Differentiation of Human Liver-Derived, Insulin-Producing Cells Toward the β-Cell Phenotype

FH-B cells were cultured as previously described (1) in Dulbecco’s modified Eagle’s medium containing 25 mmol/l glucose and supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 μg/ml streptomycin, 5 μmol/l hydrocortisone, and 5 μg/ml insulin. FH-B-TPN cells were maintained in the same culture medium, except without the hydrocortisone and insulin (complete medium). For incubation in SFM, the cells were placed in Dulbecco’s modified Eagle’s medium containing antibiotics in the presence of 10 μg/ml insulin, 5.5 μg/ml transferrin, and 5 ng/ml selenium (Sigma-Aldrich, Steinheim, Germany). Treatment with Act-A (Cytolab/PreproTech Asia, Rehovot, Israel), betacellulin (R&D Systems, Minneapolis, MN), nicotinamide (Sigma-Aldrich), exendin-4 (Sigma-Aldrich), and hepatocyte growth factor (HGF; Sigma-Aldrich) was at the concentrations detailed for each experiment. The experiments described herein used cells between passages 9 and 22 (passage 1 is defined as the first passage after neomycin selection after the introduction of PDX-1).

Insulin secretion was measured by static incubation, as previously described (4). Cells were plated in 24-well plates at 5 × 10⁴ cells per well. The cells were preincubated for 1 h in Krebs-Ringer buffer (KRB) then incubated for 30 min in KRB containing 0.5 mmol/l 1-isobutyl 3-methylxanthine and glucose at various concentrations. The cells were then extracted in acetic acid, and the amount of insulin in the buffer and cell extract was determined by radioimmunoassay (RIA) using the INSIK-5 kit (DiaSorin, Vercelli, Italy), according to the manufacturer’s instructions. This assay has <20% cross-reactivity with proinsulin. In addition, insulin content was determined using an enzyme-linked immunosorbent assay (ELISA) kit (Mercodia, Uppsala, Sweden) that recognizes only mature insulin. Insulin content was normalized to total cellular protein, measured by the Bio-Rad protein assay kit (Hercules, CA). Human C-peptide in the cell extract was determined using an RIA (DiaSorin) or ELISA (Mercodia) kit, according to the manufacturer’s instructions.

Total RNA was extracted from cultured cells and human pancreatic islets (obtained through the Juvenile Diabetes Research Foundation Islet Distribution Program) using a commercial kit (High Pure RNA isolation kit, Roche, Mannheim, Germany). Specific transcripts were analyzed with the SuperScript III (Invitrogen, Carlsbad, CA) RT-PCR kit and used according to the manufacturer’s instructions. cDNA was amplified for 40 cycles (94°C for 45 s, annealing for 45 s, and 72°C for 40 s) using the primer pairs and annealing temperatures listed in Table 1. PCR products were separated by electrophoresis in 1.5% agarose gels and visualized by ethidium bromide staining.

NOD-scid female mice (Harlan, Jerusalem, Israel), age 2 months, were made hyperglycemic by an intraperitoneal injection of 180 μg/g body wt of streptozotocin (STZ). On the day that blood glucose levels reached >300 mg/dl, mice were transplanted with 2 × 10⁶ FH-B-TPN cells in 25 μl PBS that were injected under the capsule of the left kidney using a 30-gauge needle. Blood glucose levels were monitored twice a week in samples obtained from the tail vein of fed mice using Accutrend strips (Roche). Serum levels of human C-peptide were determined in blood samples obtained from the orbital plexus of fed mice using the ELISA kit. For the glucose tolerance test, mice fasted for 6 h were injected intraperitoneally with 1 mg/g body wt glucose in saline. Blood glucose levels were monitored at the indicated time points in samples obtained from the tail vein. At the termination of the experiment, mice were injected intraperitoneally with 100 μg/g body wt 5-bromo-2-deoxyuridine (BrdU; Sigma-Aldrich). The kidney was removed 6 h later for histological analyses.

Cells plated in 24-well plates on sterilized coverslips were fixed in 4% paraformaldehyde. For nuclear antigens, cells were permeabilized with 0.25% NP-40 for 10 min. Cells were blocked for 10 min at room temperature in 1% BSA, 10% fetal bovine serum, and 0.2% saponin and incubated overnight at 4°C, with the primary antibody diluted in blocking solution (Table 2). All the antibodies were calibrated using FH-B, COS7, or 293T cells as negative controls. The bound antibody was visualized with a fluorescent secondary antibody (Biomeda, Foster City, CA) (Table 2) under a Zeiss confocal microscope. Nuclei were visualized with DAPI (Roche) staining for 5 min at room temperature. Kidney tissue was fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned. Sections were rehydrated, washed in PBS, and unmasked (if needed) in unmasking solution (Vector, Burlingame, CA), according to the manufacturer’s instructions. Sections were blocked for 2 h in 0.2% Tween 20 and 0.2% gelatin, incubated overnight at 4°C with primary antibodies and then for 2 h at room temperature with the secondary antibodies (Table 2), stained with DAPI, and mounted. BrdU staining was performed as previously described (5).

