Pancreatic pericytes of healthy humans and mice express cytokines. We hypothesized that islet inflammation, which is fundamental for β cell function and glucose regulation, is regulated by pericytes (Figure 1A). In many tissues, pericytes produce inflammatory mediators, including cytokines (32, 34–37). To determine whether this finding is valid for the pancreas, we used publicly available transcriptome analyses of human and mouse pancreatic cells (41, 48). To define cytokine expression by pericytes in human islets, we utilized single-cell RNA-seq (scRNA-seq) analysis of islets isolated from donors without diabetes or detectable autoantibodies (48). Human islet pericytes (identified by the expression of the pericytic markers PDGFRB, RGS5, and ACTA2, but not the stellate cell marker GFAP, originally annotated as “quiescent stellate” and “activated stellate”; Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/JCI179335DS1) (46, 48) express cytokine-encoding genes, including CCL2, CCL11, CXCL1, CXCL12, CXCL13, and IL6, at higher levels than macrophages and β cells (Figure 1B). Next, we used our previously published RNA-seq analysis of pancreatic pericytes from healthy young mice (41). As shown in Figure 1C, mouse pancreatic pericytes expressed an array of genes encoding cytokines, some of which were also expressed by human islet pericytes (Ccl2, Ccl11, Cxcl1, Cxcl12, and Il6; Figure 1, B and C). Although some of these cytokines have been implicated in glucose regulation (14, 15, 49–54), their pancreatic sources remain largely unknown.

Pancreatic pericytes express cytokines in a TLR4/MyD88-dependent manner. (A) A graphical model of the study’s hypothesis, with pericytes highlighted in green. (B) Heatmap showing the relative expression of selected cytokines in macrophages, β cells, and pericytes (originally annotated as “quiescent stellate [qSC]” and “activated stellate [aSC]”), employing published scRNA-seq analysis of islets from healthy human donors (48). (C) Heatmap showing the relative expression of selected cytokines in isolated pericytes and islets, employing a previously published RNA-seq analysis of mouse pancreata (41). (D and E) Bar diagram (mean ± SD) showing the results of qPCR analysis of Il1r1 (D) and Tlr4 (E) transcripts in bulk mouse pancreatic tissues, isolated islets (average set to 1), pancreatic endothelial cells (ECs; PECAM1+), pancreatic immune cells (CD45+), and pancreatic pericytes (purified based on YFP expression from Nkx3-2-Cre;R26-YFP mice). n = 3–6. *P < 0.05, ***P < 0.005 (unpaired, 2-tailed Student’s t test). Each dot represents a single sample. (F) Immunofluorescence analysis of adult mouse pancreatic tissue sections for TLR4 (green), the pericytic marker NG2 (red), insulin (white), and DAPI (blue). The right panel shows a higher magnification of the area framed in the white box on the left panel. Scale bars: 25 μm. (G) Cultured neonatal pancreatic pericytes were either treated with LPS (right) or left untreated (left; the average was set to 1) and harvested after 24 hours. Bar diagrams (mean ± SD) showing the expression levels of genes encoding selected cytokines analyzed by qPCR. n = 5. One representative of 2 independent experiments. ***P < 0.005 (unpaired, 2-tailed Student’s t test) compared with the untreated group. Each dot represents a single sample. (H) Bar diagram (mean ± SD) showing the results of qPCR analysis of Myd88 transcripts in pancreatic pericytes purified from Nkx3-2-Cre;R26-YFP (gray) or Nkx3-2-Cre;Myd88fl/fl;R26-YFP (ΔMyD88Peri; red) mice based on YFP expression. n = 4. ***P < 0.005 (unpaired, 2-tailed Student’s t test) compared with nontransgenic mice. Each dot represents a single sample. (I) RNA-seq analysis of pancreatic pericytes from control YFPPeri (Nkx3-2-Cre;R26-YFP) and YFPΔMyD88Peri (Nkx3-2-Cre;Myd88fl/fl;R26-YFP) 15-week-old mice. Volcano plot analysis showing genes upregulated (orange) and downregulated (green) in YFPΔMyD88Peri pericytes. Selected genes are annotated. n = 3.

