive
days. Depletion was maintained by administration of an additional 100 μg of GK1.5 every
3-4 days until the end of the experiment. One group of tg and non-tg mice started GK1.5
treatment before Betaferon® treatment (day -4, n=13), while other groups started after 2,
4, or 6 injections of Betaferon® (days 2, 4 or 8, n=10). A control group without CD4+ T
cell depletion received a corresponding volume of saline (100 μl) on the same days as
group that was depleted before start of Betaferon® treatment (from day -4, n=6). Depletion
efficiency was assessed for each group on two or three days after initial GK 1.5 injections
and every 6-7 days until the end of the experiment. At each time point three mice per group
were sacrificed, spleens isolated and checked for CD4+ T cell depletion by an PE anti
CD4+ antibody (clone RM4.4 eBiosciences, Austria, which recognizes a different epitope
than GK1.5), using FACS Canto II flow cytometer (BD Biosciences, The Netherlands).
Depletion efficiency was on average ~80%.
48
CD4+ T cells, Germinal Centers and Memory.
Blood was collected via cheek puncture at different time points before the start of Betaferon®
treatment (day 1), during the three treatment weeks (days 8, 12 and 19) and one week after
treatment (day 26) in lithium heparin tubes (Greiner-bio-One, the Netherlands). On day 26
animals were sacrificed by decapitation after which blood was collected.
Plasma was isolated by spinning down the blood for 10 minutes at 3000g at 4°C and stored
at -20°C until further assessment for anti- rhIFNβ antibodies.
2.3.2. Experiment 2: Immunological memory response after increased exposure to
Betaferon®.
Tg and non-tg mice (n=10), were IP injected with a Betaferon® dose of 0.1 μg, 5 μg or 20
μg per injection for 3 weeks, 5 injections/week (days 1-5, 8-12 and 15-19). Tg and non-tg
control groups (n=10) received corresponding volumes of saline (days 1-5, 8-12 and 1519). After the initial three weeks treatment period and a 6 weeks wash-out phase all mice,
including controls, were re-challenged by two IP injections of 5 μg Betaferon® in 100 μl of
PBS on days 64 and 65. Blood was collected vie cheek puncture before treatment on day -2,
during initial treatment on days 1, 5, 8, 12, 15 and 19, during washout on days 29, 42 (n=5
per group per time point to prevent oversampling), before re-challenge on days 57 (n=5) and
61 (n=3), and after re-challenge on days 67 and 73 (n=7). Plasma was isolated by spinning
down the blood for 10 minutes at 3000g at 4°C and stored at -20°C until further assessment
of anti- rhIFNβ antibodies presence.
2.3.3. Experiment 3: Formation of germinal centers.
In order to study the presence of GCs after treatment with therapeutic interferon beta, mice
received rhIFNβ-1a drug substance. This rhIFNβ product does not contain HSA (in contrast
to Betaferon®) and allows assessment of GCs specific for rhIFNβ. Low immunogenicity
of this compound, however, required an increased number of mice per group to obtain
sufficient ADA positive mice 13.
Tg and non-tg mice (n=16) were treated with 5μg of rhIFNβ-1a drug substance in 100ul of
PBS on days 1-5, 8-12 and 15-18. As control tg and non-tg mice (n=8) were treated with
Pneumovax 23® (Sanofi Pasteur MSD, Brussels, Belgium) on days 1 and 11 of experiment
as described previously 23. Pneumovax 23® is a TI antigen and should not evoke germinal
center formation. To enhance the immune response and therefore facilitate formation of
germinal centers, another two groups of tg and non-tg mice received either rhIFNβ-1a drug
substance (n=16) adsorbed aluminium hydroxide gel or Pneumovax23® (n=8) adsorbed
aluminium hydroxide gel (Sigma-Aldrich, the Netherlands) in the same doses and on the
same days as described above. A (negative) control groups of tg or non-tg mice receiving
saline on days 1-5, 8-12 and 15-18 were also included (n=4). Half of animals from each
group were sacrificed by decapitation on day 9 and the other half of the mice on day 19 of
the experiment. Spleens were isolated, snap-frozen in liquid nitrogen and stored at -80°C for
immunohistochemical analysis of GCs formation (description below). To assess antibody
formation blood was collected and plasma was isolated by spinning down the blood for 10
49
Chapter 3.
minutes at 3000g at 4°C and stored at -20°C until further analysis.
2.4. Antibody assays.
Titers of anti-rhIFNβ antibodies (ADA) were measured by a direct sandwich ELISA as
described before 21. Briefly, 96 wells microtiter plates (Greiner-Bio-One, the Netherlands)
were coated with 0.1 μg of rhIFNβ-1a per well. After an overnight incubation, plates were
washed with washing buffer (PBST: PBS + 0.05% Tween 20) and subsequently blocked with
4% milk in PBST. For the screening assay, plasma was diluted 1:100 in blocking solution
and the plate incubated for 1 hour at RT with constant orbital shaking (400rpm). Anti-hIFNβ
antibodies were detected by HRP conjugated goat anti-mouse total IgG Ab (Invitrogen, the
Netherlands). The colour reaction was initiated by addition of 3,3’,5,5’-Tetramethylbenzidine
(TMB, Invitrogen, the Netherlands) and stopped by 0,18M of H2SO4. The optical density
values were measured with a BMG Spectrostar microplate reader (Isogen Life Sciences, the
Netherlands) at a wavelength of 450 nm (OD450).
Plasma samples were defined positive if the background corrected absorbance values
were ten times higher than that of the pre-treatment sera. Titers of positive samples were
determined by subsequent ELISAs, using a twofold serial dilution series with a starting
dilution of 1:10.
Titers of anti-Pneumovax 23® antibodies were measured by a modified method as described
by Zysk et al. 24. Briefly, pneumococcal polysaccharides were conjugated with poly-Llysine and then coated on the NuncTM MaxisorpTM microtiter plate (Fisher Scientific, the
Netherlands). Plates were washed twice with distilled water and incubated for 10min with
PBS containing 0,05% Tween 20. Serial 5-fold dilutions of plasma samples in PBS were
then incubated on the plates for 2hrs at 37°C. HRP conjugated goat anti mouse IgG Ab
(Invitrogen, the Netherlands) diluted 1:4000 in PBST was added on the plate and incubated
at 4oC overnight. The colour reaction was initiated by adding TMB and terminated by
0,18M H2SO4. OD450 was measured with BMG SpectroStart microplate reader (Isogen
Life Sciences, the Netherlands).
The titers of anti-rhIFNβ antibodies and anti-Pneumovax 23® antibodies were calculated
by plotting the absorbance values of the serial dilutions against log dilution. The plots were
fitted to a sigmoidal dose-response curve using GraphPad Prism version 4.00 (GraphPad
Software, San Diego CA, USA). The reciprocal of the dilution of the EC50 value was
considered the titer of the plasma. ADA levels of the tg mice in experiment 3, “Formation
of germinal centers”, were considered positive using the described cut off point. However,
they were too low to calculate titers as described above. Therefore OD450 values using a
dilution of 1:100 were used for data analyses and figures.
2.5. Immunohistochemistry staining of GCs.
5 μm sections of frozen spleens embedded in OCT medium (CellPath, UK) were cut on a
cryostat (Leica, The Netherlands). Sections were air dried and fixed in 2% paraformaldehyde
50
CD4+ T cells, Germinal Centers and Memory.
(4°C, 2min) followed by 10min of blocking with 3% bovine serum albumin (BSA) in Tris
buffered saline (TBS) supplemented with 2% Tween 20 (TBST). Endogenous peroxidases
were inactivated by 10 min incubation with 3% H2O2 in MetOH. GC specific cells were
stained with biotin conjugated peanut agglutinin (PNA, Sigma, the Netherlands) diluted
1:200 in blocking solution (4°C, O/N). Then slides sections were incubated with Alkaline
Phosphatase conjugated Extravidin (Sigma, the Netherlands) (1:100, 30min, RT), washed
and the visualization was initialized by adding nitro blue tetrazolium/5-bromo-4-chloro3-indolyl-phosphate (NBT/BCIP) substrate (Sigma, the Netherlands). Naïve B cells were
visualized by rat antiCD45R antibody (eBiosciences, Austria) (1:200, 1hr, 37°C) followed
by anti-rat-HRP conjugate (eBiosciences, Austria) (1:200, 1hr, 37°C). The B cells were
visualized by 3-Amino-9-ethylcarbazole (AEC) substrate according to the manufacturer
protocol (Sigma, the Netherlands). Sections were analysed by Nikon Eclipse 2000-U
microscope (Nikon Corporation, Japan). The density of GCs, expressed as the number
of PNA positive regions on the section per surface of the section was calculated in the
ImageJ software (National Institute of Health, USA). The number of GC was normalized by
dividing the number of GC in spleens from treatment groups by the basal number of GCs in
the control spleens from saline treated animals.
2.6. Statistical analysis.
Because obtained data were not normally distributed non-parametric tests were used to
determine statistical differences between groups. A Kruskal-Wallis test was used to
determine the overall effect of treatments in Experiment 1. Since this test showed significant
differences between treatments a one tail Mann-Whitney test was used to assess if treatment
with GK1.5 (anti CD4+ T cell antibody) significantly decreased ADA titers compared to
the saline treated mice. In Experiment 2 a Kruskal-Wallis test was used to determine the
overall difference in ADA titers between tg and non-tg animals. To assess if ADA titers
increased significantly after re-challenge in Experiment 2 the average titer from days 57 and
61 (before re-challenge) were compared with the average titer from days 67 and 73 (after
re-challenge) using a one tailed Mann-Whitney test. Potential difference between treatment
groups in Experiment 3 was assessed by a Kruskal-Wallis test. Since the test showed
significant differences in GCs between treatment groups a two tailed Mann-Whitney test
was executed to compare GCs between tg and non-tg animals as well as to compare the
effect of adjuvant within the tg or non-tg mouse groups. All calculations were performed
using IBM® SPSS® 20.0 (IBM, The Netherlands) software and a p value equal or lower than
0.05 was considered significant.
3. Results and discussion.
3.1. Experiment 1: Timing of CD4+ T-cell involvement in the ADA response against
Betaferon®.
CD4+ T cells were depleted on different time points during Betaferon® treatment. As shown
in Figure 1, CD4+ T-cell depletion in tg and non-tg mice either before or directly after (day 2)
the start of Betaferon® treatment significantly inhibited ADA response compared to control
51
Chapter 3.
A)
* p=0.003, ** p<0.001
* p=0.005, ** p<0.001
1000
** p=0.026
4/6
8/9
Titer
6/6
9/9
100
1/6
0/6
2/6
0/6
B)
ep
le
te
d
N
on
-d
ay
D
D
D
ay
ay
4
2
-4
10
Figure 1. Formation of total antirhIFNβ IgG in CD4+ T cell depleted
animals on day 8 (A) and 19 (B) of
Betaferon® treatment. The x axis
shows on which day of experiment
the depletion of CD4+ T cells was
initiated. White bars show average
titer of tg (*) ADA positive animals
with corresponding SD. Black bars
show average titer of non-tg (**) ADA
positive animals with corresponding
SD. Numbers above bars indicate the
number of ADA positive animals out
of total amount of animals in group.
The lines above the bars indicate
significant differences between time
points for * tg and ** non-tg mice.
* p=0.003, ** p<0.001
100000
* p=0.005, ** p<0.001
4/4
10000
3/3
4/4
Titer
4/4
3/3
4/4
4/4
1000
4/4 3/4
3/4
8
on
-d
ep
le
te
d
ay
N
D
D
ay
4
2
ay
D
D
ay
-4
100
mice. When CD4+ T cell depletion was initiated after the 4th injection of Betaferon® (day 4)
tg animals developed comparable ADA titers to non-depleted controls on day 8 (p=0.776).
In contrast, in corresponding non-tg animals ADA titers on day 8 were significantly lower
when compared to the non-depleted control group. On day 19, both tg and non-tg animals
had comparable ADA levels as control mice (p=0.114 for tg and p=0.4 for non-tg). When
CD4+ T cell depletion started on day 8, no significant effect on ADA formation was observed
for either the tg and non-tg animals (day 19, p=0.229 and p=0.229 respectively). Our results
confirm the importance of CD4+ T cells for ADA formation in tg mice. They show that CD4+
T cells are recruited in anti rhIFNβ antibody responses shortly after the first administration
of Betaferon® and are vital during subsequent ADA production. Moreover, presented data
suggests a similar role of CD4+ T cells in the production of ADA in tg animals tolerant for
the hIFNβ and non-tg mice which develop immune response likely via the TD pathway at
52
CD4+ T cells, Germinal Centers and Memory.
early stage of ADA production. Next studies looking into more detail to the subsets of CD4+
cells important in forming ADA should be performed, and studies determining if the CD4+
T cells are involved in the actual trigger of the immune response or assessing if they are
recruited later for actual antibody production are needed as well.
3.2. Experiment 2: Immunological memory response after increased exposure to
Betaferon®.
This experiment was designed to study if increasing exposure to Betaferon® would lead to a
clear formation of immunological memory. As shown in Figure 2, increasing the treatment
dose resulted in higher ADA titers in both tg and non-tg mice, although overall ADA titers
of tg animals were lower than those of non-tg mice (p<0.001). During the washout phase
ADA titers started to decline, however at the end of the washout period ADA could still
be detected. After re-challenging non-tg mice with 2 injections of Betaferon®, ADA titers
rapidly increased in all treatment groups, independent of the initial treatment dose. In
contrast, in the tg mice the ADA titers did not increase significantly after re-challenge in
groups which were first injected with 0,1 μg (p=0.073) or 20 μg (p=0.094) of Betaferon®.
Surprisingly tg animals receiving 5 μg of Betaferon® during primary treatment did develop
significantly higher ADA titers after re-challenge. However, this increase was lower than
A)
Titer
8000
6000
0.1 µg
5 µg
4000
20 µg
5 µg p<0.001
2000
0
0
10
20
30
40
50
60
70
Day
B)
16000
14000
0.1µg p<0.001
5µg p=0.003
20µg p=0.041
Titer
12000
10000
0.1 µg
5 µg
8000
20 µg
6000
4000
2000
0
0
10
20
30
40
50
60
70
Day
Figure 2. Anti rhIFNβ antibody formation in tg (A) and non-tg (B) mice. The graphs show average
titer of ADA positive animals with corresponding SD. The thick black line indicates days on which
animals received Betaferon® injections.
53
Chapter 3.
in non-tg mice (~2.5 fold increase in tg mice initially treated with 5 μg Betaferon® and
~13.1 fold increase in non-tg mice treated with the same initial dose). In the non-tg control
group, which received saline instead of Betaferon® during initial treatment, two injections
of Betaferon® resulted in very limited ADA production (titer on day 73 equal to 84.5±35.6).
Control tg mice did not develop measurable ADA titers on day 73.
Our results show that non-tg mice display a clear immunological memory response,
independent of the initial dose of Betaferon®, however for the tg mice this is not as clear.
Tg mice treated with the lowest and highest Betaferon® dose have an apparent lack of
immunological memory, however the group treated with 5 ug Betaferon® appears to have
a slight immunological memory response. The latter is very much unexpected since our
previous two experiments on this topic showed that tg mice treated with this dose did not
have increased ADA titers after re-challenge. It is unlikely that this discrepancy is due to
differences in experimental procedure, as both experiments were executed in a similar
manner, however minor differences in Betaseron® batches or slight difference in animal
breeding might have contributed to this.
Nonetheless, our results show that increase of the Betaferon® dose to 20 μg of protein per
injection, despite enhancing antibody production during initial treatment when compared
with 5 μg, does not lead to a clear immunological memory response in tg mice. It appears
that increasing exposure to antigen is not sufficient to give rise to an immunological memory
response in the tg mice which is similar to the response in non-tg mice. However, to fully
understand why a clear memory response as observed in non-tg mice is not taking place in
tg mice, more studies are needed. Experiments on the cellular level, focused on the detection
of rhIFNβ specific memory cells, would be of great value.
3.3. Experiment 3: Formation of germinal centers.
The aim of this experiment was to determine if GCs are formed during the response against
therapeutic rhIFNβ. rhIFNβ-1a drug substance gave a significant ADA response in non-tg
mice and much lower ADA response in the tg mice (Figure 3). Aluminium hydroxide gel
adjuvant resulted in an increase of ADA levels in non-tg animals but not in tg animals.
Administration of Pneumovax 23® induced formation of similar antibody titers in both tg
and non-tg mice (Figure 3). No effect of adjuvant on the antibody titer was seen in animals
receiving Pneumovax 23®.
Non-tg animals treated with rhIFNβ-1a drug substance had similar number of GCs as tg
mice on days 9 and 19 (Figure 4) (p=0.352 on day 9 and p=0.318 on day 19). The number of
GCs in tg and non-tg mice was not significantly different between rhIFNβ-1a drug substance
and Pneumovax 23® treatment. For the tg mice adsorption of rhIFNβ-1a drug substance to
the adjuvant did not increase the number of GCs either on day 9 (p=0.556) or on day 19
(p=0.383). In contrast, in the non-tg mice the adjuvant significantly increased the number
of GCs by a factor of 4 on day 9 (p=0.009). On day 19 this increase was just failing to
reach statistical significance (p=0.051). Aluminum hydroxide did not have an effect on the
54
CD4+ T cells, Germinal Centers and Memory.
A)
OD 450nm
3.5
2.5
3/8
2.0
1.5
5/8
1.0
0.5
0.0
8/8
8/8
8/8
3.0
Tg IFNβ-1a
non-Tg IFNβ-1a
Tg IFNβ-1a+Adjuvant
non-Tg IFNβ-1a+Adjuvant
2/7
0/8
0/8
9
Day
B)
19
*
10000
8/8
Titer
8/8
8/8
1000
Tg IFNβ-1a
non-Tg IFNβ-1a
Tg IFNβ-1a+Adjuvant
non-Tg IFNβ-1a+Adjuvant
2/8
100
9
Day
19
C)
1000
Titer
*
*
100
10
9
Day
Tg Pneumovax23
non-Tg Pneumovax23
Tg Pneumovax23+Adjuvant
non-Tg Pneumovax23+Adjuvant
Tg Saline
non-Tg Saline
19
Figure 3. A) Effect of aluminum hydroxide on the production of total IgG anti rhIFNβ-1a antibodies
in tg and non-tg animals showed as the change of OD450 values. The antibody levels in tg animals,
although in part of them exceeded the positivity threshold (see definition in Materials and Methods
section), were too low to calculate titers according to described method. The bars show average
OD450 of all animals in group with corresponding SD. The numbers above bars indicate the number
of animals classified as positive out of total amount of mice per group. B). Effect of aluminum
hydroxide on the titer of total IgG anti rhIFNβ-1a antibodies in tg and non-tg animals. The bars show
the average titer of ADA positive animals with corresponding SD. Adjuvant significantly increased
ADA titers of non-tg animals on day 19 (*p=0.005) (C) Production of anti Pneumovax23® total IgG.
The treatment with Pneumovax 23® whether with or without adjuvant significantly increased total
IgG titers when compared to control group (*p=0.001). Bars show the average titer from all animals
in group with corresponding SD.
number of GCs formed after Pneumowax 23® treatment in the tg and non-tg mice.
Our results show that although GCs are present in the tg mice, these animals do not produce
antibodies. This observation suggests different dynamics of the maturation process of B
cells within GCs of non-tg mice, release of a higher number of mature plasma cells in nontg animals or a possible suppressing mechanism within GCs in tg mice, which prevents
production of ADA To stimulate ADA production and therefore facilitate formation of
55
A)
B)
ov
m
eu
Pn
ax
nt
23
ax
dj
uv
a
A
A
+
+
23
nt
ax
ov
ov
m
eu
Pn
0
23
+
uv
dj
t
an
uv
j
d
ax
1a
βFN
I
rh
A
23
ov
N
IF
rh
t
an
2
Pn
eu
m
0
4
dj
uv
a
2
6
Pn
eu
m
4
8
A
6
10
+
8
12
β1a
10
14
rh
IF
Nβ
-1
a
GCs relative number
12
rh
IF
N
*
14
β1a
GCs relative number
Chapter 3.
Figure 4. Formation of GCs in spleens of tg and non-tg animals 9 (A) and 19 (B) days after the
beginning of rhIFNβ-1a or Pneumovax23® treatment. The white bars show relative number of GCs
per section of the tg (white) and non-tg (black) animals’ spleen with corresponding SD. Adsorption
of rhIFNB-1a on aluminum hydroxide increased number of GCs in non-tg animals (* p=0.009) on
day 10 of Betaferon® treatment.
GCs, rhIFNβ-1a drug substance was adsorbed on aluminum hydroxide gel. This adjuvant
is approved for vaccinations and it is known to activate a Th 2 type of immune response
25,26
. Indeed, it was able to enhance ADA formation and increase GCs number in the non-tg
mice. However, it failed to increase antibody production against rhIFNβ-1a drug substance
in tg animals. The number of GCs in tg animals also remained unchanged after addition of
adjuvant. This again suggests a down regulating mechanism inhibiting production of ADAs
to therapeutic proteins in tg animals or less efficient maturation of B cells during stimulation
with “self-like” antigen.
Formation of GC in mice treated with Pneumovax 23® seems confusing as it is generally
accepted that TI antigens do not trigger GCs. However, it was shown that immunization with
certain TI type II antigens might lead to formation of GCs 27,28. Nevertheless, the difference
in the structure of TI type II antigens (polysaccharides presenting repetitive epitopes) and
rhIFNβ-1a (i.e. soluble, monomeric protein) strongly suggests TD mechanism of GCs
formation in tg animals treated with rhIFNβ-1a.
Taken together, our data suggests that the mechanisms underlying immunogenicity of
therapeutic rhIFNβ in immune tolerant animals might be an atypical TD pathway, especially
at the early stages of antibody response. Our results indicate that the early events of ADAs
production in tg animals are strongly dependent on the presence of CD4+ T cells. Moreover,
formation of GCs seems to occur in the same manner as during classical TD response.
However, significantly lower ADA production compared to non-tg control animals together
with lack of apparent memory formation suggest deviations from the TD process of B cell
56
CD4+ T cells, Germinal Centers and Memory.
maturation. Isotype switching is one of the first processes during B cell maturation and it
precedes memory formation 29,30. Therefore it might be that later events of ADA production,
such as production of memory cells, are suppressed by yet unknown mechanism. T regulatory
(Tregs) cells were suggested to play very important role in immunogenicity of rh therapeutic
proteins 31, although data supporting this hypothesis are lacking. Also the question remains
which subtype of effector CD4+ T cells are activated during ADAs production. For example,
one can hypothesize that only peripheral CD4+ T cells may be triggered during response
to aggregated therapeutic proteins. Lack of signaling from follicular CD4+ T cells, which
are crucial for B cell differentiation within GCs 29,30, might result in inability of B cells to
memory formation. It is known that when rh proteins display native, unmodified structures,
they are very low immunogenic but any physicochemical modification (e.g. aggregation)
might strongly enhance the risk of ADA occurrence. It has been shown that MZ B cells, a
typical TI B cell subset, are necessary to develop efficient response to aggregated rhIFNβ
14
. Therefore another potentially possible scenario might be that protein aggregates activate
first the MZ B cells, which subsequently attract CD4+ T cells, but somehow the MZ B cells
are incapable to fully activate a TD response. More studies are necessary, since currently no
immunological data are available to support any of these hypotheses.
4. Conclusions.
The antibody response to Betaferon® in tg immune tolerant animals cannot be classified
as typical TD response. The early stages of ADA production display features typical for
the TD mechanism that is formation of GCs and dependency on CD4+ T cells. However,
no memory seems to be induced. Moreover, response to rhIFNβ cannot be enhanced by
adjuvant (aluminum hydroxide). The future studies in immune tolerant mice are necessary
to understand the mechanisms of immunogenicity of Betaferon®.
Acknowledgements.
We acknowledge Darren P. Baker, Ph.D. from Biogen Idec for kindly providing the IFNβ1a (Avonex drug substance), which was used for animal treatment in “Experiment 3:
Formation of germinal centers” and in the ELISA to detect ADA.
57
Chapter 3.
References.
1. Schellekens H. 2002. Immunogenicity of therapeutic proteins: clinical implications and
future prospects. Clin Ther 24:1720–1740
2. McHeyzer-Williams LJ, Driver DJ, McHeyzer-Williams MG. 2001. Germinal center
reaction. Curr Opin Hematol. 8:52–59
3. Parker DC. 1993. T cell-dependent B cell activation. Annu Rev Immunol 11:331–360
4. Zhu J, Paul WE. 2008. CD4 T cells: fates, functions, and faults. Blood 112:1557–1569
5. Dong C, Flavell RA. 2001. Th1 and Th2 cells. Curr Opin Hematol 8:47–51
6. Mond JJ, Lees A, Snapper CM. 1995. T Cell independent antigens type 2. Annu Rev
Immunol 13: 655-692
7. Mond JJ, Vos Q, Lees A, Snapper CM. 1995. T cell independent antigens. Curr Opin
Immunol 7:349–354
8. Pachner AR. 2001. Measurement of antibodies to interferon beta in patients with multiple
sclerosis, Arch Neurol 58:1299–1300
9. Bertolotto A, Deisenhammer F, Gallo P, Sölberg Sørensen P. 2004. Immunogenicity of
interferon beta: differences among products. J Neurology 251:II15–II24
10. Grossberg SE, Oger J, Grossberg LD, Gehchan A, Klein JP. 2011. Frequency and magnitude
of interferon β neutralizing antibodies in the evaluation of interferon β immunogenicity in
patients with multiple sclerosis, J Interferon Cytokine Res.31:337-344
11. Sominanda A, Rot U, Suoniemi M, Deisenhammer F, Hillert J, Fogdell-Hahn A. 2007.
Interferon beta preparations for the treatment of multiple sclerosis patients differ in
neutralizing antibody seroprevalence and immunogenicity. Mult Scler. 13:208–214
12. Deisenhammer F, Reindl M, Berger T. 2001. Immunoglobulin subclasses in patients with
neutralizing and nonneutralizing antibodies against IFN-beta1b. J Interferon Cytokine Res
21:167–171
13. Van Beers MMC, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon Beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
14. Sauerborn M, van Beers MMC, Jiskoot W, Kijanka G, Boon L, Schellekens H, Brinks
V. 2013. Antibody response against Betaferon® in immune tolerant mice: involvement of
marginal zone B-cells and CD4+ T-cells and apparent lack of immunological memory. J
Clinical Immunol 32:55-63
15. Perini P, Facchinetti A, Bulian P, Massaro AR, De Pascalis D, Bertolotto A, Biasi G,
Gallo P. 2001. Interferon-beta (INF-beta) antibodies in interferon-beta1a- and interferonbeta1b-treated multiple sclerosis patients. Prevalence, kinetics, cross-reactivity, and factors
enhancing interferon-beta immunogenicity in vivo. Eur Cytokine Netw 12:56-61
16. Brinks V, Jiskoot W, Schellekens H. 2011. Immunogenicity of therapeutic proteins: the use
of animal models. Pharm Res 28:2379–2385
17. Ruotsalainen J, Martikainen M, Niittykoski M, Huhtala T, Aaltonen T Jari Heikkilä J, Bell
J, Vähä-Koskela M, Hinkkanen A. 2012. Interferon-β Sensitivity of Tumor Cells Correlates
With Poor Response to VA7 Virotherapy in Mouse Glioma Models. Mol Ther 20:1529-1539
18. Nakazawa K, Sasamoto H, Shiraki Y, Harada S, Yanagi K, Yamane K. 1991. Extracellular
production of mouse interferon beta by the Bacillus subtilis alpha-amylase secretion vectors:
antiviral activity and deduced NH2-terminal amino acid sequences of the secreted proteins.
Intervirology 32:216–227
58
CD4+ T cells, Germinal Centers and Memory.
19. Qin XQ, Beckham C, Brown JL, Lukashev M, Barsoum J. 2001. Human and mouse IFNbeta gene therapy exhibits different anti-tumor mechanisms in mouse models. Mol Ther
4:356-364
20. Willmon CL, Saloura V, Fridlender ZG, Wongthida P, Diaz RM, Thompson J, Kottke T,
Federspiel M, Barber G, Albelda SM, Vile RG. 2009. Expression of IFN-beta enhances
both efficacy and safety of oncolytic vesicular stomatitis virus for therapy of mesothelioma.
Cancer Res 69:7713–7720
21. Hermeling S, Jiskoot W, Crommelin D, Bornaes C, Schellekens H. 2005. Development of
a transgenic mouse model immune tolerant for human interferon Beta. Pharm Res 22:847–
851
22. Van Beers MMC, Sauerborn M, Gilli F, Hermeling S, Brinks V, Schellekens H, Jiskoot W.
2010. Hybrid transgenic immune tolerant mouse model for assessing the breaking of B cell
tolerance by human interferon beta, J Immunol Methods 352:32–37
23. Scheikl T, Reis B, Pfeffer K, Holzmann B, Beer S. 2009. Reduced notch activity is associated
with an impaired marginal zone B cell development and function in Sly1 mutant mice, Mol
Immunol 46:969–977
24. Zysk G, Bethe G, Nau R, Koch D, Von VCDHG, Heinz HP, Reinert RR. 2003. Immune
response to capsular and surface proteins polysaccharide in patients of streptococcus
pneumoniae with invasive pneumococcal disease. J Infect Dis 187:330–333
25. Brewer JM. 2006. (How) do aluminum adjuvants work? Immunol Lett 102:10–15
26. Korsholm KS, Petersen R V, Agger EM, Andersen P. 2010. T T-helper 1 and T-helper 2
adjuvants induce distinct differences in the magnitude, quality and kinetics of the early
inflammatory response at the site of injection, Immunology 129:75–86
27. de Vinuesa CG, Cook MC, Ball J, Drew M, Sunners Y, Cascalho M, Wabl M, Klaus GG,
MacLennan IC. 2000. Germinal Centers without T Cells. J Exp Med 191:485-494
28. Sverremark E, Fernandez C. 1998. Germinal center formation following immunization with
the polysaccharide dextran B512 is substantially increased by cholera toxin. Int Immunol
10:851-859
29. Gatto D, Brink R. 2010. The germinal center reaction, J Allergy Clin Immunol 126:898–907
30. Hollowood K, Goodlad JR, 1998. Germinal centre cell kinetics, J Pathol 185:229–233
31. De Groot AS, Scott DW. 2007. Immunogenicity of protein therapeutics, Trends Iimmunol
28:482–490
59
60
CHAPTER 4.
