Enabling Technologies for
Antibody Drug Discovery
John Babcook
Amgen British Columbia
June 8, 2009
The Evolution of Antibodyy Technology
gy
 1890 Emil von Behring – Anti-diptheria and tetanus exotoxin
antiserum
ti
th
therapy
 1897 Paul Ehrlich - Molecular basis of immunity, lock and key-type
fit was required between the antigen and cell side chains “antibody”
antibody
 1905 Clemens von Pirquet – serum sickness
 1959 Macfarlane Burnet – Clonal selection theory
 1963 Niels Jerne and Albert Nordin – demonstrated antibody
specificity
 1975 César Milstein and Georges Köhler – “magic bullet” murine
monoclonal antibodies by hybridoma technology
2
Antibodyy Technology
gy Development
p
Challenges
g
 Immunogenicity of non-human antibodies
–
–
–
–
Humanization
Phage display
Transgenic mice
SLAMTM
 Monoclonal antibody generation efficiencies
– Identification of mAbs with sufficient affinity and
functional activity to be therapeutics
3
Amgen British Columbia
Antibodyy Generation Technologies
g
XenoMouse®
SLAM
Optimized
H b id
Hybridoma
4
Sampling Immune Repertoires
# of Abs
# of
Abs
SLAM™ Technology
Classic Hybridoma
Technology
- Efficient Sampling -
- Inefficient Sampling >100 x Increase
Optimized
Hybridomas
Diversity
“Rare” Antibodies
5
SLAM™ (Selected Lymphocyte Antibody Method)
Isolation of
Immune B-cells
Single cell PCR
Babcook, J. S., Leslie, K.B., Olsen, O.A., Salmon, R.A.. and Schrader, J.W.. A novel strategy
for generating monoclonal antibodies from single, isolated lymphocytes producing antibodies of
defined specificities. Proc. Natl. Acad. Sci. USA 93:7843-7848 (1996).
6
SLAM™ (Selected Lymphocyte Antibody Method)
 Sample repertoires at the single cell level from any species.
 Unlike display
display-derived
derived antibodies
antibodies, native VH and VL pairings are
retained and in-vivo affinity maturation is exploited
 Ability
Abilit to
t select
l t best
b t antibodies
tib di from
f
l
large
panels
l off primary
i
B cells
ll
 Identification of lead antibodies for human therapy
 Reagents for diagnostics, PK-Tox and Clinical programs
 Isolation of human antibodies directly from human peripheral blood
 Infectious diseases: neutralizing antibodies to CMV
CMV, Tetanus and
Influenza have been isolated using SLAM
7
Human anti-Influenza Antibodies
 PBMCs were obtained from a recently vaccinated 33 year old
volunteer with a long clinical history of seasonal influenza-like
influenza like
infections
 A time course was p
performed p
post vaccination to look for circulating
g
peripheral plasma B-cells
 A peak was observed at day nine post immunization
 375 influenza specific plasma cells per milliliter of peripheral
blood
Single influenza specific cells were isolated and mAbs rescued by
SLAM
 Peripheral blood was collected from the same donor 2 months post
vaccination memory B
vaccination,
B-cells
cells were cultured and single influenza specific
cells were isolated and mAbs rescued by SLAM
8
Rescue of Anti-influenza mAbs byy SLAM™
Human anti-influenza
anti influenza plaque
forming cell
Transfected CHO anti-influenza
anti influenza
plaque forming cell
9
Structure of Influenza Virus
10
Anti-HA and NP Antibodies
IP of biotinylated
influenza vaccine
Western blot
of influenza vaccine
Monomer HA
(80 kD)
Monomer NP or NA
(55 kD)
11
Hemagglutination Inhibition (HI)
Influenza Cross-Specificity
p
y Analysis
y
 mAb8 (anti-H3N2) neutralizes A/Johannesburg/33/94 (1995/1996 vaccine strain) and
A/Nanchang/933/95 (1996/1997 vaccine strain)
 No evidence of cross-neutralization between different type A subtypes
 mAb-30 and mAb-37 (anti-H1N1) could also inhibit virus-mediated hemagglutination and
by three other H1N1 viruses: A/USSR/90/77, A/Chile/1/83, A/Taiwan/186 and
A/Memphis/3/96 (data not shown)
 Anti-type A mAbs did not inhibit hemagglutination by type B viruses (data not shown)
12
In-Vitro Neutralization of Influenza Infectivityy
mAb-37 on A/Texas/36/91 (H1N1)
mAb-30 on A/Texas/36/91 (H1N1)
IC50=5 ng/ml
IC50=6 ng/ml
mAb-8 on A/Johannesburg/33/94 (H3N2)
mAb-8 on A/Nanchang/933/95 (H3N2)
IC50=30 ng/ml
IC50=105 ng/ml
 Focus neutralization assays were performed to assess the ability of each mAb to neutralize influenza
virus infectivity of Madin-Darby Canine Kidney (MDCK) cells.
13
Apparent
pp
Affinities of Anti-Influenza mAbs
 All human mAbs bind the 1995/1996 vaccine with very high affinity
 mAb 8 binds the subsequent year’s H3N2 1996/1997 strain with reduced affinity, which
correlates with reduced potency of viral inhibition
14
XenoMouse® : Fully Human Antibodies
IgG1 IgG1
IgG1/ 
IgG2
IgG2/ 
IgG4 IgG4
IgG4/ 
15
XenoMouse® Technology:
Over 10 Years of Proven Success
Mouse
Ab Genes
Inactivated
Human
Ab Genes
Introduced
XenoMouse® Mice
 Vectibix®, a fully human anti-EGFR antibody, was generated through this
platform technology (Yang, X.D. et al. Cancer Res., 1999. Jackobovits, Nat
Biotechnol,, 2007))
 Denosumab, a fully human RANKL antibody, was generated using
XenoMouse® (Miller, Curr Osteoporos Rep, 2009)
16
XenoMouse® Technology:
gy Human Heavyy Chain
66VH
HPRT
y4H
y3H
DH JH C C
y2H
C
3’ Enhancer
y1H
 Contains 34 of 41 functional VH genes
 Contains entire DH (23) and JH (6) regions
 Contains C and C
 Three different versions constrained to class
switch from IgM to a single human IgG isotype:
IgG1, IgG2 or IgG4
17
XenoMouse® Technology: Human Light Chain
Human Ig Transgene
Op
Ap
Lp
B
J
Human Ig Transgene
C Kde
C
B
A
JC
3'E
3 2 1



