CXCL16-dependent scavenging of oxidized lipids by islet macrophages promotes differentiation of pathogenic CD8⁺ T cells in diabetic autoimmunity

SUMMARY

The pancreatic islet microenvironment is highly oxidative, rendering β cells vulnerable to autoinflammatory insults. Here we examined the role of islet resident macrophages in the autoimmune attack that initiates type 1 diabetes. Islet macrophages highly expressed CXCL16, a chemokine and scavenger receptor for oxidized low-density lipoproteins (OxLDL), regardless of autoimmune predisposition. Deletion of Cxcl16 in non-obese diabetic (NOD) mice suppressed the development of autoimmune diabetes. Mechanistically, Cxcl16 deficiency impaired clearance of OxLDL by islet macrophages, leading to OxLDL accumulation in pancreatic islets and a substantial reduction in intra-islet transitory CD8⁺ T cells (Tex^int) displaying proliferative and effector signatures. Tex^int cells were vulnerable to oxidative stress and diminished by ferroptosis; PD-1 blockade rescued this population and reversed diabetes resistance in NOD.Cxcl16^−/− mice. Thus, OxLDL scavenging in pancreatic islets inadvertently promotes differentiation of pathogenic CD8⁺ T cells, presenting a paradigm wherein tissue homeostasis processes can facilitate autoimmune pathogenesis in predisposed individuals.

Graphical Abstract

eTOC blurb

Pancreatic islet β cells are highly sensitive to autoinflammatory insults in part due to their highly microenvironment. Srivastava et al report that CXCL16 on resident islet macrophages scavenges OxLDL in pancreatic islets. This process, which supports tissue homeostasis, inadvertently promotes differentiation of pathogenic CD8⁺ T cells and autoimmune diabetes, underscoring a pathogenic role of tissue microenvironment in autoimmunity.

INTRODUCTION

Pancreatic islets of Langerhans are endocrine mini-organs that control glucose uptake and metabolism. Despite this important function, pancreatic islets are a frequent target of autoimmunity. In type 1 diabetes (T1D), diverse autoimmune cells invade the islets over a chronic course, causing permanent destruction of the insulin-producing β cells¹. Islets are also targeted by fulminant autoimmunity triggered by cancer checkpoint blockade immunotherapies, a phenomenon termed immune-related adverse events (irAE), leading to rapid β-cell failure^2,3. Diabetic autoimmunity is characterized by lymphocytic infiltration into the islet architecture, which only constitutes about 2% of the pancreas volume. The major exocrine tissue, however, is spared. This specific tissue destruction is thought to result from invasive antigen-specific T cell responses⁴. However, emerging evidence points toward a role of islet inherent properties in facilitating the autoimmune process⁵. The islet microenvironment at the steady state is highly oxidative⁶. β cells intrinsically express low levels of antioxidant enzymes, rendering them vulnerable to oxidative stress and autoinflammatory insults^7–10. Moreover, β cells can produce diabetogenic antigens irrespective of autoimmune predisposition. The crinophagic pathway, a physiological process by which β cells dispose of excessive insulin granules by fusing them with lysosomes, generates peptide antigens recognized by anti-islet T cells^11–13. When β cells degranulate in response to glucose stimulation, these antigenic products are released from the islets, acting as an antigen source to sensitize secondary lymphoid tissues¹⁴. Thus, the intrinsic islet microenvironment represents a pathogenic element in diabetic autoimmunity.

The islet microenvironment consists of several endocrine cell-types forming intimate interactions with the vascular network and the extracellular matrix¹⁵. This integrated ecosystem is essential for maintaining islet functions. Less appreciated is that islets also contain a local immune system organized by the resident macrophages¹⁶. Islet macrophages originate from definitive hematopoiesis, seeding islets during embryonic development¹⁷. A mature islet contains an average of 1–5 resident macrophages. As early as 3 weeks of age, islet macrophages from both C57BL/6 and NOD mice highly express MHC-II, CD68, and CD11c and produce TNFα and pro-IL-1β in the absence of stimulation^17,18. This phenotype is compatible with proinflammatory signatures described in barrier tissue macrophages¹⁹ and resembles disease-associated microglia, which increase expression of MHC-II, CD68, and CD11c during the progression of Alzheimer’s disease^20,21. Within the pancreas, only islet macrophages exhibit this proinflammatory signature, in contrast to exocrine pancreatic macrophages, which mainly exhibit an alternative activation profile¹⁷. Although this basal activation program may serve as a defensive mechanism against pathogen infections, its implication in islet autoimmunity remains unclear.

Depletion of islet macrophages substantially inhibits diabetes development in NOD mice²². Diabetes protection is most effective when islet macrophages are depleted at 3 weeks of age, a time point coinciding with the entry of the first islet-infiltrating T cells²². These findings suggest that the proinflammatory islet macrophages may have primed the islet microenvironment for the initial autoimmune attack. In this study, we aimed to identify key factors involved in this event and found a role of CXCL16-mediated OxLDL clearance in facilitating autoimmune pathogenesis.

RESULTS

Intrinsic expression of Cxcl16 in pancreatic islets promotes diabetic autoimmunity

We performed bulk RNA sequencing (RNA-seq) analysis in islet resident macrophages sourced from age- and sex-matched C57BL/6 and NOD mice. We examined female mice at pre-weaning (2-week-old), post-weaning (4-week-old), and adult (12-week-old) stages (Figure 1A). At each stage, we analyzed four biological replicates, each including ~1,500 islet macrophages (CD45⁺CD31⁻Thy1.2⁻B220⁻CD11c⁺F4/80⁺) FACS-sorted from ~800–1,000 handpicked islets from 6–8 mice (Figure S1A). In both strains, islet macrophages at all ages highly expressed genes encoding MHC-II (H2-Ab1), CD11c (Itgax), and CD68, and lineage-relevant genes including Cx3cr1 and Csf1r (Figure 1B). Notably, under all conditions, islet macrophages, exhibited high expression (top 1%) of genes involved in NFκB activation, including transcription factors (Jun, Fos), cytokines (Tnf, Il1b), and chemokines (Cxcl16) (Figure 1B). In contrast, the expression of alternative activation genes, including Arg1, Socs1, and Adora2a, was much lower (Figure 1B; Figure S1B). A set of NFκB activation genes (Junb, Atf3, Tnf, Il1b, Ccl3) showed a higher relative expression in 2-week-old C57BL/6 mice than age-matched NOD mice (Figure S1C). Although islet macrophages from 12-week-old NOD mice upregulated an additional set of interferon-responsive genes associated with autoimmunity, they sustained the high expression of NFκB activation genes (Figure S1D). Thus, islet macrophages intrinsically develop a basal activation program that is established early in life, independent of autoimmune predisposition, and persists during autoimmune progression.

Figure 1. Intrinsic expression of Cxcl16 in pancreatic islets promotes diabetic autoimmunity.

(A) Schematics of isolating islet resident macrophages from C57BL/6 and NOD mice for RNA-seq analysis.

(B) RNA-seq analysis illustrating top 1% expressed genes that are commonly expressed by islet macrophages between C57BL/6 and NOD mice of indicated ages.

(C)A heatmap depicting the gene expression of all the chemokines (ranked by their total expression level) in islet macrophages.

(D) Diabetes incidence in female NOD.Ccrl2^−/−, NOD.Ms4a7^−/−, and NOD mice.

(E) Diabetes incidence in female NOD, NOD.Cxcl16^+/−, and NOD.Cxcl16^−/− mice.

Statistics are analyzed by log-rank (Mantel-Cox) test in (D) and (E). **P < 0.01; ****P < 0.0001.

See also Figures S1 and S2.

We next sought to assess the relevance of the basal macrophage activation program in autoimmune diabetes. Besides the well-described role of MHC-II presentation and inflammatory cytokines (TNFα and IL-1β) in T1D pathogenesis^23–25, we investigated three other targets, Cxcl16, Ccrl2, and Ms4a7, which exhibited high expression levels (Figure 1B). Under all conditions, Cxcl16 was the most highly expressed chemokine gene (Figure 1C), implying a role in recruiting CXCR6⁺ T cells²⁶. Chemokine (C-C motif) receptor-like 2 (CCRL2) shares key features with atypical chemokine receptors²⁷ and regulates dendritic cell (DC) and NK cell trafficking^28,29. Ms4a7 is a member of the Ms4a gene family which encodes tetraspanin-like proteins. Although the expression of Ms4a7 is largely specific to macrophages, its function in inflammation remains unknown³⁰.

We generated NOD mice with Cxcl16, Ccrl2, and Ms4a7 individually deleted by CRISPR-Cas9 (Figure S2A). Diabetes development in female or male mice lacking either Ccrl2 or Ms4a7 was comparable to WT NOD mice (Figure 1D; Figure S2B), indicating their dispensable role in autoimmune diabetes. In contrast, we observed a significant protection against diabetes development in female NOD.Cxcl16^−/− mice during a 40-week monitoring period, whereas a majority of NOD or NOD.Cxcl16^+/− controls became diabetic (Figure 1E). Male NOD mice are generally more resistant to diabetes onset. However, male NOD.Cxcl16^−/− mice also showed significant protection (Figure S2C). Additionally, we performed glucose tolerance tests in 4-week-old male NOD and NOD.Cxcl16^−/− mice, when islet inflammation is minimal, and found no difference between the two strains (Figure S2D). The NOD.Cxcl16^−/− mice also exhibited similar serum insulin levels during ad libitum, fasting, and post glucose challenge (Figure S2E), suggesting that they maintain normal β-cell function before autoimmunity occurs. Thus, as an intrinsic component in the islet microenvironment, Cxcl16 also plays a pathogenic role in islet autoimmunity.

Diabetes protection in NOD.Cxcl16^−/− mice is mediated through an islet-specific mechanism

CXCL16 is the only chemokine known as a scavenger receptor for OxLDL³¹. CXCL16 is synthesized as a transmembrane molecule, which is transported to the cell surface to function as a scavenger receptor³² (Figure 2A). Metalloproteinases, including ADAM10 and ADAM17, cleave the transmembrane form to generate a soluble chemokine³². We used intracellular staining to determine the total expression level of CXCL16 and found that it was present in a majority of islet macrophages in 4-week-old male NOD mice (Figure 2A). About 60% of the islet macrophages also showed CXCL16 expression on the cell surface (Figure 2A), suggesting a role as a scavenger receptor. In contrast, islet macrophages from NOD.Cxcl16^−/− mice completely lacked CXCL16 expression (Figure 2A). Thus, consistent with our RNA-seq data, CXCL16 is highly expressed by islet macrophages before the manifestation of autoimmunity.

Figure 2. Cxcl16 deficiency mediates protection against autoimmune diabetes via an intraislet mechanism.

(A) Schematics (left) and FACS plots (right) showing different forms of CXCL16 expression in islet resident macrophages.

(B) RT PCR analysis of indicated genes in islet macrophages from NOD and NOD.Tnfrsf1a/1b^−/− mice.

(C) Flow cytometry analysis showing CXCL16 total expression in islet macrophages from NOD and NOD.Tnfrsf1a/1b^−/− mice.

(D) Flow cytometry analysis showing CXCL16 total expression in indicated tissue resident macrophages.

(E) Flow cytometry analysis showing CXCL16 expression in islet macrophages from three non-diabetic humans.

(F) Flow cytometry analysis showing the percentage of intra-islet CD45⁺ leukocytes.

(G) Flow cytometry analysis showing the percentage of indicated intra-islet populations.

(H) Pie charts showing the composition of intra-islet immune cells within CD45⁺ leukocytes.

