Mice used in the study were housed in strict specific pathogen-free (SPF) facilities with a 12-hour-dark/light cycle in the Yale Animal Resource Center (YARC). Pdgfra ^fl/fl C57BL/6 breeders were kindly provided by Valerie Horsley (Yale University) and Pdx1-cre⁺ C57BL/6 breeders were kindly provided by Qingchun Tong (UT Health). We bred the two mouse lines and generated Pdgfra^fl/flPdx1-cre⁺ C57BL/6 and Pdgfra^fl/flPdx1-cre^- C57BL/6 littermates for the study. The mice were fed with either autoclaved normal diet (Teklab Global, USA, 6.2% fat) or high fat diet (HFD, Research Diet, 60% fat, New Brunswick, NJ, USA) ad libitum. The use of the animals in this study was approved by the Institutional Animal Care and Use Committee of Yale University.

NIT-1 (ATCC CRL-2055), a mouse β-cell line, was purchased from ATCC (Manassas, VA, USA.). NIT-1 cells have well developed rough endoplasmic reticulum and β-granules with similar ultrastructural features to differentiated mouse β-cells (43). They were cultured in Ham’s F12K medium with 2 mM L-glutamine, 1.5 g/L sodium bicarbonate (ATCC), and 10% heat-inactivated fetal calf serum (FCS (Gemini)).

Most of the fluorochrome-conjugated monoclonal antibodies used in this study were purchased from BioLegend unless otherwise stated. The supernatants of different monoclonal antibody (mAb) hybridomas were provided by the late Charles Janeway (Yale University). RPMI-1640 medium and heat-inactivated FCS were purchased from Invitrogen and Gemini respectively. Anti-H2-K^d mAb, was affinity purified from hybridoma (clone: HB159) supernatant. Anti-Qa-2 mAb was purchased from BioLegend (clone: 659H1-9-9). The lysis buffer and other reagents for Western blot (WB) were purchased from ThermoFisher and Bio-Rad. Rabbit anti-mouse antibodies for WB were purchased from ThermoFisher and Cell Signaling.

Mice were fasted overnight with free access to water and intra-peritoneal (i.p.) glucose tolerance tests (IPGTTs) were performed by i.p. injection of glucose (1 g/kg to the mice fed with high fat diet and 2 g/kg to the mice fed with normal diet). The blood glucose was measured with a FreeStyle glucose meter (Abbott) before (time zero) and at different time points after glucose injection.

Mice were fasted for 4 h with free access to water and insulin tolerance tests (ITTs) were performed by i.p. injection of insulin (Humulin-R, 0.75 U/kg; Eli Lilly, Indianapolis, IN, USA). The blood glucose was measured with a FreeStyle glucose meter (Abbott) before (time zero) and at different time points after insulin injection.

Pancreatic islets were isolated as described (44). Briefly, mice were sacrificed by cervical dislocation before dissection. The pancreas was inflated through the bile duct with a 30G needle starting at the gall bladder with 3 ml cold collagenase (Sigma; St Louis, MO, USA) solution (0.3 mg/ml). The pancreas was dissected, weighed and placed into a siliconized glass vial containing 1 ml of 1 mg/ml collagenase solution. The sealed vials were incubated in a 37°C water bath for 12–14 min with vibration. After three washes of the digested pancreas, islets were handpicked and counted under a dissecting microscope for further experiments.

The insulin content measurement was performed as previously described (45). Briefly, the middle part of the pancreatic body was dissected, weighed and placed into an Eppendorf tube with 0.5ml ice-cold acid-ethanol buffer (1.5% concentrated HCL, 75% Ethanol and 23.5% deionized water) and homogenized by an Ultrasonic Homogenizer (VWR Scientific, Radnor, PA, USA) for 2 min with 20 pulses. After further incubating the mixture at 4°C overnight, the homogenized tissue was centrifuged at 1400g for 20 min at 4°C. Insulin content in the supernatant was measured using an insulin RIA kit (EMD-Millipore, Burlington, ME, USA).

The insulin release assay was performed as previously described (21). Briefly, isolated pancreatic islets were mixed and equally distributed to test tubes (30–60 islets/tube depending on the total islet numbers) after stabilizing in low-glucose KRB buffer for 2 hours. The islets were then stimulated with KRB containing high glucose (25 mmol/l) and the supernatant fractions were harvested every 5 min after glucose stimulation. Secreted insulin in the supernatant fractions was measured using the insulin RIA kit (EMD-Millipore, Burlington, ME, USA).

