Lymph node sharing between pancreas, gut and liver leads to immune crosstalk and regulation of pancreatic autoimmunity

Summary

Lymph nodes (LNs) are critical sites for shaping tissue-specific adaptive immunity. However, the impact of LN sharing between multiple organs on such tailoring is less understood. Here we describe the drainage hierarchy of the pancreas, liver, and the upper small intestine (duodenum), into three murine LNs. Migratory dendritic cells (migDCs), key in instructing adaptive immune outcome, exhibited stronger pro-inflammatory signatures when originating from pancreas or liver than from duodenum. Qualitatively different migDC mixing in each shared LN influenced pancreatic β cell-reactive T cells to acquire gut homing and tolerogenic phenotypes proportional to duodenal co-drainage. However, duodenal viral infections rendered non-intestinal migDCs and β cell-reactive T cells more pro-inflammatory in all shared LNs, resulting in elevated pancreatic islet lymphocyte infiltration. Our study uncovers immune crosstalk through LN co-drainage as a powerful force regulating pancreatic autoimmunity.

Introduction

Development of tissue-specific immunity is facilitated by the existence of strategically placed lymph nodes (LNs) throughout the body. Separation of immune priming sites permits potentially antagonistic immune decisions to be made simultaneously in different organs. This concept of niche-specific immunity is further exploited by the gastrointestinal tract, in which different gut segments drain to immunologically distinct LNs that accommodate the needs of each functional intestinal unit^1,2. However, an underappreciated feature of the gastrointestinal system is that several gut-draining LNs (gLNs) serve additional organs, including the pancreas and liver. Yet with few exceptions^3–5, these co-draining LNs have been primarily investigated as if solely draining liver, pancreas or duodenum in mice^6,7, overlooking the potential impact of mixed drainage. This issue is relevant in humans, where pancreaticoduodenal LNs are numerous⁸ and pancreatic diseases are frequently associated with hepatic and intestinal pathologies. Whether LN co-drainage serves as a mechanism of immune crosstalk between this triad of functionally and anatomically connected tissues in the upper digestive system is not well understood.

Given that the intestine is constantly subjected to external fluctuations, and that the liver and gut are major lymph output organs⁹, the consequences of LN sharing may be particularly salient for pancreatic immunity, which could be more susceptible to immune crosstalk in the shared LNs. The pancreatic LNs are recognized as critical sites of initiating autoreactivity to β cells in mice^10,11. Furthermore, several mechanistically ill-defined links have been discovered between gut-derived microbes and pancreatic immunity in the context of type 1 diabetes (T1D), pancreatitis and pancreatic cancer in humans and mouse models of these diseases^12–15.

The duodenal-LNs have the highest propensity amongst the gLNs to promote tolerance¹, reflected in both the tolerogenic profile of their migratory dendritic cells (migDCs) and the induction of regulatory CD4⁺ T (pTreg) cells. However, these LNs can be rendered pro-inflammatory by intestinal infections, triggering food allergies or celiac disease^1,16,17. Previous studies examining the disruption of the tolerogenic properties of the duodenal LNs have been in response to oral antigens, in which the antigen and the infection are derived from the same site: the duodenum. In contrast, here we sought to understand whether pancreatic adaptive immunity is impacted by LN sharing with the liver and gut. Leveraging the fact that the pancreas shares different LNs with liver versus gut, we uncovered the influence of the duodenum on pancreatic immunity and attribute the phenomenon to differential migDC milieus, created by the respective co-drainage, which in turn shape pancreas-reactive T cell fate.

Results

Pancreatic lymph node drainage is shared with gut and liver in a hierarchical manner.

Several LNs are embedded in the pancreas (Figure 1A) and may react to pancreatic antigen^3–7. To comprehensively identify all pancreatic LNs in mice, we used pancreatic β cell antigen-reactive CD4⁺ T cell proliferation as a proxy for pancreatic drainage into these LNs, as physiological β cell turnover is sufficient for antigen availability^10,18. We utilized B6 RIPmOVA mice that express OVA under the control of the rat insulin promoter¹⁸. After transfer of OVA reactive OT-II cells into these mice, OT-II cells divided in the liver-, celiac- and top duodenal-LNs in a semi-hierarchal manner, but not in any of the other LNs (Figures 1B, 1C, and S1A, S1B). A similar liver and celiac>duodenal LN drainage hierarchy was observed for another CD4⁺ T cell clone, BDC2.5 cells, reactive to the autoantigen Chromogranin A of the genetically diabetes prone non-obese diabetic (NOD) mouse strain^19–21 (Figures 1B, 1C, and S1A, S1B). Antigen availability in the LN drove this T cell activation, as inducing β cell death with streptozotocin (STZ) further increased proliferation (Figure S1C). To assess the degree of sharing of these pancreatic LNs with gut, we provided OVA as an oral antigen and followed OT-II cell proliferation. As expected, OT-II cells proliferated in the gLNs including the celiac-LN but little in the liver-LN (Figures 1B, 1C, and S1A, S1B). To corroborate these findings and assess the liver contribution for which we had no T cell system available, we used a strategy that permitted us to detect the origin of migratory dendritic cells (migDCs). We crossed Ptfa1^CreERT2, Alb^Cre and Vil1^CreERT2 with Rosa26^LSL-tdTomato mice, turning the exocrine pancreas, liver, and intestine, respectively, red (Ptf1a^Tomato, Alb^Tomato and Vil1^Tomato mice)^22–24. The presence of Tomato-loaded migDCs in each LN was determined, confirming the pancreatic and duodenal contributions to the liver-, celiac-, and duodenal-LNs (Figures 1D and S1D–1F), though pancreatic antigen-loaded DCs were only recovered upon artificial induction of exocrine cell death in Ptf1a^Tomato mice (diphteria toxin receptor (DTR) expression in Ptf1a^DTRTomato mice). This approach revealed that the liver predominantly drains to the liver- and to a lesser extent the celiac-LN (Figures 1D and S1E). Thus, we have identified the hierarchy of pancreatic LNs in mice and the relative contributions of liver and gut drainage (Figure 1E).

Figure 1. Pancreatic lymph node drainage is shared with gut and liver in a hierarchical manner.

(A) Images of abdominal cavity with representative locations of draining LNs circled in yellow. Scale bars indicate 0.5 cm. (B) Plots for CFSE dilution in LNs after adoptive transfer of labeled transgenic T cells representative of data in C. (C) Division index of CD45.1+ cells as in B (RIPmOVA n=4, NOD n=3, B6 n=3) 6 days (RIPmOVA and NOD) or 3 days (B6) post transfer. (D) Frequency of tdTomato⁺ DCs (CD11c⁺) from LNs of Ptf1a^DTRTomato, Alb^Tomato, and Vil1^Tomato mice. (E) Schematic of tissue drainage hierarchy to indicated LNs: liver LN (L-LN) receives lymph from liver and pancreas, celiac LN (Ce-LN) from liver, pancreas and duodenum, duodenal LN (D-LN) from pancreas and duodenum. Dotted lines indicate small potential drainage contribution of duodenum to L-LN and liver to D-LN. L = liver, Ce = celiac, D = duodenal, J = jejunal, I = ileal, C = cecal-colonic, Ac = ascending colonic, In = inguinal, LN = lymph node. In all panels: mean ± SEM, one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S1.

Migratory dendritic cells from pancreas, gut and liver are transcriptionally distinct.

Given the importance of migDC profiles in determining T cell fate in LNs^1,25, we wondered whether migDCs originating from liver, gut and pancreas are unique in nature, thereby leading to distinct immune milieus in each pancreatic LN according to the relative contributions of each organ. Flow cytometry analysis for the two main populations of migDCs, CD103⁺CD11b^– cells (migDC1s) and CD103⁺CD11b⁺ cells (migDC2s) revealed migDC1s as the main population in liver and pancreas, while the duodenum harbors both populations (Figure 2A). This was reflected in the three pancreatic LNs (Figure 2A), where migDC2s were rare in the liver-LN but numerous in the celiac-LN and duodenal-LN, consistent with the contribution of gut drainage. Further, migDC1s comprised the majority of Tomato⁺ cells in our Tomato expression systems (Figure S1F). Notably, few Tomato⁺ LN-resident DCs were observed, suggesting tissue-specific self-antigen was primarily taken up by migDCs in the organ of origin rather than acellular lymph flow (Figures S1D). Focusing in on tissue-specific migDC1s, we bulk-sorted this population from liver, pancreas, and duodenum of B6 mice and conducted RNA sequencing (RNAseq) to determine if they differed by location. Principal component analysis (PCA) revealed that migDC1s clustered strictly by tissue of origin (Figure 2B and Table S1), although hepatic and pancreatic migDC1s were more closely related (PC2, 12 % of variation) to each other than to migDC1s from the duodenum (PC1, 25% of variation). This was also reflected in the number of genes uniquely enriched in each migDC1 population (Figure 2C). Gene ontology analysis revealed more active lymphocyte activation and pro-inflammatory processes in the migDC1s from liver and pancreas compared to duodenum (Figure S2A), with interleukin-12 (IL-12) and tumor necrosis factor alpha (TNFα) production pathways amongst the most enriched. Correspondingly, genes enriched in both liver and pancreas versus duodenal migDC1s (Figure 2D, bottom right heat map) included Il12, Il15ra, and Mx1, which represent an interferon stimulated and type 1 immunity-prone signature^16,17, as well as other pro-inflammatory genes such as Alox5ap²⁶, Ccr ²⁷, and Ccl5,^28,29. By contrast, Aldh1a2 was enriched in intestinal migDC1s (Figure 2D, bottom left heat map), encoding a rate-limiting enzyme (RALDH) for production of retinoic acid (RA) from dietary vitamin A and a known critical driver of regulatory T (Treg) cell induction³⁰. Other genes indicated an overall more activated state of duodenal compared to hepatic and pancreatic migDC1s (Figure 2D, bottom left heat map, e.g., Ier2³¹, Egr1, Dusp1, Junb and Fosb³², Ppp1r15a³³, Apol7c³⁴ and Arf4³⁵). Genes uniquely enriched in liver migDC1s (Figure 2D, top left heat map) were reminiscent of an immature DC subpopulation in this organ (Id3, Ly6a, and Cd34), next to genes clearly imposed by a hepatic environment (Alb, Apoa1, Serpina gene family). Genes unique to pancreatic migDC1s (Figure 2D, top right heat map) are more typically associated with macrophages, where Anxa2, a binding partner of S100 proteins, regulates TLR4 internalization and the unfolded protein response upon bacterial infection³⁶, Fcer1g facilitates the uptake of IgE bound antigen, and Ccl6 encodes a putative CCR1 ligand in rodents³⁷.

Figure 2. Migratory dendritic cells from pancreas, gut and liver are transcriptionally distinct.

(A) Plots of CD103 and CD11b expression among DCs within the indicated tissues and LNs representative of gating for cells analyzed in B-E. Cells are pre-gated as live, (NK1.1, TCRβ, B220, CD90)–, CD11c+, MHCIIhi, F4/80–. (B) Principal component analysis (PCA) of bulk RNAseq of sorted migDC1s (n=4, pooled from 4 mice each). (C) Venn diagrams displaying the number of tissue-enriched genes (log2 > 2 & p < 0.05), based on overlap from the indicated comparisons. (D) Heatmaps displaying indicated tissue-enriched DC genes. Genes mentioned in text and further analyzed are highlighted in red. (E) Relative expression of indicated genes by qPCR on sorted migDC1s from the indicated tissues and LNs of B6 mice (n=3, 3 mice pooled per replicate). L = liver, P = pancreas, D = duodenum, Ce = celiac, LN = lymph node. In all panels: mean ± SEM, one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S2 and Table 1.

