Glucagon-Like Peptide 1 Induces Differentiation of Islet Duodenal -1–Positive Pancreatic Ductal Cells Into Insulin-Secreting Cells Hongxiang Hui,1 Chris Wright,2 and Riccardo Perfetti1,3

Glucagon-like peptide-1 (GLP-1) is an incretin hormone capable of restoring normal glucose tolerance in aging glucose-intolerant Wistar rats. Whether the antidia- ndocrine and exocrine cells originate from a betic properties of GLP-1 are exclusively due to its precursor epithelial cell during pancreatic orga- insulin secretory activity remains to be determined. A nogenesis (1,2). Various differentiation factors GLP-1–dependent differentiation of pancreatic precur- are required to achieve the mature phenotype sor cells into mature ␤-cells has recently been proposed. E ␤ The aim of this study was to investigate whether pan- characteristic of islet -cells. The use of a knockout mouse creatic ductal epithelial cells could be differentiated model for islet duodenal homeobox-1 (IDX-1) (also termed into insulin-secreting cells by exposing them to GLP-1. IPF-1/STF-1 and PDX-1) has significantly contributed to Rat (ARIP) and human (PANC-1) cell lines, both de- the elucidation of the specific role played by different rived from the pancreatic ductal epithelium, were used in the differentiation of insulin-secreting cells. Mice lacking to test this hypothesis. A major difference distinguishes IDX-1 fail to develop a pancreas (3). Islet-1, a homeodomain- these two cell lines: whereas ARIP cells spontaneously containing , is necessary for the development of the express the ␤-cell differentiation factor islet duodenal homeobox-1 (IDX-1), PANC-1 cells are characteristi- dorsal pancreas and is required for the generation of islet cally IDX-1 negative. GLP-1 induced the differentiation cells (4). Inactivation of NeuroD/Beta2 or Pax4 genes cause a of ARIP cells into insulin-synthesizing cells, although it striking reduction in the number of insulin-producing cells did not affect the phenotype of PANC-1 cells, as deter- and a failure to develop mature islets (5,6). mined by fluorescence-activated cell sorting (FACS) Growth and differentiation of islet ␤-cells is not limited analysis. Differentiation of ARIP cells by exposure to to the embryological state. A constant remodeling of size human GLP-1 occurs in a time- and dose-dependent and function of the islets of Langerhans occurs during the manner, and this is associated with an increase in IDX-1 and insulin mRNA levels. Secretion of insulin was also entire life of individuals and is likely to play an essential induced in a parallel manner, and it was regulated by the role in the prevention of diabetes. In adult rats, two inde- concentration of glucose in the culture medium. Inter- pendent pathways are used for the proliferation of pancre- estingly, PANC-1 cells, when stably transfected with atic endocrine cells. In the first pathway, new endocrine human IDX-1, gained responsiveness to GLP-1 and were cells arise from the division and differentiation of cells with- ␤ able to differentiate into -cells, as determined by FACS in the islets, whereas in the second pathway of prolifera- analysis, insulin expression, intracellular insulin content, and insulin accumulation in the culture me- tion, the islets cells originate from precursor cells located dium. Finally, we demonstrated that the for in the pancreatic ductal epithelium (7). It is likely that a GLP-1 is constitutively expressed by ARIP and PANC-1 coordinated activation of multiple differentiation factors— cells and that the mRNA level for this transcript was in a fashion similar to the sequence of events occurring increased by cellular transfection with human IDX-1. In during fetal development—is required for the cellular growth summary, our study provides evidence that GLP-1 is a of the endocrine pancreas of adults. The mechanism(s) for differentiation factor for pancreatic ductal cells and the activation of such a complex regulatory network in that its effect requires the expression of IDX-1. Diabetes 50:785–796, 2001 adulthood is not known. Recently, Xu et al. (8) demon- strated that an analog of the incretin hormone glucagon- like peptide (GLP)-1, termed exendin-4, was able to in- crease islet mass in adult animals previously subjected to From the 1Division of Diabetes, Endocrinology and Metabolism, Cedars-Sinai subtotal pancreatectomy. Similarly, we recently demon- Medical Center; the 2Division of Cell Biology, Vanderbilt University; and the 3University of California Los Angeles, Los Angeles, California. strated that the treatment of glucose-intolerant aging Address correspondence and reprint requests to Riccardo Perfetti, MD, Div. Wistar rats with GLP-1 restored normal glucose tolerance Endocrinology and Metabolism, Becker Building, Room B-131, Department of Medicine, Cedars-Sinai Medical Center, 8700 Beverly Blvd., Los Angeles, CA and induced islet cell proliferation (9). These studies 90048. E-mail: [email protected]. suggest that exogenously administered stimuli are able, in Received for publication 21 April 2000 and accepted in revised form 12 vivo, to increase the mass of insulin-secreting cells and December 2000. ANOVA, analysis of variance; FACS, fluorescence-activated cell sorting; ameliorate glucose tolerance by inducing neogenesis of FCS, fetal calf serum; FITC, fluorescein isothiocyanate; GLP, glucagon-like islet cells. In the present study, we investigated the ability peptide; GLP-R, GLP-1 receptor; IDX-1, islet duodenal homeobox-1; PBS, phosphate-buffered saline; PCR, polymerase chain reaction; RIA, radioimmu- of human recombinant GLP-1 to differentiate ductal epi- noassay; RT, reverse transcription. thelial cells into insulin-secreting cells.