The TLR/IL-1R pathway is active in pancreatic pericytes. To study the role of pericytic cytokines, we interfered with their production. One of the main pathways regulating cytokine expression is the TLR/IL-1R pathway. Mouse pancreatic pericytes express 2 of the main receptors in this pathway, Tlr4 and Il1r1 (Supplemental Figure 1). A comparison of the expression of these 2 receptors in various pancreatic cell types revealed that pericytes (purified based on their fluorescent labeling in Nkx3-2-Cre;R26-YFP mice) (41, 42, 55) expressed lower levels of Il1r1 than did endocrine cells, while they expressed Tlr4 at significantly higher levels than did the other pancreatic cell populations (Figure 1, D and E). To verify the presence of pericytic TLR4 at the protein level and independently of Nkx3-2/YFP expression, we performed immunofluorescence analysis of a wild-type pancreas, which indicated the presence of this receptor in islet-associated pericytes (Figure 1F). Finally, to determine whether the TLR/IL-1R pathway is active in pancreatic pericytes, we exposed these cells to the TLR4 agonist LPS in vitro. As shown in Figure 1G, LPS increased the expression of known target genes in cultured pancreatic pericytes. Thus, our analysis revealed an active TLR/IL-1R pathway in pancreatic pericytes, similar to pericytes in other tissues (34, 37, 56).

Pericytes express cytokines in a MyD88-dependent manner. Next, we interfered with the pericytic inflammatory response by selectively inhibiting MyD88, the canonical adaptor of the TLR/IL-1R pathway. Pancreatic pericytes expressed significantly higher levels of Myd88 than did the other analyzed pancreatic cell populations (Supplemental Figure 1). To delete this adaptor, Myd88fl mice (57) were crossed with the Nkx3-2-Cre line (58), which selectively targets mural cells, primarily pericytes, but not other pancreatic cell types, including epithelial, endothelial, and immune cells (Supplemental Figure 1) (41, 42, 44, 55), to generate ΔMyD88Peri (Nkx3-2-Cre;Myd88fl/fl) mice. Notably, all the mice used in this study were fed regular chow, housed in a specific pathogen–free barrier facility, and did not display any indication of infection. As expected, while Myd88 expression was lost in the pancreatic pericytes of ΔMyD88Peri mice, its expression was preserved in pancreatic endocrine and immune cells (Figure 1H and Supplemental Figure 1). To determine the requirement of MyD88 for pericytic cytokine production in vivo, we purified pancreatic pericytes from ΔMyD88Peri and control mice based on fluorescence labeling and profiled their transcriptome. RNA-seq analysis revealed that the loss of MyD88 resulted in the upregulation of 6 genes and the downregulation of 55 genes in pancreatic pericytes, including Myd88 itself and the cytokines Cxcl1, Cxcl13, Cxcl12, and Il6 (Figure 1I and Supplemental Table 1). Thus, the canonical TLR/IL-1R pathway is required for cytokine production by pancreatic pericytes. Furthermore, our analyses point to the basal activity of the TLR/IL-1R pathway in healthy pancreatic pericytes, which is required for proper cytokine production.

Pericytic MyD88 activity regulates the number of islet immune cells. To test our hypothesis further, we aimed to determine the role of pericytes in the regulation of islet inflammation (Figure 2A). To this end, we interfered with the pericytic inflammatory response by selectively inhibiting MyD88 in these cells and analyzed the resulting effects on islet immune cells. Islets isolated from ΔMyD88Peri (Nkx3-2-Cre;Myd88fl/fl) mice had fewer immune cells than did those isolated from littermate controls (Myd88fl/fl; Figure 2, B and C). Next, we quantified the main immune cell populations in healthy islets using flow cytometry (Figure 2D and Supplemental Figure 2). ΔMyD88Peri islets had significantly fewer macrophages, B cells, and T cells. Notably, the B cell number was one-fifth of that of the control (Figure 2D). In contrast, the sizes of the corresponding immune cell populations in the blood and spleen were comparable between transgenic and nontransgenic mice (Supplemental Figure 2).