To determine whether location-specific immunomodulatory signatures could be recovered in the draining LNs, we sorted migDC1s from liver, pancreas, duodenum and the three pancreatic LNs and conducted quantitative PCR (Q-PCR) analyses of significantly regulated genes (Figure 2E). Gut-enriched transcripts displayed a liver<celiac<duodenal-LN trend of expression (Aldh1a2, Apol7c, Apol10b), while “pancreas plus liver” (Ccl5, Il12b, Il15ra) or uniquely pancreas-enriched genes (Anxa2, Fcer1g) displayed the inverse gradient of expression, in line with the contributions of the organs to each LN we established. Other uniquely pancreas enriched genes were reduced in the LN (Figure S2B: Emp3, Ccl6) but in line with the pancreatic contribution, while uniquely liver-enriched genes were virtually lost in the LN (Figures 2B: Ly6a, and S2B: Cd34, and Id3), confirming they represented immature DCs. Gut-enriched genes indicative of a lamina propria-specific signature (Hic1)³⁸ or early response genes (Asb2, Egr1, Ifitm6, Ier2) were also undetectable in LNs, while genes associated with general migDC1 maturation state were enriched in LN compared to tissues but showed no evident duodenal-LN bias (e.g. Arf4, Ppp1r15a); the same was true for Ccr7, a DC migration marker (Figure S2B). This underscores the maturation process that all DCs undergo upon trafficking, and in line with this idea, transcripts of most genes encoding immunomodulatory proteins, including Aldh1a2, Ccl5, Il15ra, and Il12b were much higher in the LNs than tissues. Similar expression patterns were also observed in migDC1s from NOD mice (Figure S2C), as was a corresponding gradient of IL-15 protein levels in the pancreatic LNs (Figures S2D).

Collectively, these data suggest that the three pancreatic LNs differ in their migDC niche, both due to the relative presence of migDC2s contributed by the gut, and the profile of the migDC1 pool, shaped in a hierarchal manner by the liver, gut, and pancreas itself.

The relative contribution of migratory dendritic cells from pancreas, gut and liver underlies unique dendritic cell signatures in the different pancreatic lymph nodes.

Our Q-PCR results could not distinguish whether the gene expression unique to each LN reflect the cell heterogeneity due to contributions of migDC1s from each co-drained organ or if all cells, irrespective of source, homogenously adapt to a particular LN environment. To resolve this ambiguity, we conducted single cell RNA sequencing (scRNAseq) of DC-enriched (CD11c-YFP) cells from the three pancreatic LNs (Figures S3A and S3B). After identifying the myeloid cells using published gene data sets^25,39,40 (Figures S3C and S3D) and performing clustering of all migratory DCs, we identified ten distinct populations (Figure S3E), two of which represented migDC2s, one monocyte-derived DCs (moDCs), and one pre-apoptotic cells; the rest we assumed to be migDC1s (Figure S3F). Down-sampled UMAPs of all three LNs confirmed the migDC2 population was virtually absent from the liver-LN, present in the celiac-LN and most abundant in the duodenal-LN, while moDCs showed the opposite trend (Figure S3G), as expected from the LN contribution analyses in Figure 2A. Re-clustering of only the migDC1s revealed eight populations (Figure 3A). Of note, most were dominated by immune genes we already identified in our tissue migDC1 analysis as enriched in liver, pancreas, or duodenum (Figure 3B, Table S2A–H and Table S3). Furthermore, down-sampling of cells from the liver-, celiac-, and duodenal-LNs (Figure 3C) revealed that clusters 0–4 were clearly more represented in the liver-LN, clusters 5 and 7 in the duodenal-LN, while cluster 6 showed a less marked overrepresentation in the duodenal-LN; all populations were present in the celiac-LN. Since not all genes that were migDC1 cluster-defining in the scRNAseq data had been identified as indicative of tissue origin in our bulk migDC1 RNAseq, we conducted further Q-PCR analysis of select highly regulated genes on migDC1s from liver-, celiac- and duodenal-LNs and tissues. This approach identified Dusp2, Gpr183, S100a4 and S100a6 as showing a liver>celiac>duodenal LN gradient (Figure S3H), suggesting the bulk migDC1 gradients observed (see Figure 2E) resulted from differential contributions of DCs from the three organs, whereby the related clusters 0 (S100a4), 4 (Fcer1g,, Il12b), and 1 (Anxa2, Fcer1g) hinted at a pancreatic origin; clusters 2 (Ccl5, Il15ra) and 3 (Ccl5, Il12b) were likely of pancreatic or hepatic origin; clusters 5 (Ftl1, Apol10b) and 7 (Aldh1a2, Apol7c) of duodenal origin, and only cluster 6 (Cd40, Ebi3) reflected a mixed origin. Co-expression analyses (Figure 3D) further suggested that cluster 4 (Il12b, Anxa2) was of pancreatic origin.

Figure 3. The relative contribution of migratory dendritic cells from pancreas, gut and liver underlies unique dendritic cell signatures in the different pancreatic lymph nodes.

(A) UMAP of migDC1 subset containing 10,347 cells across 8 clusters. (B) Heat map of cluster defining genes. (C) UMAP of each LN down-sampled to an equivalent of 509 cells each. (D) Feature plots displaying Ccl5, Il15ra, Il12b, Anxa2, Aldh1a2 and Apol7c expression within the migDC1 UMAP. (E) Relative expression of cluster associated genes by qPCR on sorted migDC1s from Ptf1a^ZsGreen, Alb^ZsGreem, or Vil1^ZsGreen LNs (n=3–5, two mice pooled per replicate for Ptf1a^ZsGreen). (F and G) Geometric mean fluorescence intensity (gMFI) and frequency of ALDEFLUOR⁺ migDC1s within tdTomato⁺ and tdTomato⁻ populations of Vil1^Tomato (F) or Alb^Tomato (G) mice (n=5). Data pooled from two independent experiments. Cells are pre-gated on live CD45⁺CD11c⁺MHC-II^hi cells. (H) gMFI of CCL5⁺, DUSP2⁺, ANXA2⁺, and ALDEFLOUR⁺ migDC1s B6 LNs (n=5). (I) Percentage of IL-15Ra⁺, IL-12⁺, and FCER1G⁺ migDC1s from B6 LNs (n=5). See Legend Figure 1 for abbreviations. For all panels: mean ± SEM, t-test and one-way ANOVA: ns, not significant, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S3 and Tables 2–9.

To more directly pinpoint the origin of these clusters, we bred Ptfa1^CreERT2, Alb^Cre and Vil1^CreERT2 mice to Rosa26^LSL-ZsGreen mice encoding the inducible expression of non-degradable GFP, ZsGreen (Vil1^ZsGreen, Alb^ZsGreen, and Ptf1a^ZsGreen mice)⁴¹. ZsGreen was selectively expressed in each organ (Figure S3I), and ZsGreen⁺ migDC1s confirmed the migration trends of Figure 1D (Figures S3J and S3K), though in this more sensitive system a few DCs originating from the gut (Vil1^ZsGreen) were recovered in the liver-LN and some liver derived DCs (Alb^ZsGreen) in the duodenal-LN. Importantly, this approach permitted the recovery of migDC1s even in the Ptf1a system without DTR and resulted in higher percentages of ZsGreen-loaded migDCs than the Tomato system (compare Figures 1D and S1E with Figures S3J and S3K). We sorted the ZsGreen⁺ migDC1s in the three pancreatic LNs from each strain and conducted Q-PCR analysis of the genes identified both as cluster-enriched (scRNAseq of LNs) and tissue-enriched (bulk tissue RNAseq) as well as additional genes enriched in a particular LN discovered through the scRNAseq (Figures 3E and S3J). These analyses across Cre lines revealed two common features: First, genes expected to be enriched in duodenal ZsGreen⁺ migDC1s (Aldh1a2, Apol7c, Apol10c), pancreatic ZsGreen⁺ migDC1s (Anxa2, S100a4, S100a6, Gpr183, Emp3), or hepatic and pancreatic ZsGreen⁺ migDC1s (Il12b, Fcer1g) were indeed enriched, while Ccl5 and Il15ra were highest in ZsGreen⁺ cells of hepatic origin, resolving the origin of the Ccl5 clusters 2 and 3 (Figure 3A) as primarily liver-derived. Second, the expression of immunomodulatory genes was overall similar regardless of the LN a migDC1 had migrated to, indicating no major adaptation to a LN environment at homeostasis. This lack of adaptation to a LN was also underscored by pseudo-bulk comparison of the same clusters between the pancreatic LNs that showed no statistical differences (data not shown). On the functional protein level, we compared the RALDH activity of migDC1s in the three pancreatic LNs of Vil1^Tomato and Alb^Tomato mice, the two models compatible with this green fluorescent enzymatic product-based flow cytometry assay (ALDEFLUOR^™). This confirmed that Tomato⁺ migDC1 cells from Vil1^Tomato mice exhibited higher RALDH activity than the Tomato^– cells in the same LNs (Figure 3F), which was also true for migDC2s (Figure S3L). The opposite was found for Tomato+ migDC1 cells from Alb^Tomato mice (Figure 3G); in both strains the degree of activity was independent of the LN identity. Finally, we confirmed the same transcriptional trends persisted on the protein level in total migDCs1 cells from the three pancreatic LNs (and the colon-draining LN as comparison) by flow cytometry (Figures 3H, 3I, and Figures S3M–O). Some markers displayed these patterns both by intensity per cell (Mean fluorescent intensity, MFI) and percentage of cells present (Aldefluor as proxy for RALDH, DUSP2), some by MFI (ANXA2, CCL5), and some by percentage of cells positive for a protein (IL15Ra, IL-12p40, FCER1G).

Thus, these data demonstrate migDC1s retain tissue of origin imprints, without adaptation to a given LN environment under homeostatic conditions, supporting a model where the DC niche is unique to each pancreatic LN due to the relative contributions of cells from the gut, liver, and pancreas (Figure S3P).

Lymph node co-drainage with liver versus gut differentially impacts pancreatic antigen-specific T cell fate at homeostasis.

Since migDC1s appeared to retain their tissue license and primarily loaded organ-specific antigen prior to migration to the LNs, there were two possible scenarios for the consequences on antigen-specific T cells in the different pancreatic LNs. In one, the primary determinant is the organ-specific presenting DC and therefore the effect on T cells is the same in all LNs. Alternatively, the secretory milieu established by the unique mix of DCs of different origins in the T cell zone can impact T cell phenotype. We speculated that if factors secreted by DCs “in trans” could influence T cells, then LN sharing with the gut versus liver should induce differential imprints on pancreatic antigen-reactive T cells. We therefore first interrogated the consequences on CD4⁺ (BDC2.5 cells) T cell fate in the different pancreatic LNs, using T-bet, FOXP3, GATA3, RORγT, BCL6 and PD-1 as proxies for Th1, Treg, Th2, Th17 and Tfh cell skewing, respectively, as well as anergy markers FR4 and CD73 at homeostasis. BDC2.5 cells trended to be more FOXP3, GATA3 and RORγT positive the more the duodenum contributed to the LNs (Figure 4A and Figures S4A–S4D). Conversely, T-bet⁺ cells were the highest in the liver LN and decreased in celiac and duodenal LNs. Different subsets of FOXP3⁺ cells followed the T helper lineage trends (Figure S4B). Of note, T-bet and FOXP3 are known to be primarily induced by cDC1s, in contrast to GATA3 and RORγT, whose inductions reportedly depend primarily on cDC2s, suggesting the underlying reason for these trends were the different ratios of tolerance versus type 1 immunity-promoting migDC1s (See Figure 3, RALDH activity versus IL12, IL15R expression) and increasing presence of duodenum derived migDC2s from liver-, celiac- to duodenal-LN (Figure S3G), respectively. FR4, PDL-1, BCL6, and CD73 showed flatter trends. Low cell recovery and fate acquisition was observed in the colonic and inguinal control LNs (Figure S4D), likely to the lack of sufficient stimulation by antigen.

Figure 4. Lymph node co-drainage between pancreas and gut correlates with homeostatic phenotype of antigen-specific T cells.

(A) Frequency of FOXP3⁺, GATA3⁺, RORγT⁺, FR4⁺CD73⁺, BCL6⁺PD-1⁺, and T-bet⁺ among BDC2.5 T cells in NOD LNs (n=5). (B) Frequency of CD62L⁺CD44⁺, CD62L⁻CD44⁺, TNFα⁻IFNγ⁺, TNFα⁺IFNγ⁺, LAG3⁺, and PD-1⁺ among NY8.3 T cells in NOD LNs (n=5). (C) Frequency of α4β1⁺, α4β7⁺, CCR9⁺ and α4β7⁺CCR9⁺ among BDC2.5 T cells in LNs of NOD mice (n=5). (D) Frequency of α4β1⁺, α4β7⁺, CCR9⁺ and α4β7⁺CCR9⁺ among NY8.3 T cells in LNs of NOD mice (n=5). (E) Frequency of α4β1⁺, CCR9⁺, and α4β1⁺CCR9⁺ T cells in tissues of NOD mice (n=5). Data representative of two independent experiments. L = liver, Ce = celiac, D = duodenal, C = cecal-colonic, In = inguinal, SI = small intestine, P = pancreas, LN = lymph node. For all panels: mean ± SEM, one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.005, ***p < 0.0001. Please also see Figure S4.