DIABETES, VOL. 50, APRIL 2001 785 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

RESEARCH DESIGN AND METHODS TABLE 1 Cell culture. Rat (ARIP) and human (PANC-1) ductal cell lines were provided PCR information by Dr. J.M. Egan (National Institute on Aging, Baltimore, MD) or purchased Size of PCR from ATCC (American Type Culture Collection, Manassas, VA), respectively. ARIP cells were cultured in F12 medium (Gibco-BRL, Gaithersburg, MD) Target gene Primer sequence product (bp) containing 100 ␮g/ml penicillin, 50 ␮g/ml streptomycin, and 10% fetal calf Rat serum (FCS) (Gibco-BRL) at 37°C under a humidified condition of 95% air and Insulin CCTGCCCAGGCTTTTGTCAA(ϩ) 187 5% CO2. PANC-1 cells were cultured in Dulbecco’s modified Eagle’s medium CTCCAGTGCCAAGTCTGAA(Ϫ) supplemented with antibiotics and FCS, as indicated for F12 medium. Treatment GLUT2 TTAGCAACTGGGTCTGCAAT(ϩ) 343 with human GLP-1 (H-6795; Bachem, King of Prussia, PA) was carried out us- GGTGTAGTCCTACACTCATG(Ϫ) ing cells grown to 80% confluence after washing the cell layer with serum-free GK AAGGGAACAACATCGTAGGA(ϩ) 136 medium and a “wash-out” incubation for 6 h with serum-free medium. To deter- CATTGGCGGTCTTCATAGTA(Ϫ) mine dose response to GLP-1, cells were cultured with fresh serum-free medium ␤-Actin CGTAAAGACCTCTATGCCAA(ϩ) 349 containing increasing concentrations of GLP-1 (0.1, 1, 10, and 20 nmol/l) or ve- AGCCATGCCAAATGTCTCAT(Ϫ) hicle alone. At the completion of the experiment, media and cells were collected. GLP-R TCTCTTCTGCAACCGAACCT(ϩ) 350 To determine the time course of response, cells were cultured in serum-free CTGGTGCAGTGCAAGTGTCT(Ϫ) medium with GLP-1 (10 nmol/l) for 0, 12, 24, 48, 72, or 96 h. Control dishes were cultured with vehicle alone or with the peptide receptor antagonist of GLP-1, Human exendin-9 (100 nmol/l), for 72 h (provided by Dr. J.M. Egan). The glucose Insulin CTCACACCTGGTGGAAGCTC(ϩ) 212 concentration in the culture medium was 12 mmol/l for both cell lines. AGAGGGAGCAGATGCTGGTA(Ϫ) Cell transfection with human IDX-1 cDNA. PANC-1 cells were transfected GLUT2 AGCTTTGCAGTTGGTGGAAT(ϩ) 300 with a pcDNA3 construct (Invitrogen, Carlsbad, CA) harboring the wild-type AATAAGAATGCCCGTGACGA(Ϫ) full-length IDX-1 cDNA using LipoTAXI (catalog number 204110, Mammalian GK TGGACCAAGGGCTTCAAGGCC(ϩ) 207 Transfection Kit; Stratagene, La Jolla, CA). Control cells were transfected with CATGTAGCAGGCATTGCAGCC(Ϫ) the vector alone. The selection of positive (i.e., transfected) cells was carried ␤-Actin GTGGGGCGCCCCAGGCACCA(ϩ) 392 out by culturing the cells in the presence of 400 ␮g/ml of G418 sulfate (GN-04; CTCCTTAATGTCACGCACGATTTC(Ϫ) Omega, Tarzana, CA). GLP-R GTGTGGCGGCCAATTACTAC(ϩ) 347 Immunocytochemistry and immunofluorescence microscopy. Cells were CTTGGCAAGTCTGCATTTGA(Ϫ) cultured on monocoated chamber slides (Nalge Nunc International, Naper- ville, IL) in the presence of GLP-1 (10 nmol/l) or vehicle for 72 h. GK, glucokinase. For the detection of insulin, cells were washed and fixed with 3% para- formaldehyde for4hatroom temperature in phosphate-buffered saline (PBS) was used as a correction factor for determining the relative amount of solubilized with 0.1% (vol/vol) Triton X-100 in PBS for 5 min. Cells were then medium to be assayed for each individual RIA for insulin. incubated sequentially with an anti-insulin antibody and a secondary antibody, RNA isolation and Northern blot analysis. Cellular RNA was extracted as as described by the manufacturer (Biomeda, Foster City, CA). The cells were routinely described. Northern blots were hybridized with 1) a full-length rat examined using a Ziess Axiophoto microscope (Ziess, New York). insulin II cDNA probe, 2) human IDX-1 cDNA, and 3) a rat ␤-actin cDNA For the detection of IDX-1, slightly different conditions were used. Briefly, probe. All cDNA probes were labeled with [32P]dCTP (Amersham Life Science, the concentration of paraformaldehyde was decreased to 2%, and the concen- Arlington Heights, IL) by the random priming procedure using the enzyme tration of Triton to permeabilize the cells was raised to 0.2% Triton X-100. sequenace (United States Biochemical, Cleveland, OH). Hybridization and Cells were then washed with 0.01 mol/l PBS three times for 3–5 min, and washing conditions were carried out as previously described (9). Messenger nonspecific binding was inhibited by using 5% chick serum in 0.01 mol/l PBS RNA levels for individual transcripts were evaluated by densitometric analysis at room temperature for 60 min in a humid chamber. A rabbit IDX-1 antibody and normalized for the relative abundance of ␤-actin mRNA.

directed against the NH2-terminus of the frog homologue of the IDX-1 gene Reverse transcription–polymerase chain reaction and Southern blot was used as the primary antibody (1:500 diluted with 0.1% Triton X-100 and 1% analysis. After the treatment of ARIP and PANC-1 cells with GLP-1, ex- bovine serum albumin in 0.01 mol/l PBS), and slides were incubated at 4°C endin-9, or vehicle for the described length of time and the various doses, the overnight in a humid chamber. After washing, cells were incubated with a culture medium was removed and the cells were washed twice with serum- fluorescein-conjugated goat anti-rabbit IgG antibody (Molecular Probes, Eu- free medium. Total RNA was isolated using the TRiazol-method (Gibco-BRL) gene, OR) and incubated at room temperature for1hinahumid chamber. and treated with DNase (Amplification Grade, Gibco/BRL) in 20 mmol/l Tris-

