β-Cell Proliferation, but Not Neogenesis, Following 60% Partial Pancreatectomy Is Impaired in the Absence of FoxM1

Foxm1^flox/flox mice on a mixed background (129SvJ, C57BL/6) were described previously (12). Briefly, loxP sequences flank exons 4–7, yielding a null allele after Cre-mediated recombination. Pdx1^5.5kb-Cre mice (on a mixed ICR, CBA, and C57BL/6 background) were generously provided by Guoqiang Gu (Vanderbilt University) (20) and express Cre recombinase throughout the Pdx1 domain, including the entire pancreatic epithelium, by embryonic day (e) 10.5 (21,22). Foxm1 and Cre were genotyped as described (9,20). Rosa26-FoxM1 transgenic (Tg) mice (on a mixed FVB/N, C57BL/6 background) were previously described (23) and genotyped (as in 24). For embryonic analyses, the morning of vaginal plug was considered e0.5. All mice received food and drink ad libitum and were on a 12-h light-dark cycle. All mouse studies were performed in accordance with the Vanderbilt Institutional Animal Care and Use Committee guidelines under the supervision of the Division of Animal Care.

Eight-week-old female mice were anesthetized using isoflurane. A midline abdominal incision allowed exteriorization of the splenic lobe of the pancreas (between the gastroduodenal junction and the spleen). Sixty percent PPx was performed by gently denuding the pancreatic tissue from the splenic lobe using cotton-tipped swabs soaked in 0.9% saline, leaving the main pancreatic duct and splenic artery intact. For sham operations, the splenic lobe was isolated using cotton-tipped swabs and gently rubbed with a moistened gloved finger for 1 min. Abdominal muscles were sutured using 5-0 vicryl (Ethicon), and the skin was stapled using 9.0-mm staples (Reflex Skin Closure Systems). A total of 0.1 mg/kg buprenorphine-HCl in normal saline was injected subcutaneously for analgesia, and 0.8 mg/ml bromo-deoxyuridine (BrdU) (Sigma-Aldrich) was administered in drinking water in a light-safe bottle and was replaced on the 4th day. Mice were killed 1 week postoperation by cervical dislocation, and the splenic and duodenal lobes of the pancreas were harvested and analyzed separately. One day before and 7 days after the operation, intraperitoneal glucose tolerance tests (IPGTTs) were performed as described previously (9) (see Fig. 1 for experimental timeline).

RNA from whole pancreas or freshly isolated islets was extracted and DNase treated using the RNAqueous and Turbo DNA-free kits (Ambion). RNA was assessed using the ND-1000 Spectrophotometer (NanoDrop) and the 2100 Electrophoresis Bioanalyzer (Agilent). cDNA was generated from 20 ng RNA using the Superscript III First-Strand Synthesis System for RT-PCR (Invitrogen). Real-time PCR was performed using the iQ5 Real-Time PCR Detection System (Bio-Rad) with iQ Real-Time SYBR Green PCR Supermix (Bio-Rad) and the primer sets listed in online appendix Table S1 (available at http://dx.doi.org/10.2337/db08-0878). Results were quantitated using the ΔΔCt method.

Pancreatic tissue was fixed in 4% paraformaldehyde for 2–4 h at 4°C, processed for paraffin embedding, and sectioned at 5 μm. Primary antibodies (guinea pig anti-insulin or anti-glucagon [both 1:1,000; Linco], rabbit anti-cytokeratin [1:1,000; DakoCytomation], rat anti-BrdU [1:400; Accurate Chemical and Scientific], rabbit anti–phospho-histone H3 [1:250; Upstate Cell Signaling Solutions], and mouse anti–neurogenin-3 [Ngn3] [1:100; Developmental Studies Hybridoma Bank, University of Iowa]) were incubated overnight at 4°C. Secondary antibodies (peroxidase-conjugated donkey anti–guinea pig [1:250; Jackson Laboratories] and Cy2- or Cy3-conjugated donkey anti–guinea pig, anti-rabbit, or anti-rat [all 1:500; Jackson Laboratories]) were incubated for 1 h at room temperature. Detection of cytokeratin required pretreatment with 20 μg/ml proteinase-K for 5 min at room temperature, and detection of Ngn3 required microwave antigen retrieval with 10 mmol/l Tris-EGTA, pH 9.0, and tyramide signal amplification with TSA Kit no. 2 (Invitrogen).

