A Smad Signaling Network Regulates Islet Cell Proliferation

New pancreatic β-cells form in response to a loss of β-cell mass, and this formation of new β-cells is thought to be an important defense against the evolution and progression of diabetes (1). β-cell replication (1,2) and neogenesis (3–5) are the two main mechanisms that have been proposed for the formation of new β-cells. β-cell replication is generally accepted as the predominant mechanism for the generation of new β-cells in the adult islet (1), but it remains unclear how such a terminally differentiated cell can undergo proliferation (6). Specifically, although much is known about β-cell cycle regulators and responses to growth factors, very little is known about the molecular mechanics of β-cell proliferation.

Transforming growth factor-β (TGF-β) superfamily signaling has been strongly implicated in pancreatic development and postnatal growth (7,8). TGF-β signaling seems to favor endocrine cell lineage selection and maturation and seems to be antiproliferative (9). TGF-βs, growth differentiation factor 11, and activins have all been closely tied to pancreatic endocrine development (10–12) and are all thought to work through the intracellular mediators smads2 and 3 (13). Specifically, smad2/3 have been associated with pancreatic differentiation toward an endocrine phenotype (14,15). Smad activity can be blocked by inhibitory smad6 or smad7 (16,17). Smad6 is thought to specifically inhibit those smads that are canonically downstream of bone morphogenetic protein signaling (i.e., smads 1, 5, and 8) (16). Smad7 seems to be more globally active against all receptor-activated smads (i.e., smads 1, 5, and 8 plus smad2/3) (16,17). Thus, effects specifically attributable to smad7, but not to smad6, are likely due to inhibition of smad2 and/or -3 (15).

In the current study, we examined the role of smad2/3 signaling, downstream of the TGF-β receptor II (TGFβrII), along with their inhibitor, smad7, in β-cell growth after a nondiabetogenic loss of β-cells (60% partial pancreatectomy [PPx]) (18). The phosphorylated (active) forms of smad2 and smad3 (psmad2/3) are strongly present specifically in the nucleus of islet cells, but become rapidly downregulated after PPx, specifically in proliferating cells. Mice with islet-specific smad2 and -3 gene ablation had a more robust proliferative islet cell response after pancreatectomy. Interestingly, many proliferating β-cells rapidly became smad7-positive, but also became insulin-negative, and appeared to become positive for the pancreatic polypeptide (PP)-fold hormones, perhaps representing a specific dedifferentiation of β-cells as part of the proliferative process. Mice with pancreatic or β-cell–specific ablation of smad7 had little β-cell proliferation after pancreatectomy. Such dedifferentiation of β-cells has been recently described as a possible mechanism leading to β-cell failure (19). Thus, TGF-β signaling, and smad7 in particular, may represent a key mechanistic link between normal β-cell proliferation and eventual β-cell failure.

All animal experiments were performed in accordance with guidelines established by the International Animal Care and Use Committee. Smad3-exon2–null mutant mice were obtained from The Jackson Laboratory. Transgenic mice expressing Smad2^fx/fx were gifts from Dr. Erwin Bottinger (Mt. Sinai School of Medicine). Transgenic mice expressing TGF-β receptor I (TGFβrI)^fx/fx and TGFβrII^fx/fx were gifts from Prof. Stefan Karlsson (University of Lund, Lund, Sweden). Rosa LacZ mice (R26R^lacZ) have been described previously by Soriano (20) and were obtained from The Jackson Laboratory. Insulin2-cre mouse have been previously described by Postic et al. (21). Glucagon-cre (22) was obtained from the Mutant Mouse Regional Resource Centers. All transgenic mice were crossed with Pdxcre-ER (23) (Mouse Models of Human Cancers Consortium), Ngn3cre (24), and PTF1a-cre (25) (Mouse Models of Human Cancers Consortium).

In experiments with the cre-ER/LoxP system, tamoxifen (Sigma-Aldrich, St. Louis, MO) was dissolved at 20 mg/ml in corn oil (Sigma-Aldrich) and administered into adult mice intraperitoneally, 2 mg/40 g/day for 5 days, in order to induce cre recombination, and we then performed PPx 7 days after the first tamoxifen injection. When Pdxcre-ER was crossed with TGFβrI^fx/fx or TGFβrII^fx/fx, PPx was performed 4 weeks after the first tamoxifen injection due to the long half-life of TGFβrI^fx/fx and TGFβrII^fx/fx, with the same 5-day tamoxifen regimen being followed.

