Gestational Diabetes Mellitus Resulting From Impaired β-Cell Compensation in the Absence of FoxM1, a Novel Downstream Effector of Placental Lactogen

Foxm1^fl/fl and FoxM1^Δpanc mice and genotyping are described elsewhere (23,25,28). RIP-mPL transgenic mice were as previously described (17). All studies were performed with the approval of and in compliance with the Vanderbilt Institutional Animal Care and Use Committee.

Pancreata were fixed in 4% paraformaldehyde at 4°C for 4 h. Tissues were dehydrated, embedded in paraffin, and sectioned at 5 μm. Primary antibodies used are as follows: guinea pig anti-bovine insulin (Linco), 1:1,000; rat anti-bromodeoxyuridine (BrdU; Accurate Chemical & Scientific), 1:400; rabbit anti-Glut2 (Alpha Diagnostic), 1:500; mouse anti-Kip1/p27 (BD Biosciences), 1:100; mouse anti-neurogenin 3 (NGN3; Developmental Studies Hybridoma Bank), 1:100; and rabbit anti-cytokeratin (Dako), 1:1,000. Incubations were overnight in a humid chamber at 4°C. For BrdU detection, slides were treated with 1.5 N HCl for 20 min at 37°C, neutralized in sodium borate buffer for 1 min at room temperature, and treated with 0.005 mg/ml trypsin (Sigma-Aldrich) and 0.005 mg/ml CaCl₂ (in Tris buffer; pH 7.5) for 3 min at 37°C. Vectastain ABC kit (Vector Labs) was used for immunohistochemical labeling of insulin for β-cell mass measurement. The Histomouse-SP kit (Invitrogen) was used for detection of p27. NGN3 and cytokeratin immunolabeling was performed as previously described (12). Donkey anti-guinea pig cyanin 2 (Cy2) and donkey anti-rat Cy3 were used as secondary antibodies at a 1:500 dilution. Mounting medium contained 1.5 μg/ml nuclear fluorogen DAPI (Molecular Probes).

Samples were viewed under bright-field illumination or appropriate optical filters (immunofluorescence) using an Olympus BX41 microscope (Tokyo, Japan) and digital camera with the Magnafire program (Optronics). Tagged image file format (TIFF) images from each experiment were processed equivalently in Adobe Photoshop.

Entire pancreata were removed, weighed, and fixed as above. Five-micron longitudinal sections were prepared for insulin immunoperoxidase labeling and eosin counterstaining. Every 30th section (an average of 8–13 sections per pancreas) was used. Images of anti-insulin–labeled sections were scanned using a Nikon Super CoolScan 9000. Using Metamorph 6.1 software (Molecular Devices), β-cell mass was measured by obtaining the fraction of cross-sectional area of pancreatic tissue positive for insulin and multiplying this by the pancreatic weight (n = 3 animals of each genotype at each stage).

BrdU (100 mg/kg; Sigma-Aldrich) was injected intraperitoneally every 2 h three times prior to pancreas tissue processing. BrdU-labeled β-cells were detected by double immunolabeling with insulin antibodies and DAPI to visualize nuclei. Using Metamorph, BrdU-positive and -negative nuclei in insulin-positive cells were counted at ×400 magnification. At least 3,000 β-cells were counted for each of three animals of each genotype at each stage. The proportion of BrdU-positive β-cell nuclei to total β-cell nuclei was calculated and represents the percentage of β-cells that have recently gone through S-phase.

Pancreas sections were colabeled for Glut2 and insulin as described above. Every islet from one section on each slide was photographed. Using Metamorph, the area of more than 1,000 individual β-cells was determined per mouse.

Dissected pancreata were rinsed in PBS, blotted with filter paper, weighed, and homogenized (Polytron PT 10/35; Brinkmann Instruments) in 1 ml of acid alcohol. The homogenate was extracted with an additional 5 ml of acid alcohol for 48 h at 4°C and centrifuged at 2,500 rpm for 30 min. Supernatants were assayed for insulin by radioimmunoassay (30). Insulin concentrations were normalized to pancreas wet weight.

For plasma insulin content, blood samples were collected at 0 and 30 min after glucose injection during an intraperitoneal glucose tolerance test (IPGTT) from the saphenous vein. The plasma supernatant was analyzed using the Ultrasensitive ELISA kit (Alpco).

IPGTTs were performed on Foxm1^fl/fl and FoxM1^Δpanc virgin females and at gestational day (GD) 12.5 and GD15.5 as described (28). Controls for these studies were Foxm1^fl/fl mice.