Ex vivo pancreases from 18-22-week-old male mice were fixed in periodate–lysine–paraformaldehyde followed by freezing in Tissue-Tek OCT (Bayer, Elkhart, IN, USA). Each pancreas was cut in its entirety into more than two hundred 10 μm thick sections and one section was selected at a 9-section interval for staining with hematoxylin to ensure different islets were imaged. Islet mass was measured using Image J software (NIH, Bethesda, MD, USA) after photographing under the light microscope.

RNA from pancreatic islets isolated as described above was extracted with RNeasy Mini Kit (Qiagen, Hilden, Germany) and quantified by NanoDrop (ThermoFisher). Equal amounts of RNA were reverse transcribed using iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Quantitative PCR (qPCR) was performed using the Bio-Rad iQ5 qPCR detection system (Hercules, CA, USA) with the specific primers ( Supplementary Table S2 ). The level of gene expression was determined with the 2−ΔΔCt method by normalization with the reference gene gapdh.

The isolated islets were treated with Cell Dissociation Solution (Sigma) to obtain the single-cell suspension. Cells were stained with fluorochrome-conjugated monoclonal antibodies to CD45 (for immune cells, BioLegend; San Diego, CA, USA), CD140a (for all cells, BioLegend) and FluoZin-3-acetoxymethyl (for β-cells, ThermoFisher, Waltham, ME, USA) (46). The β-cell survival status was assessed with APC Annexin V Apoptosis Detection Kit with 7-AAD (BioLegend) following the manufacturer’s protocol. Cells were analyzed on a BD LSRII flow cytometer (LSRII; BD Bioscience, San Diego, CA, USA) and results were analyzed with FlowJo 10.4 software.

RNA from pancreatic islets isolated as described above was extracted with RNeasy Mini Kit (Qiagen, Hilden, Germany) and quantified by NanoDrop (ThermoFisher). RNA-sequencing (poly A) was performed at the Yale Center for Genome Analysis using NovaSeq with HiSeq paired-end, 100bp. The raw TRAPseq fastq files were processed using the fastp tool (version 0.20.0) (47). With a default setting, sequencing reads with low-quality bases were trimmed or filtered. Alignment was performed for cleaned reads using STAR (version 2.7.9) (48) and mouse reference genome (gencode version GRCm38.p6 with vM25 gene annotation). Expression quantification was performed for alignment results using featureCounts (version 2.0.0) (49). As genes with low expression levels that represent noise were to be excluded before downstream analysis, we defined low expression filtering as expressed genes with ≥ 6 read counts in at least 20% of samples. The filtered read counts matrix was then normalized by transcripts per million (TPM) method. Detection of differentially-expressed genes was performed using R package DESeq2 (version 1.30.1) (50). The Benjamini-Hochberg procedure was used for multiple test correction, and FDR ≤ 0.05 as a threshold for detection of differentially expressed genes.

Western blot (WB) analysis was performed as previously described (51). The isolated islets (100) were homogenized in 150μl lysis buffer RIPA (ThermoFisher Scientific, Pittsburgh, PA) with phosphatase and EDTA inhibitor (Roche, Mannheim, Germany). Total protein concentrations were quantified using the BCA assay (Pierce, Rockford, IL). Ten micrograms of total protein were dissolved in 2× loading buffer (Bio-Rad) and separated on a MINI-PROTEAN TGX Stain-Free Gels (4-20%, 10 well, 30μl) (Bio-Rad, Hercules, CA, USA). The proteins were then transferred to polyvinylidene fluoride membranes (Bio-Rad). Membranes were blocked with 5% nonfat milk (ThermoFisher Scientific, Pittsburgh, PA) and incubated overnight with primary antibodies ( Supplementary Table S3 ) in SDS buffer (Bio-Rad). Membranes were then washed and incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies (BioLegend) at room temperature for 2 h. After washing, membranes were incubated with West Pico Plus (ThermoFisher Scientific, Pittsburgh, PA) to visualize the proteins.

Statistical analysis was performed using GraphPad Prism software version 9.0 (GraphPad Software, San Diego, CA, US). Islet mass was analyzed using a Kolmogorov-Smirnov test. Data from other assays were analyzed with either a two-tailed Student’s t test (if data normally distributed), a two-tailed Mann-Whitney test (if data not normally distributed), multiple t tests with FDR correction, or a two-way ANOVA. P < 0.05 was considered to be significant.