PPx was conducted as described previously (18).

Insulin guinea pig 1:500 (DakoCytomation, Carpinteria, CA), glucagon rabbit monoclonal 1:2000 (Linco), Glut2 goat-polyclonal 1:50, smad7 rabbit polyclonal IgG 1:50, TGFβrI (ALK5) rabbit polyclonal anti-human 1:75, ALK1 rabbit polyclonal anti-human 1:40 (Santa Cruz Biotechnology, Santa Cruz, CA), somatostatin rat monoclonal 1:100 (LSBio), TGF-β1 ligand mouse monoclonal antibody 1:100 (R&D Systems), TGFβrII rabbit polyclonal antibody 1:100 (Millipore), p27Kip1 mouse monoclonal antibody 1:100 (Thermo Fisher Scientific), PP guinea pig antirat 1:400 (Millipore), neuropeptide Y (NPY) rabbit polyclonal antibody 1:500 (Millipore), peptide YY (PYY) rabbit polyclonal antibody 1:500 (Abcam), PDX-1 goat polyclonal 1:1000 (Abcam), Dolichos biflorus agglutinin fluorescein isothiocyanate–conjugated 1:100 (Vector Laboratories, Burlingame, CA), pSmad3 (1:400), pSmad2 (1:50), and pSmad2/3 (1:500) all rabbit polyclonal (Santa Cruz Biotechnology), Ki67 goat polyclonal 1:100 (Santa Cruz Biotechnology), and anti–BrdU rat monoclonal antibody 1:400 (Abcam) were used. Primary antibodies were incubated for 2 h at room temperature or at 4°C overnight. Biotinylated Vectastain ABC kit or AMCA/CY3/FITC fluorescent-conjugated donkey secondary antibodies (Vector Laboratories) were used for 1.5 h at room temperature. Immunoperoxidase was detected by DAB kit (DakoCytomation) or AEC (Sigma-Aldrich), and fluorescently labeled samples were imaged using a fluorescent microscope. Tissue sections were viewed on an upright AxioImager Z1 microscope (Carl Zeiss) or with an inverted Olympus Fluoview-1000 confocal microscope (Olympus) to confocally image the tissue sections. Images were captured with the AxioCamMRc5 and processed using AxioVs40V4.8.2.0 software (Carl Zeiss).

Adult mice were injected with BrdU (Sigma-Aldrich) 200 mg/kg intraperitoneally after performing PPx and subsequently once a day for 7 consecutive days. The pancreas was harvested 1 week postsurgery, fixed in 4% paraformaldehyde overnight in 4°C, and then placed in 30% sucrose overnight in 4°C. Antigen retrieval was done on the slides by treating it with 2 M HCl for 35 min followed by overnight incubation with primary antibodies. To quantify the number of BrdU-positive cells per islet in wild-type and transgenic pancreases, the whole frozen pancreas was sectioned 6-μm thick at −23°C in a cryostat and mounted on gelatin-coated glass microscope slides (Superfrost Plus; Fisherbrand), eight sections per slide. Two random slides were picked, and the average number of BrdU-positive cells per islet was determined over 20 islets using the image tracing software (Stereo Investigator; MBF Bioscience) to count and tag individual BrdU-positive/insulin-positive cells. Using the Stereo Investigator program, individual BrdU-positive cells were tagged and linked to Neurolucida software (MBF Bioscience) to quantify the number of cells.