0.8 Mb transgene
32 V genes from the
proximal
duplication
(18
functional V genes)
 All but
b t 7 V
V
off the
th
functional V repertoire of
the locus
Contains entire J region &
C gene


Contains
the
human Ig locus
complete
p
 30 functional V and 39
V pseudogenes
 100% of the V repertoire
 All 7 J and C pairs
The human light chain repertoire is 60% Ig and 40% Ig
18
Options for Antibody Generation
SLAM*
Immune Donors
Optimized
Hybridoma
XenoMouse®
*recombinant antibodies
19
XenoMouse® SLAM™ Technology – Examples
 3-8 weeks: B cell harvest to recombinant
antibody
 Identify rare antibodies with desired
specificities and characteristics
 Examples:
 Tumor specific binding
 High affinity neutralizing antibodies
20
Examples of XenoMouse® SLAM™ Programs
Program
Effort
#Binders
Affinity
IL-8
Classical Hybridoma
22
200 pM
XenoMouse SLAM
1063
610 fM*
Classical Hybridoma
1
750 pM
XenoMouse SLAM
134
63 pM
Classical Hybridoma
3
nM
XenoMouse SLAM
350
4 pM
Classical Hybridoma
277
nM
XenoMouse SLAM
>3000
9 pM
Classical Hybridoma
47
nM
XenoMouse SLAM
3800
3 pM
EGFRVIII
PTH
IL-13
TNF
* Rathanaswami P, Roalstad S, Roskos L. Su QJ. Lackie S, Babcook J. Demonstration of an
in vivo generated sub-picomolar affinity fully human monoclonal antibody to interleukin-8.
Biochem Biophys Res Commun. 334(4):1004-13 (Sep 9, 2005) .
21
The Modular Structure of Antibodies
Effector functions
•Phagocytosis
•ADCC
•CDC
Half-life
•FcRn
22
Successful Generation of Antibodies
Against
g
Diverse Target
g Types
yp
 Neutralizing Antibodies
–
–
–
–
Cytokines and Growth factors
Growth factor receptors
G-protein coupled receptors
Enzymes
 Agonistic antibodies
– TNF family receptors
– Growth factor receptors
 Highly specific antibodies
– Splice variants
– Heterodimeric proteins
23
Antibodyy Design
g Goals