Nuclei of cells stained with anti–IDX-1 antibody were visualized with Hoechst HCl (pH 8.4), 2 mmol/l MgCl2, and 50 mmol/l KCl to remove any traces of 33242 dye (Sigma, St. Louis, MO). The percentage of IDX-1–containing cells contaminating genomic DNA. RNA (2.5 ␮g) was then subjected to reverse was evaluated by counting the number of IDX-1–positive cells divided by the transcription (RT) (RT reagents; Promega, Madison, WI). RT–polymerase chain total number of cells identified by nuclear staining. reaction (PCR) was undertaken in a volume of 50 ␮l of buffer containing 50 ␮ Insulin immunofluorescence, for co-immunostaining with IDX-1, was car- mmol/l KCl, 10 mmol/l Tris-HCl, 3.5 mmol/l MgCl2, 200 mol/l each dNTPs, and 0.4 ried out using the same primary antibody described in the previous paragraph, ␮mol/l each of sense and antisense primers to rat or human insulin (depending on whereas the secondary antibody was a goat anti-guinea pig IgG (Molecular the specific cell from which the RNA was extracted). Amplification was per- Probes). The cells were examined using a using a fluorescent microscope formed for 30 cycles at a denaturing temperature of 94°C for 1 min, an annealing (E-800; Nikon, Tokyo). temperature of 60°C for 45 s, and an extension temperature of 72°C for 1 min. For Staining for insulin and IDX-1 experiments was repeated at least three the amplification of GLUT2 and glucokinase mRNAs, we used the same PCR times using independent cell cultures. Hematoxylin and eosin staining was conditions described above in the presence of gene-specific primers. For ␤-actin, used in some experiments to show the morphological changes observed under the annealing temperature was raised to 64°C for 1 min, and gene-specific primers the experimental conditions hereafter described. were used. For GLP-1 receptor (GLP-R), the annealing temperature used was Measurement of insulin secretion. PANC-1 (parental, IDX-1–transfected, 55°C. All other experimental conditions to amplify GLUT2, glucokinase, and and neomycin-transfected) cells and ARIP (only parental) cells were plated at GLP-R mRNAs were identical to those described for the amplification of in- a density of 106 cells per well in a six-well plate. Once the cells reached 80% sulin mRNA. Primer sequences for human and rat insulin, GLUT2, glucokinase, of confluence, they were washed with serum-free medium containing 12 mmol/l GLP-R, and ␤-actin are presented in Table 1. RT and PCR conditions for glucose and exposed to fresh serum-free medium for an increasing length of time human transcripts were identical to those described for rat mRNAs. (0, 12, 24, 48, 72, and 96 h) in the presence of different concentrations of GLP-1 Southern blotting with species-specific full-length cDNA probes for insulin, (0.1, 1, 10, and 20 nmol/l) or vehicle. Glucose-dependent insulin secretion was GLUT2, glucokinase, GLP-R, and ␤-actin was performed as previously de- evaluated by culturing the cells in the presence of a determined concentration scribed (9). of GLP-1 (10 nmol/l for 72 h) with an increasing concentration of glucose in Flow cytometric analysis. For fluorescence-activated cell sorting (FACS) the culture medium (0, 0.1, 1, 3, 6, 10, and 20 nmol/l). Insulin released into the analysis, cells were cultured in the presence of GLP-1 (10 nmol/l) or vehicle medium was measured by radioimmunoassay (RIA) (Linco Research, St. for 72 h and then washed with cold PBS (pH 7.4; three times) and incubated Charles, MA). Total insulin accumulation in the culture medium was then overnight on ice in PBS with 2% paraformaldehyde. After centrifugation, the normalized for total cellular protein content per each individual culture. cell pellet was resuspended in 400 ␮l of cold 0.1% Triton diluted in FACS Protein assay. Total cellular protein content was measured using the buffer (PBS with 2% FCS). After several centrifugation cycles and washes, Bradford method (Bio-Rad, Richmond, CA). The amount of measured the cells were resuspended in the assay buffer with 10 ␮l of fluorescent-

786 DIABETES, VOL. 50, APRIL 2001 H. HUI, C. WRIGHT, AND R. PERFETTI

FIG. 1. Morphological changes of ARIP and PANC-1 cells treated with GLP-1. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomycin-resistant gene) were cultured in serum-free medium with or without GLP-1 (10 nmol/l) for 72 h and stained with hematoxylin and eosin. A and B represent ARIP cells cultured with vehicle (A) or medium containing GLP-1 (B). E and F represent PANC-1 cells transfected with human IDX-1 cultured in the absence (E) or in the presence (F) of GLP-1. conjugated insulin antibody (catalog number FM205, FITC; Chromaprobe, Moun- much less evident, effect on the appearance of individual ␮ tain View, CA) in the presence of 10 l blocking antibody (anti-mouse IgG; cells. Whereas naive ARIP cells characteristically grew as Organon Teknika-Cappel, West Chester, PA). Control samples were treated with PBS, without the primary antibody, and then incubated with an isotope- individual cells, forming a fine monolayer (Fig. 1A), treat- matched fluorescein isothiocyanate (FITC)-conjugated control antibody (mu- ment with GLP-1 promoted the aggregation of cells in rine IgG; Chromaprobe). Flow cytometric analysis was performed with a small clusters (Fig. 1B). This was not a function of cell FACScan[R] cytometer, using the LYSYS II program (10). Cell viability was evaluated by the Trypan blue dye (Gibco-BRL) exclusion technique. density: treatment with GLP-1 induced ARIP cells to ag- Statistical analysis. The data were expressed as means Ϯ SE. Significance gregate in patches even when plated at a very low density. of the data were evaluated by the unpaired Student’s t test. One-way analysis Although the majority of GLP-treated cells tended to grow of variance (ANOVA) was used to evaluate statistical significance when more in semispherical patches of cells, a small percentage of cells than two data points were analyzed. Statistical analyses by unpaired Student’s t test or ANOVA are explicitly identified in the text or in the figure legends. continued to grow as sparse and isolated cells, indicating that there was heterogeneity of response to GLP-1. The RESULTS morphology of individual cells revealed some additional, Morphological changes of ARIP cells induced by although less evident, changes. These included more irreg- GLP-1. Various changes in the morphology of ARIP cells ular and variable cell shapes, with a cytoplasm less homog- resulted from the treatment with GLP-1 (10 nmol/l) for enous and rougher in appearance than control culture. 72 h. GLP-1 primarily affected the relationship between cells Morphological changes of PANC-1 cells induced by within a given culture dish, with an additional, although GLP-1. Parental (nontransfected) PANC-1 cells did not re-

DIABETES, VOL. 50, APRIL 2001 787 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