Based on the fact that the smad2/3/7 network seemed to be a potentially important regulator of embryonic pancreatic endocrine development (7,8,12,26,27), we hypothesized that this network may also play a role in islet cell growth post-PPx. We chose a 60% PPx to avoid the hyperglycemia that might have secondary affects on islet cell physiology and behavior (18). Most unperturbed wild-type pancreatic islet cells stained strongly in the nucleus for psmad2/3 (Fig. 1A). After PPx, we gave daily injections of BrdU for 1 week. In this study, we saw that many islet cells had become negative for psmad2/3, and many of those psmad2/3–negative islet cells were BrdU⁺ (Fig. 1B–D). Only rarely were psmad2/3⁺ cells also BrdU⁺. One week after PPx, 82 ± 2.5% (± SEM) of psmad2/3–negative endocrine (as defined by a nonendothelial appearance) or psmad2/3 weak cells in islets are BrdU⁺. This value is compared with 18.2 ± 2.6% of psmad2/3–positive cells being BrdU⁺. In addition, we noted that there was downregulation within 24 h of PPx of the phosphorylated (active) form of TGFβrII (pTGFβrII) in the islet such that many insulin⁺ cells that had been pTGFβrII⁺ prior to PPx had become pTGFβrII⁻ (Fig. 1Eand F). To determine whether endogenous smad7 may be mediating the suppression of smad2/3 in these BrdU⁺ islet cells, we performed immunostaining for smad7. At baseline in 10-week-old pancreas, there was only some ductal expression of smad7 (Supplementary Fig. 1A and B), but little or no expression in the islets or acinar tissue (Fig. 1G) (28). Interestingly, we found that smad7 expression appeared very early after PPx (within 24 h) in a subset of pancreatic islet cells (Fig. 1H). A spectrum of smad7 expression was seen in different islets, with some islets having over half of the cells smad7⁺. Overall, 95.3 ± 0.11% (n = 3) of islets showed smad7⁺ cells at 1 week post-PPx. In addition, we found that these smad7⁺ islet cells stained only weakly for p27, a cell-cycle inhibitor (Supplementary Fig. 1C). The smad7⁺ cells also stained for Ki67 and thus had entered the cell cycle (Supplementary Fig. 1D). Since both the smad7 and psmad2/3 antibodies were raised in rabbit, we were unable to confirm that the smad7 cells were smad2/3-negative. However, since smad2/3 cells were infrequently proliferative, whereas most smad7⁺ cells were proliferative, it seems likely that for the most part, the smad7⁺ cells were psmad2/3–negative. Furthermore, smad7 is a well-established inhibitor of smad2/3. Smad7 expression persisted in some islets for at least 4 weeks post-PPx (Fig. 1J). The smad7⁺ cells were sometimes localized to the outer (mantle) region of the islet, but could also be seen throughout many islets (Fig. 1H–L). In order to confirm that this smad7 expression acted as an inhibitor of TGF-β signaling in these islets, we stained for the ligand TGF-β1 (Fig. 1Mand O). Similar to previous reports (29), TGF-β1 was found to be distributed evenly throughout the islet at baseline (Fig. 1M). However, 24 h post-PPx, TGF-β1 appeared to accumulate extracellularly in the immediate region of smad7⁺ cells (Fig. 1O, compare with smad7 staining in the immediately adjacent histologic section in Fig. 1N). Such an extracellular accumulation of TGF-β ligand in the vicinity of inhibited signal transduction, presumably in part due to smad7 expression in TGF-β target cells, has been described in other TGF-β signaling systems (30).

In order to better characterize these smad7⁺ cells in the islet post-PPx, we performed confocal triple immunofluorescence for smad7 and various islet cell–specific proteins. Smad7⁺ cells never costained with glucagon, insulin, PDX-1, somatostatin, nor Glut2 (Fig. 2A–E). However, surprisingly, essentially all of the smad7⁺ cells stained positive for PP (Fig. 2F–H). In addition, the three members of the PP-fold group of peptides (PP, PYY, and NPY) have together been implicated as markers of immature or, perhaps in this case, dedifferentiated pancreatic endocrine cells (22,31–33). For PYY, there were occasional PYY⁺/PP⁻ cells (Fig. 2I–K). When we stained for these three peptides (PP, PYY, and NPY), along with smad7, we noted that there was essentially complete overlap of expression of smad7 with both PP and NPY (Fig. 2L–N). PP staining has caused some controversy in the past (32,34) because of the possible cross-reactivity with NPY and PYY, although the newer antibodies are purported to be more specific. Thus, we performed preincubations of the antibodies either with their cognate peptide or with the other two peptides. For all three antibodies, preincubation with its own cognate peptide blocked staining, whereas preincubation with the other two peptides had no effect on staining, thus implying that there was not significant cross-reactivity of these newer antibodies (Supplementary Figs. 2 and 3). We also have recently published successful staining for PP in embryonic and early postnatal pancreas (28).