Specificity
– Binding to target antigen or epitope of interest

Isotype
– IgG1 (antibody effector function), IgG2 or IgG4

Affinity
– Target-specific knowledge (affinity of ligand for receptor, expression level,
bioavailability etc).
bioavailability,
etc)
– Generally high affinity is required for infrequent dosing, long half-lives and low
concentrations

Cross-reactivity to orthologs
– Usually
U
ll mouse or cynomolgus
l
cross-reactivity
ti it required
i d ffor pre-clinical
li i l models
d l –
Surrogate Ab may be needed

Cross-reactivity to homologs
– Rarely needed, but important if highly conserved or cross-reactivity has toxicity
concerns

In vitro functional assays
– Target specific assays (Proliferation, R-L competition, etc)
– Non-target
g specific
p
assays
y ((ADCC,, CDC,, etc))

Toxicity assays
– Non-specific toxicity is not a significant concern, antibody toxicity is usually on target
24
Antibody Screening Cascade
Immunization of XenoMouse animals
Functional assays
Determine antigen-specific titers in sera
Fuse B cells isolated from hyperimmune animals
with myeloma to generate hybridomas or SLAM
Exhaust cultures
Binding Assay
Assa
R l i Affi
Relative
Affinity
i D
Determinations
i i
Tens to thousands of Ag-specific Abs
Run functional assay
on supernatants
Species
Cross reactivity
Cross-reactivity
 Optimized functional assays
relevant
l
t to
t target
t
t biology
bi l
 Goal is to cull Ab panel for further
characterization
Homolog
Cross reactivity
Cross-reactivity
 Soluble Antigen (equilibrium)
 Limiting antigen – Affinity
 High antigen – Quantitation
of antigen-specific antibody
 Cell based Antigen
 FACS measures relative
affinity
Sub-cloning
 Isolation of antigen-specific
hybridomas through limit dilution
 Purified antibodies are assessed
for function, affinity and potency
after sub-cloning
Relative affinity
d t
determination
i ti
Selection of panel of lead antibodies for sub-cloning
25
Anti-Cytokine
y
Antibodies – IL-13 as the Model
 Initially antibody panels against IL13 were screened for potential
neutralizing activity in:
– IL13Ra1 – IL13 competition assay
– B cell proliferation assay (> 10 nM IL13 required for bioactivity)
 Leads identified post-Ab purification were low affinity (5 – 10 nM)
and deemed not useful as therapeutic leads
 Desired antibodies with sub-100 pM affinity to drive potency and
improve likelihood of in vivo efficacy
Eotaxin-1
1 release assay on primary human dermal
 Developed an Eotaxin
fibroblast cells
– ~100 pM IL-13 as stimulus
– Cell
C ll culture
lt
media
di h
had
d no effect
ff t on th
the assay
26
Screening Overview – IL-13 Program
>3000 anti-IL13
ti IL13 antibodies
tib di
Eotaxin-1 release assay
>100 neutralizing antibodies
High antigen ELISA
Percen
nt Inhibition
Antigen-specific ELISA
High Antigen (relative Ab concentration in ng/mL)
Potency Selection
 A large panel of neutralizing antibodies were identified:
 Low concentration of IL13 favored identification of high affinity antibodies
 Relative antibody concentrations were determined with Ag-specific ELISA
 The highest potency antibodies were selected for further analysis (red circle)
27
Analysis of Purified Anti-IL-13 Antibodies
 Ab1 – Neutralizing antibody (5 nM)
from initial screening
g efforts
 Failed to neutralize in the
Eotaxin-1 assay due to low
concentration of IL13 in assay
 IL13Ra2 is a high affinity antagonist
 Screening approach with Eotaxin
assays lead to the identification of a
panel of lead antibodies
 Affinities of 3 to 60 pM
 Potency was limited by the
IL13 concentration
