Gene Therapy (2003) 10, 490–503
& 2003 Nature Publishing Group All rights reserved 0969-7128/03 $25.00
www.nature.com/gt
RESEARCH ARTICLE
Function of a genetically modified human liver cell line
that stores, processes and secretes insulin
BE Tuch1, B Szymanska2, M Yao1, MT Tabiin1, DJ Gross3, S Holman1, M Anne Swan4,
RKB Humphrey5,6, GM Marshall7 and AM Simpson2
1
Diabetes Transplant Unit, Prince of Wales Hospital and The University of New South Wales, Sydney, Australia; 2 Department of Cell
and Molecular Biology, University of Technology, Sydney; 3 Department of Endocrinology and Metabolism, Hadassah University
Hospital, Jerusalem, Israel; 4 Institute for Biomedical Research and Department of Anatomy and Histology, University of Sydney, Sydney,
Autralia; 5 School of Anatomy, The University of New South Wales, Sydney, Australia; 6 Department of Anatomical Pathology, Sydney,
Australia; and 7 Children’s Cancer Research Institute, Australia
An alternative approach to the treatment of type I diabetes is
the use of genetically altered neoplastic liver cells to
synthesize, store and secrete insulin. To try and achieve
this goal we modified a human liver cell line, HUH7, by
transfecting it with human insulin cDNA under the control of
the cytomegalovirus promoter. The HUH7-ins cells created
were able to synthesize insulin in a similar manner to that
which occurs in pancreatic b cells. They secreted insulin in a
regulated manner in response to glucose, calcium and
theophylline, the dose–response curve for glucose being
near-physiological. Perifusion studies showed that secretion
was rapid and tightly controlled. Removal of calcium resulted
in loss of glucose stimulation while addition of brefeldin A
resulted in a 30% diminution of effect, indicating that
constitutive release of insulin occurred to a small extent.
Insulin was stored in granules within the cytoplasm. When
transplanted into diabetic immunoincompetent mice, the cells
synthesized, processed, stored and secreted diarginyl insulin
in a rapid regulated manner in response to glucose.
Constitutive release of insulin also occurred and was greater
than regulated secretion. Blood glucose levels of the mice
were normalized but ultimately became subnormal due to
continued proliferation of cells. Examination of the HUH7-ins
cells as well as the parent cell line for b cell transcription
factors showed the presence of NeuroD but not PDX-1. PC1
and PC2 were also present in both cell types. Thus, the
parent HUH7 cell line possessed a number of endocrine
pancreatic features that reflect the common endodermal
ancestry of liver and pancreas, perhaps as a result of
ontogenetic regression of the neoplastic liver cell from which
the line was derived. Introduction of the insulin gene under
the control of the CMV promoter induced changes in these
cells to make them function to some extent like pancreatic b
cells. Our results support the view that neoplastic liver cells
can be induced to become substitute pancreatic b cells and
become a therapy for the treatment of type I diabetes.
Gene Therapy (2003) 10, 490–503. doi:10.1038/sj.gt.3301911
Keywords: human liver cells; insulin secretion; insulin storage; SCID mice; type I diabetes; gene therapy
Introduction
Treatment of type I diabetes is by two or more injections
of insulin daily. Despite this, it is often not possible to
achieve perfect normalization of blood glucose levels. To
obtain perfectly normal blood glucose levels attempts
have been made to replace the missing b cells by
transplantation of a whole pancreas, the islets from an
adult pancreas, or insulin-producing fetal pancreatic
cells. While transplantation of whole pancreases is of
benefit in reversing diabetes in humans, with 84% 1-year
graft survival,1 its success is limited mostly to those
people who have renal failure. Allografted adult human
islets are capable of reversing diabetes with results
improving in some2 but not all centers.3 Fetal pancreatic
Correspondence: AM Simpson PhD, Department of Cell & Molecular
Biology, University of Technology, Sydney, PO Box 123, Broadway, NSW,
2007 or BE Tuch MD, PhD, Diabetes Transplant Unit, Prince of Wales
Hospital, High Street, Randwick NSW 2031, Australia
Received 13 December 2001; accepted 22 August 2002
tissue is an alternative, the potential of which has not
been fully realized.4
An alternative approach to the treatment of type I
diabetes is the use of liver cells that have been genetically
altered to produce insulin.5–13 Results of a previous study
in our laboratory have shown that insertion of human
insulin cDNA (pC2) and the glucose transporter GLUT2
into the neoplastic human hepatoma cell line HEP G2
resulted in synthesis and storage of (pro)insulin in
structures resembling the secretory granules of normal
pancreatic b bcells.6 These HEP G2ins/g cells exhibited
regulated, near-physiological secretion of (pro)insulin in
response to glucose. In contrast, insertion of insulin into
two rat liver cell lines, FAO7 and H4-II-E8, resulted in
synthesis and release of insulin, but not its storage or
regulated secretion.
For an insulin-producing liver cell to be of maximal
benefit in vivo, it must be able to secrete a biologically
active form of insulin rapidly in response to an elevated
level of glucose. It is generally believed that this will
require storage and processing of proinsulin to insulin in
granules. Prior to this publication such granules have not
Insulin-producing liver cells
BE Tuch et al
been described in insulin-producing liver cells in vivo,
although processing has been induced.9,10 Despite the
absence of granules, blood glucose levels of diabetic
rodents in which insulin is produced in the liver can be
lowered to normal11 or near-normal9,10,12,13 levels. A
reduction of blood glucose levels, but not to normal, can
also be achieved by transplanting the rat liver cell line
H4-II-E.8 In all these rodents, even in those in which
blood glucose levels were normalized,11 secretion of
insulin in response to glucose did not occur in the rapid
manner seen from pancreatic b cells. In the study of Lee
et al,11 which employed a single-chain insulin analog,
there was very little change in blood levels of insulin
during the first hour after administration of glucose;
levels began to rise after this time and peaked at 4 h. In
normal rats, peak insulin levels occur 30 min after oral
glucose is given.
In our current study, we describe the creation of an
insulin-secreting human liver cell line by the transfection
of the insulin gene into the neoplastic HUH7 cell. These
new cells possess storage granules and cleave proinsulin
to the bioactive diarginyl insulin. They secrete the
peptide rapidly in a regulated manner in response to
glucose, but there is also some constitutive release of the
hormone. When transplanted into diabetic immunoincompetent mice, the cells lower glucose levels with
release of insulin being partly regulated, but mostly
constitutive.
Results
Gene transfer into hepatoma cells
HUH7 cells were stably transfected with the pRc CMV
vector carrying human insulin cDNA and 25 clones of
the HUH7-ins cells were isolated for further analysis.
The transfectants have remained stable for a period of
greater than 6 months in continuous culture. An analysis
of HUH7-ins cells showed the presence of both GLUT2
and glucokinase, a feature also of the untransfected cells.
GLUT 2 mRNA and protein were shown by Northern
analysis (Figure 1a) and Western blots (Figure 1b),
respectively, and glucokinase by RT/PCR (Figure 1c).
Secretion of insulin
The levels of human insulin secreted daily into the
culture supernatant from different clones of HUH7-ins
cells differed by 1.8-fold (0.1670.0 to 0.3070.10 pmol/
106 cells, n¼10). Levels of human proinsulin and its split
products, but not insulin, determined by a specific
radioimmunoassay (RIA) were 1071.2% of the total
insulin activity measured. C-peptide levels were
0.570.1% of the total insulin activity. HUH7 cells
transformed with the RcCMV vector alone did not
secrete insulin.
The effect of three well-known b cell stimuli, glucose,
theophylline and calcium, on the HUH7-ins cells (clone
2) was determined. A three-fold stimulation of insulin
secretion over basal levels (Figure 2a) was detected when
20 mM glucose was applied as a stimulus to HUH7-ins
cells for 1 h. As would be expected in regulated
secretion, on removal of the stimulus, insulin returned
to basal levels. This was confirmed by the perifusion
studies (Figure 3a). The same clone of HUH7-ins cells
showed a 46-fold response to 10 mM theophylline (Figure
491
Figure 1 Analysis of HUH7-ins cells for (a) GLUT2 mRNA, (b) GLUT2
protein and (c) glucokinase, hGLK. (a) Northern blot analysis for GLUT2
in HEP G2 — negative control (lane 1), HUH7 (lane 2), HUH7-ins, clone
2 (lane 3) and HUH7-ins cells, clone 9 (lane 4). The difference in
transcript levels of the two clones may be an artifact since clone 9 was run
(simultaneously as the other cells) on a separate gel. (b) Western blot
analysis for GLUT2 in HEP G2 — negative control (lane 1), HUH7 (lane
2), HUH7-ins clone 2 (lane 3) and HUH7-ins cells, clone 9 (lane 4).