(I) ELISA results for serum CXCL16 levels in indicated mice.

(J) Representative images of hematoxylin and eosin (H&E) staining in submandibular gland sections (top). The arrows indicate lymphocytic infiltration foci. The quantification summarizes the percentage of the infiltrated area.

(K) Representative images of H&E staining in pancreatic sections (left). The quantification summarizes insulitis scores of 100–120 islets from 4 mice per strain.

(L) Schematics (left) and diabetes incidence (right) of NOD.Rag1^−/− recipients post transfer of splenocytes from age- and sex-matched NOD or NOD.Cxcl16^−/− donors.

Data are mean ± SEM. Each dot represents individual mice (D, F, G, I, J), islets (K), or experiments (B, C). The results are from at least 2–4 independent experiments. Statistics are analyzed by Mann-Whitney test (B, F, G, J, K), paired t-test (C), One-way ANOVA with Dunnett’s multiple comparisons test (D, I), or log-rank (Mantel-Cox) test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Figure S3.

Correlating to the described alternative activation profile¹⁷, exocrine pancreatic macrophages (CD11b⁺F4/80⁺CD206⁺CD301⁺) from age- and sex-matched NOD and NOD.Cxcl16^−/− mice (4-week-old male) showed minimal CXCL16 expression (Figure S3A). To investigate the link between CXCL16 expression and NFκB activation, we assessed NOD mice lacking TNF receptors 1 and 2 (NOD.Tnfrsf1a/1b^−/−). RT-PCR analysis showed a significant decrease in the expression of Cxcl16 and other NFκB activation genes (Ilb and Fos) in islet macrophages from NOD.Tnfrsf1a/1b^−/− mice (Figure 2B), with the reduction of CXCL16 protein confirmed by flow cytometry analysis (Figure 2C). Thus, TNF signaling in islet macrophages contributes to their high expression of CXCL16.

Besides islet macrophages (Figure S3B), we also examined peritoneal, adipose tissue, lung alveolar and interstitial macrophages from 4-week-old male NOD and C57BL/6 mice (Figures S3C–S3E). Using both RT-PCR (Figure S3F) and flow cytometry (Figure 2D; Figures S3G and S3H), we found that islet macrophages in both strains exhibited the highest levels of CXCL16 total expression. These data suggest a strong association of CXCL16 expression with the islet microenvironment.

To test whether CXCL16 is also expressed by human islet macrophages, we analyzed ~2,000 islets handpicked from three individual non-diabetic donors. We identified a small portion (~1–2%) of CD45⁺ leukocytes in dispersed islet cells, which consisted of a macrophage population expressing CD11b, CD11c, and F4/80 (Figure S3I). In all three donors, we detected consistent expression (both total and surface) of CXCL16 (Figure 2E). Thus, expression of CXCL16 is conserved between mouse and human islet macrophages.

To assess how Cxcl16 deficiency may affect islet inflammation, we examined prediabetic female NOD and NOD.Cxcl16^−/− mice at 8 and 12 weeks of age (Figure 2F). In contrast to the substantial increase of CD45⁺ leukocytes in NOD mice during this effector phase, CD45⁺ leukocytes (among total islet cells) in NOD.Cxcl16^−/− mice remained significantly lower (Figure 2F). A significant reduction was also seen in intra-islet CD4⁺ and CD8⁺ T cells, B cells, and DCs in NOD.Cxcl16^−/− mice (Figure 2G). CD45⁺ leukocytes in NOD mice mainly consisted of non-resident, infiltrating populations (CD4⁺ and CD8⁺ T cells, B cells, and DCs), whereas the resident macrophage population was less prominent (Figure 2H). Conversely, the few CD45⁺ leukocytes in NOD.Cxcl16^−/− mice showed a further reduction in the composition of intra-islet CD4⁺ and CD8⁺ T cells (Figure 2H). Thus, Cxcl16 deficiency significantly restricted immune cell infiltration into pancreatic islets.

Although CXCL16 is highly expressed by islet macrophages, it can be produced by other cells in the body^34–36. To determine whether systemic CXCL16 may contribute to T1D development, we first assessed serum CXCL16 levels. While pro-inflammatory mediators like IFNγ progressively increase in serum as diabetes advances, anti-inflammatory factors such as IL-10 decline³⁷. Serum CXCL16 levels remained comparable between 4 and 10 weeks of age and were even lower in diabetic NOD mice (Figure 2I). Second, we examined whether Cxcl16 deficiency might influence the development of autoimmune sialadenitis that targets salivary glands, an autoimmune condition resembling Sjögren’s syndrome in humans. Histologic analysis revealed evident lymphocytic infiltration foci in submandibular glands of NOD.Cxcl16^−/− mice (12–16-week-old female), comparable to WT NOD mice (Figure 2J). However, in pancreatic tissues from the same mice, a significant reduction in insulitis was observed in NOD.Cxcl16^−/− mice (Figure 2K). Thus, Cxcl16 deficiency specifically inhibited islet autoimmunity with minimal impact on exocrine salivary glands. Third, we performed adoptive transfer experiments using splenocytes from 12–16-week-old female NOD or NOD.Cxcl16^−/− mice to 6-week-old NOD.Rag1^−/− recipients. Both NOD and NOD.Cxcl16^−/− donor splenocytes induced diabetes to a similar extent in the NOD.Rag1^−/− recipients (Figure 2L), indicating that Cxcl16 deficiency did not significantly affect the diabetogenic potential of the peripheral T cells. Thus, the protective effect of Cxcl16 deletion mainly occurs in pancreatic islets, with much less dependence on systemic mechanisms.

CXCL16 plays a crucial role in scavenging OxLDL in pancreatic islets

The proportions of intra-islet CXCR6⁺ CD4⁺ and CD8⁺ T cells were comparable between NOD and NOD.Cxcl16^−/− mice (Figure S4A), consistent with previous studies showing a minimal influence of CXCR6 deficiency on T cell migration to inflamed islets and diabetes development³⁸. Although CXCR6 was upregulated on islet-specific NY8.3 CD8⁺ T cells upon activation with the cognate glucose-6-phosphatase catalytic subunit-related protein 206–214 peptide (IGRP_206–214) (Figure S4B), adding recombinant CXCL16 did not affect CD44 expression or T cell proliferation in these T cells (Figure S4C). Thus, Cxcl16 deficiency does not entirely abolish the recruitment of CXCR6-expressing T cells into islets or influence their activation.

To examine whether deleting Cxcl16 may impact the proinflammatory phenotype of islet macrophages, we compared the expression of CD11c, MHC-II (I-A^g7), CD80, CD40, and PD-L1, and found no significant differences due to Cxcl16 deficiency (Figure S5A). Also, islet macrophages from NOD.Cxcl16^−/− mice were able to produce both TNFα and pro-IL-1β without stimulation (Figure S5B) and were equally viable to those from NOD mice (Figure S5C). To assess presentation of insulin, a major T1D antigen³⁹, we used dispersed islet cells from 4-week-old male mice, in which islet macrophages are the major APC. There was no significant difference in either spontaneous presentation (no antigen pulse) or presentation of exogenous insulin B-chain 9–23 peptide (Figure S5D). Collectively, Cxcl16 deficiency does not significantly alter the basal activation, survival, or the ability of islet macrophages to activate intra-islet T cells.

Based on previous studies showing that pancreatic islets contain oxidized lipids^40–43 and our observation of surface expression of CXCL16 on islet macrophages, we investigated its role in scavenging OxLDL. We used three complementary approaches to evaluate OxLDL clearance by islet macrophages from 4-week-old male NOD and NOD.Cxcl16^−/− mice. First, we examined in vitro uptake of fluorescently labeled OxLDL. Islet macrophages from NOD.Cxcl16^−/− mice showed significantly reduced uptake capacity compared to those from age- and sex-matched NOD mice (Figure 3A). Conversely, peritoneal macrophages from NOD.Cxcl16^−/− mice did not show any defect (Figure 3B), corresponding to their weak expression of CXCL16 (Figure 2D). Second, we measured endogenous OxLDL levels in vivo and found a significant decrease in islet macrophages from NOD.Cxcl16^−/− mice (Figure 3C). Third, we examined lipid peroxidation levels in islet macrophages using the BODIPY C11 reagent. Significantly lower levels of lipid peroxidation were detected in NOD.Cxcl16^−/− mice (Figure 3D). Collectively, these data indicate an important role of CXCL16 in mediating OxLDL scavenging by islet macrophages.

Figure 3. CXCL16 expression by islet macrophages is crucial for OxLDL clearance in pancreatic islets.

(A-B) Flow cytometry analysis showing in vitro uptake of fluorescently labeled OxLDL by islet (A) or peritoneal (B) macrophages from NOD or NOD.Cxcl16^−/− mice.

(C-D) Flow cytometry analysis showing endogenous OxLDL (C) or lipid peroxidation (D) levels in islet macrophages from NOD or NOD.Cxcl16^−/− mice.

(E-F) Confocal microscopy of OxLDL in intact islets from NOD or NOD.Cxcl16^−/− (E) or indicated BM chimera (F) mice. The quantification summarizes corrected OxLDL intensity per islet area. Scale bar: 10 μm.

(G) RNA-seq analysis showing the expression of several genes encoding scavenger receptors for oxidized lipids in islet macrophages.

(H-I) Flow cytometry analysis showing in vitro uptake of fluorescently labeled OxLDL by islet (H) or peritoneal (I) macrophages from C57BL/6 or B6.Cd36^−/− mice.

(J-K) Flow cytometry analysis showing endogenous OxLDL (J) or lipid peroxidation (K) levels in islet macrophages from C57BL/6 or B6.Cd36^−/− mice.

(L) Flow cytometry analysis showing endogenous OxLDL levels in intra-islet CD8⁺ and CD4⁺ T cells from NOD and NOD.Cxcl16^−/− mice.

Data are mean ± SEM. Each dot represents individual mice (A-D, H-L) or islets (E, F). The results are from at least 2–4 independent experiments. Statistics are analyzed by Mann-Whitney test (A-F, H-L). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Figures S4, S5, and S6.

To examine whether Cxcl16 deficiency may impact OxLDL levels in the entire islet microenvironment, we used confocal imaging to examine OxLDL levels in non-sectioned, intact islets from 4-week-old male NOD and NOD.Cxcl16^−/− mice. We observed a significant increase in the overall intensity of OxLDL per islet area in NOD.Cxcl16^−/− mice (Figure 3E), indicating that impaired OxLDL clearance by islet macrophages resulted in OxLDL accumulation in islets. To confirm these results, we performed bone marrow (BM) chimera experiments. Specifically, BM stem cells from NOD or NOD.Cxcl16^−/− donors were transplanted into lethally irradiated NOD.Cxcl16^−/− or NOD recipients, respectively. Because islet macrophages in lethally irradiated hosts are completely replaced by donor BM cells within 7 to15 weeks¹⁷, this system allowed us to evaluate islets from NOD mice reconstituted with Cxcl16-deficient islet macrophages (KO→WT) and vice versa (WT→KO). We performed confocal microscopy in intact islets isolated between 8~10 weeks after reconstitution and found that islets sourced from the KO→WT group had significantly higher total OxLDL levels than those from the WT→KO group (Figure 3F). Thus, reconstituting islets with Cxcl16-deficient macrophages sufficiently led to OxLDL accumulation, a phenotype resembling NOD.Cxcl16^−/− mice.