FIG. 2. Immunocytochemistry for insulin of ARIP and PANC-1 cells treated with or without GLP-1. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomycin-resistant gene) were cultured with or without GLP-1 (10 nmol/l) for 72 h and subjected to immunostaining with an anti-human insulin antibody. A and B represent ARIP cells cultured in serum-free medium in the absence (A) or presence (B) of GLP-1. C and D represent nontransfected PANC-1 cells cultured in the absence (C) or presence (D) of GLP-1. E and F represent PANC-1/IDX-1 cells cultured in the absence (E) or presence (F) of GLP-1. PANC-1/IDX-1 cells, treated with GLP-1 (10 nmol/l for 72 h) and stained solely with secondary .(antibody, were used as a negative control (G). Rat pancreas was used as a positive control (H) (original magnification ؋200 spond to GLP-1 (10 nmol/l for 72 h) with any morpholog- Immunocytochemistry for insulin. Treatment with GLP-1 ical change (Fig. 1C for vehicle alone; Fig. 1D for GLP-1 induces the differentiation of ductal epithelial cells into treatment). Transfection with human IDX-1 induced clear insulin-producing cells. Figure 2 illustrates a series of cell changes in the shape of individual cells as well as in the cultures grown with or without GLP-1 (10 nmol/l) for 72 h. relationship between cells (Fig. 1D). IDX-1–transfected Using anti-insulin antibody, a positive immunoreactivity cells grew in patches rather than in isolation and were for insulin was detected in GLP-1–treated ARIP cells (Fig. surrounded by a large amount of extracellular matrix (Fig. 2B); in contrast, no insulin immunoreactivity was observed in 1E). Treatment of IDX-1–transfected PANC-1 cells with ARIP cells cultured with vehicle alone (Fig. 2A). Preab- GLP-1 (10 nmol/l for 72 h) further promoted the tendency sorption of the antibody with an excess of human recom- of forming aggregates with few cells (Fig. 1F). In addition, binant insulin prevented the staining of insulin-positive we observed that GLP-1 induced an increase in the size of GLP-1–treated cells (data not shown). Although treatment individual cells (Fig. 1E and F). with GLP-1 turned the majority of ARIP cells into insulin-

788 DIABETES, VOL. 50, APRIL 2001 H. HUI, C. WRIGHT, AND R. PERFETTI

FIG. 3. Immunofluorocytochemistry for IDX-1 of ARIP and PANC-1 cells treated with or without GLP-1. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomycin-resistant gene) were treated with or without GLP-1 (10 nmol/l) for 72 h and subjected to immunostaining with an anti–IDX-1 antibody (1) and with the nuclear dye Hoechst 33242 (2). A1/A2 and B1/B2 represent ARIP cells cultured in serum-free medium in the absence (A1/A2) or presence (B1/B2)of GLP-1. C1/C2 and D1/D2 represent nontransfected PANC-1 cells cultured in the absence (C1/C2) or presence (D1/D2) of GLP-1. E1/E2 and F1/F2 represent PANC-1/IDX-1 cells cultured in the absence (E1/E2) or presence (F1/F2) of GLP-1. G–I show PANC-1/IDX-1 cells stained for insulin alone (G), IDX-1 alone (H), or costained for both IDX-1 and insulin (I).

producing cells, a minority of cells never acquired these features in response to GLP-1 and never gained the ability to synthesize insulin. No positive insulin immunostaining was observed with parental PANC-1 cells cultured in the presence or absence of GLP-1 (Fig. 2C and D). Transfection of human PANC-1 cells with human IDX-1 was able to render these cells capable of synthesizing insulin when exposed to GLP-1 (Fig. 2F). In contrast, PANC-1 cells cultured with vehicle alone were in- sulin negative, even when transfected with human IDX-1 (Fig. 2E). A negative control was obtained by solely using the secondary antibody to stain a culture of PANC-1/IDX- 1–transfected cells cultured in the presence of GLP-1 (10 nmol/l) for 72 h (Fig. 2G). A section of rat pancreas was used as a positive control for insulin immunostaining (Fig. 2H). Immunofluorochistochemistry for IDX-1. ARIP and PANC-1 (parental- and IDX-1–transfected) cells were cul- tured as described for insulin immunostaining and sub- jected to immunofluorescence study for IDX-1. Control nuclear staining was performed for all culture conditions. Using anti–IDX-1 antibody, a positive immunoreactivity for IDX-1 was detected in ARIP cells treated with vehicle alone (Fig. 3, A1 and A2) or GLP-1 (Fig. 3, B1 and B2). No positive

DIABETES, VOL. 50, APRIL 2001 789 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

FIG. 4. Dose- and time-dependent insulin accumulation in the culture medium of ARIP cells. A: Dose-dependent insulin secretion into the medium of ARIP cells cultured in serum-free medium containing 12 mmol/l glucose in the presence or absence of various concentrations of GLP-1 for 72 h. First lane on the right: fresh unused culture medium; other lanes: the insulin level in the culture media from ARIP cells cultured in the presence of the indicated concentrations of GLP-1. B: Time course of insulin secretion in medium containing 12 mmol/l glucose in the presence of GLP-1 (10 nmol/l) for various lengths of time. Values represent the amount of insulin in the medium after an indicated time. Each experiment was repeated at least four times, and ؎ the data plotted on the graph represent the mean SD. Statistical FIG. 5. Dose- and time-dependent insulin accumulation in the culture significance of the data was evaluated by unpaired Student’s t test: medium of PANC-1 cells. A: Dose-dependent insulin secretion in the *P < 0.05, **P < 0.01; ***P < 0.0001. In A, the significance of the data medium of PANC-1 cells cultured in serum-free medium containing 12 was evaluated by comparison of each sample with untreated ARIP cells; mmol/l glucose in the presence of various concentrations of GLP-1. in B, the data were evaluated by analyzing the significance of the curve First lane on the right: fresh unused culture medium; other lanes: the itself by ANOVA. insulin level in the culture media from PANC-1 cells cultured in the presence of the indicated concentrations of GLP-1. B: Dose-dependent insulin secretion of PANC-1 cells transfected with human IDX-1 and IDX-1 staining was observed with parental PANC-1 cells cultured in serum-free medium containing 12 mmol/l glucose in the presence of various concentrations of GLP-1 for 72 h. First lane on the cultured with vehicle alone or GLP-1 (Fig. 3, C1, C2, D1, right: fresh unused culture medium; other lanes: the insulin level in the and D2). Transfection of human PANC-1 cells with human culture media from cells cultured in the presence of various concen- IDX-1 induced, as expected, the expression of the coun- trations of GLP-1. C: Time course of insulin secretion into the medium of PANC-1 cells transfected with human IDX-1 and cultured in serum- terpart protein (Fig. 3, E1, E2, F1, and F2). Treatment with free medium containing 12 mmol/l glucose in the presence of GLP-1 (10 GLP-1 (10 nmol/l for 72 h) promoted a further increase the nmol/l). Each experiment was repeated at least four times and the data expression level of IDX-1 for both ARIP and PANC-1/IDX- plotted on the graph represent the mean ؎ SD. Statistical significance of the data were evaluated by unpaired Student’s t test: *P < 0.05, 1–transfected cultures (Fig. 3, pictures B and F). It ap- **P < 0.01; ***P < 0.0001. No statistical significance was determined peared that GLP-1 increased the level of expression of for the data presented in A.InB, the significance of the data was evaluated by comparison of each sample with untreated PANC-1 cells; IDX-1 rather than number of IDX-1–expressing cells. in C, the data were evaluated by analyzing the significance of the curve Counting of 400 cells from several independent cultures of itself by ANOVA. ARIP and PANC-1/IDX-1 cells treated with GLP-1 or vehi- cle revealed that ϳ70% of ARIP and 100% of PANC-1/IDX-1 nmol/l for 72 h) demonstrated that the two proteins were cells expressed IDX-1. coexpressed under the experimental conditions described Double immunofluorescence for insulin and IDX-1 of (Fig. 3G, staining for insulin; Fig. 3H, staining for IDX-1; PANC-1/IDX-1–transfected cells cultured with GLP-1 (10 Fig. 3I, double immunostaining for insulin and IDX-1).