Pdgfra plays a role in age-dependent β-cell proliferation and expansion (16), but how Pdgfrα affects islet β-cell function and glucose metabolism in obesity and type 2 diabetes was previously unknown. To study this, we generated Pdgfra^fl/fl Pdx1-Cre⁺ C57BL/6 mice (KO mice) and Pdgfra^fl/fl Pdx1-Cre⁻ C57BL/6 littermates (WT mice). We confirmed that Pdgfrα protein (CD140a) was reduced in islet β-cells by flow cytometry ( Supplementary Figure S1A ). We also examined the expression of CD140a at different mouse ages - young (3–4 wks) and adults (10–12 wks). As expected, KO mice have lower CD140a expression ( Supplementary Figure S1B ); however, it is interesting that the difference between KO and control mice was much greater in young mice, and the proportion of CD140a-expressing islet β-cells is also higher (over 10%, see Y axis) and less than 1% (see Y axis) in adult KO and control mice respectively ( Supplementary Figure S1B ). We first monitored the body weight (BW) of the KO and WT male mice fed with a normal diet ( Supplementary Figure S2 ) and observed that the KO mice gained more body weight ( Figures 1A,B ) compared to the control mice on normal diet, although there was no difference in food intake ( Supplementary Figure S3 ). Next, we assessed glucose metabolism in KO mice and control mice, fed with normal diet by performing IPGTT at different ages (6–7 and 14 weeks), and KO mice in both age groups exhibited impaired glucose tolerance ( Figures 1C,D ). To further evaluate the secreted/synthesized insulin by β-cells in pancreatic islets, we performed insulin release assays and found that islets from KO mice secreted less insulin compared to the control mice ( Figure 1E ) and that the pancreata from KO mice contained less insulin ( Figure 1F ). In addition, both age groups of KO mice had higher fasting blood glucose compared to the age-matched control mice ( Figures 1G,H ). Interestingly, although KO mice secreted less insulin, we found that there was a trend to higher fasting blood insulin levels in KO mice, but these differences between KO and control mice were not statistically significant at any age tested ( Figures 1I,J ).

To assess insulin resistance, we performed insulin tolerance tests (ITT) and found that there was no significant difference between adolescent KO and control mice ( Figure 1K ), but the adult KO mice showed pronounced insulin resistance ( Figure 1L ). This suggests that although decreased insulin secretion was present from an early age, the insulin resistance was age-related in our model system. In line with the body weight, KO mice fed with normal diet had more abdominal fat including epididymal fat, ( Figure 1M ), with the ratio of white adipose tissue (WAT), calculated using epididymal fat to body weight, also significantly higher in KO mice, compared to the control mice ( Figure 1N ). Moreover, we found more fat in the liver of KO mice fed with normal diet than WT mice ( Figure 1O ), and there was a higher ratio of liver weight to body weight in the KO mice than the control mice ( Figure 1P ).

To assess if Pdgfra deficiency in pancreatic islet β-cells affects islet architecture and the composition of islet cells, especially islet α cells that also regulate blood glucose, we performed pancreatic histology and histochemical staining of KO and control mice. We found that the size of the islets from the KO mice were generally smaller than in control mice but there was no alteration of islet architecture ( Supplementary Figure S4A ). Supporting our insulin release data ( Figure 1E ), the staining of insulin in the islets of KO mice was weaker than in control islets ( Supplementary Figure S4B ). Interestingly, it appears that there were more glucagon-positive α cells in the islets of KO mice, compared to the control mice ( Supplementary Figure S4C ); however, as the islets from the control mice were generally larger than those from the KO mice, it is conceivable that the “absolute” number of islet α cells was comparable between the KO and control mice. Thus, our results demonstrate that Pdgfra deficiency in pancreatic islet β-cells results in significantly impaired glucose metabolism, increased body weight, body fat and fatty liver in C57BL/6 mice fed with normal chow.