(Lanes 1 and 2 were run simultaneously on a separate gel.) The molecular
weight of GLUT2 is 55 kDa. (c) RT/PCR expression of human glucokinase
(hGLK), 155 bp product, in HEP G2 (lane 1), HUH 7 (lane 2), HUH7-ins,
clone 2 (lane 3), HUH7-ins cells, clone 9 (lane 4), human liver (lane 5) and
human skin fibroblasts (lane 6).
2b) and an eight-fold response to 10 mM calcium over
basal levels (Figure 2c). Insulin secretion in response to
20 mM glucose plus 10 mM theophylline resulted in a
significant (Po0.01) increase in insulin secretion to
2.770.3 pmol insulin/106 cells, above that seen for
theophylline alone (2.2 7 0.3) (Figure 2d). Similar results
to these were seen for two other clones (9 and 13).
In response to varying concentrations of glucose from
0 to 20 mM, a dose–response curve for insulin secretion
was generated with clone 2 (Figure 3b). While glucose
responsiveness commenced at a slightly lower concentration than normal islets, a secretion curve approaching
normal physiological conditions was generated.
To assist in confirming that secretion of insulin from
the HUH7-ins cells was regulated, two extra sets of
experiments involving static stimulation were performed. Firstly, calcium was removed from the basal
medium before 20 mM glucose was added. This abolished the response of the cells to glucose (Figure 2e), just
as it did in MIN6 cells, which are normally glucose
responsive (Ref.14 and own results not shown). Secondly,
BFA, an agent that blocks movement of proteins in the
Golgi apparatus,15 and therefore constitutive release of
insulin, was added at the same time as 20 mM glucose.
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
492
b
**
20mM Gluc.
(n=4)
0.08
0.04
0.00
B1
***
(n=4)
2
1
0
B1
B2 Stim B3
d
0.4
Control
10mM Ca 2+
(n=4)
pmoles insulin/ 106 cells
0.5
***
0.3
0.2
0.1
0.0
3
Control
pmoles insulin/ 10 6 cells
Glucose w/o Ca2+
(n=4)
1.5
1.0
0.5
10
20
30
40
50
Fraction Number
b
0.25
(n=4)
**
1
B2 Stim B3
**
0.20
0.15
0.10
0.05
*
0.08
0.00
0
5
10
15
20
mM Glucose
0.04
0.00
B1
B2
Stim
B3
Figure 2 Secretion of insulin from HUH7-ins cells in response to (a)
20 mM glucose, (b) 10 mM theophylline, (c) 10 mM calcium, (d) 20 mM
glucose+10 mM theophylline, (e) 20 mM glucose in the absence of calcium,
or in the presence of 10 mg/ml BFA. Cells were incubated in basal medium
for two consecutive 1 h periods before being exposed to the stimulus for 1 h
(solid bars); cells in the control group were treated throughout with basal
medium (unfilled bars). n: number of experiments; B1: first basal period;
B2: second basal period; B3: basal period following stimulation; Stim:
stimulation period. Values are expressed as means+s.e. Significantly
different from B2, *Po0.05, **Po0.01, ***Po0.001.
Secretion of insulin from the HUH7-ins cells was
inhibited 30% by BFA (Figure 2e), whereas it completely
eliminated secretion of proinsulin from MRC5-VI cells
(human lung fibroblasts) transfected with the insulin
gene that release insulin constitutively (results not
shown). BFA had no effect on glucose-induced insulin
secretion from MIN6 cells, secretion being enhanced by
20 mM glucose six-fold in the presence or absence of BFA.
These data indicate that there is a small amount of
constitutive release of insulin from HUH7-ins cells in
vitro.
Insulin degradation
We examined whether the HUH7 liver cell line has the
ability, like primary hepatocytes,16 to degrade insulin.
HUH7 cells cultured in DMEM medium containing 10%
FCS metabolized 0.1870.04 pmol insulin/106 cells in 1 h
and 0.6170.09 and 1.0970.03 pmol insulin/106 cells in 6
and 24 h, respectively (n¼6). When the HUH7 cells were
maintained in Basal medium, the cells metabolized
0.1270.04 and 0.5870.05 pmol insulin/106 cells in 1
Gene Therapy
2.0
0
e
0.12
2.5
2
B1
Glucose
Glucose + BFA
Stimulus removed
3.0
(n=4)
B2 Stim B3
0.16
Stimulus applied
(20 mM Glucose)
3.5
0.0
***
Theo. + Gluc.
0
B1
4.0
Control
10mM Theo.
B2 Stim B3
c
pmoles insulin/ 10 6 cells
a
3
pmoles insulin / 10 6 cells
0.12
Control
Insulin Secretion
(relative to pre-stimulation levels)
0.16
pmoles insulin/ 10 6 cells
pmoles insulin/ 10 6 cells
a
Figure 3 Secretion of insulin from HUH7-ins cells, clone 2. (a) Dynamic
perifusion system. Data have been normalized to the average insulin
secretion measured in samples between 0 and 20 min (0.1170.001 pmol
insulin/108 cells), during which the cells were perfused with basal medium
without any stimuli. Data are representative of three independent
experiments. (b) Dose-response curve to varying concentrations of glucose,
1.5–20 mM. Values are expressed as means+s.e. for each point.
and 6 h, respectively (n¼6). By comparison the amount
of insulin degraded by NIH 3T3 cells (negative control)
was negligible. It has been known for some time that the
endosomal compartment of hepatocytes contains an
acidic endopeptidase or endosomal acidic insulinase,
which has recently been confirmed to be cathepsin D.17
The activity of cathepsin D is almost completely
inactivated by protease inhibitors, such as pepstatin
A.17,18 We incubated Huh7 cells over a 24-h period in
DMEM medium and a 6-h period in basal medium, in
the presence of pepstatin A, and confirmed that insulin
degradation was negligible in the presence of the
protease inhibitor.
Insulin content
The insulin content of the HUH7-ins clones, determined
in acid ethanol extracts, varied by 1.6-fold from 4.270.6
to 6.970.8 pmol/106 cells. Proinsulin was not detectable
in the cell extracts, as determined by an RIA specific for
proinsulin.6 Clones 2, 9 and 13, which were used for the
acute stimulation experiments described previously,
contained the largest amount of insulin, 6.970.8,
5.970.3 and 4.771.0 pmol/106 cells (n¼5), respectively.
The percent of insulin in each cell of these three clones
secreted each day was 3.6, 5.1 and 4.3, respectively.
Insulin-producing liver cells
BE Tuch et al
Figure 4 (a) Qualitative Western blot analysis for PC1, PC2 in MIN6
(lane 1), HUH7 (lane 2), untransplanted HUH7-ins (lane 3) and grafted
HUH7-ins cells (lane 4). ProPC1 has a molecular weight of 87 kDa, PC1
75 and the truncated carboxy-terminal form of PC1 62 kDa. ProPC2 has a
molecular weight of 75 kDa, and PC2 63 kDa. (b) Qualitative Western blot
analysis for the b cell transcription factors, NeuroD and PDX-1 in NIT-1
(lane 1), HUH7 (lane 2), untransplanted HUH7-ins (lane 3) and grafted
HUH7-ins cells (lane 4). The molecular weight of NeuroD is 55 kDa and
that of PDX-1 46 kDa.
Proinsulin converting enzymes
Western blot analysis of extracts from HUH7-ins and
HUH cells showed the presence of both PC1 and PC2,
enzymes in b cells which cleave proinsulin (Figure 4a).
Levels of these enzymes were upregulated by transfection with the vector containing insulin cDNA, but not
with an empty vector (data not shown). PC1 was
detected at a molecular weight of 75 kDa in the liver
cell lines, as compared to bands of size 87 and 62 kDa in
MIN6 cells, which correspond to proPC1 and truncated
carboxy terminal PC1, respectively. The different sizes of
PC1 may be because of the variation in glycosylation
related to different cell requirements for convertase
processing. Two molecular weight bands of PC2 were
detected in the liver cell lines, 73 and 50 kDa, as
compared to 73 and 63 kDa in MIN6 cells. The former
is proPC2 and the latter a cleaved version.