We then similarly monitored the consequences for pancreatic-antigen specific CD8⁺ T cells (NY8.3 cells⁴², recognizing pancreatic islet derived autoantigen IGRP). Flow cytometry analysis of the pancreatic LNs three days after adoptive transfer of naïve NY8.3 CD8⁺ T cells revealed the highest frequency of IFNγ⁺, TNFα⁺ but also LAG3⁺ cells in the liver- and celiac-LNs followed by the duodenal-LN (Figure 4B, Figure S4E–S4G). CD44⁺CD62L⁺ and PD-1⁺ cells were the most frequent in the liver-LN, correlating with the more pro-inflammatory migDC1s present in the liver-LN.

In addition to T cell polarization, DCs give homing instructions to T cells through signals like RA to induce the gut-homing receptors CCR9 and α4β7⁴³. We therefore wondered how duodenal co-drainage would affect homing receptor expression of pancreatic antigen reactive T cells. We first confirmed in our system that gut antigen-reactive CD4⁺ and CD8⁺ T cells followed this rule (Figures S4H–S4N). Both CD4⁺ (OT-II, Figure S4K and S4L) and CD8⁺ (OT-I, Figure S4M and S4N) T cells received CCR9 and α4β7 imprints following orally administered OVA antigen in a manner reflective of gut contribution to the pancreatic LNs. This pattern was similar to α4β1⁴⁴ and even observed for the polyclonal T cell pools (Figures S4I and S4J). Adoptively transferred BDC2.5 cells also displayed CCR9 expression and higher frequencies of α4β7⁺ cells uniquely in the LNs shared with the duodenum but not with the liver (Figures 4C and S4H); the same was seen for NY8.3 cells (Figures 4D and S4H). Notably, while α4β1⁺ T cells were enriched, CCR9⁺ T cells were present in the pancreas at similar or even higher frequencies than gut (Figure 4E), in line with previous reports⁴⁴, indicating that gut homing markers do not prevent T cells from seeding the pancreas.

Taken together, these results support the conclusion that pancreatic antigen-reactive T cell fate and homing instruction differs according to the organs co-drained by the LNs in which T cells encounter antigen, whereby LNs dominated by liver (and pancreas) facilitate type 1 immunity and those by the duodenum a tolerogenic program.

Pancreatic antigen reactive CD4⁺ T cell fate is impacted by intestinal viral infections.

We next wondered if pancreatic adaptive immunity is susceptible to a break in tolerance due to gastrointestinal infections as previously demonstrated for the duodenal-LN¹. We chose viral perturbations as these require behavioral switches in cDC1s, the DC type most implicated in our crosstalk. We first tested multiple doses of various gastrointestinal viruses, including reovirus strains type 1 Lang (T1L)¹⁶ and T3SA+⁴⁵ and acute murine norovirus strain (MNV-CW3)¹⁷ for their capacity to induce polyclonal T-bet⁺ CD4⁺ T cell (Th1 cells) in the gLNs of adult NOD mice. Only T1L elicited robust T-bet expression (Figures S5A–S5C). Therefore, we chose this virus to study the effect of immune perturbation of the duodenal-LNs on pancreas reactive CD4⁺ T cells using adoptively transferred BDC2.5 T cells. Indeed, the cells showed a much higher percentage of T-bet⁺FOXP3^– and T-bet⁺FOXP3⁺ cells and a reduction of T-bet^–FOXP3⁺ cells following T1L infection (Figures 5A–5C and S5D–S5F, S5L). This effect was observed in all LNs but, among the pancreatic LNs, it was strongest in the LNs shared only with the duodenum. T1L infects intestinal epithelial cells⁴⁶ that are subsequently engulfed by DCs, where it does not further replicate⁴⁷, but we formally excluded that the (minimal) effect of T1L in the liver-LN was due to viral spread to pancreas or liver (Figure S5G). To confirm that our effect on pancreas reactive T cells was due to co-drainage rather than systemic inflammation, we undertook several approaches: first, to determine if our T cell phenotype upon exclusively pancreas derived antigen was comparable to antigen sourced from the site of highest T1L infection, i.e. the gut, we provided the BDC2.5 antigenic peptide orally. The peptide antigen elicited the same proximal to distal gradient of T-bet^–FOXP3⁺ cell induction along the gLNs (Figures S5H–S5J, S5M) as reported for dietary protein¹. Similar to studies using pooled gLNs upon dietary OVA antigen¹⁶, T1L infection elicited a decrease in T-bet^–FOXP3⁺ cells and an increase in T-bet⁺FOXP3^– cells among BDC2.5 cells in all gLNs (Figures S5H–S5J, S65M). Notably, the shift of fate was comparable to that seen with the endogenous BDC2.5 antigen (Figures 5A–5C), suggesting we observed an on-target effect. Next, we provided antigen (OVA) systemically to make it accessible to DCs in every LN and assessed OT-II T cell fate in multiple locations, reasoning that if gastrointestinally-introduced T1L acted through a systemic effect, we should see increased T-bet⁺ and decreased FOXP3⁺ frequencies everywhere. However, these phenomena remained confined to the gLNs (Figure 5D–5F, S5N) despite robust seeding and proliferation (Figure S5P and S5Q) in all additional lymphoid tissues tested (axillary and inguinal LNs and spleen). Finally, we gave both OVA and T1L systemically, which demonstrated that other sites are in principle capable of mounting robust T-bet expression (Figures 5G and 5H, S5O, S5R and S5S). Although it is possible that the gut contributes minorly to the liver LN also (see Figure S3K, Vil1^ZsGreen cells), we attributed the effect of high dose T1L in NOD mice to the combination of high pancreatic antigen abundance in this LN (Figures 1B–1D), the baseline skewing of the liver LN toward type 1 immunity (Figures 2D and 2E, Figure 4A), and higher interferon sensitivity of the liver of NOD compared to B6 mice (Figure S5T and S5U). Furthermore, an overall “all-or-none” response to infection induced interferon signaling seems to occur, supported by the fact that in this condition we do not observe graded responses between the celiac- and duodenal-LNs.

Figure 5. Gastrointestinal viruses shift pancreatic antigen reactive CD4 T cell fate through lymph node sharing.

(A) Plots of T-bet and FOXP3 expression among BDC2.5 T cells in Ce-LN of NOD mice, representative of gating applied in B-C. (B and C) Frequency of single FOXP3⁺ (B) and single T-bet⁺ (C) BDC2.5 T cells in NOD mice 96 h after cell transfer and infection with reovirus T1L (n=5). Data representative of two independent experiments. (D) Plots of T-bet and FOXP3 expression among OT-II T cells in D-LN of B6 mice representative of gating applied in E-H. (E and F) Frequency of single FOXP3⁺ (E) and single T-bet⁺ (F) OT-II T cells in B6 mice 96 h after cell transfer and intragastric (i.g.) infection with T1L and intravenous (i.v.) OVA (n=5). (G and H) Frequency of single FOXP3⁺ (G) and single T-bet⁺ (H) OT-II T cells in B6 mice 96 h after cell transfer and i.v. infection with T1L and i.v. OVA (n=5). Ax = auxiliary, Sp = spleen, see Legend Figure 1 for additional abbreviations. In all panels: mean ± SEM, t-test, one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S5.

These findings suggest that pancreatic-antigen reactive CD4⁺ T cell fate not only differs at baseline depending on the LN in which it is induced but also can be changed to a more pro-inflammatory state by gut viral infections due to shared LN drainage.

Intestinal viral infection leads to a pro-inflammatory profile of dendritic cells from liver and pancreas in the shared lymph nodes.

We sought to define the DC biology that underlies the gradient of type 1 immunity between the pancreatic LNs following intestinal viral infections using T1L as a model. MigDC1s sorted from the pancreatic LNs or strictly gLNs of T1L and mock infected B6 and NOD mice were subjected to Q-PCR analysis of interferon stimulated genes (ISGs, Figures 6A and S6A, B). This analysis revealed that most ISGs peaked at 48 h post infection and followed the established gradient of gut contribution to pancreatic LN drainage, although the liver-LN showed no strong transcript elevation. In the tissues of NOD mice, migDC1s showed no ISG activation in the pancreas upon T1L; some upregulation was observed in the gut but also in the liver (Figure 6B), in line with the suspected interferon sensitivity in this mouse strain’s liver (see Figures S5T and S5U). To assess the functional output of ISG upregulation, we isolated migDC1s from pancreatic, colonic and, as a peripheral control, inguinal LNs and stained them for activation (CD80, CD86), migration (CCR7) and cytokine production (IL-12) upon T1L infection. Infection led to a consistent increase in migDC1 activation and migration (Figure 6C) and immune cell and migDC1 counts (Figures S6C and S6D) only in the LNs draining the gut (D, C), while an increase in IL-12⁺ DC frequencies was extended to the liver-LN (Figure 6D). To directly determine whether this shift to a pro-inflammatory program was due to alteration of gut-derived migDC1s, we analyzed the DCs of Vil1^Tomato mice in the pancreatic LNs for IL-12 and CD80. As expected, IL-12⁺CD80⁺ cells were enriched in both the celiac- and duodenal-LNs amongst the Tomato⁺ migDC1s (Figures 6E, 6F and Figures S6E–S6G). While the contribution of type 1 cytokines by gut-derived DCs would explain our pancreatic antigen-reactive T cell data through a bystander mechanism^48–50, we wondered whether DCs originating from co-drained organs were also affected by intestinal viral infection, e.g., through paracrine interferon signaling between DCs^51,52. We therefore repeated the former experiment using Alb^Tomato mice (which have improved Tomato⁺ cell recovery compared to Ptf1a^Tomato mice). This experiment revealed that indeed an increased frequency of liver-derived migDC1s was IL-12⁺CD80⁺ in the pancreatic LNs (Figures 6G and S6H–S6J), although to a lesser extent than gut-derived DCs (Figures 6H and S6J, compare Ce-LN to Figures 6F and S6G, respectively). Finally, we sorted ZsGreen⁺ migDC1 cells from Vil1^ZsGreen, Alb^ZsGreen and Ptf1a^ZsGreen mice in the pancreatic LNs following T1L- versus mock-infection and conducted QPCR analysis of ISGs. This analysis underscored the finding that liver- and pancreas-derived DCs acquire an ISG program in the LNs shared with gut (Figures 6I) despite an uninfected origin (Figure S5F).

Figure 6. Intestinal viral infection leads to a pro-inflammatory profile of dendritic cells from liver and pancreas in the shared lymph nodes.

(A-B) Relative expression of ISGs by qPCR of sorted migDC1s from (A) B6 LNs 48 or 96 h after T1L infection (n=3, 2 mice pooled per replicate) and (B) NOD mice 48 h after T1L infection. (C-D) Frequency of (C) CD80⁺CD86⁺ and CCR7⁺ migDC1s and (D) IL-12⁺CD80⁺ migDC1s in indicated LNs of B6 mice 48 h after T1L infection. (E) Frequency of IL-12⁺CD80⁺ tdTomato⁺ migDC1s from Vil1^Tomato mice 48 h after T1L infection (n=5). (F) Ratio analysis of IL-12⁺CD80⁺ within tdTomato⁺ and tdTomato^– migDC1s from Vil1^Tomato mice (n=5). (G) Frequency of IL-12⁺CD80⁺ tdTomato⁺ migDC1s from Alb^Tomato mice 48 h after T1L infection (n=5). (H) Ratio analysis of IL-12⁺CD80⁺ within tdTomato⁺ and tdTomato^– migDC1s from Alb^CTomato mice (n=5). Data pooled from two independent experiments (C-H). (I) Relative expression of ISGs by qPCR of sorted ZsGreen⁺ migDC1s from Vil1^ZsGreen, Alb^ZsGreen, and Ptf1a^ZsGreen mice 48 h after T1L infection (n=3). See Legend Figure 1 for abbreviations. In all panels: mean ± SEM, t-test: *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S6.

