Pancreatic Pericytes Support β-Cell Function in a Tcf7l2-Dependent Manner

Type 2 diabetes has a strong genetic component, with a number of genetic variations associated with an increased risk to develop this disease (1,2). In particular, polymorphism in TCF7L2 (TCF4) is associated with increased risk to diabetes (3). This gene encodes a member of T-cell factor/lymphoid enhancer factor (TCF/LEF) transcription factors family, which functions downstream of the canonical Wnt signaling pathway by recruiting β-catenin to target genes (4). Diabetes-associated alleles of TCF7L2, such as the T allele of the single-nucleotide polymorphism in rs7903146, are associated with impaired glucose-stimulated insulin secretion (GSIS) and insulin production but intact hepatic function and insulin sensitivity (3,5–8). The T allele of the rs7903146 variant was predicted to result in an inactive protein lacking its DNA-binding domain (9). However, how TCF7L2 functions to regulate glucose homeostasis remains an open question.

To date, the use of mouse systems to determine the cellular mechanism(s) through which abnormal Tcf7l2 activity contributes to β-cell dysfunction has produced conflicting results. As opposed to humans, hepatic phenotypes dominate the abnormal glucose levels observed upon body-wide deregulation of Tcf7l2 expression in mice (10–12). β-Cell–specific inference with Tcf7l2 activity using mouse genetic tools yielded discrepant results, with some studies showing reduced β-cell mass and glucose intolerance and others showing normal glucose response (12–17). This contradiction could partially stem from the use of different approaches to interfere with Tcf7l2 activity, such as knocking down the endogenous gene (12,14,15) versus overexpressing a dominant-negative (DN) form (13,16). Tcf7 (Tcf1), a member of the TCF/LEF family with high homology to Tcf7l2, was recently shown to play a central role in maintaining β-cell mass (18), raising the possibility that overexpressing a DN Tcf7l2 interferes with the activity of other TCF/LEF proteins in β-cells. Interestingly, although β-cell–selective deletion of Tcf7l2 resulted in their reduced mass (15), selective deletion of this transcription factor DNA-binding domain affected neither β-cell function nor mass (12). Accordingly, nonautonomous roles of Tcf7l2 in regulating β-cell function were suggested (17).

β-Cells rely on extrinsic cues, including those provided by cells of the islet microenvironment, for their proper function (19,20). We and others recently showed that pericytes, which together with endothelial cells make the dense islet capillary network, support β-cell function and glucose homeostasis (21,22). Although abnormalities in islet pericytes were implicated in obesity and type 2 diabetes (23), whether impaired pericyte function contributes to β-cell dysfunction and disease progression remains an open question.

Profiling gene expression of pancreatic pericytes revealed the expression of Tcf7l2 in these cells. We hypothesized that Tcf7l2 activity in pancreatic pericytes is required for their ability to properly support β-cell function. To test our hypothesis, we selectively inactivated this transcription factor in these cells by combining two transgenic mouse lines: Tcf7l2^flox, which allows Cre-mediated deletion of this transcription factor DNA-binding domain (24), and Nkx3.2-Cre (25), which selectively targets mural cells of the pancreas (21). Our results show an impaired glucose tolerance but intact insulin sensitivity in male mice homozygous for mutated Tcf7l2. Our analysis pointed to impaired GSIS and reduced expression of genes required for β-cell function and maturity upon inactivation of pericytic Tcf7l2. Lastly, we linked Tcf7l2-dependent pericytic expression of bone morphogenetic protein 4 (BMP4) to glucose regulation. To conclude, our results indicate that pericytic Tcf7l2 activity is required for β-cell function and glucose homeostasis. Our findings further point to the contribution of abnormal pericytes activity to diabetes progression.