Sections 250 μm apart (5–15 sections per sample) were incubated with anti-insulin primary antibody followed by peroxidase-conjugated secondary antibody, visualized using a DAB Peroxidase Substrate Kit (Vector Laboratories), and counterstained with eosin. Digital images were created using a Nikon Coolscan 9000 and NikonScan version 4.0.2. Total pancreatic and insulin⁺ areas of each section were measured using MetaMorph version 6.1 (Molecular Devices). β-Cell mass equaled the ratio of total insulin⁺ area to total pancreatic area of all sections, multiplied by tissue wet weight.

Sections 250–750 μm apart (three to eight sections per sample) were labeled with anti-insulin and anti-BrdU primary antibodies. Nuclei were labeled with 1.5 μg/ml 4′,6′-diamidino-2-phenylindole (DAPI) (Molecular Probes) in mounting media. All insulin⁺ cells on each section were imaged at ×400 magnification on an Olympus BX41 microscope with a digital camera using Magnafire (Optronics). Percent proliferating β-cells equaled the number of insulin/BrdU double-positive cells divided by the total number of insulin⁺ cells. α-Cell proliferation was measured similarly, using anti-glucagon labeling. Ductal cells were identified by cytokeratin labeling on sections 250–750 μm apart, and images were captured at ×200 magnification. Acinar cells were identified by morphological characteristics on three separate ×200 images from sections 750 μm apart. BrdU labeling was previously described (9). Phospho-histone H3 labeling of sections 125 μm apart (seven to nine sections per sample) required microwave antigen retrieval in 10 mmol/l sodium citrate, pH 6.5, followed by 0.2% Triton-X-100 in PBS.

Using images collected for β-cell proliferation, the cross-sectional area of each insulin⁺ cell grouping was circumscribed and measured using MetaMorph. Average cross-sectional β-cell size per islet equaled the total area of the islet divided by the number of β-cell nuclei within it. β-Cell nuclear size was measured by circumscribing the DAPI⁺ cross-sectional area within insulin⁺ cells.

Data were analyzed by unpaired t test or two-way ANOVA with Bonferroni's posttests, as appropriate, using GraphPad Prism version 5.01. P < 0.05 was considered significant. Specific analyses performed and sample sizes are included in the figure legends.

FoxM1 mRNA and protein are relatively low in adult islets (data not shown and 9), but it was unknown whether upregulation occurs upon stimulation of endocrine cell proliferation. Indeed, Foxm1 expression was upregulated twofold in control islets 6 days following PPx versus sham (Fig. 2), suggesting that FoxM1 plays a role in the regenerative response of endocrine cells. Importantly, Foxm1 upregulation corresponded with the peak of β-cell proliferation following PPx (25).

Additionally, we observed significant upregulation of known FoxM1 targets, including Plk1, Cenp-a, Birc5/survivin, and Ccnb1 (Fig. 2). Other targets, including Skp2, Cks1, and Cdc25b, displayed a trend of upregulation following PPx versus sham but did not reach significance. A similar trend was observed for Ccna2 but not for Ccnd2, both of which have been linked to β-cell proliferation (17,18,26), but neither of which is a known FoxM1 target.

To determine whether FoxM1 is necessary for endocrine cell regeneration following pancreatic injury, FoxM1^Δpanc and control female littermates underwent PPx or sham operation at 8 weeks of age. Previous reports showed that 60% PPx does not induce hyperglycemia (25), unlike 90% PPx (27). IPGTTs performed 7 days following 60% PPx or sham confirmed that control and FoxM1^Δpanc female mice remained glucose tolerant (Fig. 3).