Histology and electron microscopy
Light microscopic analysis of the HUH7-ins cells showed
the presence of insulin in the cytoplasm (Figure 5a).
Electron microscopy (Figure 6a) showed the appearance
typical of a secretory cell, with rough endoplasmic
reticulum, Golgi complex, a few large vacuoles of 0.5–
4 mm diameter and secretory vesicles containing dense
granules. Large vacuoles were also present in another
human insulin-producing cell we have created, HEP
G2ins/g cells.6 No secretory vesicles containing granules
were observed in HUH7 cells, but there were a small
number of vacuoles. The secretory vesicles present in
HUH7-ins cells, at 170–230 nm diameter, were smaller
than the structures termed membrane sacs 300 nm in
diameter, found in normal b cells.19,20 The dense granules
in HUH7-ins cells were 85–100 nm in diameter, smaller
than in a normal b cell in which mature granules are
240 nm in diameter.19 As in normal b cells, the granules
in the HUH7-ins cells were surrounded by an electronlucent area. Many granules in HUH7-ins cells had a pale
area in the center of the dense granule (Figure 6b). In
MIN6 cells processed in the same way for electron
microscopy, there were both large secretory vesicles of
300 nm containing dense granules of 180 nm, and
smaller vesicles with granules of the same appearance
and dimensions as in the HUH7-ins cells (Figure 6c). In
both HUH7-ins and MIN6 cells, the central pale area was
presumably a feature of the granule in early stage
secretory vesicles. Figure 7a is an example of a secretory
vesicle in an HUH7-ins cell in which the dense granule
(diameter 100 nm) had a central pale area which was
only 5 nm in diameter. This is presumed to be a more
mature granule than those with larger central pale areas.
Immunoelectron microscopy (IEM) revealed that insulin was stored in the granules within the small
secretory vesicles of HUH7-ins cells (Figures 7b and c)
and in both the large and small secretory granules of
MIN6 cells (not shown). These granules were surrounded by the pale, electron-lucent region of the
secretory vesicle. The membrane was unstained as only
uranyl acetate without lead staining was used in these
preparations to avoid confusion with gold particles.
Processing for IEM did not alter the dimensions of the
secretory vesicles in either HUH7-ins or MIN6 cells. NIT
1 insulinoma cells previously gave similar results with
this immunogold technique.6 There was no non-specific
labeling in the nucleus or elsewhere in the cell. Control
HUH7 cells were negative as were HUH7-ins and MIN6
cells in the absence of the primary antibody.
493
(Pro)insulin synthesis
In response to varying concentrations of glucose, 0–
20 mM, a dose–response curve was generated for
(pro)insulin synthesis (Figure 8). Half-maximal synthesis
occurred at 5.4 mM glucose and the Vmax was
42.971.4 103 cpm/mg DNA (n¼6), not significantly
different from that for human islets.21,22 The ratio of
(pro)insulin synthesis to total protein synthesis was 0.065
at 2.5 mM glucose, 0.196 at 10 mM glucose and 0.213 at
20 mM glucose.
b cell transcription factors
To try and understand how it was possible for the
introduction of the insulin gene into the liver cell line to
induce formation of storage granules and regulated
secretion of insulin, both the untransfected and transfected cells were analyzed for the presence of b cell
transcription factors. Western blot analysis of extracts
showed the presence of the transcription factor NeuroD
but not PDX-1 (Figure 4b). Both factors were present in
the control b cell line.
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
494
Figure 5 Photomicrographs of (a) HUH7-ins cells (clone 2), (b) HUH7 cells (negative control), (c) HUH7-ins grafted beneath the renal capsule and (d) the
nearby kidney (negative control). Immunoperoxidase for insulin. Positively stained cells are either red (a) or brown (c) (magnification 348).
Transplantation into diabetic SCID mice
Transplanted cells coalesced into a reddish friable mass
at the site of grafting and grew rapidly, both beneath the
renal capsule and subcutaneously. Subcutaneous grafts
were visible 1–2 weeks after being transplanted. The
blood glucose levels of all 18 diabetic mice that were
transplanted became normal, 5.570.3 mM, in 5–26 days
after the cells were implanted. Animals receiving cells
beneath the renal capsule became normoglycemic faster
than those grafted subcutaneously, median of 12 days
(range 5–12) versus 18 days (range 12–26 days) (Figure
9a). Whether cells were injected or transplanted in a
plasma clot did not alter the outcome (data not shown).
Blood glucose levels continued to decline reaching
subnormal levels, 1.770.2 mM, a median of 3 days
(range 1–13 days) after blood glucose levels were
normalized. Removal of the grafts at this time resulted
in a prompt increase in the blood glucose levels to
hyperglycemic values, 16.773.5 mM (Figure 9b).
Intraperitoneal glucose tolerance test
The fasting blood glucose levels of transplanted mice,
which had visible subcutaneous grafts, were lower than
those of the controls, for example, in the fasting state,
2.670.4 mM versus 4.970.5 mM, Po0.001 (Figure 10).
Glucose levels peaked at 20 min in these grafted mice
with human C-peptide detectable in the blood of these
mice and rising from a fasting level of 0.1370.05 to
0.2970.08 nM (n¼4 in each group) at 20 min. Glucose
levels peaked at 20–40 min in the control mice, in the
blood of which human C-peptide was undetectable.
Analysis of grafts
Histological analysis of the grafts showed the presence of
insulin in the cytoplasm of cells (Figure 5c). Electron
Gene Therapy
microscopic analysis showed the same features observed
in the untransplanted cells, namely granules containing
insulin (Figure 7d). These granules were of size 95–
143 nm, similar to those in untransplanted cells. HPLC
analysis of graft extracts revealed that they contained
insulin-related peptides, the main one being human
carboxy extended diarginyl insulin (Figure 11). There
were two minor peaks detectable in the insulin RIA
corresponding to split31,32 human proinsulin and intact
human proinsulin. Only a very small peak was seen to
elute at the position of mature human insulin. Western
blot analysis of protein extracted showed the presence of
PC1 and PC2 (Figure 4a). PC1 and PC2 were in two
forms, both as the pro-enzyme and as a cleaved product.
Examination of the graft for b cell transcription factors
gave the same results as that observed in the untransplanted cells, namely the presence of NeuroD, although
in lower amounts, but not PDX-1 (Figure 4b). The
reduction of expression of NeuroD in grafted HUH7ins cells also was observed in HUH-7 cells transplanted
into immunoincompetent mice (data not shown).
Insulin content and secretion
The total insulin content of grafts removed from the mice
was 1.570.3 pmol/100 mg, which is less than the insulin
content of the pancreas of control immunoincompetent
mice, 83715 pmol/mg. This suggests that regulated
insulin secretion in these grafts was small, with lowering
of blood glucose levels in the animals being largely the
result of constitutive release of insulin. This functional
difference between the cells in vitro and in vivo may be
caused by the effects of transplantation, such as lower
levels of oxygen,23 which are critical for insulin secretion.
When the cells in the grafts were removed from the mice
and cultured in vitro, the insulin content was similar to
that of untransplanted cells, 5.1–5.6 pmol/106 cells. The
Insulin-producing liver cells
BE Tuch et al
amount of hormone secreted per day in vitro varied from
0.28 to 0.34 pmol/106 cells, that is, 5.5–6.1% of the insulin
content. When cells isolated from the grafts were
challenged with 20 mM glucose, insulin secretion was
enhanced 1.9 to 2.6-fold, similar to that in untransplanted
cells (Figure 2a). These data indicate that the transplanted HUH7-ins cells retained their ability to secrete
insulin in a regulated manner.
495
Discussion
Figure 6 Transmission electron micrograph of (a) an HUH7-ins cell
showing secretory vesicles with dense granules (g) surrounded by a pale
halo, large vacuoles (v), rough endoplasmic reticulum and nucleus (n),
bar¼1 mm, (b) an HUH7-ins cell showing a cluster of secretory vesicles, of
approximately 200 nm diameter, containing secretory granules (g) of
approximately 100 nm diameter surrounded by a pale halo. These granules
contain central pale areas ranging from 15 to 50 nm diameter,
bar¼200 nm. (c) Two MIN6 cells. In the cell on the left of the micrograph
is a secretory vesicle containing a dense granule (g) with a central pale
area. The granule is surrounded by a pale halo. This secretory vesicle and
the granule are of the same appearance and dimension as those in the
HUH7-ins cells. In the cell on the right, secretory vesicles of diameter
300 nm and granules of 180 nm diameter are seen. Each granule is
surrounded by a pale halo, bar¼250 nm.