The impact of adjuvants and immune
suppressants on the production of anti rh
interferon beta antibodies in immune tolerant
mice.
Grzegorz Kijanka
Huub Schellekens
Vera Brinks
Manuscript in preparation.
61
Chapter 4.
Abstract.
Anti drug antibodies (ADA) reduce the efficiency of therapeutic proteins Impurities and
leachates from product’s containers have been suggested to act as adjuvants and increasing
the level of ADA formation. We have tested several adjuvants, known to enhance the
classical T cell dependent (TD) response, for their effect on the ADA formation against
recombinant human interferon beta (rhIFNβ) in immune tolerant mice. Most were incapable
of enhancing antibody production. Only conjugation of rhIFNβ to cholera toxin subunit B
resulted in enhanced production of ADA. However, no clear formation of immunological
memory was observed upon re-challange with rhIFNβ. In addition we have tested the
potency of several immune suppressants in blocking of ADA formation. We found that
among all tested compounds only rapamycin effectively decreased ADA development.
62
Impact of immunemodulators on immunogenicity of rhINFb.
1. Introduction.
Protein drugs have revolutionized treatment of many severe diseases like diabetes,
hemophilia, multiple sclerosis (MS) and anemia. Unfortunately, the therapy outcome is
often affected by the production of anti drug antibodies (ADA) 1–3. They can neutralize the
biological activity of the drug, resulting in a decreased or even abolished efficacy of therapy.
Moreover, in rare cases ADA can also cross-react with endogenous proteins leading to
severe, even life threating side effects 4,5. Many factors have been suggested to increase the
risk of immunogenicity. Treatment regimen, sequence mutations, (lack of) posttranslational
modifications, aggregation, additives used in the formulation and impurities might all
contribute to ADA production 6–8. Although a number of studies have confirmed that the
presence of aggregates in the formulated product is a major risk factor, it has also been
suggested that host cell products or leachates from rubber stoppers of syringes might act as
adjuvants and increase the risk of ADA. Both bacterial compounds and leachates have been
shown to enhance antibody formation against a T cell dependent (TD) antigen (ovalbumin) in
mice 9,10. However, an increasing number of reports show that the immunological processes
underlying immunogenicity significantly differs from a classical adaptive immune response
11,12
. Moreover, recombinant human insulin was shown to be not immunogenic in immune
tolerant mice when administered with Freund’s adjuvant (FA) 13. Thus the question has been
raised if, or how these, TD adjuvants might enhance ADA response against therapeutic
proteins.
Because known risk factors contributing to the development of ADA might be very
difficult to eliminate from the end product, e.g. trace amount of aggregates, it has been
proposed that co treatment with an immune suppressor could be a possible way of lowering
immunogenicity. In fact, the immune suppressors methotrexate (MTX) and azathioprine
have been shown to decrease the ADA formation for different therapeutic proteins 14.
Moreover, a cyclosporine have been successfully used in therapy of antibody mediated
pure red cell aplasia (PRCA) induced by rh erythropoietin treatment 14–16. These examples
indicate that immune suppression might be used to lower ADA responses.
We have studied the effect of adjuvants and immune suppressants on the immunogenicity of
recombinant human interferon beta-1a (rhIFNβ-1a) in immune tolerant mice.
2. Materials and methods.
2.1. Animals.
Heterozygous transgenic (tg) C57Bl/6 mice carrying the human interferon beta (hIFNβ)
gene originally developed by Hermeling et al. 17 were bred at the Central Laboratory Animal
Institute (Utrecht University, the Netherlands). Tg C57Bl/6 males were crossbred with nontransgenic (non-tg) FVB/N mice (Janvier, BioServices, The Netherlands) 18. The hIFNβ
genotype of the offspring (F1) was determined by PCR using chromosomal DNA isolated
from ear tissue. Both PCR positive (tg) and PCR negative (non-tg) littermates of 6-10
weeks of age were used for the experiments. Mice were housed in standard perspex cages
63
Chapter 4.
and given access to food (Hope Farms, Worden, The Netherlands) and water (acidified) ad
libitum. All experiments were performed according to the Institutional Ethical Committee
Regulations of Utrecht University, the Netherlands.
2.2. Recombinant human interferon Beta.
RhIFNβ-1a (Avonex® drug substance) was supplied by Biogen Idec (Cambridge, MA,
US) and was formulated to a concentration of 270 μg/ml in 100 mM sodium phosphate
pH 7.2, 200 mM sodium chloride. RhIFNβ-1b (Betaferon®) was obtained from Schering
(Berlin, Germany). Lyophilised powder containing 300 μg of rhIFNβ, 15 mg of human
serum albumin (HSA) and 15 mg of mannitol was reconstituted in 1 ml of 10 mM sodium
phosphate pH 7.4, 137 mM sodium chloride (PBS). Before injection, rhIFNβ-1a and
rhIFNβ-1b were further diluted in PBS to a concentration of 50 μg/ml.
2.3. Impact of adjuvants on the immunogenicity of rhIFNB in immune tolerant mice.
2.3.1. Preparation of rhIFNB-1a/adjuvant solutions.
Complete and Incomplete FA (CFA/IFA), Aluminum hydroxide gel (Alum), Saponin from
Quillaja saponaria Molina, CuCl2 and Cholera Toxin Subunit B (CTB) were purchased
from Sigma Aldrich (The Netherlands). Montanide ISA 50 V2 was obtained from SEPPIC
(France).
CFA/IFA or Montanide ISA 50 V2 were mixed with rhIFNβ-1a solution in a 1:1 ratio and
vortexed for 30 sec to obtain a homogenous emulsion. Alum was added to the rhIFNβ-1a
solution to a final concentration of 9% (vol/vol). The mixture was incubated on the roller
mixer for 30 min at RT in order to allow rhIFNβ-1a adsorbing onto the Alum. Saponin
was added to rhIFNβ-1a to a final concentration of 0.2%. CuCl2 was dissolved in PBS to a
final concentration of 0.08M (stock solution). To avoid metal induced oxidation of rhIFNβ1a and therefore formation of protein aggregates, the stock solution of CuCl2 was mixed
with rhIFNβ-1a in a 1:1 ratio directly before injection. The final concentration of CuCl2 of
0.04M was chosen to be identical as concentration used in the protocol of metal induced
oxidation of rhIFNα described by Hermeling et al 19. CTB was conjugated to rhIFNβ-1a
using the 3-(2-Pyridyldithio)propionic acid N-hydroxysuccinimide ester (SPDP, Sigma,
The Netherlands) as a linker (see supporting materials for details). All solutions were stored
at 4°C before applying them to the mice.
2.3.2. Animal experiments.
In all animal experiments performed to study impact of adjuvants on the immunogenicity of
rhIFNβ-1a mice were treated with rhIFNβ-1a according to a modified regimen described by
Hermeling et al 17. Briefly, tg and non-tg mice (unless indicated elsewhere n=8 per group)
received 15 intraperitoneal injections of rhIFNβ-1a, 5 μg/100 μl of PBS per injection, over
a time period of three weeks (days 1-5, 8-12, 15-19).
In Experiment I the following adjuvants were tested: CFA/IFA, Saponin, Aluminum
hydroxide and Montanide. The CFA was administered together with rhIFNβ-1a on day 1
of the experiment and IFA on days 3 and 5. On the other days the mice from this group
64
Impact of immunemodulators on immunogenicity of rhINFb.
received rhIFNβ-1a treatment only. Mice co-injected with Montanide received the rhIFNβ1a/Montanide emulsion on days 1-5. On days 8-12 and 15-19 these mice were injected
with rhIFNβ-1a only. Saponin and Alum were co-injected with rhIFNβ-1a during the whole
period of the experiment (days 1-5, 8-12 and 15-19). A control group was treated with
rhIFNβ-1a without the addition of an adjuvant.
The mice were euthanized by decapitation under anesthesia 3 days after last the injection (day
22). The development of ADA was monitored in blood samples collected submandibularly
on days -2, 10 (before the rhIFNβ-1a injection of that day) and 22 (during euthanasia).
Plasma was isolated by spinning down the blood for 10 minutes at 3000g at 4°C and was
stored at -20°C until further assessment for anti- rhIFNβ antibodies.
Experiment II was performed to examine the impact of CTB and CuCl2 on immunogenicity
of rhIFNβ-1a. Experimental groups consisted of Tg mice (n=8) and non-Tg controls (n=7
in groups treated with CTB and CuCl2 and n=6 in group injected with rhIFNβ-1a alone).
Mice received rhIFNβ-1a with or without adjuvant (control) on days 1-5, 8-12 and 15-19
and were euthanized on day 22 by decapitation under anesthesia. The development of ADA
was monitored in blood samples collected submandibularly on days -2, 10 (before rhIFNβ1a injection of that day) and 22 (during euthanasia). Plasma was isolated by spinning down
the blood for 10 minutes at 3000g at 4°C and stored at -20°C until further assessment for
anti- rhIFNβ antibodies presence.
Because in Experiment II CTB was found to be a potent enhancer of ADA production in
the Tg mice, Experiment III was performed to determine if this enhanced immunogenicity
could lead a clear formation of immunological memory in the Tg mice. Tg (n=8) and non-Tg
(n=7) mice were treated with rhIFNβ-1a with or without (controls) CTB for three weeks as
described in the previous paragraph. After the initial treatment a 6 week washout phase was
applied after which the mice were re-challenged with two injections of 5 μg of rhIFNβ-1a
on days 64 and 65. On day 73 the animals were euthanized by decapitation under anesthesia.
The anti rh-IFNB antibodies were determined in blood samples collected submandibularly
on days -2, 10 (before rhIFNB-1a injection of that day), 22, 50, 61, 68 and 73 (during
euthanasia). Plasma was isolated by spinning down the blood for 10 minutes at 3000g at
4°C and stored at -20°C until further assessment for anti- rhIFNβ antibodies.
2.4. The impact of immune suppressants on the immunogenicity of rhIFNβ-1b in
immune tolerant animals.
2.4.1. Preparation of immune suppressants and rhIFNβ-1b solutions.
Prednisone, Methotrexate, Cyclosporine and Azathioprine were purchased from Sigma
(the Netherlands). Rapamycin was obtained from Tebu-Bio (the Netherlands). All immune
suppressants were dissolved in DMSO to a concentration of 12.5 mg/ml. The final solutions
were obtained by diluting DMSO stocks of immune suppressants 50x in PBS containing
rhIFNβ-1b shortly before injection. All immune suppressants were used in a dose of 25 μg
per injection, which corresponds to ~ 1 mg/kg of body weight.
65
Chapter 4.
2.4.2. Animal experiment.
Tg and non-Tg mice (n=8 per group) were treated with 100 μl of rhIFNβ-1b solution (10 μg/
ml) with or without immune suppressants for three weeks (days 1-5, 8-12, 15-19). 3 days
after the last injection (day 22) mice were euthanized by decapitation under anesthesia.
Blood was collected before the start of rhIFNβ-1b treatment (day -2), every week during
treatment (days 8 and 15) and during euthanasia (day 22).
Plasma was isolated by spinning down the blood for 10 minutes at 3000g at 4°C and stored
at -20°C until further assessment for anti- rhIFNβ antibodies.
2.5. Antibody assays.
Titers of anti-rhIFNβ antibodies (ADA) were measured by a direct sandwich ELISA as
described before 17. Before measuring the ADA titer plasma samples were screened to
determine the presence of ADA. Briefly, 96 wells microtiter plates (Greiner-Bio-One, the
Netherlands) were coated with 0,1μg of rhIFNβ-1a per well. After an overnight incubation,
plates were washed with washing buffer (PBST: PBS + 0,05% Tween 20) and subsequently
blocked with 4% milk in PBST. For the screening assay, plasma was diluted 1:100 in
blocking solution and the plate incubated for 1 hour at RT with constant orbital shaking (400
rpm). Anti-hIFNβ antibodies were detected by HRP conjugated goat anti-mouse total IgG
Ab (Invitrogen, Bleiswijk, the Netherlands). The colour reaction was initiated by addition of
3,3’,5,5’-Tetramethylbenzidine (TMB, Invitrogen, the Netherlands) and stopped by 0,18M
of H2SO4. The optical density values were measured with a BMG Spectrostar microplate
reader (Isogen Life Sciences, the Netherlands) at a wavelength of 450 nm (OD450).
Plasma samples were defined positive if the background corrected absorbance values
were ten times higher than that of the pre-treatment sera. Titers of positive samples were
determined by subsequent ELISAs, using a twofold serial dilution series. The starting
dilution was adjusted for every positive plasma sample on the basis of the screening ELISA
and it was in the range of 1:20 - 1:200.
The titers of anti-rhIFNβ antibodies were calculated by plotting the absorbance values of the
serial dilutions against the log dilution. The plots were fitted to a sigmoidal dose-response
curve using GraphPad Prism version 4.00 (GraphPad Software, USA). The reciprocal of the
dilution of the EC50 value was considered the titer of the plasma. In the majority of mice
for which ADA titers could not be determined by described approach an increase of OD450
was observed in screening ELISA for day 22 when compared with basal OD450 (on day -2).
That suggests very low production of ADA and therefore for statistical calculations and for
graphs they were given arbitrary titer of 1.
2.6. Statistical Analysis.
All obtained data were checked for normal distribution by a Schapiro-Wilk test. Because data
were not normally distributed non-parametric tests were used to determine the differences
between groups. A one-side Mann-Whitney test was used to determine if the adjuvants
66
Impact of immunemodulators on immunogenicity of rhINFb.
increased the antibody formation in treatment groups when compared with controls. The
same test was used to evaluate if the immune suppressants decreased the ADA titers. A P
value equal to or lower than 0.05 was considered significant. All calculations were done
using the IBM® SPSS® Statistics 20 software (IBM, USA).
3. Results.
3.1. Impact of adjuvants on the immunogenicity of rhIFNB in immune tolerant mice.
The production of ADA on day 22 of Experiment I in tg and non-tg mice is shown in Figure
1A. Treatment of tg mice with rhIFNβ-1a did not induce production of ADA on day 22.
Moreover, none of the tested adjuvants was able to trigger ADA production in those mice.
In contrast, all non-tg mice treated with rhIFNβ-1a developed ADA. Montanide ISA 50
V2, saponin and Alum significantly increased the ADA titers on day 22 in non-tg mice.
Among all tested adjuvants saponin seemed to be the most potent as the average ADA titer
in non-tg mice was 60 fold higher than in control non-tg mice receiving only rhIFNβ-1a
(p<0.0001). Alum induced a 8.7 fold increase and Montanide ISA 50 V2 elevated the titers
by 2 fold (p=0.002 and p=0.021, respectively). No impact of CFA/IFA on antibody titers
was observed in the non-tg mice.
In Experiment II mice were treated with rhIFNβ-1a conjugated to CTB or mixed with
CuCl2. Although rhIFNβ-1a/CTB conjugates did not result in a significant increase in ADA
A)
Tg
1000
100
10
*
1000
100
10
C
M
M
Tg
Sa
on
po
ta
ni
ni
n
de
Al
IS
um
A
50
in
um
V2
hy
dr
ox
id
e
C
on
tro
l
FA
/IF
A
Sa
on
po
ta
ni
ni
n
de
Al
I
S
um
A
50
in
um
V2
hy
dr
ox
id
e
C
on
tro
l
FA
/IF
A
non-Tg
**
Anti rhIFNβ titer
100000
10000
1000
100
10
1
10000
1000
100
10
2l
uC
C
TB
on
tro
l
C
C
on
tro
l
2l
uC
C
C
TB
1
C
C
**
1
100000
Anti rhIFNβ titer
***
10000
1
B)
non-Tg
100000
Anti rhIFNβ titer
Anti rhIFNβ titer
10000
Figure 1. Production of anti rhIFNβ-1a ADA in tg and non-tg mice in A) Experiment I and B)
Experiment II. *p≤0.05, **p≤0.01, ***p≤0.001.
67
A)
Tg
10000
1000
100
10
1
10
B)
Day
1000
100
10
1
22
10
Tg
22
10000
1000
100
10
1
Day
Non-Tg
10000
Anti rhIFNβ titer
Anti rhIFNβ titer
Non-Tg
10000
Anti rhIFNβ titer
Anti rhIFNβ titer
Chapter 4.
10
Day
22
1000
100
10
1
10
Day
22
Figure 2. Production of anti rhIFNβ-1a ADA in tg and non-tg mice on days 15 and 22 (primary
treatment) of Experiment III. A) Mice treated with rhIFNβ-1a, B) Mice treated with CTB/rhIFNβ1a conjugates
titers in the tg mice, 3 out of 8 mice did develop measurable ADA titers (Figure 1B). CuCl2
induced ADA in 1 tg mouse. In non-tg mice administration of rhIFNB-1a/CTB conjugates
resulted in a significant increase of ADA titers, when compared to control mice receiving
only rhIFNβ-1a (p=0.034). The average ADA titer in non-tg mice receiving rhIFNβ-1a with
or without CuCl2 did not differ significantly. However, in group treated with rhIFNβ-1a
mixed with CuCl2 all non-tg mice were classified as responders, whereas in group treated
only with rhIFNB-1a ADA formation was found in only 4 out of 6 mice.
Because CTB was found in Experiment II to be an effective adjuvant for rhIFNβ-1a in tg
mice a follow up study (Experiment III) was performed to determine if CTB may lead to
a complete breaking of tolerance and, in consequence, to the formation of immunological
memory against rhIFNβ. To test this hypothesis, animals were administered with rhIFNβ-1a
conjugated with CTB according to the same regime as for Experiment II. As shown in the
Figure 2, all tg and non-tg mice treated with rhIFNβ-1a and CTB developed ADA on day
22 of the experiment. Moreover ADA titers in mice treated with CTB/rhIFNβ conjugates
were significantly higher than in control mice (p<0.001 and p=0.017 for tg and non-tg
mice respectively). Interestingly, injections of CTB resulted in a faster ADA onset in both
the tg and non-tg mice compared to the non-tg mice receiving only rhIFNβ-1a. The initial
treatment was followed by a 6 week washout phase, after which mice were re-challenged
with rhIFNβ-1a without CTB. At the end of the washout phase (day 61) the ADA were
68
Impact of immunemodulators on immunogenicity of rhINFb.
non-Tg
Tg
A)
Anti rhIFNβ titer
10000
1000
100
10
1
Anti rhIFNβ titer
B)
Before
10000
1000
100
10
Tg
Before
10000
1000
100
10
Before
After
non-Tg
***
100000
100000
1
**
100000
1
After
Anti rhIFNβ titer
Anti rhIFNβ titer
100000
After
10000
1000
100
10
1
Before
After
Figure 3. Titers of anti rhIFNβ-1a ADA in tg and non-tg mice before (day 61) and after (days 68 and
73) re-challenge with rhIFNβ-1a in Experiment III. A) Mice primarily treated with rhIFNβ-1a, B)
Mice primarily treated with CTB/rhIFNβ-1a conjugates. **p≤0.01, ***p≤0.001.
found only in 3 out of 7 tg animals. In contrast, ADA were still detected in all non-tg mice
irrespective of the primary treatment they received. On days 63 and 64 mice received two
injections with rhIFNβ-1a. Re-challenge resulted in a fast and robust increase of ADA titers
in both groups of non-tg mice, regardless of the initial treatment (p<0.001 and p=0.008 for
mice treated initially with or without CTB, respectively). In tg mice, however, ADA titers
did not increase after re-challenge.
3.2. The impact of immune suppressants on the immunogenicity of rhIFNβ-1b in
immune tolerant animals.
Figure 4 shows the results of the combined treatment with rhIFNβ-1b and immune
suppressants on the production of ADA on days 15 and 22 of the experiment. On day 15 only
rapamycin decreased ADA titers significantly in both tg and non-tg mice (p=0.008 for both
tg and non-tg mice). On day 22 ADA titers were significantly lower only in tg mice treated
with rapamycin (p=0.0035). In contrast, on day 22 in non-tg mice a reduction in ADA titers
was observed in groups receiving rapamycin (p<0.001) and cyclosporine (p=0.003). MTX
seemed to decrease the ADA titers in the tg on both days, however its impact failed to reach
significance (p=0.065 on both days).
Discussion.
Nearly all therapeutic proteins trigger ADA formation in at least a part of patients. Many
69
Chapter 4.
factors have been suggested to contribute to ADA development. Some of them have been
suggested to act as adjuvants, such as host cells impurities, leachates from storage containers
and excipients 8,10. For instance, the adjuvant properties of leachates from uncoated rubber
stoppers have been suggested as the cause of the increased frequency of antibody induced
pure red cell aplasia in patients treated with erythropoietin supplied in ready-to-use prefilled syringes (Eprex®, Johnson&Johnson, USA) 20. Leachates extracted from Eprex®
syringe’s uncoated rubber stoppers were found to enhance the immune response against
ovalbumin in mice 10. Also trace amounts of bacterial compounds were found to significantly
increase the immune response against ovalbumin and rh erythropoietin in Balb/c mice 9.
However, in these reports an impact of the leachates and bacterial compounds had been
tested using foreign, T cell dependent antigens which are known to be susceptible to the
effect of adjuvants. Direct prove showing the capability of these compounds to enhance
the production of ADA against a “self” protein in immune tolerant animals or humans is
however lacking.
The most commonly used adjuvants are mineral salts, water/oil emulsions and bacterial
compounds. In this study we evaluated several adjuvants some of which are commonly
used in animal and human vaccines. In addition, we tested the adjuvant properties of copper
salt, which was used as a catalyst used for the oxidation and aggregation of rhIFNα. These
studies showed that aggregates induced by cupper oxidation are indeed immunogenic in
the immune tolerant mice, suggesting that cupper might also have adjuvant properties 21.
The mode of action of the tested adjuvants used remains unclear but may be explained by
the formation of a depot at the site of injection and enhanced antigen uptake by dendritic
cells 22–25, the stimulation of pattern recognition receptors 26, induction of cell membrane
permeabilization attracting T 27 and increased uptake by antigen presenting cells by binding
to the GM1 ganglioside 28.
Our results clearly show that most adjuvants used in vaccines fail to induce ADA formation
towards the “self” antigen rhIFNβ in the immune tolerant mice. Although some of them,
i.e. saponin, were extremely potent in enhancing the immune response in non-tg mice (TD
antigen). Only CTB, when conjugated with rhIFNβ-1a, significantly increased both the
number of tg mice with detectable ADA and ADA titers. This might be caused by several
potential mechanisms: i) CTB may enhance the uptake of rhIFNβ-1a by antigen presenting
cells (APCs), ii) CTB might attract T helper cells to the immune response and as a foreign
protein conjugated to rhIFNβ-1a may by-pass T cell tolerance, iii) CTB-rhIFNβ conjugates
might form aggregate-like oligomers which enhance ADA production. Especially the
last explanation is attractive because it has been shown that protein therapeutics, when
aggregated, are considerably more immunogenic than monomeric proteins 21,29,30. RhIFNβ1a, when conjugated to CTB, formed oligomers of molecular weight exceeding 170 kDa
(supplementary data). However, the exact mechanisms causing this adjuvant effect of CTB
has to be elucidated.
70
Impact of immunemodulators on immunogenicity of rhINFb.
The re-challenge of tg mice with rhIFNβ-1a revealed that the production of ADA during
primary treatment with CTB/rhIFNβ conjugates did not lead to immunological memory.
It strongly suggests that tolerance of tg mice towards hIFNβ was not broken by adjuvant
usage. These results are in accordance with our previous observations showing that
aggregated rhIFNβ-1b, although very potent in triggering ADA production, failed to
induce a clear immunological memory response after re-challenge in tg mice 11. Our data
suggest possible existence of suppressing mechanism, which prevents memory formation.
For example it could be hypothesized that regulatory T cells (Tregs), which are known to
prevent autoimmunity by down regulation of effector T cells 31, might inhibit formation of
memory. However, further studies are needed to determine the mechanism responsible for
apparent lack of memory.
Immune suppressants had limited influence on the ADA production in both tg and non-tg
mice. Only rapamycin decreased ADA titers on both days 15 and day 22 in both tg and nontg mice. Rapamycin blocks the proliferation of T and B cells by inhibition of their response
to interleukin 2 (IL-2) 32. IL-2 is mainly produced by Th1 lymphocytes and it promotes
B cell proliferation 32–34. However, it is also required for differentiation of Th2 cells from
precursor Th0 cell 32,35. Interestingly, rapamycin seemed to have a higher impact on the
immune response in non-tg mice than in tg. On day 22, 2 out of 8 non-tg animals were
classified as non-responders in contrast to tg mice, which all developed detectable levels of
ADA.
The second immune suppressant effective in non-tg mice on day 22 was cyclosporine.
Similarly to rapamycin, cyclosporine also alters IL-2 production and signaling. Similarly to
rapamycin, cyclosporine also seemed to be less potent in tg mice as no significant difference
between cyclosporine treated and control groups was observed (p=0.1). Taken together,
rapamycin and cyclosporine effects seem to suggest that production of ADA in tg mice
is less dependent on IL-2. However future, more detailed studies are necessary to explain
wherever Th1 cell activation is more pronounced in non-tg mice compared to tg, whether
Th2 cells are more important in non-tg mice or other activity of IL-2 (e.g. maintenance of
Treg) is necessary for TD response.
On days 15 and 22 MTX seemed to reduce the average ADA titer in tg mice, but the p value
just failed to reach significance (p=0.065 on both days). The optimization of the MTX dosage
regimen might increase its effect. Interestingly MTX seemed not to have much effect on
lowering ADA titers in non-tg mice (p=0.4). MTX blocks the formation of nucleic acids by
inhibiting the metabolism of folic acids 36. As such, it blocks non-specifically cell division.
Thus future studies are needed to confirm stronger effect of MTX on ADA production in tg
than non-tg mice. Prednisone and azathioprine did not show any effect on ADA formation
in tg and non-tg mice. However, one has to keep in mind that the dosage regimen used in
current study might not be optimal for all immune suppressants. Nevertheless, our study
showed that immune suppressants might be effective in decreasing the ADA formation,
71
A)
Tg
non-Tg
*
1000
100
10
10000
Tg
100
10
Pr
ed
ni
so
et
ho ne
tr
ex
C
yc
at
e
lo
sp
or
A
za
in
e
th
io
pr
in
R
e
ap
am
yc
in
C
on
tr
ol
M
non-Tg
**
100000
**
1000
100
10
Pr
ed
ni
so
M
et
ne
ho
tr
ex
C
yc
at
e
lo
sp
or
A
za
in
e
th
io
pr
in
R
e
ap
am
yc
in
C
on
tr
ol
1
1000
***
10000
1000
100
10
1
Pr
ed
ni
so
et
ne
ho
tr
ex
C
yc
at
e
lo
sp
or
A
za
in
e
th
io
pr
in
R
e
ap
am
yc
in
C
on
tr
ol
Anti rhIFNβ titer
100000
Anti rhIFNβ titer
B)
**
1
Pr
ed
ni
so
M
et
ne
ho
tr
ex
C
yc
at
e
lo
sp
o
A
rin
za
e
th
io
pr
in
R
e
ap
am
yc
in
C
on
tr
ol
1
10000
M
Anti rhIFNβ titer
10000
Anti rhIFNβ titer
Chapter 4.
Figure 4. Titers of anti rhIFNβ-1b ADA on day 15 (A) and 22 (B) in tg and non-tg mice treated with
or without immune suppressants. *p≤0.05, **p≤0.01, ***p≤0.001.
however future studies are needed to optimize treatment regimen.
Conclusions.
We have shown that most of adjuvants known to increase antibody response to classical TD
antigens are not effective in enhancing the ADA production towards rh protein in immune
tolerant mice. Moreover, we showed that ADA production triggered by CTB, the only
effective adjuvant in tg mice, seemed not to induce complete break of immune tolerance and
formation of immunological memory. We have also shown that concomitant treatment with
immune suppressants might be effective in blocking ADA formation. However, influence
of some immune suppressants (cyclosporine) might be lower than in classical TD response.
Acknowledgments.
We acknowledge Darren P. Baker, Ph.D. from Biogen Idec for kindly providing the IFNβ-1a
(Avonex drug substance) in adjuvant study and in the ELISA to detect ADA.
72
References.
Impact of immunemodulators on immunogenicity of rhINFb.
1.
West RL, Zelinkova Z, Wolbink GJ, Kuipers EJ, Stokkers PC, van der Woude CJ. 2008.
Immunogenicity negatively influences the outcome of adalimumab treatment in Crohn’s
disease. Aliment Pharmacol Ther 28:1122–1126
2.
Schellekens H. 2005. Factors influencing the immunogenicity of therapeutic proteins.
Nephrol Dial Transplant 20:vi3–9
3.
Barbosa MD, Celis E. 2007. Immunogenicity of protein therapeutics and the interplay
between tolerance and antibody responses. Drug Discov Today. 12:674–681.
4.
Schellekens H. 2002. Immunogenicity of therapeutic proteins: clinical implications and
future prospects. Clin Ther 24:1720–1740
5.
Cournoyer D, Toffelmire EB, Wells GA, Barber DL, Barrett BJ, Delage R, Forrest DL,
Gagnon RF, Harvey EA, Laneuville P, Patterson BJ, Poon MC, Posen GA, Messner HA;
Canadian PRCA Focus Group. 2004. Anti-erythropoietin antibody-mediated pure red cell
aplasia after treatment with recombinant erythropoietin products: recommendations for
minimization of risk. J Am Soc Nephrol 15:2728–2734
6.
Jahn EM, Schneider CK. 2009. How to systematically evaluate immunogenicity of
therapeutic proteins - regulatory considerations. N Biotechnol 25:280–286
7.
Kijanka G, Jiskoot W, Schellekens H, Brinks V. 2013. Effect of treatment regimen on the
immunogenicity of human interferon Beta in immune tolerant Mice. Pharm Res 30:15531560
8.
Sharma B. 2007. Immunogenicity of therapeutic proteins. Part 2: impact of container
closures. Biotechnol Adv 25:318–324
9.
Verthelyi D, Wang V. 2010. Trace levels of innate immune response modulating impurities
(IIRMIs) synergize to break tolerance to therapeutic proteins. PloS One 5:e15252
10.
Ryan MH, Heavner GA, Brigham-Burke M, McMahon F, Shanahan MF, Gunturi SR,
Sharma B, Farrell FX. 2006. An in vivo model to assess factors that may stimulate the
generation of an immune reaction to erythropoietin. Int Immunopharmacol 6:647–655
11.
van Beers MM, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon Beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
12.