Similar to FoxM1^Δpanc male mice (9), FoxM1^Δpanc female mice also exhibited ∼50% reduced β-cell mass at 9 weeks of age compared with littermate controls (data not shown), although FoxM1^Δpanc female mice remained normoglycemic. This ∼50% reduction in β-cell mass was evident in the splenic, but not duodenal, lobe and was unchanged after a sham operation (Fig. 4B and C). Total pancreatic weight and weight of the individual lobes were not significantly different between control and FoxM1^Δpanc mice (Fig. 4A) (data not shown), indicating a specific defect in the endocrine compartment of the pancreas. Consistent with previous findings (28), a significant majority of β-cell mass was located within the splenic versus duodenal lobe of control mice after a sham operation (P < 0.01).

PPx eliminated the majority of splenic lobe tissue, and within 7 days a significant and equivalent amount of tissue regenerated in both FoxM1^Δpanc and control mice (Fig. 4A) (online appendix Fig. S1A–D), suggesting that overall pancreas regeneration during this time period was not impaired by Foxm1 deletion. β-Cell mass regeneration within the splenic lobe, however, was impaired in FoxM1^Δpanc mice, as β-cell mass in these mice did not significantly increase within 7 days following PPx, in contrast to control littermates (Fig. 4B).

Because β-cells are replenished primarily by self-replication (4), and FoxM1 is important for postnatal β-cell proliferation (9), it seemed likely that reduced β-cell proliferation contributed to impaired β-cell mass regeneration in FoxM1^Δpanc mice. Indeed, PPx dramatically stimulated β-cell proliferation in both pancreatic lobes of control mice, and these effects were significantly blunted in FoxM1^Δpanc mice (Fig. 5A–F). Moreover, β-cell proliferation within the duodenal lobe of FoxM1^Δpanc mice did not significantly increase after PPx.

β-Cell proliferation in the splenic lobe of FoxM1^Δpanc mice did increase significantly following PPx versus sham, despite Foxm1 deletion (Fig. 5E), due to increased proliferation within small endocrine cell clusters (eight or fewer insulin⁺ cells) and definitive islets (nine or more insulin⁺ cells) (Fig. 5G and I). However, β-cell proliferation within definitive islets was significantly reduced in FoxM1^Δpanc versus control mice (Fig. 5I). Because there was no significant impairment in β-cell proliferation in small endocrine cell clusters in FoxM1^Δpanc mice, we hypothesized that these clusters represented newly formed β-cells that did not require FoxM1 for proliferation. This hypothesis was supported by the finding that embryonic β-cell proliferation did not require FoxM1 (online appendix Fig. S2A and B). Additionally, proliferation of β-cells within small endocrine cell clusters after PPx was specifically stimulated in the regenerating splenic (Fig. 5G), but not duodenal (Fig. 5H), lobe. We hypothesize that just as perinatal β-cells can proliferate in the absence of FoxM1, newly differentiated β-cells in the adult can undergo several rounds of FoxM1-independent replication before switching to FoxM1-dependent proliferation.

PPx stimulated β-cell proliferation within definitive islets of control mice to the same extent (sevenfold) in both the splenic and duodenal lobes versus sham (Fig. 5I and J). Proliferation in the splenic lobe of FoxM1^Δpanc mice increased only fivefold, while the increase within the duodenal lobe did not reach significance, suggesting that proliferation of preexisting β-cells was impaired in the absence of FoxM1. The increased BrdU incorporation observed within definitive islets in the splenic lobe of FoxM1^Δpanc mice 7 days after PPx versus sham (Fig. 5I) may represent proliferation that occurred at an earlier stage of new islet development. These results are summarized in Table 1. α-Cells similarly required FoxM1 for PPx-stimulated proliferation (Fig. 5K). In contrast, acinar and ductal cell proliferation was not impaired in FoxM1^Δpanc mice (online appendix Fig. S3A and B), revealing that FoxM1 is specifically required for endocrine cell proliferation.