These findings support a model in which migDC1s directly responding to an insult, in the case of T1L from the intestine, can influence other migDC1s, which together modify T cell responses in the same anatomical niche.

Pancreas-reactive cytotoxic CD 8⁺ T cells are induced upon gastrointestinal viral infections

CD8⁺ T cells are the essential executioners of T cell mediated autoimmunity. We therefore screened the endogenous CD8⁺ T cell response of NOD mice to enteric viruses and found that T1L, T3SA⁺ and MNV-CW3 at the lower dose tested induced GZMB (only expressed upon infections) and IFNγ (expressed at baseline but increased upon infections) expression, confirming a lower activation threshold for CD8^+50,53,54 than CD4⁺ T cells (Figures S7A–S7C). We therefore next tested the effect of lower dose T1L on NY8.3 T cell fate in the pancreatic LNs. We observed a robust conversion to cytotoxic CD8⁺ T cells that followed the gradient of LN sharing between pancreas and duodenum (Figures 7A–7D and S7D, S7E). Again, we observed some effect of the infection in the liver LN despite low gut contribution to this LN. The same conversion held true when STZ was omitted (Figures S7F and S7G), when MNV-CW3 and T3SA were used (Figures S7L–S7O), or when an entirely different pancreatic β cell reactive CD8⁺ T cell clone, G9C8⁵⁵, recognizing insulin, was used (Figures S7H and S7I). We wished to again address the dependence of these effects on specific co-drainage. In this case, we could not directly compare dietary (as proxy for entire length of gut) and endogenous β cell derived antigen, as the IGRP peptide recognized by NY8.3 cells is insoluble. To establish this reference point, we charactered OT-I cells following oral OVA in B6 mice infected with T1L or MNV-CW3, the viruses known to break oral tolerance in CD4⁺ T cells^16,17, and found that indeed, OT-I cells became cytotoxic both in the LNs shared with the pancreas and in LNs unique to the gut (Figures S7J and S7K). When we monitored OT-I conversion to cytotoxic T cells upon systemic OVA delivery, we found that the LNs sensitive to oral T1L infection remained the same (Figure 7E and Figures S7P, S7Q), while if we delivered T1L systemically, the OT-I cells became cytotoxic in all sites monitored (Figure 7F and Figures S7R, S7S). Finally, we wondered if this activation was of consequence in the pancreatic tissue. First, we established that both CCR9 and α4β1 were upregulated by GZMB⁺ NY8.3 cells after T1L infection (Figure 7G). Next, we transferred total splenocytes from NY8.3 mice (containing a majority of autoreactive CD8⁺ T cells critical for islet autodestruction⁵⁶ and a minority of necessary CD4⁺ T cells⁵⁷) to female NOD Rag1^−/− mice, which displayed no pancreatic islet infiltration, and concomitantly infected with T1L or not (Figure 7G). Two weeks after cell transfer, we indeed found an enrichment for GZMB⁺β1⁺ NY8.3 cells in the pancreas upon T1L infection but not gut (Figure 7I), while CCR9⁺ cells were unaltered in the pancreas but trended to be enriched in the gut upon infection (Figures S7T and S7U). Pancreatic islet infiltration by lymphocytes was significantly increased 4 weeks following viral infection compared to non-infected controls (Figures 7J and S7V), which were predominantly CD8⁺ T cells, as expected (Figure S7W). In this experimental setup, we did not observe full-blown β cell destruction or diabetes (Figures S7X and S6Y), processes that likely require breaking additional self-tolerance mechanisms.

Figure 7. Gastrointestinal viruses activate cytotoxic pancreatic antigen reactive CD8 T cells in shared lymph nodes and lead to islet lymphocyte infiltration.

(A) Plots of IFNγ and GZMB expression among NY8.3 T cells in Ce-LN of NOD mice representative of gating applied in B-C. (B-C) Frequency of IFNγ⁺GZMB⁺ (B), total GZMB⁺ (C) and single IFNγ⁺ (D) NY8.3 T cells in indicated LNs of NOD mice 72 h after cell transfer and infection with T1L (n=5). Data representative of two independent experiments. (E-F) Frequency of GZMB⁺ OT-I T cells in indicated LNs of B6 mice (E) 72 h after intragastric T1L infection and i.v. OVA (n=4) and (F) 72 h after intravenous T1L infection and i,v, OVA. (G) Frequency of CCR9⁺GZMB⁺ and α4β1⁺GZMB⁺ NY8.3 T cells in indicated LNs of NOD mice 72 h after T1L infection (n=5). (H) Schematic of diabetes development experiments in NOD Rag−/− mice. (I) Frequency of β1⁺GZMB⁺ NY8.3 T cells in tissues of NOD Rag1^−/− mice 2 wk after cell transfer and T1L infection. (J) Average infiltration score of pancreatic islets of NOD Rag1^−/− mice 4 weeks after T1L infection. Data pooled from two independent experiments (n=4–5 per experiment). See Legend Figure 5 for abbreviations. In all panels: mean ± SEM, t-test and one-way ANOVA: *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001. Please also see Figure S7.

These findings show that intestinal infection has the power to alter pancreas-reactive T cell responses within shared LNs, with correlating consequences for pancreatic tissue immunity.

Discussion

Through our investigations of LN sharing in the upper gastrointestinal tract, we have uncovered a previously underappreciated means of influencing tissue-specific adaptive immunity in the upper digestive system that likely applies to other situations of organ co-drainage. The discovery adds to the growing list of ways in which the gut influences systemic immunity and offers LN co-drainage as a possible mechanism by which gastrointestinal infections can lower the threshold for autoimmunity in the pancreas (or liver) without infection of these organs themselves.

The comparison of liver, pancreas and gut migDC1s highlights the concept that migDCs are instructed distinctly in each tissue milieu, an imprinting that will shape the nature of the immune outcome in lymphoid organs including the propensity to develop autoimmunity. Through our RNAseq analyses, we identified tissue-specific signatures of migDC1s within the duodenum, pancreas, and liver and showed how the immunomodulatory gene signatures are retained following migration to the LNs despite major transcriptional changes associated with DC maturation and migration. Of note, Batf3-dependent cDC1s have previously been ascribed a critical role in licensing pancreatic β cell autoreactive T cells⁵⁸, and are the DC subtype specialized in taking up dead cells. These findings in combination with our new data suggest that cDC1s are the DC subtype deciding between peripheral self-tolerance and autoimmunity⁵⁹, akin to their role with respect to non-self antigens such as dietary proteins. Our data suggest that duodenal co-drainage offsets the more pro-inflammatory effect of liver- and pancreas-derived migDC1s and supports self-tolerance to these organs. However, during intestinal viral infection, we found that non-intestinal-derived migDC1s could be affected in trans with the potential to alter T cell responses accordingly. While we focused on migDC1 dependent processes, it is possible that migDC2s can influence T cell fate in situations of co-drainage and type 2 or 3 immunity-inducing stimuli. Finally, we also discovered that DC heterogeneity in a LN can be due to the relative contribution of DCs of different origins; it remains to be established if LN sharing represents an exception or a much more common phenomenon that has been ignored in single organ-focused studies.

Despite the potential for activating auto-reactive T cells, shared LNs between the pancreas and duodenum may still be protective overall due to the strong tolerogenic profile of the duodenal-LN¹. Notably, gastrointestinal infections and intestinal microbiota composition can either delay or accelerate diabetes in NOD mice in a plethora of studies^13,15, leaving the field with no unifying consensus. The mechanistic prediction from our study is that any perturbation leading to type 1 immunity in the duodenum and coinciding with β cell antigen presentation in the co-drained LNs will facilitate autoimmunity.

Our studies take advantage of the dual LN sharing between pancreas and liver versus duodenum to query the impact of co-drainage on antigen-specific T cell fate.

LN sharing between duodenum, liver and head of the pancreas is likely linked to their common ventral foregut origin during embryonic development and parallel lymphangiogenesis. Our study focused on the potential dangers of shared LNs in the context of inflammatory stimuli. The tolerogenic influence of the gut on the shared LNs, while beneficial for preventing autoimmune attack, may also have negative consequences such as supporting the aggressiveness of hepatic or pancreatic cancers. This endowment of the gut with direct regulation of autoimmunity in other tissues is likely an evolutionary tradeoff of the potentially much larger benefit of aligning the duodenal adaptive immune response with that of liver or pancreas: these organs will occasionally encounter gut-derived antigens via the leakage into the liver or pancreas through the portal vein⁶⁰ or pancreatic ducts¹², respectively. LN sharing can ensure the reciprocal homing of antigen specific T cells and save these vital organs from fatal infection or inappropriate responses. From a practical standpoint, the anatomical accessibility of the gut offers the potential of non-invasively targeting the hepatic or pancreatic immune system, via shared LN drainage with the gut, for therapeutic purposes.

Limitations of Study

Our studies are solely performed in mice, and our T cell fate analyses largely rely on adoptively transferred self-reactive clones. For proof-of-principle and due to the intuitive relevance, we focused on the fate of pancreatic β cell reactive T cells and environmental influences that we know are gut-derived. We ignored the potential active influence of liver-derived DCs in the shared LNs on these T cells, the impact that different LN milieus may have on liver- or exocrine pancreas-reactive T cells, and the effect of perturbations in the liver or pancreas. We thus also did not explore what the consequence is of potentially misdirecting pancreas-reactive T cells to the gut in the case of exocrine self-antigens, which as opposed to the endocrine antigens investigated here will be highly abundant in the duodenum. Finally, our findings reflect the situation in which antigen, likely in the form of dead self-cells, is predominantly captured in the tissue, and seek DC-centric explanations for T cell fate. It is possible that lymph-borne bioactive molecules or antigens lead to enhanced crosstalk and engage additional DC subtypes, thus that the impact of co-drainage may be even larger than described here.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Daria Esterházy (desterhazy@bsd.uchicago.edu).

Materials availability

This study generated no new materials.

Data and code availability

• RNA-seq data have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the key resources table.

• This paper does not report original code.