All experiments were performed according to protocols approved by the Tel Aviv University Committee on Animal Research. Nkx3.2-Cre (Nkx3–2^tm1(cre)Wez) (25) and Tcf7l2^flox (Tcf7l2^tm2.1Cle) (24) mice were gifts from Warren Zimmer (Texas A&M University, College Station, TX) and Hans Clevers (Hubrecht Institute, Utrecht, the Netherlands), respectively. R26-yellow fluorescent protein (YFP; enhanced [E]YFP) (Gt(ROSA)26Sor^tm1(EYFP)Cos) mice were obtained from The Jackson Laboratory. For diet-induced obesity, mice were fed a high-fat diet (HFD; 60% fat [kCal]; Teklad) beginning at 6 weeks of age. Mice were intraperitoneally injected with dextrose (Sigma-Aldrich), insulin (Lilly), or mouse recombinant BMP (rBMP4; R&D) when indicated. For analysis of functional vasculature, fluorescein-labeled tomato lectin (Vector) was injected intravenously and allowed to circulate for 5 min before the animal was euthanized.

Cell isolation was performed as described (26). Cells were stained with primary antibodies (Supplementary Table 1) when indicated and collected using FACSAria (BD Biosciences) or analyzed using a Gallios flow cytometer (Beckman Coulter) and Kaluza software (Beckman Coulter).

Islets were isolated according to standard protocols (21). For insulin secretion, after overnight culture, isolated islets were incubated in RPMI medium supplemented with glucose for 1 h. Pancreas and islet insulin was extracted by overnight incubation in 1.5% HCl and 70% ethanol mixture. Hormone levels were determined using mouse Ultrasensitive Insulin ELISA (Alpco), mouse Proinsulin ELISA (Alpco), and glucagon-like peptide 1 (GLP-1[7-36]) Active Elisa kit (Millipore).

Dissected tissues were fixed in paraformaldehyde (4%), followed by cryosectioning. Tissue sections were stained with primary antibodies (Supplementary Table 1), followed by secondary fluorescent antibodies (Alexa Fluor; Invitrogen). After tissues were stained with anti-Tcf7l2 antibody, the TSA Fluorescein System (PerkinElmer) was used. The In Situ Cell Death Detection Kit (Roche) was used for TUNEL assay. Images were acquired using BZ-9000 BioRevo (Keyence) and SP8 confocal (Leica) microscopes.

Analysis of islet vasculature was performed as described (21). For measurement of β-cell mass, immunostained paraffin-embedded tissue sections were counterstained with HCS CellMask Stain (Invitrogen) to label the whole tissue sections. Sections were automatically imaged using IN Cell 2000 analyzer (GE Healthcare) and analyzed by developer software (GE Healthcare).

RNA was extracted using PureLink RNA Micro kit (Invitrogen). For RNA deep sequencing, amplification, cDNA library preparation, sequencing, and bioinformatics analysis were performed using commercial services (Otogenetics). Gene expression data have been deposited in ArrayExpress (https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-5325/). For quantitative (q)PCR analysis, TaqMan and SYBR Green assays (Invitrogen) (Supplementary Table 2) were used, normalized to GAPDH and cyclophilin expression, respectively.

Paired data were evaluated using two-tailed Student t test.

We recently showed that pericytes support β-cell function (21). Here, we set to elucidate the molecular basis of pericyte activity. To this end, we used the Nkx3.2-Cre mouse line (25) to manipulate pancreatic mural cells. Nkx3.2 (Bapx1) is expressed in the mesenchymal compartment of the embryonic gut, stomach, and pancreatic buds, as well as in skeletal somites (27). We recently showed that in the adult pancreas, Nkx3.2-Cre line targets mural cells, including islet pericytes and vascular smooth muscle cells (vSMCs), but no other pancreatic cell types, including epithelial and endothelial cells (21) (Supplementary Fig. 1). Pericytes in the exocrine pancreas, identified by expression of neural/glial antigen 2 (NG2) and desmin, were also targeted by this Cre as apparent fluorescent labeling of these cells in the pancreas of Nkx3.2-Cre;R26-YFP mice (Supplementary Fig. 1). Of note, this mouse line does not target hepatic pericytes, which were shown to regulate insulin response (28) (Supplementary Fig. 1); thus, our analysis indicated that the Nkx3.2-Cre line targets mural cells in the endocrine and exocrine pancreas.