Sham-operated FoxM1^Δpanc mice exhibited larger average β-cell size in both pancreatic lobes (Fig. 6A and B) versus control littermates, due in part to increased β-cell nucleus size (Fig. 6C and D). However, hypertrophy also contributed, likely as a compensatory measure due to reduced β-cell number. PPx did not alter β-cell size in control mice. However, average β-cell size was significantly reduced specifically within the splenic lobe of FoxM1^Δpanc mice after PPx, likely due to new β-cell formation.

Although PPx reduced total β-cell number within the splenic lobe, islets present within this lobe 7 days after PPx were of the same average size as after the sham operation in control mice (Fig. 7A). However, FoxM1^Δpanc islets were significantly smaller 7 days after PPx compared with sham. Similarly, the average islet size within the duodenal lobe of control mice trended toward an increase following PPx versus sham, but no expansion was observed in FoxM1^Δpanc mice (Fig. 7B). These data suggest that Foxm1 deletion impairs islet growth after PPx. Of note, islets within the splenic lobe of control mice after sham operation averaged two times larger than islets within the duodenal lobe (P < 0.05), which may be due to a greater proportion of small endocrine cell clusters versus definitive islets normally found in the duodenal versus splenic lobe (P < 0.05) (Fig. 7C and D). Interestingly, after PPx FoxM1^Δpanc mice exhibited an increased proportion of small endocrine cell clusters versus definitive islets within their splenic lobe in comparison with control mice (Fig. 7C), which likely contributed to their reduced average islet size. No differences were observed in the proportion of small endocrine cell clusters between control and FoxM1^Δpanc mice after sham operation or within the duodenal lobe under any condition (Fig. 7D).

The average number of insulin⁺ cell groupings in the splenic lobe of both FoxM1^Δpanc and control mice trended toward an increase 7 days after PPx versus day 0 (Fig. 7E), while no differences were observed in the duodenal lobes of either genotype (Fig. 7F). Insulin⁺ cells were observed within foci of regeneration in the splenic lobes of both groups of mice (online appendix Fig. 1B and D), suggesting that generation of new islets, or neogenesis, occured specifically in the splenic lobe after PPx and was not impaired in FoxM1^Δpanc mice. These findings were supported by observation of Ngn3, a marker of endocrine progenitor cells (29), in ductal epithelial cells in both control (Fig. 7G) and FoxM1^Δpanc (Fig. 7H) mice following PPx. Importantly, Ngn3 expression was only observed in ductules within the regenerating portion of the splenic lobe but not within differentiated pancreas tissue. No difference in the number of Ngn3⁺ ductal epithelial cells was detected between control and FoxM1^Δpanc mice following PPx.

To determine whether Foxm1 overexpression affects pancreas and β-cell regeneration following PPx, we utilized Rosa26-FoxM1 Tg mice that constitutively and ubiquitously overexpress human FoxM1 (23). Despite FoxM1 overexpression within islets (online appendix Fig. S4A), Tg mice exhibited no significant differences in glucose tolerance (online appendix Fig. S4B) or β-cell mass following PPx (online appendix Fig. S4C) compared with wild-type littermates. Furthermore, we observed no differences in pancreatic tissue weight (online appendix Fig. S4D), average islet size (online appendix Fig. S4E), or islet density (online appendix Fig. S4F) between Tg and wild-type mice in either lobe.

FoxM1 is highly expressed within pancreatic endocrine cells during embryogenesis and is downregulated with age (9), corresponding with reduced β-cell proliferation. We previously showed that mice with Foxm1 deleted throughout the pancreatic epithelium (FoxM1^Δpanc) exhibit impaired growth and maintenance of postnatal β-cell mass secondary to reduced postnatal β-cell proliferation, associated with increased nuclear-localized p27^Kip1 and premature cellular senescence (9). Several of FoxM1's direct transcriptional targets inhibit p27^Kip1 by promoting its ubiquitination and subsequent degradation (Skp2, Cks1) (14) or its nuclear export (kinase-interacting stathmim) (30). Here, we show that FoxM1^Δpanc mice have impaired β-cell mass regeneration following PPx due to reduced β-cell proliferation but that neogenesis proceeds normally.

