SOX4 Allows Facultative β-Cell Proliferation Through Repression of Cdkn1a

Insulin-producing pancreatic β-cells are generated through differentiation during development and through replication postnatally (1–3). Previous work has demonstrated that cyclins and cyclin-dependent kinases are sufficient to drive β-cell proliferation, whereas loss of Ccnd2 leads to a reduction in β-cell mass (4). Furthermore, cell-cycle inhibitors have been implicated in the suppression of β-cell proliferation that accompanies disease. For example, genome-wide association studies have identified diabetes risk associated single nucleotide polymorphisms in Cdkn2a, Cdkn2b, and Cdkn1c (5,6). However, there have been few direct regulators of cell-cycle genes defined in β-cells.

Previously, we demonstrated that Sox4 is the predominant Sox family member involved in the specification and differentiation of β-cells (7,8) and that Sox4 expression is maintained in mature islets. Interestingly, genome-wide association studies on diabetes associated risk variants have identified single nucleotide polymorphisms in Sox4 regulatory elements, located in the Cdkal1 genomic region (5,9,10). In addition, SOX4 may also be important for regulating insulin secretion (11,12). However, there has yet to be an in vivo loss-of-function study that determines the physiological role for Sox4 in β-cells and explains its association with type 2 diabetes. In this study, we show that by directly modulating Cdkn1a (p21) expression, SOX4 is essential for the high-fat diet (HFD)–stimulated proliferation of β-cells in mice. These studies demonstrate that reductions in Sox4 expression may increase diabetes risk by impairing normal β-cell replication.

Chemicals were purchased from Sigma-Aldrich (Oakville, Ontario, Canada) or Thermo Fisher Scientific (Waltham, MA). Oligonucleotides were purchased from IDT (Coralville, IA). Tissue culture reagents were from HyClone Laboratories (Logan, UT) and cultureware from Falcon or Corning (Thermo Fisher Scientific).

All experiments were approved by the University of British Columbia Animal Care Committee. Animals were housed under a 12-h light/dark cycle, fed ad libitum with standard chow diet (5010; Lab Diets) or HFD (D12331; 58% fat; 7 weeks of age onwards; Research Diets). Mouse strains used were on the C57BL/6 background. Throughout, males were studied. Control mice were littermate Sox4^flox/flox, whereas experimental mice were Sox4^flox/flox;Pdx1-CreER^+/0 (13,14). All mice were administered 8 mg tamoxifen in corn oil (60 mg/mL) by oral gavage every other day beginning at 6 weeks of age. Human islet studies were approved by the BC Children’s and Women’s Hospital Research Ethics Board. Islets were provided by the Alberta Diabetes Institute Islet Core or University of Alberta Clinical Islet Transplantation Program.

Following a 10-h fast during the dark cycle, mice were weighed and fasting saphenous vein blood glucose levels obtained using a OneTouch UltraMini glucometer (Lifescan). Two grams per kilogram of 40% weight for volume d-glucose was delivered via intraperitoneal (i.p.) injection and blood glucose levels determined at 15, 30, 45, 60, and 90 min. Blood was collected at fasting and 10 min following i.p. glucose for determination of plasma insulin levels using ELISA (STELLUX; ALPCO Salem, NH). Blood glucose levels greater than the glucometer detection limit were reported as 33.3 mmol.

Intraperitoneal insulin tolerance test was performed following a 3-h fast during the light cycle with 1 units/kg i.p. insulin injection.

Immunostaining was performed on 5-μm paraffin sections as described (8). Primary antibodies used included: anti-SOX4 (1:200; 54020; Abcam); anti-insulin (1:500; A0564; DakoCytomation); anti-glucagon (GCG) (1;2,000; G2654; Sigma-Aldrich). Secondary antibodies were from Jackson ImmunoResearch Laboratories (706-096-148 and 715-586-151) or Vector Laboratories (BA-1000). 5-Ethynyl-2’-deoxyuridine (EdU) staining on sections was performed as described (8). A total of 0.5 mg EdU was administered (i.p.) twice a day for 7 days, beginning at 7 weeks of age.