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

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
APC anti-mouse CD25 (clone PC61.5)	eBioscience	Cat#17-0251-82; RRID: AB_469366
APC anti-mouse CD11c (clone N418)	eBioscience	Cat#17-0114-82; RRID: AB_469346
APC anti-mouse Granzyme B (clone QA16A02)	Biolegend	Cat#372204; RRID: AB_2687028
APC anti-mouse T-bet (clone 4B10)	Biolegend	Cat#644814; RRID: AB_10901173
R718 anti-mouse Bcl6 (clone K112-91)	BD Biosciences	Cat#567413; RRID: AB_2869985
APC/Cyanine7 anti-mouse TCRβ (clone H57-597)	Biolegend	Cat#109220; RRID: AB_893624
APC/Cyanine7 anti-mouse B220 (clone RA2-6B2)	Biolegend	Cat#103224; RRID: AB_313007
APC/Cyanine7 anti-mouse CD90.2 (clone 53.21)	BD Biosciences	Cat#561641; RRID: AB_10898013
APC/Cyanine7 anti-mouse NK1.1 (clone PK136)	Biolegend	Cat#108724; RRID: AB_830871
APC/eFlour 780 anti-mouse CD11b (clone M1/70)	Invitrogen	Cat#47-0112-82; RRID: AB_1603193
APC/Fire 750 anti-mouse FR4 (clone 12A5)	Biolegend	Cat#125013; RRID: AB_2721484
PE anti-mouse FoxP3 (clone FJK-16S)	eBioscience	Cat#12-5773-82; RRID: AB_465936
PE anti-mouse I-Ak (clone 10-3.6)	Biolegend	Cat#109908; RRID: AB_313457
PE anti-mouse LPAM-1 (α4β7) (clone DATK32)	Biolegend	Cat#120605; RRID: AB_493268
PE anti-mouse TNFα (clone MP6-XT22)	BD Biosciences	Cat#554419; RRID: AB_395380
PE anti-mouse CCL5 (clone 2E9)	Biolegend	Cat#149103; RRID: AB_2564405
PE anti-mouse Gata3 (clone TWAJ)	eBioscience	Cat#12-9966-42; RRID: AB_1963600
PE anti-mouse CD103 (clone M290)	BD Biosciences	Cat#557495; RRID: AB_396732
PE Rat anti-mouse CD8α	Biolegend	Cat#100708; RRID: AB_312747
PE/Dazzle 594 anti-mouse CD215 (IL-15Rα) (clone 6B4C88)	Biolegend	Cat#153511; RRID: AB_2922479
PE/FireTM 640 anti-mouse Lag3 (clone C9B7W)	Biolegend	Cat#125247; RRID: AB_2924464
PE/Cyanine5 anti-mouse CD11c (clone N418)	Biolegend	Cat#117316; RRID: AB_493566
PE/Cyanine7 anti-mouse CD199 (CCR9) (clone CW-1.2)	Biolegend	Cat#128711; RRID: AB_10901176
PE/Cyanine7 anti-mouse CD45.1 (clone A20)	eBioscience	Cat#25-0453-82; RRID: AB_469629
PE/Cyanine7 anti-mouse CD45 (clone 30-F11)	eBioscience	Cat#25-0451-82; RRID: AB_2734986
PE/Cyanine7 anti-mouse IL-12p40 (clone C17.8)	eBioscience	Cat#25-7123-82; RRID: AB_2543528
PE/Cyanine7 anti-mouse CD11c (clone N418)	eBioscience	Cat#25-0114-82; RRID: AB_469590
PE/Cyanine7 anti-mouse CD73 (clone TY/11.8)	Biolegend	Cat#127223; RRID: AB_2716102
FITC anti-mouse Vβ8.1/8.2 (clone KJ-16)	Biolegend	Cat#118406; RRID: AB_1227786
FITC anti-mouse CD4 (clone RM4-5)	Biolegend	Cat#100510; RRID: AB_312713
FITC anti-mouse Vβ6 (clone RR4-7)	BD Biosciences	Cat#553193; RRID: AB_394700
Spark BlueTM 550 anti-mouse CD8α (clone 53-6.7)	Biolegend	Cat#100779; RRID: AB_2832268
PerCP/eFluorTM710 anti-mouse I-Ab (MHCII) (clone M5/114)	eBioscience	Cat#46-5321-82; RRID: AB_1834439
PerCP/eFluorTM710 anti-mouse Vβ5.1/5.2 (clone MR9-4)	eBioscience	Cat#46-5796-82; RRID: AB_10853817
PerCP/Cyanine5.5 anti-mouse CD90.1 (clone OX-70)	Biolegend	Cat#202516; RRID: AB_961437
BV421 anti-mouse Vβ4 (clone KT4)	BD Biosciences	Cat#743019; RRID: AB_2741216
BV421 anti-mouse CD103 (clone OX-62)	BD Biosciences	Cat#566297; RRID: AB_2739670
BV421 anti-mouse TNFα (clone MP6-XT22)	BD Biosciences	Cat#563387; RRID: AB_2738173
BV421 anti-mouse CD45 (clone 30-F11)	BD Biosciences	Cat#563890; RRID: AB_2651151
BV421 anti-mouse RORγT (clone Q31-378)	BD Biosciences	Cat#562894; RRID: AB_2687545
BV421 anti-mouse CCR7 (clone 4B12)	BD Biosciences	Cat#562675; RRID: AB_2737716
Pacific BlueTM Anti-rabbit IgG	Invitrogen	Cat#P-10994; RRID: AB_2539814
BV480 anti-mouse CD4 (GK1.5)	BD Biosciences	Cat#746475; RRID: AB_2743777
BV510 anti-mouse CD62L (clone MEL-14)	Biolegend	Cat#104441; RRID: AB_2561537
BV570 anti-mouse CD44 (clone IM7)	Biolegend	Cat#103037; RRID: AB_10900641
BV605 anti-mouse CD8α (clone 53-6.7)	Biolegend	Cat#100744; RRID: AB_2562609
BV605 anti-mouse CD4 (clone RM4-5)	BD Biosciences	Cat#563151; RRID: AB_2687549
BV605 anti-mouse IFNγ XMG1.2)	Biolegend	Cat#505839; RRID: AB_2561438
BV605 anti-mouse CD86 (GL-1)	Biolegend	Cat#105037; RRID: AB_11204429
BV711 anti-mouse TCRβ (H57-597)	Biolegend	Cat#109243; RRID: AB_2629564
BV711 anti-mouse B220 (clone RA2-6B2)	Biolegend	Cat#103255; RRID: AB_2563491
BV711 anti-mouse CD80 (clone 16-10A)	Biolegend	Cat#104743; RRID: AB_2810338
BV785 anti-mouse PD-1 (clone 29F.1A12)	Biolegend	Cat#135225; RRID: AB_2563680
BV785 anti-mouse CD8α (clone 53-6.7)	Biolegend	Cat#100749; RRID: AB_11218801
BV786 anti-mouse CD45 (clone 30-F11)	BD Biosciences	Cat#564225; RRID: AB_2716861
BV786 Streptavidin	BD Biosciences	Cat#563858; RRID: AB_2869529
AF488TM anti-mouse CD11c (clone N418)	eBioscience	Cat#53-0114-82; RRID: AB_469903
AF488TM anti-mouse F4/80 (clone BM8)	Biolegend	Cat#123120; RRID: AB_893479
AF488TM anti-mouse Anxa2 (clone EPR13052)	Abcam	Cat#Ab200791; RRID: AB_2938814
AF647TM anti-mouse CD11b (clone M1/70)	Biolegend	Cat#101218; RRID: AB_389327
AF647TM anti-mouse CD29 (β1) (clone HA2/5)	BD Biosciences	Cat#562153; RRID: AB_10896298
AF647TM anti-mouse Fcer1g (clone E12)	Santa Cruz	Cat#Sc-390222; RRID: AB_2938815
AF700TM anti-mouse CD11c (clone N418)	eBioscience	Cat#56-0114-82; RRID: AB_493992
Anti-mouse DUSP2	Thermo Fisher	Cat#PA5-26093; RRID: AB_2543593
Biotin anti-mouse B220 (clone RA3-6B2)	BD Biosciences	Cat#553086; RRID: AB_394616
Biotin anti-mouse CD11c (clone HL3)	BD Biosciences	Cat#553800; RRID: AB_395059
Biotin anti-mouse NK1.1 (clone PK136)	BD Biosciences	Cat#553163; RRID: AB_394675
Biotin anti-mouse CD4 (clone GK1.5)	BD Biosciences	Cat#553728; RRID: AB_395012
Biotin anti-mouse CD8α (clone 53-6.7)	BD Biosciences	Cat#553029; RRID: AB_394567
Biotin anti-mouse CD11b (clone M1/70)	BD Biosciences	Cat#553309; RRID: AB_394773
Biotin anti-mouse CD25 (clone 7D4)	BD Biosciences	Cat#553070; RRID: AB_394602
Biotin anti-mouse TER-119 (clone TER-119)	BD Biosciences	Cat#553672; RRID: AB_394985
Biotin anti-mouse CD44 (clone IM7)	Biolegend	Cat#103003; RRID: 312954
Biotin anti-mouse CD49d (a4) (clone R1-2)	Biolegend	Cat#103603; RRID: AB_313034
Guinea pig anti-mouse insulin	Invitrogen	Cat#PA1-26938; RRID: AB_794668
Rabbit anti-mouse CD3	Abcam	Cat#ab5690; RRID: AB_305055
AF488TM donkey anti-guinea pig	Jackson Immunoresearch	Cat#706-545-148; RRID: AB_2340472
CyTM3 donkey anti-rat	Jackson Immunoresearch	Cat#712-165-153; RRID: AB_2340667
AF647TM donkey anti-rabbit	Invitrogen	Cat#A31573; RRID: AB_2536183
Bacterial and virus strains
Type 1 Lang (T1L)	Dermody Lab (University of Pittsburgh)	N/A
T3SA+	Dermody Lab (University of Pittsburgh)	N/A
MNV CW3	Randall Lab (University of Chicago)	N/A
Biological samples
Bovine serum albumin (BSA)	Thermo FIsher	Cat#BP1600
Normal donkey serum (NDS)	Jackson ImmunoResearch	Cat#017-000-121
Newborn calf serum (NCS)	GeminiBio	Cat#100-504
Penicillin-streptomycin solution	Corning	Cat#30001CI
Collagenase, Type IV	Thermo Fisher	Cat#17104019
Collagenase, Type D	Millipore Sigma	Cat#11088866001
Collagenase, Type VIII	Millipore Sigma	Cat#C2139
Albumin from chicken egg white	Millipore Sigma	Cat#A5378
Ovalbumin EndoFitTM	InvivoGen	Cat#vac-pova
Chemicals, peptides, and recombinant proteins
TRIzolTM Reagent	Thermo Fisher	Cat#15596018
DNasel	Millipore Sigma	Cat#10104159001
Betaine (5M)	Millipore Sigma	Cat#B0300
Histopaque-1077	Millipore Sigma	Cat#10770
RNAlater	Millipore Sigma	Cat#R0901
Omeprazole	Millipore Sigma	Cat#O104
Tamoxifen	Millipore Sigma	Cat#T5648
Diphtheria toxin	Millipore Sigma	Cat#D0564
Streptozotocin (STZ)	Caymen Chemicals	Cat#13104
Low melt agarose	Goldbio	Cat#A20405
Clear mounting media	Thermo Fisher	Cat#OB010020
2-mercaptoethanol	Sigma Aldrich	Cat#M7522
RPMI 1640, powder	Thermo Fisher	Cat#31800105
Percoll	GE Healthcare	Cat#17089109
HEPES Buffer	Corning	Cat#25060CI
HIP2.5 peptide (LQTLALWSRMD)	Lifetein Technologies	N/A, custom order
LIVE/DEADTM Fixable Near-IR	Thermo Fisher	Cat#L10119
LIVE/DEADTM Fixable Aqua	Thermo Fisher	Cat#L34965
Zombie NIRTM	Biolegend	Cat#423105
CellTraceTM Violet	Thermo Fisher	Cat#C34557
CellTraceTM CFSE	Thermo Fisher	Cat#C34554
Trypan blue solution, 0.4%	Thermo Fisher	Cat#15250061
Buffer TCL	Qiagen	Cat#1031576
Critical commercial assays
Fixation/permeabilization Concentrate	Thermo Fisher	Cat#00-5123-43
Fixation/permeabilization Diluent	Thermo Fisher	Cat#00-5223-56
Cytofix/cytopermTM Fixation and Permeabilization Solution	BD Biosciences	Cat#554722
Permeabilization Buffer (10X)	Thermo Fisher	Cat#00-8333-56
GolgiPlugTM Protein Transport Inhibitor	BD Biosciences	Cat#555029
eBioscienceTM Cell Stimulation Cocktail	Thermo Fisher	Cat#00-4975-93
SuperscriptTM III One-Step RT-PCR System with PlatinumTM Taq High Fidelity DNA Polymerase	Thermo Fisher	Cat#12574035
Maxima H Minus Reverse Transcriptase (200 U/μL)	Thermo Fisher	Cat#EP0753
RNasinTM Plus RNase Inhibitor	Promega	Cat#N2611
HiFi HS Ready Mix 2X	KAPA Biosystems	Cat#K2601K
RNAClean XP beads	Beckman Coulter	Cat#A63987
Ultra-Sensitive Mouse Insulin ELISA kit	Crystal Chem	Cat#90080
Mouse IL-15 DuoSet ELISA	R&D Systems	Cat#DY447-05
AldefluorTM Kit	Stem Cell Technologies	Cat#01700
Anti-biotin Microbeads	Miltenyi Biotec	Cat#130090485
Deposited data
Bulk RNA-seq	This study	GEO: GSE209701
scRNA-seq	This study	GEO: GSE209702
Experimental models: Organisms/strains
Mouse: C57BL/6J (B6)	Jackson Laboratory	MGI:3028467
Mouse: B6.Cg-Tg(TcraTcrb)425Cbn/J (OT-II)	Mucida Lab (Rockefeller University)	MGI:2174541
Mouse: C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-I)	Jackson Laboratory	MGI:2174540
Mouse: B6.SJL-Ptprca Pepcb/BoyJ (CD45.1)	Jackson Laboratory	MGI:2164701
Mouse: C57BL/6-Tg(Ins2-TFRC/OVA)296Wehi/WehiJ (RIPmOVA)	Jackson Laboratory	MGI:3522666
Mouse: B6.Cg-Tg(Itgax-Venus)1Mnz/J (CD11c-YFP)	Nussenzweig Lab (Rockefeller University)	MGI:3839312
Mouse: B6.129S6(Cg)-Ptf1atm2(cre/ESR1)Cvw/J (Ptf1aCreERT2)	Jackson Laboratory	MGI:6719359
Mouse: B6.Cg-Tg(Vil1-cre/ERT2)23Syr/J (VillinCreERT2)	Jackson Laboratory	MGI:6278020
Mouse: B6.Cg-Tg(Vil1-cre)1000Gum/J (Villincre)	Jackson Laboratory	MGI:5474782
Mouse: B6.Cg-Speer6-ps1Tg(Alb-cre)21Mgn/J (Albcre)	Jackson Laboratory	MGI:2164673
Mouse: B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (TdTomato)	Jackson Laboratory	MGI:3813512
Mouse: B6.Cg-Gt(ROSA)26Sortm6(CAG-ZsGreen1)Hze/J (ZsGreen)	Jackson Laboratory	MGI:3813510
Mouse: C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J (iDTR)	Jackson Laboratory	MGI:3772860
Mouse: NOD/ShiLtJ (NOD)	Jackson Laboratory	MGI:2162056
Mouse: NOD.NON-Thy1a/1LtJ (NOD Thy1.1)	Jackson Laboratory	MGI:2386298
Mouse: NOD.129S7(B6)-Rag1tm1Mom/J (NOD Rag1−/−)	Jackson Laboratory	MGI:2165365
Mouse: NOD.Cg-Tg(TcraBDC2.5,TcrbBDC2.5)1Doi/DoiJ (NOD BDC2.5)	Jackson Laboratory	MGI:2183909
Mouse: NOD.Cg-Tg(TcraTcrbNY8.3)1 Pesa/DvsJ (NOD NY8.3)	Jackson Laboratory	MGI:3603341
Mouse: NOD.G9C8	Chervonsky Lab (University of Chicago)	N/A
Oligonucleotides
Aldh1a2_Fwd	Integrated DNA Technologies	ATGGATGCGTCTGAAAGAGG
Aldh1a2_Rev	Integrated DNA Technologies	TGACTCCCTGCAAATCGATG
Ly6a_Fwd	Integrated DNA Technologies	GGATGGACACTTCTCACACTAC
Ly6a_Rev	Integrated DNA Technologies	GCAGGTAATTGATGGGCAAG
Anxa2_Fwd	Integrated DNA Technologies	GTCTACTGTCCACGAAATCCTG
Anxa2_Rev	Integrated DNA Technologies	ACTCCTTTGGTCTTGACTGC
Ccl6_Fwd	Integrated DNA Technologies	TCTTTATCCTTGTGGCTGTCC
Ccl6_Rev	Integrated DNA Technologies	ATGGGATCTGTGTGGCATAAG
Ccl5_Fwd	Integrated DNA Technologies	GGGTACCATGAAGATCTCTGC
Ccl5_Rev	Integrated DNA Technologies	TCTAGGGAGAGGTAGGCAAAG
Apol7c_Fwd	Integrated DNA Technologies	TGTATGCACAGATGGTCACTG
Apol7c_Rev	Integrated DNA Technologies	CAGGCAGAGGACTTAATAGGTG
36B4_Fwd	Integrated DNA Technologies	GCCGTGATGCCCAGGGAAGAC
36B4_Rev	Integrated DNA Technologies	CATCTGCTTGGAGCCCACGTT
Isg15_Fwd	Integrated DNA Technologies	GGTGTCCGTGACTAACTCCAT
Isg15_Rev	Integrated DNA Technologies	TGGAAAGGGTAAGACCGTCCT
Stat1_Fwd	Integrated DNA Technologies	AGGGGCCATCACATTCACAT
Stat1_Rev	Integrated DNA Technologies	AGATACTTCAGGGGATTCTC
Mx1_Fwd	Integrated DNA Technologies	GACCATAGGGGTCTTGACCAA
Mx1_Rev	Integrated DNA Technologies	AGACTTGCTCTTTCTGAAAAGCC
Il15ra_Fwd	Integrated DNA Technologies	CTACATCGGTCCTCTTGGTTG
Il15ra_Rev	Integrated DNA Technologies	GCCCTCACAGTCATTGGTAC
Ifit3_Fwd	Integrated DNA Technologies	CTGAAGGGGAGCGATTGATT
Ifit3_Rev	Integrated DNA Technologies	AACGGCACATGACCAAAGAGTAGA
Fcer1g_Fwd	Integrated DNA Technologies	TATAGCCAGCCGTGAGAAAGC
Fcer1g_Rev	Integrated DNA Technologies	AGCCAAGCACGTCTGTTCTG
Il12b_Fwd	Integrated DNA Technologies	ACTCCCCATTCCTACTTCTCC
Il12b_Rev	Integrated DNA Technologies	CATTCCCGCCTTTGCATTG
S100a10_Fwd	Integrated DNA Technologies	GCACTAGCCTCATCGTGG
S100a10_Rev	Integrated DNA Technologies	CTGCAAACCTGTGAAACGTAAG
Hic1_Fwd	Integrated DNA Technologies	GCATCTTGCTCCCGTCTTCC
Hic1_Rev	Integrated DNA Technologies	CCCAGCACACTCTCCCGATTTA
Asb2_Fwd	Integrated DNA Technologies	TGCAGAGAACACCTGGATTG
Asb2_Rev	Integrated DNA Technologies	GTATCGCACCAATATCCTCACC
Egr1_Fwd	Integrated DNA Technologies	TGAGCACCTGACCACAGAGTC
Egr1_Rev	Integrated DNA Technologies	TGAAAAGGGGTTCAGGCCAC
Ifitm6_Fwd	Integrated DNA Technologies	ACATCTACTCGGTGAAGTCCAGG
Ifitm6_Rev	Integrated DNA Technologies	GGCGGTTGAAGCATGGGATT
Ier2_Fwd	Integrated DNA Technologies	AAGAGGAAGTGCTGCGAGTC
Ier2_Rev	Integrated DNA Technologies	TAGACGGGCCTTCTTGCTTG
Arf4_Fwd	Integrated DNA Technologies	GCTGTGCTGCAGAAAATGCTTC
Arf4_Rev	Integrated DNA Technologies	AGCACAAGTGGCTTGGACAT
Ppp1r15a_Fwd	Integrated DNA Technologies	GCCTGTGAAACATTGCGTCC
Ppp1r15a_Rev	Integrated DNA Technologies	CCATGTGTCTGGGCGGC
Ccr7_Fwd	Integrated DNA Technologies	AAAGCACAGCCTTCCTGTGT
Ccr7_Rev	Integrated DNA Technologies	AGTCCACCGTGGTATTCTCG
Id3_Fwd	Integrated DNA Technologies	AGCTTTTGCCACTGACCC
Id3_Rev	Integrated DNA Technologies	AGATCGAAGCTCATCCATGC
Cd34_Fwd	Integrated DNA Technologies	CTGACTTGAGAAAGCTGGGGAT
Cd34_Rev	Integrated DNA Technologies	AGCCATCAAGGTTCCAGCTC
Emp3_Fwd	Integrated DNA Technologies	CCTGTCCTTCATCCTCTTCATG
Emp3_Rev	Integrated DNA Technologies	GGTGTGGATGGCATAGATGAG
Apol10b_Fwd	Integrated DNA Technologies	GGAGCCTGATAACTGAAGATGG
Apol10b_Rev	Integrated DNA Technologies	CTCCTGTGCTAAACTCTCCTTC
Dusp2_Fwd	Integrated DNA Technologies	GCGGTTTCAAAAGCTTCCAG
Dusp2_Rev	Integrated DNA Technologies	TAGGGCAAGATTTCCACAGG
S100a4_Fwd	Integrated DNA Technologies	TGAACAAGACAGAGCTCAAGG
S100a4_Rev	Integrated DNA Technologies	GAAGACACAGTACTCCTGGAAG
S100a6_Fwd	Integrated DNA Technologies	ACAAGTACTCTGGCAAGGAAG
S100a6_Rev	Integrated DNA Technologies	GATCCTTGTTACGGTCCAGATC
Gpr183_Fwd	Integrated DNA Technologies	GCCTATCACAGTCATTCTCCTG
Gpr183_Rev	Integrated DNA Technologies	CACAGGATGAACACGACAATG
Fth1_Fwd	Integrated DNA Technologies	TCAACCGCCAGATCAACC
Fth1_Rev	Integrated DNA Technologies	TCAGTTTCTCGGCATGCTC
Ftl1_Fwd	Integrated DNA Technologies	CAGCCATGACCTCTCAGATTC
Ftl1_Rev	Integrated DNA Technologies	CCACGTCATCCCGATCAAAA
Ebi3_Fwd	Integrated DNA Technologies	CAAGGAACAGAGCCACAGAG
Ebi3_Rev	Integrated DNA Technologies	GGGATACCGAGAAGCATGG
Cd40_Fwd	Integrated DNA Technologies	CGGTCCATCTAGGGCAGTGT
Cd40_Rev	Integrated DNA Technologies	CTGGCTGGCACAAATCACAG
Ifnar1_Fwd	Integrated DNA Technologies	TCTCTGTCATGGTCCTTTATGC
Ifnar1_Rev	Integrated DNA Technologies	CTCAGCCGTCAGAAGTACAAG
Ifnar2_Fwd	Integrated DNA Technologies	GTGACAGATAAGTGGTTGGAGG
Ifnar2_Rev	Integrated DNA Technologies	ACGATCTCAAATTCTGGCGG
Il15_Fwd	Integrated DNA Technologies	CAT ATGGAATCCAACTGGAT AGATGTAAGATA
Il15_Rev	Integrated DNA Technologies	CATATGCTCGAGGGACGTGTTGATGAACAT
T1L_Fwd	Thermo Fisher	CGCTTTTGAAGGTCGTGTATCA
T1L_Rev	Thermo Fisher	CTGGCTGTGCTGAGATTGTTTT
T1L_probe	Thermo Fisher	FAM-AGCGCGCAAGAGGGATGGGA-BNFQ
Software and algorithms
GraphPad Prism (v9.3.1)	GraphPad	www.graphpad.com
FlowJo (v10.8.1)	Tree Star	https://www.flowjo.com/
Adobe Illustrator (v25.4.1)	Adobe	www.adobe.com/products/illustrator
R (v4.0.5)	The R Foundation	https://www.r-project.org/
Kallisto	Bray et al.	https://pachterlab.github.io/kallisto/about
DE-Seq2 (v1.30.1)	Love et al.	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
CellRanger (v3.0)	10X Genomics	https://www.10xgenomics.com/support
Seurat (v4.1.0)	Hao et al.	https://github.com/satijalab/seurat
Devices
BD LSRII	BD Biosciences	N/A
4-laser Aurora spectral flow cytometer	Cytek Bioscience	N/A
Vibrating blade microtome	Leica	N/A
EVOSTM FL fluorescent microscope	Invitrogen	N/A
SP5 confocal microscope	Leica	N/A