Several other scavenger receptors, including CD36, macrophage scavenger receptor (MSR1), scavenger receptor class B, type I (SR-BI), and lectin-like OxLDL receptor 1 (LOX-1), take up OxLDL by tissue-resident macrophages^44–46. In our bulk RNA-seq analysis, the expression of Msr1, Scarb1 (encoding SR-BI), and Olr1 (encoding LOX-1) was largely negative in islet macrophages from both C57BL/6 and NOD mice across different ages (Figure 3G). We found a weak expression of Cd36, which was about 30–40-fold lower than Cxcl16 (Figure 3G). We used the same functional assays to examine whether islet macrophages might also employ CD36 for OxLDL clearance. By measuring in vitro uptake of fluorescently labeled OxLDL, islet macrophages from B6.Cd36^−/− mice showed comparable levels to WT C57BL/6 mice (Figure 3H). Conversely, peritoneal macrophages from B6.Cd36^−/− mice exhibited a significant impairment (Figure 3I), consistent with previous studies showing an important role of CD36 in OxLDL uptake by peritoneal macrophages^47–51. Also, we did not find notable changes in either endogenous OxLDL levels (Figure 3J) or lipid peroxidation (Figure 3K) in islet macrophages from B6.Cd36^−/− mice. Moreover, OxLDL levels in intact islets were similar between WT C57BL/6 and B6.Cd36^−/− mice (Figure S6A). Thus, islet macrophages did not show a strong dependence on CD36 for OxLDL uptake, a phenomenon distinct from peritoneal macrophages.

Next, we assessed OxLDL scavenging in islets undergoing active autoimmunity and found that intra-islet DCs in 8-week-old female NOD mice exhibited a mild level of total CXCL16 expression, which was significantly less than that in islet macrophages (Figure S6B). However, surface CXCL16 expression was minimal in these intra-islet DCs (Figure S6C), suggesting a limited role as a scavenger receptor. Confirming this, both endogenous OxLDL (Figure S6D) and lipid peroxidation (Figure S6E) levels in intra-islet DCs were comparable between NOD and NOD.Cxcl16^−/− mice. Thus, Cxcl16 deficiency primarily affects OxLDL scavenging by islet macrophages rather than DCs.

Islet-infiltrating T cells showed no detectable CXCL16 (Figure S6B), suggesting that they did not directly use CXCL16 for OxLDL uptake. However, the observed OxLDL accumulation in islets of NOD.Cxcl16^−/− mice suggested that intra-islet T cells might be indirectly affected in this altered tissue microenvironment. We detected significantly higher levels of endogenous OxLDL in intra-islet CD8⁺ T cells from NOD.Cxcl16^−/− mice than WT NOD mice (Figure 3L), whereas the intra-islet CD4⁺ T cells were not significantly affected (Figure 3L). Thus, Cxcl16 deficiency led to OxLDL accumulation in pancreatic islets, which, in turn, increased OxLDL exposure to islet-infiltrating CD8⁺ T cells.

Cxcl16 deficiency modulates intra-islet CD8⁺ T cell differentiation

The marked reduction in intra-islet CD4⁺ and CD8⁺ T cells in NOD.Cxcl16^−/− mice led us to further examine how Cxcl16 deficiency may impact on their cellular heterogeneity by single-cell RNA sequencing (scRNA-seq) analysis. To assess comparable numbers of T cells between the two strains, we included ~200 and ~1200 handpicked islets from 12-week-old female NOD and NOD.Cxcl16^−/− mice, respectively. We analyzed 1934 CD4⁺ and 1379 CD8⁺ T cells pooled from the NOD and NOD.Cxcl16^−/− samples. For CD4⁺ T cells, we classified clusters corresponding to naïve, Th1-like, anergy, type 1 regulatory (Tr1), and regulatory T cell (Treg) phenotypes (Figure S7A), by respective signature genes (Figures S7B and S7C). We did not observe major differences in the distribution of these CD4⁺ T cell clusters between NOD and NOD.Cxcl16^−/− mice (Figure S7D). However, there was a moderate increase of Tregs in NOD.Cxcl16^−/− mice. We then verified this result by flow cytometry, which showed a consistent increase in the frequency of Foxp3⁺ Tregs among intra-islet CD4⁺ T cells in NOD.Cxcl16^−/− mice (Figure S7E). Moreover, this phenotype was specifically observed in intra-islet Tregs but not in pLN (Figure S7E). We then treated NOD.Cxcl16^−/− mice with the anti-CD25 (clone PC61) antibody and observed accelerated diabetes development in both NOD and NOD.Cxcl16^−/− mice (Figure S7F). This result suggests that removal of the Treg control can overcome diabetes resistance in NOD.Cxcl16^−/− mice. However, since the anti-CD25 antibody treatment mainly targets Tregs in the periphery⁵³, accelerated diabetes development in NOD.Cxcl16^−/− mice may largely involve mobilization of peripheral T cells.

We identified four major CD8⁺ T cell clusters with distinctive transcriptional profiles (Figure 4A). CD8–1 represented naïve CD8⁺ T cells, with high expression of Il7r, Sell (encoding CD62L), and Lef1 (Figures 4B and 4C; Figure S8A). Except for CD8–1, all other clusters expressed Pdcd1 (encoding PD-1) and Tox (Figure 4B). CD8–2 showed high expression of Tcf7 (encoding TCF1) and Slamf6 (Figures 4B and 4C; Figure S8A), which are markers of stem-like CD8⁺ T cells (Tex^prog). In contrast, CD8–3 and CD8–4 had low expression of Tcf7 and Slamf6 but showed increased expression of effector genes, such as Gzmb and Ifng (Figures 4B and 4C). Notably, CD8–3 was further distinguished by its strong expression of genes involved in cell cycling (Stmn1, Birc5, Mki67) (Figures 4B and 4C). These signatures closely resemble transitory (Tex^int) CD8⁺ T cells (TCF1⁻PD-1⁺CD101⁻Tim3⁺) in chronic LCMV infection, which proliferate and execute effector functions during viral control⁵⁴. Compared to CD8–3, CD8–4 retained the expression of effector genes, downregulated proliferative genes, and upregulated genes encoding several inhibitory molecules, including Tim3, LAG3, TIGIT, and CTLA4 (Figure 4C), suggesting that CD8–4 represented a population that was further differentiated into exhausted-like T cells with effector functions (Tex^eff). By gene set enrichment analysis (GSEA), differentially expressed genes (DEGs) between CD8–3 and CD8–4 were significantly enriched for the transitory (CD101⁻Tim3⁺) and exhausted (CD101⁺Tim3⁺) CD8⁺ T cells described in LCMV infection⁵⁴, respectively (Figure 4D). Notably, in NOD.Cxcl16^−/− mice, we observed a marked reduction (~40%) in the proportion of the CD8–3 Tex^int population, with a moderate decrease (~17%) also seen in CD8–4 (Figure 4E). Thus, aside from a substantial decrease in total CD8⁺ T cell numbers, the remaining intra-islet CD8⁺ T cells in NOD.Cxcl16^−/− mice exhibited a modified program of their local differentiation.

Figure 4. Cxcl16 deficiency leads to a substantial decrease in intra-islet Tex^int CD8⁺ T cells.

(A) A UMAP plot depicting intra-islet CD8⁺ T cell clusters, merged between NOD and NOD.Cxcl16^−/− mice.

(B) Feature plots depicting expression of several signature genes classifying CD8⁺ T cell heterogeneity.

(C) Volcano plots depicting DEGs in indicated CD8⁺ T cell clusters.

(D) GSEA of DEGs between CD8–3 and CD8–4 clusters with DEGs between transitory and exhausted CD8⁺ T cells from a published dataset (BioProject: PRJNA497086).

(E) UMAP plots (upper) depicting CD8⁺ T cell clusters between NOD and NOD.Cxcl16^−/− mice. The bar graph (lower) shows relative frequency of indicated clusters.

(F-H) Representative FACS plots (F) and quantification of intra-islet TCF1⁻TOX⁺ CD8⁺ T cells (gated on CD45⁺CD3⁺CD8⁺) among total CD8⁺ T (G) or islet (H) cells.

(I-K) Representative FACS plots (I) and quantification of intra-islet SLAMF6^lowCD39⁺ CD8⁺ T cells (gated on CD45⁺CD3⁺CD8⁺PD1⁺) among PD1⁺CD8⁺ T (J) and total islet (K) cells.

(L) Representative FACS plots showing K^d:IGRP_206–214 tetramer-binding cells.

(M-O) Representative FACS plots (M) and quantification of IGRP_206–214-reactive SLAMF6^lowCD39⁺ CD8⁺ T cells among PD1⁺CD8⁺ T (N) and total islet (O) cells.

Data are mean ± SEM. Each dot represents individual mice (G, H, J, K, N, O). The results are from at least 2–4 independent experiments. Statistics are analyzed by Benjamini-Hochberg correction for multiple tests (C) or Mann-Whitney test (G, H, J, K, N, O). *P < 0.05; **P < 0.01; ***P < 0.001. See also Figures S7 and S8.

To validate our transcriptional data, we performed flow cytometry analysis on individual 12-week-old female NOD and NOD.Cxcl16^−/− mice (~200 handpicked islets per mouse). Based on the gene signature, we identified three subsets in total intra-islet CD8⁺ T cells by differential expression of transcription factors TCF1 and TOX: TCF1⁺TOX⁻ cells corresponding to the naive CD8–1 cluster, TCF1⁺TOX⁺ cells representing the CD8–2 Tex^prog cluster, and TCF1⁻TOX⁺ cells indicative of a combination of CD8–3 and CD8–4 (Figure 4F). Furthermore, the TCF1⁻TOX⁺ population showed a significant reduction in the proportion of total intra-islet CD8⁺ T cells in NOD.Cxcl16^−/− mice (Figure 4G), consistent with the decrease observed in CD8–3 and CD8–4 (Figure 4E). Notably, when assessed by the frequency of the total islet cells, the TCF1⁻TOX⁺ population was barely present in the islets of NOD.Cxcl16^−/− mice (Figure 4H).

Next, we utilized a panel of surface markers to further classify intra-islet PD-1⁺CD8⁺ T cells (containing CD8–2, CD8–3, and CD8–4). We identified a PD-1⁺SLAMF6⁺CD39⁻ subset corresponding to the CD8–2 cluster (Figure 4I), which also expressed CXCR5 (Figure S8B), a marker of the Tex^prog cells. The other subset, PD-1⁺SLAMF6^lowCD39⁺ (Figure 4I), was consistent with the gene expression profile shared by CD8–3 and CD8–4 (Figures 4B and 4C). Indeed, we observed a significant reduction in the frequency of the SLAMF6^lowCD39⁺ population in PD-1⁺CD8⁺ T cells (Figure 4J) or total islet cells (Figure 4K) from NOD.Cxcl16^−/− mice. We attempted to further classify the CD8–3 Tex^int and the CD8–4 Tex^eff cells within the SLAMF6^lowCD39⁺ population by examining CX₃CR1 and CD101 expression as described in LCMV infection^54,55. However, we could not detect the expression of either marker in intra-islet CD8⁺ T cells (Figure S8C).

To extend this analysis to β-cell-specific CD8⁺ T cells, we labeled islet infiltrates with the MHC-I (K^d) tetramer incorporating the IGRP_206–214 peptide⁵⁶, which allowed the identification of intra-islet CD8⁺ T cells reactive to IGRP_206–214 (Figure 4L). The proportion of intra-islet IGRP_206–214-reactive T cells among total CD8⁺ T cells was comparable between WT and NOD.Cxcl16^−/− mice (Figure S8D). However, in both tetramer-negative (Figure S8E) and IGRP_206–214-reactive (Figure 4M) T cells, the SLAMF6^lowCD39⁺ subset in NOD.Cxcl16^−/− mice showed a significant reduction in the frequency among PD-1⁺CD8⁺ T cells (Figure 4N) or total islet cells (Figure 4O). Thus, Cxcl16 deficiency disrupts intra-islet differentiation of pathogenic CD8⁺ T cells, which could have a crucial role in providing protection against islet autoimmunity.