Sixteen sections (250-μm intervals) of each pancreas were imaged using a BX61 microscope and tiled using the cellSens Dimension software (Olympus). CellProfiler v2.1.1 (15) was used to quantify images, and counts were normalized to total pancreatic nuclei. For EdU quantification, sections were imaged using confocal microscopy (Leica SP8; Leica Microsystems) and counted as above. All EdU⁺ insulin⁺ (INS⁺) cells were normalized to the total number of INS⁺ cells counted.

RNA was isolated with TRIzol, DNase treated (Turbo DNAse Free; Thermo Fisher Scientific), and reverse transcribed with Superscript III (Thermo Fisher Scientific) as previously described (8). TaqMan quantitative PCR (qPCR) and low-density TaqMan arrays were performed as described (8) and according to the manufacturer’s instructions, respectively. Primers used were: mouse Sox4 (forward [F], 5′-TCCTCTCCTGCCTCTTGG-3′; reverse [R], 5′-TCGAATTCCCGGACTATTGC-3′; and probe 5′-CCGCGCTCCCTTCAGTAGGTG-3′) and human SOX4 (F, 5′-GAGTCCAGCATCTCCAACC-3′; R, 5′-ACTCTTCGTCTGTCCTTTTCG-3′; and probe 5′-CGCGCCCTTCAGTAGGTGAAAAC-3′)

Chromatin was isolated from MIN6 cells and fragmented using an M220 focused ultrasonicator (Covaris), and chromatin immunoprecipitation was carried out as previously described using SOX4 antibodies (54020; Abcam) (8). Primer pairs used for qPCR included: Cdkn1a-A (F, 5′-CCCTGTACACTCTTGCTTACTT-3′ and R, 5′-CTGTCTAGGTCGTGTGTGATTC-3′); Cdkn1a-B (F, 5′-CCTCCCAAGTACCGTGATTTC-3′ and R, 5′-CGGATGTTCCTGACAGACATAG-3′); and Cdkn1a-C (F, 5′-CAGCTGAGCCTCAAAGGAATA-3′ and R, 5′-CAAGGGACCAAGGGAGATATAAG-3′).

The rate of change in β-cell number is equal to the difference in proliferation and apoptotic rates. TUNEL staining was identical in both genotypes (data not shown); thus, apoptotic rates were simplified (Eq. 1) from this analysis. Experimental data of EdU⁺; INS⁺/INS⁺ cells were used to determine the fraction of β-cells proliferating over 1 week (Eq. 2). Assuming a constant β-cell rate of change over 24 weeks, the number of newly derived β-cells was calculated using an exponential compound formula (Eq. 3).

Western blots were performed as described (16). Primary antibodies used included: anti-SOX4 (1:500; 19114.2; Custom HuCal; AbD Serotec); anti-CDKN1A (1:40; SCBT sc6246); and anti-GAPDH (1;100,000; G8795; Sigma-Aldrich). Secondary horseradish peroxidase–conjugated antibodies were from Jackson ImmunoResearch Laboratories (115-035-174 and 109-036-097).

All statistical analyses were performed with Prism5 (GraphPad Software, La Jolla, CA). Two-tailed unpaired Student t tests, Mann–Whitney U tests, one-way ANOVA with post hoc Tukey tests, Kruskal-Wallis H tests with post hoc Dunn tests, or multiple t tests with Holm-Sidak multiple-comparison correction were performed as appropriate. A P value <0.05 was considered significant. Error bars represent SEM.

Previous work demonstrated high levels of Sox4 expression in adult islets (8). Corroborating this with published RNA-sequencing data from both humans and mice (17–22), Sox4 was found to be the most highly expressed Sox family member (Fig. 1A). Type 2 diabetes in humans is correlated with reduced β-cell proliferation. In human islet samples, SOX4 expression was diminished (Fig. 1B). Thus, we sought to understand how dysregulated expression of Sox4 might contribute to progression of type 2 diabetes.