We next characterized pancreatic mural cells by profiling their gene expression. To this end, cells were sorted from pancreatic tissues of Nkx3.2-Cre;R26-YFP mice based on their fluorescent labeling (Supplementary Fig. 1). Of note, vast majority of labeled cells express platelet-derived growth factor receptor-β (PDGFR-β), which is expressed by pericytes but not vSMCs (Supplementary Fig. 1) (29). RNA was extracted from sorted cells and subjected to deep sequencing (Supplementary Table 3). Gene Ontology term analysis revealed that pancreatic mural cells were enriched with components of Wnt signaling (Fig. 1A; 126 genes). Interestingly, these cells expressed two of the four mammalian TCF/LEF transcription factors, Tcf7l1 and Tcf7l2 (Fig. 1B). Considering the association of polymorphism in TCF7L2 with β-cell dysfunction and diabetes, we focused our analysis on this transcription factor.

To validate Tcf7l2 expression, we performed qPCR, Western blot, and immunofluorescence analyses. Tcf7l2 transcript and protein were detected in purified pancreatic mural cells (Fig. 1C and Supplementary Fig. 2). Notably, Tcf7l2 mRNA levels in mural cells were 5-fold higher than in islets and 22-fold higher than in bulk pancreatic tissue (Fig. 1C). Tcf7l2 protein was detected in the nuclei of pancreatic pericytes, including those associated with islets (Fig. 1D and E), but not in the nuclei of vSMCs, identified by expression of α-smooth muscle actin (α-SMA) and localization around large blood vessels (Fig. 1F). In agreement with previous studies reporting low Tcf7l2 transcript and protein levels in pancreatic endocrine cells (30–33), we did not detect this transcription factor in β-cells and isolated islets (Fig. 1E and Supplementary Fig. 2). To conclude, our analyses revealed the expression of Tcf7l2 by pancreatic pericytes.

To test the requirement of pericytic Tcf7l2 for glucose regulation, we set to interfere with this transcription factor activity in these cells. The diabetes-associated T allele of rs7903146 variant was predicted to result in an inactive Tcf7l2 protein lacking its DNA-binding domain (9). We therefore used a previously described transgenic mouse line, Tcf7l2^flox, allowing Cre-mediated deletion of the endogenous Tcf7l2 DNA-binding domain, rendering it inactive (24). Of note, all splice variants are present in mice carrying this transgene (24). To selectively inactive this transcription factor in pancreatic pericytes, the Tcf7l2^flox transgenic mouse line was crossed with the Nkx3.2-Cre line. To verify recombination of the Tcf7l2 locus, we analyzed its transcript levels by using primers providing detection of wild-type Tcf7l2 but not its recombined form (12,24). To allow isolation of pancreatic mural cells by flow cytometry (as described in Supplementary Fig. 1), a R26-YFP transgene was included to generate Nkx3.2-Cre;R26-YFP;Tcf7l2^flox/+ and Nkx3.2-Cre;R26-YFP;Tcf7l2^flox/flox mice, as well as Nkx3.2-Cre;R26-YFP control mice. As shown in Fig. 2A, wild-type Tcf7l2 was nearly absent from pancreatic mural cells of homozygous mice and was significantly reduced in cells of heterozygous mice compared with control mice. Of note, the expression levels of wild-type Tcf7l2 in islets isolated from homozygous and heterozygous mice were comparable to those of controls (Fig. 2B).

We analyzed glucose response in mice by performing intraperitoneal glucose tolerance tests (IPGTTs) on three groups of mice: homozygous for the inactive Tcf7l2 allele (Nkx3.2-Cre;Tcf7l2^flox/flox), heterozygous for this allele (Nkx3.2-Cre;Tcf7l2^flox/+), and nontransgenic controls (Cre negative: Tcf7l2^flox/+ or Tcf7l2^flox/flox). Of note, mice expressing the Nkx3.2-Cre transgene by itself (i.e., do not carry the Tcf7l2^flox transgene) displayed comparable glucose response to Cre-negative control mice (Supplementary Fig. 3). As shown in Fig. 2C, our analysis revealed that 13-week-old homozygous, but not heterozygous male mice, display an impaired glucose response compared with littermate controls (Fig. 2C). The TCF7L2 rs7903146 T-allele has a modest effect on β-cell function, which becomes more evident when insulin action decreases (34). We tested whether metabolic stress aggravates glucose intolerance of transgenic mice by feeding mice the HFD (60% fat) to induce obesity. Our analysis revealed that heterozygous and homozygous obese animals were glucose intolerant (Fig. 2D). Thus, our findings indicate that reduced levels of active Tcf7l2 in pericytes (in heterozygous mice) were sufficient to maintain glucose response in lean but not obese mice and that its complete loss (in homozygous mice) induced glucose intolerance in both lean and obese animals.