 Ab2 is one of the panel of
antibodies that met the affinity
and potency design goals.
28
Cell Culture and Automation Group
p
 Responsible for standard and high throughput cell
culture
 Very high volume - both 96 and 384 well format
 Currently aseptically processing ~8000 plates/month
 Extensive
E t
i use off specialized
i li d automated
t
t d systems
t
iis
required
 D
Developed
l
d and
d continue
ti
to
t optimize
ti i the
th hybridoma
h b id
process
 H
Harvest,
t purify
if and
d culture
lt
off B-cells
B ll for
f hybridoma
h b id
and SLAM
29
High Throughput Sterile Liquid Handling Robots
Air filter
2 carts
Tray based plate storage and transport
360 plate capacity (96 or 384 well format)
30
Molecular Biology
gy Group
p
 Antigen generation
 So
Soluble,
ub e, cell
ce su
surface,
ace, fusions
us o s ((HIS,
S, myc,
yc, TCE,
C , Fc,
c, GS
GST e
etc.)
c)
 Orthologue cloning
 Single cell PCR
 SLAM or hybridoma recombinant antibody generation
 V-gene cloning, class switching and vector shuttling
 Expression
E
i vectors
t
 Sequencing
 Genotyping
 Epitope Mapping and Binning
 High resolution Affinity measurements
 KinExA
31
Antibody Generation:
High
g Throughput
g p Recombinant Antibodyy Cloning
g
PCR amplify the V heavy and
light from 2-3 subclones/line
Sequence heavy and light chains
 Establishing clonality
 Identifying unique antibodies
Sequence analysis
Light
g Chain
FR1
FR2
CDR1
FR3
CDR2
CDR3
FR4
B3/Jk4
A20/Jk4
A27/Jk3
A30/Jk1
Potential N-glycosylation site
Heavy Chain
FR1
CDR1
FR2
FR3
CDR2
CDR3
FR4
VH4-34/D6-13/JH3B
VH1-2/D5-5/JH6B
VH3-30/D6-19/JH4B
VH3-23/D3-3/JH6B
Potential Asp isomerization sites
5 unique Abs
Belong to 4 different V
gene groups
No unpaired Cys
32
W in CDR3
Affinityy Consideration for Lead Antibodyy Selection
Correlation of Affinity and Potency
100
Dose (mg/kg
g/3 weeks)
10
1
0.1
3 pM Ag
30 pM Ag
0.01
300 pM Ag
3 nM Ag
0 001
0.001
0.1
1
10
100
1000
Affinity (pM)
 Theoretical Effect of mAb Affinity on Potency
•
The dose (mg/kg/3 weeks) of mAb required to suppress Ag levels in vivo by at
least 90% at steady-state was simulated as a function of mAb affinity (Kd) at four
baseline levels of endogenous Ag. When the Kd of the mAb is less than 1/10th
the baseline Ag levels, further improvements in affinity will not provide any
improvements in potency.
Rathanaswami P, Roalstad S, Roskos L. Su QJ. Lackie S, Babcook J. Demonstration of an
in vivo generated sub-picomolar affinity fully human monoclonal antibody to interleukin-8.
Biochem Biophys Res Commun. 334(4):1004-13 (2005)
33
Antibody Affinity Measurements
 Kd Measurements
 Soluble antigens
g
• KinExA
Measurements of very high (fM range) affinities
Partially
y active material*
Native, unpurified, unquantified material*
 On-cell membrane bound antigens*
• FACS
• KinExA
 Kon Measurements
M
t
 Soluble antigens - KinExA
*Difficult to measure with standard ELISAs and Biacore
34
KinExA Affinityy Measurements
Soluble Antigens
L
R
RLRL
L
LA R
L L L
L L
L
L
L L LL LL L
Equilibrium
Prepare
samples
Antigen
(Ag) Is
L
+
R
LRa
Titrated with
(L=Ag,
Constant
R=mAb) Amount
of Antibody (Ab)
3 nM Ag
2 nM Ag
1 nM Ag
0.1 nM Ab
0.1 nM Ab
0.1 nM Ab
Mixture Allowed
to Equilibrate
Measure Free
Ab
35% Free Ab
Sample
Cy5 20 Ab
Buffer
Seconds
Volts
[Free R]
35
< 0.5 sec
Bead Column

C t capturing
Coat
t i reagentt on the
th beads
b d

The short contact time between the
sample and solid phase kinetically
excludes the possibility of significant
dissociation of Ag*Ab
Ag Ab.

Only free Ab is bound to the beads.

The solid phase is used only as a
probe.