790 DIABETES, VOL. 50, APRIL 2001 H. HUI, C. WRIGHT, AND R. PERFETTI

FIG. 7. Northern blot analysis for islet cell transcripts. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomy- cin-resistant gene) were cultured in serum-free medium with or with- out GLP-1 (10 nmol/l for 72 h) and subjected to Northern blot analysis for the detection of IDX-1, insulin, and ␤-actin mRNAs. Rat pancreas was used as a positive control and rat kidney as a negative control. FIG. 6. Glucose-dependent insulin secretion. ARIP (A) and PANC-1 Lanes 1 and 2: ARIP cells cultured in the absence (1) or presence (2) cells (transfected with human IDX-1) (B) were cultured in serum-free of GLP-1; lanes 3 and 4: PANC-1 cells cultured in the absence (3)or medium with GLP-1 (10 nmol/l for 72 h), or vehicle, in the presence of presence (4) of GLP-1; lanes 5 and 6: PANC-1 cells transfected with various concentrations of glucose (0, 0.1, 1, 3, 6, 10, and 20 mmol/l). human IDX-1 and cultured in the absence (5) or presence (6) of GLP-1; Each experiment was repeated three times, and the data plotted on the lane 7: rat kidney; lane 8: rat pancreas. Each experiment was repeated .graph represent the mean ؎ SD. Insulin levels were normalized for three times using RNA samples from independent cultures total protein content in each individual sample of culture medium. Statistical significance of the data was evaluated by ANOVA. and to promote a dose-dependent accumulation of insulin Approximately 85% of IDX-1–containing cells were also in the culture medium (Fig. 5B). The time course of the positive for the presence of intracellular insulin. All insu- GLP-1 response revealed a peak secretion within 48 h, lin-containing cells were positive for IDX-1 staining; the with a plateau at 72 h and an early decline after 96 h from anti-insulin antibody did not stain some weakly IDX-1– the first exposure to GLP-1 (Fig. 5C). positive cells (ϳ20% of PANC-1/IDX-1 cells). Cell culturing in the presence of an increasing concen- Insulin release in the culture medium. ARIP cells cul- tration of glucose with a constant concentration of GLP-1 tured in the presence of GLP-1 exhibited a dose-dependent (10 nmol/l for 72 h), or vehicle, revealed that both ARIP response of insulin secretion (Fig. 4A). The minimum and PANC-1/IDX-1–transfected cells were able to release concentration of GLP-1 required to transform rat pancre- insulin in a glucose-dependent manner when exposed to atic (ARIP) ductal cells into insulin-producing cells was 1 GLP-1 (Fig. 6A and B). No insulin secretory response was nmol/l. A linear increase of insulin accumulation into the observed with wild-type PANC-1 cells cultured in the culture medium was observed with increasing doses, and presence of GLP-1 (10 nmol/l for 72 h) with an increasing a plateau of this response was detected with 20 nmol/l of concentration of glucose in the culture medium (data not GLP-1. Analysis of the time course of the insulin secretory shown). For both ARIP and PANC-1/IDX-1 cells, the lowest response of ARIP cells cultured in the presence of GLP-1 concentration of glucose required to induce the secretion (10 nmol/l) revealed that the maximal secretion was ob- of insulin was 3 mmol/l. A linear increase of insulin served at 48 h, with a plateau at 72 h, followed by an early accumulation into the culture medium was observed with decline at later time points (Fig. 4B). increasing doses (P Ͻ 0.001), and a plateau of this re- The results described above for ARIP cells treated with sponse was detected with glucose concentrations between GLP-1 were not confirmed when a different ductal cell line 10 and 20 mmol/l (Fig. 6A and B). (human PANC-1) was used to perform a similar set of Messenger RNA levels of ␤-cell–specific genes. Rat experiments. PANC-1 cells did not secrete insulin in (ARIP) and human (PANC-1) ductal epithelial cells were response to GLP-1 (Fig. 5A). However, cellular transfec- subjected to Northern blot analysis for detection of IDX-1, tion of PANC-1 cells with the human ␤-cell differentiation insulin, and ␤-actin mRNA levels. Whereas ARIP cells factor IDX-1 rendered the human ductal cells capable of showed that the IDX-1 gene was constitutively tran- responding to GLP-1 and induced the synthesis and secre- scribed, PANC-1 cells were IDX-1–negative (Fig. 7). Hy- tion of insulin. A 1 nmol/l concentration of GLP-1 in the bridization of the same blot with insulin cDNA probe was culture medium was required to induce insulin secretion negative for both ARIP and PANC-1 cells (Fig. 7).