We have genetically modified the neoplastic human liver
cell line, HUH7, to synthesize, process, store and secrete
a bioactive form of insulin in a regulated manner to the
physiological stimulus glucose. These cells are not
normal b cells since they also possess the capacity to
release insulin in a constitutive manner. When the cells
are transplanted into immunodeficient diabetic mice,
blood glucose levels are lowered to normal and subsequently subnormal levels with secretion of insulin
continuing to be partly regulated by glucose, but mostly
constitutive. Our experience with the HUH7-ins cell line
extends our results with another human liver cell line,
HEP G2ins/g, which also synthesized, stored and
secreted insulin in a regulated manner when exposed
to glucose in tissue culture.6 The ability of the transplanted HEP G2ins/g cells to secrete (pro)insulin in
response to glucose was impaired, with the cells having
no immediate effect on the blood glucose levels of the
diabetic recipient mice.24 This was different from the
results we obtained with transplanted HUH7-ins cells
(Figure 10). Our results with the HUH7-ins cells support
the conclusion of other researchers,7–13 namely that liver
cells, or at least their precursors,9 can be genetically
altered to act as b cells, and replace those which are lost
in type I diabetes. What is novel about the data reported
in this paper is that secretion of insulin from liver cells
that possess a b cell transcription factor in vivo is
regulated by glucose, albeit to a small extent. Most
genetically altered liver cells that synthesize (pro)insulin
do not secrete the peptide in response to glucose.7–10,12,13
The two exceptions are HEP G2ins/g6 and rodents
infected with an adenovirus containing the cDNA for
pyruvate kinase promoter and a single-chain insulin
analogue.11 The HEP G2ins/g cells do secrete (pro)insulin rapidly in response to glucose in vitro but the cells are
unstable when transplanted.24 Secretion of the singlechain insulin analog from rats does occur when glucose
is administered, but is delayed with the peak effect at 4
hours, presumably because control is at a translational
level.
Rapid regulated secretion of (pro)insulin requires prior
storage of this peptide in cytoplasmic granules. Such
storage is a feature of HUH7-ins, HEP G2ins5 and HEP
G2ins/g cells.6 This has been demonstrated by immunogold electron microscopy, immunohistochemistry and
extraction of cell contents with acid ethanol. To the best
of our knowledge, no other insulin-producing liver cell
has yet been described as having this feature. In some,
insulin can be observed immunohistochemically,8,9,11,13
but no supporting proof of storage has been provided.
The granules in the HUH7-ins cells were up to 100 nm in
size, smaller than those observed in HEP G2ins/g cells,6
pancreatic b cell lines,6 and primary b cells19 but similar
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
496
Figure 7 Transmission electron micrograph (a) of a higher magnification of an HUH7-ins cell showing a secretory vesicle of 220 nm diameter containing a
secretion granule (g) of 100 nm diameter with a central pale area of approximately 5 nm, bar¼200 nm. Immunoelectron micrograph (b) of a secretory vesicle
and granule (g) in an HUH7-ins cell, labeled with 10 nm gold particles to show the location of insulin. The gold label is seen only over the secretion granule,
which is surrounded by a pale halo. The membrane of the secretory vesicle is unstained and the granule is pale, as only uranyl acetate staining was
performed, bar¼40 nm. Immunoelectron micrograph (c) of a lower magnification of an HUH7-ins cell with 10 nm gold particles marking the location of
insulin, bar¼200 nm. Transmission electron micrograph (d) of HUH7-ins cells transplanted into SCID mice, showing many granules of similar appearance
to that in untransplanted material, Bar¼1 mm.
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
Figure 8 (Pro)insulin synthesis of Huh7-ins cells in response to varying
concentrations of glucose, 0–20 mM. Values are expressed as means7s.e.
n¼4 for each point.
to small secretory granules in MIN6 cells. A feature
common to all these cells was a halo between the
electron-dense center of the granule and its limiting
membrane. The roundness of the electron-dense region
inside the granules of the HUH7-ins cells, as compared
to the crystalloid appearance of adult human b cells, is
probably a reflection of the storage of diarginyl insulin in
the liver cells as compared to mature insulin in the
pancreatic cells. The number of storage granules we
observed in the HUH7-ins cells was fewer than those
seen in pancreatic b cells. This is consistent with there
being some constitutive release of insulin as well as
degradation of insulin by the liver cells.16,17 Neither of
these features occurs in a pancreatic b cell, but insulin
degradation does occur in a normal liver cell. Constitu-
tive release is not a feature of a pancreatic b cell, but a
small amount of insulin is degraded in these cells, within
the secretory granules. Insulin degradation is a prominent feature of a normal liver cell.25
What induced the formation of storage granules in the
HUH7-ins cells, and for that matter, the transfected HEP
G2 cells, is unknown. Such information is highly relevant
in understanding how to make a substitute b cell since it
is the storage of insulin in such granules that allows the b
cell to secrete insulin so rapidly when needed. This
feature is a hallmark of all endocrine cells. A clue to the
answer lies in the analysis of the cell lines for b cell
transcription factors. Both HUH7-ins cells and the parent
cell line HUH7 contain NeuroD, a transcription factor
essential for the formation of pancreatic b cells.26 HEP
G2ins/g cells also possess this factor (data not shown).
This agent is not normally found in liver cells, being
restricted to neural and pancreatic tissues. This result
indicates that the human liver cell line we have used in
the current study, HUH7, already had at least one feature
characteristic of pancreatic b cells before it was transfected. This is not unexpected since pancreas and liver
are both derived from endodermal cells, and neoplastic
cells are known to regress ontogenetically when they
form.27 The absence of PDX-1, although surprising, can
be explained since not all b cells need to express this
transcription factor for glucose-stimulated secretion of
insulin.28 The introduction of the insulin gene under the
control of the CMV promoter appears to be responsible
for activating the gene(s) required for the formation of
storage granules in this ‘pancreatic’ liver cell. The
combination of these conditions is probably required
for the creation of granules. These structures form
neither in other cell lines transfected with the insulin
gene,7 nor in HUH7 cells in which the cDNA for insulin
is introduced under the control of an inducible promoter,
sheep metallothionein.29 The plasticity of adult cells has
been demonstrated on many occasions, for example, bile
duct cells in the liver can be induced to form hepatocytes
following bile duct ligation and treatment with chemicals
such as furan or carbon tetrachloride.30,31 Furthermore,
497
Figure 9 Blood glucose levels of mice transplanted with HUH7-ins cells. Data are presented in all mice (a) or in selected mice (b). The cells were placed
either beneath the renal capsule (n¼8) or subcutaneously (n¼10). Grafts placed beneath the renal capsule functioned more efficiently than those implanted
subcutaneously. In all cases, the elevated blood glucose values first became normal and then subnormal. Removal of (R/O) the graft resulted in an immediate
rise in the blood glucose levels (b).
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
498
Figure 10 Intraperitoneal glucose tolerance test in mice transplanted with
HUH7-ins cells: Pp0.004; **transplanted mice odiabetic mice and nondiabetic controls; *transplanted mice and non-diabetic controls odiabetic
mice.
250
1
2
3
4
5
200
IRI pmol/l
150
100
50
0
0
10
20
30
40
50
60
70
80
90
Fraction #
Figure 11 HPLC analysis of acid ethanol extracts of grafted HUH7-ins
cells. Arrows designate the elution positions of: (1) intact human insulin;
(2) carboxy extended diarginyl insulin; (3, 4) split human proinsulin; (5)
intact human proinsulin. The profile is representative of eight qualitatively
similar experiments.
hepatocytes arise in the pancreas under conditions of
copper depletion/repletion,32 D-galactosamine treatment33 or repeated injections of cadmium.34 This
plasticity may play an important role in the ability of
the liver to respond to abnormal mitogenic activity or to
a pathological insult caused by external factors and when
fully understood could possibly be harnessed in future
genetic manipulation studies as a cure for disease.
The HUH7-ins cells cleaved proinsulin to diarginyl
insulin, a process that requires the converting enzymes
Gene Therapy
PC1 and PC2, which were present in the HUH7-ins cells.
They were also present in the parent cell line, a reflection
of the pancreatic nature of this human liver cell line.