We hypothesized that these smad7⁺ cells may be β-cells that recently turned off insulin, even within 24 h post-PPx. This possibility would be consistent with lineage-tagging studies that found that new β-cells during regeneration derive only from pre-existing β-cells (1,2), but might suggest that in order for a highly differentiated mature cell such as a β-cell to undergo cell division, it may have to temporarily dedifferentiate. To test whether the smad7⁺ cells represent recently insulin⁺ cells that turned off insulin, we performed a lineage-tracing strategy in the PPx model to trace the origin of these smad7⁺cells. We performed PPx on mice that were double-positive for Ins2-Cre and R26R-lsl-tomato-red reporter (21,35). The reporter activation frequency of Ins2-cre in β-cells has been reported to be 88% (19). Pancreatectomy was performed in 10-week-old mice and the pancreas harvested 1 week later (Fig. 3A–D). In the Ins2-Cre;R26R-lsl-tomato red mice, these insulin-negative cells would be lineage-tagged and still be tomato-red⁺. We found 1 week post-PPx that the smad7⁺ cells were again insulin-negative as seen earlier, but importantly, smad7⁺ cells were tomato red–positive in Ins2-cre;R26R-lsl-tomato mice (Fig. 3A–D). These results strongly suggest that the smad7⁺ cells had recently been positive for insulin (Fig. 3A–D). Thus, presumably as a result of PPx, some insulin⁺ cells turned off insulin and became positive for smad7, and, based on the fact that all or most of the smad7⁺ cells are PP⁺, the former insulin-positive cells also turned on PP. To investigate this latter point, we stained for PP using Ins2-Cre;R26R-lsl-tomato red mice and harvested 1 week post-PPx. There we also found PP⁺ staining with tomato red cells, indicating that insulin⁺ β-cells appear to have turned into PP⁺ cells (Fig. 3E–H).

Recently, glucagon⁺ α-cells have been identified as a potential source of new β-cells in the islet in models of severe β-cell injury (5). To test whether smad7⁺ cells may have derived from α-cells that recently turned off glucagon, we crossed a glucagon-cre mouse (22) with an R26R-lsl-tomato red reporter mouse strain (22,36). One week post-PPx in these glucagon-cre;R26R-lsl-tomato mice, we did not see any tomato-labeled smad7⁺ cells, implying that the smad7⁺ cells in the islet did not derive from α-cells that had turned off glucagon expression (Fig. 3I).

In order to determine whether suppression of psmad2 and psmad3 plays a key role in islet cell proliferation post-PPx, we obtained smad2-conditional (Smad2^fx/fx) (37) and smad3 global mutant (exon 2 deletion) mice (38). Smad3^−/− (exon 2) global null mutant mice are viable and fertile (38), and to our knowledge, no smad3 conditional-mutant mice exist. Smad2^fx/fx mice were crossed with a pdx1-cre-ERT mouse to create tamoxifen-inducible smad2 conditional mutants for the pancreas. We then performed PPx 1 week after tamoxifen treatment in pdx1-cre-ERT;smad2^fx/fx mice, smad3^−/− global null mutant mice, and double smad2/3 mutant mice. For the pdx1-cre-ERT;smad2^fx/fx mice, 1 week post-PPx, there was a 50% increase in the number of BrdU-labeled islet cells (Fig. 4A–D, quantification in Fig. 5G). In smad3^−/− mutants, there was a tripling of the number of BrdU⁺ islet cells (Figs. 4Eand F and 5G). At baseline in the smad2 mutants, there was a slightly increased proliferation rate compared with littermate controls (compare Fig. 4A and D and pdx-cre-ERT; smad2^fx/fx sham in Fig. 5G). Interestingly, in the smad3^−/− null mutant mice, although not quantified, we noticed that there was near-absent proliferation of nonislet pancreatic tissue (Fig. 4F). For double smad2/3 mutant mice, when crossed with a PTF1a-cre strain (25), it showed an enhanced islet proliferation rate similar to the smad3 mutants, but with no additive effect, thus suggesting that smad2 may somehow work via smad3, explaining the greater effect of smad3 mutation (28) on islet cell replication (Figs. 4G and 5G).