Loss of pericytic MyD88 interferes with the number and phenotype of islet immune cells. Fifteen-week-old ΔMyD88Peri (red) and nontransgenic littermates (Cre-negative; ‘‘non tg’’; gray) male mice were analyzed. (A) A graphical model of the study’s hypothesis, with immune cells highlighted in green. (B) Representative dot plot showing immune cells (CD45+ cells) among the dispersed islet cells. (C and D) Bar diagrams (mean ± SD) showing the total number of immune cells (CD45+; C), macrophages (MФ; CD45+CD11c+CD64+; D), B cells (CD45+CD19+; D), T cells (CD45+CD3+; D), and DCs (CD45+CD11c+CD64–; D) in 100 isolated islets. *P < 0.05, **P < 0.001, ***P < 0.005 (unpaired, 2-tailed Student’s t test) compared with nontransgenic samples. Each dot represents a single sample. n = 4–12. (E, F, H, and I) scRNA-seq analysis of islet immune cells from ΔMyD88Peri and nontransgenic mice. Shown are UMAP visualization and Seurat clusters with cell annotation (E), a heatmap representative of selected macrophage marker expression (F), the relative portion of the different macrophage clusters (H), and a heatmap representative of selected differential gene expression in cells of clusters 0 and 2 (I). (G) Immunofluorescence analysis of nontransgenic mouse pancreatic tissue sections for the macrophage marker Iba1 (green), the pericytic marker NG2 (red), insulin (white), and DAPI (blue). The right panel shows a higher magnification of the area framed in the white box on the left panel. Scale bars: 25 μm.

Abnormal composition and phenotype of immune cells in ΔMyD88Peri islets. To further define the potential differences between islet immune cells from ΔMyD88Peri and those from nontransgenic mice, we performed scRNA-seq analysis of these cells. Macrophages (clusters 0, 2, 7, and 10), DCs (cluster 6), B cells (clusters 1 and 8), T cells (cluster 3), and a small population of type 2 innate lymphoid cells (ILC2s; cluster 11) were detected in the islets (Figure 2E and Supplemental Table 2). The presence of the 4 islet macrophage populations was consistent with previous reports on the heterogeneity of these cells (22, 59, 60). Based on their transcriptome, the macrophages in cluster 10 resembled M2 macrophages (expressing Cd163, Mrc1 [CD206], Lyve1, Clec10a [CD301], Timd4, and Folr2), whereas the cells in group 7 did not express any of the known macrophage subtype markers (Figure 2F and Supplemental Table 2). Cells in clusters 0 and 2 represent classical M1-like macrophages (expressing Lyz2, Csf1r, Cd14, Fcgr1 [CD64], Itgax [CD11c], Aif1 [Iba1], Cx3cr1, and Ccr2; Figure 2F) (22, 60). M2-like macrophages (cluster 10) expressed Igf1, and classical M1-like macrophages (clusters 0 and 2) and DCs (cluster 6) expressed Il1b (CD11c+ cells; Supplemental Figure 2). Thus, islet macrophage subpopulations also differ in the expression of secreted factors shown to affect β cells (8, 26). Classical macrophages (identified as Iba1+ cells), which encompass most islet macrophages (22, 60), are located within the islets in proximity to both β cells and pericytes (Figure 2, G and H).

As changes in macrophage composition are associated with destructive islet inflammation (8), we compared the relative proportions of the 4 macrophage populations in ΔMyD88Peri and the control islets. Most of the macrophages in the nontransgenic islets were grouped into cluster 0 when their relative proportions were lower in the transgenic islets (72% vs. 50%; Figure 2H). In contrast, the proportion of macrophages in cluster 2 was higher in ΔMyD88Peri islets (38%) than in control (10%; Figure 2H). As detailed above, macrophages belonging to these 2 clusters expressed similar markers of classical macrophages; however, cells in cluster 0 expressed an array of genes not expressed by cells in cluster 2 (Figure 2I and Supplemental Table 2). Among these differentially expressed genes are Tgfbr1 and Zeb2, which are associated with tissue-specific macrophage differentiation (61, 62). Thus, the cells in clusters 0 and 2 may represent macrophages at different differentiation stages when ΔMyD88Peri islets are potentially populated with a higher proportion of immature macrophages.

Pericytic MyD88 is required for β cell function and glucose regulation. To test our hypothesis further, we examined whether interfering with pericyte-regulated islet inflammation affects β cell function and glucose homeostasis (Figure 3A). ΔMyD88Peri and littermate control adult mice displayed comparable weight and basal glucose levels (Supplemental Figure 3). Previous studies have demonstrated that Nkx3-2-Cre expression alone does not affect the response to glucose, and is mainly restricted to cells in the gastrointestinal and skeletal systems, with no targeting of hepatic pericytes (41, 58, 63). As expected, ΔMyD88Peri and control mice displayed comparable insulin sensitivity (Figure 3B). However, ΔMyD88Peri mice (both females and males) exhibited an impaired response to glucose challenge (Figure 3C and Supplemental Figure 3). Thus, our analysis indicated that pancreatic pericytic MyD88 is required for glucose regulation during homeostasis.