Liver cells do not normally possess these enzymes, but
they can be induced by the introduction of the b cell
transcription factor PDX-1.9 Diarginyl insulin was capable of lowering blood glucose levels in the transplanted
mice since it is equal in biological potency to insulin.35
Secretion of insulin from the HUH7-ins cells was
partially regulated, presumably being derived from the
granules. This was shown in vitro in response to glucose,
calcium and a stimulus of cyclic AMP, and in vivo to
glucose. Secretion was tightly controlled with secretion
rapidly turned on when glucose was added, and rapidly
turned off when the stimulus was removed. Further
support of regulated secretion was shown by the loss of
glucose stimulation in the absence of calcium. In a
pancreatic b cell it is the influx of extracellular calcium
through voltage-activated calcium channels on the cell
surface that initiates movement of insulin granules to the
cell membrane and release of their contents. The step
prior to opening of such channels in a b cell is
depolarization of the cell membrane caused by closure
of ATP-dependent potassium channels. It is possible that
these channels are present in HUH7-ins cells since they
are present in HEP G2ins/g cells (unpublished data).
Combining all of this information leads us to hypothesize that glucose causes secretion of insulin from HUH7ins cells by the classical pathway described in b cells.
Metabolism of glucose results in an increase in the ratio
of ATP:ADP, closure of the ATP-dependent potassium
channels on the cell membrane, opening of the voltageactivate calcium channels, and then secretion of insulin.
Brefeldin A (BFA) is an agent that blocks passaging of
proinsulin from the trans-Golgi network to granules, but
has no effect on processing of proinsulin in, or secretion
of insulin from the granules.15 Addition of this agent
partially blocked secretion of insulin in response to
glucose, whereas it had no effect on glucose-induced
insulin secretion from the pancreatic b cell line MIN6.
These results imply that at least some of the insulin
released, 30% (Figure 2e), was constitutive. In vivo the
situation was different. While there was some glucoseregulated secretion of insulin, most of the insulin
released was constitutive. This is concluded from the
low insulin content of the grafts and the occurrence of
hypoglycemia in recipient mice. The glucose doseresponse curve for insulin secretion from HUH7-ins cells
was near-physiological, and similar to that obtained with
the HEP G2ins/g cell line.6 Enhancement of insulin
secretion commenced at 2.5 mM glucose, not quite the
same as that in pancreatic b cells, 4–5 mM,36 but much
less sensitive than a pituitary cell line transfected with
the insulin gene, 10 mMM.37 It is possible that modulation
of the expression of GLUT2 in the HUH7-ins cells will
shift the dose-response curve to the right, to become
physiological. Expression of GLUT2 in insulin-producing
cell lines, for example, HEP G2ins cells, is responsible for
the induction of glucose responsiveness.6 In contrast to
the dose-response curve for insulin secretion, that for
insulin synthesis was exactly the same as that which
occurs in pancreatic b cells.21,22 This implies that different
metabolites of glucose are responsible for the secretion of
insulin and translation of the insulin message from the
nucleus.
Insulin-producing liver cells
BE Tuch et al
In summary, we have described the creation of a
primitive insulin-secreting human liver cell line, HUH7ins. The cells possess storage granules, cleave proinsulin
to diarginyl insulin, and secrete this rapidly in a
regulated manner in response to glucose and other b
cell stimuli. There is also constitutive release of the
hormone. When transplanted into diabetic mice, the cells
retain their properties of storage, processing and
regulated secretion of insulin, and normalize blood
glucose levels. However, constitutive release of insulin
is greater than its regulated secretion, and this together
with unrestricted proliferation of cells resulted in
hypoglycemia. A number of changes will be required
before cells of this type could potentially be used
clinically. These include a decrease of sensitivity to
glucose, reduction in constitutive release of insulin,
removal of the oncogenic potential, introduction of a
suicide gene and encapsulation of the cells to prevent
their rejection. Regardless, our results offer hope that
primitive liver cells can be induced to become substitute
pancreatic b cells and be used as a therapy for the
treatment of type I diabetes.
Materials and methods
Plasmid construction
The full-length 0.6-kb human insulin cDNA, pC2 was
cloned into the multi-cloning site of pRcCMV (cytomegalovirus promoter) (Invitrogen, San Diego, CA, USA),
as previously described.6 The pRcCMV vector expresses
resistance to the eucaryocidal antibiotic neomycin/G418.
HUH7 cells were transfected with 20-mg of the recombinant plasmid and vector. Transfection was accomplished
by electroporation at 400 V and 500-mF in a Biorad
(Biorad, Hercules, CA, USA) gene pulser at a cell
concentration of 5 106 cells/cuvette, in Dulbecco’s
modification of Eagles’s medium (DMEM) (Trace Biosciences, Melbourne, Australia) containing no fetal calf
serum (FCS). To obtain stable transfectants of HUH7 cells
containing insulin cDNA, 48 h later 0.55 mg/ml of the
G418 antibiotic (Gibco Laboratories, Grand Island, NY,
USA) was added to the culture medium. Medium plus
drugs were changed every 2–3 days. After 3–4 weeks of
selection, colonies were picked and screened for production of insulin by RIA, Selected clones were expanded
into mass cultures and maintained in G418-containing
media throughout the course of all experiments.
Cell culture
HUH7 and HUH7-ins cells were grown as monolayers in
DMEM supplemented with 10% FCS (Trace Biosciences)
in 5% CO2 in air at 371C. For the latter cell line, the
antibiotic G418 0.55 mg/ml was also added. MIN6 cells
were grown in DMEM supplemented with 15% FCS.
NIH3T3 cells were grown in DMEM with 10% FCS.
MRC5-VI cells transfected with the insulin gene were
grown in DMEM, supplemented with 10% FCS and
1 mg/ml G418.38
Extraction of peptides
Insulin was extracted from liver cells with 0.18 N HCl in
70% ethanol for 18 h at 41C. For measurement of insulin
content, samples were diluted before being placed in the
RIA. For high-performance liquid chromatographic
analysis (HPLC), the ethanol was evaporated, the
samples lyophylized and then reconstituted in 50 ml of
0.1% trifluoroacetic acid/0.1% bovine serum albumin
(BSA).
499
Insulin secretion
This was carried out in two ways: (a) static stimulation
and (b) perifusion.
Static stimulation39: Before stimulation, tissue culture
plates were thoroughly washed with basal medium
(Phosphte-buffered saline (PBS) containing 1 mM CaCl2
and supplemented with 20 mM HEPES and 2 mg/ml
BSA, 1.0 mM glucose (unless otherwise stated)) to
remove culture medium and FCS. Monolayers were
incubated in the basal medium at pH 7.4 for two
consecutive periods of 1 h to stabilize the basal secretion
of insulin. Monolayers were then exposed to stimuli for
1 h. Glucose, 1.25–20 mM, and theophylline 10 mM
(Sigma, St Louis, MI, USA) were dissolved in basal
medium. When BFA (Sigma) was added, it was present
at 10 mg/ml (35 mM) and was diluted from a 10 mg/ml
(35 mM) stock made up in absolute ethanol and stored at
201C. Calcium (10 mM) was dissolved in basal medium
without phosphate. Three controls were used – basal
medium alone, basal medium without phosphate and
basal medium with ethanol. The static stimulation
performed in the absence of calcium had 0.15 mM EGTA
in the basal medium to chelate calcium from the cell.
Perifusion: Cells were grown on cytodex 3 beads and
cultured in spinner flasks in normal culture medium for
4 days.6 Beads with cells (approximately 108) attached
were transferred to a column, connected to a peristaltic
pump and fraction collector, and basal medium was
pumped through the system for 40 min. This was
followed by a 30 min stimulus with glucose and a
further 30 min basal period. Samples were collected
every 2 min and assayed for insulin.
Insulin degradation
Huh7 and NIH3T3 cells (negative control) were seeded
at 106 cells/well in DMEM, 10% FCS and incubated
overnight. The medium was then replaced with DMEM
or basal medium supplemented with 0.6 pmol/ml
insulin and the cells were incubated for 1, 6 or 24 h
(cells in basal medium for 1 and 6 h only). Prior to
collection, the culture supernatant was acidified to pH
4.0 with acetic acid for 5 min. The pH of the collected
supernatants was neutralized with NaOH before the
samples were placed in the RIA. Insulin degradation was
also measured in Huh7 cells following 24 h incubation
with pepstatin A (Sigma) (0.1–5 mg/ml)17 in DMEM
medium and a 6 h incubation in basal medium.