Sauerborn M, van Beers MM, Jiskoot W, Kijanka GM, Boon L, Schellekens H, Brinks
V. 2012. Antibody response against Betaferon® in immune tolerant mice: involvement of
marginal zone B-cells and CD4+ T-cells and apparent lack of immunological memory. J
Clin Immunol 33:255-263
13.
Ottesen JL, Nilsson P, Jami J, Weilguny D, Dührkop M, Bucchini D, Havelund S, Fogh JM.
1994. The potential immunogenicity of human insulin and insulin analogues evaluated in a
transgenic mouse model. Diabetologia 37:1178-1185
14.
Krieckaert CL, Bartelds GM, Lems WF, Wolbink GJ. 2010. The effect of immunomodulators
on the immunogenicity of TNF-blocking therapeutic monoclonal antibodies: a review.
Arthritis Res Ther 12:217
15.
Verhelst D, Rossert J, Casadevall N, Krüger A, Eckardt KU, Macdougall IC. 2004. Treatment
of erythropoietin-induced pure red cell aplasia: a retrospective study. Lancet 363:1768–
1771
16.
Casadevall N, Nataf J, Viron B, Kolta A, Kiladjian JJ, Martin-Dupont P, Michaud P, Papo T,
Ugo V, Teyssandier I, Varet B, Mayeux P. 2002. Pure red-cell aplasia and antierythropoietin
antibodies in patients treated with recombinant erythropoietin. N Engl J Med 346:469–475
73
Chapter 4.
17.
Hermeling S, Jiskoot W, Crommelin D, Bornaes C, Schellekens H. 2005. Development of
a transgenic mouse model immune tolerant for human interferon Beta. Pharm Res 22:847–
851
18.
van Beers MM, Sauerborn M, Gilli F, Hermeling S, Brinks V, Schellekens H, Jiskoot W.
2010. Hybrid transgenic immune tolerant mouse model for assessing the breaking of B cell
tolerance by human interferon beta. J Immunol Methods 352:32–37
19.
Hermeling S, Aranha L, Damen JM, Slijper M, Schellekens H, Crommelin DJ, Jiskoot W.
2005. Structural characterization and immunogenicity in wild-type and immune tolerant
mice of degraded recombinant human interferon alpha2b. Pharm Res 22:1997–2006
20.
Sharma B, Bader F, Templeman T, Lisi P, Ryan M, Heavner GA. 2004. Technical
investigations into the cause of the increased incidence of antibody-mediated pure red cell
aplasia associated with Eprex® EJHP 5:86–91
21.
Hermeling S, Schellekens H, Maas C, Gebbink MF, Crommelin DJ, Jiskoot W. 2006.
Antibody response to aggregated human interferon alpha2b in wild-type and transgenic
immune tolerant mice depends on type and level of aggregation. J Pharm Sci 95:1084–1096
22.
Mastelic B, Ahmed S, Egan WM, Del Giudice G, Golding H, Gust I, Neels P, Reed SG,
Sheets RL, Siegrist CA, Lambert PH. 2010. Mode of action of adjuvants: implications for
vaccine safety and design. Biologicals 38:594–601
23.
Vogel FR. 2000. Improving vaccine performance with adjuvants. Clin Infect Dis 30:S266–
270
24.
Carter D, Reed SG. 2010. Role of adjuvants in modeling the immune response. Curr Opin
HIV AIDS 5:409–413
25.
Wang W, Singh M. 2011. Selection of Adjuvants for Enhanced Vaccine Potency. WJV
01:33–78
26.
Chiarella P, Massi E, De Robertis M, Signori E, Fazio VM. 2007. Adjuvants in vaccines and
for immunisation: current trends. Expert Opin Biol Ther 7:1551–1562
27.
George-Chandy A, Eriksson K, Lebens M, Nordström I, Schön E, Holmgren J. 2001.
Cholera toxin B subunit as a carrier molecule promotes antigen presentation and increases
CD40 and CD86 expression on antigen-presenting cells. Infect Immun 69:5716-5725
28.
Peachman KK, Mclean DM, Tong JC, Alving CR, Rao M. 2009. Ganglioside GM1 binding
peptides: a potential adjuvant for transcutaneous immunization. Open Immunol J 2:94–102
29.
Barnard JG, Babcock K, Carpenter JF. 2013. Characterization and Quantitation of Aggregates
and Particles in Interferon-β Products : Potential Links Between Product Quality Attributes
and Immunogenicity. J Pharm Sci 102:915–928
30.
Fradkin AH, Carpenter JF, Randolph TW. 2009. Immunogenicity of Aggregates of
Recombinant Human Growth Hormone in Mouse Models. J Pharm Sci 98:3247–3264
31.
Sakaguchi S, Wing K, Onishi Y, Prieto-Martin P, Yamaguchi T. 2009. Regulatory T cells:
how do they suppress immune responses? Int Immunol 21:1105–1111
32.
Gaffen SL, Liu KD. 2004. Overview of interleukin-2 function, production and clinical
applications. Cytokine 28:109-123
33.
Carter LL, Dutton RW. 1996. Type 1 and type 2: a fundamental dichotomy for all T-cell
subsets. Curr Opin Immunol 8:336–342
34.
Malek TR. 2008. The Biology of Interleukin-2. Annu Rev Immunol 26:453-479
35.
Zhu J, Paul WE. 2008. CD4 T cells: fates, functions, and faults. Blood 112:1557–1569
36.
Prey S, Paul C. 2009. Effect of folic or folinic acid supplementation on methotrexate‐
associated safety and efficacy in inflammatory disease: a systematic review. Br J Dermatol
74
160:160-168
Impact of immunemodulators on immunogenicity of rhINFb.
75
Chapter 4.
Supplementary data.
S1. Material and methods.
S1.1. Conjugation of rhIFNB-1a to cholera toxin subunit B.
Before conjugation 1 mg of CTB was dialyzed overnight (O/N) against PBS pH 7.4 in order
to replace Tris buffer with PBS. After dialysis the CTB solution was supplemented with
EDTA to the final concentration of 1 mM. The conjugation was initiated by adding 20 ul
SPDP (25 mM in DMSO) to 1 ml of CTB solution. Reaction was executed for 60 min on
the bench top roller mixer at RT. The unbound SPDP was removed by dialysis against PBS.
After another supplementation with EDTA (final concentration 1 mM) 5 ml of rhIFNβ-1a
solution was added to CTB and incubated O/N on the roller mixer at RT. The conjugates
were diluted to a final concentration of 50 μg rhIFNβ-1a/ml.
S1.2. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis (SDS-PAGE).
The SDS-PAGE analysis was performed with the X-Cell Sure Lock@ Mini-Cell system
(Invitrogen, the Netherlands). 1 μg of rhIFNβ-1a or rhIFNB-1a/CTB conjugates were
loaded on the NuPAGE® 4-12% Bis-Tris Mini Gels of 1mm thickness (Novex®, Invitrogen,
the Netherlands). Protein samples were separated under non-reducing and reducing (with
5mM β-mercaptoethanol) conditions at 130 V for 60 minutes. Proteins were visualized by
silver staining.
S1.3.Western-Blot Analysis.
Proteins were transferred from SDS-PAGE gel onto the nitrocellulose membrane by iBlot®
Gel Transfer Device (Life technologies, the Netherlands). After transfer the membrane was
blocked in 4% milk in PBS supplemented with 0.1% Tween 20 (PBST/M). The rhIFNβ
was detected by polyclonal rabbit anti hIFNβ Abs (Promokine, Germany) diluted to a
concentration of 0.1μg/ml in PBST/M. Membrane was incubated with Abs for 1 hr at RT
and subsequently washed three times with PBST (each wash 5 min). Then goat anti rabbitHRP antibodies were applied (1:2500 in PBST/M, Cell Signaling Technology, Inc., USA)
and incubated for 1 hr at RT. Unbound Abs were removed by 4 washing steps with PBST
and one with PBS. The blot was developed by 3-Amino-9-ethylcarbazole (AEC) Staining
Kit according to manufacturer’s protocol (Sigma Aldrich, the Netherlands).
S2. Results.
The conjugation of rhIFNβ-1a and CTB resulted in formation of multiple species with the
minimal molecular weight of ~35kDa, which corresponds to rhIFNβ-1a/CTB conjugates
consisting of 1:1 rhIFNβ-1a CTB ratio (Supplementary Figure 1). Higher molecular weight
species contained multiple CTB and rhIFNβ-1a subunits. SDS-PAGE analysis revealed
presence of unconjugated CTB and rhIFNβ-1a in the final formulation. CTB and rhIFNβ1a were conjugated via disulfide bridges and addition of B-mercaptoethanol resulted in
degradation of conjugates and increase of CTB and rhIFNβ-1a detected on gel and blot
confirming successful conjugation.
76
M
1
Impact of immunemodulators on immunogenicity of rhINFb.
2
3
4
M
1
2
3
4
170 kDa
130 kDa
100 kDa
70 kDa
55 kDa
40 kDa
35 kDa
25 kDa
15 kDa
10 kDa
Supplementary Figure 1. A) The SDS-PAGE analysis of CTB and rhIFNβ-1a conjugation. Samples’
order: non-reducing conditions 1) rhIFNβ-1a, 2) CTB/rhIFNβ-1a conjugates, reducing 3) rhIFNβ-1a,
4) CTB/rhIFNβ-1a conjugates. B) Western blot staining confirming the presence of rhIFNβ-1a in
conjugates.
77
Chapter 4.
78
CHAPTER 5.
Recombinant human interferon beta forms
complexes with bacteria - potential risk factor
for immunogenicity?
Grzegorz Kijanka
Vera Brinks
Inga Jensch
Claudia Weber
Huub Schellekens
Andreas Greinacher
Krystin Krauel
79
Chapter 5.
Abstract.
Recombinant human interferon beta (rhIFNβ) is the first protein drug authorized for the
treatment of multiple sclerosis. However its clinical effect may be diminished by the
production of anti drug antibodies (ADA). Many factors have been suggested to contribute
to the development of ADA, however some reports suggest that bacterial infections may
play a role. Here, we have studied whether rhIFNβ binds to bacteria, and therefore if this
binding might increase the risk of ADA development.
RhIFNβ, aggregated or in its native form, binds effectively to different species of bacteria.
The binding is driven by electrostatic interactions, however rhIFNβ binding to Gram
positive bacteria is more pronounced than to Gram negative ones. Future studies should be
performed to determine if this binding could contribute to ADA formation against rhIFNβ.
80
1. Introduction.
Interaction of rhINFb and bacteria.
Multiple sclerosis (MS) is a debilitating disorder of the central nervous system that affects
about 2-3 million people worldwide (2-150 MS patients per 100 000 people depending on
the country) 1. The use of recombinant DNA techniques and the subsequent development
of recombinant human interferon beta (rhIFNβ) have revolutionized the treatment of MS.
RhIFNβ products decrease the number of and severity of relapses and lead to a slower
disease progression 2,3. Unfortunately, the effect of rhIFNβ therapy is often abrogated by
the development of anti-drug antibodies (ADA) 4–6. These antibodies may reduce efficacy
and even lead to treatment failure. Among many factors which contribute to the formation
of ADA, the presence of protein aggregates in the formulated drug has been found to be of
particular importance 7–9, however other factors might play a role as well.
In recent years, there has been increasing evidence that bacterial infections play an important
role in the development and progression of MS 10–14. For example, the cerebrospinal fluid of
MS patients has significantly higher prevalences of Chlamydia pneumoniae DNA and anti C.
pneumoniae antibodies compared to healthy individuals 15–17. Also, MS relapses have been
associated with the presence of Staphyloccoccus aureus enterotoxin A 18. Moreover, super
antigens produced by S. aureus were successfully used to induce experimental autoimmune
encephalomyelitis, a mouse model of MS 19. Additionally Schrijver et al 20 demonstrated the
presence of bacterial peptidoglycan in the antigen presenting cells in brains of MS patients.
We hypothesise that presence of bacteria might be associated with immunogenicity of
rhIFNβ.
For antibody formation against heparin, such a mechanism has been described. After
heparin treatment, some patients acquire heparin induced thrombocytopenia (HIT), an
immunogenicity reaction against complexes of platelet factor 4 (PF4) and heparin. Critical
in HIT is that administered heparin binds to PF4 and forms complexes of multiple PF4
molecules per one heparin chain. Patients suffering from infections might have a higher
risk to develop HIT, apparently because bacteria can form immunogenic complexes with
endogenous PF4 21. These PF4/bacteria complexes trigger antibodies which are crossreactive with PF4/heparin complexes 21–24.
We have studied whether bacterial infections might also be a risk factor for immunogenicity
of rhIFNβ products by determining whether rhIFNβ can bind to bacteria and form complexes.
A series of experiments were designed to first assess binding of (monomeric and aggregated)
rhIFNβ to bacteria, followed by a study to determine whether binding is preferred to Gram
positive or negative bacteria and by which mechanism this binding could take place. In
addition, we tested if human serum albumin (HSA), a stabilizing agent used in some rhIFNβ
products, could bind to bacteria. We have also studied the interaction of rhIFNβ with PF4.
2. Material and methods.
2.1. Recombinant human interferon beta.
Three formulations of rhIFNβ were prepared. The rhIFNβ-1a referred to as bulk rhIFNβ was
81
Chapter 5.
supplied by Biogen Idec (Cambridge, MA, US). Before use, bulk rhIFNβ was formulated
to a concentration of 270 μg/ml in 100 mM sodium phosphate pH 7.2, 200 mM sodium
chloride. To remove trace amounts of bacterial-size aggregates (>1000nm), biotinylated
bulk rhIFNβ was centrifuged (3700 g, 10 min, 4°C) directly before binding studies and
the supernatant was used for experiments. This supernatant was referred to as centrifuged
rhIFNβ. The third rhIFNβ formulation was obtained via metal catalysed oxidation of bulk
rhIFNβ as described in paragraph 2.2, and is referred to as stressed rhIFNβ.
2.2. Biotinylation and metal catalysed oxidation of rhIFNβ and HSA.
Before biotinylation, bulk rhIFNβ was freeze-dried and reconstituted in phosphate buffered
saline (PBS) to a concentration of 1 mg/ml. Lyophilized human serum albumin (SigmaAldrich, the Netherlands) was reconstituted in PBS to a concentration of 1 mg/ml. To 1
ml of either rhIFNβ or HSA solution a volume of 200 μl biotin solution (1 mg/ml, Sigma
Aldrich, the Netherlands) and 0.8ml of 0.1M NaHCO3 pH=9.6 was added. The biotinylation
was performed for 2 hours at RT, and unbound biotin was removed by overnight dialysis
against PBS. Biotinylation was confirmed via DotBlot (data not shown). After biotinylation
the concentrations of rhIRNβ-Biot and HSA-Biot were adjusted to 300 μg/ml.
Metal catalysed oxidation of rhIFNβ and rhIFNβ-Biot involved adding both CuCl2 and
ascorbic acid to the rhIFNβ solutions to a final concentration of 0.04 mM and 4 mM,
respectively. The reaction was executed at RT for 3 hours, and terminated by addition of
ethylenediaminetetraacetic acid (EDTA) to a final concentration of 1 mM. The oxidized
samples were dialyzed overnight against PBS.
Before initiation of metal catalysed oxidation of HSA-Biot with FeCl3 the HSA-Biot
formulation was dialyzed overnight against oxidation buffer (50mM HEPES, 100mM KCl,
10mM MgCl2, pH 7,4). Oxidation was initiated by adding FeCl3 and ascorbic acid to a final
concentration of 100 mM and 2.5 mM, respectively. The reaction was executed at 37°C for
one hour, after which it was terminated by addition of EDTA to a final concentration of 1
mM. After oxidation samples were dialyzed against PBS.
The concentration of stressed (metal catalyzed oxidized) and bulk rhIFNβ, stressed and bulk
rhIFNβ-Biot and stressed and unstressed HSA-Biot was determined by MicroBCA® assay
according to the manufacturer’s protocol (Thermo Fisher Scientific, The Netherlands).
2.3. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis.
SDS-PAGE was performed with the X-Cell Sur Lock® Mini-Cell system (Invitrogen, the
Netherlands). 2 μg of samples were loaded on the NuPAGE® 4-12% Bis-Tris Mini Gels of 1
mm thickness (Novex®, Invitrogen, the Netherlands). Protein samples were separated under
non-reducing conditions at 130 V for 60 minutes. Protein bands were visualized by silver
staining.
2.4. FITC labelling of bacteria.
Because the size of metal catalysed oxidation-induced HSA-Biot and rhIFNβ-Biot
82
Interaction of rhINFb and bacteria.
aggregates was similar to the expected size of bacteria (≥1000 nm), bacteria were labelled
with fluorescein isothiocyanate (FITC) in order to distinguish between complexes composed
of bacteria/protein and the protein aggregates. Bacteria (species indicated in paragraphs 2.5
and 2.6) were UV inactivated and diluted in PBS to a final concentration of 109 cells/ml. The
proper pH for labelling (pH=9) was obtained by adding 20 μl of 1M NaHCO3 pH 9 to 200
μl of a solution containing the inactivated bacteria. FITC (Life Technologies, Germany) was
dissolved in dimethylosulfoxide (DMSO, Life Technologies, Germany) to a concentration
of 7.4 mg/ml directly before starting the labelling reaction. Labelling was initiated by adding
10 μl of FITC to the solutions containing bacteria. The reaction was executed at RT for 1
hour with constant stirring (500 rpm). Free dye was removed a by series of 3 washing steps
with 10 ml of PBS. After washing, the bacterial pellet was suspended in 1 ml PBS.
2.5. Binding of rhIFNβ-Biot and HSA-Biot to FITC labelled bacteria.
2.5.1. Binding of rhIFNβ-Biot and HSA-Biot to a selection of FITC labeled bacteria.
Gram negative bacterial species (Escherichia coli JM109, Neisseria maningitidis MC58_
siaD, Escherichia coli K-12 wiltype, Escherichia coli K-12 waaA) and gram positive
bacteria (Staphylococcus aureus SA113spa, Streptococcus pneumoniae NCTC 10319,
Listeria monocytogenes L028 Sr 1/2c) were FITC labelled as described in section 2.4.
Stressed, unstressed and centrifuged rhIFNβ-Biot, unstressed HSA-Biot or unstressed PF4Biot (kindly provided by dr Krystin Krauel) dissolved in PBS supplemented with 0.5% of
BSA (PBS/B) were mixed with FITC labelled bacteria and incubated for 30 min at 4°C
with constant orbital shaking (800 rpm/min). The final concentration of proteins was 5 μg/
ml and bacteria 1.5x106 cells/ml. The mixture was then centrifuged (3000 g, 6 min, RT) and
washed with 3 ml of PBS/B was performed to remove unbound proteins. Bound rhIFNβBiot, HSA-Biot or PF4-Biot was detected by streptavidin-PE Cy5.5 (1:100 in PBS/B, BD
Biosciences, Germany) added to the pellet and incubated for 30 min at 4°C with constant
orbital shaking. After incubation, the free streptavidin was removed by wash with PBS/B,
after which bacteria were fixed in 1% paraformaldehyde in PBS. Binding was analysed on
a Cytomics FC 500 flow cytometer (Beckman Coulter, Germany).
2.5.2. Binding of rhIFNβ-Biot and HSA-Biot to an extended panel of bacteria.
To elaborate on the results of previous experiment and to validate the findings, a follow
up study was performed to test the binding of rhIFNβ-Biot and HSA-Biot to additional
bacteria strains. In addition, the experimental conditions of this experiment were improved
for FACS testing. The following strains were used: Gram negative Escherichia coli JM109,
Escherichia coli 536, Salmonella typhimurium rfa+ SL3770, Neisseria meningitidis
MC58ΔsiaD, Haemophilus influenza and Gram positive Staphylococcus aureus SA113,
Streptococcus pneumoniae NCTC 10319, Streptococcus mutans UA159, Streptococcus
pyogenes A158, Listeria monocytogenes L028 Sr 1/2c. As the low concentration of bacteria
after FITC labelling resulted in very time consuming FACS measurements (2.5.1), the final
concentration of bacteria used for binding was increased to ~2x107/ml (adjusted after FITC
labelling). The final concentration of proteins used for analysis (stressed and bulk rhIFNβ-
83
Chapter 5.
Biot, stressed and unstressed HSA-Biot and unstressed PF4-Biot) was also increased to 20
μg/ml. The binding was analyzed according to the protocol described in paragraph 2.5.1.
2.6. Mechanism underlying binding of rhIFNβ to bacteria – competition assays.
After showing that biotinylated rhIFNβ binds to bacteria in a similar manner as PF4 two
studies were performed to determine the nature of rhIFNβ binding to bacteria. The first
study was designed to determine if both rhIFNβ and PF4 recognize the same binding partner
on the bacterial surface. In the second study we tested if rhIFNβ binding to bacteria is due
to electrostatic interactions as shown for PF4 21.
In the first study a solution containing UV inactivated E. coli JM109 (final concentration
1.5x106/ml) was incubated with either PF4 or stressed rhIFNβ in a final concentration
range of 5 to 40 μg/ml. Solutions were then washed with PBS/B to remove any unbound
PF4 or stressed rhIFNβ. Then (i) stressed rhIFNβ-Biot (final concentration 5 μg/ml) was
added to the E. coli JM109 solution that was incubated with PF4 and (ii) PF4-Biot (final
concentration 5 μg/ml) was added to the E. coli JM109 solution that was incubated with
stressed rhIFNβ. Binding of either PF4-Biot or stressed rhIFNβ-Biot to bacteria was
measured with streptavidin-PE Cy5.5 as described in the paragraph 2.5.1.
To test if binding of rhIFNβ-Biot to E. coli JM109 was based on electrostatic interactions
a polyaninon (heparin) was used as competitor during the incubation of rhIFNβ-Biot with
E.coli JM109 during a binding assay. Heparin, as strong polyanion, blocks binding of
positively charged proteins to the negatively charged surface of bacteria. Stressed rhIFNβBiot, centrifuged rhIFNβ-Biot or PF4-Biot (final concentration 5 μg/ml) were incubated
with E.coli JM109 in the presence or absence of heparin (final concentration 100 U/ml).
After washing away unbound proteins, the binding of rhIFNβ-Biot and PF4-Biot to E. coli
JM109 was assessed by adding the streptavidin-PE Cy5.5 (protocol described in section
2.5.1).
2.7. Interaction of PF4 with rhIFNβ.
As the competition study (2.6.) suggested a possible interaction of rhIFNβ with PF4, an
ELISA was performed in order to test binding of these two proteins. NuncTM Maxisorp
(Thermo Scientific, Germany) plates were coated with 100 μl (2 μg/ml) of either bulk
rhIFNβ, stressed rhIFNβ or beta-2-glycoprotein (β2-GPI, a protein known to bind to PF4)
in 0.05M carbonate-bicarbonate buffer pH=9.6. After an overnight incubation the plate was
washed 5 times with PBS supplemented with 0.1% tween 20 (PBST). Unless indicated
otherwise, all reagents were incubated on the plate for 1 hour at RT. Each step was preceded
by 5 washing steps with PBST.
After coating, the plate was blocked with 2% BSA in PBST. Then PF4 diluted in blocking
solution was added to the wells at different concentrations (0-10 μg/ml). The bound PF4
was detected with 100 μl of anti PF4 rabbit antibody (1:680 in blocking buffer; Dianova,
Germany). Afterwards 100 μl of anti rabbit HRP (1:4000 in blocking buffer; Cell Signalling
Technology, Inc., USA) was added to the wells. The colour reaction was initiated by the
84
Interaction of rhINFb and bacteria.
addition of 3,3’,5,5’-Tetramethylbenzidine (TMB, Invitrogen, the Netherlands) and stopped
by 0.5M H2SO4. The OD450 was measured with ParadigmTM Detection Platform (Beckman
Coulter, Germany).
2.8. Potential cross-reactivity of anti PF4/bacteria antibodies with rhIFNβ/bacteria
complexes.
Anti PF4/heparin complexes antibodies were shown to be poly-reactive and not to bind
specifically to heparin or PF4 25. To test if PF4/bacteria complexes have similar epitopes
compared to rhIFNβ/bacteria complexes, cross-reactivity of anti PF4/heparin complex
induced antibodies with rhIFNβ/bacteria complexes were tested. Antibodies against PF4/
bacteria complexes and antibodies against rhIFNβ/bacteria complexes were affinity purified
by the “adsorption/elution” assay described by Krauel et al 21. Briefly, E. coli JM109 and S.
pneumoniae NCTC 10319 were mixed with either bulk rhIFNβ, stressed rhIFNβ or PF4 and
incubated for 30 min at 4°C with constant orbital shaking (800 rpm/min) to allow binding of
proteins to bacteria. The final concentration of proteins was 50 μg/ml and bacteria 1.25x106/
ml in PBS/B. The unbound proteins were washed away by 1 ml of PBS/B followed by
centrifugation (3000 g, 4°C, 5 min). Then a volume of 100 μl of human serum containing
antibodies against PF4/heparin complexes (prediluted 1:50 in PBS/B) were added to the
pellets, mixed and incubated for 30 min at 4°C with constant orbital shaking. After incubation
the samples were centrifuged and supernatants were collected (unbound fraction). In order
to dissociate antibodies bound to rhIFNβ/bacteria or PF4/bacteria complexes 200 μl 0.1M
glycine buffer pH=2.7 was added to the remaining pellets. After 5 min incubation at RT,
samples were centrifuged. Subsequently 180 μl of supernatants were transferred to tubes
containing 20 μl of Tris buffer pH=9 in order to neutralize the pH and avoid irreversible
damage of collected antibodies. Supernatants collected after acidic dissociation are referred
as “bound fraction”.
Bound and unbound fractions collected in the adsorption/elution assay using both rhIFNβ/
bacteria and PF4/bacteria complexes, were analyzed via ELISA to detect anti PF4/heparin
complex antibodies as described previously by Juhl et al 26. Briefly, NuncTM Covalink
Microtiter plates (Thermo Scientific, Germany) were coated with PF4 (20 μg/ml) or PF4/
heparin complexes (20 μg/ml of PF4 mixed with 0.5 U/ml of heparin). A volume of 100
μl of the unbound and bound fractions collected during the “adsorption/elution” assay
were applied on a washed plate and incubated for 1 hour at RT. To prove that antibodies
are specific to PF4/heparin complexes and not to monomeric PF4, the coated complexes
were dissociated by an excess of heparin (100 U/ml). After washing, antibodies bound to
the coated PF4/heparin complexes or PF4 were detected with goat anti-human IgG HRP
conjugates (Jackson ImmunoResearch, USA). The colour reaction was initiated by addition
of TMB and stopped by 0.5M H2SO4. The OD450 was measured with a ParadigmTM
Detection Platform (Beckman Coulter, Germany).
2.9. Statistical analysis.
85
Chapter 5.
First, data was assessed for normal distribution using a Shapiro-Wilk test. Because the data
were not normally distributed a Mann-Whitney or Kruskal-Wallis non-parametric test was
used to determine differences between experimental groups.
In the studies 2.5.1. and 2.5.2. differences in binding of bulk rhIFNβ, centrifuged rhIFNβ
and stressed rhIFNβ to bacteria were assessed by a Kruskal-Wallis test. Additionally, in
study 2.5.2., a Mann-Whitney test was used to assess differences in binding of bulk rhIFNβ,
centrifuged rhIFNβ and stressed rhIFNβ to the gram positive and gram negative bacteria.
In the study 2.6. a Mann-Whitney test was used to determine the effect of competitors (PF4
or Heparin) on the binding of rhIFNβ to E. coli JM109. In the study 2.7. differences in the
binding of PF4 to the different rhIFNβ formulations were assessed by a Mann-Whitney test.
3. Results.
3.1. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis.
SDS-PAGE analysis showed that biotinylation of HSA did not induce significant changes.
The intensity of the bands was similar for both HSA-Biot and HSA. No additional bands
were observed in HSA-Biot compared with native HSA (Figure 1). In contrast to HSA,
biotinylation of rhIFNβ appeared to result in a slightly higher level of trimers and tetramers
compared to non-biotinylated rhIFNβ. The CuCl2 catalysed oxidation of rhIFNβ resulted
in the formation of high molecular weight species visible on the top of the gel as a smear.
A)
B)
1
170 kDa
130 kDa
100 kDa
70 kDa
2
3
4
5
6
7
1
2
3
170 kDa
130 kDa
100 kDa
70 kDa
55 kDa
55 kDa
40 kDa
35 kDa
25 kDa
15 kDa
40 kDa
35 kDa
25 kDa
15 kDa
10 kDa
Figure 1. SDS PAGE analysis of rhIFNβ and HSA after biotinylation and oxidation. Proteins (2 μg)
were separated under non-reducing conditions and visualized by silver staining. Samples order A) 1.
Marker, 2. Bulk rhIFNβ, 3. Bulk rhIFNβ-Biot, 4. stressed rhIFNβ, 5. stressed rhIFNβ-Biot, 6. HSA,
7. HSA-Biot, B) 1. Marker, 2. HSA-Biot, 3. stressed HSA-Biot.
86
Interaction of rhINFb and bacteria.
No differences in SDS-PAGE analysis of batches of stressed rhIFNβ used in studies 2.5.1.
and 2.5.2. were detected (data not shown). Also the oxidation of HSA-Biot resulted in the
formation of a significant amount of high molecular weight species (Figure 1).
3.2. Binding of rhIFNβ and HSA to FITC labelled bacteria.
3.2.1. Binding of rhIFNβ-Biot and HSA-Biot to a selection of FITC labelled bacteria.
As shown in Figure 2A and 2B, bulk, centrifuged and stressed rhIFNβ-Biot all bind to
the tested bacteria, although not to the same degree. Especially bulk and stressed rhIFNβBiot gave rise to the formation of large IFNβ/bacterial complexes. Also, stressed rhIFNβ
appeared to bind stronger to gram positive bacteria than to gram negative ones. When bacteria
were incubated with centrifuged rhIFNβ free of aggregates, the maximum fluorescence
was much lower compared to that of bulk or stressed rhIFNβ. Binding of the different
rhIFNβ formulations to E. coli K-12 waa12, which lacks oligosaccharide component of
lipopolysaccharide, revealed higher binding capacity to the bacterial cell wall when lipid A
is fully exposed. In contrast to rhIFNβ-Biot, unstressed and stressed HSA-Biot did not show
an interaction with the tested bacteria.