OxLDL exposure diminishes effector functions of anti-islet CD8⁺ T cells

Based on previous studies demonstrating a prominent role of oxidized lipids in inducing T cell dysfunction^57–61, we hypothesized that intra-islet Tex^int cells may be intrinsically more susceptible to OxLDL-induced damage. To test this, we analyzed DEGs that were commonly upregulated in the CD8–3 Tex^int cluster compared to each of the other clusters. In addition to strong signatures of cell cycling and effector functions, these DEGs were also enriched for cell death, mTORC1 activation, metabolic reprogramming to glycolysis, and oxidative stress (Figure 5A). Specifically, the CD8–3 Tex^int cluster highly expressed a set of genes (e.g., Ppia, Top2a, Idi1, Prdx1, Fdps) known to be involved in peroxisome, lipid peroxidation, and responses to oxidative stress^62–65 (Figure 5B), suggesting susceptibility to oxidative stress.

Figure 5. OxLDL exposure reduces intra-islet Tex^int CD8⁺ T cells susceptible to oxidative stress.

(A) Enrichment of genes highly expressed by the CD8–3 cluster versus other subsets for MSigDB pathways.

(B) Heatmaps depicting oxidative stress-related genes that are highly expressed by CD8–3.

(C) Schematics of the experimental design (left), representative FACS plots (middle), and quantification (right) of the percentage of SLAMF6^lowCD39⁺ CD8⁺ T cells during exposure to native LDL or OxLDL.

(D) Schematics of the experimental design (left) and quantification (right) of the percentage of SLAMF6^lowCD39⁺ CD8⁺ T cells exposed to OxLDL with addition of islet macrophages from NOD and NOD.Cxcl16^−/− mice.

(E) Representative FACS plots (left) and quantification (right) showing intra-islet TCF1⁻TOX⁺ CD8⁺ T cells (gated on CD45⁺CD3⁺CD8⁺) in NOD and NOD.Cxcl16^−/− mice post HFD or SFD.

(F-H) NY8.3 CD8⁺ T cells were activated in vitro and then exposed to OxLDL. (F) The frequency of CD44⁺CD8⁺ T cells. (G) Proliferation (CFSE dilution) in NY8.3 CD8⁺ T cells. (H) Intracellular cytokine staining of IFNγ and TNFα upon restimulation with PMA and ionomycin.

Data are mean ± SEM. Each dot represents individual mice (E) or biological replicates (C, D, F, G, H) from three independent experiments using 4–10 mice per experiment. Statistics are analyzed by RM one-way ANOVA with Tukey’s multiple comparisons test (C), Two-way ANOVA with Tukey’s multiple comparisons test (D, E), or paired t test (F, G, H). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Figures S9 and S10.

For functional validation, we developed an in vitro culture system to assess the impact of OxLDL on intra-islet CD8⁺ T cells. We hypothesized that exposing intra-islet CD8⁺ T cells from NOD mice to OxLDL may recapitulate the reduction of the Tex^int CD8⁺ T cells observed in NOD.Cxcl16^−/− mice. Specifically, we isolated ~1,000 islets from 12–16-week-old female NOD mice and cultured equal aliquots of dispersed islet cells in media alone or with the addition of native LDL or OxLDL (Figure 5C). In comparison to cells cultured in media, OxLDL caused a consistent and significant reduction in the frequency of the SLAMF6^lowCD39⁺ CD8⁺ T cell subset, whereas exposure to native LDL did not show notable differences (Figure 5C). Thus, OxLDL exposure can directly diminish intra-islet Tex^int CD8⁺ T cells.

To further determine the relationship between the reduction of Tex^int cells and OxLDL clearance by islet macrophages, we enriched islet macrophages from dispersed islet cells of 4-week-old male NOD or NOD.Cxcl16^−/− mice and co-cultured them with intra-islet CD8⁺ T cells undergoing OxLDL exposure (Figure 5D). Islet macrophages from NOD mice consistently rescued the decrease of the SLAMF6^lowCD39⁺ subset caused by OxLDL exposure, whereas those from NOD.Cxcl16^−/− mice were unable to do so (Figure 5D; Figure S9A). To confirm these results, we generated a Hoxb8-derived macrophage line expressing Cxcl16 (Hoxb8-Cxcl16). This enabled the expression of CXCL16 protein on the surface of the Hoxb8-Cxcl16 cells (Figure S9B). Notably, OxLDL-mediated reduction of the SLAMF6^lowCD39⁺ CD8⁺ T cells was only rescued by the Hoxb8-Cxcl16 but not the control cells (Hoxb8-Ctrl) (Figure S9C). Thus, CXCL16 expression sufficiently cleared OxLDL in the culture to prevent the reduction of the Tex^int CD8⁺ T cells.

Next, we investigated whether phenotypes observed in NOD.Cxcl16^−/− mice could be reproduced in vivo when OxLDL is increased systemically. High fat diet (HFD) elevates serum OxLDL levels^66,67. Moreover, NOD mice given a HFD are unexpectedly prevented from diabetes onset⁶⁸. To explore whether this phenotype relates to changes in the intra-islet Tex^int CD8⁺ T cells, we fed female NOD and NOD.Cxcl16^−/− mice either a HFD or a control standard fat diet (SFD) starting at 4 weeks of age. By 12 weeks of age, histologic analysis revealed less insulitis development in islets from NOD mice fed HFD (Figure S10A). While NOD mice on the HFD showed noticeable weight gain, intriguingly, HFD-fed NOD.Cxcl16^−/− mice did not exhibit significant weight gain (Figure S10B). Moreover, islet infiltration of CD45⁺ leukocytes (Figure S10C) and total CD8⁺ T cells (Figure S10D) in NOD.Cxcl16^−/− mice remained low regardless of HFD or SFD, suggesting that the baseline anti-inflammatory phenotype in NOD.Cxcl16^−/− mice may have attenuated the suppressive effects of HFD feeding. To assess changes in CD8⁺ T cell subsets, we examined the TCF1⁺TOX⁻ (CD8–1), TCF1⁺TOX⁺ (CD8–2), and TCF1⁻TOX⁺ (CD8–3 and 4) populations in total intra-islet CD8⁺ T cells. Notably, HFD significantly reduced the proportion of the TCF1⁻TOX⁺ subset in NOD mice to a level comparable to NOD.Cxcl16^−/− mice fed either HFD or SFD (Figure 5E). Thus, HFD-induced protection against diabetes may be related to the reduction of the TCF1⁻TOX⁺ CD8⁺ T cells, which is similar to the phenotype observed in NOD.Cxcl16^−/− mice.

To assess how OxLDL might affect islet-specific effector CD8⁺ T cells, we exposed activated NY8.3 CD8⁺ T cells to OxLDL and found a substantial decrease in the CD44^hi effector population (Figure 5F). To evaluate T cell proliferation, we measured CFSE dilution in NY8.3 CD8⁺ T cells and found significantly reduced cell division upon exposure to OxLDL (Figure 5G). To further examine effector functions, we assessed the production of IFNγ and TNFα by intracellular cytokine staining in NY8.3 CD8⁺ T cells upon restimulation. Remarkably, OxLDL exposure significantly reduced the production of both IFNγ and TNFα by NY8.3 CD8⁺ T cells (Figure 5H). Thus, OxLDL can directly dampen the effector responses of anti-islet CD8⁺ T cells and reduce their pathogenicity.

Ferroptosis contributes to OxLDL-mediated reduction of intra-islet Tex^int CD8⁺ T cells

During increased uptake of fatty acids, effector T cells undergo cell death by ferroptosis^58,60,61. We referred to a ferroptosis database⁶⁹ and found a higher expression of 210 ferroptosis driver genes in the CD8–3 Tex^int cluster relative to other CD8⁺ T cell subsets (Figure 6A). Notably, we observed upregulation of Acsl4 (encoding long chain acyl-CoA synthetase 4), a signature enzyme that not only predicts cellular sensitivity to ferroptosis but is also required for its execution⁷⁰, in the CD8–3 cluster (Figure 6A). At the protein level, we observed consistently higher expression of ACSL4 in the PD-1⁺SLAMF6^lowCD39⁺ CD8⁺ T cell population than in the naïve (PD-1⁻CD39⁻) and Tex^prog (PD-1⁺SLAMF6⁺CD39⁻) T cells from islets of 12-week-old female NOD mice (Figure 6B), suggesting a higher level of ferroptosis in the intra-islet Tex^int CD8⁺ T cells.

Figure 6. Ferroptosis is a contributing mechanism by which OxLDL exposure diminishes intra-islet CD8⁺ T cells.

(A) Violin plots depicting the expression of ferroptosis-related genes in CD8–3 versus other clusters.

(B) Flow cytometry analysis of ACSL4 expression in indicated intra-islet CD8⁺ T cell subsets.

(C-D) Flow cytometry analysis of intra-islet SLAMF6^lowCD39⁺ CD8⁺ T cells exposed to OxLDL in the presence of indicated inhibitors. (C) Schematics of the experimental design. (D) Representative FACS plots (left) and quantification (right).

Data are mean ± SEM. Each dot represents an animal from three independent experiments (B) or individual experiments using 6–8 mice per experiment (D). Statistics are analyzed by RM one-way ANOVA with Tukey’s multiple comparisons test (B) or Two-way ANOVA with Tukey’s multiple comparisons test (D). **P < 0.01; ***P < 0.001; ****P < 0.0001.

To functionally test the role of ferroptosis, we cultured islet infiltrates from 8–12-week-old NOD mice with OxLDL, while including the ferroptosis inhibitors Ferrostatin-1⁷¹ and Liproxstatin-1⁷², the pan-caspase apoptosis inhibitor Z-VAD-FMK, and the necrosis inhibitor Necrostatin-1 (Figure 6C). Treatment with OxLDL significantly reduced the proportion of the SLAMF6^lowCD39⁺ subset compared to media alone. However, the addition of either Ferrostatin-1 or Liproxstatin-1 to the culture with OxLDL largely restored the proportion of the SLAMF6^lowCD39⁺ subset to a level comparable to that observed in the media-only condition (Figure. 6D). Notably, this rescue was not observed when OxLDL was added along with either Z-VAD-FMK or Necrostatin-1 (Figure. 6D). Thus, ferroptosis represents an important mechanism underlying the reduction of intra-islet SLAMF6^lowCD39⁺ CD8⁺ T cells during increased OxLDL exposure.