Polymorphism in TCF7L2 is associated with an increased risk of diabetes in women and men (8); however, we did not observe differences in glucose response in transgenic and control female mice (Supplementary Fig. 3). Glucose metabolism differs in female and male mice (35); thus, sex-dependent differences may underlie the distinct phenotype observed in female and male Nkx3.2-Cre;Tcf7l2^flox/flox mice.

The Nkx3.2-Cre mouse line has nonpancreatic expression in the gastrointestinal mesenchyme and skeleton (25,27). We therefore analyzed for potential changes in function of these tissues in Nkx3.2-Cre;Tcf7l2^flox/flox and Nkx3.2-Cre;Tcf7l2^flox/+ mice that could contribute to their glucose intolerance. The three analyzed mouse groups show comparable body weight when fed both regular chow and HFD, indicating normal food uptake and digestion (Supplementary Fig. 4). Next, we analyzed for GLP-1 production by analyzing gut expression of Pcsk1 and Gcg (encoding prohormone convertases 1/3 and proglucagon, respectively) and measuring serum GLP-1 levels, and found them comparable in Nkx3.2-Cre;Tcf7l2^flox/flox and control mice (Supplementary Fig. 4). Finally, insulin sensitivity was comparable between Nkx3.2-Cre;Tcf7l2^flox/flox, Nkx3.2-Cre;Tcf7l2^flox/+, and control (Cre-negative) male mice fed the normal diet and HFD (Supplementary Fig. 4).

To conclude, our results implicate that Tcf7l2 activity in pancreatic pericytes is required for glucose regulation in vivo, without affecting insulin sensitivity. Furthermore, our analysis suggests that the requirement of pericytic Tcf7l2 activity for glucose regulation is more evident upon metabolic stress.

Pericytes support endothelial cell function and blood flow (29). We therefore analyzed whether Tcf7l2 inactivation in pancreatic pericytes interferes with islet vascularization. To this end, Nkx3.2-Cre;Tcf7l2^flox/flox and control mice were intravenously injected with tomato lectin to label functional vessels. Our analysis revealed intact islet vascularization distribution and density in transgenic mice (Fig. 3A and B). In agreement, expression levels of the hypoxia gene Hif1a was comparable in islets isolated from the two mouse groups (Fig. 3C). Thus, our analysis indicates that loss of pericytic Tcf7l2 activity did not interfere with functionality of islet vasculature.

We previously showed that islet pericytes support β-cell function (21). We therefore set to determine whether β-cell dysfunction underlies the observed glucose intolerance upon loss of Tcf7l2 activity in pancreatic pericytes. We tested whether the impaired glucose response is associated with insufficient GSIS by measuring glucose-stimulated serum insulin levels and found they were lower in Nkx3.2-Cre;Tcf7l2^flox/flox mice (Fig. 4A). Next, we analyzed whether β-cell mass or function, or both, were affected in Tcf7l2-deficient mice. Our analysis indicated normal islet morphology and comparable pancreatic and β-cell mass in transgenic mice (Fig. 4B–D). In agreement with these findings, we observed neither β-cell death nor expression of genes associated with β-cell stress in islets of transgenic mice (i.e., Chop, Atf4) (Supplementary Fig. 5). To test for potential changes in β-cell function, we analyzed GSIS of isolated islets and found that those of transgenic mice secreted less insulin in response to high glucose levels compared with control islets (Fig. 4E). Thus, our analysis points to β-cell dysfunction upon inactivation of pericytic Tcf7l2.