3.2.2. Binding of rhIFNβ-Biot and HSA-Biot to an extended panel of bacteria.To
test the hypothesis that rhIFNβ-Biot has higher binding affinity/capacity to gram positive
bacteria compared to gram negative species a study in which additional bacterial species
were included was performed (Figure 2 C). In this study only bulk and stressed formulations
of rhIFNβ-Biot were used, and the number of bacteria cells for experiment was adjusted to
improve FACS analysis. Figure 3C shows that in contrast to the first study, bulk rhIFNβBiot has higher binding to all tested bacteria than stressed rhIFNβ (p<0.001). Both bulk
and stressed rhIFNβ-Biot formulations showed higher binding to the gram positive bacteria
than to the gram negative ones (p<0.001 and p=0.002 for bulk and stressed rhIFNβ-Biot,
respectively). Again no significant binding of HSA-Biot to the tested bacteria, either stressed
or unstressed, was observed.
3.3. Mechanism underlying binding of rhIFNβ to bacteria – competition assays.
In order to determine if rhIFNβ recognizes the same binding location as the PF4 on the
bacterial surface a competition assay was performed. As shown in Figure 3A, pre-incubation
of E. coli JM109 with PF4 did not inhibit the binding of stressed rhIFNβ-Biot to bacteria,
but instead significantly increased the binding (p=0.048). Increasing PF4 concentrations
resulted in higher binding of rhIFNβ-Biot to E. coli JM109. The inverse setup of the assay,
in which pre-incubation with rhIFNβ was followed by addition of PF4-Biot, resulted in
similar results. The binding of PF4-Biot was also increased when compared to the control
in which no rhIFNβ was used (p=0.048).
To determine the mechanism responsible for the binding of rhIFNβ-Biot to bacteria a
competition assay using an excess of heparin in the binding buffer was performed. Heparin,
as a polyanion, is known to compete with bacteria for binding to positively charged
molecules, like PF4-Biot 21. As shown in Figure 3B, heparin significantly affected the
87
Chapter 5.
binding of all tested proteins to E.coli JM109 (p<0.001). It completely blocked binding
of centrifuged rhIFNβ-Biot and PF4-Biot to bacteria. In contrast, although the heparin
significantly decreased binding of stressed rhIFNβ-Biot to E.coli JM109, about 20% of
initial binding was still observed.
3.4. Interaction between PF4 and rhIFNβ.
As shown in Figure 4, PF4 can bind to rhIFNβ. The binding efficiency strongly depended
on the type of rhIFNβ formulation used, with stressed rhIFNβ binding to PF4 significantly
lower than binding of bulk rhIFNβ to PF4 (p=0.038).
3.5. Potential cross-reactivity of anti PF4/bacteria antibodies with rhIFNβ/bacteria
complexes.
To test if antibodies formed against PF4/bacteria complexes can recognize the complexes
formed by rhIFNβ and bacteria an “adsorbtion/elution” test was performed. As shown in
Figure 5 antibodies against PF4/heparin complexes were found in “bound fractions” only
in samples collected from Escherichia coli JM109 and Streptococcus pneumoniae NCTC
10319 incubated with PF4. These antibodies were complex specific since no antibody binding
was detected when complexes were dissociated by an excess of heparin. No antibodies were
found in the “bound fractions” eluted from bacteria coated with bulk rhIFNβ or stressed
rhIFNβ.
A)
Buffer
IFNβ bulk
IFNβ stressed IFNβ centrifuged
HSA
PF4
Escherichia coli
JM109
Staphylococcus aureus
SA113spa
PE
Streptococcus pneumoiae
NCTC 10319
Listeria monocytogenes
L028 Sr 1/2c
Neisseria maningitidis
MC58_siaD
Escherichia coli
K-12 wiltype
Escherichia coli
K-12 waaA
FITC
88
Interaction of rhINFb and bacteria.
B)
Protein’s binding (GMFI x %)
10000
*
*
1000
*
*
*
*
100
*
10
IFNβ bulk
IFNβ stressed
IFNβ centrifuged
PF4
1
0,1
C)
Protein’s binding (GMFI x %)
10000
*
1000
100
10
1
IFNβ bulk
IFNβ stressed
HSA
HSA stressed
PF4
0,1
Figure 2. A) Density plots showing the binding of bulk rhIFNβ-Biot, stressed rhIFNβ-Biot,
centrifuged rhIFNβ-Biot, HSA-Biot and PF4-Biot to different bacteria in the experiment 2.5.1.
Only FITC positive events (bacteria) are presented. B) Adsorption of proteins to different bacteria
in the study 2.5.1. The binding is expressed as Cy5.5 geometric mean of fluorescence intensity of
all FITC positive events (bacteria) multiplied by the percentage of Cy5.5 positive events (GMFI *
%). Bars show the average binding ± SD of three independent experiments. Binding of rhIFNβ-Biot
to bacteria significantly differed between bulk, stressed and centrifuged rhIFNβ-Biot (* p=0.024).
C) Adsorption of proteins to different bacteria in the 2.5.2. study. Bars show the average binding ±
SD of three independent experiments. Binding of bulk and stressed rhIFNβ-Biot to Gram positive
bacteria was significantly higher than to Gram negative species (* p<0.001 for bulk rhIFNβ-Biot and
** p=0.002 for stressed rhIFNβ-Biot).
89
Chapter 5.
4. Discussion.
Immunogenicity of therapeutic proteins is a major factor limiting treatment efficiency and
can drastically increase therapy costs. Although many factors contributing to the production
of ADA have been identified, there is still a lack of understanding on why ADAs are formed.
It is unclear why the immune system is activated against a self-like antigen, and why ADA
production is triggered.
In this report we show that rhIFNβ, either monomeric or stressed, can bind to bacteria.
However, the binding differs between monomeric, bulk and stressed rhIFNβ formulations.
Moreover, we observed differences in binding of different batches of stressed rhIFNβ
between experiments 2.5.1. and 2.5.2.. Stressed rhIFNβ used in experiment 2.5.1. in general
bound better to bacteria than bulk rhIFNβ, in experiment 2.5.2., however, bulk rhIFNβ
binding was much more efficient than that of stressed formulation. It suggests that small
variation between batches of stressed rhIFNβ, which cannot be detected by SDS-PAGE
analysis, might affect binding of this formulation to bacteria. However, altered binding
efficiency might be also caused adjustments of rhIFNβ and bacteria concentrations in
experiment 2.5.2..
Our finding that rhIFNβ binds to bacteria, is similar to previous reports by Greinacher
and Krauel, in which an interaction between PF4 with bacteria was identified. They
additionally show that anti PF4/bacteria complexes antibodies might cross-react with PF4/
heparin complexes and contribute to HIT development 21,24. Binding of rhIFNβ to bacteria
therefore opens a new, hypothetical mechanism that could contribute to the induction of
immunogenicity. Our results clearly show that rhIFNβ is capable of binding to different
bacterial species and its adsorption seems to be more efficient to Gram positive bacteria. The
main mechanism underlying the binding is an electrostatic interaction as the net charge of
rhIFNβ at physiological pH is positive and bacterial wall carries negative charge 27. Indeed,
we showed that heparin drastically lowered binding of rhIFNβ to bacteria. Additionally,
HSA, which is negatively charged at pH 7.4, did not show any adsorption to the bacterial
wall.
It could be speculated that the interaction between rhIFNβ and bacteria might affect the
immunogenicity of rhIFNβ products. The study by Krauel et al suggested that the interaction
between PF4 and bacteria during infection can lead to the formation of antibodies, which
are cross-reactive with PF4/heparin complexes 24. One might imagine that rhIFNβ/bacteria
complexes might also raise an antibody response. It is commonly accepted that protein
aggregates are one of the most potent risk factors leading to an ADA response. Hypothetically,
rhIFNβ adsorbed to the bacterial surface might form an “aggregate-like” structure, which
could enhance the risk of immunogenicity in the same manner as protein aggregates. Uptake
of the rhIFNβ together with bacteria might hypothetically lead to simultaneous processing
of rhIFNβ and bacterial proteins. This might result in the presentation of rhIFNβ peptides
to the CD4+ T cells. In such sense the bacterial component of rhIFNβ/bacteria complexes
90
Interaction of rhINFb and bacteria.
A)
*
9000
*
8000
Gm*%
7000
6000
5000
IFNβ-Biot, PF4
4000
PF4-Biot, IFNβ
3000
2000
1000
0
B)
0
40
20
10
concentration of unbiotinylated PF4 or rhIFNβ [µg/ml]
10000
*
*
5
*
1000
Gm*%
IFNβ centrifuged
IFNβ stressed
PF4
100
10
No heparin
Heparin
Figure 3. A) Results of competition assay to study if PF4 and rhIFNβ recognize the same binding
area on the bacterial surface. Graph shows the binding of stressed rhIFNβ-Biot with or without pretreatment of bacteria with PF4 (black bars) and binding of PF4-Biot with or without pre-treatment of
bacteria with stressed rhIFNβ (white bars). The bars show average binding (GMFI * %) ± SD of three
independent experiments. Pre-treatment of bacteria with PF4 increased binding of stressed rhIFNβBiot to bacteria (* p=0.048). Pre-treatment of bacteria with stressed rhIFNβ also increased binding
of PF4-Biot to E.coli JM109 (* p=0.048). B) Binding of stressed rhIFNβ-Biot, centrifuged rhIFNβ
and PF4-Biot to E. coli JM109 is inhibited by the heparin (** p<0.001). Bars show average binding
(GMFI * %) ± SD of proteins to bacteria (n=3).
might work as adjuvant. As many of MS patients suffer from chronic bacterial infections,
such rhIFNβ/bacteria co-uptake might possibly lead to higher risk of ADA development.
Moreover, such mechanism might be an elegant explanation on why ADA formation occurs
in patients treated with rhIFNβ formulations containing low aggregate amounts. However,
future studies on this topic should be performed to test this hypothesis.
Although the binding of rhIFNβ to bacteria seems to be mainly electrostatically driven the
question remains unanswered why rhIFNβ binds more effectively to Gram positive bacterial
species. There are several explanations; i) teichoic acids may carry more negative charge
91
Chapter 5.
3.5
3
OD450
2.5
*
2
IFNβ stressed
IFNβ bulk
B2-GPIl
1.5
1
0.5
0
-0.5
0
2
4
6
8
10
12
PF4 [μg/ml]
Figure 4. ELISA results showing binding of PF4 to stressed rhIFNβ, bulk rhIFNβ and β2-GPI. Data
show average OD450 of duplicates ± SD. PF4 binding seemed to be preferable to bulk rhIFNβ than
to stressed rhIFNβ (* p=0.038).
than lipopolysaccharide (LPS) and therefore facilitate stronger binding of rhIFNβ than LPS,
ii) other components of Gram positive bacteria cell wall, e.g. peptidoglycan, might also bind
rhIFNβ-1a and/or, iii) the capacity of rhIFNβ binding might be species dependent and the
more effective binding to Gram positive species could have been a result of the selection
of bacteria for the experiment. Future studies, with more bacterial species, are needed to
confirm the preferable binding of rhIFNβ to Gram positive bacteria. Also, as binding of
stressed rhIFNβ-1a to E. coli JM109 was partially preserved in the presence of heparin, the
nature of binding should be investigated in more detail.
The biological importance of the interaction between rhIFNβ and bacteria remains unclear.
Krauel et al suggested that binding of PF4 to bacteria might be an ancient defence mechanism
against microorganisms 24. Adsorption of PF4 onto the bacterial surface induces antibodies,
which are specific for PF4/bacteria complexes, but do not recognize the soluble PF4 25.
Moreover they seem to be independent on the species of bacterial component forming
complexes 21. Therefore the formation of PF4/bacteria complexes could possibly protect
from infection of multiple bacterial species. One could hypothesize that IFNβ might have
a similar role. IFNβ has been shown to be produced not only during viral, but also during
bacterial infections 28,29. Therefore, hypothetically, PF4 and IFNβ might form an ancient
anti-microbial defence system. Such hypothesis seems to be indirectly supported by the
nature of IFNβ. Although IFNβ shares many activities with different IFNα subtypes, it has
not been completely replaced by them in the evolution process, and therefore indicates a
evolutionary relevant function of IFNβ over IFNα. Moreover, IFNβ was found to be crucial
for anti-viral responses 30,31. Potential participation of IFNβ in an ancient defence mechanism
might be a clue to understanding why IFNβ is very conserved. However, data showing the
92
OD450
E. coli JM109
4
3.5
3
2.5
2
1.5
1
0.5
0
Interaction of rhINFb and bacteria.
Buffer
PF4
IFNβ bulk
IFNβ stressed
Bound fraction
Bound fraction
+ heparin
Unbound fraction
PF4/Heparin coating
Bound fraction
PF4 coating
OD 450
S. pneumoniae
4
3.5
3
2.5
2
1.5
1
0.5
0
Buffer
PF4
IFNβ bulk
IFNβ stressed
Bound fraction
Bound fraction
+ heparin
Unbound fraction
PF4/Heparin coating
Bound fraction
PF4 coating
Figure 5. Result of ELISA for testing the binding of anti rhIFNβ/E.coli and rhIFNβ/S.pneumoniae
affinity purified antibodies to PF4/heparin complexes. Bars show the average OD450±SD of two
independent experiments.
existence of anti-IFNβ/bacteria complex antibodies in humans, similar to the presence of
anti PF4/bacteria complex antibodies, is lacking and future studies are necessary to test such
hypothesis.
To examine if rhIFNβ and PF4 share the binding partner we performed a competition
assay. To our surprise we discovered that rhIFNβ is capable to bind to PF4. Moreover, we
found that bulk rhIFNβ is more susceptible to bind to PF4 than stressed rhIFNβ is. The
possible biological significance of this finding is unknown, although it might suggest a
complementary role of IFNβ and PF4 in a defence mechanism against bacteria. However,
lack of cross-reactivity between anti PF4/heparin complex antibodies with rhIFNβ/heparin
complexes, might suggest otherwise. Thus, more detailed studies need to be conducted to
determinate the potential meaning of the interaction between IFNβ and bacteria.
93
Chapter 5.
5. Conclusions.
We found that rhIFNβ is capable of binding to bacteria. This binding is most likely driven
by electrostatic interactions. Moreover, we did observe more efficient binding of rhIFNβ
to Gram positive bacteria compared to Gram negative bacteria. As many of MS patients
suffer from chronic bacterial infections, the interaction between rhIFNβ and bacteria may
influence immunogenicity of rhIFNβ.
94
References.
Interaction of rhINFb and bacteria.
1. Rosati G. 2001. The prevalence of multiple sclerosis in the world: an update. Neurol Sci
22:117–139
2. Cohen BA, Rivera VM. 2010. PRISMS: the story of a pivotal clinical trial series in multiple
sclerosis. Curr Med Res Opin 26:827–838
3. Paty DW, Li DK. 1993 Interferon beta-lb is effective in relapsing-remitting multiple
sclerosis. II. MRI analysis results of a multicenter, randomized, double-blind, placebocontrolled trial. Neurology 43:662-667
4. Kappos L, Clanet M, Sandberg-Wollheim M, Radue EW, Hartung HP, Hohlfeld R, Xu J,
Bennett D, Sandrock A, Goelz S; European Interferon Beta-1a IM Dose-Comparison Study
Investigators. 2005. Neutralizing antibodies and efficacy of interferon beta-1a: a 4-year
controlled study. Neurology 65:40–47
5. Van der Voort LF, Gilli F, Bertolotto A, Knol DL, Uitdehaag BM, Polman CH, Killestein
J. 2010. Clinical effect of neutralizing antibodies to interferon beta that persist long after
cessation of therapy for multiple sclerosis. Arch Neurol 67:402–407
6. Cohen BA, Oger J, Gagnon A, Giovannoni G. 2008. The implications of immunogenicity
for protein-based multiple sclerosis therapies. J Neurol Sci 275:7–17
7. Van Beers MM, Jiskoot W, Schellekens H. 2010. On the role of aggregates in the
immunogenicity of recombinant human interferon beta in patients with multiple sclerosis. J
Interferon Cytokine Res 30:767–775
8. Barnard JG, Babcock K, Carpenter JF. 2013. Characterization and Quantitation of Aggregates
and Particles in Interferon-β Products : Potential Links Between Product Quality Attributes
and Immunogenicity. J Pharm Sci 102:915–928
9. Van Beers MM, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon Beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
10. Kakalacheva K, Lünemann JD. 2011. Environmental triggers of multiple sclerosis. FEBS
Lett 585:3724–3729
11. Casetta I, Granieri E. 2000. Clinical infections and multiple sclerosis: contribution from
analytical epidemiology. J Neurovirol 6:S147–S151
12. Nicolson GL, Haier J. 2009. Role of chronic bacterial and viral infections in
neurodegenerative, neurobehavioral, psychiatric, autoimmune and fatiguing illnesses: Part
1. BJMP 2:20–28
13. Milo R, Kahana E. 2010. Multiple sclerosis: geoepidemiology, genetics and the environment.
Autoimmun Rev 9:A387–A394
14. Nicolson GL, Haier J. 2010. Role of chronic bacterial and viral infections in
neurodegenerative, neurobehavioral, psychiatric, autoimmune and fatiguing illnesses: Part
2 BJMP 3:301
15. Sriram S, Mitchell W, Stratton C. 1998. Multiple sclerosis associated with Chlamydia
pneumoniae infection of the CNS. Neurology 50:571–572
16. Yao SY, Stratton CW, Mitchell WM, Sriram S. 2001. CSF oligoclonal bands in MS include
antibodies against Chlamydophila antigens. Neurology 56:1168–1176
17. Sriram S, Stratton CW, Yao S, Tharp A, Ding L, Bannan JD, Mitchell WM. 1999. Chlamydia
pneumoniae infection of the central nervous system in multiple sclerosis. Ann Neurol 46:6–
14
18. Mulvey MR, Doupe M, Prout M, Leong C, Hizon R, Grossberndt A, Klowak M, Gupta A,
95
Chapter 5.
Melanson M, Gomori A, Esfahani F, Klassen L, Frost EE, Namaka M. 2011. Staphylococcus
aureus harbouring Enterotoxin A as a possible risk factor for multiple sclerosis exacerbations.
Mult Scler 17:397–403
19. Brocke S, Gaur A, Piercy C, Gautam A, Gijbels K, Fathman CG, Steinman L. 1993
Induction of relapsing paralysis in experimental autoimmune encephalomyelitis by bacterial
superantigen. Nature 365:642-644
20. Schrijver IA, van Meurs M, Melief MJ, Wim Ang C, Buljevac D, Ravid R, Hazenberg MP,
Laman JD. 2001. Bacterial peptidoglycan and immune reactivity in the central nervous
system in multiple sclerosis. Brain 124:1544–1554
21. Krauel K, Pötschke C, Weber C, Kessler W, Fürll B, Ittermann T, Maier S, Hammerschmidt
S, Bröker BM, Greinacher A. 2011. Platelet factor 4 binds to bacteria, [corrected] inducing
antibodies cross-reacting with the major antigen in heparin-induced thrombocytopenia.
Blood 117:1370-1378
22. Greinacher A. 2009. Heparin-induced thrombocytopenia. J Thromb Haemost 7:9–12
23. Franchini M. 2005. Heparin-induced thrombocytopenia: an update. Thromb J 3:14
24. Krauel K, Weber C, Brandt S, Zähringer U, Mamat U, Greinacher A, Hammerschmidt S.
2012. Platelet factor 4 binding to lipid A of Gram-negative bacteria exposes PF4/heparinlike epitopes. Blood 120:3345–3352
25. Greinacher A, Michels I, Mueller-Eckhardt C. 1992. Heparin-associated thrombocytopenia:
the antibody is not heparin specific. Thromb Haemost 67:545–549
26. Juhl D, Eichler P, Lubenow N, Strobel U, Wessel A, Greinacher A. 2006. Incidence and
clinical significance of anti-PF4/heparin antibodies of the IgG, IgM, and IgA class in
755 consecutive patient samples referred for diagnostic testing for heparin-induced
thrombocytopenia. Eur J Haematol 76:420-426
27. Silhavy TJ, Kahne D, Walker S. 2010. The bacterial cell envelope. Cold Spring Harb
Perspect Biol 2:a000414
28.
Monroe, K. M., McWhirter, S. M. & Vance, R. E Monroe KM, McWhirter SM, Vance RE.
2010. Induction of type I interferons by bacteria. Cell microbiol 12:881–890
29.
Solodova E, Jablonska J, Weiss S, Lienenklaus S. 2011. Production of IFN-β during Listeria
monocytogenes infection is restricted to monocyte/macrophage lineage. PloS One 6:e18543
30.
Theofilopoulos AN, Baccala R, Beutler B, Kono DH. 2005. Type I interferons (α/β) in
immunity and autoimmunity. Annu Rev Immunol 23:307-336
31.
Samuel CE. 2001. Antiviral Actions of Interferons. Clin Microbiol Rev 14:778–809
96
CHAPTER 6.
Effect of treatment regimen
on the immunogenicity of human interferon
beta in immune tolerant mice.
Grzegorz Kijanka
Wim Jiskoot
Huub Schellekens
Vera Brinks
Pharm Res. 2013. 30(6):1553-60.
97
Chapter 6.
Abstract.
Interferon beta is commonly used as therapeutic in the first line of therapy for multiple
sclerosis. However, depending on the product, it induces an antibody response in up to 60%
of patients. This study evaluated the impact of therapy related factors like dose, route of
administration and administration frequency on the immunogenicity of one of the originator
interferon beta drugs (Betaferon®) in an immune tolerant transgenic mouse model.
Immune tolerant transgenic mice received injections with Betaferon® via different routes,
doses and injection frequencies. Anti-drug antibodies (ADA) production was measured by
ELISA to assess immunogenicity.
A single injection of Betaferon® was found to be sufficient for the induction of ADA. The
antibody titer was enhanced with increasing dose and treatment frequency. Among the
tested administration routes, the intravenous route was the most immunogenic one, which
is in contradiction with one of the dogma in immunogenicity research according to which
subcutaneous administration is the most immunogenic route. Intramuscular, intraperitoneal
and subcutaneous injections resulted in comparable immunogenicity.
This study shows that treatment related factors affect significantly immunogenicity of
Betaseron® and therefore substantiate the need for further studies on these factors in patients.
98
1. Introduction.
Impact of treatment regimen on immunogenicity.
Recombinant human interferon beta (rhIFNβ) is an approved drug for the treatment of
multiple sclerosis (MS). It improves the quality of patients’ lives by reducing relapses and
disease progression 1. However, rhIFNβ may lead to an antibody response that abolishes
its therapeutic effect 2,3. Of the rhIFNβ products of which the immunogenic properties
are relatively well investigated, Betaferon® (rhIFNβ-1b) has been found to be the most
immunogenic followed by Rebif® and Avonex® (rhIFNβ-1a) 4-7. Although aggregates are
the main factors causing immunogenicity of different rhIFNβ products, other variables,
including treatment factors, may also contribute to the development of anti-drug antibodies
(ADA) 4,5,8,9. Nonetheless, clear evidence indicating which treatment variables increase
the incidence of ADA and magnitude of antibody response is lacking. The majority of
clinical studies compare immunogenicity of different dosing, frequency of injections and
route of administration, while different products are used. Vice versa, studies that compare
immunogenicity of the different products, do not take into account that they have their
distinct route/dose and frequency of administration. Moreover, the various studies are
difficult to compare because of a lack of standardization of the ADA assays 10,11. As a result,
conclusions on how treatment-related factors affect immunogenicity of rhIFNβ products are
often contradictory 8,12-15.
In this study we performed a series of experiments to evaluate to what extent the dose,
frequency, and route of administration would affect immunogenicity of a single rhIFNβ
product, Betaferon®. For this we used the immune tolerant transgenic mouse model
developed by Hermeling et al 6 which was later optimized by van Beers et al 16. We found that
all investigated treatment factors may significantly modify ADA production. To compare
effect of treatment variables on the ADA production during immunogenicity reaction with
classical immune response against foreign protein a non-transgenic mice were included in
the study.
2. Materials and methods.
2.1. Interferon Beta.
Betaferon® (rhIFNβ-1b) was obtained from Schering (Berlin, Germany). Lyophilised
powder containing 300 μg of rhIFNβ, 15 mg of human serum albumin (HSA) and 15 mg of
mannitol was reconstituted in 1 ml of 10 mM sodium phosphate pH 7.4, 137 mM sodium
chloride (PBS). Before injection, Betaferon® was further diluted to desired concentrations
in PBS (see sections below). RhIFNb-1a (Avonex® drug substance) was supplied by Biogen
Idec and was formulated before use at a concentration of 270 μg/ml in 100 mM sodium
phosphate pH 7.2, 200 mM sodium chloride.
2.2. Animals.
Heterozygote transgenic C57Bl/6 mice carrying the human interferon beta (hIFNβ) gene
were crossed with wildtype FVB/N mice (Janvier, BioServices,Uden, The Netherlands).
The genotype of the offspring (F1) was determined using PCR showing the presence or
99
Chapter 6.
absence of the hIFNβ gene in chromosomal DNA isolated from ear tissue. Both PCR positive
(transgenic) and PCR negative (non-transgenic) littermates of 6-10 weeks of age were used
for the experiments. Mice were housed in standard perspex cages and given access to food
(Hope Farms, Woerden, The Netherlands) and water (acidified) ad libitum. All testing was
in accordance with the European guidelines on animal experiments.
2.3. Experimental setup.
Experiment 1: Number and frequency of injections.
Both non-transgenic (non-tg) and transgenic (tg) mice were divided into 4 experimental
groups which were administered either 1, 2, 3, or 4 intraperitoneal (IP) injections of
Betaferon® (n=6 per group, dose: 5 μg/injection). The injections were given on consecutive
days starting at day 1. Blood was collected before injections on day 0, 8 and 15. On day
21, the mice were sacrificed by cervical dislocation preceded by blood collection via heart
puncture under isoflurane anesthesia.
In addition, non-Tg and Tg mice were subjected to either 1 or 2 IP injections of Betaferon®
per week for 6 weeks (n=6 per group, dose 5 μg/injection). Animals receiving one injection
per week were treated on days 1, 8, 16, 22, 29, and 36. Mice given two injections per week
were additionally treated on days 3, 11, 18, 25, 32, and 39. Blood was collected by cheek
puncture before injections (day 0) and every week during the first 3 immunization weeks
(days 7, 14, and 21). At 6 weeks after the first injection (day 42), the mice were sacrificed
by cervical dislocation preceded by blood collection via heart puncture under isoflurane
anesthesia.
Plasma was isolated by spinning down the blood (3000 g at 4°C for 10 min) and was stored
at -20°C until further analysis.
Experiment 2: Route of administration.
Both non-tg and tg mice were injected with Betaferon® for 3 weeks (n=12 per group, dose
5 μg/injection), on days 1 to 5, 8 to 12, and 15 to 18 via the IP, subcutaneous (SC, neck),
intramuscular (IM, top part of hind legs) or intravenous route (IV, tail vein). Blood was
collected by cheek puncture before (day 0) and during the treatment weeks (days 5, 8,
12, and 15). To prevent interference of ADA-drug complexes, blood was collected before
injections. On day 19 the mice were sacrificed by cervical dislocation preceded by blood
collection via heart puncture under isoflurane anesthesia.
Plasma was isolated and stored as described above.
Experiment 3: Administration dose.
Tg and non-tg animals (n=10 per group) received 5 IP injections of Betaferon® per week
for 3 weeks (days 1 to 5, 8 to 12, and 15 to 19). They were treated with a dose of either 0.1
μg, 0.5 μg, 1 μg, 2.5 μg, 5 μg or 20 μg Betaferon® per injection. Blood was collected by
cheek puncture before the start of injections (day 0) and twice a week during the 3 week
treatment (days 5, 8, 12, 15, and 19). During the treatment weeks, blood was collected
100
Impact of treatment regimen on immunogenicity.
before injections to prevent interference of ADA-drug complexes.
Plasma was isolated and stored as described above.
2.4. Antibody assay.
ADA titers were measured by a modified direct sandwich ELISA described previously
6
, using rhIFNb-1a (Avonex®) to coat ELISA plates and peroxidise labeled antimouse IgG (Invitrogen, Bleiswijk, The Netherlands) as detecting antibodies, and TMB
[3,3’,5,5;-TetraMethylBenzidine] (Invitrogen, Bleiswijk, The Netherlands) as substrate.
100-fold diluted plasma was screened and defined positive if the background corrected
absorbance values were ten times higher than the average absorbance of the pre-treatment
sera. The titer of ADA-positive plasma was determined by plotting the absorbance values at
450 nm (OD450) of a serial dilution against log dilution. The plots were fitted to a sigmoidal
dose-response curve using GraphPad Prism version 4.00 (GraphPad Software, San Diego
CA, USA). The reciprocal of the dilution of the EC50 value was considered the titer of the
serum. Since ADAs’ levels of the animals in experiment 1 were considered positive using
the cutoff as described before, but were too low to calculate titers as described, OD450
values using a dilution of 1:100 were used for calculations.
2.5. Statistical analysis.
Before determining statistical differences in antibody titers all data were tested for normal
distribution. Because antibody titers were not normally distributed, non-parametric MannWhitney or Kruskal-Wallis tests were used to assess statistical differences between groups.
All calculations were performed using SPSS 16.0 software and a p value ≤0.05 was
considered significant.
3. Results.
Experiment 1: Number and frequency of injections.
In order to assess the minimal number of injections necessary to induce ADA formation,
animals were injected with Betaferon® one, two, three, or four times on consecutive days.
As shown in Figure 1, a single injection was sufficient to evoke an ADA response in both the
tg and non-tg animals. However, for both types of animals, the ADA production was much
lower than in other experiments in which treatment lasted for at least 3 weeks (instead of at
most 4 days). Therefore OD450 values were used as a measuret of ADA production. For all
treatment groups the OD450 values in the non-tg mice were significantly higher on days 14
and 20 compared to the tg animals (p=0.008 at day 14 and p<0.001 at day 21). Increasing
the number of injections did not results in significantly higher OD450 levels in the tg mice
(p=0.412) or in the non-tg (p=0.765).