To identify the impact of Pdx-Cre-driven Pdgfra deficiency on glucose metabolism in a metabolically stressed condition, we fed 6-week-old KO and control mice with high fat diet (HFD) for 8 weeks ( Supplementary Figure S5 ). The body weight gain in the KO mice was greater compared to the control mice ( Figures 2A, B ) and the KO mice had much higher fasting blood glucose and significantly worse glucose tolerance, assessed by IPGTT, compared to control mice ( Figures 2C, D ). Fasting blood insulin levels were also significantly higher in KO mice than in the control mice ( Figure 2E ). In line with the KO mice on normal diet ( Figure 1L ), but more pronounced, the KO mice on HFD had higher insulin resistance, assessed by ITT, compared to the control mice ( Figure 2F ). Given the fact that KO mice had decreased insulin synthesis and/or secretion ( Figures 1E, F ), this suggested that insulin resistance was, in fact, even more severe. Furthermore, we found that the ratio of white adipose tissue (WAT) to body weight and the ratio of liver to body weight in KO mice were significantly higher, compared to the control mice ( Figures 2G, H ). Taken together, our results from high fat diet-fed mice agreed with those from normal chow-fed mice, indicating that high-fat diet exacerbated the metabolic dysfunction.

As the KO mice without metabolic stress showed impaired metabolism and insulin resistance, to further verify the impaired beta cell function and insulin resistance, we focused on studies with normal diet. We calculated index of Homeostasis Model Assessment 2 (HOMA2-%S) and HOMA IR and found decreased insulin sensitivity but increased insulin resistance in KO mice fed with normal diet, compared to the control mice ( Figures 3A, B ). Skeletal muscles make up the most important organ for whole-body glucose homeostasis, and insulin resistance in skeletal muscle is an important feature of obesity and T2D (22). We assessed inflammation by qPCR, in skeletal muscle from KO mice fed with normal diet, compared to the control mice. Supporting the observation of insulin resistance in KO mice, we found higher expression of inflammatory cytokines, including TNFα, IFNγ, IL-1β, IL-6 and IL-22, in the muscles of KO mice fed with normal diet, compared to the control mice ( Figures 3C–G ). Furthermore, in keeping with the higher levels of inflammatory cytokines tested, higher expression of inflammatory chemokines such as CCL2, CCL5 and SLAFM7 were also found in the muscles of KO mice, compared to the controls ( Figures 3H–J ). Our results indicated that Pdgfra deficiency in pancreatic β-cells not only results in impaired β-cell function and glucose metabolism but also leads to heightened inflammation in muscle tissue, which likely contributes to the insulin resistance and body weight gain.

Next, we investigated whether Pdx-Cre-driven Pdgfra deficiency affected islet development and function in C57BL/6 mice. We first evaluated islet mass by measuring the islet area under light microscopy on pancreatic tissue sections of the entire pancreas from KO and control mice. We found that islet mass in KO mice was lower than in control mice using Image J analysis ( Figure 4A ). Kolmogorov-Smirnov test analysis indicated that there were significantly fewer islets and reduced islet area in the pancreata of KO mice compared with controls ( Figure 4B ). We also measured the net islet number in the same pancreatic weight of KO and control mice and analyzed the number of islets per milligram pancreas weight. We found fewer islets per milligram of pancreas in the KO mice, compared with the control mice, both of which were fed with normal diet and the experiments were performed in parallel ( Figure 4C ). The low islet number led us to hypothesize that β-cells in KO mice could be prone to apoptosis. To test our hypothesis, we performed Annexin V staining using ex vivo freshly prepared dispersed islet cells. Islet cells were co-stained with a β cell marker FluoZin and the immune cell marker CD45, in addition to Annexin V staining, prior to analyzing by flow cytometry, with β-cells gated as CD45 negative and FluoZin positive (CD45-FluoZin+). We found that there were more live β-cells (Annexin5-7AAD-) in control mice than in KO mice ( Figures 4D, E ), and conversely, more apoptotic β-cells (Annexin5 + 7AAD-) in KO mice than in control mice ( Figures 4D, F ). Thus, our data indicated that Pdgfrα deficiency in pancreatic β-cells enhanced β-cell apoptosis, which may contribute to the reduced islet number and function observed in the KO mice.

Next, we assessed pro-apoptotic (Fas and Caspase9) and anti-apoptotic (Bcl2) gene expression in the islets. To our surprise, there were no significant differences in the pro-apoptotic gene expression, regardless of the diet ( Figures 4G–J ). In contrast, the expression of the anti-apoptotic gene Bcl2 was significantly reduced in KO mice compared to the control mice, and this was diet independent ( Figures 4K, L ). This suggested that Pdgfra deficiency in islet β-cells did not increase pro-apoptotic signaling in β-cells but rather affected the ability of β-cells to survive.