DIABETES, VOL. 50, APRIL 2001 791 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

FIG. 8. RT-PCR analysis of islet ␤-cell–specific transcripts. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomycin-resistant gene) were cultured in serum-free medium and treated with GLP-1 (10 nmol/l), vehicle alone, or exendin-9 (100 nmol/l) and subjected to RT-PCR for insulin, glucokinase (GK), the ␤-cell glucose transporter GLUT2, and ␤-actin. A: Lane 1: ARIP cells treated with vehicle alone; lane 2: ARIP cells cultured in the presence of exendin-9 (100 nmol/l) for 72 h; lanes 3–6: ARIP cells treated with GLP-1 for 24 h (lane 3), 48h(lane 4), 72 h (lane 5), and 96 h (lane 6). B: Lanes 1 and 2: PANC-1 cells treated with vehicle alone (lane 1) or GLP-1 (lane 2); lanes 3–7: PANC-1 cells transfected with human IDX-1 and treated with exendin-9 for 72 h (lane 3) or GLP-1 for0h(lane 4), 24 h (lane 5), 48 h (lane 6), 72h(lane 7), and 96 h (lane 8). Each experiment was repeated three times using RNA samples from independent cultures.

To analyze the ability of GLP-1 to induce the differenti- after the appearance of insulin mRNA, indicating that the ation of ductal cells into insulin-producing cells, we began glucose-sensing ability of insulin-secreting cells was a late RT-PCR analysis to achieve maximum sensitivity for the event in the differentiation process. Glucokinase mRNA detection of insulin mRNA, as well as the mRNA for other levels remained unchanged after the 72-h detection. No ␤-cell–specific transcripts. RT-PCR analysis was per- RT-PCR products were detected in the negative control or formed using RNA isolated from at least five different in non–GLP-1–treated cells. This leaves the sequence of cultures for each experimental condition, and each PCR as GLUT2 at 24 h, insulin at 48 h, and was repeated more than two times. glucokinase at 72 h. RT-PCR for ␤-actin was used as a RT-PCR analysis of ARIP cells demonstrated that the control for RNA loading. lowest concentration of GLP-1 required for the initial RT-PCR for insulin using RNA extracts obtained from detection of insulin mRNA was that of 1 nmol/l (data not PANC-1 cells treated for 96 h with GLP-1 (10 nmol/l) were shown). To evaluate the time course of response to GLP-1, consistently negative (Fig. 8B). Transfection of PANC-1 ARIP cells were cultured in serum-free medium with cells with human IDX-1 was not able per se to induce GLP-1 (10 nmol/l) for an increasing length of time (Fig. insulin gene transcription; however, it was sufficient to 8A). A clear band at 187 bp corresponding to rat insulin I render PANC-1 cells responsive to GLP-1. The lowest dose and II mRNA was detected first in ARIP cells cultured with of GLP-1 required to detect insulin mRNA was 1 nmol/l GLP-1 for 48 h. This was followed by a plateau at 72 h, with (data not shown). This was followed by a progressive insulin mRNA levels remaining constant up to 96 h after dose-dependent increase of insulin mRNA levels, reaching the first GLP-1 exposure. No RT-PCR products were a peak at 10 nmol/l and a plateau at 20 nmol/l (data not detected in the negative control or in non–GLP-1–treated shown). The time course of response in PANC-1 cells cells. RT-PCR for GLUT2 revealed the presence of this transfected with human IDX-1 and cultured with GLP-1 (10 transcript (343 bp) after only 24 h of treatment with GLP-1, nmol/l) revealed the earliest detection of insulin after 48 h, preceding the earliest detection of insulin mRNA by ϳ1 followed by a plateau at 72 h. RT-PCR for ␤-actin was used day. After the initial expression at 24 h, GLUT2 mRNA as a control for RNA loading. The pattern of expression for remained constant over time, in a fashion similar to GLUT2 and glucokinase mRNAs was similar to that de- insulin. Glucokinase mRNA was detectable at 72 h, 24 h scribed for ARIP cells treated with GLP-1, with GLUT2

792 DIABETES, VOL. 50, APRIL 2001 H. HUI, C. WRIGHT, AND R. PERFETTI

FIG. 9. GLP-R expression in ARIP and PANC-1 cells. ARIP and PANC-1 cells (transfected with human IDX-1 or solely with the neomycin-resistant gene) were cultured in serum-free medium and treated with GLP-1 (10 nmol/l), or vehicle alone, for 72 h. RT-PCR for GLP-R and ␤-actin was performed with a cycle-titration PCR (15, 20, 25, and 30 PCR cycles) to assess the relative abundance of GLP-R transcript. Primers for both transcripts were added together in a single PCR for each individual sample. A: ARIP cells cultured with vehicle alone or GLP-1; B: PANC-1 cells cultured with vehicle alone or GLP-1; C: PANC-1/IDX-1 cells culture with vehicle alone or GLP-1. Each experiment was repeated four times, using RNA samples from independent cultures. The blots on the left represent one individual experiment, whereas the graphs on the right are the average of four independent experiments. The numeric values in the graphs represent the ratio of GLP-R mRNA to actin mRNA. preceding the earliest insulin gene expression by 24 h and the experimental conditions described in the present study glucokinase appearing only 24 h after the first detection of was specifically induced by GLP-1. insulin mRNA. This leaves the sequence of ␤-cell–specific Detection of GLP-R by RT-PCR with gene-specific prim- genes in PANC-1 as GLUT2 at 24 h, insulin at 48 h, and ers revealed that both ARIP and PANC-1 cells constitu- glucokinase at 72 h. tively (i.e., before the treatment with GLP-1 or transfection Treatment of either ARIP or PANC-1/IDX-1 cells with with the IDX-1 gene) expressed the receptor for GLP-1. the GLP-R antagonist exendin-9 inhibited the expression Cellular transfection of PANC-1 cells with the IDX-1 gene of GLUT2, insulin, and glucokinase (Fig. 8A and B), dem- increased the mRNA level for GLP-R (P Ͻ 0.01 comparing onstrating that the ␤-cell–like phenotype observed under PANC-1/IDX-1 cells with either wild-type PANC-1 or ARIP

DIABETES, VOL. 50, APRIL 2001 793 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