PD-1 blockade rescues intra-islet TCF1⁻TOX⁺ CD8⁺ T cells and diabetes development in NOD.Cxcl16^−/− mice

We next sought to test whether rescuing intra-islet TCF1⁻TOX⁺ CD8⁺ T cells could reverse diabetes protection in NOD.Cxcl16^−/− mice. Anti-PD-1 antibody treatment in NOD mice leads to significant expansion of intra-islet TCF1⁻TOX⁺ CD8⁺ T cells⁵². Although this finding suggests PD-1 blockade as a potential approach, this expansion could be due to mobilized infiltration of peripheral CD8⁺ T cells, rather than a direct impact on intra-islet T cells. To determine the specificity of the anti-PD-1 treatment on intra-islet versus peripheral CD8⁺ T cells, we administered anti-PD-1 or control antibody to 12-week-old female NOD mice and adoptively transferred their splenocytes to NOD.Rag1^−/− hosts. Even though many of the anti-PD-1-treated donor mice were already diabetic (not shown), their splenocytes transferred diabetes with an incidence comparable to splenocytes from diabetes-free donors given the control antibody (Figure 7A). Additionally, we administered FTY720, an agonist of sphingosine 1-phosphate (S1P)⁷³, to 6-week-old female NOD mice, when the majority of the islets are free from infiltration, to block the trafficking of T cells from secondary lymphoid tissues to islets. FTY720 treatment substantially reduced T cells in the peripheral blood (Figure 7B) and prevented acute diabetes development induced by anti-PD-1 (Figure 7C). This phenomenon was distinct in 12-week-old female NOD mice, which already have evident islet infiltration. Although FTY720 significantly restricted circulating T cells in these mice (Figure 7D), it did not suppress diabetes development induced by anti-PD-1 (Figure 7E). Furthermore, we serendipitously found that the anti-CD8β antibody (clone 53–5.8), which only depleted CD8⁺ T cells in secondary lymphoid tissues but not in the islets (Figure 7F), was unable to protect NOD mice from diabetes induced by anti-PD-1 (Figure 7G). In contrast, the anti-CD8α antibody (clone YTS169.4) largely depleted intra-islet CD8⁺ T cells (Figure 7F) and abolished diabetes development (Figure 7G). Thus, the anti-PD-1 treatment specifically targets intra-islet CD8⁺ T cells with minimal impact on peripheral T cells.

Figure 7. PD-1 blockade rescues intra-islet TCF1⁻TOX⁺ CD8⁺ T cells and reverses diabetes resistance in NOD.Cxcl16^−/− mice.

(A) Schematics (upper) and diabetes incidence (lower) of NOD.Rag1^−/− recipients given splenocytes from donor NOD mice treated with anti-PD-1 or control antibody.

(B-E) FTY720 or control saline was administrated to female NOD mice at 6 (B, C) or 12 (D, E) weeks of age followed by PD-1 blockade. (B, D) Flow cytometry analysis of the frequency of T cells in peripheral blood. (C, E) Diabetes incidence post PD-1 blockade.

(F) Flow cytometry analysis showing CD8⁺ T cell depletion by the anti-CD8α or anti-CD8β antibody. The bar graphs summarize the ratio between CD8⁺ and CD4⁺ T cells in individual mice (each dot).

(G) Diabetes incidence in 8-week-old NOD female mice on day 20 post treatment with indicated antibodies.

(H) Diabetes incidence in female NOD and NOD.Cxcl16^−/− mice post PD-1 blockade.

(I) Representative FACS plots showing the intra-islet TCF1⁻TOX⁺ CD8⁺ T cells on day 7 post PD-1 blockade.

Data are mean ± SEM. The results are representative of two independent experiments (I) or pooled from at least 2–4 experiments (A-H). Statistics are analyzed by log-rank (Mantel-Cox) test (A, C, E, H) or Mann-Whitney test (F). **P < 0.01; ****P < 0.0001.

We then treated cohorts of age- and sex-matched NOD or NOD.Cxcl16^−/− mice with anti-PD-1 or control antibody. A majority (> 80%) of both the NOD and NOD.Cxcl16^−/− mice promptly became diabetic with comparable incidences (Figure 7H). Some of the NOD.Cxcl16^−/− mice showed slower disease onset (Figure 7H), possibly due to their lower levels of islet infiltration prior to the anti-PD-1 treatment. Additionally, on day 7 post anti-PD-1 treatment, the TCF1⁻TOX⁺ subset became a dominant population among the intra-islet CD8⁺ T cells in in both NOD and NOD.Cxcl16^−/− mice (Figure 7I). Thus, rescuing the effector CD8⁺ T cells specifically in the islets of NOD.Cxcl16^−/− mice overcomes their resistance to diabetes development.

DISCUSSION

Our study indicates that the tissue microenvironment plays an active role in fostering autoimmunity, rather than being a passive target. The incidence of T1D is among the highest in tissue-specific autoimmune diseases, while autoimmunity in the exocrine pancreas is much less common. Self-reactive T cells have been identified in individuals without autoimmune propensity^74–78. Immune checkpoint inhibitors can mobilize these anti-self T cells in cancer patients, triggering irAEs affecting tissues including pancreatic islets⁷⁹. However, cases of irAEs targeting the exocrine pancreas are rare². These findings implicate that the classic paradigm based on genetic predisposition may not fully explain the susceptibility of pancreatic islets to autoimmunity. In this study, we demonstrate that a proinflammatory resident macrophage may create a tissue microenvironment that fosters onward autoimmunity.

Dysregulation of lipid metabolism and accumulation of oxidized lipids in the islets plays an important role in the sequelae of both type 1 and type 2 diabetes^40–42. The intrinsic expression of CXCL16 may reflect a metabolic adaptation of the islet macrophages to the oxidative microenvironment. Functionally, we show that CXCL16 is crucial for islet macrophages to adequately scavenge OxLDL. In addition to CXCL16, islet macrophages may also use other mechanisms to scavenge OxLDL. It is possible that the weak expression of CD36, as well as macropinocytosis and efferocytosis⁴⁶, may also contribute to OxLDL clearance. These compensatory mechanisms may explain why NOD.Cxcl16^−/− mice maintain normal levels of insulin production before autoimmunity occurs, as a severe defect may lead to oxidative stress causing β-cell dysfunction.

Our analysis of islet-infiltrating CD8⁺ T cells bodes well with previous studies that have characterized the differentiation/exhaustion program of intra-islet CD8⁺ T cells^33,52,80,81. It is plausible that islets provide constant antigen exposure, appropriate APC network and cytokine milieu to support local CD8⁺ T cell differentiation. This program also involves communication with secondary lymphoid tissues. Elegant studies have uncovered a stem-like epigenetic feature that is retained by β-cell-specific CD8⁺ T cells in humans⁸². Moreover, a stem-like IGRP-specific CD8⁺ T cell population was identified in the pLN of NOD mice, which can migrate to the islets and differentiate into effector T cells⁸³. Thus, as one of the few spontaneous autoimmune models, the NOD strain represents a proper tool for studying CD8⁺ T cell exhaustion over a chronic autoimmune process.

Our data highlight key differences between autoimmune CD8⁺ T cell differentiation and canonical exhaustion. A recent fundamental study uncovered an exhausted-like program in autoimmune CD8⁺ T cells that is restrained by the expression of LAG3⁸⁰. Interestingly, autoimmune CD8⁺ T cells appear to retain effector/cytotoxic functions despite upregulating inhibitory molecules. In our analysis, the CD8–4 cluster exhibits the strongest expression of genes encoding inhibitory molecules without downregulating effector genes. These Tex^eff autoimmune cells may still drive β-cell killing, differing from dysfunctional, fully exhausted CD8⁺ T cells observed in chronic viral infection and cancer^{54,55,84–92}. Furthermore, we identified an oxidative stress-related signature in intra-islet Tex^int CD8⁺ T cells, which may reflect a tissue imprint of the oxidative islet microenvironment.

Finally, our study reveals an important role of oxidized lipids in the regulation of autoimmune CD8⁺ T cells. NOD mice lacking IL-27 or IL-27Rα are resistant to diabetes development⁹³. Moreover, CD8⁺ T cells from Il-27rα^–/– NOD mice mostly became SLAMF6⁺ progenitors but were largely unable to differentiate into the SLAMF6^– effectors⁸¹. In addition to these T-cell-intrinsic mechanisms, we demonstrate that excessive OxLDL exposure sufficiently disrupts intra-islet CD8⁺ T cell differentiation. Overall, our study provides new perspectives into the complex interplay between T-cell-intrinsic and extrinsic environmental factors in autoimmune CD8⁺ T cell differentiation and may have implications for developing targeted therapies for autoimmune diseases.

Limitations of the study

We observed considerable variation in OxLDL levels across individual islets, possibly reflecting the well-described heterogeneity in islet autoimmunity. Our attempts to correlate OxLDL levels to islet inflammation using confocal imaging have not been successful. Single-islet transcriptomic studies may yield useful information. We attempted to profile OxLDL content in islets using mass spectrometry but were hindered by limited sample materials. Although our data indicate that Cxcl16 deficiency primarily impairs OxLDL clearance by islet macrophages, further investigation would require conditional knockout models. Despite that islet macrophages highly express Lyz2, we encountered poor deletion efficiency using NOD.Lyz2-Cre mice. Due to low numbers of intra-islet Tex^int CD8⁺ T cells, we were limited for conducting functional experiments on oxidized lipids uptake (receptors involved) and oxidative signatures (expression of ROS). Lastly, our observation of CXCL16 expression by human islet macrophages suggests future studies examining the biological relevance of OxLDL clearance in human islets.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiaoxiao Wan (wanx@wustl.edu).

Materials availability

Mouse models and other reagent generated in this study will be available upon request.

Data and code availability

• Bulk RNA sequencing (GSE141786) and single-cell RNA sequencing (GSE262101) data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus database.

• This paper does not report an original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR METHODS

EXPERIMENTAL AND PARTICIPANT DETAILS

Mouse models

NOD/ShiLtJ (NOD), NOD.129S7(B6)-Rag1tm1Mom/J (NOD.Rag1^−/−), NOD.Tnfrsf1a/1b^−/−, NOD.Tcr NY8.3 (NOD.Cg-Tg(TcraTcrbNY8.3)1Pesa/DvsJ), C57BL/6, and B6.Cd36^−/− (B6.129S1-Cd36tm1Mfe/J), mice were originally obtained from The Jackson Laboratory. NOD.Cxcl16^−/−, NOD.Ccrl2^−/−, and NOD.Ms4a7^−/− mice were generated by CRISPR/Cas9 on the NOD background at the Genome Engineering Center at Washington University in St Louis. Each resulting founder line was genotyped by NGS, which confirmed the generation of mice carrying deletional mutations in one allele of Cxcl16 (a 13 bp deletion in exon 2), Ccrl2 (a 23 bp deletion in exon 3), and Ms4a7 (a 11 bp deletion in exon 3). Each founder line was crossed to WT NOD mice for two generations to generate heterozygous mice. Mice with homozygous deletion were generated by intercrossing the heterozygous mice. All mice were bred, maintained, and experimented in our pathogen-free animal facility in accordance with the Division of Comparative Medicine of Washington University School of Medicine (Association for Assessment and Accreditation of Laboratory Animal Care accreditation no. A3381–01).

Islet isolation

Islets were harvested as described previously¹⁷. Briefly, following clamping of the bile duct leading to the duodenum, 5 ml of type XI collagenase (0.4 mg/ ml; Sigma Aldrich) in isolation buffer (1x HBSS, 10 mM HEPES, 1 mM MgCl₂; pH 7.4) was injected through the common bile duct to perfuse the pancreas. The inflated pancreas was carefully removed and digested at 37°C for 12-14 min. Crude islets were collected and washed several times with the wash buffer (1x HBSS, 10 mM HEPES, 1 mM MgCl_2, 1 mM CaCl₂; pH 7.4). Islets were hand-picked under a dissection microscope. Purified islets were then dispersed into single cell suspension using a non-enzymatic cell dissociation solution (Sigma Aldrich) for 3 min at 37°C.