Sixty percent PPx is a normoglycemic pancreatic injury model that stimulates regeneration by differentiation and proliferation and hyperplasia of preexisting tissue. We found that Foxm1 and some of its targets were upregulated within islets following PPx during a period of enhanced α- and β-cell proliferation, and we hypothesized that FoxM1 was required for injury-stimulated endocrine cell proliferation. Indeed, we found that FoxM1^Δpanc mice exhibited impaired α- and β-cell proliferation and impaired β-cell mass regeneration following PPx. PPx stimulated β-cell proliferation in both the splenic (regenerating) and duodenal (expanding) lobes of control mice but not in the duodenal lobe of FoxM1^Δpanc mice. Thus, hyperplasia of preexisting β-cells after injury requires FoxM1.

Interestingly, PPx-stimulated β-cell proliferation was blunted, but not completely inhibited within the splenic lobe of FoxM1^Δpanc mice, due at least in part to impaired proliferation in definitive islets, but not small endocrine cell clusters, in the absence of FoxM1. These data suggest that another pathway(s), which may mimic that utilized during embryonic pancreas development, compensates for loss of FoxM1 during regeneration, as FoxM1 was also not required for embryonic β-cell proliferation. Thus, preexisting β-cells seemingly exhibit a stronger requirement for FoxM1 than new β-cells formed by neogenesis. Although it is currently unclear which factors are required for β-cell proliferation during embryogenesis and following islet neogenesis, our studies support the hypothesis that neogenesis in the adult is similar to embryonic β-cell development, both with regard to reexpression of Ngn3 and initial FoxM1-independent proliferation.

Because BrdU was administered over 7 days, we may have underestimated the effects of Foxm1 deletion on endocrine cell proliferation. Previous studies have shown that Foxm1^−/− cells can undergo multiple rounds of DNA synthesis without karyo- or cytokinesis (endoreduplication), resulting in polyploid nuclei (16). We found evidence of polyploid β-cells and exocrine cells in FoxM1^Δpanc mice. These cells would have been counted as proliferating cells in our analyses but would not have actually resulted in increased cell number. This phenomenon could partially explain the increased BrdU incorporation in β-cells without significant β-cell mass regeneration within the splenic lobe of FoxM1^Δpanc mice after PPx compared with sham, although it most likely would not be restricted to one lobe.

Additionally, incomplete Foxm1 recombination could have resulted in mosaic expression, allowing some β-cells to continue proliferating. However, we think that this is unlikely because we previously showed that Pdx1^5.5kb-Cre recombined the Rosa26-LacZ reporter allele within all endodermally derived pancreatic cells and yielded a significant reduction in FoxM1 protein by e15.5 (9). We also found that Foxm1 transcripts were significantly reduced in FoxM1^Δpanc neonatal pancreas (online appendix Fig. S5A) and even more dramatically in adult islets (online appendix Fig. S5B). These tissues contain nonendodermally derived mesenchymal, endothelial, and neuronal cells, which would not undergo Foxm1 recombination and thus likely account for remaining transcript expression.

Proliferation of α- and β-cells, but not acinar or ductal cells, was reduced in FoxM1^Δpanc mice versus control littermates following PPx, suggesting that FoxM1 is required specifically within endocrine cells. These data concur with other studies (17–19,31) showing that endocrine cells are particularly susceptible to perturbations in cell cycle regulation. The current study utilized quantitative RT-PCR on islet RNA to assess changes in Foxm1 expression following PPx. Unfortunately, antibodies to detect FoxM1 protein in mouse tissue do not currently exist, precluding precise identification of the cell population(s) in which FoxM1 expression or localization changes with time and after PPx. Such studies will allow for better understanding of cell type–specific requirements for FoxM1 and may help to explain the differing effects of FoxM1 absence that we observe between endocrine and exocrine cells following PPx. However, our expression analyses identified several potentially important regulators of pancreatic endocrine cell proliferation and regeneration that are themselves regulated by FoxM1 (Plk1, Cenp-a, Birc5/survivin, and Ccnb1).