Bound Ab can be detected by
secondary reagents Cy5 labeled

The signal is directly proportional to
and linearly related to [free Ab] in the
sample
sample.
50% Free Ab
V
Volts
15 % Free Ab
L = Ag
R = Ab
KinExA – Calculation of Kd
Perform data analysis to produce a binding curve and
error graphs
% Free A
Ab
Kd
79.09pM
ABC
78.55fM
Signal 100%
0.69
NSB
0.22
Ratio
0.001
%E
%Error
21
2.1
ABC = Active Ab conc
NSB = Non-specific binding
[Ag0]
Kd
79.09pM
95% confidence interval
Kd High
91.56pM
Kd Low
54.74pM
ABC
78.55fM
95% confidence interval
ABC
C High Greater
G
than 21.74pM
ABC Low Less than 0.28fM
Kd
[Ab0]
36
Affinityy Measurement of Human Anti-hIL8 mAbs
 mAbs were generated using XenoMouse®
 Goal was to generate an Ab with better affinity than ABX-IL8 (Kd <200pM)
KinExA Kd
Anti-hIL-8
mAb
Kd (pM)
Kd Range (pM)
mAb
conc.
(pM)
33
280
150-420
10
142
400
190-680
25
203
190
64-340
25
215
360
230-450
50
469
870
640-1010
200
809
2.2
0.36-4.8
23
837
11
0.054-31
90
861
2.9
0.010-8.2
25
928
0.057
<0.010-1.8
20
1064
54
29-72
50
Rathanaswami et al. Biochem. Biophys. Res. Commun. 334:1004-1013, 2005
37
The Kd was determined using
an unpurified HEK293
supernatant
supernatant.
•Tight Kd ranges were
obtained for subnanomolar
affinity Abs
For the mAbs with affinities in
low pM – fM range, the 95%
confidence interval tends to be
broader
•mAb conc. used in the
equilibrium mixture is far
higher than Kd
•Warrants KinExA signal
optimization and multiple
curve analysis
Kd Measurements for the Highest Affinity antihIL-8 mAbs byy KinExA
% Free mAb
b
mAb 928
120
Curve 1: Kd controlled curve
100
•Kd = 870 fM (500fM – 1.3pM)
80
95% confidence interval is narrow when the curve is
generated by using an Ab concentration at or below Kd
60
Curve 2
Curve 1
40
20
0
1.00E-15
1.00E-12
1.00E-09
Concentration of Antigen (M)
The accuracy is further increased if one experiment is
done using the mAb at or near the Kd concentration and
another experiment is performed at a mAb concentration
10 fold higher than Kd and n-curve analysis is then done
N-curve analysis
Anti-hIL-8
mAb
Kd (pM)
Kd (pM) Range
mAb conc.
(pM)
809
3.3 ((2.2))
1.9-5.2 ((0.36 -4.8))
4.6,, 27 ((23))
837
16 (11)
9.3-25 (0.054 – 31)
18, 120 (90)
861
3.0 (2.9)
2.0-4.2 (0.010 – 1.8)
1.3, 13 (25)
928
0 61 (0.057)
0.61
(0 057) 0.38-0.94
0 38 0 94 (<0.010
( 0 010 – 1.8)
1 8)
0 68 2.0,
0.68,
2 0 14 (20)
Rathanaswami et al. Biochem. Biophys. Res. Commun. 334:1004-1013, 2005
38
Amgen British Columbia (ABC)
 Located in Burnaby, BC, Canada
 Wholly-owned subsidiary of Amgen
 56 full-time employees working in areas of antibody generation,
preclinical development, bioinformatics and robotics/automation
 Operations housed in a 56,000 sq. ft. “state-of-the-art” research facility
39
Acknowledgments
Kyla Currie
Ian Foltz
Margaret Karow
Chad King
Stephanie Masterman
Swami Rathanaswami
Stefan Sonderhoff
-.Anita Badh
Ken Dale
Sodi Kang
Nishreen Leila
Karen Richmond
-.Brad Hedberg
Kathy Manchulenko
Laura Sekirov
Laura Taylor
Kelly Berry
Brian Chan
Brandon Clavette
Vicki Fleming
Oscar Pan
Lisa Perkins
Stephanie Simmons
-.Lydia Baggott
Connie Chan
Desiree Lim
Angelica Moksa
-.Heather Sweet
Tina Wang
Ester Leng
Dawn Weishuhn
Taiyo Hara
40
Shawna Campbell
Katie Carr
Niki Cuthbert
Zoe Dunbar
Crystal Harris
Margo Mattinson
Erin McMillan
Andrea Munoz
Lauren Rose
Cathy Zeiler
-.Tom Boone
-.St. Jude Children's
Research Hospital:
Robert Webster
-.Biomedical Research
Centre, UBC
John Schrader
Kevin Leslie