Insulin synthesis
After trypsinization, 106 HUH7-ins cells were incubated
at 371C (5% CO2/95% O2) in vials containing 2 ml KrebsRinger bicarbonate buffer supplemented with 10 mM
HEPES, 2 mg/ml BSA (KRB/BSA), 50 mCi/ml L-[4,
5-3H] leucine and 2.5–20 mM glucose. After 2 h, the cells
were washed with KRB–BSA containing 0.05% nonradioactive leucine and disrupted by sonication in 200 ml
distilled water. The amount of labeled insulin was
determined by an immunoprecipitation technique as
previously described.40
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
500
RNA preparation and Northern (RNA) blot
Total cellular RNA was isolated by the guanine thiocyanate method.41 Northern blot analysis was carried out as
previously described.6 Blots were hybridized with 32Plabeled human GLUT 2.
Analysis of gene expression by RT/PCR
RNA (1 mg) was reverse transcribed using Moloney
murine leukemia virus reverse transcriptase and random
hexaneucleotide primers42 (Perkin/Elmer, Norwalk, IN,
USA). The PCR reaction was performed using a cDNA
amount representing 100 ng total RNA and 1 unit of
Amplitaq polymerase (Perkin/Elmer) in a reaction mix
containing 10 mM Tris–HCl (pH 8.3), 50 mM KCl, 1.5 mM
MgCl2, 0.01% (w/v) gelatin, 250 mM dNTP’s and 20 pmol
of each primer. The PCR was performed in a Corbett
Thermal Sequencer: following an initial denaturation
step of 2 min at 941C, a 35 cycle program consisting of
941C for 45 s, 611C for 1 min and 721C for 45 sec was
performed. This was followed by a final cycle at 721C for
5 min, prior to cooling to 41C. Following PCR, a 10 ml
aliquot was subjected to 12% polyacrylamide gel
electrophoresis, prior to ethidium bromide staining and
photography. The gene specific human glucokinase (gift
from M. Permutt, St Louis, Missouri, USA) forward PCR
primer was 50 CTACTTGGAACAAATCGCCC, and the
reverse PCR primer was 30 CCTCCTCCTGGACTTCTTCC
(primers synthesized by Gibco Laboratories). This
primer pair generated a 155 bp PCR product from bp
244–379 of the published human glucokinase sequence.43
Several negative control reactions were included in each
experiment. Some of the negative controls contained
water instead of cDNA, and other control reaction
mixtures were prepared without the addition of RNA.
All necessary precautions against contamination of PCRs
were rigorously observed.
Source of mice
Male, inbred SCID mice were obtained from the Walter
and Eliza Hall Institute of Medical Research in Melbourne, Victoria, Australia. They were housed in
sterilized polycarbonate cages covered with autoclavable
filter tops, six to a cage, and had unlimited access to
sterilized tap water and gamma-irradiated mouse food.
The temperature of the room in which they were housed
was 171–241C and the relative humidity 53–76%.
Induction of diabetes
Diabetes was induced in the mice by injection of multilow-dose streptozotocin 75 mg/kg/day for five consecutive days.44 This technique was used in preference to a
single dose of streptozotocin because of lower morbidity.
Blood glucose levels were measured on 5 ml blood
obtained from a tail vein with the blood being placed
on glucose reagent strips (Bayer Diagnostic, Mulgrave,
Victoria, Australia) and the blood glucose level read in a
glucometer (Bayer Diagnostics). The blood glucose levels
of the 18 mice transplanted were 21.471.2 mM.
Transplantation of HUH7-ins cells
HUH7-ins cells were trypsinized from culture flasks,
their concentration determined with a hemocytometer,
and aliquots of 1 107 cells were transplanted into the
mice either by direct injection or in a plasma clot.
Gene Therapy
For the direct injection, the HUH7-ins cells were
initially aspirated into polyethylene tubing (internal
diameter 0.48 mm; external diameter 0.96 mm) attached
to a 23-gauge needle placed on a Hamilton syringe. The
end of the tubing was then inserted either beneath the
left renal capsule of the mice45 or into the subcutaneous
tissues above the right scapula before the plunger of the
syringe was rotated to allow injection of the cells into the
mice. Six mice were transplanted in this way, three
beneath the renal capsule and three subcutaneously. In a
further five diabetic mice the cells were injected
subcutaneously using a 1 ml syringe attached to a 23gauge needle without the polyethylene tubing.
A further seven mice were subsequently transplanted
with cells placed in a plasma clot,46 five having grafts
placed beneath the renal capsule and two subcutaneously.
Blood glucose levels and weight of the mice were
monitored every 1–3 days. Insulin was not administered
to the mice at any time, as is sometimes done to reduce
mortality and morbidity. Grafts were removed from the
mice when the blood glucose levels became low. Blood
glucose levels of the mice were monitored thereafter, an
increase expected if the graft had been responsible for
reversal of the diabetes.
Intraperitoneal glucose tolerance test
An intraperitoneal glucose tolerance test was performed
in mice transplanted with HUH7-ins cells, both before
and after blood glucose levels became normal. This was
carried out on mice fasted overnight. A total of 8.3 mmol/
g body weight of a 1.67 M glucose solution was injected
intraperitoneally, blood glucose levels measured at 0, 10,
20, 40, 60, 90 and 120 min thereafter,47 and human Cpeptide levels at 0 and 20 min.
RIA
Levels of human proinsulin (hPI) were measured by an
RIA specific for this peptide and its split products as
described previously.6,48 Iodinated hPI was prepared by
the chloramine-T method. Specificity was established by
showing o0.05% cross-reactivity with insulin. The
insulin RIA was carried out using guinea pig insulin
antibody (Wellcome, England) and 125I-labeled insulin
prepared by the chloramine-T method (gift from Dr Paul
Williams, University of Sydney). Cross-reactivity with
hPI was 73%. Human insulin standards were used to
assay the product from HUH7-ins cells and rat insulin
standards for MIN6 cells.
A commercial RIA (Eurodiagnostica, Malmö, Sweden)
was used to measure levels of human C-peptide. The
antibody used in the assay does not detect human insulin
but there is 41% cross reactivity with human proinsulin.
Membrane preparation, gel electrophoresis and
Western blotting
Cell membranes were prepared from HEP G2, HUH7,
HUH7-ins and MIN6 cells by suspending cells in ice-cold
buffer (20 mM K2HPO4, 1 mM EDTA and 110 mM KCl)
containing protease inhibitors, and sonicating cells on
ice. Supernatants were prepared by centrifugation at
16 000 g in a refrigerated microfuge and the protein
concentration determined by the method of Bradford.49
A total of 5 mg of protein was suspended in an equal
Insulin-producing liver cells
BE Tuch et al
volume of 2 sample buffer (100 mM Tris, 4% SDS, 0.2%
bromophenol blue, 20% glycerol, 10% b-mercaptoethanol, pH 6.80), heated at 951C for 5 min, and electrophoresed on a 6.5% polyacylamide gel. Proteins were
transferred to PVDF membranes (Bio Rad, Hercules,
CA, USA) and blocked with 4–5% dry milk in Trisbuffered saline with Tween (TBST: 10mM Tris, 150 mM
NaCl, 0.05% Tween 20, pH 8.0). The blot was incubated
overnight at 41C with primary antibody in TBST plus 1%
skim milk, then washed for 1 h in TBST, and bands
visualized using the ECL chemiluminescent Western
blotting kit (Amersham International, Amersham, Bucks,
UK). The primary antibodies used were to human GLUT
2 1:500 (gift from M Permutt, St Louis, MI, USA), PC1
1:1000 (gift from I Lindberg, New Orleans, LA, USA),
PC2 1:1000 (gift from C Rhodes, Seattle, WA, USA),
carboxypetidase E 1:1000 (gift from L Fricker, New York,
USA), PDX-1 (gift from H Edlund, Umea, Sweden) and
NeuroD (Santa Cruz Biotehcnology Inc., Santa Cruz, CA,
USA).
Immunohistochemical analysis
HUH7 and HUH7-ins cells were collected after trypsinization, or from transplanted mice, and fixed in 10%
formalin at 41C for 2 h. They were then washed in PBS
and loaded onto slides coated with poly-L-lysine (Sigma).