Antibodies

The following fluorescently conjugated antibodies were purchased from BioLegend: anti-mouse CXCL16 (12–81), anti-human CXCL16 (22-19-12), anti-PD1 (29F.1A12), anti-B220 (RA3–6B2), anti-CD11c (N418), anti-CD4 (RM4–5), anti-CD45 (30-F11), anti-F4/80 (BM8), anti-CD3 (145–2C11), anti-SIRPα (P84), anti-XCR1 (ZET), anti-human CD45 (2D1), anti-human CD11b (ICRF44), and anti-human CD11c (Bu15). Anti-CD8 (53–6.7), anti-CD39 (24DMS1), anti-Foxp3(FJK-16s) and anti-OxLDL (BS-1698R) were purchased from eBioscience. Anti-TCF-7/TCF1 (S33–966) and anti-Ly108 (SLAMF6) (13G3) were purchased from BD Biosciences. Anti-TOX (REA473) was purchased from Cell signaling technology.

Flow cytometry

Single cell suspension of islets was incubated in a buffer containing the FcR blocking antibody (2.4G2) for 10 min at 4°C. Surface staining was performed with fluorescently labeled antibodies (1:200 [vol/vol]) by incubating at 4°C for 30 min. Analysis of CD8⁺ T cells specific to IGRP_206–214 was performed using the MHC-I (K^d) tetramer loaded with the IGRP_206–214 peptide⁵⁶. Briefly, islet cells were labeled with fluorescently conjugated tetramer for 20 min at room temperature. Cells were then stained with the surface markers. For intracellular CXCL16 staining, cells were incubated (DMEM, 10% FBS) at 37°C with brefeldin A for 3 h. For intracellular cytokine (TNFα and Pro-IL-1β) staining, cells were incubated with brefeldin A in DMEM with 10% FBS for 4h at 37°C in the absence of stimulation. Following incubation cells were washed twice and surface staining was performed as described above. Cells were fixed (Cytofix/Cytoperm; BD Biosciences) for 30 min at 4°C and then washed twice with the perm buffer (Cytofix/Cytoperm; BD Biosciences). Following fixation and permeabilization, cells were incubated with the anti-CXCL16 antibody for 30 min at 4°C. Staining of transcription factors TCF1, TOX and FOXP3 was performed using Foxp3/transcription factor staining buffer set (Thermo Fisher Scientific) following the manufacturer’s instructions. For intracellular OxLDL staining, cells were fixed after surface staining with 4% paraformaldehyde (Sigma Aldrich) in 1X PBS for 15 min on ice. Cells were washed with 1X PBS and 1% BSA and permeabilized with 2% saponin (Sigma-Aldrich) in 1X PBS for 5 min at 4°C, and then incubated with the anti-OxLDL antibody for 20 min at 4°C. Samples were examined using BD FACSCanto II Cell Analyzer (BD Biosciences) and the sorting was performed on FACSAria II (BD Biosciences). Data were analyzed using FlowJo 10.8.0 software (TreeStar). The adjusted MFI was calculated by subtracting the autofluorescence background.

Bulk RNA sequencing analysis

Islets were isolated from 2-, 4- and 12-week-old female WT NOD or C57BL/6J mice. Four biological replicates per condition were used. Each biological replicate contains ~1,500 islet macrophages (CD45⁺CD31⁻B220⁻CD90⁻CD11c⁺F4/80⁺) FACS-sorted from 6–8 mice. mRNA was isolated with RNAqueous-Micro Kit (Thermo Fisher Scientific), and cDNA was prepared using SeqPlex RNA Amplification kit (Sigma-Aldrich). Illumina NovaSeq-2500 was used for sequencing. Libraries were prepared and sequenced at the Genome Technology Access Center (Washington University in St. Louis, MO). Raw reads were aligned with the STAR aligner, counts were generated with htseq-count utility from the HTSeq Python library, and differential gene expression was performed using DESeq2 R package with Benjamini⁻Hochberg adjusted P = 0.05 as a threshold. The global gene expression level was determined by reads per kilobase of transcript per million reads mapped (RPKM). DEG and GSEA analysis was performed using Phantasus v1.21.5 provided by the Artyomov lab (Washington University School of Medicine; https://artyomovlab.wustl.edu/phantasus/).

scRNA-seq library preparation

Intra-islet live CD45⁺CD31⁻ cells were sorted using islets from 12-week-old female NOD and NOD.Cxcl16^−/− mice. Libraries for single-cell gene expression and sequencing were prepared at the McDonnell Genome Institute (MGI) at Washington University using the 3v3.1 methods. Specifically, cDNA was prepared after the GEM generation and barcoding, followed by the GEM-RT reaction and bead cleanup steps. Purified cDNA was amplified for 11–13 cycles before being cleaned up using SPRIselect beads. Samples were then run on a Bioanalyzer to determine the cDNA concentration. GEX libraries were prepared as recommended by the 10x Genomics Chromium Single Cell 3’ Reagent Kits User Guide (v3.1 Chemistry Dual Index) with appropriate modifications to the PCR cycles based on the calculated cDNA concentration. For sample preparation on the 10x Genomics platform, the Chromium Next GEM Single Cell 3’ Kit v3.1, 16 rxns (PN-1000268), Chromium Next GEM Chip G Single Cell Kit, 48 rxns (PN-1000120), and Dual Index Kit TT Set A, 96 rxns (PN-1000215) were used. The concentration of each library was accurately determined through qPCR utilizing the KAPA library Quantification Kit according to the manufacturer’s protocol (KAPA Biosystems/Roche) to produce cluster counts appropriate for the Illumina NovaSeq6000 instrument. Normalized libraries were sequenced on a NovaSeq6000 S4 Flow Cell using the XP workflow and a 50×10×16×150 sequencing recipe according to manufacturer protocol. A median sequencing depth of 50,000 reads/cell was targeted for each Gene Expression Library.

scRNA-seq data analysis

The NOD and NOD.Cxcl16^−/− libraries were counted using Cell Ranger (v7.0.1). Low quality UMIs were filtered and mapped to the mouse genome (mm10). Both datasets were further aggregated using the Cell Ranger aggr pipeline (v7.0.1). The Cell Ranger aggr pipeline automatically equalizes the average read depth between samples. The gene expression from both datasets was filtered, normalized and clustered, and the resulting Cloupe file was created and imported in Loupe Browser for further analyses and visualization. To filter out multiplets, low quality cells, and empty droplets, filtering was performed as follows: Features were normalized to include log2 counts between 10–12, UMI included log2 count in the range of 11.5–14, mitochondrial genes <5%, and PTPRC log2 expression >1. This removed 52% of barcodes. PCA was performed to reduce the dimensionality of the dataset, and the principal components were visualized by tSNE or UMAP plots. The differentially expressed genes between clusters or libraries were identified using default algorithms. Bonferroni-adjusted p-values were used to determine significance at an FDR<0.05.

Diabetes monitoring

All the female and male mice of each genotype produced in the animal facility over a year were included for continuous monitoring until 40 weeks of age. Urine glucose levels in mice were checked weekly (AimStrip US-G; Germaine Laboratories). For all the anti-PD-1 antibody treated mice, urine glucose levels were monitored daily until day 20 or 30 following treatment. Mice were considered diabetic after two consecutive readings of >250 mg/dl.

Glucose tolerance test

NOD or NOD.Cxcl16^−/− male mice (4-week-old) were fasted for 12 h and then injected with D-glucose (2g/kg body weight). Blood glucose was measured using Quintet AC Glucometer at 0, 15-, 30-, 60- and 90-min post glucose challenge.

Serum ELISA

For serum insulin ELISA, blood samples were collected ad libitum, fasting and 1 h post glucose challenge from NOD WT and NOD.Cxcl16^−/− mice. Serum insulin ELISA was performed as described previously⁹⁴. Briefly, the mouse anti-insulin capture antibody (clone 3A6; Novus Biologicals) was coated on a 96 well plate and incubated overnight at 4°C. Plates were washed and then blocked (5% FBS) at RT for 1 h. Serum samples were added following blocking for 2 h at 37°C. Plates were washed and detection antibody mouse anti-insulin antibody (1 mg/mL) (clone 8E2- HRP; Novus Biologicals) was added for 2 h at 4°C. OptEIA TMB substrate (BD Biosciences) was used to visualize the antibody binding. The reaction was stopped using a diluted phosphoric acid, and absorbance was read at 450 nm. CXCL16 ELISA was performed using Mouse CXCL16 ELISA kit (Invitrogen EMCXCL16) according to the manufacturer’s protocol.

Human pancreatic islets

De-identified human primary islets isolated from deceased donors were purchased from Prodo Laboratories (Aliso Viejo, CA). Three independent experiments were performed using islets from three donors: donor 1 (Male, 26 years, BMI 21.35), donor 2 (Male, 44 years, BMI 25.7), donor 3 (male, 32 years, BMI 26.9). The purity of the islets was between 85 and 98%. The islets were cultured overnight in CMRL medium supplemented with 10% FBS for recovery. Following incubation, islets were hand-picked and dispersed using non enzymatic cell dissociation solution (Sigma Aldrich) for 3 min at 37°C. Intracellular and surface staining of CXCL16 was performed as described above.

Isolation of tissue-resident macrophages

For adipose tissue macrophage, visceral adipose tissues were finely minced and digested in Collagenase IV buffer (Collagenase IV (conc); 1% BSA in 1XPBS) for 20 min at 37°C with shaking at 200 rpm in an orbital shaker. Digested tissue was filtered through the 70-μm filter and centrifuged at 500 g for 10 min. Floating adipocytes were removed, and the pellet containing stromal vascular fraction (SVF) was resuspended with ACK lysis buffer to remove RBCs. Single cell suspension of SVF was then washed and used for flowcytometry. For lung macrophage isolation, lungs were perfused by injecting 10 ml of ice-cold PBS into the right ventricle. Lungs were then harvested, finely minced, and digested using Collagenase D (1.5 mg/ml) and DNase (50 μg/ml) in PBS for 30 min at 37°C in an orbital shaker. Digested tissue was filtered through a 70-μm filter, RBCs were removed using ACK lysis buffer and the single cell suspension was washed and used for flow cytometry. Peritoneal macrophages were analyzed from the peritoneal exudate cells collected using peritoneal lavage with 10 ml of ice-cold PBS.

Antigen presentation assay

For antigen presentation assay, islet cells from 2–3 WT NOD or NOD.Cxcl16^−/− mice were pooled together. In a 96-well plate, 5×10⁵ islet cells were cocultured with 5×10⁵ CD4⁺ T cell hybridoma (clone IIT-3) in DMEM media containing 10% FBS. In some wells, 10 μM insulin B-chain 9–23 peptide was added for probing presentation with exogenous insulin peptide pulse. Culture supernatants were collected after 24 h and the levels of IL-2 in the supernatant were measured by culturing with CTLL-2, an IL-2⁻dependent cell line. The proliferation of CTLL-2 was determined by the incorporation of ₃[H] thymidine.

In vivo treatment

For PD-1 checkpoint blockade, NOD or NOD.Cxcl16^−/− mice were injected i.p. with three doses of 250 μg of the anti-PD-1 (RMP1–14) or the isotype control Rat IgG2a antibody on days 0, 3 and 6 in sterile 1X PBS. For impairing Treg function, female WT NOD or NOD.Cxcl16^−/− mice were given two injections (500 μg per injection) of rat anti-mouse CD25 monoclonal antibody (clone PC61) on day 14 and 21 of age by i.p. For CD8⁺ T cell depletion, the anti-CD8α (YTS169.4) or anti-CD8β (53–5.8) antibody (250 μg per injection) was injected i.p. three days before the first injection of anti-PD-1 and during the 3 doses of anti-PD-1 injections. The percentage of CD8⁺ T cells post depletion was examined by flow cytometry using a different anti-CD8α (53–6.7) antibody clone. All the antibodies used for treatment were purchased from Leinco Technologies. For FTY720 treatment, mice were treated with FTY720 (100 μg; Sigma-Aldrich) or saline intravenously starting a day before anti-PD-1 antibody injection and the treatment was continued until the end of the monitoring period for 20 days.