FIG. 10. Flow cytometric analysis. ARIP and PANC-1 (parental- and IDX-1–transfected) cells were treated with GLP-1 (10 nmol/l) or vehicle for 72 h and stained with fluorescent antibodies against insulin. A–G show a representative experiment for each individual ex- perimental condition of five replicate analyses that were performed. A and B: ARIP cells cultured in the presence of vehicle (A) or GLP-1 (B). C and D: parental PANC-1 cells cultured in the presence of vehicle (C) or GLP-1 (D). E and F: PANC-1 cells transfected with the human IDX-1 gene cultured in the presence of vehicle (E) or GLP-1 (F). G: PANC-1 cells cultured in the presence of GLP-1 and stained solely with an unspecific isotype-matched FITC-con- jugated antibody without the primary antibody. A summary of independent experiments is presented in the lower section of the figure. cells after 25 PCR cycles). Treatment with GLP-1 (10 tion with IDX-1 was not able, per se, to differentiate nmol/l for 72 h) promoted a further modest increase (not PANC-1 cells into insulin-producing cells; however, treat- statistically significant) in GLP-R mRNA levels in both ment with GLP-1 (10 nmol/l for 72 h) converted 61.7% of transfected and nontransfected cells (Fig. 9). the cultured cells from insulin-negative to insulin-positive Flow cytometric analysis. The ability of GLP-1 to pro- (Fig. 10). FACS analysis of PANC-1 cells treated with mote the differentiation of IDX-1–positive ductal epithelial GLP-1 and stained solely with an unspecific isotype- cells into insulin-synthesizing cells was further confirmed matched FITC-conjugated antibody, without the primary by FACS analysis. This procedure was used to provide a antibody, was used as a negative control (Fig. 10). quantitative measure of the ability of GLP-1 to induce the differentiation of insulin-secreting cells. ARIP cells cultured for 72 h in the presence of GLP-1 (10 DISCUSSION nmol/l) were able to transcribe and translate the insulin The present study demonstrates that the treatment of gene, such that 72.6% of them reacted with an anti-human pancreatic ductal epithelial cells with the gastrointestinal insulin antibody, demonstrating that they were able to incretin hormone GLP-1 promotes their differentiation into synthesize insulin (Fig. 10). PANC-1 cells cultured in the pancreatic ␤-like cells. GLP-1 requires the gene expression presence or absence of GLP-1 showed that only 1.4% of the of the islet differentiation factor IDX-1 to exert its differ- entire culture population contained insulin in the cyto- entiation promoting activity. plasm—a percentage equivalent to the background level of Insulin, GLUT2, and glucokinase mRNAs (the three the assay. Stable transfection with human IDX-1, although main gene transcripts that define the physiology of normal not capable of inducing cells to synthesize insulin, was ␤-cells) were transcribed by epithelial ductal cells after sufficient to render them responsive to GLP-1. Transfec- exposure to GLP-1. In conjunction with the induction of

794 DIABETES, VOL. 50, APRIL 2001 H. HUI, C. WRIGHT, AND R. PERFETTI the gene expression for insulin, we showed that GLP-1– genetically susceptible to acquiring that phenotype, and treated cells contain and secrete the counterpart protein, IDX-1 is a key player in this process. as demonstrated by immunocytochemistry, RIA, and FACS Our interest in studying pancreatic ductal cells derives analysis. GLUT2 was the first ␤-cell–specific transcript from studies demonstrating that it is from this cell type we detected in GLP-1–treated pancreatic ductal cells. that pancreatic endocrine cells derive (1,7,13). This has This was followed by insulin and, finally, by the glucose- been proposed to occur both during pancreatic organo- phosphorylating enzyme glucokinase. The specificity of genesis and islet cell proliferation after injury in mature the effect of GLP-1 was validated by experiments demon- animals. In adult animals, the removal of 90% of the strating that treatment of cells (ARIP or PANC-1) with the pancreas (8,14), or the treatment with the toxic agent streptozotocin (15,16), are two known models to induce GLP-R antagonist exendin-9 inhibited the expression of ␤ ␤-cell–specific genes. hyperglycemia by dramatically reducing the islet -cell mass. In both of these experimental models, the destruc- The mechanisms regulating proliferation and differenti- tion of islet mass is followed by a compensatory attempt to ation of the pancreatic hormone-producing cells and the replace the normal population of insulin-secreting cells. chronology of these biological events are still largely New ␤-cells are formed from existing islets and from undetermined. The sequence of events hereby described ductal epithelial cells (7). The latter source has greater leads us to speculate that the ability to regulate glucose intrinsic biological relevance. Indeed, the possibility of uptake by the islet-specific glucose transporter GLUT2 is differentiating insulin-secreting cells from nonendocrine the first step necessary for the “sensitization” of the cells supports the hypothesis that the biological source regulatory region(s) of the insulin gene to glucose. This (pancreatic ductal epithelium) for this compensatory would then promote the transcription of insulin mRNA. mechanism may be present even in the setting of a GLP-1–dependent activation of IDX-1 would further com- generalized destruction of the entire population of islet mit these cells toward a ␤-cell–like pathway of differenti- ␤-cells. This is strongly supported by recent studies dem- ation by inducing the synthesis of glucokinase, the chief onstrating that primary cultures of epithelial ductal cells element of the “glucose-sensing machine” of the islets of (from human and mouse pancreas) are susceptible to Langerhans. undergoing differentiation into endocrine cells (17,18). The homeodomain protein IDX-1 is an insulin gene In the normal ductal epithelium, it remains to be deter- expressed in the early pancreatic mined whether there are different populations of cells, gland of the embryo. During pancreatic islet development, some of which are capable of differentiating into endo- IDX-1 plays an important role in determining islet cell crine cells, whereas others have merely a structural role in differentiation (11). It is the early IDX-1 gene expression defining the epithelial wall. Alternatively, it could be during embryogenesis, coupled with the activation of speculated that all pancreatic ductal epithelial cells could other transcription factors (i.e., NeuroD/Beta 2, Pax 4, represent a not-fully-differentiated population of cells ca- etc.), that determine the pancreatic endocrine hormone pable of acquiring a new phenotype under specific stimuli. production (5,6). In adult (mature) animals, the expression In summary, pancreatic ductal cell lines are capable, of IDX-1 is repressed in the majority of pancreatic cells, under specific stimuli, of converting into pancreatic endo- with the exception of the ␤- and ␦-cells (somatostatin- crine cells. In this study, we demonstrated that human secreting cells) of the islets of Langerhans (12). GLP-1 is capable, when acting on IDX-1–positive cells, of ␤ In this study, we demonstrated that only those pancre- promoting a -cell–like phenotype. This model system atic epithelial cells that express IDX-1 are susceptible to may provide a basis for elucidating the minimum biologi- cal requirements for a non–␤-cell to become a fully func- undergoing differentiation into insulin-secreting cells once ␤ they are treated with GLP-1. Interestingly, the overexpres- tioning -cell. sion of human IDX-1, by means of stable cellular transfec- tion, is not sufficient per se to induce the differentiation of ACKNOWLEDGMENTS these cells into a ␤-cell–like phenotype. It is only when This study was supported in part by the American Feder- IDX-1–positive cells are exposed to GLP-1 that they ac- ation for Aging Research. quire the ability to synthesize insulin. Similarly, it is not the We would like to thank Rita Velikina and Dr. Run Yu for presence of receptors for GLP-1 per se that allows the their technical support. We are very grateful to Patricia differentiation of GLP-1–treated cells into insulin-secreting Merkel for the critical reading of the manuscript. cells. Indeed, although only parental ARIP cells were able to differentiate into insulin-secreting cells after treatment REFERENCES with GLP-1, both ARIP and PANC-1 cells constitutively 1. Teitelman G, Lee JK: Cell lineage analysis of pancreatic islet cell develop- ment: glucagon and insulin cells arise from catecholaminergic precursor expressed receptors for GLP-1. PANC-1 cells responded to present in the pancreatic duct. Dev Biol 121:454–466, 1987 GLP-1 by means of differentiation into insulin-secreting 2. Pictet RL, Clark WR, Williams RH, Rutter WJ: An ultrastructual analysis of cells only when transfected with human IDX-1. A potential the developing embryonic pancreas. Dev Biol 29:436–467, 1972 interplay between IDX-1 and GLP-1 is further suggested by 3. Jonsson J, Carlsson L, Edlund T, Edlund H: Insulin-promoter factor 1 is required for pancreas development in mice. Nature 371:606–609, 1994 our data demonstrating that treatment with GLP-1 in- 4. Ahlgren U, Pfaff SL, Jessell TM, Edlund T, Edlund H: Independent creases IDX-1 mRNA levels and transfection with IDX-1 requirement for ISL1 in formation of pancreatic mesenchyme and islet induces an increase in GLP-R mRNA levels. Although cells. Nature 385:257–260, 1997 these findings may require additional studies for further 5. Naya FJ, Huang HP, Qiu Y, Mutoh H, DeMayo FJ, Leiter AB, Tsai MJ: Diabetes, defective pancreatic morphogenesis, and abnormal enteroendo- characterization, they clearly indicate that GLP-1 is able to crine differentiation in BETA2/NeuroD-deficient mice. Genes Dev 11:2323– induce a ␤-cell–like phenotype only in cells that are 2334, 1997