The endogenous peroxidase was inactivated by treatment of cells with 3% hydrogen peroxide. Prior to
antibody labeling, 10% goat serum was applied to the
samples to prevent non-specific antibody labeling. The
cells were incubated with the primary antibody (guinea
pig polyclonal insulin antibody (1:1000)) overnight at
41C. After washing in PBS the cells were allowed to react
with the secondary antibody (guinea pig, anti-rabbit
immunoglobulin (1:500)) for 30 min at room temperature. This was followed by sequential labeling with
biotinylated anti-rabbit immunoglobulin and streptavidin–peroxidase conjugate. Coloration was produced
with substrate solution containing hydrogen peroxidase
and 3-amino-9-ethylcarbazole (red) or diamino-benzidine (brown). For the negative control, primary antibody
was omitted by replacing the antibody with PBS. Pig
pancreas was immunohistochemically stained to serve as
the positive control. After immunostaining the nuclei
were counterstained with Mayer’s hematoxylin (Sigma).
All immunoglobulins and AEC substrate were purchased from Dako (Santa Barbara, CA, USA).
Electron microscopy
After trypsinization, single cell suspensions were allowed to settle between changes of solutions as follows:
control HUH7 cells and HUH7-ins cells were fixed for
2 h at room temperature in 3% glutaraldehyde and 1.5%
formalin with 30 mM betaine in 0.17 cacodylate buffer,
pH 7.4, osmolarity 1200 mOsM, followed by 1.5%
osmium tetroxide in 1 M cacodylate buffer at room
temperature for 30 minutes. Subsequent processing was
as detailed by Swan.50
Engrafted cells: 1 mm3 fragments of grafts from the
transplanted mice were fixed in 2% glutaraldehyde and
1% paraformaldehyde for 1 h at 41C in 0.1 M cacodylate
buffer at pH 7.4, followed by 1% osmium tetroxide for
1 h at 41C. A block stain was carried out in aqueous 2%
uranyl acetate at room temperature for 1 h. Tissue was
dehydrated in solutions of graded alcohol and em-
bedded in Durcupan (Fluka, Switzerland). Ultra-thin
sections (80 nm) were cut on a Reichert Ultracut-S and
counterstained with Reynold’s lead citrate and examined
at 80 kV on a JEOL 100C electron microscope. Granule
size was measured directly from the image.
For immunoelectronmicroscopy, a post-embedding
immunogold procedure was used to localize intracellular
insulin. Single cell suspensions of control HUH7, HUH7ins and MIN6 cells (mouse pancreatic b cell line, used as
a positive control) were fixed for 2 h at room temperature
in 4% formalin in 0.1 M phosphate buffer containing
40 mM betaine, pH 7.4. Cells were allowed to settle
between washes of PBS, 0.9% sodium chloride, ethanol
series to 70% followed by embedding in freshly made LR
White. Cells were centrifuged gently at 0.2 rcf for 10 min.
in a MiniSpin Plus Eppendorf centrifuge for the last
embedding step only. Hardening was at 481C for 48 h.
Sections were cut using a MT-1 microtome and labeling
procedures were carried out immediately after sectioning as described by Saccomano et al,51 with particular
details as follows: non-specific binding was blocked
using 1% BSA and 1% glycine in PBS at room
temperature for 2 h. A goat anti-mouse IgG (EMGMHL
10) 10 nm gold probe 1:50 (BBInternational, Australian
Laboratory Services Pty Ltd, Sydney, Australia) incubated for 2 h at room temperature was directed against a
monoclonal mouse anti-human insulin (1:70) primary
antibody (BioGenex, San Ramon, CA, USA) incubated
overnight at 41C. Primary antibody was replaced by PBS
to assess the level of non-specific binding of the gold
probe. Sections were stained with 1.5% aqueous uranyl
acetate alone for 9 min before observing at 80 kV in a
JEOL 100C electron microscope. Lead citrate staining
was not performed, to avoid confusion between possible
lead precipitates and gold particles.
501
Reverse phase high-performance liquid
chromatography (RP-HPLC) for separation of insulinrelated peptides
The HPLC system consisted of a 1050 Hewlett-Packard
pump with a Merck (Darmstadt, Germany) LiChrospher
100 RP-18 (5 mm) 250 4 mm column with a guard
cartridge. Cell extract samples (100 ml each) were loaded
on the HPLC column and eluted at a rate of 1 ml/min
with the following two buffer system: Buffer A: 3.4 ml
concentrated H3PO4, 14 g NaClO4, 2.2 g heptanesulfonic
acid monohydrate sodium salt brought to pH 3; Buffer B:
90% acetonitrile in H2O. Solution B was held at 38% for
30 min for insulin elution and then linearly increased to
42% over 65 min for elution of proinsulin-related peptides; for column wash solution B was then increased to
60% over 10 min, held at 60% for 5 min and then linearly
decreased over 10 min to 38%. Absorption at 217 nm was
monitored with a spectrophotometric detector (Jasco,
Tokyo, Japan). Fractions of 1 ml were collected in tubes
containing 0.1 ml 0.5 M boric acid, 0.1% BSA. For
determination of insulin and its related peptides, the
samples were dried in a Speed-Vac apparatus and
reconstituted for radioimmunoanalysis in 0.5 ml PBS/
0.1% BSA. Human insulin was determined by standard
RIA as described.52 Elution positions of human insulin,
diarginyl insulin, split 31,32 proinsulin, split 65,66
proinsulin and intact proinsulin were determined with
Gene Therapy
Insulin-producing liver cells
BE Tuch et al
502
appropriate standards (kindly provided by Eli Lilly &
Co, Indianapolis, IN, USA).
Statistical analysis
Data were analyzed with the computer statistical
package NCSS.53 Statistical significance of differences
between groups was tested by Student’s paired t-test or,
if there were more than two groups, by one-way analysis
of variance after log transformation of the data. Where
the P-value for the latter test was o0.05, the differences
between individual groups were evaluated with Duncan’s multiple range test.
Acknowledgements
This work was supported by a Project Grant from the
National Health and Medical Research Council of
Australia. Three of the authors were recipients of
Postgraduate Scholarships, one from the University of
Technology, Sydney (BZ); one from The University of
New South Wales (MTT); and one an Australian
Postgraduate Award from the Department of Education,
Training and Youth Affairs, Australia (RKBH). We wish
to thank Dr Murray Smith for his constructive comments
on the manuscript, Dr Anil Amaratunga, Mr Appavoo
Mathiyalagan and Dr Chang Tao for technical assistance
with Western blots, Mrs Pauline Khoury for assistance
with the transplants, Ms Robyn Baum and Ms Lillian Tan
for technical assistance with the proinsulin and Cpeptide assay, and Mr Roland Smith, Anatomy, Sydney
University for printing the electron micrographs in
Figures 6 and 7a–c.
References
1 Bland BJ (ed). International Pancreas Transplant Registry Newsletter.
2001, p. 13.
2 Shapiro AMJ et al. Islet transplantation in seven patients with
type 1 diabetes mellitus using a glucocorticoid-free immunosuppressive regimen. N Engl J Med 2000; 343:230–238.
3 Brendel MD, Hering BJ, Schultz AO, Bretzel RG. International
Islet Transplant Registry, Vol. 9, 2001.
4 Tuch BE et al. Recovery of human fetal pancreas after one year of
implantation in the diabetic patient. Transplantation 1988; 46:
865–870.
5 Simpson AM et al. Functional expression of the human insulin
gene in a human hepatoma cell line (HEP G2). Gene Ther 1995; 2:
223–231.
6 Simpson AM et al. Gene therapy of diabetes: glucose-stimulated
insulin secretion in a human hepatoma cell line (HEP G2ins/g).
Gene Ther 1997; 4: 1202–1215.
7 Vollenweider F et al. Processing of proinsulin by transfected
hepatoma (FAO) cells. J Biol Chem 1992; 267: 14 629–14 636.
8 Short DK, Okada S, Yamauchi K, Pessin JE. Adenovirusmediated transfer of a modified human proinsulin gene reverses
hyperglycaemia in diabetic mice. Am J Physiol 1998; 275: E748E756.
9 Ferber S et al. Pancreatic and duodenal homeobox gene 1
induces expression of insulin genes in liver and ameliorates
streptozotocin-induced hyperglycaemia. Nature Medicine 2000; 6:
568–572.
10 Mitanchez D et al. Regulated expression of mature human
insulin in the liver of transgenic mice. FEBS Lett 1998; 421: 285–
289.