To understand how Sox4 regulates adult β-cell function, male Pdx1-CreER^+/0;Sox4^flox/flox mice and littermate Sox4^flox/flox mice were administered tamoxifen at 6 weeks of age to generate knockout (βS4KO) or control mice, respectively. Tamoxifen administration led to a significant decrease in Sox4 mRNA expression by 8 weeks of age in the βS4KO mice (Fig. 1C) and significantly reduced SOX4 immunoreactivity(+) at the experimental end point (30 weeks of age; Fig. 1D and E), suggesting effective knockout without recovery over the experimental time course (Fig. 1F).

To ascertain whether Sox4 was required for normal glucose homeostasis, weekly body weight and biweekly glucose tolerance measurements were taken. Chow-fed mice exhibited no significant differences in body weight or glucose tolerance up to the end point at 30 weeks of age (Supplementary Fig. 1). In another cohort of mice, HFD feeding was initiated at 7 weeks of age to stimulate β-cell compensation and model human prediabetes. HFD-fed control and βS4KO mice gained weight at the same rate (Fig. 2A). However, HFD-fed βS4KO mice developed significantly worse glucose intolerance and diabetes by 30 weeks of age (Fig. 2B–F). No significant differences in glucose disposal following i.p. insulin injection were observed (Supplementary Fig. 2), suggesting the defect was rooted in the β-cell and not the periphery.

To determine how glucose tolerance was impaired in βS4KO mice, plasma insulin levels were measured following glucose challenge. HFD-fed βS4KO mice had significantly reduced stimulated plasma insulin levels (Fig. 2G). Reduced plasma insulin levels can result from reductions in either insulin secretion or β-cell mass. To test the former, ex vivo glucose-stimulated insulin secretion assays were performed on groups of 50 size-matched islets isolated from 30-week-old HFD-fed mice. No significant differences were observed in glucose or KCl-stimulated insulin secretion (Fig. 2H) or insulin content (Fig. 2I). Taken together, these data demonstrated that Sox4 is required for β-cell compensation to excess nutrients. In addition, loss of Sox4 did not impair β-cell function, suggesting that the reduction in plasma insulin level was because of decreased β-cell mass.

Quantification of β-cell numbers at 30 weeks of age revealed a 40% reduction in the number of INS⁺ cells in HFD-fed βS4KO pancreases versus controls (Fig. 3A–C). No significant difference in the number of GCG⁺ cells was observed (Fig. 3D). As proliferation is essential to balance β-cell attrition with age and prevent glucose intolerance and diabetes, we quantified β-cell proliferation in βS4KO mice during the first week of HFD feeding as previously described (23). After 1 week, 7% of total β-cells were EdU-labeled in control mice (Fig. 3E and G), similar to published cumulative labeling results in β-cells (1). Importantly, βS4KO mice had a significant reduction in the number of EdU⁺INS⁺ cells (4%; Fig. 3F and G), suggesting that Sox4 is essential for HFD-stimulated β-cell replication.

In agreement with previous observations (23), the number of β-cells did not differ between genotypes following 1 week of HFD (Fig. 3H). Thus, we used this time point to identify potential SOX4 target genes that may regulate β-cell proliferation prior to reductions in β-cell mass. Among an array of cell-cycle genes, only Cdkn1a and Mcm10 had significant twofold increases in expression (Fig. 4A and B). Cdkn1a has been associated with β-cell cycle arrest and reductions in β-cell proliferation (24). Consistent with the gene expression changes, CDKN1A protein was also significantly increased in βS4KO islet cell-lysates after 1 week of HFD feeding (Fig. 4C). Further analysis of the Cdkn1a genomic region revealed multiple SOX4 consensus DNA-binding motifs (Supplementary Fig. 3A) (25). Chromatin immunoprecipitation from MIN6 cell extracts using SOX4 antibodies demonstrated enrichment of SOX4 at the Cdkn1a genomic regions containing these putative SOX4 binding sites (Fig. 4D–F). Together, these studies revealed a previously unappreciated role for SOX4 in promoting β-cell proliferation through the repression of the cell-cycle inhibitor Cdkn1a.