Next, we measured insulin content in isolated islets and pancreatic tissues from Nkx3.2-Cre;Tcf7l2^flox/flox and control animals and found it was significantly reduced in homozygous animals (Fig. 4F and G). Although gene expression analysis indicated that the levels of Ins1 and Ins2 transcripts, encoding insulin, were on average lower in transgenic islets, this reduction was not statistically significant (Fig. 4H). We detected a statistically significant reduction in expression levels of Pcsk1, required for posttranslational processing of insulin, in Nkx3.2-Cre;Tcf7l2^flox/flox islets (Fig. 4H), but the ratio between proinsulin and insulin levels was comparable in transgenic and control islets (Fig. 4I). In addition, Gcg levels were comparable in Nkx3.2-Cre;Tcf7l2^flox/flox and control islets (Fig. 4J). To conclude, we observed reduced islet and pancreatic insulin content upon inactivation of pericytic Tcf7l2.

Nkx3.2-Cre;Tcf7l2^flox/flox islets secrete a smaller portion of their insulin content in response to glucose challenge, indicating an impaired GSIS independent of abrogated insulin production (Fig. 4E’). To determine whether the observed impaired GSIS in Nkx3.2-Cre;Tcf7l2^flox/flox islets is associated with changes in expression levels of genes required for glucose sensing and insulin secretion, we analyzed islets for expression of Glut2, Kir6.2, and Sur1. As shown in Fig. 4K, all three genes were expressed at lower levels in Nkx3.2-Cre;Tcf7l2^flox/flox islets than in control islets. In contrast, expression of Glpr1, encoding the GLP-1 receptor, was comparable in transgenic and control islets (Fig. 4L). β-Cell function and gene expression have been shown to depend on transcription factors, including MafA, Pdx1, and NeuroD1 (2). We observed reduced transcripts level of all three in islets from Nkx3.2-Cre;Tcf7l2^flox/flox mice compared with control islets (Fig. 4M). Thus, inactivation of pericytic Tcfl72 is associated with impaired expression of β-cell genes.

To conclude, our findings indicate that although β-cell mass was unaffected in Tcf7l2 transgenic mice, their functionality was impaired. Furthermore, our analysis suggests that normal expression of β-cell genes associated with their function and maturity is dependent on Tcf7l2 activity in pancreatic pericytes.

Our findings point to a Tcf7l2-dependent activity of pancreatic pericytes in supporting β-cell function and gene expression. We therefore analyzed potential changes in pancreatic pericytes upon inactivation of Tcf7l2. Our immunofluorescence analysis revealed that pericytes are localized in proximity to endothelial cells in control and in transgenic islets (Fig. 5A). However, morphometric analysis revealed that pericyte coverage was mildly reduced in islets of Nkx3.2-Cre;Tcf7l2^flox/flox mice compared with control islets (by 7%) (Fig. 5B).

Next, we analyzed whether loss of Tcf7l2 activity affects their gene expression. To this end, pancreatic mural cells were purified by flow cytometry (as described in Supplementary Fig. 1) from Nkx3.2-Cre;R26-YFP;Tcf7l2^flox/flox and control Nkx3.2-Cre;R26-YFP mice, and islets were isolated from nontransgenic mice and their RNA was extracted and sequenced. We hypothesized that pericytes secrete factors to support β-cells under normal conditions and that the expression of some of these factors is Tcf7l2 dependent. We thus focused our analysis on genes that encode secreted ligands that are expressed by nontransgenic pancreatic mural cells but not by isolated islets (Supplementary Table 3). The expression levels of seven of these genes were significantly lower in Tcf7l2-transgenic pancreatic mural cells (Fig. 5C and Supplementary Table 3): Bmp4, Ccl2, Ccl7, C7, Il6, Fam150b, and Nmb. Tcf7l2 was previously shown to directly regulate the expression of first three genes (36–38). Importantly, β-cells were shown to express the receptors for BMP4, neuromedin B, and interleukin (IL)-6 (encoded by Bmp4, Nmb, and Il6, respectively), and these three factors were implicated in β-cell function (39–41).