Gene Therapy
11 Lee HC et al. Remission in models of type 1 diabetes by gene
therapy using a single-chain insulin analogue. Nature 2000; 408:
483–488.
12 Kolodka TM, Finegold M, Moss L, Woo SLC. Gene therapy for
diabetes mellitus in rats by hepatic expression of insulin. Proc
Natl Acad Sci USA 1995; 92: 3293–3297.
13 Valera A et al. Regulated expression of human insulin in the liver
of transgenic mice corrects diabetic alterations. FASEB J 1994; 8:
440–447.
14 Miyazaki J-I et al. Establishment of a pancreatic beta cell line that
retains glucose-inducible insulin secretion: special reference to
expression of glucose transporter. Endocrinology 1990; 127: 126–
132.
15 Huang XF, Arvan P. Formation of the insulin-containing
secretory granule core occurs within immature b-granules. J
Biol Chem 1994; 269: 20 838–20 844.
16 Duckworth WC. Insulin degradation: mechanisms, products and
significance. Endocr Rev 1998; 9: 319–345.
17 Authier F et al. Endosomal proteolysis of internalized insulin at
the C-terminal region of the B chain by cathepsin D. J Biol Chem
2002; 277: 9437–9446.
18 Terman A et al. Decreased apoptotic response of inclusion-cell
disease fibroblasts: a consequence of lysosomal enzyme missorting? Exper Cell Res 2002; 274: 9–15.
19 Alberts B et al. The Molecular Biology of the Cell. Garland
Publishing Inc: New York, 1994, p. 629.
20 Orci L. Macro- and micro-domains in the endocrine pancreas.
Diabetes 1982; 31: 538–565.
21 Schuit FC, IN’T Veld PA, Pipeleers DG. Glucose stimulates
proinsulin biosynthesis by a dose-dependent recruitment of
pancreatic beta cells. Proc Natl Acad Sci USA 1988; 85: 3865–3869.
22 Simpson AM, Tuch BE. Control of insulin biosynthesis in the
human fetal beta cell. Pancreas 1995; 11: 48–54.
23 Carlsson PO, Palm F, Andersson A, Liss P. Markedly decreased
oxygen tension in transplanted rat pancreatic islets irrespective
of the implantation site. Diabetes 2001; 50: 489–495.
24 Tuch BE et al. Transplantation of genetically engineered insulin
producing hepatocytes into immunoincompetent mice. Transplant Proc 1998; 30: 473.
25 Orci L. The insulin factory: a tour of the plant surroundings and
a visit to the assembly line. Diabetologia 1985; 28: 528–546.
26 Naya FJ et al. Diabetes, defective pancreatic morphogenesis, and
abnormal enteroendocrine differentiation in BETA2/neuroDdeficient mice. Genes Dev 1997; 11: 2323–2334.
27 Tian YW, Smith PG, Yeoh GC. The oval-shaped cell as a
candidate for a liver stem cell in embryonic, neonatal and
precancerous liver: identification based on morphology and
immunohistochemical staining for albumin and pyruvate
kinase isoenzyme expression. Histochem Cell Biol 1997; 107:243–
250.
28 Leibowitz G et al. IPF1/PDX1 deficiency and beta-cell dysfunction in Psammomys obesus, an animal with type 2 diabetes.
Diabetes 2001; 50:1799–1806.
29 Koina ME, Simpson AM, Tuch BE, Marshall GM. Inducible
insulin expression in two human hepatoma cell lines. Proc Aust
Diabetes Soc 1998; 134: (Abstract P15).
30 Sirica AE, Gainey TW, Mumaw VR. Ductal hepatocytes.
Evidence for a bile ductular cell origin in furan-treated rats.
Am J Pathol 1993; 145: 375–383.
31 Elmore LW, Sirica AE. Phenotypic characterization of metaplastic intestinal glands and ductular hepatocytes in cholangiofibrotic lesions rapidly induced in the caudate liver lobe of rats
treated with furan. Cancer Res 1991; 51: 5752–5759.
32 Rao MS, Subbaraq V, Reddy JK. Induction of hepatocytes in the
pancreas of copper-depleted rats following copper depletion.
Cell Differ 1986; 18: 109–117.
33 Dabeva MD, Shafritz DA. Activation, proliferation and differentiation of progenitor cells in to hepatocytes in the D-
Insulin-producing liver cells
BE Tuch et al
503
34
35
36
37
38
39
40
41
42
43
galactosamine model of liver regeneration. Am J Pathol 1993;143:
1606–1620.
Konishi N, Ward JM, Waalkes MP. Pancreatic hepatocytes in
Fischer and Wistar rats induced by repeated injections of
cadmium chloride. Toxicol Appl Pharmacol 1990;18: 678–686.
Monti LD et al. Intravenous infusion of diarginyl insulin an
insulin, analogue – effects on glucose turnover and lipid levels in
insulin-treated type-II diabetic patients. Metab – Clin Exp
1992;41:540–544.
Matchinsky FM, Glaser B, Magnuson MA. Pancreatic b-cell
glucokinase: closing the gap between theoretical concepts and
experimental realities. Diabetes 1998; 47: 307–315.
Hughes SD, Johnson JH, Quaade C, Newgard CB. Engineering
of glucose-stimulated insulin secretion and biosynthesis in nonislet cells. Proc Natl Acad Sci USA 1992; 89: 688–692.
Simpson AM et al. Transformation of pituitary and fibroblast cell
lines using human insulin cDNA and a dexamethasoneinducible promoter. Transplant Proc 1993; 25: 2915–2916.
Tuch BE, Maitland JE, Turtle J. Culture and perifusion of human
fetal pancreas. In: Larner J, Pohl S (eds). Methods in Diabetes
Research. Vol. 1: Laboratory Methods, Part B. John Wiley & Sons:
New York, 1984. pp 153–163.
Maldonato A, Renold AE, Sharp GWG, Cerasi E. Glucoseinduced proinsulin biosynthesis. Diabetes 1977; 26: 538–545.
Chomczynski P, Sacchi N. Single-step method of RNA isolation
by acid guanidinium thiocyanate–phenol–chloroform extraction.
Anal Biochem 1987; 162: 156–159.
Noonan KE et al. Quantitative analysis of MDR1 (multidrug
resistance) gene expression in human tumours by polymerase
chain reaction. Proc Natl Acad Sci USA 1990; 87: 7160–7164.
Tanizawa Y, Koranyi LI, Welling CM, Permutt MA. Human liver
glucokinase gene: cloning and sequence determination of two
44
45
46
47
48
49
50
51
52
53
alternatively spliced cDNAs. Proc Natl Acad Sci USA 1991; 88:
7294–7297.
Tuch BE, Chen J. Resistance of the human fetal b cell to the toxic
effect of multiple low dose streptozotocin. Pancreas 1993; 8: 305–
311.
Tuch BE, Osgerby KJ, Turtle JR. Normalization of blood glucose
levels in nondiabetic nude mice by human fetal pancreas after
induction of diabetes. Transplantation 1988; 46: 608–611.
Simeonovic CJ, Brown DJ, Teirrinen KUS, Wilson JD. Preparation
and transplantation of fetal proislets. In: Ricordi C (ed).
Pancreatic Islet Cell Transplantation. RG Landes Co: Austin, TX,
1992, pp 238–248.
Davalli AM et al. Function, mass, and replication of porcine and
rat islets transplanted into diabetic nude mice. Diabetes 1995; 44:
104–111.
Tuch BE, Roberts EC, Darby KB. Release of proinsulin from the
human fetal beta cell. J Endocrinol 1992; 132: 159–67.
Bradford MM. A rapid and sensitive method for the quantitation
of microgram quantities of protein utilising the principle of
protein-dye binding. Anal Biochem 1976; 72: 248–254.
Swan MA. Improved preservation of ultrastructural morphology in human spermatozoa using betaine in the primary fixative.
Int J Androl 1997;20: 45–54.
Saccomanno K et al. Immunoelectron microscopy: a new method
for detection of insulin antibodies. J Histochem Cytochem 1993; 41:
1233–1239.
Gross DJ, Leibowitz G, Cerasi E, Kaiser N. Increased susceptibility of islets from diabetes-prone Psammomys obesus to the
deleterious effects of chronic glucose exposure. Endocrinology
1996; 137: 5610–5615.
Hintze JL. Number Cruncher Statistical System: Version 5.01.
Kaysville, Utah, 1991.
Gene Therapy