In the second part of this experiment, the impact of the frequency of injection on ADA
production was tested. tg and non-tg mice received either one or two IP injections of
Betaferon® per week for a total time period of 6 weeks. The two injections per week
schedule resulted in higher ADA levels compared to one injection per week in both tg and
101
Chapter 6.
non-Tg
A)
OD 450nm
3
*
**
1 injection
2
2 consecutive injections
3 consecutive injections
4 consecutive injections
1
0
0
7
B)
Day
14
21
*
**
Tg
OD 450nm
3
2
1 injection
2 consecutive injections
3 consecutive injections
4 consecutive injections
1
0
0
7
Day
14
21
Figure 1. Anti-rhIFNβ antibody levels in non-tg (A) and tg (B) mice after 1, 2, 3, and 4 consecutive IP
injections (dose 5 μg per injection). Bars represent average OD 450nm for 6 animals and corresponding
SEM. A higher number of injections did not result in significantly higher ADA production in either tg
or non-tg animals. However, ADA levels were significantly higher in non-Tg than Tg animals on day
14 (p=0.008 * ) and day 21 (p<0.001 **).
non-tg mice (Figure 2) with p<0.001 and p=0.009, respectively. Interestingly, ADA titers
in tg animals remained almost unchanged and relatively low during the treatment period,
whereas in non-tg mice production of ADAs seemed to intensify over time. On day 42 the
ADA titers in non-tg mice were 6-10 times higher than on day 21 (p=0.017).
Experiment 2: Route of administration. As shown in Figure 3, ADA production strongly depended upon the route of administration
and the effect of administration route differed between tg and non-tg mice. For the non-tg
mice the least immunogenic route was SC administration, resulting in the lowest ADA titers
102
Impact of treatment regimen on immunogenicity.
(SC vs IM p=0.012, SC vs IP p=0.025, SC vs IV p=0.001). In addition, it appears that this
route resulted in the lowest number of ADA-positive non-tg animals during the treatment
period. Administration of Betaseron® via the IV route, which is commonly believed to be the
least immunogenic one 17,18, resulted in ADA titers in non-tg mice that were comparable to
those induced via the IM and IP routes of administration (p=0.133 and p=0.848 respectively).
In the tg animals, the SC route also seemed to be the least immunogenic, but the difference
in ADA titers between SC/IM and SC/IP routes was not significant (p=0.377 and p=0.335,
respectively). However, the onset of the ADAs response in the tg animals seemed to be
slower for the SC-treated mice compared to the animals injected via the IM or IP route.
Surprisingly, the administration route leading to the highest ADA production in the tg
mice was IV (IV vs SC p=0.001, IV vs IM p=0.002, and IV vs IP p=0.018). Injection of
Betaseron® via IP and IM routes resulted in similar ADA titers (p=0.692).
Experiment 3: Administration dose.
Animals were treated with 1 of 6 different doses of Betaferon® in the range of 0.1 μg – 20
μg per IP injection (5 injections per week for 3 weeks). As shown in Figure 4 and in Table 1,
ADA production depended upon the dose administered. The higher the dose used, generally
the higher the ADA production in both Tg (p<0.01) and non-Tg (p<0.01) mice. Moreover,
increasing the dose resulted in a faster onset of ADA response and a higher number of ADA
positive animals (p=0.0028 for tg animals, p=0.0167 for non-tg animals). In addition, as
shown in Table 1 and Figure 4, the onset of ADAs appeared to be similar in tg and non-tg
mice. The overall ADA titers in non-tg mice were significantly higher than in the tg mice
(p<0.01).
4. Discussion.
The treatment regimen used for the administration of rhIFNβ during MS therapy has been
proposed to be one of the major factors influencing ADA production 17,19. However, it is not
known which treatment component, i.e. dose, frequency, or route of administration, is the
Anti IFNβ titer
25000
** *
20000
Tg 1 injection
Tg 2 injections
non-Tg 1 injection
non-Tg 2 injections
15000
10000
5000
0
** *
14
** *
21
Day
42
Figure 2. Anti-rhIFNβ antibody titers in animals injected once or twice per week for 6 weeks with
Betaferon. Bars show average titer of anti-rhIFNβ IgG and corresponding SEM (n=6). Two instead
of 1 injections per week significantly increased ADA production (Tg p<0.001 **, non- Tg p=0.009 *).
103
Tg
Non-Tg
8000
*
*
4000
2000
0
6/6 6/6
0/6 3/6
5/6 6/6
*
12/12 12/12
8
12
Day
15
Anti INFβ titer
**
**
**
6000
6/6 6/6
4000
0
12/12 12/12
5/6 5/6
8
E)
8
12
Tg
Non-Tg
Day
18/19
F)
Tg
Non-Tg
8000
6/6 6/6
6000
4000
0
18/19
p=0,002
p=0,018
15
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
12/12 12/12
6/6 6/6
3/6 6/6
8
12
Day
15
18/19
p=0,001
p=0,025
p=0,012
Anti INFβ titer
p=0,001
15
Day
IV
2000
Anti INFβ titer
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
3/6 6/6
10000
6/6 6/6
12
5/6 6/6
D)
10000
8000
12/12 12/12
4000
18/19
Tg
Non-Tg
*
6/6 6/6
6000
2000
IM
*
*
8000
0
C)
2000
10000
Anti INFβ titer
Anti INFβ titer
10000
6000
IP
B)
SC
A)
Anti INFβ titer
Chapter 6.
SC
IM
IP
IV
SC
IM
IP
IV
Figure 3. Effect of route of administration on ADA response. Development of ADAs over time in
animals injected SC (A), IM (B), IP (C), and IV (D). Bars show averages of titer of anti-rhIFNβ total
IgG with corresponding SEM. Antibody responses in tg mice were statistically lower compared to
non-tg mice for SC, IM, and IP administration (SC and IP p<0.05; IM p<0.001). Comparison of ADA
production within the treatment period in Tg (E) and non-Tg (F) animals.
most prominent in reducing or enhancing ADA production to rhIFNβ. Reports are often
contradictory and therefore the impact of treatment related factors on ADA production is
not clear 12,13,15.
The experiments described here show that different treatment-related factors affect the
antibody response against rhIFNβ. We show that even a single administration of Betaferon®
can lead to formation of ADA in our immune tolerant mice. Interestingly, increasing the
104
Impact of treatment regimen on immunogenicity.
non-Tg
Anti IFNβ titers
0.1µg
0.5µg
1µg
2.5µg
5µg
20µg
10000
5000
0
8
12
Day
15
Tg
B)
15000
19
15000
Anti IFNβ titers
A)
0.1µg
0.5µg
1µg
2.5µg
5µg
20µg
10000
5000
0
8
12
Day
15
19
Figure 4. Formation of anti-rhIFNβ antibodies in tg (A) and non-tg (B) animals injected with 0.1,
0.5, 1, 2.5, 5 and 20 μg Betaferon. Bars represent average titer of ADA-positive animals and
corresponding SEM.
number of daily injections did not increase ADA levels significantly. This suggests that the
magnitude of ADA response might depend on the treatment duration and that a treatment
duration limited to a maximum of 4 days is too short to strongly activate the immune system
in the immune tolerant mice. In the second part of experiment 1, animals were given 1 or 2
injections of Betaferon® per week for 6 weeks. In non-tg animals the ADA levels intensified
over time, which suggests that injections given to these mice in the second and later weeks
might have resulted in a booster effect as found after vaccine administration. In contrast,
in tg animals ADA titers from day 14 onward remained unchanged. The cumulative results
from the two parts of experiment 1 show that the probability of an immune response and
magnitude of ADA production against Betaferon® increases with higher administration
frequency and/or higher number of injections and longer treatment duration. However,
in tg animals the impact of treatment duration seems to be more complex than in non-tg
mice; longer treatment duration in the second part of experiment 1 resulted in higher ADA
production but a lack of an booster effect led to stable ADA titers over time.
These data partially correlate with clinical observations. We show here that in both tg and
non-tg mice, even a single application of Betaferon® is sufficient to induce ADAs In contrast,
in MS patients ADA production begins 3 to 12 months after the start of treatment 20. On the
other hand, stable ADA levels found in tg mice correspond to clinical data, where ADA titers
have a tendency to persist unchanged for a long period of time or even show a tendency
to decline 8,20. Moreover, 2 injections of Betaferon® per week resulted in higher ADA
production in both types of mice. Also in patients, products administered more frequently
seem to be more immunogenic. Betaferon® and Rebif® which require administration 3-4
times per week are more immunogenic in terms of neutralizing antibodies than Avonex®
which is injected once a week. However, more extensive studies in humans focusing only
on one product are necessary to fully confirm the data presented here.
105
Chapter 6.
The most commonly accepted hierarchy of the immunogenicity of administration routes
places IV as the least and SC as the most immunogenic one. Moreover, this classification
seems to be confirmed by pre-clinical and clinical data 21,22. However, the finding that the IV
route is highly immunogenic corresponds to at least two other reports on immunogenicity
of biotherapeutics. An anti-TNFα antibody was found to induce higher antibody titers in
cynomolgus monkeys when injected IV than after SC application 23. Similarly, a PEGylated
form of Factor VIII was more immunogenic in haemophilia A mice when administrated IV
24
. This study, together with these reports, shows that immunogenicity is dependent on the
route of administration and that immunogenicity due to different routes might not always
be in accordance with generally believed dogma in clinical practice 4,17. Clinical testing
in patients should therefore confirm which route of application is most safe, ideally for
different therapeutic protein drugs.
Why the different routes of administration are more or less immunogenic will probably
be a combination of product characteristics, such as the presence of aggregates, the
microenvironment present at the site of injection, and kinetics of drug disposition. A higher
aggregate content may result in prolonged exposure at the site of injection, giving more time
for activation of local immune cells.
High immunogenicity due to SC administration is usually explained by an elongated exposure
of the antigen to anigen-presenting cells, increasing the probability of their activation
and subsequent stimulation of T- and B cells 25. The low immunogenicity of the SC route
observed in our experiment might be the effect of a difference in skin-muscle connections
in mice compared to humans, resulting in faster spreading of injected solution under the
skin than in humans. That might lead to faster clearance of protein and as a consequence
different activation of skin immune cells. However, the low immunogenicity of rhIFNβ
after SC injection might be an intrinsic property of this protein. IM and IP administration
may create a drug deposit inside the muscle or peritoneal cavity that slowly releases the
protein. Slow release of protein from the injection site, together with close proximity to
draining lymph nodes might explain relatively high immune response in non-tg mice after
administration of proteins via those routes. Since Betaferon® is a foreign protein for non-tg
animals it is expected to activate immune cells more efficiently than in tg mice. Indeed, tg
animals developed very similar ADA titers after SC, IM and IP.
A high content of aggregates may be responsible for the high immunogenicity of Betaferon®
after IV administration. Large aggregates present in Betaferon® may be easily recognized
and taken up by splenic macrophages and therefore might lead to activation of other
immune cells e.g. marginal zone B cells. Frequent injection of such structures may lead to
sustained activation of those cells and therefore promotion of ADAs generation. However,
due to a lack of convincing immunological data supporting this hypothesis, further studies
are needed to investigate potentially higher activation of splenic macrophages by frequent
injections of aggregated proteins, and the effect of the proximity of draining lymph nodes.
106
Impact of treatment regimen on immunogenicity.
The high treatment dose used for Betaferon® (dose 250 μg/injection) compared to the dose
given for Avonex® and Rebif® (30 μg/injection or 22-44 μg/injection, respectively), has been
suggested as one of the reasons why the former is more immunogenic in patients. However,
data from clinical studies do not always confirm that a higher dose of protein correlates
with higher immunogenicity 4,23. In our study, we show that the dose is an important factor
affecting immunogenicity in our immune tolerant mice. The higher the dose, the more
ADAs were produced by both tg and non-tg animals. Moreover, our results and clinical data
suggest that the impact of dose on immunogenicity might be strongly protein dependent.
More detailed human studies are necessary to fully understand the influence of dose on
ADA production.
5. Conclussions.
This report describes the results of a study focused on the impact of treatment-related factors
on the immunogenicity of therapeutic rhIFNβ in immune tolerant mice, with the focus on the
rhIFNβ-1b product; Betaferon®. We studied the effect of protein dose, number of injections,
treatment frequency, and route of administration on ADA response, experiments which are
difficult to do in patients. We found that all these factors may have a significant impact on
the risk of ADA development. One has to keep in mind that direct translation of our results
from animals to patients is not possible and the results shown in this report should be, as far
as possible, confirmed in humans.
Acknowledgments.
We acknowledge Darren P. Baker, Ph.D. from BiogenIdec for kindly providing the IFNβ-1a
(Avonex drug substance) which we used in the ELISA to detect ADAs. We thank Abdul
Basmeleh for performing the ELISAs described in this article.
107
Chapter 6.
References.
1.
Paty D, Li DK. 1993. Interferon beta-1b is effective in relapsing-remitting multiple sclerosis.
II. MRI analysis results of a multicenter, randomized, double-blind, placebo-controlled
trial. UBC MS/MRI Study Group and the IFNB Multiple Sclerosis Study Group. Neurology
43:662-667
2.
Deisenhammer F, Reindl M, Harvey J, Gasse T, Dilitz E, Berger T. 1999. Bioavailability
of interferon beta 1b in MS patients with and without neutralizing antibodies. Neurology
52:1239-1243
3.
Bertolotto A, Gilli F, Sala A, Malucchi S, Milano E, Melis F, Marnetto F, Lindberg RL,
Bottero R, Di Sapio A, Giordana MT. 2003 Persistent neutralizing antibodies abolish the
interferon beta bioavailability in MS patients. Neurology 60:634-639
4.
Ross C, Clemmesen KM, Svenson M, Sørensen PS, Koch-Henriksen N, Skovgaard GL,
Bendtzen K. 2000. Immunogenicity of interferon-beta in multiple sclerosis patients:
influence of preparation, dosage, dose frequency, and route of administration. Danish
Multiple Sclerosis Study Group. Ann Neurol 48:706-712
5.
Bertolotto A, Deisenhammer F, Gallo P, Sölberg Sørensen P. 2004. Immunogenicity of
interferon beta: differences among products. J Neurol 251:II15-II24
6.
Hermeling S, Jiskoot W, Crommelin D, Bornaes C, Schellekens H. 2005. Development of
a transgenic mouse model immune tolerant for human interferon Beta. Pharm Res 22:847851
7.
Van Beers M, Jiskoot W, Schellekens H. 2010. On the role of aggregates in the
immunogenicity of recombinant human interferon beta in patients with multiple sclerosis. J
Interferon Cytokine Res 30:767-775
8.
Hartung H, Polman C, Bertolotto A, Deisenhammer F, Giovannoni G, Havrdova E,
Hemmer B, Hillert J, Kappos L, Kieseier B, Killestein J, Malcus C, Comabella M, Pachner
A, Schellekens H, Sellebjerg F, Selmaj K, Sorensen PS. 2007. Neutralising antibodies to
interferon beta in multiple sclerosis: expert panel report. J Neurol 254:827-837
9.
Runkel L, Meier W, Pepinsky RB, Karpusas M, Whitty A, Kimball K, Brickelmaier M,
Muldowney C, Jones W, Goelz SE. 1998. Structural and functional differences between
glycosylated and non-glycosylated forms of human interferon-β (IFN-β). Pharm Res
15:641-649.
10.
Sorensen P. 2008. Neutralizing antibodies against interferon-beta. Ther Adv Neurol Disord
1:62-68
11.
Bendtzen K. 2010. Critical Review: Assessment of interferon-β immunogenicity in multiple
sclerosis. J Interferon Cytokine Res 30:759-766
12.
1999. Evidence of interferon beta-1a dose response in relapsing-remitting MS: the OWIMS
Study. The Once Weekly Interferon for MS Study Group. Neurology 53:679-686
13.
Clanet M, Radue EW, Kappos L, Hartung HP, Hohlfeld R, Sandberg-Wollheim M,
Kooijmans-Coutinho MF, Tsao EC, Sandrock AW. 2002. European IFNbeta-1a (Avonex)
Dose-Comparison Study Investigators. A randomized, double-blind, dose-comparison
study of weekly interferon beta-1a in relapsing MS. Neurology 59:1507-1517
14.
Freedman M, Francis GS, Sanders EA, Rice GP, O’Connor P, Comi G, Duquette P, Metz L,
Murray TJ, Bouchard JP, Abramsky O, Pelletier J, O’Brien F. 2005. Once Weekly Interferon
beta-1alpha for Multiple Sclerosis Study Group; University of British Columbia MS/MRI
Research Group. Randomized study of once-weekly interferon beta-1la therapy in relapsing
multiple sclerosis: three-year data from the OWIMS study. Mult Scler 11:41-45
15.
Cohen B, Rivera VM. 2010. PRISMS: the story of a pivotal clinical trial series in multiple
sclerosis. Curr Med Res Opin 26:827-838
108
Impact of treatment regimen on immunogenicity.
16.
Van Beers M, Sauerborn M, Gilli F, Hermeling S, Brinks V, Schellekens H, Jiskoot W.
2010. Hybrid transgenic immune tolerant mouse model for assessing the breaking of B cell
tolerance by human interferon beta. J Immunol Methods 352:32–37
17.
Schellekens H. 2005. Factors influencing the immunogenicity of therapeutic proteins.
Nephrol Dial Transplant 20:vi3–vi9
18.
Singh S. 2011. Impact of product-related factors on immunogenicity of biotherapeutics. J
Pharm Sci 100:357-387
19.
Barbosa M, Celis E. 2007. Immunogenicity of protein therapeutics and the interplay between
tolerance and antibody responses. Drug Discov Today 12:674-681
20.
Perini P, Facchinetti A, Bulian P, Massaro AR, Pascalis DD, Bertolotto A, Biasi G, Gallo
P. 2001. Interferon-beta (INF-beta) antibodies in interferon-beta1a- and interferon-beta1btreated multiple sclerosis patients. Prevalence, kinetics, cross-reactivity, and factors
enhancing interferon-beta immunogenicity in vivo. Eur Cytokine Netw 12:56-61
21.
Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggregates seem to play a key role
among the parameters influencing the antigenicity of interferon alpha (IFN-α) in normal and
transgenic mice. Pharm Res 14:1472-1478
22.
Schellekens H, Jiskoot W. 2006. Erythropoietin-Associated PRCA: Still an unsolved
mystery. J Immunotoxicol 3:123-130
23.
Administration. FaD. Adalimumab. 2002. Pharmacology Reviews: Adalimumab product
approval information—licensing action.; 2002 December 31. Available from: http://www.
accessdata.fda.gov/drugsatfda_docs/nda/2008/125057s110TOC.cfm
24.
Peng A, Kosloski M, Nakamura G, Ding H, Balu-Iyer SV. 2011. PEGylation of a factor
VIII-phosphatidylinositol complex: pharmacokinetic and immunogenicity in hemophilia A
mice. AAPS J 14:35-42
25.
Van Regenmortel M, Boven K, Bader F. 2005. Immunogenicity of biopharmaceuticals: An
example from erythropoietin. BioPharm Int 18:36–52
109
110
CHAPTER 7.
Influence of aggregation and route of
injection on the biodistribution of mouse
serum albumin.
Grzegorz Kijanka
Malgorzata Prokopowicz
Huub Schellekens
Vera Brinks
Submitted for publication.
111
Chapter 7.
Abstract.
Protein aggregates are a major risk factor for immunogenicity. Until now most studies on
aggregate-driven immunogenicity have focused on linking physicochemical features of the
aggregates to the formation of antidrug antibodies. Lacking is however, basic knowledge on
the effect of aggregation on the biodistribution and clearance of therapeutic proteins in vivo.
The aim of current study was to get insight into the effect of aggregation on biodistribution in
mice using different routes of administration. Fluorescently labeled stressed and unstressed
mouse serum albumin was injected via different routes in mice and detected via in vivo
fluorescence imaging up to 48 hrs post-injection. We found that biodistribution of stressed
MSA significantly differed from its unstressed counterpart. Subcutaneous and intramuscular
administration resulted in accumulation of protein at the site of injection, from which
clearance of stressed MSA was considerably slower than clearance of unstressed MSA.
Upon intravenous and intraperitoneal injection of stressed MSA, fluorescent “hot spots”
were observed in the spleen, liver and lungs. Further and more detailed examination of
biodistribution after intraperitoneal injection showed higher fluorescence in most of tested
organs suggesting more efficient diffusion and/or lymphatic uptake from peritoneum of
unstressed MSA than the stressed formulation.
112
1. Introduction.
Biodistribution of MSA aggregates.
Therapeutic proteins have revolutionized the therapy of many diseases like multiple
sclerosis, rheumatoid arthritis, Crohn’s disease and many others. Unfortunately, therapeutic
proteins are immunogenic and cause the production of anti-drug antibodies (ADA) in some
of the patients. These ADAs can decrease the treatment efficiency and can lead to severe
side effects 1–3.
Among the many risk factors that could induce the production of ADA, protein aggregates
seems to be critical. An increasing number of reports link the presence of protein
aggregates in the formulated product with an increased risk of ADA formation 4–9. Various
physicochemical features including aggregate size, molecular weight , composition and
rigidity have been studied to determine which are critical in immunogenicity 6–8. However,
data on what happens to aggregates after their administration into patients is very limited.
Filipe et al. showed that incubation of human monoclonal IgG aggregates in plasma for 24
hrs resulted in alteration of the total number of aggregates, led to different aggregate size
and changed their structure 10. These results indicate that aggregates can undergo significant
modifications after coming in contact with biological fluids.
Many reports, both from clinical and animals studies, have shown that the route of injection
might have a significant impact on immunogenicity of therapeutic proteins 11–14. One of the
explanations of this phenomenon is distinct biodistribution of the drug after administration
via different routes 15,16. However, studies comparing biodistribution of (aggregated) proteins
after given via different routes are lacking. Since the physicochemical characteristics of
aggregates and monomers differ significantly, it seems likely that the biodistribution of
these species is different too. In fact, existing literature seems to suggest differences in
biodistribution of protein monomers and aggregates. For example, it has been shown that
uptake of proteins after subcutaneous (SC) injection is mainly via lymphatic transportation
which could carry macromolecules and particulates up to 100 nm in diameter 17. However,
as aggregates often exceed this size, one could imagine that clearance of aggregates from the
injection site upon SC administration will be slower than that of monomers. Decomposition
of protein aggregates might be necessary prior to their removal. One could also hypothesize
that after intravenous (IV) injection protein aggregates are cleared from circulation by the
reticuloendothelical system (RES) as it has been shown for liposomes 18. However, these
hypotheses need to be confirmed.
This report describes a series of experiments designed to study the biodistribution of
aggregated proteins after administration into mice. In order to obtain an autologous model
system mimicking the human situation we used mouse serum albumin (MSA) as a protein,
which was labeled with an infrared fluorescence probe to allow detection in vivo and exvivo. In a first experiment we administered unstressed or stressed (aggregated) MSA via four
different injection routes: intraperitoneal (IP), IV, SC and intramuscular (IM) and assessed
fluorescence over a period of 48 hours. In a follow up study we determined in more detail
113
Chapter 7.
the biodistribution of unstressed and stressed MSA over time after IP injection.
2. Materials and methods.
2.1 Animals.
FvB/N females (5-6 weeks old) were purchased from Charles River (The Netherlands).
Animals were housed in the standard perspex cages with free access to food and water
(acidified). To reduce food-induced fluorescence in the intestinal track, animals received
chlorophyll free diet (Harlan Laboratories, The Netherlands). In order to eliminate
autofluorescence by the fur, mice were completely shaven and treated with Veet cream (local
pharmacy) to remove remaining hair. All testing was performed according to Institutional
Ethical Committee Regulations of Utrecht University, the Netherlands.
2.1 Protein labeling and stressing.
MSA (Sigma Aldrich, the Netherlands) was reconstituted in phosphate buffered saline
(PBS) to a concentration of 5 mg/ml. It was labeled with Alexa Fluor 700 carboxylic acid
succinimidyl ester (Invitrogen, the Netherlands) according to the manufacturer protocol.
The molar ratio of dye to protein was 10:1. After labeling, the free dye was removed by
overnight dialysis against PBS. The degree of protein labeling was determined according
to the manufacturer’s protocol and was equal to 0.3 Alexa700 dye molecules per one MSA
molecule.
Stressed MSA-Alexa700 was acquired via metal catalyzed oxidation. First, MSA-Alexa700
conjugates were dialyzed overnight (O/N) against 50mM HEPES, 100mM KCl, 10mM
MgCl2, pH 7,4 (oxidation buffer). After dialysis MSA was diluted to a concentration of ~1,2
mg/ml and the oxidation procedure was initiated by adding FeCl3 and ascorbic acid to final
concentrations of 100 mM and 2,5 mM, respectively. The solution was kept at 37°C for
one hour, after which the oxidation reaction was terminated by the addition of EDTA to a
final concentration of 1 mM. Afterwards, stressed MSA-Alexa700 was again O/N dialyzed
against PBS. The protein concentration was determined by a MicroBCA® assay according to
the manufacturer’s protocol (Thermo Fisher Scientific, The Netherlands). MSA-Alexa700
solutions (stressed and unstressed) were diluted to a final concentration of 1mg/ml, aliquoted
and stored at 4˚C (short term storage, < week) or -20˚C (long term storage, > week).
2.2 Characterization of unstressed and stressed MSA.
2.2.1. Visual inspection.
Stressed samples were visually inspected to determine the presence of visible particles.
Unstressed MSA-Alexa700 formulation was used as a reference.
2.2.2. Size-Exclusion Chromatography (SEC).
SEC was performed on a Waters 2699 Aliance (Waters, USA) equipped with a WATERS
2487 Dual λ absorbance detector and a WATERS 2475 Multi λ fluorescence detector. Proteins
were separated on a Superdex 200 column (GE Health Care Life Sciences, Belgium). The
elution buffer was composed of 50 mM phosphate buffer, 200 mM NaCl and pH 7.4. Before
114
Biodistribution of MSA aggregates.
analysis both the buffer and the MSA-Alexa700 samples were filtered through a 0.45 µm
Nylon filter. A volume of 100 μl stressed or unstressed MSA-Alexa700 solution was applied
to the column and the separation was performed with a flow rate of 0.5 ml/min. UV detection
was performed at a wavelength of 280 nm (A280) and fluorescence was induced at 702 nm
and detected at 720 nm (F720).
In order to calculate the amount of free dye in the formulations the area under the curve
(AUC) was calculated for the F720 signal. The percentage of oligomers, monomers and
fragments of MSA present in both stressed and unstressed formulations were calculated
for the AUC measured at A280. To calculate the recovery of the stressed MSA-Alexa700
after filtration, the AUC of stressed MSA-Alexa700 was compared with the AUC of the
unstressed MSA-Alexa700, which was referred to as 100%.
2.2.3. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis (SDS-PAGE).
The SDS-PAGE analysis was performed with the X-Cell Sure Lock@ Mini-Cell system
(Invitrogen, the Netherlands). 0.5 μg or 1 μg of stressed and unstressed MSA-Alexa700 were
loaded on the NuPAGE® 4-12% Bis-Tris Mini Gels of 1mm thickness (Novex®, Invitrogen,
the Netherlands). Protein samples were separated under non-reducing conditions at 130 V
for 60 minutes. Proteins were visualized by silver staining and by detection of fluorescence
using the infrared imager Odyssey (LiCore, Germany).
2.2.4. Dynamic Light Scattering (DLS) and NanoParticle Tracking Analysis (NTA).
DLS measurements were performed on a Malvern ALV CGS-3 goniometer (Malvern
Instruments, Malvern, UK) coupled to an LSE-5003 autocorrelator and a He\Ne laser (25
mW, output power, λ=632.8 nm). Data were obtained at 25°C and the laser was operating
at an 90° angle. Data were captured by the ALV correlator software (Malvern, UK) and
analyzed using the DTS (Nano) software (Malvern, UK).
In addition, both stressed and unstressed MSA-Alexa700 were analyzed for the number
of particles using the NanoSize LM14 (NanoSight Ltd., UK) equipped with an EMCCD/
Andor camera and a 532 nm laser. Particle count and size distribution were analyzed by
three 60 s measurements.
2.3 Biodistribution of unstressed and stressed MSA upon injection via different routes.
A volume of 50 µl (50 µg) of unstressed or stressed MSA-Alexa700 was injected via one of
following routes: IP, IV, IM (right hind leg) or SC (neck), (n=5). Fluorescence was measured
by the BioSpace Photon ImagerTM (Biospace Lab, France) before injection, directly post
injection (p.i.), every 10 min within the first hour p.i. and after 3, 5, 8, 24, and 48 hrs
p.i. The fluorescence was excited at 696 nm and emission was measured at 720 nm for
10 s. Using the autofluorescence measured for all animals before the injection of MSA, a
fluorescence threshold value was determined at 10 counts per s. Also, regions of interest
(ROIs) were drawn around the injection site of animals treated SC and IM. The maximal
fluorescence measured at the neck (corresponding to SC injection site) and the right hind
115
Chapter 7.
leg (corresponding to IM injection site) of mice administered IP and IV, was used to set the
fluorescence threshold for these ROIs (≥70 counts per s).
After 48 hours all mice were euthanized and following organs and tissues were collected for
ex vivo analysis of fluorescence: bladder, spleen, kidneys, liver, lungs, heart, skin (from site
of injection of animals injected SC and IM), muscle (IM) and plasma. Organs were stored
at -80˚C and plasma at -20˚C prior to ex vivo analysis.
Ex vivo fluorescence of spleens, livers and lungs, was measured on an infra-red imager
(Odyssey, LiCore, Germany). The distribution of MSA-Alexa700 conjugates in all collected
organs and tissues was determined using an adopted method for the measurement of anti
tumor nanobody biodistribution described by Oliveira et al 19. Briefly, organs were first
weighed and homogenized in a radioimmunoprecipitation buffer (RIPA, 50 mM Tris, 150
mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate and 1% Triton X 100) supplemented with
protease inhibitors (Sigma, the Netherlands). Homogenized solutions were transferred onto
96-well plates (Greiner bio one, The Netherlands) and fluorescence was measured on the
infra-red imager. The concentration of MSA-Alexa700 in every organ was calculated from
a standard curve created by serial, two-fold dilutions of known concentrations of unstressed
or stressed MSA-Alexa700 conjugates in RIPA buffer.
2.4 Detailed biodistribution of unstressed and stressed MSA upon IP administration.
Mice (n=5 per group) were injected IP with either unstressed or stressed MSA-Alexa700.