A number of soluble factors known to contribute to β-cell differentiation were evaluated for their ability to promote differentiation of FH-B-TPN cells. As seen in Table 3, of the four factors analyzed for their effect in complete medium (Act-A, betacellulin, nicotinamide, and exendin-4), only Act-A increased cellular insulin content significantly (about fourfold) compared with cells cultured in regular medium, as judged by RIA. Exendin-4, and to a lesser extent nicotinamide, induced a decrease in the insulin content of FH-B-TPN cells. Combined treatment with Act-A plus betacellulin plus nicotinamide did not result in an additive effect compared with Act-A alone. Similarly, the effect of Act-A combined with HGF was somewhat worse than that of Act-A alone, in contrast to previous reports showing HGF to contribute to the differentiation of insulin-producing cells at a higher concentration (6). The optimal effect of Act-A was achieved at a concentration of 3 nmol/l (Fig. 1).

To evaluate the effect of Act-A in combination with SFM treatment, which has previously been shown to increase insulin content in FH-B-TPN cells, the cells were incubated in different protocols with various combinations of the two conditions (Table 4). The effect of a 3-day treatment with 3 nmol/l Act-A was more pronounced in SFM compared with in complete medium. When the 3-day Act-A treatment in SFM was preceded by a 6-day incubation in SFM in the absence of Act-A, insulin content was greatly increased, to 33 times that of cells grown in complete medium. This represented a 2.6-fold increase over the insulin content of cells incubated for the same combined 9-day period in SFM in the absence of Act-A. This insulin content represented 6% of the cellular protein content and ∼60% of the insulin content of normal human pancreatic islets. When the order was reversed, with the Act-A treatment preceding the incubation in SFM, the resulting insulin content was not much higher than with Act-A alone. The RIA results were confirmed by ELISA analysis, which detects only mature insulin, showing that 95% of stored insulin was in the form of mature protein. The human C-peptide levels in these cells, as analyzed by ELISA, were 5,153 ± 180 ng/10⁶ cells. The levels of human C-peptide secreted into the culture medium were 43.5 ± 1.3 ng/10⁶ cells. No insulin was detected in FH-B cells, which did not express PDX-1, when treated under these conditions, which suggests that insulin expression could not be induced by these treatments in the absence of PDX-1.

The change in insulin content was accompanied by changes in the expression of other genes, as revealed by RT-PCR analyses. An increase in insulin mRNA levels was induced by all three treatments (Fig. 2: 3 nmol/l Act-A for 3 days in complete medium [lane 6], SFM for 6 days [lane 7], and SFM for 6 days followed by 3 nmol/l Act-A for 3 days in complete medium [lane 8]). Among the transcription factor genes analyzed, NeuroD transcripts were induced by all three treatments, most notably by SFM followed by Act-A, and Nkx2.2 was highly induced by SFM and to a lesser extent by SFM followed by Act-A. In contrast, Nkx6.1 expression was detected in all the conditions studied. The expression of paired box homeotic gene 6 (PAX-6) decreased after incubation in all three media, particularly the two lacking serum. Expression of the prohormone convertase (PC) 1/3 was significantly elevated only by treatment in SFM followed by Act-A, whereas PC2 transcript levels were not affected. Glucokinase expression was detectable in complete medium, in contrast to our previous findings, which were obtained with different primers and a different RT-PCR kit (1). All three treatments resulted in a significant increase in glucokinase transcript levels. Of the two non-β islet cell genes analyzed, glucagon expression was decreased in the presence of Act-A and elevated by SFM lacking Act-A, whereas PP transcripts as well as those of the hepatic gene α1-antitrypsin (α1-AT) increased in response to Act-A alone but decreased in the two conditions lacking serum. Most of the genes analyzed were not expressed in similarly treated FH-B cells, indicating that in the absence of PDX-1, the culture medium conditions alone were not sufficient to induce their expression. Notable exceptions were the activation of NeuroD expression after treatment with Act-A and PC1/3 induction by SFM treatment followed by Act-A (Fig. 2). These findings were reproducible in multiple independent experiments. The large number of cycles in the RT-PCR analyses was required to generate clearly visible bands in most of the samples. As seen in Fig. 2, there were great differences in intensities between different transcripts and samples, indicating that a plateau was not reached. This number of cycles ensured that negative results can be trusted to represent a complete lack of expression of specific genes, while weak positive results needed to be confirmed by other methods, such as immunostaining.