Islets were isolated from virgin, GD14.5, and postpartum day 8 (P8) females by intraductal collagenase perfusion. At least three mice were used per group. Total islet RNA (75–125 islets/mouse) was extracted using the RNAqueous kit (Ambion). RNA concentration and integrity were assessed using the ND-1000 Spectrophotometer (NanoDrop) and the 2100 Electrophoresis Bioanalyzer (Agilent).

cDNA was synthesized using the Superscript III First-Strand synthesis system (Invitrogen). Reactions were carried out in duplicate with iQ SYBR Green supermix (Bio-Rad) at an annealing temperature of 58°C. Data were collected using an iCycler iQ Real-Time PCR Detection System (Bio-Rad) and software (Bio-Rad). Primers optimized by melting curve analysis were generated as follows: Foxm1 (forward, cacttggattgagaccactt; reverse, gtcgtttctgctgtgattcc), Bcl-xl (forward, ccttggatccaggagaacg; reverse, caggaaccagcggttgaa), and hypoxanthine-guanine phosphoribosyltransferase (Hprt) (forward, agtcaacgggggacataaaa; reverse, tgcattgttttaccagtgtcaa). Expression levels were normalized against the levels of Hprt, a ubiquitously expressed gene that shows little change during cellular growth or differentiation. Results were analyzed using the 2^−ΔΔCt method (31).

Islets from 8- to 12-week-old virgin and GD15.5 Foxm1^fl/fl and FoxM1^Δpanc females were lysed and sonicated with a Virsonic 100 sonicator (Virtis Company). The supernatant was quantified using the Bio-Rad DC protein assay (Bio-Rad). Western blotting was done as previously described using 10 μg islet protein/lane from one animal (32). Membranes were probed with primary antibodies diluted in 3% nonfat milk in Tris-buffered saline (TBS) overnight at 4°C: goat anti-Menin (Bethyl Laboratories) 1:1,000; rabbit anti-Pdx1 (a gift from Dr. Chris Wright, Vanderbilt University), 1:10,000; mouse anti-p27 (BD Biosciences), 1:2000; and mouse anti–β-actin (Santa Cruz Biotechnology), 1:10,000. Peroxidase-conjugated species-specific secondary antibodies (Jackson ImmunoResearch Laboratories) were diluted to 1:5,000 in 3% nonfat milk in TBS and incubated for 1 h at room temperature. Protein levels in individual lanes were quantified using ImageJ 1.42 (National Institutes of Health [NIH]) and normalized to the level of β-actin signal in corresponding lanes. Three to five animals per group were analyzed.

Islets from 8- to 12-week-old Foxm1^fl/fl virgin females were cultured for 4 days in 1× RPMI 1640 supplemented with 10% horse serum, 11 mmol/l glucose, 100 units/ml penicillin, and 100 μg/ml streptomycin with or without 500 ng/ml human PL (National Hormone and Peptide Program). An average of 50–60 islets were plated in one 60 × 15-mm non–tissue culture–treated Petri dish at 37°C in 5% CO₂. Cultures were replenished with fresh medium with or without PL after 48-h incubation. After 4 days, RNA was extracted and subjected to quantitative (q)RT-PCR analysis.

Transferase-mediated dUTP nick-end labeling was performed on paraffin sections of adult pancreata using the In Situ Cell Death Detection kit (Roche) and followed by immunofluorescent staining of insulin. Cleaved caspase-3 was detected using an antibody raised in rabbit (Cell Signaling Technology) at a dilution of 1:1,500. Immunoreactivity was detected using the Vectastain ABC and DAB kits (Vector Labs).

Data were analyzed by Student t test or two-way ANOVA with Bonferroni post-tests, using GraphPad Prism V5.01. A value of P ≤ 0.05 was considered significant.

In pregnant mice, maternal β-cell replication peaks at GD14.5 (18) (supplementary Fig. 1, which is available in the online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0050/DC1). β-Cell mass expansion during pregnancy is due mainly to stimulation of proliferation by elevated circulating pregnancy hormones, including, GH, Prl, and PL (6,15,16), as well as via cell hypertrophy (8,33). We previously showed that FoxM1 is required for postnatal β-cell mass expansion and proliferation and is also necessary for β-cell mass regeneration after PPx (12,28). In the liver, FoxM1 activity is required for GH-mediated stimulation of liver regeneration (29). Thus, we hypothesized that FoxM1 would be required for the hormone-induced increases in β-cell proliferation and β-cell mass during pregnancy.