Before and after the injection the syringes were weight to precisely determine the injected
dose of MSA-Alexa700 conjugates. In vivo fluorescence was assessed before injection,
directly p.i. and before euthanasia (15 min, 1, 3, 8 and 24 hrs p.i) using the BioSpace Photon
ImagerTM. Mice were euthanized by decapitation under isoflurane anesthesia. Afterwards,
following organs and tissues were collected for ex vivo analysis: blood, urine, spleen, liver,
kidney (left), stomach, small and large intestine, bladder, muscle and skin (from right hind
leg) lung, thymus, heart and brain.
Ex vivo fluorescence measurements and analysis of homogenized tissues and organs was
performed as described above.
2.4. Statistical analysis.
All data were first assessed for the normal distribution using the Shapiro-Wilk test. Because
data from the first experiment 2.3 was not normally distributed, a Mann-Whitney nonparametric test was used to assess differences between in vivo fluorescence of the skin and
muscle areas (ROIs) of animals injected IM and SC with either stressed or unstressed MSAAlexa700. The same test was also used to analyze the difference in ex vivo MSA-Alexa700
signal between the different organs collected from mice treated with either stressed or
unstressed MSA-Alexa700. To compare the difference in ex vivo MSA-Alexa700 signal
between the organs collected using different injection routes a Kruskal-Wallis test was
employed.
116
Biodistribution of MSA aggregates.
The data obtained during the experiment 2.4. was also not normally distributed, therefore also
here non-parametric tests were used for analysis. A Mann-Whitney test was used to assess
the difference in overall fluorescent signal (cumulated for all time points) between organs
from animals injected with either stressed or unstressed MSA-Alexa700. A KolmogorovSmirnov test was employed to check if the distribution of the fluorescence measured in
different organs ex vivo differed over time between stressed and unstressed MSA treatment.
3. Results.
3.1. Visual inspection.
Stressing the MSA-Alexa700 formulation via metal catalyzed oxidation resulted in the
formation of insoluble, rapidly sedimending aggregates that were easily detectible by eye
(Figure 1A).
3.2 Size exclusion chromatography.
B)
A)
C)
D)
E)
S
US
US
F)
S
US
Size [nm]
2908±98
Stressed
Unstressed 12553±17179
Channel
Stressed
Unstressed
A 280nm
F 720nm
A 280nm
F 720nm
DLS
S
UN
PDI
0.485±0.083
0.885±0.163
AUC
[uV*sec]
2052761
30753514
5475202
75679564
≥ Trimer
7.95
0.86
2.13
1.03
Size [nm]
257±51
131±27
SEC
SD
108±15
63±18
S
US
NanoSight
Average 108/ml
1220±751
66±4
Dimer
Monomer
Fragments
Free dye
8.80
3.84
10.7
7.42
44.00
51.13
76.55
82.59
39.25
44.17
10.63
8.96
8.82
2.36
S
Ratio
18,67
Recovery yield
[%]
37.5
40.6
100
100
Figure 1. A) Photography of stressed and unstressed MSA-Alexa700 formulations. B) Fluorescence
of unstressed (US) and stressed (S) MSA-Alexa700 formulations measured using an in vivo imager.
C) Photography of 0.45 μm nylon filters surface after filtration of MSA-Alexa700 formulation prior
to SEC analysis. D) Silver stained results of SDS-page analysis of the stressed and unstressed MSAAlexa700 conjugates under non-reducing conditions. Samples order: marker, 0.5 μg and 1 μg of
unstressed MSA-Alexa700, 0.5ug and 1ug of stressed MSA-Alexa700. E) Fluorescent analysis of
protein separated on the SDS-page gel. Samples order: 0.5 μg and 1 μg of unstressed MSA-Alexa700,
0.5 μg and 1 μg of stressed MSA-Alexa700. The stressed formulation contains aggregates too big to
enter the gel and stuck in the wells (yellow rectangles). F) Summarized results of DLS, NTA and SEC
analysis of both stressed and unstressed MSA formulations.
117
Chapter 7.
Before SEC analysis both stressed and unstressed MSA-Alexa700 solutions were filtered
through a 0.45 μm nylon filter. Inspection of the filter revealed that a significant amount
of MSA-Alexa700 aggregates was retained on the filter used for the stressed solution
(Figure 1C). This observation was confirmed by the total AUC of stressed MSA-Alexa700
which showed a protein loss of ~60% after filtration when compared with total AUC of
unstressed MSA-Alexa700 (Figure 1F). SEC analysis revealed a reduced percentage of
protein monomers in the stressed MSA-Alexa700 sample when compared to the unstressed
MSA-Alexa700 formulation (76% unstressed vs 44% stressed formulation). The percentage
of protein dimers was similar in both formulations (8,8% and 10,7%), but the percentage
of trimers and higher molecular weight species was slightly increased in the stressed MSAAlexa700 solution (7,9% vs 2,3%). Interestingly, SEC analysis revealed degradation of
MSA-Alexa700 during the stress procedure since the amount of protein fragments in the
stressed formulation was three fold higher compared to the unstressed formulation (30% vs
10%, respectively).
SEC analysis also allowed estimating the content of free dye in the stressed and unstressed
MSA-Alexa700 solutions. As shown in Figure 1F only 2.4% of the fluorescent signal came
from unbound dye in the unstressed formulation. In contrast, in the stressed MSA-Alexa700
the content of free dye seemed to be higher (8,8%), however after correcting for the protein
loss during filtration of this stressed solution the amount of free dye was estimated to be 3%.
3.3. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis (SDS-PAGE).
Silver staining revealed the presence of multiple multimeric entities (up to the estimated
weight of MSA tetramers) in the unstressed MSA-Alexa700 formulation. However for this
unstressed formulation, monomers were the most abundant species (Figure 1E). Also, no
aggregates or fragments were detected (Figure 1D). In contrast, the stressed formulation
contained significant amounts of both MSA fragments and aggregates (Figure 1D). In
fact, the stressed MSA-Alexa700 contained aggregates which were too big to enter the gel
(Figure 1E).
3.4. DLS and NTA.
Both DLS and NTA analyses confirmed that the stressed MSA-Alexa700 formulation
contained a higher amount of aggregates compared to the unstressed formulation (Figure
1F). For the stressed MSA-Alexa700 DLS analysis showed the presence of particles with
an average size of 2908±98 nm and a PDI of 0.485±0.083. For the unstressed formulation
DLS analysis gave inconclusive results as the measured size varied from about 400 nm
to 20.000 nm depending on the measurement, and the PDI values were very high (0.7 to
1). NTA measurements showed the presence of particles in both stressed and unstressed
formulations with an average size of 257±51 nm (stressed) and 131±27 nm (unstressed).
Moreover, NTA analysis showed an ~18 fold increase in particle count in the stressed MSAAlexa700 solution compared to the unstressed one.
3.4. Biodistribution of unstressed and stressed MSA upon injection via different routes.
118
Basal
0.5h
1h
3h
8h
24h
48h
7
Stressed
6
5
4
Unstressed
3
x10² Count/sec
IV
0h
Biodistribution of MSA aggregates.
2
1
7
Stressed
5
4
Unstressed
3
x10² Count/sec
IM
6
2
1
6
5
4
Unstressed
3
x10² Count/sec
IP
Stressed
7
2
1
7
Stressed
5
4
Unstressed
3
2
1
IM
SC
12
10
8
6
4
2
0
0
Cm2
Cm2
14
12
10
8
6
4
2
0
0
x10² Count/sec
SC
6
p=0.001
10
20
30
Time [h]
40
p<0.0001
10
20
30
Time [h]
40
Figure 2. A) In vivo distribution of fluorescence signal from MSA-Alexa700 conjugates upon
injection via IV, IM, IP and SC route in time. B) Change of size of ROIs drawn around the injection
site of animals injected IM or SC. The black line shows results obtained in animals injected with
unstressed MSA-Alexa700 and red from animals treated with stressed protein.
119
Chapter 7.
In vivo and ex vivo analysis of MSA-Alexa700 conjugates upon injection revealed significant
differences in biodistribution between the stressed and non-stressed formulations.
Intravenous injection
As showed in Figure 2 IV injection led to a rapid distribution of MSA-Alexa700 conjugates
throughout the body for both stressed and unstressed formulation. At 10 minutes p.i.
accumulation of the fluorescent signal was observed in the area of the liver and 20 min
p.i. high fluorescence was seen in the area of the bladder. At 24 hrs p.i., mice injected with
stressed MSAAlexa700 displayed fluorescence only in the liver and bladder areas. In contrast, animals
injected with unstressed MSA-Alexa700 displayed fluorescence in other parts of their body
as well at this time point. Both in vivo and ex vivo analyses showed that IV injection resulted
in similar accumulation of fluorescent signal in the livers of mice receiving either stressed
or unstressed MSA-Alexa700. Moreover, IV administration resulted in the highest amount
of fluorescence in the liver area of all injection routes studied in this experiment (p<0.001
for IV vs IP/IM/SC). The accumulation of MSA-Alexa700 48 hrs p.i. in other organs than
the liver was negligibly low.
An interesting observation made during ex vivo measurements was the presence of “hotspots”
with very high fluorescence in the lungs and spleens of mice treated with stressed MSAAlexa700 (Figure 3). Organs collected from animals receiving the unstressed formulation
did not show these “hotspots”.
Intramuscular injection
After IM injection, fluorescence could be measured at the site of injection until the end
of the experiment (Figure 2). Moreover, the intensity of fluorescence measured in vivo at
the site of injection 48 hrs p.i. seemed to be similar for mice treated with either stressed
or unstressed MSA-Alexa700. However, ex vivo analysis revealed that mice treated with
stressed MSA-Alexa700 displayed higher fluorescence signal in the muscle compared
to mice treated with the unstressed formulation (p=0.016). Also clear differences in the
distribution kinetics at the site of injection were observed. Unstressed MSA-Alexa700
seemed to diffuse or be removed from the injection spot much faster than the stressed
formulation. The spread of fluorescence signal (increase of ROIs) around the injection spot
in animals receiving unstressed MSA-Alexa700 seemed to be faster and the surface area
of ROIs was significantly higher compared to the mice receiving the stressed formulation
(p=0.001). At 3 hrs p.i. fluorescence was found throughout the body, for both stressed and
unstressed formulations. This body-wide fluorescence signal was still measurable 8 hrs p.i.
Ex vivo analysis revealed low accumulation of fluorescent signal of stressed and unstressed
MSA-Alexa700 in the liver 48 hrs p.i..
Intraperitoneal injection
120
Biodistribution of MSA aggregates.
Directly after IP injection of either stressed or unstressed MSA-Alexa700 conjugates a
fluorescent signal was detectible in the whole peritoneum (Figure 2). No significant changes
in the fluorescent signal (i.e. intensity or distribution) were observed within the first hour
p.i.. At 3 hours p.i. fluorescence was detected throughout the body for both stressed and
unstressed MSA-Alexa700 treatments, with a particular strong signal in the area of the
liver and in the bladder. Up to 8 hrs p.i. fluorescence was detected throughout the body. 24
hours p.i. of stressed and unstressed MSA-Alexa700 a weak fluorescent signal could still
be detected at the area of the peritoneum and the liver. However, this signal seemed more
pronounced in animals receiving the unstressed formulation. At the end of the experiment
(48 hrs p.i.) residual fluorescence could be measured only in the liver area of mice receiving
unstressed MSA-Alexa700. In contrast, mice injected with stressed MSA-Alexa700
displayed residual fluorescence in area of the liver and peritoneum.
Ex vivo analysis showed highest fluorescence signal in the isolated livers (Figures 3), with
similar intensity in mice treated with stressed and unstressed formulations. However, in
livers and spleens isolated from the mice injected with stressed MSA-Alexa700 “hotspots”
of high fluorescence were detected. In contrast, in mice treated with unstressed MSAAlexa700 no “hotspots” were observed in the liver and the fluorescent signal was equally
µg MSA/g tissue
100
80
60
40
20
100
60
40
20
0
-20
-20
80
60
40
20
120
IP
100
*
IM
80
0
120
µg MSA/g tissue
120
IV
µg MSA/g tissue
µg MSA/g tissue
120
100
80
*
SC
*
60
40
20
0
0
-20
-20
Figure 3. Distribution of fluorescence signal in different organs 48 p.i. via IV, IM, IP and SC routes.
The bars show average accumulation of MSA-Alexa700 conjugates ± SD. The black bars show results
obtained from animals injected with stressed protein and white bars indicate unstressed protein.
121
Chapter 7.
IV
Spleen
Liver
IP
Spleen
Unstressed
Stressed
Lung
Figure 4. A) IV injection of stressed MSA-Alexa700 resulted in the formation of fluorescent “hot
spots” in the lungs and spleens. B) In livers and spleens of animals IP injected with stressed protein
clumps of highly fluorescent signal were observed.
distributed throughout the liver (Figure 4). However the total fluorescence signal in spleens
isolated from mice treated with stressed MSA-Alexa700 was higher than the signal from
mice treated with the unstressed formulation (p=0.001).
Subcutaneous Injection
SC injections gave a similar distribution of fluorescent signal to IM injections. After SC
injection, the spread of fluorescent signal from the injection site seemed to be faster in mice
treated with unstressed than with stressed MSA-Alexa700. However, in contrast to the IM
route of injection, a clear difference in the intensity of fluorescence at the site of injection
between stressed and unstressed formulations was still visible 48hrs p.i.. Ex vivo analysis
confirmed a significantly higher amount of MSA-Alexa700 at the site of injection (skin) of
animals receiving the stressed formulation (p=0.045).
3.5. Detailed biodistribution of unstressed and stressed MSA upon IP administration.
To determine the biodistribution over time after IP administration animals were treated with
stressed or unstressed MSA-Alexa700 conjugates and euthanized at designated time points
p.i.. Selected organs and tissues were collected for the ex vivo analysis. The in vivo and
ex vivo measurement showed that some animals were injected incorrectly as shortly after
the injection the fluorescent signal was not equally distributed through the peritoneum and
post mortem analysis revealed injection solution in the intestines or stomach of these mice.
These mice were therefore excluded from the analysis.
122
Biodistribution of MSA aggregates.
In vivo fluorescence measured up to 24 hrs p.i. was found similar to the previous experiment
(data not shown). Description of results is therefore limited to ex vivo analysis of
fluorescence. As shown in Figure 5 the ex vivo analysis revealed similar fluorescent signal
(i.e. intensity over time) in the plasma, urine, kidneys, thymus, small intestine and skin
isolated at different time points for mice injected with either stressed or unstressed MSAAlexa700. However for the spleen, bladder, liver, heart, lungs, large intestine, brain, muscle
and stomach, significant differences in distribution of fluorescent signal over time were
found. Similar to the first experiment, the highest fluorescent signal was found in the liver
at 3 hrs p.i., and this signal was almost 6 times higher in mice treated with the unstressed
formulation compared to animals injected with stressed MSA-Alexa700 (1091.4±154.6%
ID/g tissue and 209.7±55.8 ID/g tissue, respectively). Accordingly, the distribution over
time (p=0.001) and magnitude of fluorescent signal (p=0.004) differed in the livers of mice
injected with unstressed and stressed MSA-Alexa700.
Although no significant difference was found between the overall fluorescence signals
of spleens of mice treated with stressed or unstressed MSA-Alexa700, the distribution of
fluorescent signal over time differed between these groups (p=0.041). Within the first 8
hrs p.i. a higher fluorescent signal was present in the spleens of mice receiving unstressed
MSA-Alexa700 compared to mice that received the stressed formulation. At 3 hrs p.i., the
fluoresce measured in the spleens of mice injected with unstressed MSA-Alexa700 was
3 times higher than fluorescence of spleens from animals treated with stressed MSAAlexa700. However, at 24 hrs p.i. the fluorescence measured in the spleens of mice
injected with stressed MSA-Alexa700 was found to be 3 times higher than in spleens of
mice injected with unstressed MSA-Alexa700. Moreover, in the mice treated with stressed
MSA-Alexa700 again “hotspots” with high fluorescence signal were present (like observed
in the first experiment).
While in most of organs the fluorescence was increasing until 3 hrs p.i., the signal in the
intestines of mice injected with stressed MSA-Alexa700, already reached a maximum 15
min p.i.. Interestingly the fluorescence in the large intestines of mice treated with both
stressed and unstressed MSA-Alexa700 was much higher than in the small intestines.
4. Discussion.
Presence of protein aggregates in the formulation of therapeutic proteins has been identified
as a major risk factor for immunogenicity. However, there is almost no data showing
the fate of these aggregates after injection. It has been hypothesized that aggregates can
persist at the site of injection, especially when applied SC or IM, much longer than protein
monomers. Therefore aggregates may trigger the activation of immune cells located at the
site of injection more efficient than monomers.
Current experiments were designed to study the biodistribution of unstressed, mainly
monomeric and stressed, mostly aggregated MSA in mice when given via different routes.We
first showed that metal catalyzed oxidation is an effective way to obtain a MSA formulation
123
Chapter 7.
Spleen
300
Ctrl 15min 1hr
3hrs
8hrs
24hrs
Stressed
250
200
Unstressed
150
100
MW p=0.441
KS p=0.315
0
5
10
15
Urine
5000
0
5
10
Urine bladder
15
1400
1200
1000
MW p=0.022 800
KS p=0.032 600
400
200
0
20
0
15
140
120
100
80
60
40
20
0
0
2000
1500
1000
500
0
0
5
30
10
Heart
15
25
20
15
MW p=0.096
KS p=0.048
10
5
0
0
140
120
100
80
60
40
20
00
5
10
20
Intrestine small
MW p=0.109
KS p=0.153
5
10
250
15
Skin
20
200
150
100
MW p=0.182
KS p=0.0159
50
0
0
5
10
15
20
25
400
350
300
250
200
150
100
50
0
0
800
700
600
500
400
300
200
100
0
0
20
Liver
5
10
Lungs
25
10
15
3.5
3
2.5
2
1.5
1
0.5
0
0
Intrestine large
20
25
MW p=0.006
KS p=0.032
5
10
0
0
15
MW p=0.004
KS p=0.003
5
1000
180
160
140
120
MW p=0.004 100
KS p<0.0001 80
60
40
20
0
20
0
25
15
Thymus
20
25
50
45
40
35
30
25
20
15
10
5
0
20
25
MW p=0.302
KS p=0.287
2000
2500
% ID/g tissue
Plasma
4000
MW p=0.151
KS p=0.041 3000
50
0
14
12
10
8
6
4
2
0
5
10
Kidney
15
20
25
MW p=0.205
KS p=0.153
5
10
Brain
15
20
25
MW p<0.0001
KS p<0.0001
5
10
15
Muscle
20
25
MW p=0.037
KS p=0.029
0
5
250
10
15
Stomach
20
25
200
150
MW p=0.102 100
KS p=0.241
50
5
10
15
20
25
0
0
MW p=0.01
KS p=0.25
5
10
15
20
25
T [h]
Figure 5. The biodistribution of the MSA-Alexa700 over time in animals IP injected with
stressed (red line) and unstressed (black line) formulations. The accumulation of fluorescent
signal is expressed as percent of injected dose per gram of tissue (% ID/g tissue) or percent
of injected dose per ml of plasma or urine (% ID/ml).
containing high amounts of aggregates. In addition we showed that stressed MSA-Alexa700
persists longer at the site of injection compared to unstressed MSA-Alexa700 after SC
and IM administration. The time frame in which the experiment was performed (up to
48 hrs), did not allow complete clearance of fluorescence at the injection site, however
it could be hypothesized that longer experiment would further magnify these differences
between stressed and unstressed MSA Alexa700 treatment. Nonetheless, our observation
that stressed MSA-Alexa700 persist longer at the site of injection is in agreement with the
hypothesis that aggregates are cleared slower than monomers when given SC. And might
explain why for some drugs SC appears the most immunogenic route.
124
Biodistribution of MSA aggregates.
High immunogenicity of proteins when injected IV has been described for rhIFNβ 11,
PEGylated Factor VIII 12 and antiTNFα antibody 20. In our previous manuscript we suggested
that aggregates, when injected directly into the blood stream, might be rapidly removed by
the macrophages of the reticuloendothelial system (RES) in the spleen, in a similar manner
as for liposomes exceeding 100 nm in diameter 18. In current experiment we induced the
formation of aggregates exceeding 100 nm, and small highly fluorescent “hotspots” in the
spleens were present. These spots might be aggregates caught by macrophages or might
be aggregates that are stuck in the blood vessels. Considering these spots in the spleen and
potential importance of macrophage uptake it would be worth performing an in depth study
on what causes these “hotspots”, not only for IV administration, but for IP treatment as well.
Another very interesting finding was that “hotspots” were also found in the lungs of mice
treated IV with stressed MSA-Alexa700. Since the lungs are composed of very small blood
capillaries it appears likely that aggregates might have been blocked in them. It is possible
that macrophages present in the lung tissue take up these particles and therefore increase
immunogenicity.
The biodistribution of MSA-Alexa700 upon IP administration differs depending on the
amount of aggregates in the formulation. In general, the accumulation of fluorescently
labeled and stressed MSA seems lower in the majority of organs compared to accumulation
of unstressed MSA. This could be due to a lower diffusion rate of the stressed protein through
the peritoneum as well as due to slower uptake of aggregates by the lymphatic transportation
system. A very interesting observation was the longer persistence of fluorescent signal and
its inhomogeneous distribution (hotspots) in the spleens of animals injected with stressed
MSA-Alexa700. The mechanism by which these hotspots are formed is not clear. Their size
and number suggests that the origin of “hotpots” in spleens from mice injected IP differs
from those found in mice administered IV. IV injection with MSA-Alexa700 induced a high
number of small “hotspots”, while IP administration resulted in a low number of “hotspots”
with a relatively large surface. Further studies on what composes these hotspots and how
they are formed are therefore needed.
Although the data presented in this manuscript seems to explain some observations described
in the literature on how the route of administration affects immunogenicity of protein drugs,
one should be aware of the limitations of the employed assays and mouse model. Because
we aimed at obtaining an autologous system we have chosen MSA as protein model.
However, as MSA receptors are located in liver and the liver being the main organ of MSA
metabolism 21,22, the elevated fluorescence signal in this organ might not be representative
for other proteins. Moreover, choosing a self-protein as a model did not allowed for direct
detection of injected protein by ELISA for example, because of high background signal
from endogenous MSA. Therefore the detection of injected MSA could only be done with
the use of a label, in this case Alexa-700. A possible issue that could arise is that the dye
separates from the protein, resulting in the detection of free dye and not of the protein itself.
125
Chapter 7.
We showed that our injected MSA formulations contained less than 5% of free dye, which
should be low enough to allow detection of the conjugated MSA-Alexa700. However, our
data also suggests that MSA-Alexa700 might be degraded quite rapidly in vivo (within
hours). Because the MSA-Alexa700 conjugates are too big to be filtered out by kidneys, the
fluorescence found in the urine most likely originates from MSA-Alexa700 degradation.
The fluorescent signal in the urine was high (reaching in the maximum almost 3000% of
ID/ml) and therefore supports degradation of MSA-Alexa conjugates and suggesting that
degradation product might differ in spectral properties from conjugates.
Conclusions.
In this report we show that in vivo florescence imaging, despite some drawbacks, is a valuable
method to study the biodistribution of protein (aggregates) upon injection. We showed (i)
that biodistribution of MSA differs between the stressed and unstressed formulation and
(ii) that the biodistribution of MSA strongly depends on the application route. Also IV and
IP injection of stressed MSA-Alexa700 resulted in the formation of fluorescent “hot spots”
in spleens (IP and IV), livers (IP) and lungs (IV) suggesting entrapping of MSA-Alexa700
aggregates in the microenvironment that might enhance induction of antibody responses.
Acknowledgments.
We thank dr Anton Martens, Willy Noort and Miranda van Amersfoort for the access to the
Biospace Photon ImagerTM and M3VisionTM software. We also thank dr Paul van Bergen en
Henegouwen for the access to the Odyssey infrared imager.
126
References.
1.
Biodistribution of MSA aggregates.
Bertolotto A, Malucchi S, Milano E, Castello A, Capobianco M, Mutani R. 2000.
Interferon beta neutralizing antibodies in multiple sclerosis: neutralizing activity and crossreactivity with three different preparations. Immunopharmacology 48:95–100
2.
Schellekens H, Casadevall N. 2004. Immunogenicity of recombinant human proteins:
causes and consequences. J Neurol 251:II4–9
3.
Singh SK. 2011. Impact of product‐related factors on immunogenicity of biotherapeutics. J
Pharm Sci 100:4–16
4.
Moore WV, Leppert P. 1980. Role of aggregated human growth hormone (hGH) in
development of antibodies to hGH. J Clin Endocrinol Metab 51:691–697
5.
Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggregates seem to play a key role
among the parameters influencing the antigenicity of interferon alpha (IFN-alpha) in normal
and transgenic mice. Pharm Res 14:1472–1478
6.
Hermeling S, Schellekens H, Maas C, Gebbink MF, Crommelin DJ, Jiskoot W. 2006.
Antibody response to aggregated human interferon alpha2b in wild-type and transgenic
immune tolerant mice depends on type and level of aggregation. J Pharm Sci 95:1084–1096
7.
Van Beers MMC, Jiskoot W,Schellekens H, 2010. On the role of aggregates in the
immunogenicity of recombinant human interferon beta in patients with multiple sclerosis. J
Interferon Cytokine Res 30:767–775
8.
Filipe V, Hawe A, Schellekens H, Jiskoot W. 2010. Aggregation and immunogenicity of
therapeutic proteins. In Wang W, Roberts CJ editors. Aggregation of therapeutic proteins,
Hoboken NJ: John Wiley & Sons, Inc. p 403-433
9.
Rosenberg AS. 2006. Effects of protein aggregates: an immunologic perspective. AAPS J
8:E501–507
10.
Filipe V, Poole R, Oladunjoye O, Braeckmans K, Jiskoot W. 2012. Detection and
characterization of subvisible aggregates of monoclonal IgG in serum. Pharm Res 29:2202–
2212
11.
Kijanka G, Jiskoot W, Schellekens H, Brinks V. 2013. Effect of treatment regimen on the
immunogenicity of human interferon beta in immune tolerant mice. Pharm Res 30:15531560
12.
Peng A, Kosloski MP, Nakamura G, Ding H, Balu-Iyer SV. 2012. PEGylation of a factor
VIII-phosphatidylinositol complex: pharmacokinetics and immunogenicity in hemophilia A
mice. AAPS J 14:35–42
13.
Ross C, Clemmesen KM, Svenson M, Sørensen PS, Koch-Henriksen N, Skovgaard GL,
Bendtzen K. 2000. Immunogenicity of interferon-beta in multiple sclerosis patients:
influence of preparation , dosage , dose frequency , and route of administration. Danish
Multiple Sclerosis Study Group. Ann Neurol 48:706–712
14.
Schellekens H. 2005. Factors influencing the immunogenicity of therapeutic proteins.
Nephrol Dial Transplant. 20 Suppl 6:vi3–9
15.
Srinivas NR, Shyu WC, Weiner RS, Warner G, Comereski C, Tay LK, Greene DS,
Barbhaiya RH. 1997. Assessment of dose proportionality, absolute bioavailability, and
immunogenicity response of CTLA4Ig (BMS-188667), a novel immunosuppressive agent,
following subcutaneous and intravenous administration to rats. Pharm Res 14:911–916
16.
Laursen T, Møller J, Jørgensen JO, Orskov H, Christiansen JS. 1996. Bioavailability and
bioactivity of intravenous vs subcutaneous infusion of growth hormone in GH-deficient
patients. Clin endocrinol 45:333–339
17.
Porter CJ, Edwards GA, Charman SA. 2001. Lymphatic transport of proteins after s.c.
127
Chapter 7.
injection: implications of animal model selection. Adv Drug Deliv Rev 50:157–171
18.
Drummond DC, Meyer O, Hong K, Kirpotin DB, Papahadjopoulos D. 1999. Optimizing
liposomes for delivery of chemotherapeutic agents to solid tumors. Pharmacol Rev 51:
691–743
19.
Oliveira S, van Dongen GA, Stigter-van Walsum M, Roovers RC, Stam JC, Mali W, van
Diest PJ, van Bergen En Henegouwen PM. 2011. Rapid visualization of human tumor
xenografts through optical imaging with a near-infrared fluorescent anti-epidermal growth
factor receptor nanobody. Mol Imaging 11:33-46
20.
Administration. FaD. Adalimumab. 2002. Pharmacology reviews: Adalimumab product
approval information—licensing action.; 2002 December 31.
Available
from:
nda/2008/125057s110TOC.cfm
http://www.accessdata.fda.gov/drugsatfda_docs/
21.
Ockner RK, Weisiger RA, Gollan JL. 1983. Hepatic uptake of albumin-bound substances:
albumin receptor concept. Am J Physiol 245:G13–18
22.
Brunskill NJ. 2004. Albumin receptors – structure and function. In Duttaroy AK, Spener F.
editors. Cellular proteins and their fatty acids in health and disease. Weinheim, FRG: WileyVCH Verlag GmbH & Co. KGaA, p 79-94
128
CHAPTER 8.
Summary.
129
Chapter 8.
1. Summary.
Since the 1990s, when recombinant human (rh) proteins were introduced onto the drug
market, the availability and safety of protein based therapies have improved significantly
compared to classical therapies using animal derived proteins. Unfortunately, the
efficiency of current rh proteins is often altered by anti drug antibodies (ADA), which may
significantly increase therapy costs and negatively affect patient’s well being (Chapter 2).
Many factors have been correlated with an increased risk of ADA production 1,2, however
an increasing number of studies highlights the significance of protein aggregates presence
in the formulated drug 3–6. Nevertheless, full understanding of the mechanism by which
aggregate-induced immunogenicity takes place is lacking. Classical therapeutic proteins
isolated from the animal tissues, were foreign to the patient immune system, and triggered
ADA production via a classical T cell dependent (TD) mechanism 7.
Most of the therapeutic proteins currently available on the market are human homologues
produced using recombinant DNA technologies. Such proteins have identical or very
similar primary sequences to their endogenous counterparts and as such, they shouldn’t be
recognized by the patient immune system as foreign. However, despite the primary sequence
similarities, rh proteins still trigger ADA formation. As mentioned before, aggregates are a
risk factor for immunogenicity. It has been suggested that they present repetitive epitopes
that induce clustering of B cell receptors and promote activation of B cells via a T cell
independent (TI) type 2 mechanism.