We also presented evidence of β-cell neogenesis following 60% PPx, including Ngn3⁺ duct cells, insulin⁺ cells within regenerating tissue, and increased islet density. These findings were equivalent between control and FoxM1^Δpanc mice, revealing that β-cell neogenesis does not require FoxM1. Peshavaria et al. (25) showed increased small endocrine cell clusters by 2 days post-PPx, which normalized by 7 days. Our data are consistent with this finding, as control mice at 7 days post-PPx exhibited a normal proportion of small endocrine cell clusters. However, FoxM1^Δpanc mice had an increased proportion of small endocrine cell clusters at 7 days post-PPx, indicating that formation of these clusters is not impaired in the absence of FoxM1, but significant growth is. These data support our hypothesis that initial proliferation of newly differentiated β-cells does not require FoxM1, but these cells later switch to a FoxM1-dependent state. Unfortunately, current technology precludes real-time monitoring of β-cell proliferation and islet growth. However, this study highlights the importance of β-cell proliferation as a mechanism of β-cell mass regeneration, as our data show that neogenesis, without adequate proliferation, is not sufficient to regenerate a significant amount of β-cell mass after 60% PPx.

As β-cell regeneration after partial duct ligation mimics embryonic β-cell differentiation with Ngn3 reexpression (6), it is not surprising that similar processes occur following 60% PPx, although perhaps to a lesser extent. Previous studies (5,32) asserting the lack of reactivation of the embryonic program, and thus neogenesis, following PPx were performed in a manner that likely precluded true regeneration, with surgical excision of the splenic lobe and pancreatic duct, while other studies (4) did not analyze the splenic lobe specifically, possibly missing the contribution of neogenesis and regeneration to new β-cell formation. The current study removed tissue from the splenic lobe, leaving the main pancreatic duct intact. This lobe was analyzed separately, and only specific foci of regeneration exhibited evidence of neogenesis. Thus, this study does not contradict but adds to the current knowledge regarding the potential for pancreas regeneration and β-cell neogenesis.

FOXM1 overexpression did not affect β-cell regeneration following PPx, suggesting that FoxM1 is not sufficient to induce β-cell proliferation, emphasizing that the β-cell cell cycle is under tight negative regulatory control, even after injury. As reviewed (33), β-cells express all of the cell cycle “brakes” studied thus far but only some “accelerators.” Thus, it may be more difficult than initially anticipated to stimulate β-cell proliferation in vivo or in vitro as a therapeutic option for diabetic patients. Because FoxM1 positively regulates cell cycle accelerators and negatively regulates cell cycle brakes, it is of particular interest. However, better understanding of posttranslational and inhibitory regulators is necessary to successfully manipulate FoxM1 activity in β-cells.

Islets were isolated by Zhongyi Chen and Anastasia Golovin of the Vanderbilt Islet Procurement and Analysis Core, supported by the Diabetes Research and Training Center (P60 DK20593). RNA quality was assessed by the Vanderbilt Microarray Shared Resource, supported by the Vanderbilt Ingram Cancer Center (P30 CA68485), the Vanderbilt Digestive Disease Center (P30 DK58404), and the Vanderbilt Vision Center (P30 EY08126). A.A.M. is supported by an American Diabetes Association Medical Scholars Training Award (7-07 MS-01) and by the Vanderbilt Medical Scientist Training Program (T32 GM007347). M.G. is supported by the National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK071052).

We thank Elizabeth Tweedie Ables, Michelle Guney, and Renuka Menon for critical reading of the manuscript and helpful suggestions. We also thank Young Ah Oh, Christine Pope, Xue Feng, Usa G. Kopsombut, Andre Boustani, and Jacob Kaufman for their technical assistance.

This article is dedicated to R.H.C., who passed away during the course of this study.