Antibodies that specifically recognize FoxM1 in mouse tissue are not available. Thus, we examined whether Foxm1 mRNA expression in maternal islets correlated with β-cell proliferation during pregnancy. Based on the maternal β-cell proliferation profile during pregnancy, qRT-PCR analysis was performed on isolated control islets at virgin, GD14.5, and P8 stages. Maternal Foxm1 expression in control islets was elevated more than 50% at GD14.5 and returned back to prepregnancy levels at P8 (Fig. 1). Thus, Foxm1 expression correlated with the dynamics of β-cell proliferation during pregnancy.

We predicted that lack of FoxM1 would lead to decreased β-cell proliferation and mass during pregnancy. Significantly fewer proliferating β-cells were observed in FoxM1^Δpanc females compared with controls (Fig. 2) at all stages. Controls displayed an approximately threefold increase in β-cell proliferation at GD15.5, with no corresponding increase in FoxM1^Δpanc mice, clearly indicating that FoxM1 is important for increased maternal β-cell proliferation during pregnancy.

Our previous studies revealed an increase in nuclear p27 in neonatal islets in the absence of FoxM1 (28). Examination of islets from virgin and pregnant females revealed a decrease in intensity of nuclear p27 expression in control islets at GD15.5 (Fig. 3,A and B). In contrast, islets from virgin and GD15.5 FoxM1^Δpanc females showed increased nuclear p27 compared with control females (Fig. 3,C and D). There was no change in total p27 protein as detected by Western blotting (data not shown). In addition, we examined expression of Menin in control and FoxM1^Δpanc islets at virgin and GD15.5 stages. Islets from pregnant control animals showed a decrease in Menin expression compared with virgin mice, consistent with published results (18). However, in FoxM1^Δpanc islets from pregnant females, Menin protein levels remained elevated (Fig. 3 E).

The decrease in β-cell proliferation observed in FoxM1^Δpanc females translated into decreased β-cell mass. Virgin FoxM1^Δpanc females had a significant decrease in β-cell mass compared with controls (Fig. 4,A), although we did not detect a significant difference in total pancreatic insulin content at baseline (Fig. 4,B). A dramatic decrease in β-cell mass in FoxM1^Δpanc females was also observed at GD15.5 (Fig. 4,A). Pregnancy neither stimulated β-cell mass expansion nor increased total pancreatic insulin content in FoxM1^Δpanc females compared with virgin females (Fig. 4 A and B).

To determine whether loss of FoxM1 might also affect β-cell differentiation, we examined Pdx1 expression by immunolabeling and Western blotting. Pdx1 expression was unchanged in FoxM1^Δpanc islets from either virgin or pregnant females compared with controls (data not shown). Sections from virgin and pregnant mice control and FoxM1^Δpanc pancreata were also examined for insulin expression. Islets from control animals showed uniform insulin reactivity (Fig. 4,C and data not shown). In contrast, although insulin expression in virgin FoxM1^Δpanc islets resembled controls, many cells labeled weakly in islets from female FoxM1^Δpanc at GD15.5 (Fig. 4 D and E). These low-level insulin-expressing cells were not undergoing apoptosis, as there was no increase in apoptotic β-cells in pregnant FoxM1^Δpanc females (data not shown). Thus, loss of FoxM1 results in both a decrease in the number of proliferating maternal β-cells during pregnancy as well as a decrease in insulin expression in individual β-cells. Overall, these data suggest that FoxM1 is a key factor promoting maternal β-cell compensation during pregnancy.