While smad7 appeared to be a potential marker of proliferative, dedifferentiated, insulin-negative β-cells, we hypothesized that smad7 expression may also be necessary for proliferation of these β-cells. To test this possibility, we generated a smad7^fx/fx knockin mouse. These mice were created with loxP-sites flanking exon5 (the TGF-β receptor–interacting domain) of the smad7 locus (Supplementary Fig. 4). We then conditionally deleted smad7 in the pancreas by crossing the smad7^fx/fx mice with three different cre strains: pdx1-cre-ERT, PTF1a-cre, and ngn3-cre (23–25). Again, PPx was performed in these mice, and in the case of the pdx1-cre-ERT mice, tamoxifen was given daily for 5 days, starting 1 week prior to pancreatectomy. In all three crosses, there was essentially complete blockage of pancreatectomy-induced islet proliferation, with no difference seen in the number of BrdU⁺ islet cells among any of these three strains post-PPx compared with sham-operated cre-negative littermates (Fig. 5A–C and G), thus confirming that smad7 expression in islet cells is necessary for proliferation in response to a β-cell loss. The ngn3-cre cross showed the lowest overall proliferative rate. This result may reflect that ngn3-cre leads to a more effective deletion of both smad7 alleles in β-cell progenitors. Interestingly, in the PTF1a-cre;smad7^fx/fx animals, there was very little proliferation seen throughout the entire pancreas, including the exocrine tissue (see Fig. 5A, with a heavily BrdU⁺ lymph node as an internal positive control). However, we noticed that some islets in the PTF1a-cre;smad7^fx/fx mice did have several proliferative cells. We suspected that there may be some mosaicism of cre recombination for PTF1a-cre mice, especially in the islets, so we crossed these mice with the R26R-lsl-lacZ conditional reporter mice (20) (Fig. 5D–F). We chose this reporter in this study rather than the tomato red reporter used earlier because the floxed sequence of the R26R-lsl-LacZ mouse is closer in size to the floxed smad7 segment. One week post-PPx in these PTF1a-cre;smad7^fx/fx;R26R-lsl-lacZ mice, we again saw a few islets that had several BrdU⁺ cells, but essentially all of the BrdU⁺ cells in the islet were β-gal–negative and therefore were either not of the PTF1a lineage (e.g., endothelial cells) or were not cre lineage-tagged due to mosaicism (Fig. 5D–F, quantified in Fig. 5H). These data suggest that the effect of smad7 deletion in inhibiting islet cell proliferation is cell autonomous (i.e., a given islet cell with smad7 deletion is less likely to divide than an adjacent islet cell with smad7 intact).

Although smad2/3 were found to be important potential regulators of islet proliferation post-PPx (Fig. 4), multiple TGF-β superfamily ligands and receptors can work through smad2/3. As described above, we noted that the TGF-β1 ligand and activated TGFβrII staining pattern was rapidly altered in wild-type islets post-PPx, suggesting that downregulation of signaling specifically through TGFβrII may permit islet proliferation (Fig. 1Eand F, M and O). In order to test this possibility, we performed PPx in mice with conditional inactivation of the TGFβrI (TGFβrI^fx/fx, also called Alk-5, the canonical binding partner with TGFβrII) and/or (TGFβrII^fx/fx) (39,40). These receptor-floxed mice were crossed with either PTF1a-cre mice or pdx1-cre-ERT mice. One week after PPx, we found that mice with TGFβrII ablation (PTF1a-cre;TGFβrII^fx/fx or pdx1-cre-ERT;TGFβrII^fx/fx) had elevated numbers of BrdU⁺ islet cells, in a range similar to islets after PPx in smad2 and/or smad3 mutant mice (Fig. 4Iand J, quantification in Fig. 5G). Mice with ablation of only TGFβrI (PTF1a-cre;TGFβrI^fx/fx or pdx1-cre-ERT2;TGFβrI^fx/fx) had no difference in proliferation compared with controls (not shown), and there was no additive increased proliferation in the TGFβrII-ablated mice when additionally crossed to make a double TGFβrI/TGFβrII mutant mouse (PTF1a-cre;TGFβrI^fx/fx;TGFβrII^fx/fx or pdx1-cre-ERT2;TGFβrI^fx/fx;TGFβrII^fx/fx) (Fig. 4H). These results suggest that TGFβrII is a key suppressor of islet cell proliferation after PPx, likely through smad2 and smad3 phosphorylation. Interestingly, in this study, we found that increased proliferation in the pdx-cre-ERT2;TGFβrII^fx/fx mice required that the tamoxifen be given at least 1 month prior to PPx. If the tamoxifen was given just 1 week ahead, as in the floxed smad mice, we saw no increase in proliferation. This requirement may reflect a relatively long half-life of the TGFβrII in the islet cells.