The studies described in this thesis aim to (in part) elucidate the mechanisms underlying
immunogenicity of rh therapeutic proteins. In order to model immunogenicity in experimental
animals, I have treated rh interferon beta immune tolerant (rhIFNβ transgenic, tg) and nontransgenic (non-tg) control mice with (commercially available) rh interferon beta products
Betaseron® and Avonex®. RhIFNβ-1b (Betaseron ®) is a highly immunogenic product
that contains about 60% aggregates, wherease rhIFNβ-1a (Avonex ®) is only minimally
immunogenic and contains about 5% aggregates.
The experiments in Chapter 3 are a continuation of recent findings by Sauerborn et al 8
and van Beers et al 4. They found that CD4+ T cells are crucial for ADA formation against
rhIFNβ-1b (Betaferon®) and that no apparent immunological memory was formed against
rhIFNβ-1b or rhIFNβ-1a (Avonex®). We first performed a study to determine when CD4+ T
cells are recruited in the immune processes leading to ADA formation. CD4+ T cells were
removed at different time points after the start of rhIFNβ-1b treatment. We found that ADA
formation could be inhibited only if CD4+ T cells were removed before or shortly after the
start of rhIFNβ-1b treatment, suggesting that CD4+ T cells are involved in the onset of ADA
responses or are attracted by other immune cells at very early stage of ADA production. In
the second experiment we assessed the presence of germinal centers (GCs) in the spleen
during rhIFNβ-1a treatment. These GCs are characteristic for a classical T- cell dependent
immune response and usually coincide with the formation of immunological memory.
130
Summary.
Results show similar numbers of GCs in tg and non-tg animals during rhIFNβ-1a treatment,
however, in tg mice the formation of GCs was not correlated with ADA levels. Lastly, we
tested if increasing doses of rhIFNβ-1b might lead to a break of immune tolerance and
subsequent immunological memory formation. Even at the highest dose used we did not
observe a clear immunological memory response.
Contaminations of formulated protein drug or additives used to increase protein stability
have been suggested to have adjuvant activities and therefore could contribute to the
development of an ADA response 9,10. To test this hypothesis a series of experiments has
been performed, in which the potency of commonly used vaccine adjuvants to increase ADA
titers was determined (Chapter 4). Among all tested compounds only cholera toxin subunit
B (CTB) was found to efficiently enhance ADA production in tg mice. In contrast, in nontg controls, saponin, Montanide ISA 50 V2, aluminum hydroxide and CTB were found to
significantly increase ADA titers. Because CTB increased ADA titers and potentially broke
tolerance in the tg mice a follow-up study was performed to test if conjugation of rhIFNβ1a with CTB could lead to the induction of immunological memory. Surprisingly, even
though CTB/rhIFNβ-1a conjugates triggered a very efficient primary immune response, rechallenge with rhIFNβ-1a did not increase ADA titers, indicating a lack of immunological
memory formation. In addition, the potency of several immune suppressants in reducing
ADA formation in mice receiving rhIFNβ-1b was tested. In tg mice only rapamycin was
found to significantly block ADA formation. In non-tg controls rapamycin and cyclosporine
were found to be effective.
Presence of protein aggregates in the formulated product is a well-known risk factor for
immunogenicity 6,11. However, other factors may also contribute to the development of ADAs.
Greinacher et al suggested that immunogenicity observed in patients receiving heparin
infusions might be correlated with bacterial infections preceding the heparin treatment.
They suggested that bacteria might form complexes with platelet factor 4 (PF4) and induce
antibody responses, which might cross-react with heparin/PF4 complexes. The experiments
described in Chapter 5 were performed to test the hypothesis that rhIFNβ might also interact
with bacteria. As the multiple sclerosis (MS) patients suffer from bacterial infections more
often than healthy individuals 12–14, formation of rhIFNβ/bacteria complexes could be a risk
factor potentially leading to enhanced immunogenicity of rhIFNβ. We have tested several
species of bacteria for binding to monomeric and stressed (aggregated) rhIFNβ. We found
that rhIFNβ binds well to all tested bacteria. However, the capacity of binding might strongly
depends on the rhIFNβ formulation. Interestingly, even though rhIFNβ binding to bacteria
seem to be electrostatically driven, rhIFNβ binds preferably to gram positive bacteria. Our
data show that bacterial infection might increase the risk of ADA in MS patients treated with
rhIFNβ, however this hypothesis has to be confirmed in an animal or patient study.
The treatment regimen is commonly believed to influence the risk of ADA formation. Lacking
are however detailed studies showing how differences in dose, route of administration and
131
Chapter 8.
frequency of administration affect immunogenicity of therapeutic proteins. For example, it
is commonly accepted that intravenous (IV) route is least immunogenic, followed by the
intramuscular (IM) and subcutaneous (SC) route. However, this assumption is based mainly
on results from vaccine studies or from studies in which immunogenicity of different rh
protein products, distinct routes of administration, schedule and dose are all compared.
Chapter 6 describes the results of a study designed to investigate the impact of the frequency
of administration, dose and route of administration of rhIFNβ on the production of ADA.
Surprisingly, in contrast to common belief, we found the IV route of administration to be
the most immunogenic in our tg mice. Moreover, we found that ADA production increases
with increasing number of rhIFNβ injections, increasing administration frequency and dose.
Among all risk factors affecting immunogenicity the presence of protein aggregates seems
to be of the highest risk 3,6,15–17. Numerous studies have been performed to characterize the
physicochemical properties of immunogenic aggregates 3,11,18. However, no studies have
been performed that focus on what happens to the aggregates upon their administration.
Chapter 7 summarizes the results of experiments in which we have tested the fate of
protein aggregates in vivo. We took advantage of in vivo and ex vivo fluorescence imaging
techniques, which allow tracing labeled proteins in mice. We found that upon IM and SC
injection protein aggregates are retained at the site of injection longer than non-aggregated
ones. Moreover, we show that injection of aggregated protein IV results in the formation of
highly fluorescent spots in lungs and spleens, which are probably big aggregates trapped in
blood capillaries or protein taken up by splenic or lung macrophages. Similar fluorescent
structures were also observed in livers and spleens of animals administered with aggregated
protein IP.
Taken together all data described in this thesis emphasize that the immune process underlying
immunogenicity of rhIFNβ is unique and not yet described before. It seems to share some
features with a CD4+ T cell dependent (TD) response, e.g. dependency of ADA production
on CD4+ T cells, time of CD4+ T cells recruitment into response and formation of germinal
centers. However, the results described here also show clear differences from a TD immune
response. Immunogenicity seems to require “danger signals” for triggering ADA production
that are different from the danger signals associated with a TD response, as TD adjuvants
commonly used to induce immune responses failed to trigger ADA formation in the tg mice.
Moreover, the lack of apparent memory formation, even after co-stimulation with very potent
adjuvant (CTB), suggests: i) the existence of a suppressing mechanism blocking the late
events of B cell maturation and differentiation, or ii) activation of subsets of CD4+ T cells
and/or B cells that are not capable of differentiation into memory cells, or iii) depletion of
formed memory cells. In addition to these differences the immunogenicity of rh therapeutic
proteins seems to be less dependent on the interleukin-2 (IL-2). Moreover, obtained results
clearly show that dogmas of classical immunology dogmas and conclusions from studies
on immune response against foreign proteins might not be valid for the immunogenicity of
rh proteins.
132
Summary.
2. Future perspectives.
2.1. Immunological mechanism underlying immunogenicity.
The immunological mechanism underlying immunogenicity of therapeutic proteins
has been studied in our group for several years. In order to obtain a system that closely
resembles patients treated with rh proteins we used immune tolerant transgenic mice and
treated them with similar rh proteins as being used in patients. Early hypotheses suggested
that protein aggregates resemble bacterial polysaccharide antigens by presenting repetitive
epitopes. However, both animal studies and clinical reports revealed that the chain of
reactions leading to the formation of ADA is much more complex. In fact, immunogenicity
was shown to display features of both a T cell dependent and T cell independent pathway.
Sauerborn et al showed that marginal zone B cells, which are known to be key players
in the Tind response, are crucial in triggering ADA production 8. Simultaneously, she
showed that ADA development is strongly impaired when the immune system is devoid of
functional CD4+ T cells 8. Moreover, as shown by van Beers et al, involvement of CD4+ T
cells apparently does not lead to the development of a classical TD characteristic, namely an
immunological memory response 4. Studies described in Chapters 3-5 of this thesis provide
further information on the mechanism underlying immunogenicity. However, we are still
lacking a full understanding of the immune processes that precede the formation of ADA.
2.1.1. CD4+ T regulatory cells.
One of the most intriguing questions regarding the mechanism underlying immunogenicity
is why CD4+ T cells are being activated to respond against rh proteins. As sequences of rh
proteins are (almost) identical as of endogenous proteins, CD4+ T cells should not recognize
their epitopes, or any response towards these epitopes should have been blocked by a
regulatory mechanism. The most important class of regulatory cells are CD4+ T regulatory
cells (Tregs) They prevent unwanted immune response and tranquillize the immune response
to prevent tissue damage 19. Therefore the question remains if and how protein aggregates
by-pass this regulatory mechanism. One could hypothesize that protein aggregates may
present a strong “danger signal” that could overcome the suppressory activity of Tregs.
This scenario, however, does not explain why immunological memory seems to be lacking.
Another possibility is that Tregs are pushed by aggregates to become “effector-like” cells.
Future studies focusing on the role of Tregs in immunogenicity will be of great importance
to determine their potential role in the formation of ADA. It would be of great value to
block or remove Tregs in tg mice by specific antibodies and determine how this influences
the ADA production.
2.1.2. Other CD4+ T cells subsets.
Although Tregs seem to be a promising class of CD4+ T cells that could “guide”
immunogenicity, they are not the only candidates. As shown by the studies on the effect of
immune suppressants on immunogenicity of rhIFNβ described in chapter 4, the balance of
CD4+ T cells subsets involved in ADA responses most likely differs between tg and nontg mice. In 1980s Mosmann et al noticed that murine CD4+ T cells are not a homogenous
133
Chapter 8.
population and introduced the Th1/Th2 dichotomy of CD4+ T cells 20. Current classification
includes additionally Th17, Th9, Th22 and follicular T cells. We depleted the entire CD4+
T cell pool (chapter 3), but these different subsets might have distinct involvement in
immunogenicity. More detailed studies on the potential involvement of different CD4+ T cell
subsets might be necessary to fully understand the mechanism underlying immunogenicity of
therapeutic proteins. This can be accomplished by several potential approaches: i) depletion
of separate subsets, ii) determination which of those CD4+ T cells are being activated during
ADA production, iii) isolation of antigen specific CD4+ T cells and characterization of their
phenotype and iv) adoptive transfer of isolated and separated CD4+ T cells subsets into
irradiated animals.
2.1.3. B cells and other antigen presenting cells.
In previous paragraphs much focus has been given to the involvement of CD4+ T cells in
the production of ADAs. However, other immune cells beside CD4+ T cells are also taking
part in the immune responses against rh proteins. For example MZ B cells were shown to
be important players. Sauerborn et al showed that after MZ depletion the ADA formation
is impaired. She proposed an immunogenicity mechanism, in which MZ B cells are being
activated by protein aggregates and later on CD4+ T cells and/or other subsets of B cells
become involved 7. Data presented in Chapter 3 suggest a rapid recruitment of CD4+ T cell
in immunogenicity as in this study no significant delay in CD4+ recruitment was observed
when compared to the TD response observed in non-tg mice. However, the minimal time
difference which could be detected by our experimental setup was ~24 hours. Therefore
future studies focusing on the first hours of response are necessary to confirm the hypothesis
of Sauernborn. Moreover, as depletion of MZ B cells did not lead to complete inhibition
of ADA production, other immune cells might be activated simultaneously to MZ B cells.
Dendritic cells, macrophages or other B cell subsets might also attract CD4+ T cells to
become involved in antibody production. More detailed studies are necessary to understand
the successive steps from trigger to actual ADA responses.
2.2. Biodistribution of aggregates upon injection.
Increasing number of reports correlate the presence of aggregates in the formulation
with an increased risk of ADA development 3,6,15. At the same time our knowledge on the
physicochemical characteristic of immunogenic aggregates increases. Lacking is, however,
data showing if physicochemical characteristic of aggregates found in formulated drug
are preserved upon administration. We do not know if their size and/or number change,
if they interact with other blood borne proteins, or how aggregates are distributed upon
injection. In Chapter 7 we show that in vivo and ex vivo fluorescence imaging techniques
are very powerful techniques for studying the fate of protein aggregates upon injection. We
showed that biodistribution strongly depends on the aggregation state of the protein, and
that upon IM and SC injection the highly aggregated formulation persists longer at the site
of injection compared to the nonstressed formulation containing mainly monomer protein.
Also, we detected the formation of fluorescent “hot-spots” in the livers, lungs and spleens
134
Summary.
of animals administered with highly aggregated protein either IV or IP. Future studies
should be performed to determine how the size of aggregates affects their clearance from
the site of injection and their biodistribution. Also, experiments studying if the activation of
immune cells in organs with “hot-spots” differs from activation of cells in similar organs,
without these “hot spots” should be performed. Moreover, fluorescence imaging techniques
might allow studying the degradation of aggregates in vivo. In the experiments described in
Chapter 7 one dye was used to label protein monomers and aggregates, which did not allow
detecting either of these species separately in vivo. Fluorescence resonance energy transfer
(FRET) or the use of fluorescence quenchers would allow to assess which form of the protein
(aggregated or monomeric) is removed from the site of injection and would allow following
the aggregates and monomers administered simultaneously. FRET would require labeling
protein monomers (components of aggregates) with two different fluorophores (donor and
acceptor). If the donor and acceptor were in close proximity (1-10nm), the energy absorbed
by donor would be transferred to acceptor inducing fluorescence. It would allow to measure
if aggregates upon injection remain non-degraded and when the degradation starts. Labeling
a protein with a fluorophore together with its quencher would allow to assess the time
it takes for the aggregate to degrade as only products of degradation, monomers labeled
only with fluorophore, would be measurable. Another intriguing question is if aggregates of
different proteins, when displaying similar physicochemical characteristic, are distributed
in a similar way in vivo. Moreover, many protein products contain HSA as the stabilizing
compound. It would be of great interest to study if addition of HSA in the formulation
affects biodistribution of rh proteins.
2.3. Determine how universal the mechanism is that underlies immunogenicity of
therapeutic proteins.
Most data presented in this thesis have been obtained using rhIFNβ. One could argue
whether these results might be translated to other rh therapeutic proteins. Protein drugs
are very heterogeneous molecules with different sizes, structures, posttranslational
modifications, potencies etc. Clinical and animal studies seem to suggest that the main
components of immunogenicity mechanism might be universal for all therapeutic proteins.
For example, for both rhIFNβ and rhIFNα aggregates were shown to be the most potent
triggers of immunogenicity 4,11. Moreover, absence of rapid ADA production upon rechallenge suggesting lack of apparent memory formation was observed for both rhIFNβ
and antiTNF in clinical studies 8,21,22. However, future studies with other protein models
and corresponding immune tolerant animals are necessary to determine how similar the
mechanism is that underlies immunogenicity of rh therapeutic proteins.
3. Final remarks.
Therapeutic proteins are one of the fastest growing group of pharmaceuticals with respect to
the number of new products and their market value. Moreover, as patents covering the first
rh originator products are expiring, biosimilars are being introduced on the marker. Taken
together, the amount of rh proteins being brought onto the market will expand greatly,
135
Chapter 8.
meaning that immunogenicity will remain an important issue that needs to be tackled. A lot
of efforts have been performed to develop tools that predict immunogenicity preclinically.
However, without understanding the immune processes leading to ADA development,
none of the methods are fully reliable. More and more reports strengthen the suggestion
that the mechanism underlying immunogenicity of therapeutic proteins is different from
a classical adaptive immune response. Therefore immunogenicity prediction tools, which
are in majority based on current knowledge about classical adaptive immune responses,
are prone to over or underestimate the risk of ADA production. It is clear that in order to
improve the safety and performance of rh therapeutic proteins a better understanding of
immunogenicity is needed.
136
References.
Summary.
1.
Büttel IC, Chamberlain P, Chowers Y, Ehmann F, Greinacher A, Jefferis R, Kramer D,
Kropshofer H, Lloyd P, Lubiniecki A, Krause R, Mire-Sluis A, Platts-Mills T, Ragheb JA,
Reipert BM, Schellekens H, Seitz R, Stas P, Subramanyam M, Thorpe R, Trouvin JH, Weise
M, Windisch J, Schneider CK. 2011.Taking immunogenicity assessment of therapeutic
proteins to the next level. Biologicals 39:100–109
2.
Schellekens H. 2005. Factors influencing the immunogenicity of therapeutic proteins.
Nephrol Dial Transplant 20:vi3–9
3.
Barnard JG, Babcock K, Carpenter JF. 2013. Characterization and quantitation of aggregates
and particles in interferon-β products : potential links between product quality attributes and
immunogenicity. J Pharm Sci 102:915–928
4.
Van Beers MM, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
5.
Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggregates seem to play a key role
among the parameters influencing the antigenicity of interferon alpha (IFN-alpha) in normal
and transgenic mice. Pharm Res 14:1472–1478
6.
Fradkin AH, Carpenter JF, Randolph TW. 2009. W. Immunogenicity of aggregates of
recombinant human growth hormone in mouse models. J Pharm Sci 98:3247–3264
7.
Sauerborn M, Brinks V, Jiskoot W, Schellekens H. 2010. Immunological mechanism
underlying the immune response to recombinant human protein therapeutics. Trends
Pharmacol Sci 31:53–59
8.
Sauerborn M, van Beers MM, Jiskoot W, Kijanka GM, Boon L, Schellekens H, Brinks
V. 2012. Antibody response against Betaferon® in immune tolerant mice: involvement of
marginal zone B-cells and CD4+ T-cells and apparent lack of immunological memory. J Clin
Immunol 33:255-263
9.
Verthelyi D, Wang V. 2010. Trace levels of innate immune response modulating impurities
(IIRMIs) synergize to break tolerance to therapeutic proteins. PloS One 5:e15252
10.
Sharma B. 2007. Immunogenicity of therapeutic proteins. Part 2: impact of container
closures. Biotechnol Adv 25;318–324
11.
Hermeling S, Schellekens H, Maas C, Gebbink MF, Crommelin DJ, Jiskoot W. 2006.
Antibody response to aggregated human interferon alpha2b in wild-type and transgenic
immune tolerant mice depends on type and level of aggregation J Pharm Sci 95:1084–1096
12.
Pawate S, Sriram S. 2010. The role of infections in the pathogenesis and course of multiple
sclerosis. Ann Indian Acad Neurol 13:80–86
13.
Mulvey MR, Doupe M, Prout M, Leong C, Hizon R, Grossberndt A, Klowak M, Gupta A,
Melanson M, Gomori A, Esfahani F, Klassen L, Frost EE, Namaka M. 2011. Staphylococcus
aureus harbouring Enterotoxin A as a possible risk factor for multiple sclerosis exacerbations.
Mult Scler 17:397–403.
14.
Nicholson G, Haier J. 2010. Role of chronic bacterial and viral infections in neurodegenerative,
neurobehavioural, psychiatric, autoimmune and fatiguing illnesses: Part 2. Br J Med Practit
3:103-110
15. Rosenberg AS. 2006. Effects of protein aggregates: an immunologic perspective. AAPS J
8:E501–507
16.
Schellekens, H. 2002. Immunogenicity of therapeutic proteins: clinical implications and
future prospects. Clin Ther 24:1720–1740
17.
Singh, SK. 2011. Impact of product‐related factors on immunogenicity of biotherapeutics. J
137
Chapter 8.
Pharm Sci 100:354-87
18.
Mahler HC, Friess W, Grauschopf U, Kiese S. 2009. Protein aggregation : pathways,
induction factors and analysis J Pharm Sci 98:2909–2934
19.
Bluestone JA, Abbas AK. 2003. Natural versus adaptive regulatory T cells. Nat Rev Immunol
3:253–257
20.
Mosmann TR, Cherwinski H, Bond MW, Giedlin MA, Coffman RL. 1986. Two types of
murine helper T cell clone. I. Definition according to profiles of lymphokine activities and
secreted proteins. J Immunol 136:2348-2357
21.
Perini P, Facchinetti A, Bulian P, Massaro AR, DePascalis D, Bertolotto A, Biasi G, Gallo P.
2001. Interferon-beta (INF-b ) antibodies in multiple sclerosis patients. Prevalence, kinetics,
cross-reactivity, and factors enhancing interferon-b immunogenicity in. Eur Cytokine Netw
12:56-61
22.
Ben-Horin S, Mazor Y, Yanai H, Ron Y, Kopylov U, Yavzori M, Picard O, Fudim E, Maor Y,
Lahat A, Coscas D, Eliakim R, Dotan I, Chowers Y. 2012. The decline of anti-drug antibody
titres after discontinuation of anti-TNFs: implications for predicting re-induction outcome in
IBD. Aliment Pharmacol Ther 35:714-722
138
Appendices
Nederlandse samenvatting.
139
Appendices.
1. Samenvatting.
Sinds recombinante humane (rh) therapeutische eiwitten op de markt zijn gebracht in de jaren
90, zijn de beschikbaarheid en veiligheid van eiwitmedicijnen significant verbeterd. Jammer
genoeg wordt de effectiviteit van de hedendaagse rh therapeutische eiwitten geplaagd door
de vorming van antilichamen gericht tegen deze rh eiwitten. De antilichamen kunnen leiden
tot immuun-gerelateerde bijwerkingen, tot verlaagde effectiviteit en tot levensbedreigende
auto-immuunreacties. De zogenoemde immunogeniciteit van rh therapeutische eiwitten
schaadt dus het patiënten welzijn en verhoogt de kosten van de therapie (Hoofdstuk 2).
Er zijn verschillende factoren die de kans op antilichaamvorming kunnen verhogen 1, 2.
De meeste studies laten echter zien dat eiwitaggregaten aanwezig in de rh therapeutische
eiwitten een belangrijke rol spelen 3-6. Desalniettemin is het niet bekend hoe deze aggregaten
immunogeniciteit veroorzaken.
De eerste therapeutische eiwitten werden geïsoleerd uit dierlijk weefsel en werden herkend
als lichaamsvreemd door het immuunsysteem van de patiënten. De antilichaamrespons die
de therapeutische eiwitten van dierlijke herkomst induceerden was via een klassiek T cel
afhankelijk (TA) immuunmechanisme 7.
Hedendaagse therapeutische eiwitten worden geproduceerd via recombinante DNA
technieken, en zijn vooral (homologen van) humane eiwitten. Ze hebben een identieke of
zeer vergelijkbare aminozuursequentie vergeleken met endogeen geproduceerde eiwitten
in mensen, en zouden daarom niet als lichaamsvreemd herkend moeten worden. Echter,
ondanks de gelijkenis in aminozuursequenties induceren rh therapeutische eiwitten nog
steeds een antilichaamrespons in patiënten. Zoals eerder aangegeven zijn aggregaten een
risicofactor voor immunogeniciteit. Een van de mogelijkheden is dat aggregaten herhaalde
epitopen bevatten die door binding aan B cel receptoren en het clusteren van B cel receptoren,
B cellen activeren via een T cel onafhankelijk (TO) type 2 immuunmechanisme.
De studies beschreven in dit proefschrift hebben tot doel het beter begrijpen van het
immuunmechanisme dat verantwoordelijk is voor immunogeniciteit van rh therapeutische
eiwitten. Het model dat ik hiervoor heb gebruikt zijn immuuntolerante transgene (tg) en
controle muizen (n-tg), die behandeld zijn met (commercieel beschikbare) rh interferon
beta producten met verschillende typen en hoeveelheden aggregaten. Deze zijn rhIFNβ1b (Betaferon ®), een zeer immunogeen product dat ongeveer 60% aggregaten bevat, en
rhIFNβ-1a (Avonex ®), een product met ~5% aggregaten, en wat minimaal immunogeen is.
De experimenten beschreven in Hoofdstuk 3 zijn een vervolg op recente resultaten van
Sauerborn en collega’s 8, en van Beers en collega’s 4. Hun experimenten laten zien dat CD4+
T cellen cruciaal zijn in de antilichaamproductie tegen geaggregeerd rh interferon beta -1b
(rhIFNβ-1b, Betaferon ®) en dat er geen immunologisch geheugen gevormd lijkt te worden
tegen rhIFNβ-1b of rhIFNβ-1a (Avonex ®). Het eerste experiment beschreven in Hoofdstuk
3 had als doel het bepalen wanneer CD4+ T cellen betrokken zijn bij de antilichaamrespons
tegen geaggregeerd rhIFNβ-1b. Hiervoor werden CD4+ T cellen op verschillende
140
Nederlandse samenvatting.
tijdspunten na de start van rhIFNβ-1b behandeling geëlimineerd. De resultaten laten zien
dat de antilichaamrespons alleen verhindert kan worden als CD4+ T cellen geëlimineerd zijn
voor de start van behandeling of daags na de start van de rhIFNβ-1b behandeling. CD4+ T
cellen lijken dus vooral betrokken te zijn bij de eerste fasen van de antilichaamproductie.
In het tweede experiment hebben we gekeken naar de aanwezigheid van germinale centra
(GC) in de milt tijdens rhIFNβ-1a behandeling. GC worden vooral gevormd als onderdeel
van een klassieke TA immuunrespons, en de vorming ervan valt samen met de vorming van
immunologisch geheugen. De data laat zien dat er gelijke aantallen GC zijn gevormd in tg
en n-tg muizen gedurende rhIFNβ-1a behandeling, echter in de tg muizen is de vorming van
GC niet gecorreleerd aan de antilichaamrespons tegen rhIFNβ-1a. In een derde experiment
hebben we gekeken of het verhogen van de rhIFNβ-1b dosis zou leiden tot het doorbreken
van immuuntolerantie en tot de vorming van immunologisch geheugen in tg muizen. Zelfs
bij de hoogste rhIFNβ-1b dosis getest is geen antilichaamrespons gemeten die indicatief is
voor de vorming van immunologisch geheugen.
Verontreinigingen, onzuiverheden en stabilisatoren in de formulering zouden mogelijk
als adjuvans kunnen werken en zo de antilichaamrespons tegen rh eiwitten kunnen
versterken 9, 10. Om deze hypothese te testen is een serie experimenten uitgevoerd waarin
het effect van veelgebruikte adjuvans op de antilichaamrespons tegen rhIFNβ-1a is getest
(Hoofdstuk 4). Van alle geteste adjuvans is alleen cholera toxine subunit B (CTB) in
staat om de antilichaamrespons tegen rhIFNβ-1a effectief te verhogen in tg muizen. In de
n-tg muizen verhogen saponine, montanide, ISA 50V2, aluminium hydroxide en CTB de
antilichaamrespons significant. In een vervolgexperiment is gekeken of CTB de tolerantie
voor rhIFNβ-1a in tg muizen had doorbroken. In dit experiment werden tg muizen behandeld
met CTB en rhIFNβ-1a en vervolgens werd getest of immunologisch geheugen was gevormd.
Hoewel een sterke primaire immuunrespons aanwezig was, was er geen antilichaamrespons
indicatief voor de vorming van immunologisch geheugen. In het laatste experiment is getest
of veelgebruikte immunosuppressieve middelen de antilichaamrespons tegen rhIFNβ-1b
kunnen verlagen. In tg muizen leidt alleen rapamycine tot een lagere antilichaamrespons, en
in n-tg muizen kunnen zowel rapamycine en cyclosporine de antilichaamrespons blokkeren.
Aggregaten zijn zeer sterke triggers van immunogeniciteit tegen rh therapeutische eiwitten
6, 11
. Echter, andere factoren zouden ook bij kunnen dragen aan de vorming van een
antilichaamrespons. Greinacher en collega’s hebben laten zien dat de immunogeniciteit
in patiënten behandeld met heparine infusies gecorreleerd zou kunnen zijn aan bacteriële
infecties. Bacteriën zouden een complex kunnen vormen met platelet factor 4 (PF4), en dit
bacteriële/PF4 complex zou een antilichaamrespons kunnen induceren dat tevens gericht is
tegen heparine/PF4 complexen. De experimenten in Hoofdstuk 5 hadden als doel het testen
of bacteriën tevens een complex kunnen vormen met rhIFNβ. Omdat patiënten behandeld
met rhIFNβ (patiënten met multiple sclerosis-MS) vaker bacteriële infecties hebben dan
gezonde mensen 12-14, zou de vorming van bacteriële/ rhIFNβ complexen misschien kunnen
141
Appendices.
leiden tot verhoogde kans op immunogeniciteit tegen rhIFNβ. We hebben de binding van
verschillende typen bacteriën met zowel monomeer als geaggregeerd rhIFNβ gemeten, en
alle geteste bacteriële stammen binden aan rhIFNβ-1b. Echter de binding lijkt in sterke mate
af te hangen van het type rhIFNβ. Ook lijkt de binding tussen bacteriën en rhIFNβ plaats
te vinden door middel van elektrostatische krachten, hoewel rhIFNβ sterker lijkt te binden
aan gram positieve bacteriën dan gram negatieve bacteriën. De data laat zien dat bacteriële
infecties mogelijk bij kunnen dragen aan immunogeniciteit van rhIFNβ, echter, aanvullende
studies in dieren en patiënten zijn nodig om deze hypothese te testen.
Het behandelingsregime zou tevens een rol kunnen spelen bij de vorming van antilichamen
tegen therapeutische eiwitten. Er zijn echter geen gedetailleerde studies uitgevoerd waarbij
de invloed van verschillen in dosis, administratieroute en frequentie van administratie
zijn gemeten. Er wordt gedacht dat intraveneuze administratie het minst immunogeen is,
gevolgd door intramusculaire administratie en subcutane administratie. Deze aannames
zijn gebaseerd op vaccinatiestudies (met lichaamsvreemde stoffen en adjuvans), en
op studies waarbij immunogeniciteit van verschillende rh therapeutische eiwitten met
verschillende toedieningsroutes, toedieningsschema’s en dosissen werden vergeleken. De
experimenten in Hoofdstuk 6 beschrijven hoe verschillen in toedieningsfrequentie, dosis en
toedieningsroute bij kunnen dragen aan immunogeniciteit van een enkel therapeutisch eiwit
met hoge mate van aggregatie (rhIFNβ-1b). In tegenstelling tot de beschreven aannames,
wijzen de resultaten erop dat intraveneuze administratie het meest immunogeen is in de tg
muizen. Tevens neemt de antilichaamrespons toe bij een hogere frequentie van toediening
en een hogere dosis.