DIABETES, VOL. 50, APRIL 2001 795 GLP-1 IS A DIFFERENTIATION FACTOR FOR ISLET ␤-CELLS

6. Sosa-Pineda B, Chowdhury K, Torres M, Oliver G, Gruss P: The Pax4 gene 12. Habener JF, Stoffers DA: A newly discovered role of transcription factors is essential for differentiation of insulin-producing ␤ cells in the mamma- involved in pancreas development and the pathogenesis of diabetes lian pancreas. Nature 386:399–402, 1997 mellitus. Proc Assoc Am Phys 110:12–21, 1998 7. Bonner-Weir S, Baxter LA, Schuppin GT, Smith FE: A second pathway for 13. Bouwens L: Transdifferentiation versus stem cell hypothesis for the regeneration of adult exocrine and endocrine pancreas: a possible reca- regeneration of islet ␤-cells in the pancreas. Microsc Res Tech 43:332–336, pitulation of embryonic development. Diabetes 42:1715–1720, 1993 1998 8. Xu G, Stoffers DA, Habener JF, Bonner-Weir S: Exendin-4 stimulates both 14. Jonas JC, Sharma A, Hasenkamp W, Ilkova H, Patane G, Laybutt R, ␤ ␤ -cell replication and neogenesis, resulting in increased -cell mass and Bonner-Weir S, Weir GC: Chronic hyperglycemia triggers loss of pancreatic improved glucose tolerance in diabetic rats. Diabetes 48:2270–2276, 1999 ␤ cell differentiation in an animal model of diabetes. J Biol Chem 9. Wang Y, Perfetti R, Greig NH, Holloway HW, DeOre KA, Montrose- 274:14112–14121, 1999 Rafizadeh C, Elahi D, Egan JM: Glucagon-like peptide-1 can reverse the 15. Finegood DT, Weir GC, Bonner-Weir S: Prior streptozotocin treatment age-related decline in glucose tolerance in rats. J Clin Invest 99:2883–2889, 1997 does not inhibit pancreas regeneration after 90% pancreatectomy in rats. 10. Elstner E, Linker-Israeli M, Le J, Umiel T, Michl P, Said JW, Binderup L, Am J Physiol 276:E822–E827, 1999 ␤ Reed JC, Koeffler HP: Synergistic decrease of clonal proliferation, induc- 16. Wang RN, Bouwens L, Kloppel G: -Cell growth in adolescent and adult tion of differentiation, and apoptosis of acute promyelocytic leukemia cells rats treated with streptozotocin during the neonatal period. Diabetologia after combined treatment with novel 20-epi vitamin D3 analogs and 9-cis 39:548–557, 1996 retinoid acid. J Clin Invest 99:349–360, 1997 17. Ramiya VK, Maraist M, Arfors KE, Schatz DA, Peck AE, Conrnelius JG: 11. Kaneto H, Miyagawa J, Kajimoto Y, Yamamoto K, Watada H, Umayahara Y, Reversal of insulin-dependent diabetes using islets generated in vitro from Hanafusa T, Matsuzawa Y, Yamasaki Y, Higashiyama S, Taniguchi N: pancreatic stem cells. Nat Med 6:278–282, 2000 Expression of heparin-binding epidermal growth factor-like growth factor 18. Bonner-Weir S, Taneja M, Weir GC, Tatarkiewicz K, Song KH, Sharma A, during pancreas development: a potential role of PDX-1 in transcriptional O’Neil JJ: In vitro cultivation of human islets from expanded ductal tissue. activation. J Biol Chem 272:29137–29143, 1997 Proc Natl Acad SciUSA97:7999–8004, 2000

796 DIABETES, VOL. 50, APRIL 2001