To conclude, our analysis showed that lack of Tcf7l2 activity in pancreatic pericytes led to mildly reduced islet pericyte density and further resulted in impaired expression of secreted ligands by these cells, some of which were implicated in β-cell function.

β-Cells were shown to depend on the activity of the BMP4 receptor, BMPR1A, for their proper function in vivo (40). In addition, elevating BMP4 levels in vivo (by pancreatic expression of a Bmp4 transgene or systemic administration of rBMP4) improved the glucose response in mice and promoted the mature β-cell phenotype (40). Importantly, Tcf7l2 was shown to directly promote Bmp4 expression through its identified binding sites on this gene promotor (36,42). We therefore hypothesized that BMP4 produced by pericytes in a Tcf7l2-dependent manner supports β-cell function. We tested our hypothesis by first validating pericytic Bmp4 expression by qPCR analysis, which revealed reduced transcript levels in Tcf7l2-deficient mural cells to a third of the level found in control cells (Fig. 6A).

To study the contribution of reduced BMP4 production by pericytes to the observed phenotype in Nkx3.2-Cre;Tcf7l2^flox/flox mice, we intraperitoneally injected rBMP4 to homozygous mice (40). As shown in Fig. 6, glucose tolerance and GSIS of rBMP4-treated animals were significantly improved compared with untreated transgenic animals. Importantly, glucose response and insulin secretion of rBMP4-treated Nkx3.2-Cre;Tcf7l2^flox/flox mice were comparable to that of nontransgenic control animals (Fig. 6B and C).

To analyze whether the improved GSIS upon rBMP4 treatment of Nkx3.2-Cre;Tcf7l2^flox/flox mice is associated with improvement in their β-cell mature phenotype, we cultured isolated islets in the presence or absence of this recombinant protein. Treating Nkx3.2-Cre;Tcf7l2^flox/flox islets with rBMP4 for 24 h promoted the expression of Pdx1 and MafA, encoding transcription factors associated with mature β-cell phenotype (Fig. 6D). Of note, in agreement with previous studies (43,44), we could not observe changes in the genes expression levels in similarly treated wild-type islets (Supplementary Fig. 6).

To conclude, our analysis showed that extrinsic BMP4 was sufficient to rescue glucose response, GSIS, and mature β-cell gene expression in mice deficient of pericytic Tcf7l2. Thus, our findings suggest that pericytes support β-cell function through Tcf7l2-depedent production of BMP4.

Here, we show that pericytes support β-cell function and glucose response in a Tcf7l2-dependent manner. To analyze the requirement of this transcription factor for pericyte function, we selectively expressed a mutated form of Tcf7l2, lacking its DNA-binding domain, in pancreatic pericytes. Loss of pericytic Tcf7l2 activity interfered with glucose regulation in mice due to impaired β-cell function (Figs. 2 and 4). Our analysis further indicated Tcf7l2-dependent pericytic expression of secreted factors that were previously implicated in β-cell function, including BMP4 (Fig. 5). Finally, we showed that treatment of Tcf7l2-deficient mice with exogenous rBMP4 was sufficient to rescue their glucose intolerance and impaired GSIS phenotype (Fig. 6). Thus, we suggest that pancreatic pericytes express secreted factors in a Tcf7l2-dependent manner to support β-cell function and glucose response. Our findings propose that impaired pericytic activity perturbs β-cell function, thus potentially contributing to disrupted glucose regulation and diabetes progression.

Abnormalities in islet pericytes were implicated in obesity and diabetes (23). We, along with others, showed that pericytes directly support β-cell function and glucose regulation (21,22). Interestingly, SORCS1, a type 2 diabetes–associated gene that encodes a PDGF-binding protein, was suggested to regulate pericyte function (23,45). However, whether impaired pericyte activity contributes to diabetes progression remains an open question. The findings of this study suggest that diabetes-associated changes in pancreatic pericytes lead to impaired glucose regulation. This raises the possibility that abnormalities in the islet microenvironment in individuals with diabetes contribute to β-cell dysfunction and loss of glucose regulation.