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

B6 (C57BL/6J), NOD (NOD/ShiLtJ), OT-I (C57BL/6-Tg(TcraTcrb)1100Mjb/J), CD45.1 (B6.SJL-Ptprca Pepcb/BoyJ), RIPmOVA (C57BL/6-Tg(Ins2-TFRC/OVA)296Wehi/WehiJ), iDTR (C57BL/6-Gt(ROSA)26Sor^{tm1(HBEGF)Awai}/J), Ptf1a^CreERT (B6.129S6(Cg)-Ptf1a^{tm2(cre/ESR1)Cvw}/J), Vil1^CreERT2 (B6.Cg-Tg(Vil1-cre/ERT2)23Syr/J), Vil1^Cre (B6.Cg-Tg(Vil1-cre)1000Gum/J), Alb^Cre (B6.Cg-Speer6-ps1^{Tg(Alb-cre)21Mgn}/J), TdTomato (B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J), ZsGreen (B6.Cg-Gt(ROSA)26Sor^{tm6(CAG-ZsGreen1)Hze}/J), NOD BDC2.5 (NOD.Cg-Tg(TcraBDC2.5,TcrbBDC2.5)1Doi/DoiJ), NOD NY8.3 (NOD.Cg-Tg(TcraTcrbNY8.3)1Pesa/DvsJ), NOD Thy1.1 (NOD.NON-Thy1^a/1LtJ), and NOD Rag1−/− (NOD.129S7(B6)-Rag1tm1Mom/J) were purchased from the Jackson Laboratories and maintained in house. CD11c-YFP (B6.Cg-Tg(Itgax-Venus)1Mnz/J) mice were provided by M. Nussenzweig (The Rockefeller University). OT-II TCR-transgenic were provided by D. Mucida (The Rockefeller University). NOD.G9C8⁵⁵ mice were provided by A. Chervonsky (University of Chicago). TCR-transgenic mice were maintained as heterozygotes. Vil1^CreERT2, Vil1^Cre, Alb^Cre, and Ptf1a^CreERT2 mice were maintained as homozygotes and bred to homozygous TdTomato or ZsGreen mice to generate Vil1^{Tomato/ZsGreen}, Alb^{Tomato/ZsGreen}, and Ptf1a^{Tomato/ZsGreen} mice. Male and female mice (on a B6 background) were used between 6–12 weeks of age. Female NOD mice between 6–12 weeks of age were exclusively used for this study due to their increased propensity for diabetes compared to male NOD mice⁶¹. Mice were maintained at the University of Chicago animal facilities under specific pathogen-free conditions.