Replenishment of β-cell numbers relies on cell division (26). Mirroring the neuronal system (27), our studies established pancreatic SOX4 as a factor that initially specifies a terminally differentiated β-cell state (8) and then supports β-cell proliferation. This is in agreement with experimental evidence that SOX4 can act as an oncogene (28) as well as regulate cell-cycle genes and proliferation (29,30).

HFD-fed βS4KO mice exhibited glucose intolerance with reduced plasma insulin levels. Quantification of pancreatic sections revealed a reduction in β-cell numbers in βS4KO pancreas. Moreover, β-cell proliferation was reduced in βS4KO mice when assessed at 8 weeks of age, following 1 week of HFD feeding. To estimate the impact of a reduction in the β-cell proliferation rate from 7 to 4% over a period of 24 weeks, we used the experimentally derived rate from the 1-week EdU-labeling period in a mathematical model of growth, compounded over 24 weeks: Cells_final = Cells_initial(1 + Proliferation)^{24 weeks} (see research design and methods). Assuming a steady state of β-cell proliferation, this estimated a 46% decrease after 24 weeks relative to controls, roughly accounting for our observed 40% reduction in β-cell number. Caveats to this simplified model include a non–steady-state proliferation rate over time; in fact, previous estimates suggest that β-cell proliferation rate is significantly higher at birth with an exponential decay (1,31). In addition, we did not observe significant TUNEL staining at 8 weeks of age (data not shown) in our model, but may have missed a burst of β-cell death. Therefore, values were simplified in our estimation, as it suggested changes in cell survival would not affect relative β-cell mass in βS4KO versus controls.

Our screen of potential target genes implicated in cell-cycle control identified Cdkn1a as the best candidate, as it was among the most differentially expressed genes and contained multiple SOX4-binding motifs in its promoter. Cdkn1a is a cyclin-dependent kinase inhibitor that has been correlated with endocrine specification and exit from the cell cycle (24). Cdkn1a expression has also been demonstrated to increase with age, a setting correlated with reduced proliferation (32). Although a germline Cdkn1a gene knockout did not lead to increased proliferation because of possible redundancy and developmental compensation by other family members (33), recent evidence suggested that CDKN1A inhibits β-cell cycle progression and targeting the protein with small-molecule inhibitors may improve human islet cell proliferation (34). Mcm10 may also contribute to reduced β-cell replication, as Mcm10 may be coexpressed with Cdkn1a during late G₁ phase (35), although this awaits further exploration in β-cells. In sum, the evidence points toward SOX4 as a positive modulator of nutrient-stimulated β-cell proliferation through suppression of Cdkn1a expression.

In conclusion, Sox4 remains important following β-cell differentiation to maintain effective cellular replication rates through modulating the expression of Cdkn1a. Although Sox4 is not required for β-cell survival, restoring normal expression in those with prediabetes may maintain β-cell mass and prevent development of frank type 2 diabetes.

Acknowledgments. The authors thank V. Lefebvre (Lerner Research Institute, Cleveland, OH) for mouse lines that enabled this work, David Lin (BC Children's Hospital Research Institute, Vancouver, British Columbia, Canada) for experimental expertise, and the members of the Lynn Laboratory (Vancouver, British Columbia, Canada) for technical support, discussion, and critical reading of the manuscript.

Funding. This work was supported by operating grants from the Canadian Institutes of Health Research (MOP142222 to F.C.L.) and JDRF (2-2011-91 to F.C.L.). Salary was supported by the Michael Smith Foundation for Health Research (5238 BIOM), the Canadian Diabetes Association, and the BC Children’s Hospital Research Institute (to F.C.L.). Fellowship support was provided by the Canadian Institutes of Health Research–BC Transplantation Trainee Program (to E.E.X. and T.S.) and the Manpei Suzuki Diabetes Foundation (to S.S.).

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

Author Contributions. E.E.X., S.S., T.S., and C.N. generated and analyzed data. E.E.X. and F.C.L. designed experiments and drafted the manuscript. All of the authors approved the version of the manuscript to be published. F.C.L. 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.