Our analysis revealed that pericytes produce BMP4 in a Tcf7l2-dependent manner to support glucose response (Figs. 5 and 6). Treatment of Nkx3.2-Cre;Tcf7l2^flox/flox mice with rBMP4 was sufficient to rescue their glucose tolerance, strengthening the importance of this factor in supporting proper β-cell function. Although the BMP4-BMPR1A pathway was shown to promote β-cell function and gene expression in vivo (40), treating isolated islets with rBMP4 did not affect β-cells or impair their survival and function (43,44), pointing to the requirement of regulated BMP4 levels for proper β-cell function. Loss of BMPR1A function in β-cells resulted in a more severe glucose intolerance phenotype than the one we observed upon loss of pericytic Tcf7l2 (40) (Fig. 2). These differences might reflect the presence of residual pericytic BMP4 in transgenic mice or additional pancreatic BMP4 sources, or both, as previously suggested (40).

Loss of pericytic Tcf7l2 activity was associated with a reduction in pancreatic and islet insulin content (by 50 and 38%, respectively) (Fig. 4). Together with the unaffected insulin-to-proinsulin ratio, this observation indicates an impaired insulin biosynthesis in Nkx3.2-Cre;Tcf7l2^flox/flox mice. Ins1 and Ins2 transcript levels were an average of 39 and 33%, respectively, lower in transgenic islets (Fig. 4), but this reduction was not statistically significant. Thus, reduced insulin content accompanying the loss of pericytic Tcf7l2 could potentially result from an impaired posttranscriptional regulation of insulin biosynthesis, such as compromised proinsulin translation (46).

Tcf7l2 was proposed to support β-cell function in a cell-autonomous manner (14,15). Whether impaired Tcf7l2 activity in β-cells is sufficient to drive diabetes progression remains controversial, however (12,17,47,48). Accordingly, Tcf7l2 was suggested to regulate β-cell function in a nonautonomous manner (12,17); for example, pancreatic and nonpancreatic incretin production was recently shown to depend on this transcription factor (49,50). Our findings provide additional evidence for a nonautonomous role of Tcf7l2 in β-cell function through regulation the pericyte/β-cell axis.

Polymorphism in TCF7L2 has a strong correlation to type 2 diabetes in humans (3,5,6). In particular, multiple studies have shown that the T allele of rs7903146 increases by 30% the risk of developing this disease (5). Individuals harboring one or two copies of this allele display reduced basal and glucose-stimulated insulin secretion but maintain hepatic function (6,7,34). As shown in this study, mice lacking Tcf7l2 activity in their pancreatic pericytes were glucose intolerant but did not develop diabetes, even when fed the HFD (Fig. 2). These discrepancies could reflect the evident difference in the physiologies of humans and mice. Alternatively, TCF7L2 may act in multiple cell types to regulate glucose homeostasis, and loss of its activity in a single cell type is insufficient to drive diabetes progression. Lastly, although impaired TCF7L2 activity increases the risk of developing diabetes, additional genetic and environmental factors contribute to disease progression (1). Nevertheless, impaired β-cell function upon loss of pericytic TCF7L2 activity proposes a cellular mechanism through which mutations in this transcription factor predispose individuals to diabetes.

Acknowledgments. The authors thank Maya Avraham and Shani Mizrachi (Tel Aviv University) for technical assistance.

This work was performed in partial fulfillment of the requirements for a PhD degree for L.S. from the Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel.

Funding. This work was supported by European Research Council starting grant (336204) to L.L.

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

Author Contributions. L.S. conducted experiments, acquired and analyzed data, and wrote the manuscript. E.R., A.E., and H.C.G. conducted experiments and acquired and analyzed data. S.W.-A. analyzed data. M.L., L.K.-M., A.H., and D.B. conducted experiments and acquired data. A.P. and M.W. provided reagents. L.L. designed and supervised research, analyzed data, and wrote the manuscript. L.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.

Prior Presentation. Parts of this study were presented in abstract form at the 2nd Joint Meeting of the European Association for the Study of Diabetes (EASD) Islet Study Group and Beta Cell Workshop, Dresden, Germany, 7–11 May 2017, and at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.