METHOD DETAILS

Infections

Reovirus strains T1L and T3SA+ were recovered using reverse genetics, plaque-purified, and titered by plaque assay as described^45,46. Mice were infected with 2×10⁸ (for NY8.3 and OT-I transfer experiments) or 2×10⁹ PFU T1L (for DC phenotyping, BDC2.5 and OT-II transfer experiments), or 4×10⁹ PFU T3SA+ diluted in PBS by oral gavage or retro-orbital injection. MNV CW3 was provided by the Randall Lab (University of Chicago). CW3 was recovered by a reverse genetics by transfection of 293T cells and then further amplified in BV2 cells as described⁶². After removing cell debris, supernatants were ultracentrifuged to concentrate MNV, and the titer was assessed by TCID₅₀. Mice were infected with 5×10⁷ TCID₅₀ units of MNV CW3 diluted in plain RPMI by oral gavage. Mice were harvested at the following time points after infection unless otherwise stated: 48 h for dendritic cell phenotyping, 72 h for CD8 T cell phenotyping, and 96 h for CD4 T cell phenotyping.

Antigen administration

Prior to peptide gavage, mice were given omeprazole dissolved in PBS containing 1% Tween-80. Omeprazole was administered intraperitoneally (i.p.) 24 h before first gavage and 15 min before each consecutive gavage. Ovalbumin (OVA) was diluted in sterile PBS or H₂O and filter sterilized prior to administration. Mice received 50 mg OVA in 200 μl PBS by oral gavage daily for 3 days, and OVA was added to drinking water at a concentration of 5 mg/ml until harvest. HIP2.5 peptide was diluted in PBS containing 1% Tween-80, and mice received 20 mg HIP2.5 in 200 μl PBS per day for three days by oral gavage.

For intravenous antigen administration, endotoxin-free ovalbumin (InvivoGen) was dissolved in sterile PBS. Each mouse received 1 mg OVA administered by retro-orbital injection.

Streptozotocin (STZ), Diptheria toxin (DT), and Tamoxifen administration

Prior to STZ treatment, mice were fasted for 4–5 hours. STZ was dissolved in sodium citrate buffer immediately prior to i.p. injection. Sodium citrate buffer was prepared by adding 1.47 g sodium citrate to 50 ml H2O and adjusting pH to 4.5 with citric acid. Mice were either given consecutive low doses (20–40 mg/kg) for NY8.3 experiments or a single high dose (150 mg/kg) for BDC2.5 experiments.

DT was dissolved in PBS and administered i.p. at a dose of 1000 ng/mouse.

Tamoxifen was dissolved first as 1 g/ml in methanol and then diluted to 10 mg/ml in corn oil. Mice received consecutive doses of 1 mg daily for 5 days administered i.p.

Adoptive T cell transfer

Naïve T cells were isolated from pooled lymph nodes and spleens by negative selection using biotinylated antibodies against NK1.1, B220, CD11c, CD11b, TER119, CD8α, CD4, and CD44 and anti-biotin MACs beads (Miltenyi Biotec). Purity of transgenic cells was validated by flow cytometry (CD45.1⁺Vβ5⁺CD25^– for OT-II and OT-I cells, Thy1.1⁺Vβ8.1⁺ for NY8.3 cells, Thy1.1⁺Vβ4⁺ for BDC2.5 cells, CD45.1⁺ Vβ6⁺ for G9C8 cells) and viability was evaluated using Trypan blue staining. T cells were labeled with Cell Trace^™ Violet or CFSE Cell Proliferation Kit (Life Technologies) by incubating cells for 3–5 mins in 37°C water bath. For all adoptive T cell transfers, 750k cells were given by retro-orbital injection under isoflurane anesthesia. For BDC2.5 and OT-II experiments, mice were harvested 4 days after transfer later unless otherwise stated. For NY8.3, OT-I, and G9C8 experiments, mice were harvested 3 days after transfer. In case of infection, mice received cells and were infected on the same day.

Splenocyte transfer into Rag KO mice

Whole splenocytes were isolated from a NY8.3 donor mouse and characterized by flow cytometry using the following markers: B220, TCRβ, CD4, CD8α, Vβ8.1/2, Thy1.1. Splenocytes were adjusted to contain approximately 7,000 NY8.3 T cells. Cells were transferred retro-orbitally under isoflurane anesthesia.

APC isolation from lymph nodes

For APC analysis, lymph nodes were dissected into RPMI containing 2% NCS, 1% HEPES. Lymph nodes were finely chopped and digested with 2.5 mg/ml Collagenase D for 30 min at 37°C. Cells were used for downstream surface marker staining immediately. Alternatively, for intracellular cytokine staining, pellets were resuspended in RPMI containing 2% NCS, 1% HEPES and Golgi Plug at 1:1000 and incubated at 37°C for 6 h prior to staining.

T cell stimulation

Lymph nodes were dissected into RPMI containing 2% NCS and 1% penicillin/streptomycin. Lymph nodes were smushed between frosted glass slides and transferred to 96 well plates for centrifugation. Pellets were resuspended in the same media as above containing cell stimulation cocktail at 1:500 and incubated at 37°C for 4 h. After incubation, cells were spun down and washed once with fresh media before proceeding to cell staining.

Cell isolation from small intestine

The small intestine was removed from the mesentery and Peyer’s patches and feces were removed. The small intestine was roughly divided into four quarters and the proximal ¼ was taken as duodenum. Intestines were cut longitudinally and washed three times in PBS containing 1 uM DTT to remove mucus. The tissue was cut into 1 cm pieces and incubated in PBS with 30 mM EDTA for 10 min at 37°C at 230 rpm with vigorous shaking after the incubation. Tissues were transferred to a new tube containing fresh PBS and incubated for 10 min at 37°C at 230 rpm. Tissues were washed with PBS on a metal sieve, finely chopped and digested in RPMI containing 2% FCS, 1% HEPES, and 2 mg/ml Collagenase VIII, 200 μg/ml DNaseI for 30 min at 37°C at 80 rpm. Digests were quenched in cold RPMI with 2% FCS. Pellets were resuspended in 40% Percoll and separated by centrifugation in a discontinuous Percoll gradient (80%/40%) at 2300 rpm for 23 min at room temperature without brakes. APCs and lymphocytes were isolated from the interphase, washed, and stained for FACS analysis.

Cell isolation from pancreas and liver

The pancreas was perfused via the bile duct with 2.5 ml HBSS containing 1 mg/ml collagenase IV (Thermo Fisher) and 20 μg/ml DNaseI (Millipore Sigma). Lymph nodes were removed, and the pancreas was placed into a conical tube with remaining collagenase mixture. The liver was perfused via the portal vein with RPMI containing 2.5 mg/ml collagenase D and 20 μg/ml DNaseI. Several lobes of the liver were removed, chopped into small pieces, and placed into a conical tube with remaining collagenase solution. Pancreas and liver samples were then incubated in a water bath at 37°C for 25 min. The digestion was quenched with 20 ml RPMI with 10% NCS and tubes were shaken vigorously for 1 min before passing each sample through a mesh filter. Liver samples were spun for 1 min at 75 x g, discarding the pellet to remove parenchymal cells. Samples were passed through a mesh filter and spun at 800 x g for 2 min. Cells were washed with 20 ml RPMI with 10% FCS and spun once more at 800 x g for 2 min. Pellets were resuspended in 5 ml Histopaque 1077 and 10 ml of plain RPMI was carefully laid over the top. Samples were then spun at room temperature at 900 x g for 18 min without brakes. The white layer of the gradient was transferred to a new tube and washed with 35 ml RPMI with 10% FCS. Cells were spun at 800 x g for 2 min and then transferred to an Eppendorf tube for downstream staining.

DC sorting

To sort dendritic cells from B6 and NOD mice, 1000 migDC1s (live CD45⁺LIN-CD11c⁺MHCII^hiCD103⁺CD11b⁻) were sorted from the liver-, celiac-, duodenal-, and colonic-LNs (n=3, 2 mice pooled per replicate). For ZsGreen sorts, 100 ZsGreen⁺ migDC1s were sorted from the liver-, celiac-, and duodenal-LNs of Vil1^ZsGreen, Ptf1a^ZsGreen, and Alb^ZsGreen mice (n=3, 2 mice pooled per replicate for Ptf1a^ZsGreen mice). Cells were sorted into 25 μl TCL buffer containing 1% 2-mercaptoethanol, immediately spun down, and snap frozen on dry ice. Samples were kept at −80°C until further processing. All sorts were conducted at the University of Chicago Flow Cytometry Core using a BD FACSAria Fusion Cell Sorter.

Bulk RNAseq library preparation

Cells were sorted into TCL buffer containing 1% 2-mercaptoethanol. RNA was isolated using RNAClean XP beads (Agentcourt) on a magnetic stand. Reverse transcription primers were: P1-RNA-TSO: Biot-rArArUrGrArUrArCrGrGrCrGrArCrCrArCrCrGrArUrNrNrNrNrNrNrGrGrG, P1-T31: Biot-AATGATACGGCGACCACCGATCG31T, P1-PCR: Biot-GAATGATACGGCGACCACCGAT. RNA was eluted for 1 min in RT- cDNA synthesis mix 1 (0.5 μl P1-T31 (20 μM), 0.3 μl RNasin plus (Promega), 1.5 μl 10mM dNTP, 3.5 μl 10mM Tris pH 7.5–0.5% IGEPAL CA-630 (Sigma) and 1.7 μl RNase free ddH₂O) and pipetted up and down to mix. The eluted sample was then incubated for 3 min at 72°C, followed by 1 min on ice, then 7.5 μl of mix 2 was added (3 μl 5X RT Buffer, 0.375 μl 100 mM DTT, 0.375 μl RNasin plus, 0.5 μl P1-RNA-TSO (40 μM), 0.75 μl Maxima RT Minus H (Thermo Scientific), 1.8 μl 5M Betaine (Sigma), 0.9 μl 50 mM MgCl2 and 0.175 μl RNase free ddH₂O and mixed well. Samples were placed in a thermocycler and subjected to the following protocol: 42°C for 90 s, (50°C for 2 min, 42°C for 2 min) x10 cycles, 70°C for 15 min. The cDNA was then amplified using 13.5 μl cDNA, 20 μl KAPA HiFi 2x Mix, 1.5 μl P1-PCR, and 5 μl H2O. Samples were subjected to the following protocol: 98°C for 3 min, (98°C for 15 s, 67°C for 20 s, 72°C for 6 min) x12 cycles, 72°C for 5 min. cDNA was cleaned up using RNA Clean XP beads and eluted in water. The concentration was calculated using a Qubit fluorometer and the average fragment length of 1500–1800 was determined by Bioanalyzer. Samples were normalized to 0.1 ng/uL using ddH₂O, and 2.5 μl cDNA were tagmented using Nextera XT Index Kit according to the manufacturer’s protocol, except that all volumes were used at 0.5x of the indicated volumes. The concentration of the eluted samples was calculated by Qubit. Samples were pooled at 10 nM and handed off to the UChicago Genomics Core for sequencing using paired end 50 bp reads on a Novaseq S2 flowcell.