The evidence that new adult β-cells only originate from pre-existing β-cells is fairly strong (1,2), but a general tenet of cell biology is that mature, highly differentiated cells do not readily divide. Controversy exists as to whether β-cells are able to proliferate while in a fully differentiated, insulin-expressing state (39,41–43). While mature β-cells have been shown to directly undergo cell division in certain systems (44), it remains unclear whether such an event is part of a normal physiologic program. In addition, a dedifferentiation program in β-cells, resulting in an empty β-cell, has been recently shown to be associated with β-cell failure (19). In this study, we found that a rapid and transient dedifferentiation and proliferation of β-cells occurs after a loss of β-cell mass. These proliferative cells are marked by and dependent upon expression of smad7. These smad7⁺ proliferative islet cells may mirror the insulin-negative empty β-cells described by Talchai et al. (19).

Our findings implicate smads2/3/7 in regulating the proliferative state of β-cells after a loss of β-cells. TGF-β superfamily signaling has been shown to play an important role in regulating many developmental and physiologic processes in the pancreas and in the β-cell (8). Others have shown a specific role for TGF-β superfamily signaling in the developing pancreas (7,8,11,12). We previously found that TGFβrII-mediated signals to the embryonic ductal structures suppress the recruitment of endocrine progenitors and suppress their proliferation (12). The canonical downstream effectors of TGFβrII are smads 2 and 3 (30). Others have shown that smads 2 and 3 are potential regulators of transdifferentiation of a duct cell line (AR42J cells) into β-cells in vitro (26,45), though these smad-mediated effects were thought to be representative of activin-ligand activity rather than TGF-βs. In keeping with our results, a previous report showed that heterozygous smad2 global-null mutant embryos (smad2^+/−) have increased numbers of ngn3⁺ progenitor cells (10) and increased numbers of nkx2.2⁺ and nkx6.1⁺ progenitor cells (10,46). Bonner-Weir et al. (47) showed that TGF-β ligand localizes to the periductal region, and they proposed that TGF-β may act as a break on ductal proliferation post-PPx. In contrast with our findings of a fourfold increase in baseline islet cell proliferation in pdx1-cre-ERT;smad2^fx/fx islets 1 week post–tamoxifen treatment, an adult mouse with pancreas-specific (pdx1-cre, not tamoxifen-inducible) ablation of smad2 was found to have poor insulin secretion and poor β-cell proliferation (48). These differences may well reflect the ablation of smad2 beginning during the embryonic period versus acutely before the PPx. The 1.5- to threefold increase in islet cell proliferation that we saw post-PPx in the smad2 and/or smad3 mutant mice supports a role for either TGF-β or activin signaling in suppressing β-cell proliferation, since both TGF-β and activin work through smads 2 and 3. Collectively, our data showing the rapid suppression of phosphorylated TGFβrII (Fig. 1Eand F), the rapid accumulation of TGF-β ligand in the vicinity of smad7 expression (Fig. 1M–O), and enhanced islet proliferation in the TGFβrII mutant mice (Figs. 4H–J and 5G) all suggest that it is specifically TGF-β signaling, rather than activin signaling, that mitigates islet cell proliferation post-PPx.

Our results also show an important regulatory role for smad7, and not smad6, in endocrine differentiation and proliferation. Importantly, when smad6 was expressed under the pdx1 promoter, no pancreatic phenotype was seen (14). Smad7, however, when ectopically expressed under the pdx1 promoter, led to an increase in α-cells, but a decrease in β-cells at the time of birth by 80–90% (15). The authors ascribed this phenotype to an inability of endocrine progenitor cells to mature.

We found that normal endogenous smad7 expression was associated with dedifferentiation of β-cells in the islet within 24 h of PPx and that this smad7 expression was necessary for β-cell proliferation. We feel that this phenomenon is not simply degranulation of β-cells, since the cells also begin to express PP-fold peptides. Along similar lines, β-cell dedifferentiation accompanied by the onset of expression of other peptide hormones, including PP, has recently been described in a model of β-cell failure (19). The authors attributed the dedifferentiation to β-cell exhaustion due to loss of FoxO1 in β-cells. There, some β-cells dedifferentiated into α-cells. We recently published that different injury models (pancreatic duct ligation vs. PPx) will result in different types of regenerative responses (49), which could explain why in our PPx model we did not observe β-cell dedifferentiation to α-cells. In our PPx model, however, the dedifferentiation appears to be part of a β-cell self-duplication process, with a dedifferentiation and then redifferentiation. The expression of PP-fold peptides in these dedifferentiated β-cells may reflect the established developmental lineage relationship between β-cells and PP cells (31–33).