Next, we evaluated islet β-cell proliferation, staining dispersed islet cells with Ki67, co-staining with anti-mouse CD45 and FluoZin and analyzing by flow cytometry. After gating islet β-cells (CD45-FluoZin+), we analyzed the proportion of Ki67 positive (Ki67+) cells. It is interesting that islets from KO mice had many fewer proliferating Ki67+ β-cells compared with control mice ( Supplementary Figure S6 ).

To further investigate the molecular mechanism by which Pdgfra alters β-cell survival, we studied the transcriptome of the islets from KO and control mice by total RNA sequencing. The principal component analysis showed that the two groups of samples were clearly separated, indicating that the transcriptomes of the two groups of samples were very different ( Supplementary Figure S7A ). Transcriptome analysis showed that there were 45 genes expressed significantly differently between the two groups, of which 17 were upregulated, and 28 were downregulated in the islets from KO mice compared to the control mice ( Figures 5A, B ). We found that atf5 (activating transcription factor 5) was the most downregulated gene while gadd45b (Growth Arrest and DNA Damage Inducible Beta), Foxi3 (Forkhead box i3) and Naps4 (neuronal PAS domain protein 4) were among the highest upregulated genes in the islets from KO mice ( Figures 5A, B ). The down-regulation of atf5 and up-regulation of gadd45b were further confirmed by qPCR of the islets of mice fed with normal diet or HFD ( Figures 5C–F ; Figures S7B, C) . Interestingly, both atf5 and gadd45b genes are associated with apoptosis caused by cellular stress. Atf5 is a cellular pro-survival transcription factor (23, 24) whereas Gadd45b is associated with apoptosis, often through Fas-mediated apoptosis (25). The gene enrichment analysis revealed that the most significantly enriched functional gene set was the tRNA aminoacylation for protein translation, which is also associated with ER stress ( Figure 5G ; Supplementary Table S1 ). Next, we assessed the levels of protein expression of Atf5 and Gadd45b in the islets by Western blot. Supporting the gene expression profile by RNA-seq and qPCR, Atf5 protein expression was significantly lower ( Figure 5H ), whereas Gadd45b protein expression was significantly higher ( Figure 5I ) in the islets of KO mice compared with the control mice. Atf5 expression is associated with PI3K (26, 27), which is an important intracellular kinase regulating many cell functions especially cell growth and survival (28). Interestingly, it is also a target of Pdgfrα (29, 30). Thus, we investigated the protein expression of PI3K by Western blot. Our results showed that the protein expression of PI3K regulatory subunit p85 and the phosphorylated P-PI3K p85 in the islets of KO mice were both significantly reduced compared to the islets from control mice ( Figures 5J, K ). Taken together, we found two altered genes, Atf5 and Gadd45b, which have not been reported before, contributing to β-cell apoptosis and survival in association with Pdgfrα through the PI3K pathway.

To verify that our findings using islets, in vivo and/or ex vivo, were indeed changes in β-cells, we tested the β-cell line, NIT-1 cells. We cultured NIT-1 cells, after the cells became confluent, in the presence of different concentrations of CP673451, a potent Pdgfr inhibitor (31), and solvent for 12 hours. The supernatants and NIT-1 cells were collected for insulin content and the NIT-1 cells were subject to Annexin V assay. Some NIT-1 cells were also prepared for RNA extraction ( Supplementary Figure S8 ). In line with our in vivo findings in KO mice ( Figures 1E, F ), inhibition of Pdgfrα in NIT-1 β-cells resulted in reduced insulin secretion ( Figure 6A ). The reduction did not appear to be inhibitor dose dependent ( Figure 6A ). Also supporting our ex vivo finding in KO mice ( Figures 4D–F ), inhibiting Pdgfrα in NIT-1 β-cells led to a dose-dependent decrease in live cells (negative for Annexin 5 & 7AAD, Figures 6B, C ) but increased apoptotic cells ( Figures 6B, D ). At the molecular level, we found that inhibition of Pdgfrα in NIT-1 β-cells led to decreased expression of atf5 ( Figure 6E ), but increased gadd45b expression ( Figure 6F ), as well as reduced expression of pi3k ( Figure 6G ). Thus, the results from our in vitro investigation using NIT-1 β-cells with the Pdgfrα inhibitor validated our findings in vivo using mice with a β-cell targeted deletion of Pdgfrα. Taken together, using different experimental systems, we have revealed that Atf5 and Gadd45b are most likely the main players contributing to the impaired glucose metabolism seen in the KO mice.