Van alle risicofactoren die bij kunnen dragen aan immunogeniciteit van rh therapeutische
eiwitten, lijken aggregaten het meest belangrijk te zijn 3,7,15-17. Verschillende studies
hebben gekeken naar de fysisch chemische eigenschappen van eiwitaggregaten en hoe
deze eigenschappen samenhangen met immunogeniciteit 3, 11, 18. Echter, er zijn nog geen
studies uitgevoerd naar wat er gebeurt met de eiwitaggregaten na toediening. Hoofdstuk 7
beschrijft een experiment waarbij de biodistributie van eiwitaggregaten in vivo is bestudeerd.
Door middel van in vivo en ex vivo fluorescentie visualisatie hebben we gevonden dat
na intramusculaire en subcutane injectie de eiwitaggregaten langer op de injectieplek
verblijven dan niet geaggregeerd eiwit. Tevens zijn er “hotspots” met hoge fluorescentie
aanwezig in de longen en milt na intraveneuze administratie van eiwitaggregaten. Deze
hotspots kunnen eiwitaggregaten zijn in de capillairen van de longen of macrofagen in de
milt. Na intraperitoneale injectie van eiwitaggregaten zijn dezelfde “hotspots” zichtbaar in
de lever en milt.
De data beschreven in dit proefschrift benadrukken dat de immuunrespons tegen rhIFNβ
uniek is en een nog niet beschreven type immuunrespons reflecteert. Deze immuunrespons
heeft karakteristieken van een TA respons zoals een afhankelijkheid van CD4+ T cellen,
de timing waarbij deze cellen betrokken zijn, en de vorming van GC. Echter de resultaten
142
Nederlandse samenvatting.
beschreven in dit proefschrift laten ook verschillen zien ten opzichte van een TA respons. Zo
lijken de signalen benodigd voor het induceren van een antilichaamrespons anders dan de
signalen geassocieerd met een TA respons omdat TA adjuvans geen effect lijken te hebben
op immunogeniciteit in de tg muizen. Immunogeniciteit van rh therapeutische eiwitten lijkt
minder afhankelijk van interleukine 2 (IL-2). Tevens lijkt er geen immunologisch geheugen
gevormd te worden, ondanks stimulatie met CTB. Dit wijst erop dat (i) er een suppressief
mechanisme is dat de processen betrokken bij B cel maturatie en differentiatie blokkeert, (ii)
de betrokken T cel en B cel subsets niet differentiëren in geheugencellen of (iii) gevormde
geheugencellen inactief worden.
De verkregen resultaten laten duidelijk zien dat de dogma’s van de klassieke immunologie
en de conclusies getrokken met behulp van vaccinstudies niet toepasbaar zijn op het
mechanisme dat verantwoordelijk is voor immunogeniciteit van rh therapeutische eiwitten.
2. Perspectieven.
2.1 Immunologisch mechanisme verantwoordelijk voor immunogeniciteit.
Onze onderzoeksgroep bestudeert het immunologische mechanisme dat verantwoordelijk
is voor immunogeniciteit van rh therapeutische eiwitten. Hierbij behandelen we
immuuntolerante transgene muizen met therapeutische eiwitten waarmee patiënten
behandeld worden. Dit model is zeer geschikt om de processen te bestuderen die tot
antilichaamrespons tegen therapeutische eiwitten leiden in patiënten. Vroegere hypothesen
beschreven dat eiwitaggregaten aanwezig in de producten zouden kunnen lijken op bacteriële
polysachariden, en dus dat aggregaten herhaalde structurele epitopen bevatten die B cellen
kunnen activeren. Echter, zowel dierstudies als klinische studies laten zien dat de vorming
van antilichamen tegen therapeutische eiwitten complexer is. Zo heeft immunogeniciteit
van therapeutische eiwitten karakteristieken van zowel een TO als een TA immuunrespons.
Sauerborn en collega’s hebben laten zien dat marginale zone B cellen, een van de celtypen
betrokken bij een TO respons, cruciaal zijn bij het induceren van antilichamen 8. Van Beers
en collega’s hebben echter laten zien dat de immuunrespons zeer beperkt is als CD4+ T
cellen inactief worden gemaakt 8. De betrokkenheid van CD4+ T cellen garandeert echter
niet de vorming van immunologisch geheugen 4. De studies beschreven in Hoofdstukken
3-5 van dit proefschrift geven meer inzicht in het mechanisme verantwoordelijk voor
immunogeniciteit van therapeutische eiwitten, echter, een volledig inzicht in dit mechanisme
is nog niet aanwezig.
2.1.1 CD4+ regulatorische T cellen.
Een van de meest intrigerende vragen betreffende het mechanisme dat immunogeniciteit
onderligt is waarom CD4+ T cellen betrokken zijn bij de antilichaamrespons tegen rh
eiwitten. Omdat de aminozuursequenties van rh therapeutische eiwitten (vrijwel) identiek
zijn aan endogene eiwitten, zouden de CD4+ T cellen geen epitopen moeten herkennen in
deze eiwitten, of de respons tegen de epitopen zou moeten worden geblokkeerd door een
regulatoir mechanisme. De belangrijkste groep van regulatorische cellen zijn de CD4+ T
143
Appendices.
regulatorische cellen (Tregs). Deze cellen verhinderen ongewenste immuunresponsen en
verlagen aanwezige immuunresponsen om zo weefselschade te beperken 19. De vraag is of,
en zo ja, hoe eiwitaggregaten deze Tregs omzeilen. Dit scenario van het omzeilen van Tregs
verklaart echter nog niet waarom immunologisch geheugen niet gevormd lijkt te worden.
Een andere mogelijkheid is dat de aggregaten Tregs stimuleren om effector cellen te worden.
Verdere studies naar de rol van Tregs in immunogeniciteit zullen belangrijk zijn om de te
bepalen hoe belangrijk deze CD4+ T cel subset is in de formatie van antilichamen. Een
mogelijke studie is het blokkeren of verwijderen van Tregs in de immuuntolerante muizen
door specifieke antilichamen, en vervolgens bepalen of dit immunogeniciteit beïnvloedt.
2.1.2. Andere CD4+ T cel subsets.
Tregs zijn veelbelovende CD4+ T cellen die mogelijk betrokken zijn bij immunogeniciteit van
rh therapeutische eiwitten, echter andere CD4+ T cel subsets zouden ook belangrijk kunnen
zijn. De studies beschreven in Hoofdstuk 4 waarin immunosuppressieve middelen zijn
gebruikt laten zien dat de relatieve betrokkenheid van CD4+ T cel subsets mogelijk verschilt
tussen tg en n-tg muizen. In 1980 hebben Mosmann en collega’s gevonden dat muizen CD4+
T cellen een heterogene populatie zijn, en hebben toen (het dualisme van) Th1 en Th2 CD4+
T cellen geïntroduceerd 20. Tegenwoordig zijn ook Th17, Th9, Th22 en folliculaire T cellen
geïdentificeerd. In onze studies (Hoofdstuk 3) hebben we de gehele CD4+ T cel populatie
verwijderd, en het is dus mogelijk dat verschillende CD4+ T cel subsets in verschillende
mate betrokken zijn bij immunogeniciteit. Gedetailleerde studies naar de betrokkenheid van
de verschillende subsets is nodig om het mechanisme leidend tot immunogeniciteit van rh
therapeutische eiwitten te verduidelijken. Verschillende studies kunnen worden uitgevoerd:
(i) depletie van verschillende CD4+ T cel subsets, (ii) bepalen welke subsets geactiveerd
worden gedurende antilichaamproductie, (iii) isolatie van antigeen specifieke CD4+ T cellen
en karakterisatie van hun fenotype en (iv) adoptive transfer van CD4+ T cel subsets uit
antilichaam producerende muizen naar bestraalde muizen.
2.1.3. B cellen en andere antigeen presenterende cellen.
In de vorige paragrafen is vooral ingegaan op de betrokkenheid van CD4+ T cellen bij
de vorming van antilichamen. Echter, naast CD4+ T cellen zijn ook andere immuuncellen
betrokken bij de immuunresponsen tegen rh therapeutische eiwitten. Marginale zone B
cellen zijn bijvoorbeeld belangrijke spelers. Sauerborn en collega’s hebben laten zien dat
de depletie van marginale zone B cellen resulteert in verminderde antilichaamproductie.
Sauerborn heeft een immuunmechanisme voorgesteld waarin marginale zone B cellen
geactiveerd worden door eiwitaggregaten, gevolgd door activatie van CD4+ T cellen en/of
andere B cel subsets 7. De resultaten gepresenteerd in Hoofdstuk 3 suggereren dat CD4+ T
cellen relatief vroeg in de immunogeniciteitsreactie betrokken zijn, en dat de timing hiervan
hetzelfde is tussen tg, en n-tg muizen die een TA respons hebben. Echter, het experiment
beschreven in Hoofdstuk 3 kijkt naar de betrokkenheid van CD4+ T cellen in een interval
van 24 uur, dus vervolgstudies waarin van uur tot uur wordt gekeken naar de betrokkenheid
van CD4+ T cellen zijn nodig om de hypothese van Sauerborn en collega’s te testen. Omdat
144
Nederlandse samenvatting.
depletie van marginale zone B cellen niet tot complete inhibitie van de antilichaamrespons
leidt, is het waarschijnlijk dat andere immuuncellen geactiveerd worden naast deze marginale
zone B cellen. Zo zouden dendritische cellen, macrofagen, of andere B cel subsets de CD4+
T cellen kunnen betrekken bij de antilichaamrespons. Meer gedetailleerde studies naar
het stapsgewijze verloop van antilichaamrespons zijn nodig om immunogeniciteit van rh
therapeutische eiwitten te begrijpen.
2.2 Biodistributie van aggregaten na injectie.
In een groeiend aantal artikelen wordt de aanwezigheid van aggregaten in de formulering
gelinkt aan een verhoogde kans op antilichaamvorming 3,6,15. Tegelijkertijd is onze kennis
over de fysisch chemische eigenschappen van immunogene aggregaten aan het vergroten.
Wat echter niet bekend is, is of de fysisch chemische eigenschappen van de aggregaten
veranderen na toediening. We weten niet of de aggregaatgrootte of aantallen veranderen,
of de aggregaten een interactie aangaan met eiwitten aanwezig in het bloed, of hoe
de aggregaten distribueren na toediening. In Hoofdstuk 7 laten we zien dat in vivo en
ex vivo beeldvorming door middel van fluorescente labeling van eiwit (aggregaten) zeer
geschikt is om de distributie van aggregaten te volgen na injectie. We hebben laten zien
dat de biodistributie bepaald wordt door de mate van aggregatie. Bovendien leidt injectie
van aggregaten via de intramusculaire en subcutane route tot langere ophoping van
aggregaten op de plaats van injectie vergeleken met injectie van monomeer eiwit. Ook
hebben we fluorescente “hotspots” in de levers, longen en milten gevonden van muizen
behandeld met geaggregeerd eiwit na intraveneuze en intraperitoneale injecties. Een
interessante toekomstige studie is het bestuderen of de activatie van immuuncellen in de
organen met “hotspots” anders is dan de activatie van immuuncellen in organen zonder
“hotspots”. Andere toekomstige studies naar de invloed van aggregaatgrootte op de klaring
van deze eiwitten, en op de biodistributie zullen tevens uitgevoerd moeten worden. In de
experimenten beschreven in Hoofdstuk 7, is 1 kleuring gebruikt om eiwitmonomeren en
aggregaten te labelen. Het gevolg hiervan is dat er geen onderscheid tussen monomeren en
aggregaten gemaakt kan worden in vivo. FRET (fluorescence resonance energy transfer) of
het gebruik van fluorescente “quenchers” zouden gebruikt kunnen worden om onderscheid
te maken tussen geaggregeerd en monomeer eiwit. Deze methoden zouden dus gebruikt
kunnen worden om beide afzonderlijk te volgen in vivo en zo te bepalen welke van de
twee lang op de injectieplek blijft. FRET houdt in dat eiwitmonomeren (de bouwstenen van
aggregaten) met twee verschillende fluoroforen gelabeld worden (donor en acceptor). Zodra
de donor en acceptor in nabijheid van elkaar zijn (1-10 nm), zal er energie van de donor
naar de acceptor gaan en zo een fluorescent signaal afgeven. Deze methode zal geschikt zijn
om te bestuderen of aggregaten na injectie uiteen vallen of intact blijven. Het eiwit kan ook
gelabeld worden met een fluorofoor en quencher, en in dat geval zou tevens de degradatie
van aggregaten gemeten kunnen worden; alleen degradatieproducten (monomeer gelabeld
met fluorofoor) zijn meetbaar.
Een andere vraag is of aggregaten van verschillende eiwitten, maar met dezelfde fysisch
145
Appendices.
chemische eigenschappen op dezelfde manier gedistribueerd worden in vivo. Omdat veel rh
therapeutische eiwitten tevens het eiwit human serum albumine bevatten als stabilisator, zou
het ook zeer interessant zijn om te bestuderen of dit eiwit de distributie van rh therapeutische
eiwitten beïnvloedt.
2.3. Bepalen hoe universeel het mechanisme is dat de immunogeniciteit van rh
therapeutische eiwitten onderligt.
De data gepresenteerd in dit proefschrift is vooral gericht op het gebruik van rh interferon
beta als therapeutisch eiwit. Een legitieme vraag is of de resultaten verkregen met dit eiwit
ook toepasbaar zijn op andere rh therapeutische eiwitten, sinds eiwitten erg heterogene
moleculen zijn met verschillende grootte, structuur, posttranslationele modificatie,
potentie etc. Zowel klinische en dierstudies lijken erop de wijzen dat de mechanismen
verantwoordelijk voor immunogeniciteit in grote lijnen overeenkomen. Zo zijn voor zowel
rhIFNβ en rhIFNα aggregaten potentiele triggers voor immunogeniciteit 4, 11 . Ook is er voor
rhIFNβ en antiTNF aangetoond dat immunologisch geheugen niet gevormd lijkt te worden
in patiënten 8, 21, 22. Echter, studies met andere rh therapeutische eiwitten en bijbehorende
immuuntolerante tg muizen moeten worden uitgevoerd om de mate van overlap in de
immuunmechanismen leidend tot antilichaamproductie tegen rh therapeutische eiwitten te
bepalen.
3. Conclusies.
Therapeutische eiwitten zijn een van de snelst groeiende groep van medicijnen met
betrekking tot het aantal nieuwe producten en marktaandeel. Bovendien verlopen de
patenten van de eerste rh therapeutische eiwitten waardoor biosimilars geïntroduceerd
worden op de farmaceutische markt. Het aantal therapeutische eiwitten zal dus enorm
toenemen, en immunogeniciteit van deze producten zal dus een groot probleem blijven.
Veel onderzoek is erop gericht om preklinische tests te ontwikkelen die immunogeniciteit
van rh therapeutische eiwitten kunnen voorspellen. Echter we weten nog niet waarom
therapeutische eiwitten immunogeen zijn. Een toenemend aantal studies laat zien dat
het mechanisme verantwoordelijk voor immunogeniciteit van rh therapeutische eiwitten
niet een klassieke adaptieve immuunrespons is. De preklinisch voorspellende tests, die
gebaseerd zijn op een klassieke adaptieve immuunrespons, zijn daarom geneigd om het
risico op immunogeniciteit te onder- of overschatten. Om de veiligheid en effectiviteit
van rh therapeutische eiwitten te verbeteren is meer inzicht in de immuunmechanismen
verantwoordelijk voor immunogeniciteit van deze medicijnen nodig.
146
Referenties.
Nederlandse samenvatting.
1.
Büttel IC, Chamberlain P, Chowers Y, Ehmann F, Greinacher A, Jefferis R, Kramer D,
Kropshofer H, Lloyd P, Lubiniecki A, Krause R, Mire-Sluis A, Platts-Mills T, Ragheb JA,
Reipert BM, Schellekens H, Seitz R, Stas P, Subramanyam M, Thorpe R, Trouvin JH, Weise
M, Windisch J, Schneider CK. 2011.Taking immunogenicity assessment of therapeutic
proteins to the next level. Biologicals 39:100–109
2.
Schellekens H. 2005. Factors influencing the immunogenicity of therapeutic proteins.
Nephrol Dial Transplant 20:vi3–9
3.
Barnard JG, Babcock K, Carpenter JF. 2013. Characterization and quantitation of aggregates
and particles in interferon-β products : potential links between product quality attributes and
immunogenicity. J Pharm Sci 102:915–928
4.
Van Beers MM, Sauerborn M, Gilli F, Brinks V, Schellekens H, Jiskoot W. 2010. Aggregated
recombinant human interferon beta induces antibodies but no memory in immune-tolerant
transgenic mice. Pharm Res 27:1812–1824
5.
Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggregates seem to play a key role
among the parameters influencing the antigenicity of interferon alpha (IFN-alpha) in normal
and transgenic mice. Pharm Res 14:1472–1478
6.
Fradkin AH, Carpenter JF, Randolph TW. 2009. W. Immunogenicity of aggregates of
recombinant human growth hormone in mouse models. J Pharm Sci 98:3247–3264
7.
Sauerborn M, Brinks V, Jiskoot W, Schellekens H. 2010. Immunological mechanism
underlying the immune response to recombinant human protein therapeutics. Trends
Pharmacol Sci 31:53–59
8.
Sauerborn M, van Beers MM, Jiskoot W, Kijanka GM, Boon L, Schellekens H, Brinks
V. 2012. Antibody response against Betaferon® in immune tolerant mice: involvement of
marginal zone B-cells and CD4+ T-cells and apparent lack of immunological memory. J Clin
Immunol 33:255-263
9.
Verthelyi D, Wang V. 2010. Trace levels of innate immune response modulating impurities
(IIRMIs) synergize to break tolerance to therapeutic proteins. PloS One 5:e15252
10.
Sharma B. 2007. Immunogenicity of therapeutic proteins. Part 2: impact of container
closures. Biotechnol Adv 25;318–324
11.
Hermeling S, Schellekens H, Maas C, Gebbink MF, Crommelin DJ, Jiskoot W. 2006.
Antibody response to aggregated human interferon alpha2b in wild-type and transgenic
immune tolerant mice depends on type and level of aggregation J Pharm Sci 95:1084–1096
12.
Pawate S, Sriram S. 2010. The role of infections in the pathogenesis and course of multiple
sclerosis. Ann Indian Acad Neurol 13:80–86
13.
Mulvey MR, Doupe M, Prout M, Leong C, Hizon R, Grossberndt A, Klowak M, Gupta A,
Melanson M, Gomori A, Esfahani F, Klassen L, Frost EE, Namaka M. 2011. Staphylococcus
aureus harbouring Enterotoxin A as a possible risk factor for multiple sclerosis exacerbations.
Mult Scler 17:397–403.
14.
Nicholson G, Haier J. 2010. Role of chronic bacterial and viral infections in neurodegenerative,
neurobehavioural, psychiatric, autoimmune and fatiguing illnesses: Part 2. Br J Med Practit
3:103-110
15. Rosenberg AS. 2006. Effects of protein aggregates: an immunologic perspective. AAPS J
8:E501–507
16.
Schellekens, H. 2002. Immunogenicity of therapeutic proteins: clinical implications and
future prospects. Clin Ther 24:1720–1740
17.
Singh, SK. 2011. Impact of product‐related factors on immunogenicity of biotherapeutics. J
147
Appendices.
Pharm Sci 100:354-87
18.
Mahler HC, Friess W, Grauschopf U, Kiese S. 2009. Protein aggregation : pathways,
induction factors and analysis J Pharm Sci 98:2909–2934
19.
Bluestone JA, Abbas AK. 2003. Natural versus adaptive regulatory T cells. Nat Rev Immunol
3:253–257
20.
Mosmann TR, Cherwinski H, Bond MW, Giedlin MA, Coffman RL. 1986. Two types of
murine helper T cell clone. I. Definition according to profiles of lymphokine activities and
secreted proteins. J Immunol 136:2348-2357
21.
Perini P, Facchinetti A, Bulian P, Massaro AR, DePascalis D, Bertolotto A, Biasi G, Gallo P.
2001. Interferon-beta (INF-b ) antibodies in multiple sclerosis patients. Prevalence, kinetics,
cross-reactivity, and factors enhancing interferon-b immunogenicity in. Eur Cytokine Netw
12:56-61
22.
Ben-Horin S, Mazor Y, Yanai H, Ron Y, Kopylov U, Yavzori M, Picard O, Fudim E, Maor Y,
Lahat A, Coscas D, Eliakim R, Dotan I, Chowers Y. 2012. The decline of anti-drug antibody
titres after discontinuation of anti-TNFs: implications for predicting re-induction outcome in
IBD. Aliment Pharmacol Ther 35:714-722
148
Appendices
Affiliations of collaborating authors.
149
Appendices.
Affiliations of collaborating authors.
Loois Boon
Bioceros, Utrecht, the Netherands
Vera Brinks
Department of Pharmaceutics, Utrecht Institute for Pharmaceutical Sciences (UIPS),
Utrecht University, Utrecht, the Netherlands
Andreas Greinacher
Institut für Immunologie und Transfusionsmedizin, Ernst-Moritz-Arndt-Universität,
Greifswald, Germany
Inga Jensch
Zentrum für Innovationskompetenz-Humorale Immunreaktionen bei kardiovaskulären
Erkrankungen, Ernst-Moritz-Arndt-Universität, Greifswald, Germany
Wim Jiskoot
Department of Drug Delivery Technology, Leiden/Amsterdam Centre for Drug Research
(LACDR), Leiden University, Leiden, the Netherlands
Krystin Krauel
Zentrum für Innovationskompetenz-Humorale Immunreaktionen bei kardiovaskulären
Erkrankungen, Ernst-Moritz-Arndt-Universität, Greifswald, Germany
Malgorzata Prokopowicz
Department of Pharmaceutics, Utrecht Institute for Pharmaceutical Sciences (UIPS),
Utrecht University, Utrecht, the Netherlands
Melody Sauerborn
Department of Pharmaceutics, Utrecht Institute for Pharmaceutical Sciences (UIPS),
Utrecht University, Utrecht, the Netherlands
Huub Schellekens
Department of Pharmaceutics, Utrecht Institute for Pharmaceutical Sciences (UIPS),Utrecht
University, Utrecht, the Netherlands
Department of Innovation Management, Utrecht University, the Netherlands
Claudia Weber
Abteilung Genetik der Mikroorganismen, Interfakultäres Institut für Genetik und Funktionelle
Genomforschung, Ernst-Moritz-Arndt-Universität, Greifswald, Germany
150
Appendices
Curriculum Vitae.
151
Appendices.
Grzegorz Kijanka was born on 27th of February 1984 in Chrzanów,
Poland. In 2007 he obtained a Bachelor degree in Biotechnology
after successful defence of his bachelor dissertation written under
the supervision of prof. Arkadiusz Kozubek and dr Maria Stasiuk.
He continued his education on the Medical Biotechnology.
His master project was performed under the supervision of
dr M. Stasiuk. During his master studies he participated in the
SOCRATES/ERASMUS program and he did a 5 months internship
at the Department of Pharmaceutics of Utrecht University. During
this internship he worked with dr Melody Sauerborn on a project
focusing on the immunogenicity of therapeutic interferon beta.
In 2009 he obtained master degree. Shortly after master thesis defence, in August 2009,
he started his PhD project at the Department of Pharmaceutics of Utrecht University under
the supervision of prof. Huub Schellekens and dr Vera Brinks. As a part of his project he
spent 2 months as a guest researcher at The Center for Innovation Competence - Humoral
Immune Response in Cardiovascular Diseases (ZIK HIKE) at the University of Greifswald,
Germany. During his PhD he studied multiple problems connected with immunogenicity of
protein drugs with the main focus on the recombinant human interferon beta. The results
of his project are presented in this dissertation. From September 2013 he works as a post
doc fellow at Leiden Academic Center for Drug Research in the group of prof. Wim Jiskoot.
152
Appendices
List of publications.
153
Appendices.
Publications from this thesis.
Grzegorz Kijanka, Wim Jiskoot, Melody Sauerborn, Huub Schellekens, Vera Brinks.
Immunogenicity of therapeutic proteins in pharmaceutical formulation development of
peptides and proteins, CRC Press, 2012: pp 297-322
Grzegorz Kijanka, Wim Jiskoot, Huub Schellekens, Vera Brinks. Effect of treatment
regimen on the immunogenicity of human interferon beta in immune tolerant mice. Pharm.
Res. 2013:30;1553-1560
Grzegorz Kijanka, Melody Sauerborn, Louis Boon, Huub Schellekens, Vera Brinks.
Immunogenicity of recombinant human interferon beta in immune tolerant mice is
characterized by early involvement of CD4+ T cells, formation of germinal centers, but no
apparent immunological memory. Submitted for publication
Grzegorz Kijanka, Malgorzata Prokopowicz, Huub Schellekens, Vera Brinks. Influence of
aggregation and route of injection on the biodistribution of mouse serum albumin. Submitted
for publication
Grzegorz Kijanka, Huub Schellekens, Vera Brinks. The impact of adjuvants on the production
of anti rh interferon beta antibodies and formation of immune memory in immune tolerant
mice. Manuscript in preparation
Other publications.
Melody Sauerborn, Miranda van Beers, Wim Jiskoot, Grzegorz Kijanka, Louis Boon,
Huub Schellekens, Vera Brinks. Antibody response against Betaferon® in immune tolerant
mice: involvement of marginal zone B-cells and CD4+ T-cells and apparent lack of
immunological memory. J Clin Immunol. 2012:33;255-263
Maria Stasiuk, Grzegorz Kijanka, Arkadiusz Kozubek. Transformations of erythrocytes
shape and their regulation. Postepy Biochem. 2009:55;425-433
Selected abstracts.
Immunogenicity of therapeutic interferon beta; impact of route of administration and
dose in immune tolerant mice. Grzegorz Kijanka, Vera Brinks, Huub Schellekens. Oral
presentation, 3rd European Immunogenicity Platform Symposium, 8th-9th December 2010,
Ghent, Belgium
In vivo fate of protein aggregates upon injection – the use of fluorescence optical in vivo
imaging. Grzegorz Kijanka, Vasco Filipe, Wim Jiskoot, Vera Brinks, Huub Schellekens.
Oral presentation,4th European Immunogenicity Platform Symposium, 7th-8th February
2012, Copenhagen, Denmark
Immune mechanisms underlying immunogenicity of therapeutic proteins in immune
tolerant mice. Grzegorz Kijanka, Melody Sauerborn, Louis Boon, Huub Schellekens,
Vera Brinks. Oral presentation, 5th Open Scientific European Immunogenicity Platform
154
Symposium, 26th-27th February 2013, Munich, Germany
List of publications.
In vivo imaging – a useful technique to trace the fate of therapeutic protein aggregates
upon injection? Grzegorz Kijanka, Vasco Filipe, Wim Jiskoot, Vera Brinks, Huub
Schellekens. Poster presentation, Workshop on Protein Aggregation and Immunogenicity,
9th-12th July 2012, Breckenridge, USA
Tracking of the protein aggregates fate upon injection via different routes. Grzegorz
Kijanka, Malgorzata Prokopowicz, Huub Schellekens, Vera Brinks. Poster presentation,
Meeting of the Belgian-Dutch Biopharmaceutical Society, 9th November 2012, Utrecht,
The Netherlands
155
Appendices.
156
Appendices
Acknowledgements.
157
Appendices.
Acknowledgements.
Some people, including me, find it difficult to express themselves simply by words.
I am very grateful to everyone who has ever helped me with my work, however
putting acknowledgements in words is a task which overwhelmedme... The following
acknowledgements only partially express how grateful I am to every one who has ever
helped me during my PhD project.
In the first place I want to thank my wife Marta. Without your influence and neverending, every day support completing of this thesis wouldn’t be possible.
I thank my parents (Janina and Bernard), and my parents-in-low (Malgorzata and
Grzegorz) for help and support before and during the PhD project period.
Last but not least, I thank my sister Magdalena for great design of the book cover and the
illustrations for the chapters’ title pages.
I thank my promoter, professor Huub Schellekens. Dear Huub, you took the risk and you
hired me, a person with only basic immunology knowledge, on an immunological project.
I hope I haven’t disappointed you! Thanks for showing me what science is about. Thanks
for freedom you gave me in planning and performing my experiments, which was a great
possibility to develop myself as scientist.
Finishing of my project and writing this thesis wouldn’t be possible without my copromotor and daily supervisor doctor Vera Brinks. Dear Vera, you helped me so many
times and in so many aspects that I simply cannot list all of them here. I would need
another book for that  This book is a result of equally mine and your work. So allow me
to write simply: THANK YOU FOR EVERYTHING! By the time this book is printed you will
have probably known if you stay in academia or not, but I really hope you will. You would
be a great professor 
I thank my ERASMUS supervisor, dr Melody Sauerborn. Melody, without your kind word
to Huub I wouldn’t be considered by Huub as a candidate for PhD project. I will always be
grateful for that.
I also want to thank my colleagues from our small immunogenicity groups in Utrecht
(Andhyk, Haffid, Mohadeseh) and Leiden (Miranda, Riccardo and Vasco) for great
atmosphere in lab, fruitful discussions and help in practical work.
I thank my current boss professor Wim Jiskoot for all very interesting and valuable
discussions during “Utrecht/Leiden Immunogenicity meetings”.
Working with animals is a difficult piece of science. This is why I want to acknowledge the
Biofarmacy (Ebel) and GDL animal caretakers (Sabine, Anja, Ajit, Helma and many others).
Many thanks for all your help!
Special thanks for Mies. You are the hardest working man I have ever met. You were
constantly busy, but somehow you always found time to help me or to answer even the
stupidest question. And, what is the most astonishing, you were always in a good mood.
Also I would like to express my gratitude to ALL colleagues from the Department of
Pharmaceutics. I had great time working with you!
For hosting me in Greifswald and for great co-operation many thanks to Inga, Krystin,
158
Acknowledgements.
Christine and professor Andreas Greinacher. I had great time in Greifswald and I really
enjoyed working with you. We obtained very promising data and I hope the collaboration
between our labs will be continued.
Nothing in life is for free. Well… in most cases it’s true, but fortunately (for me) not
always. I thank Darren Baker and Louis Boon for supplying me with materials essential
for my work and for comments to my manuscripts.
Many thanks to Miranda van Amersfoot for help with the imaging equipment and
software.
If I forgot to mention someone, please forgive me! Please keep in mind that now, when I
am writing these acknowledgements, I am still able to think only about my child Alicja. I
am sorry, but those of you who are a parent for sure know what I mean 
159