Loss of pericytic MyD88 causes β cell dedifferentiation and glucose intolerance. Fifteen-week-old ΔMyD88Peri (red) and nontransgenic littermates (Cre-negative; ‘‘non tg’’; gray) male mice were analyzed. (A) Graphical model of the study hypothesis, with β cells highlighted in green. (B) Intraperitoneal insulin tolerance test (ITT). The mean (± SEM) blood glucose levels are presented. n = 8–10. (C) Intraperitoneal glucose tolerance test (IPGTT). Shown are mean (± SEM) blood glucose levels (left) and area under the curve (AUC, right). n = 9–12. (D) Bar diagram (mean ± SD) showing the glucose-stimulated insulin secretion (GSIS) of isolated islets. n = 4–5. (E) Bar diagram (mean ± SD) showing the insulin content of the isolated islets. n = 4–5. (F, H, and I) Bar diagrams (mean ± SD) showing β cell gene expression analyzed by qPCR. The average levels in the control islets were set to 1. n = 5–9. (G) Bar diagrams (mean ± SD) showing comparable β cell mass in transgenic and control mice. n = 4. *P < 0.05, **P < 0.01, ***P < 0.005; NS, not significant (unpaired, 2-tailed Student’s t test) compared to control.

Next, we aimed to determine the underlying causes of the impaired glucose response in mice lacking pericytic MyD88. Vascular coverage and pericyte density were comparable between the transgenic and control islets (Supplemental Figure 4). Nevertheless, to study β cell function independently of blood flow, we measured glucose-stimulated insulin secretion (GSIS) in islets isolated from ΔMyD88Peri and control mice. While basal insulin secretion was intact, islets lacking pericytic MyD88 secreted less insulin in response to glucose challenge (Figure 3D). This impaired insulin secretion correlated with decreased insulin levels in ΔMyD88Peri islets (Figure 3E) and lower expression of the insulin-encoding genes Ins1 and Ins2 (Figure 3F). Thus, the lack of pericytic MyD88 affects insulin production and subsequent secretion.

To further define how β cells are affected in ΔMyD88Peri mice, we analyzed their phenotype. The β cell mass was unaffected in ΔMyD88Peri mice (Figure 3G). In agreement with these findings, transgenic islets did not upregulate the β cell stress genes Atf4 and Ddit3 (Chop) (Supplemental Figure 4). Furthermore, islet cytoarchitecture and β-to-α cell ratios were similar in transgenic and control mice, as were the Gcg and Sst expression levels (Supplemental Figure 4). However, ΔMyD88Peri islets expressed significantly lower levels of genes encoding components of the GSIS machinery, including Slc2a2 (Glut2), Sur1, and Kcnj11 (Kir6.2) (Figure 3H). The expression levels of MafA, NeuroD1, Pdx1, and Ucn3, which are all required for the mature functional β cell phenotype, were lower in the ΔMyD88Peri islets (Figure 3I). Thus, our analysis points to β cell dedifferentiation in the absence of pericytic MyD88 activity.

Next, we investigated whether pericytic MyD88 deficiency influences postnatal β cell development. During the postnatal period, transgenic and control pups exhibited comparable weight gain and pancreatic mass (Supplemental Figure 5). Furthermore, islet morphology and pancreatic insulin content were similar between ΔMyD88Peri and littermate control pups (Supplemental Figure 5). Notably, while adult transgenic mice were glucose intolerant (Figure 3C), the glucose responsiveness of pre-adult ΔMyD88Peri mice (6 and 10 weeks of age) was comparable to control (Supplemental Figure 5). These findings collectively indicate that the loss of pericytic MyD88 does not impede β cell development and that ΔMyD88Peri mice develop glucose intolerance in adulthood.

In conclusion, pericytic MyD88 is required for glucose regulation by supporting the mature β cell phenotype and insulin production during adulthood, potentially by regulating islet inflammation.

Cxcl1 rescued the β cell phenotype and glucose response of ΔMyD88Peri mice. Our analysis indicated that pericytes regulate islet inflammation and glucose homeostasis. Therefore, we aimed to elucidate the potential underlying molecular mechanism by focusing on a single pericytic cytokine (Figure 4A). A potential candidate is CXCL1, which is expressed in both human and mouse pericytes (Figure 1, B and C). Pancreatic pericytes secrete Cxcl1 (Figure 4B) and are the primary source of this cytokine in the pancreas (Figure 4C). Importantly, our analysis indicated that pericytic Cxcl1 levels depended on the TLR4/MyD88 pathway (Figure 1, G and I, and Figure 4, B and D). Thus, we hypothesized that the loss of pericytic Cxcl1 contributes to MyD88-dependent abrogation of islet inflammation and β cell function.