Immunofluorescence analyses comparing cells in complete medium with cells treated in SFM followed by Act-A showed an increase in the staining intensity for insulin and C-peptide (Fig. 3). In addition, a great increase was observed in the number of cells stained for NeuroD and NKX2.2, thereby confirming the RNA analyses. Conversely, this culture condition resulted in the disappearance of PAX-6 and PP immunostaining in all of the cells. PC1/3 and glucokinase immunostaining was present in all the cells grown in complete medium; however, the staining intensity for PC1/3 increased after treatment with SFM plus Act-A. The difference in PC1/3 staining intensity was smaller than expected from the more pronounced difference in RNA signals. No glucagon or somatostatin staining was visible in cells in either condition.

The cell doubling time, as determined by cell counting, was not affected by Act-A treatment in complete medium. Incubation in SFM increased doubling time by twofold, resulting in a slower proliferation rate compared with cells growing in complete medium, whereas the SFM plus Act-A treatment increased it fourfold. No apoptotic cells were detected by a TUNEL assay in any of the conditions (data not shown).

Insulin secretion, which was previously shown to be glucose responsive in the physiological concentration range in FH-B-TPN cells in both complete medium and SFM (1), was also normal in cells grown in the presence of Act-A, both during 3 days in complete medium and after Act-A treatment during the last 3 of 9 days in SFM (Fig. 4). The maximal secretion at 20 mmol/l glucose of cells treated with SFM plus Act-A represented 1.1% of their insulin content, similar to that in normal islets.

To evaluate the dependence of the differentiated cell phenotype after SFM plus Act-A treatment on continuous culture under these conditions, cells were shifted after the treatment to a 10-day period in complete medium. No significant changes were observed in transcripts of Nkx2.2, glucokinase, PC1/3, or PC2 (Fig. 2, lanes 8 and 9), and no reappearance of glucagon transcripts was observed. In contrast, there was a reduction in the levels of insulin, NeuroD, and Nkx6.1 transcripts and a reappearance of PAX-6, PP, and α1-AT transcripts. Insulin content was reduced by 41% to 3,626 ± 207 ng/10⁶ cells.

To assess the functional stability of the FH-B-TPN cells in vivo after in vitro treatment with SFM plus Act-A, cells were transplanted under the renal capsule of STZ-induced diabetic NOD-scid mice. As seen in Fig. 5A, blood glucose levels were lowered starting 2 days after transplantation. Glycemia was normalized thereafter, and stable blood glucose levels were maintained for >2 months, until the experiment was terminated for histological analyses. No hypoglycemia developed by the end of the experiment. Before the mice were killed, they were subjected to a glucose tolerance test, which demonstrated a normal rate of glucose clearance (Fig. 5B). Human C-peptide ELISA detected serum levels of 0.31–0.84 ng/ml (compared with 0.27 ng/ml in a human serum control and no detectable signal in normal mouse serum). Histological analyses detected insulin and human C-peptide immunofluorescence staining in cells positively identified as human using human-specific anti−heat shock protein 27 antibodies with no cross-reactivity to mouse (Fig. 5C). No BrdU-labeled cells were detected in the transplants, indicating that little or no cell replication occurred in the transplanted cells at this time point (Fig. 5C).

Our findings demonstrated that FH-B-TPN cells can be differentiated toward the β-cell phenotype by incubating them in the presence of Act-A in SFM. These conditions activate expression of the β-cell transcription factors NeuroD and Nkx2.2 and downregulate expression of the α-cell transcription factor PAX-6. Changes in the expression of other genes may be the result of this shift in transcription factor profile or may be directly affected by the inductive conditions. The resulting upregulation of glucokinase and PC1/3 expression and the downregulation of PP as well as the hepatic marker α1-AT bring the phenotype of FH-B-TPN cells closer to that of normal β-cells. The immunofluorescence analyses demonstrated that the phenotype of the FH-B-TPN cell population is quite uniform, which is consistent with its clonal origin (1,2).

In one significant finding, the cellular insulin content was increased by up to 33-fold to over 6% of cellular protein content, which represents ∼60% of the content of normal human pancreatic islets. To our knowledge, this is the highest content reported to date for human insulin–producing cells derived from stem/progenitor cells. These amounts of insulin result from biosynthesis in the cells rather than uptake from the medium, as judged by the following criteria: 1) no insulin was detected in FH-B cells cultured in the same conditions, 2) insulin was detected in FH-B-TPN cells cultured in complete medium that was not supplemented with insulin, 3) insulin mRNA was detected in the cells, and 4) human C-peptide was detected in the cells by ELISA and immunofluorescence, as well as in the culture medium and in the serum of mice transplanted with these cells. The modified cells maintained a normal glucose-induced insulin secretory profile in the physiological concentration range, which was already observed in FH-B-TPN cells cultured in complete medium (1). The insulin content of isolated human islets is known to vary widely (3). Eizirik et al. (7) cited a value of 22.9 ng/islet, which would represent 11.45 μg/10⁶ cells, assuming an islet contains an average of 2,000 cells. Here we used a convenient measuring stick of 10 μg/10⁶ cells for comparison.