Confocal imaging of OxLDL in intact islets

Islets were isolated and fixed in 2% PFA for 15 min at room temperature. After fixation, unsectioned intact islets were blocked and permeabilized with a PBS-based buffer containing 0.1% BSA and 0.1%Tritonx100 for 30 min at room temperature. Following washing, primary staining with OxLDL antibody (1:100 in PBS containing 0.1% BSA /0.1% Tritonx100) was performed at room temperature for 2 h. The islets were washed and a secondary antibody (anti Rabbit Alexa Fluor 555; 1:500) was added overnight at 4°C. Islets were stained with anti-mouse CD45 Alexa Fluor 647 antibody (1:100) overnight at 4°C. Following washing, Hoechst (1:500) was added for 15 min. Washed islets were then dehydrated using ethanol (sequentially 50%/70%/100%) and methyl salicylate for 5 min. Islets were washed and loaded onto the cavity slide, covered with a coverslip. Images of entire islets were taken on a Leica TCS SP8 X White Light Laser Confocal Microscope using a 63x objective with a NA of 1.4. 2D snapshots were acquired using Imaris image analysis software. The intensity of OxLDL per islet area was quantified using the ImageJ software after correction by subtracting the autofluorescent background in the islet area from the same channel.

Bone marrow chimera

Bone marrow cells were isolated from the femur and tibia of NOD and NOD.Cxcl16^−/− donor mice. After RBC lysis of bone marrow cells, 15×10⁶ cells were adoptively transferred to lethally irradiated (1000 rads; split dose) NOD and NOD.Cxcl16^−/− recipient mice. Eight weeks post BM transplant, islets were isolated and examined for accumulation of OxLDL using confocal microscopy as described above.

Adoptive transfer

Single cell suspensions from spleens were prepared after red blood cell lysis and were resuspended in sterile 1X PBS. For transfer of spleen cells from 12-week-old NOD and NOD.Cxcl16^−/− female mice, a total of 1.5×10⁷ cells were injected i.v. into each NOD.Rag1^−/− recipient. For transfer of spleen cells from NOD mice with anti-PD-1 or isotype control antibody treatment, a total of 10⁷ spleen cells were transferred i.v. to each NOD.Rag1^−/− recipient.

High fat diet feeding

Female NOD or NOD.CXCL16^−/− mice were fed high fat diet (Research Diets Inc. # D12492; irradiated, 60 kcal% Fat) or standard chow (SFD) ad libitum beginning at 4 weeks of age. Body weight was monitored weekly until 12 weeks of age. Islet analysis for immune cell infiltration and CD8⁺ T cell subsets was performed at 12 weeks of age.

Lipid peroxidation assay

For determining lipid peroxidation in the islet resident macrophage, dispersed islet cells were first stained with the surface markers. Following surface staining, cells were washed twice with warm 1XPBS and then incubated in 5 μM BODIPY 581/591 C11 reagent (ThermoFisher) in PBS for 30 min at 37°C. Cells were washed twice and analyzed by flow cytometry.

In vitro uptake of OxLDL

OxLDL uptake was measured in islet resident and peritoneal macrophages obtained from dispersed islet cells and peritoneal exudate respectively. Cells were incubated with OxLDL-DyLight^™-488 (Cayman Chemical) in PBS (1:20 diluted) for 30 min at 37°C. Following incubation cells were washed, surface stained, and the samples were analyzed by flowcytometry.

Ex vivo culture with exogenous lipids

Islets isolated from WT NOD mice were pooled and non-enzymatically dispersed. Islet cells were cultured in DMEM containing 10% FBS and treated with OxLDL (50 μg/ml) and LDL (50 μg/ml) for 24h. For the experiment involving various cell death pathway inhibitors including ferroptosis inhibitors, Ferrostatin-1(10 μM) and liproxstatin-1 (1 μM), the apoptosis inhibitor Z-VAD-FMK (100 μM), and the necroptosis inhibitor, Nec-1 (100 μM) were added with OxLDL (50 μg/ml) for 24 h. Following incubation, cells were washed, and CD8⁺ T cell subsets were analyzed by flowcytometry.

Islet macrophage-T-cell co-culture

Dispersed islet cells were prepared from pooled 4-week-old male NOD or NOD.Cxcl16^−/− mice and were plated in tissue culture plates in DMEM containing 10% fetal bovine serum. The cells were cultured for 24 h to allow the islet macrophages to adhere. The culture supernatant containing floating cells was removed to enrich islet macrophages¹¹. Intra-islet CD8⁺ T cells sourced from heavily inflamed islets isolated from 12–16-week-old female NOD mice, when T cell infiltration dominates islet cell populations, were added to NOD and NOD.Cxcl16^−/− islet macrophages in the presence of OxLDL (50 μg/ml). Following 24 h of co-culture, CD8⁺ T cell subset was analyzed by flow cytometry.

Generation of Hoxb8-Cxcl16 expressing cell line

The Hoxb8 immortalized monocytic cell line was subjected to lentiviral transduction to exogenously express murine Cxcl16. Briefly, the lentiviral plasmid backbone (Addgene 91798) was used to introduce the Cxcl16 cDNA sequence under the expression of the hPGK promoter. First, the plasmid was digested with MluI-HF (NEB Cat# R3198) and NheI-HF (NEB Cat# R3131), and then purified from the agarose gel with the gel isolation kit (NEB Cat# T1010). Next, the linearized plasmid and the synthesized Cxcl16 cDNA were subjected to the In-Fusion reaction with the In-Fusion cloning kit (Takara Cat# 638944) according to the manufacturer’s instructions. Subsequently, the plasmid was packaged in HEK293T using TransIT-Lenti (#MIR 6600) and transfecting the packaging pSPAX (Addgene #12260) and the envelope pMD2.G (Addgene #12259) plasmids. At 48 h, the lentivirus-containing media were collected, and spinoculation was performed at 800 xg for 90 min using undifferentiated Hoxb8 cells. The following day, the transduced Hoxb8 cells were subjected to puromycin (Cat# A1113803) antibiotic pressure for 7 days before assessing the expression of CXCL16 protein by flow cytometry. On day 7, Hoxb8-Ctrl and Hoxb8-Cxcl16 differentiated macrophages were plated in a 96-well plate (1×10⁴ cells/well). The supernatant was removed the next day, and infiltrated islet cells isolated from 12–16 weeks old female NOD mice were added to the Hoxb8-Ctrl and Hoxb8-Cxcl16 macrophages in the presence of OxLDL (50 μg/ml). After 24 h of co-culture, the CD8⁺ T cell subset was analyzed by flow cytometry.

In vitro T cell culture

Splenocytes isolated from NOD.Tcr NY8.3 mice were activated in DMEM containing 10% fetal bovine serum, 20 U/ml IL-2 with the indicated dose of IGRP_206–214 peptide for 48 h. The activated CD8⁺ T cells were then exposed to OxLDL (50 μg/ml) for 24 h. For T cell proliferation experiments, naïve NOD.Tcr NY8.3 splenocytes were incubated with 5 μM CFSE (CFSE Cell Division Tracker Kit; Biolegend) for 20 min at 37°C prior to in vitro activation. For the detection of intracellular cytokines, activated CD8⁺ T cells were restimulated with PMA (50 ng/ml) and ionomycin (0.5 μg/ml) in the presence of Brefeldin A (1 mg/ml) for 6h. The levels of TNFα and IFNγ in activated CD8⁺ T cells were measured by intracellular flow cytometry as described above. For the experiments involving the culture of NOD.Tcr NY8.3 splenocytes with the soluble CXCL16 chemokine, cells were first labeled with CFSE (5 μM). Following labeling, cells were washed and cultured in DMEM containing 10% fetal bovine serum, 20 U/ml IL-2 with the indicated dose of IGRP_206–214 peptide in the presence of recombinant murine CXCL16 (50 ng/ml; Peprotech). After 48 h, CD8⁺ T cell activation and proliferation were assessed by flow cytometry.

Histologic analysis

For histologic analysis, pancreata and submandibular glands were isolated from 12–16-week female old NOD or NOD.Cxcl16^−/− mice, fixed in 10% paraformaldehyde. After fixation, samples were embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Microscopy imaging of stained sections was performed using an Eclipse E800 microscope (Nikon) equipped with CFI Plan Apo Lambda DM 20× air objective, X-Cite 120PC light source (Excelitas Technologies), EXi blue fluorescence. microscopy camera, and QCapture 64-bit v2.9.13 acquisition software. (QImaging). Insulitis scoring was performed on individual islets from at least three non-serial pancreatic sections per mouse. The area of autoimmune sialadenitis foci was calculated by the ImageJ software in full-print submandibular gland sections from individual mice. For each mouse, at least 15 foci were examined from a section; the cumulative area of lymphocytic infiltration foci was calculated from at least non-serial sections per mouse.

RNA isolation and real time PCR

RNA from the sorted cells was isolated using the Ambion RNAqueous- Micro kit (Thermo Fisher Scientific) following the manufacturer’s instructions. The quantification of RNA was performed through OD260 measurement using NanoDrop (Thermo Fisher Scientific). Subsequently, cDNA was synthesized using iScript Reverse Transcription Supermix for qRT-PCR (Bio-Rad Laboratories). qRT-PCR was performed using the 5′ nuclease (TaqMan) chemistry using iTaq Universal Probes Supermix (Bio-Rad Laboratories). The primers and probes were designed and procured from IDT. All quantitative PCRs were carried out on a StepOnePlus Real-Time PCR system (Thermo Fisher Scientific) utilizing StepOne 2.3 Software (Thermo Fisher Scientific).