Cxcl1 treatment rescues the glucose intolerance in ΔMyD88Peri mice. (A) A graphical model of the study’s hypothesis, with Cxcl1 highlighted in green. (B) Cultured neonatal pancreatic pericytes were either treated with LPS (right) or left untreated (left), and their supernatant was collected after 48 hours. Bar diagrams (mean ± SD) showing Cxcl1 protein concentration in the supernatant. n = 5. ***P < 0.005 (unpaired, 2-tailed Student’s t test) compared with the untreated group. Each dot represents a single sample. (C) Bar diagram (mean ± SD) showing the results of qPCR analysis of Cxcl1 transcripts in the indicated pancreatic cell types (as detailed in Figure 1, the average levels in islets were set to 1). n = 3–6. ***P < 0.005 (unpaired, 2-tailed Student’s t test) relative to the islets. Each dot represents a single sample. (D) Bar diagram (mean ± SD) shows qPCR analysis of Cxcl1 transcripts in pancreatic pericytes of ΔMyD88Peri mice (red) and nontransgenic (“non tg”; gray; the average was set to 1) mice. n = 3–6. ***P < 0.005 (unpaired, 2-tailed Student’s t test) compared with nontransgenic mice. Each dot represents a single sample. (E–G) Analyses of ΔMyD88Peri mice treated with recombinant Cxcl1 (rCxcl1; blue), PBS-treated ΔMyD88Peri (red), or PBS-treated nontransgenic (black line and gray bars) 15-week-old mice, 1 week after treatment. (E) Bar diagram (mean ± SD) showing the number of macrophages (MФ; CD45+CD64+ cells) and B cells (CD45+CD19+ cells) in 100 islets. n = 6–10. (F) IPGTT. Shown are the mean (± SEM) blood glucose levels (left) and the AUC (right). n = 7–8. (G) Bar diagrams (mean ± SD) showing expression of indicated genes in isolated islets. n = 4–7. *P < 0.05, ***P < 0.005; NS, not significant (1-way ANOVA with Tukey’s post hoc test). Each dot represents a single sample.

To test this hypothesis, we analyzed the impact of exogenous Cxcl1 on islet immune cells of ΔMyD88Peri mice. As shown in Figure 4E, recombinant Cxcl1 (rCxcl1) administration increased the number of macrophages and B cells within transgenic islets to the level observed in nontransgenic control islets. This finding suggests that pericytic Cxcl1 regulates the number of immune cells in the islets, and the decrease in Cxcl1 in ΔMyD88Peri mice likely contributes to the abnormal islet inflammation observed in these mice.

Next, we tested the contribution of Cxcl1 loss to glucose intolerance in ΔMyD88Peri mice. Treatment with rCxcl1 rescued the glucose response of transgenic animals, which became comparable to that of nontransgenic control mice (Figure 4F). Similar treatments did not significantly improve the glucose response in wild-type mice (Supplemental Figure 6) (64). Moreover, administration of rCxcl1 did not influence food intake or weight gain in ΔMyD88Peri mice for up to 4 weeks after treatment. To determine whether the observed reversal of glucose intolerance is associated with the correction of the β cell phenotype in transgenic mice, we analyzed the islets of rCxcl1-treated ΔMyD88Peri mice for genes associated with β cell maturity. As shown in Figure 4G, rCxcl1 treatment increased the expression of Ins1, MafA, and Unc3 in ΔMyD88Peri islets, making their levels comparable to those in nontransgenic controls. Thus, the low levels of pericytic Cxcl1 in ΔMyD88Peri mice likely contributed to their β cell failure and glucose intolerance.

In conclusion, our analysis suggested that pericytes produce Cxcl1 in a MyD88-dependent manner to regulate islet inflammation and support the β cell phenotype and glucose regulation.