Virgin FoxM1^Δpanc female mice remain euglycemic (12,28) despite defects in β-cell mass. Thus, IPGTT was performed on 8- to 12-week-old virgin and pregnant females at GD12.5 and GD15.5 to determine whether pregnant female FoxM1^Δpanc mice could maintain glucose homeostasis despite defects in β-cell proliferation and β-cell mass expansion. FoxM1^Δpanc pregnant females showed no dramatic differences in fasting blood glucose levels compared with control females at either time point examined. However, mutant females displayed impaired glucose tolerance at GD12.5 (Fig. 5,A) and overt GDM at GD15.5 (Fig. 5,B). In humans, GDM is diagnosed if patients fail two or more values on the 100-g glucose tolerance test. The normal cutoff values are below 95 mg/dl at fasting, 180 mg/dl at 1 h, 155 mg/dl at 2 h, and 140 mg/dl at 3 h (2). An elevated fasting plasma glucose concentration is not required for a diagnosis of GDM (2). One hour after glucose challenge, blood glucose levels in FoxM1^Δpanc females were greater than 300 mg/dl. In addition, the blood glucose concentration 2 h after glucose challenge was ∼200 mg/dl (Fig. 5 B). The mean plasma insulin level in control mice at GD15.5 was 231.7 pg/ml at fasting and 977.2 pg/ml at 30 min (n = 2). In contrast, mean plasma insulin level in FoxM1^Δpanc females was 304.6 pg/ml at fasting and 361.3 pg/ml at 30 min (n = 4). These results demonstrate that FoxM1 is absolutely essential for the maintenance of glucose homeostasis during pregnancy and that loss of FoxM1 results in a reduced capacity for β-cell compensation in the face of increasing insulin resistance as pregnancy proceeds.

Because prior GDM elevates the risk for type 2 diabetes later in life, we examined whether there were any deleterious lasting consequences of pregnancy in FoxM1^Δpanc females. Eight days after giving birth, FoxM1^Δpanc females tended to have higher blood glucose levels during an IPGTT than postpartum control females; however, these did not reach statistical significance (supplementary Fig. 2). The impaired β-cell proliferation observed in FoxM1^Δpanc virgin females was still observed at P8 (Figs. 2 and 6). In addition, the islets in P8 FoxM1^Δpanc pancreata were smaller than those in controls (Figs. 6,A and B and 7,A). Sections from P8 control islets contained an average of 21 β-cells per islet, whereas sections from P8 FoxM1^Δpanc islets contained an average of only 9 β-cells per islet. After parturition, FoxM1^Δpanc females exhibited a significant increase in the proportion of small endocrine cell clusters (1–8 cells) versus definitive islets (>8 cells) in comparison with control mice (Fig. 6,C), likely contributing to the differences observed in average islet size (Fig. 7 A). Although it is possible that small endocrine clusters represent sections through the “top” of an islet of normal size, we observe this increase in small islet clusters only in postpartum FoxM1^Δpanc females; no differences were observed in the proportion of small endocrine cell clusters between control and FoxM1^Δpanc virgin females (12) (data not shown).

Despite the decrease in islet size, total β-cell mass was normal in FoxM1^Δpanc females after parturition (Fig. 7,B and C). This is likely due to a combination of increased individual β-cell size (Fig. 7,E) and an increase in the actual number of islet clusters in FoxM1^Δpanc pancreata (Fig. 7,D), possibly due to neogenesis. Postpartum FoxM1^Δpanc pancreata contained rare NGN3-expressing cells adjacent to ducts at P2 and P4 and juxtaductal insulin-positive cells at P8 (Fig. 6 D and E). Thus, NGN3 expression preceded insulin expression, suggestive of neogenesis. NGN3-positive cells were not observed in control pancreata at any time point examined (virgin, GD15.5, P2, P4, and P8) or in FoxM1^Δpanc pancreata from virgin, GD15.5, or P8 females.

PL is the primary lactogenic hormone that stimulates β-cell proliferation during pregnancy (34). To examine whether FoxM1 might act downstream of PL to promote β-cell proliferation during pregnancy, Foxm1 expression was examined in isolated Foxm1^fl/fl female islets exposed to PL in culture. As shown in Fig. 8,A, PL treatment increased expression of Bcl-xl, a known target of PL signaling. In addition, Foxm1 expression was increased approximately threefold. As a more direct test of whether FoxM1 functions downstream of PL, RIP-mPL transgenic mice were interbred with FoxM1^Δpanc mice. RIP-mPL mice express PL under the control of the rat insulin promoter and have a twofold increase in β-cell proliferation (17). FoxM1 was absolutely essential for the PL-mediated increase in β-cell proliferation in RIP-mPL transgenic mice (Fig. 8 B).

During pregnancy, β-cell mass increases by ∼50% due to stimulation of proliferation by GH, Prl, and most importantly, PL (6,15,16). Foxm1 expression strongly correlated with the dynamics of maternal β-cell proliferation during pregnancy and was induced in response to PL in cultured islets. In contrast to the partial requirement we recently described for FoxM1 in β-cell regeneration and proliferation after PPx (12), we found that FoxM1 is absolutely essential for β-cell mass expansion during pregnancy. These data, combined with the fact that FoxM1 is necessary for PL-induced β-cell proliferation in RIP-mPL transgenic mice, point to FoxM1 as a downstream target of PL signaling and an essential mediator of PL activity in islets. Because pregnancy is a physiological stimulus of β-cell proliferation, understanding the molecular pathways involved in pregnancy-induced β-cell mass expansion could facilitate identification of therapeutic targets for enhancing β-cell mass in vitro or in vivo.