Statistical analysis

For all the experiments age- and gender-matched mice were used. The Mann-Whitney test was used to determine the statistical significance in unpaired biological replicates between two experimental groups. The Paired Student t test was used to determine the statistical significance in paired biological replicates between two experimental groups. One-way ANOVA with Dunnett’s multiple comparisons test was used to determine the statistical significance in unmatched samples of three or more groups. RM one-way ANOVA with Tukey’s multiple comparisons test was used to determine the statistical significance in matched/paired samples of three or more groups. Two-way ANOVA with Tukey’s multiple comparisons test was used to determine the statistical significance in samples of three or more groups, when two factors may affect the results. Log-rank (Mantel-Cox) test was performed to compare the diabetes incidence in mice. Error bars in all the data represent SEM, and the P values were calculated using GraphPad PRISM 10.1.1

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mouse CXCL16 PE	BD Pharmingen	Cat# 566740; RRID: AB_2869842
Anti-human CXCL16 PE	Biolegend	Cat# 360803; RRID: AB_2563010
Anti-mouse PD1 BV421	Biolegend	Cat# 135221; RRID: AB_2562568
Anti-mouse/human B220 PerCP	Biolegend	Cat#103234; RRID: AB_893353
Anti-mouse CD11c PEcy7	Biolegend	Cat# 117318; RRID: AB_493568
Anti-mouse CD4 PEcy7	Biolegend	Cat# 100528; RRID: AB_312729
Anti-mouse CD4 Pacific Blue	Biolegend	Cat# 100531; RRID: AB_493374
Anti-mouse CD45 BV510	Biolegend	Cat# 103138; RRID: AB_2563061
Anti-mouse F4/80 APC	Biolegend	Cat# 123116; RRID: AB_893481
Anti-mouse/rat XCR1 APC	Biolegend	Cat# 148206; RRID: AB_2563932
Anti-mouse CD3ε FITC	Biolegend	Cat# 100306; RRID: AB_312671
Anti-mouse CD172a (SIRPa) Alexa Fluor® 488	Biolegend	Cat# 144024; RRID: AB_2650815
Anti-mouse Foxp3 PE	eBioscience	Cat# 12-5773-82; RRID: AB_465936
Anti-mouse CD11b PECy7	Biolegend	Cat# 101216; RRID: AB_312799
Anti-mouse TCF-7/TCF-1 PE	BD Pharmingen	Cat# 564217; RRID: AB_2687845
Anti-mouse CD39 PE	eBioscience	Cat# 12-0391-82; RRID: AB_1210740
Anti-mouse/human OxLDL	eBioscience	Cat# BS-1698R;
Anti-Mouse Ly-108 AF647	BD Pharmingen	Cat# 561547; RRID: AB_10712759
Anti-mouse F4/80 PerCP	Biolegend	Cat# 123126; RRID: AB_893483
Anti-mouse CD19 APCcy7	Biolegend	Cat# 115530; RRID: AB_830707
Anti-mouse CD11b APCcy7	Biolegend	Cat# 101226; RRID: AB_830642
Anti-mouse CD11c APCcy7	Biolegend	Cat# 117324; RRID: AB_830649
Anti-mouse CD44 BV421	Biolegend	Cat# 103040; RRID: AB_2616903
Anti-mouse CD3 PEcy7	Biolegend	Cat# 100220; RRID: AB_1732057
Anti-mouse Thy1.2 FITC	Biolegend	Cat# 140304; RRID: AB_10642812
Anti-mouse CD11b APC	Biolegend	Cat# 101212; RRID: AB_312795
Anti-mouse/human Tox	Miltenyi Biotec	Cat# 130-126-455; RRID: AB_2801785
Anti-mouse PDL1 APC	Biolegend	Cat# 124312; RRID: AB_10612741
Anti-mouse CD80 FITC	Biolegend	Cat# 104706; RRID: AB_313127
Anti-mouse CD40 PE	BD Pharmingen	Cat# 553791; RRID: AB_395055
Anti-mouse I-Ag7 PB	inhouse
Anti-human CD11c Percp	eBioscience	Cat# 46-0116-42; RRID: AB_10596368
Anti-human CD11b BV421	Biolegend	Cat# 301324; RRID: AB_11219589
Anti-human CD45 BV510	Biolegend	Cat# 368525; RRID: AB_2687376
Anti-mouse CD36 PE	BD Pharmingen	Cat# 562702; RRID: AB_2737732
Anti-mouse F4/80 FITC	Biolegend	Cat# 123108; RRID: AB_893502
Anti-mouse IFN-γ APC	Biolegend	Cat# 505810; RRID: AB_315404
Anti-mouse TNF-α FITC	Biolegend	Cat# 506304; RRID: AB_315425
IL-1 beta (Pro-form) PerCP-eFluor710	eBioscience	Cat# 46-7114-82; RRID: AB_2573835
Anti-mouse CD8 PerCP-eFluor™ 710	eBioscience	Cat# 46-0081-82; RRID: AB_1834433
Anti-mouse CXCR6 APC	eBioscience	Cat# 17-9186-82; RRID: AB_2734900
Anti-mouse CD103 PE	Biolegend	Cat# 121406; RRID: AB_1133989
Anti-mouse Ly6c V450	BD Pharmingen	Cat# 560594; RRID: AB_1727559
Anti-Mouse CD8 (Clone YTS 169.4) – Purified	Leinco Technologies	Cat# C2850; RRID: AB_2829606
Anti-Mouse CD8b.2 - Purified (Clone 53–5.8)	Leinco Technologies	Cat# C2836; RRID: AB_2829594
Anti-Mouse PD-1(Clone RMP1–14) – Purified	Leinco Technologies	Cat# P372; RRID: AB_2749820
Anti-Mouse CD25 - Purified (Clone PC61)	Leinco Technologies	Cat# C1194; RRID: AB_2737451
Biological samples
Human primary islets	Prodo Laboratories	n/a
Chemicals, peptides, and recombinant proteins
OxLDL	Kalenbiomed	Cat# 770252–7
LDL	Kalenbiomed	Cat# 770200–4
HDL	Kalenbiomed	Cat# 770300–4
Ferrostatin-1	Sigma-Aldrich	Cat# SML0583
Z-VAD-FMK	Tocris	Cat# 2163
Necrostatin-1	Tocris	Cat# 2324
Liproxstatin-1	Apexbio	Cat# B4987
Collagenase Type XI	Sigma-Aldrich	Cat# C7657
Collagenase Type IV	Sigma-Aldrich	Cat# C5138
Collagenase D	Roche	Cat# 45–11088866001
7-AAD Viability Staining Solution	Biolegend	Cat# 420403
IGRP 206–214 Peptide	Peptide 2.0 Inc.	Custom made synthetic oligos
InsB 9–23 Peptide	Peptide 2.0 Inc	Custom made synthetic oligos
PMA	Sigma	Cat# P8139
Ionomycin	Sigma	Cat# 10634
iScript Reverse Transcription Supermix for RT-qPCR	Biorad	Cat# 1708840
iTaq™ Universal Probes Supermix	Biorad	Cat# 1725131
GolgiPlug	BD Bioscience	Cat# 555029
Recombinant Murine CXCL16	Peprotech	Cat# 250–28
PBS	Gibco	Cat#10010023
Puromycin	Gibco	Cat# A1113803
MluI-HF	NEB	Cat# R3198
NheI-HF	NEB	Cat# R3131
Fetal Bovine Serum	Gibco	Cat# 16000044
BSA	Sigma	Cat# A2153
Tritonx100	Sigma	Cat# T8787
FTY720	Sigma Aldrich	Cat# SML0700
DMEM	Gibco	Cat# 12100–061
Hoechst 33258	Invitrogen	Cat# H3569
3[H] thymidine	Perkin Elmer	Cat# NET027A005MC
Cell Dissociation Solution Non-enzymatic	Sigma Aldrich	Cat# C5914–100ML
OptEIA TMB substrate	BD Biosciences	Cat# 555214; RRID: AB_2869044
D Glucose	Sigma Aldrich	Cat# G7021
Saponin	Sigma Aldrich	Cat# 47036
Paraformaldehyde	Thermo Fisher	Cat# 043368–9M
BD Cytofix/Cytoperm™ Fixation/Permeabilization Kit	BD Bioscience	Cat#554714
eBioscience Foxp3/Transcription factor staining kit	eBioscience	Cat# 00-5523-00
BODIPY 581/591 C11 reagent	Invitrogen	Cat# D3861
TransIT-Lenti Transfection Reagent	Mirus Bio	Cat# MIR 6600
eBioscience™ Fixable Viability Dye eFIuor™ 780	Invitrogen	Cat# 65-0865-14
Critical commercial assays
Oxidized LDL Uptake Assay Kit	Cayman Chemical	Cat# 601180
RNAqueous-Micro Total RNA Isolation Kit	Invitrogen	Cat# AM1931
CFSE Cell Division Tracker Kit	Biolegend	Cat# 423801
CXCL16 Mouse ELISA Kit	Invitrogen	Cat# EMCXCL16
Monarch® Plasmid DNA Miniprep Kit	NEB	Cat# T1010
In-Fusion cloning kit	Takara	Cat# 638944
SeqPlex RNA Amplification Kit	Sigma Aldrich	Cat#SEQR
Deposited data
Bulk RNA sequencing data (Islet Macrophage NOD, B6)	Zakharov et al.33	GEO: GSE141786
Single-cell RNA sequencing data (NOD & NOD.Cxcl16−/− CD45 cells)	This paper	GEO: GSE262101
Gene expression of CD8+ T cell subsets from chronic LCMV infection	Hudson et al.54	BioProject: PRJNA497086
Experimental models: Cell lines
HEK 293T	ATCC	Cat# CRL-3216
IIT3 T cell hybridoma	In house	n/a
CTLL-2 cytotoxic T cell line	ATCC	TIB-214
Experimental models: Organisms/strains
Mouse: NOD/ShiLtJ	The Jackson Laboratory	JAX# 001976
Mouse: NOD.129S7(B6)-Rag1tm1Mom/J	The Jackson Laboratory	JAX# 003729
Mouse: NOD.Tnfrsf1a/1b−/−	The Jackson Laboratory	JAX# 024314
Mouse: NOD.Tcr NY8.3	The Jackson Laboratory	JAX# 005868
Mouse: C57BL/6	The Jackson Laboratory	JAX# 000664
Mouse: B6.Cd36−/−	The Jackson Laboratory	JAX# 019006
Mouse: NOD.Cxcl16−/−	This Paper	n/a
Mouse: NOD.Ccrl2−/−	This Paper	n/a
Mouse: NOD.Ms4a7−/−	This Paper	n/a
Oligonucleotides
Actb	IDT	Primer1:GACTCATCGTACTCCTGCTTG Primer2:GATTACTGCTCTGGCTCCTAG Probe:/56FAM/CTGGCCTCA/ZEN/CTGT CCACCTTCC/3IABkFQ/
ILb	IDT	Primer1:CTCTTGTTGATGTGCTGCTG Primer2:GACCTGTTCI IIGAAGTTGACG Probe: /56- FAM/TTCCAAACC/ZEN/TTTGACCTGG GCTGT/3IABkFQ/
Cxcl16	IDT	Primer1:TTCCCATGACCAGTTCCAC Primer2:ATCAGGTTCCAGTTGCAGTC Probe:/56- FAM/TCTTGGCTT/ZEN/CCCCCACACA CG/3IABkFQ/
Recombinant DNA
pLEX_305-C-dTAG	Addgene	Cat# 91798
psPAX2	Addgene	Cat# 12260
pMD2.G	Addgene	Cat# 12259
Software and algorithms
FlowJo 10.8.0	BD Bioscience	https://www.flowjo.com/
Graphpad Prism v10.2.0	Graphpad	https://www.graphpad.com/
ImageJ	NIH, Bethesda, MD	https://imagej.nih.gov/ij/
Phantasus v1.21.5	Artyomov Lab	https://artyomovlab.wustl.edu/phantasus/
BioRender	Science Suite Inc.	https://www.biorender.com/
Loupe browser 6	10x genomics	https://www.10xgenomics.com/support/software/loupe-browser/latest
Imaris	Oxford Instruments	https://imaris.oxinst.com/
CellRanger version 7.0.1	10x genomics	https://www.10xgenomics.com/support/software/cell-ranger/latest
Other
Quintet AC Glucometer	McKesson Brand	Cat# 5055
Quintet AC Blood Glucose Test Strips,	McKesson QUINTET AC	Cat#5059
AimStrip® US-G, 100/bottle	Germaine Laboratories	Cat# 50100
High fat diet	Research Diets Inc.	Cat# D12492

Highlights.

• CXCL16 in islet resident macrophages is crucial for OxLDL clearance in islets

• Cxcl16 deletion in NOD mice suppresses T1D via an islet-specific mechanism

• OxLDL exposure diminishes intra-islet transitory CD8⁺ T cells by ferroptosis

• PD-1 blockade rescues transitory CD8⁺ T cells and reverses diabetes resistance