The differentiated phenotype achieved by these culture conditions was partly stable in vitro after the return of the cells to complete medium for 10 days. However, a more relevant test of phenotypic stability is in vivo cell function. As judged by assays performed >2 months after transplantation, the transplanted cells were capable of maintaining euglycemia and responding to a glucose tolerance test in a normal time course. In addition, intense immunofluorescence for insulin and human C-peptide demonstrated that a stable, highly differentiated state of the cells was maintained for prolonged periods in a nonpancreatic site in vivo. In one interesting finding, the immunostaining analyses also revealed the absence of cell replication in vivo, as judged by the lack of BrdU incorporation, indicating that the constitutive expression of hTERT did not overcome signals of contact inhibition by itself.

The molecular mechanism by which cell culture in the presence of Act-A in SFM leads to FH-B-TPN cell differentiation remains to be elucidated. The activity of Act-A required the expression of PDX-1, as Act-A by itself did not induce differentiation of FH-B cells. It was interesting that the other factors tested (betacellulin, nicotinamide, and HGF) did not significantly increase FH-B-TPN cell differentiation, as judged by insulin content, and did not potentiate the effect of Act-A at the concentrations used here. The effect of SFM is likely attributable to serum removal rather than the presence of insulin, transferrin, and selenium in the medium. Serum removal may eliminate both growth factors and inhibitors of cell differentiation. The changes in cell doubling time suggest that further cell differentiation is coupled with withdrawal from the cell cycle. Act-A is a cytokine of the transforming growth factor-β superfamily, which is responsible for pleiotropic effects, including cell proliferation, differentiation, and apoptosis (8). The factor and its receptor are expressed in multiple cell types in the embryo and the adult, including islet α- and δ-cells (9). Follistatin, an antagonist of Act-A, has been shown to block development of the rat endocrine pancreas and promote exocrine pancreas development (10). Mouse models expressing a dominant-negative form of Act-A receptor II manifest abnormalities in endocrine pancreas development (11,12). Act-A together with betacellulin (13) or HGF (6) convert the exocrine pancreas cell line AR42J into insulin-producing cells and induce differentiation of human fetal pancreas endocrine cells (14). Taken together, these findings support a role for Act-A in β-cell development and differentiation. In addition, Act-A inhibits hepatocyte growth (15,16), which is confirmed by our findings of an increase in cell doubling time obtained with Act-A in SFM. The activity of the Act-A receptor is mediated through intracellular proteins known as Smads, which translocate to the cell nucleus and modulate gene transcription (8). In comparing the pattern of gene expression obtained with SFM to that of SFM plus Act-A (Fig. 2, lanes 7 and 8), the major changes observed were the increase in NeuroD and PC1/3 and the decrease in Nkx2.2 expression induced by Act-A. It is possible that these changes are responsible for the close to threefold increase in insulin content found in the cells treated with Act-A. The fact that treatment with Act-A after a 6-day incubation in SFM resulted in the highest insulin content obtained in this study suggests a need for a sequential order of events, in which an increase in Nkx2.2 and a decrease in PAX-6 expression is followed by a marked increase in NeuroD expression, to maximize the cellular insulin biosynthesis and storage capacities. Microarray analyses of the entire spectrum of changes in gene expression affected by this treatment may provide further mechanistic insights. Nevertheless, the properties of FH-B-TPN cells after the differentiation affected by Act-A in SFM render them excellent candidates for the development of an abundant cell source for β-cell replacement therapy of type 1 diabetes. The presence of a constitutive hTERT gene in FH-B-TPN cells may represent an increased risk of neoplasia compared with cells lacking hTERT expression. However, FH-B cells were extensively tested both in vitro and in vivo and exhibited no oncogenicity (2). FH-B-TPN cells manifest a decreased proliferation rate compared with FH-B cells, undergo contact inhibition in vitro (Fig. 3), and do not label with BrdU in vivo (Fig. 5). Nevertheless, the use of cell encapsulation or suicide genes may be needed to increase the safety of their transplantation.

This work was supported by National Institutes of Health Grant DK-52956-06 to S.E.

We thank Mark Magnuson and Donald Steiner for antibodies. The Nkx2.2 monoclonal antibody developed by T. Jessell was obtained from the Developmental Studies Hybridoma Bank.

This work was performed in partial fulfillment of the requirements for the doctoral degree of Michal Zalzman.