Our data show that FoxM1 negatively regulates protein levels of the cell cycle inhibitor, Menin. We do not yet know whether this is at the level of transcription. However, the effect of FoxM1 on p27 in our model is mainly at the level of protein localization because total p27 levels were unchanged in FoxM1^Δpanc pancreata. FoxM1 promotes p27 nuclear export in other systems (27). Interestingly, Menin expression is negatively regulated by lactogens (18). In the future, it will be interesting to examine whether Stat5 directly regulates the Foxm1 gene in β-cells in response to PL and to determine the mechanism whereby FoxM1 regulates Menin in conjunction with pregnancy hormones.

In addition to enhancing β-cell mass and β-cell proliferation, PL and other pregnancy hormones enhance β-cell function. Functional changes in maternal β-cells include a lowered threshold for glucose-stimulated insulin secretion, increased glucokinase activity, enhanced insulin biosynthesis, increased gap junctional communication, and increased islet blood flow (6,34,,–37). Islets from FoxM1^Δpanc pregnant females showed reduced insulin labeling per β-cell and reduced total pancreatic insulin content. Thus, β-cell defects in the absence of FoxM1 may not be limited to a dramatic decrease in β-cell proliferation. Preliminary data from our laboratory suggest that loss of FoxM1 impairs β-cell function when animals are exposed to a high-fat diet (A. Ackermann Misfeldt, U.G. Kopsombut, and M.G., unpublished observations).

After parturition, a combination of decreased proliferation, increased apoptosis, and decreased cell size normally restores β-cell mass to prepregnancy levels within 10 days (38). It was striking to us that 8 days after giving birth, islets in FoxM1^Δpanc pancreata were significantly smaller than those in controls, although glucose homeostasis and β-cell mass were restored. Our data strongly suggest that in the absence of β-cell proliferation, postpartum FoxM1^Δpanc pancreata generate new β-cells via neogenesis, a process we previously showed does not require FoxM1 in embryos or adults (12).

GDM results from an interaction between genetic and environmental risk factors. Common variants in several candidate genes increase the risk of GDM in humans (e.g., KCNJ11, GK, and HNF4a) (4); however, very few mouse models of GDM currently exist. Heterozygous leptin receptor–deficient mice (db/+) spontaneously develop GDM due to increased peripheral insulin resistance (39). In addition, β-cell–specific inactivation of HNF4α resulted in impaired glucose tolerance in virgin females that worsens during pregnancy, although these animals do not manifest definitive GDM (40). HNF4α is required for increased β-cell proliferation and β-cell mass during pregnancy (40). Similarly, transgenic induction of Menin expression in islets of pregnant females resulted in moderate hyperglycemia by GD9.0 that worsened as pregnancy progressed (18); however, even the control females in this study showed elevated blood glucose. The model presented here is more similar to human GDM in that a phenotype is manifested specifically at midgestation; virgin and P8 FoxM1^Δpanc females show no overt phenotype. Human GDM patients likely have an underlying undetected β-cell defect that becomes apparent only under the physiological stress of pregnancy. Future studies will examine whether prior GDM predisposes FoxM1^Δpanc females to type 2 diabetes later in life, a phenotype that never manifests in FoxM1^Δpanc virgin females.

R.C.V. is supported by NIH/National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (RO1 DK072264). M.G. is supported by NIH/NIDDK (R01 DK071052).

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

Parts of this study were presented at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009.

We thank Dr. Amanda Ackermann Misfeldt and members of the Gannon laboratory for critical reading of the article and helpful suggestions. We also thank Young Ah Oh, Nikul Patel, Shidrokh Ardestani, Obi Umunakwe, Paige Cooper, David Lowe, and Usa G. Kopsombut for their technical assistance and Gregory Ayers in the Department of Biostatistics for assistance with data analysis. Islets were isolated with the help of Dr. Yanyun Gu and Dr. Zhongyi Chen and Anastasia Golovin of the Vanderbilt Islet Procurement and Analysis Core (DRTC 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).

This article is dedicated to the memory of the late Dr. Robert H. Costa, who was instrumental in initiating our studies of FoxM1.