IL-1β production by islet macrophages depends on pericytic MyD88 and Cxcl1. Next, we aimed to determine how pericytic MyD88 deficiency affects the ability of macrophages to support β cells (Figure 5A). As macrophages were shown to directly affect β cell function through the secretion of IL-1β (8), we tested the hypothesis that the loss of pericytic MyD88 interfered with the production of this cytokine. To determine whether Il1b expression was affected in ΔMyD88Peri mice, we compared its expression in nontransgenic and transgenic pancreatic macrophages and DCs (i.e., CD11c+ cells). As shown in Figure 5B, Il1b expression in pancreatic CD11c+ cells from ΔMyD88Peri mice was an order of magnitude lower than that in cells from control mice, supporting our hypothesis.

Pericytic MyD88 is required for immune IL-1β production. (A) A graphical model of the study’s hypothesis, with IL-1β highlighted in green. (B) Bar diagram (mean ± SD) showing reduced Il1b expression in pancreatic macrophages and DCs of transgenic mice. Shown is a qPCR analysis of Il1b transcripts in CD11c+ pancreatic cells from ΔMyD88Peri (red) and nontransgenic (gray) 15-week-old mice. n = 4–5. ***P < 0.005 (unpaired, 2-tailed Student’s t test). Each dot represents a single sample. (C) Cxcl1 induces Il1b expression in pancreatic cells in vivo. ΔMyD88Peri (blue) and nontransgenic (empty bars) adult mice were treated with mouse rCxcl1 (1 μg/g body weight), and their pancreatic immune cells (CD45+ cells) were analyzed 1 week later and compared to those of PBS-treated nontransgenic mice (gray bars, the average was set to 1). Bar diagrams (mean ± SD) showing Il1b expression analyzed by qPCR. n = 4–6. ***P < 0.005; NS, not significant (1-way ANOVA with Tukey’s post hoc test). Each dot represents a single sample. (D) IPGTT of rIL-1β–treated ΔMyD88Peri (purple) and PBS-treated nontransgenic (black line and gray bars) 15-week-old mice 1 day after treatment. The mean (± SEM) blood glucose levels (left) and area under the curve (AUC, right) are shown. n = 4–5. NS, not significant (1-way ANOVA with Tukey’s post hoc test). (E) qPCR analysis of islets isolated from rIL-1β–treated ΔMyD88Peri (purple), PBS-treated ΔMyD88Peri (red), and nontransgenic (gray) 15-week-old mice 1 day after treatment. Bar diagrams (mean ± SD) show expression of indicated genes. n = 4–6. *P < 0.05, ***P < 0.005; NS, not significant (1-way ANOVA with Tukey’s post hoc test). Each dot represents a single sample.

Cxcl1 induces IL-1β production by bone marrow–derived macrophages (65). To test whether Cxcl1, expressed by pancreatic pericytes in a TLR/MyD88-dependent manner, has similar effects on pancreatic immune cells in vivo, we treated mice with this cytokine. As shown in Figure 5C, rCxcl1 treatment boosted Il1b expression in pancreatic immune cells. Interestingly, rCxcl1 had a similar effect on cells isolated from wild-type and transgenic mice (Figure 5C). Thus, we suggest that pericytes regulate the local production of IL-1β by islet macrophages and DCs in healthy individuals in a MyD88-dependent manner, potentially through the production of Cxcl1.

IL-1β rescued the glucose intolerance of ΔMyD88Peri mice. Next, we tested the potential contribution of IL-1β to glucose intolerance in the ΔMyD88Peri mice. To this end, we injected transgenic animals with recombinant murine IL-1β (rIL-1β). One day after treatment, the glucose response of treated transgenic animals was comparable to that of nontransgenic controls (Figure 5D). Similar treatment with rIL-1β did not affect the glucose response in wild-type mice (Supplemental Figure 7). To define potential effects on β cell gene expression, we analyzed rIL-1β–treated ΔMyD88Peri islets. Similarly to Cxcl1, rIL-1β treatment in vivo increased the expression of Ins1, MafA, and Unc3 in ΔMyD88Peri islets, making their levels comparable to those in nontransgenic controls (Figure 5E). An increase in the expression of these 3 genes was also observed when islets isolated from ΔMyD88Peri mice were treated with rIL-1β in culture, suggesting a direct effect on β cells (Supplemental Figure 7). Thus, lower levels of IL-1β in ΔMyD88Peri mice likely contributed to β cell failure and glucose intolerance. In conclusion, our analysis suggested that pericytes produce Cxcl1 in a MyD88-dependent manner to support the β cell phenotype and glucose regulation, likely by facilitating IL-1β production by islet macrophages and DCs.