Although we have not shown directly that these dedifferentiated smad7⁺ cells then redifferentiate after proliferation, logic would suggest that such a redifferentiation must occur. Specifically, since we saw multiple BrdU⁺/insulin⁺ cells 1 week post-PPx, but little or no BrdU⁺ cells in the smad7-ablated islets 1 week post-PPx (Fig. 5A–C), it would strongly suggest that the source of these insulin⁺/BrdU⁺ cells is β-cells that became insulin-negative/smad7⁺, then divided (became BrdU⁺), and then turned off smad7 expression and re-expressed insulin. An alternative explanation could be that the smad7⁺ cells instead influence other cells in the islet to divide through a paracrine, non–cell-autonomous mechanism. However, we feel that such a non–cell-autonomous mechanism is less likely since in the PTF1a-cre;smad7^fx/fx;R26R-lsl-lacZ islets that showed a mosaic penetration of cre recombination in islet cells, non–cre-recombined cells were found to be the predominant cells that were BrdU⁺ (Fig. 5D–Fand H). Thus, the absence of smad7 in the majority of the islet cells did not prevent proliferation of those few cells that did not cre-recombine. Although we found smad7⁺ cells to be present in ∼95% of islets post-PPx, there was a large interislet variability within the same pancreas in terms of their smad7 expression pattern (Fig. 1H–Kand N). A similar variability was observed by Talchai et al. (19) in the different patterns of insulin immunoreactivity or insulin-negative cells in different islets.

We found that smad7⁺ dedifferentiated islet cells were derived from insulin⁺ cells. In addition, these dedifferentiated β-cells had become positive for the PP-fold family of molecules (PP, NPY, and PYY). The expression of these PP-fold molecules in the smad7⁺ cells suggests that the β-cells have dedifferentiated along a developmental cell lineage pathway in which PP-fold–expressing cells represent a more immature cell prior to redifferentiation to form β-cells. PP-fold peptides have been shown to be expressed in early embryonic endocrine cells and have been found to be frequently coexpressed with other islet hormones in the early embryo, but not late in gestation (31,32,34). Using cell ablation strategies, PP-fold–expressing cells were found to be uniquely necessary for the formation of β-cells and δ-cells, again consistent with our findings that β-cells, and possibly δ-cells, but not α-cells, appeared to have dedifferentiated to form smad7⁺, PP-fold–expressing cells (33,50). Furthermore, using conditional cell lineage-tagging strategies, β-cells, but not α-cells, were shown to derive from a PP-expressing lineage (22), again consistent with our findings.

Thus, we have identified an important and necessary regulatory point for β-cell replication after β-cell loss, in the form of a smad2/3/7 network. In addition, our results begin to explain a concept of how a highly differentiated cell such as a β-cell could undergo cell division. The dedifferentiation that we see, apparently guided in part by smad7 expression, seems to represent a recapitulation of an embryonic phenotype with PP-fold expression. TGF-β signaling may represent a target pathway for regulating β-cell proliferation in humans.

Acknowledgments. The authors thank Lauren Brink, Jessica Thomas, and Sean-Paul Williams for technical assistance and mouse breeding. The authors also thank Farzad Esni for helpful discussions, and Christine Kalinyak, Anne Meinert, and Tamara Daviston for administrative assistance.

Funding. This study was supported by the National Institutes of Health (grants R01-DK-064952 and R01-DK-083541-01 to G.G.), and the Children's Hospital of Pittsburgh. M.D. was supported by Project Z01-ES-101603 from the Division of Intramural Research of the National Institutes of Health (National Institute of Environmental Health Sciences).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.E. and S.T. performed the experiments and wrote the manuscript. J.W., P.G., and C.W. performed the experiments. K.P., J.P., C.S., and X.X. shared in discussion of the manuscript. Y.W. provided reagents and helped with the discussion of the manuscript. M.D. generated the Smad7^fx/fx mouse. G.G. designed the overall experiment and wrote the manuscript. G.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.