Single cell RNA sequencing library preparation

Between 30,000–100,000 YFP⁺ cells were sorted from the liver-, celiac- and duodenal-LNs of CD11c^YFP mice (n=3, 10 mice pooled per replicate). Cells were processed by the UChicago Genomics Core for library preparation and sequencing. In brief, cells were loaded onto sChromium NextGEM Chip (10x Genomics) for 3’ end library preparation using dual indexes according to manufacturer’s protocol (10x Genomics). The single cell RNAseq libraries were sequenced on a Novaseq SP100 flowcell.

Bulk RNA sequencing analysis

Raw fastq files were pseudo-aligned to the mouse reference transcriptome M26 (GRCm39) using Kallisto⁶³. The resulting gene counts were input to R and transcripts per million (tpm) was scaled by gene length. Initial filtering removed genes with expression under 3 counts per million (cpm). The un-normalized filtered gene counts were then used as input for the R package DESeq2 using the default parameters (Wald test) to compare genes differentially expressed across locations, correcting for replicate and isolation day batch effects⁶⁴. The apeglm method was used for log fold change shrinkage to better visualize effect sizes⁶⁵. Genes with log2 fold changes greater than 2 or less than −2 and a false discovery rate (FDR) of 0.05 were considered significant for downstream studies. To generate tissue-specific gene sets the significant genes for both pairwise comparisons (ex: Duodenum vs liver, duodenum vs pancreas) were overlapped in a Venn diagram using the R package Venn Diagram. The overlapping genes were extracted and used to generate heatmaps using the R package pheatmap. PCA plots were generated using tpm values using the R package PCAtools.

For Gene Ontology (GO) pathway analysis, genes that were found to be significantly different in liver and pancreas compared to duodenum were used as input to the Gene Ontology website (http://geneontology.org). The top 30 significantly enriched pathways are shown.

Single cell RNA sequencing analysis

Gene counts were calculated by aligning raw fastq files to the mm10 genome using CellRanger (10xGenomics). The counts were used as input to the R package Seurat⁶⁶. Cells were initially filtered to remove doublets and poor-quality cells based on the unique molecular identifiers (UMIs), number of features, and percentage of mitochondrial genes (<0.25). To normalize gene counts, Seurat’s LogNormalize function was used with default parameters (scale factor = 10,000). Variable genes were then identified using the Seurat function FindVariableFeatures with default parameters. At this point, samples were integrated into one object using the function FindIntegrationAnchors for all further downstream analysis. The data was scaled using ScaleData and then subjected to dimensionality reduction by PCA using RunPCA. The cells were clustered using Seurat’s FindNeighbors followed by FindClusters prior to visualization by RunUMAP. Clusters were identified using FindAllMarkers and the resulting gene lists defining each cluster were used as input into ImmGen database to identify the highest correlating cell type. We next subset the antigen presenting cells and repeated the steps from FindVariableFeatures through FindAllMarkers to generate a new UMAP. In addition to the ImmGen database, previous bulk RNA sequencing sets and APC subset markers^25,39,40 were used to call the different APC clusters. From this clustering, we further subset the data to include all migratory DCs (mig cDC1s, mig cDC2s, moDCs and one cluster that could not be differentiated as mig cDC1 or 2). Reclustering this subset of data yielded 9 clusters with each DC population corresponding to specific clusters. Each lymph node sample was downsampled to visualize the relative contribution of each cell type to the LN. Finally, we subset only migratory cDC1s for further downstream analysis. Heatmaps were generated using DoHeatmap using cluster averages for each gene. Specific gene expression was plotted using FeaturePlot, using blend to overlay two genes.

qPCR

DCs were sorted into TCL buffer containing 1% 2-mercaptoethanol. cDNA was prepared as described above for the RNAseq library preparation except without tagmentation and with minor changes: cDNA was amplified for 25 cycles at the amplification step, and samples were eluted in 60 μl after bead cleanup. Samples were normalized to 1 ng/μl and 1 μl was used as input for qPCR. Master mixes were prepared using Power Sybr Green (Invitrogen) containing 7.5 μl Sybr, 0.15 μl 10 μM forward and reverse primer mix, and water up to 14 μl. Samples were multichanneled into 384 well plates in duplicate and run on an Applied Biosystems QuantStudio 6 Flex machine. Primers were designed using IDT RealTime qPCR Primer Design tool. The Ct values obtained were confirmed to be in the linear range of a dilution curve and the primers were validated as generating a single product by confirming the melting curve only had one peak. Delta Ct was calculated by subtracting the average of duplicate values for each gene from the average duplicate values for the housekeeping gene 36B4 for each sample. Relative expression values were then calculated by the equation: 2^-ΔCT. Data are presented as 2^-ΔCTx10,000.

For reovirus detection in tissues and lymph nodes, mice were infected with T1L at a dose of 2×10⁹ PFU and harvested 48 h post-infection. Jejunum, pancreas, liver, and kidney samples were placed into RNAlater until RNA extraction with TRIzol reagent according to manufacturer’s instructions. Briefly, samples were homogenized in TRIzol using a bead beater, and RNA was extracted by adding chloroform to homogenate. After centrifuging, RNA was precipitated with isopropanol then washed with 75% EtOH. Final product was resuspended in ultrapure H₂O. For qPCR, reverse transcription was conducted using SuperScript III One-Step RT-PCR system (Invitrogen) and forward primer (S4-CGCTTTTGAAGGTCGTGTATCA) binding in the S4 region of the genome for 15 min at 50°C followed by termination of the reaction by incubation at 95°C for 3 min. Reverse primer (S4-CTGGCTGTGCTGAGATTGTTTT) and probe (FAM-AGCGCGCAAGAGGGATGGGA-BNFQ) were added to the reaction mix for qPCR amplification. Samples were run in duplicate, and qPCR was conducted with pre-incubation for 10 min at 95°C followed by 45 cycles (95°C for 15 s, 60°C for 30 s). Delta Ct was calculated by subtracting the average of duplicate values for each gene from the average duplicate values for the housekeeping gene GAPDH for each sample.

Flow cytometry

Cells were first stained for viability using LIVE/DEAD^™ Dead Cells Stain kits or Zombie NIR^™ Fixable Viability kit. For surface epitopes, cells were resuspended in antibody cocktail and incubated for 20 min at 4°C. Cells were incubated in Cytofix/cytoperm^™ solution for 20 min prior to staining with intracellular cytokine antibodies overnight. For transcription factor staining, cells were incubated in Fixation/permeabilization solution for 30 min, then stained for nuclear epitopes overnight. Intracellular and nuclear antibodies were diluted in 1x Permeabilization buffer rather than FACS buffer.

The Aldefluor (Stem cell technologies) assay was conducted according to manufacturer’s protocol following lymph node digestion in collagenase D. Briefly, a cell count was conducted, and cell concentration was adjusted to 1e6 cells/ml in Eppendorf tubes. 5 μl Aldefluor reagent was added to each tube prior to incubation at 37°C for 30 min. After centrifugation, cells were immediately stained for surface markers. Surface stain cocktail was prepared in Assay buffer to prevent efflux of Aldefluor reagent from cells.

Flow cytometry was conducted on an LSRII or 4-laser Aurora spectral flow cytometer and analyzed using FlowJo Software. Cell division index was calculated using the FlowJo formula (http://www.flowjo.com/v765/en/proliferation.html), whereby the index represents the fraction of total cell divisions over the calculated total starting cells.

Histology

Pancreatic tissue was fixed overnight in 4% PFA on a shaker. Samples were processed, paraffin-embedded, sectioned, and stained with hematoxylin/eosin by the University of Chicago Histology Core. Islets were scored for T cell infiltration by two independent lab members who were blinded to infection status. At least 20 islets were scored per pancreas. The following scoring system was used: no infiltration (0), peri-insulitis (1), less than 50% infiltrated (2), more than 50% infiltrated (3). >20 islets were scored per mouse to calculate average infiltration score.

Immunofluorescent microscopy

For ZsGreen system validation, duodenum, liver, and pancreas tissues were harvested from Alb^Cre-ZsGreen, Ptf1a^Cre-ZsGreen, and Vil1^Cre-ZsGreen mice and fixed in 2% PFA in PBS for 2 h at room temperature. Tissues were embedded in 4% low melt agarose in PBS, and 50 μm sections were cut with a vibrating blade microtome into an ice cold PBS bath. Sections were placed onto glass slides with mounting media and sealed under glass coverslips with nail polish. Images were taken with EVOS^™ FL fluorescent microscope.

For islet infiltration characterization, pancreata were harvested from NY8.3 mice and processed as described above. 50 μm sections were incubated in permeabilization (perm) wash containing 0.5% TritonX in PBS 3x for 20 min, then blocked for at least 3 h in blocking buffer containing 5% NDS and 5% BSA in perm wash. Sections were stained overnight in primary antibodies (Guinea pig anti-insulin, PE rat anti-CD8α, rabbit anti-CD3,) at 4°C. Sections were washed 3x for 20 min in perm wash, then incubated overnight in secondary antibodies (AF488^T donkey anti-guinea pig, Cy^™3 donkey anti-rat, AF647^™ donkey anti-rabbit) at 4°C. Sections were again washed 3x for 20 min in perm wash then mounted onto glass slides with mounting media and sealed under glass coverslips with nail polish. Images were taken using the SP5 confocal microscope.

ELISA

To measure total pancreatic insulin content, whole pancreata were homogenized in acid ethanol using tissue douncers. Following overnight incubation at −20°C, samples were neutralized with Tris-HCl (pH = 7.5) and diluted to appropriate concentration in sample diluent. Insulin ELISA was conducted according to the manufacturer’s instructions for wide-range assay. For cytokine ELISAs, whole lymph nodes were snap frozen on dry ice then homogenized in hypotonic cell lysis buffer containing 1% TritonX, 20 mM HEPES, 300 mM NaCl, 1.5 mM MgCl2, and 0.2 mM EDTA. Lysates were spun down and the supernatant was transferred to a new tube for analysis. Mouse IL-15 DuoSet ELISA kit was used to measure cytokine concentrations and conducted according to manufacturer’s instruction.

QUANTIFICATION AND STATISTICAL ANALYSIS

Except for the RNA sequencing data (see above) all data were analyzed with Prism software (GraphPad). Data is presented as average ± SEM. Multivariate data was analyzed by applying one-way ANOVA and Tukey’s multiple comparison post hoc test, comparison between two treatment conditions by two-tailed unpaired Student’s t-test assuming a Gaussian distribution. Gaussian distribution tests appropriate for small sample sizes were applied for all datasets (Shapiro-Wilk and Kolmogorov-Smirnov test); all data sets passed the requirements for Gaussian distribution besides some infected sets which came out to be non-significant. Survival curve was analyzed by Mantel-Cox, Gehan-Breslow-Wilcoxon and Logrank. P-values or representative symbols are noted when differences are less than or equal to 0.1; all other differences were not found to be significant (n.s., p > 0.1). *p < 0.05, **p < 0.01, *** p < 0.001, ****p < 0.0001). One-way ANOVA was used for multivariant comparisons between more than two groups. Two-way ANOVA with multiple comparisons was used to calculate significance in Figure 5G and 5H. Two-tailed t-test was used to compare two groups (i.e. NI vs. infected LNs).

Supplementary Material

Table S2

Table S2. Genes defining migDC clusters 0–7 (Table S2A-H, respectively) from scRNAseq, related to Figure 3A–C.

Table S3

Table S3. Cluster averages for all genes for the 8 clusters found in the migDC1 UMAP, related to Figures 3A–C.

Table S1

Table S1. Transcripts per million (TPM) of all genes from bulk RNAseq of purified migDC1s from liver, pancreas and duodenum, related to Figure 2B–D.