The XCL1–XCR1 axis supports intestinal tissue residency and antitumor immunity

Upon antigen recognition, CD8⁺ T cells undergo proliferation and differentiate into cytotoxic effector cells capable of killing infected cells. Effector T cells are heterogeneous and include a subset that establishes a memory population, which can persist for a lifetime following antigen clearance (Chang et al., 2014). The majority of memory T cells are not found in circulation but rather lodged in tissues (Steinert et al., 2015). These tissue-resident memory T cells (T_RM) facilitate rapid responses via their localization in non-lymphoid tissues (NLT), providing sentinel activity during a secondary response (Schenkel et al., 2014). Cell fate specification requires the coordinated activity of multiple cell-intrinsic and environmental factors, reflective of adaptations supporting NLT trafficking, differentiation, and recall response. T_RM display a hybrid of memory and effector cell characteristics, retaining cytotoxicity and recall potential despite the expression of factors associated with differentiated T cell subsets (Milner and Goldrath, 2018). Studies of T_RM suggest that these cells may represent an ideal population for therapeutic targeting in cancer (Okła et al., 2021). Indeed, we have previously shown that modulation of T_RM regulatory factors can improve antitumor immunity (Milner et al., 2017; Reina-Campos et al., 2023).

Recently, single-cell RNA sequencing (scRNA-seq) has revealed similar heterogeneity within tumor-infiltrating lymphocyte (TIL) populations, reflecting a spectrum of cell states linked to tissue surveillance, memory potential, and cytotoxicity (van der Leun et al., 2020; Giles et al., 2023). Upon differentiation, CD8⁺ TIL lose self-renewal capacity and polyfunctionality until reaching exhaustion, a hypofunctional T cell state driven by chronic antigen stimulation and immunosuppressive signals (Wherry et al., 2003). Progenitor-exhausted CD8⁺ T cells (T_PEX) are characterized by high Tcf7 expression and proliferate upon anti-PD1 treatment, giving rise to effector cells that eventually succumb to exhaustion (Hudson et al., 2019; Jansen et al., 2019; Miller et al., 2019; Siddiqui et al., 2019). The balance of these subsets has varying effects on the overall immune response depending on tumor type (Sade-Feldman et al., 2018; Miller et al., 2019; Escobar et al., 2023). Evidence from human patients and mouse in vivo studies suggests that differential interactions with antigen-presenting cells (APC) may regulate these states (Jansen et al., 2019; Stoltzfus et al., 2021; Wang et al., 2023). CD8⁺ TIL which resemble T_RM have been identified in both mouse and human studies and are consistently associated with improved prognoses (Ganesan et al., 2017; Savas et al., 2018; Park et al., 2019; Milner et al., 2020; Banchereau et al., 2021; Okła et al., 2021). While the degree of regulatory overlap between T cell exhaustion and residency is unclear, defining the signals that endow TIL with T_RM properties will enable the development of new immunotherapies (Schenkel and Pauken, 2023).

CD8⁺ T cell fates are conditioned by APC interactions during priming (Mempel et al., 2004). Conventional dendritic cells (DC) are segregated into two subsets found in mice and humans, cDC1 and cDC2. cDC1 are identified via expression of Clec9a, Itgae, Xcr1, and BDCA3 (human) and are key for antigen cross-presentation and pro-inflammatory cytokine production, while cDC2 promote CD4⁺ T cell responses (Wculek et al., 2020). Recent reports have studied the role of cDC1 in shaping T cell differentiation during chronic viral infection (Dähling et al., 2022; Domenjo-Vila et al., 2023). It was demonstrated that cDC1 are critical for the maintenance of the TCF1^hi state during anti-PD-L1 treatment. Loss of cDC1 led to a reduction of CXCR5 and CCR7 ligands, disrupting the localization of splenic T cells. It has become evident that DC subtypes modulate T cell differentiation in lymphoid and NLT, via cytokines (Iborra et al., 2016), and expression of TGFβ-activating integrins (Mani et al., 2019). In humans, CD1c⁺ CD163⁺ DCs preferentially induce T_RM features during in vitro co-culture, enabling accumulation in human epithelial grafts in mice (Bourdely et al., 2020). Whether CD8⁺ T cells utilize chemokines to facilitate DC contacts during T_RM formation is not well established.

Chemokines are secreted proteins that guide the spatial organization of immune cells (Ozga et al., 2021). In human tumors, CXCL13-producing T cells associate in tertiary lymphoid structures, lymph node–like aggregates that mediate DC–T cell interactions (Sautès-Fridman et al., 2019). The IFN-γ-inducible protein CXCL10 is also produced by CD8⁺ T cells during lymph node priming to amplify the effector response (Peperzak et al., 2013). While studies have observed differential expression of chemokines in T cell subsets during infection and cancer (Miller et al., 2019; Di Pilato et al., 2021), their contribution to the development of tissue residency is not well appreciated.

Here, we found that XCL1 promotes intestinal CD8⁺ T_RM formation and TIL activity utilizing acute lymphocytic choriomeningitis virus (LCMV-Arm) and melanoma (B16), respectively. Perturbation of this axis influences the distribution and differentiation of antigen-specific CD8⁺ T cells. Human data support modulating cDC1-CD8⁺ T cell interactions to improve the antitumor immune response. Collectively, our study demonstrates a pivotal role for XCL1 in facilitating intestinal T_RM development and supports efforts to study T_RM factors in oncology.

To identify genes relevant in T_RM biology, we overlayed gene lists derived from published bulk RNA-seq studies. First, we isolated genes induced in CD8⁺ T cells by TGFβ, an established regulator of T_RM differentiation (Zhang and Bevan, 2013; Nath et al., 2019). Second, we included the core T_RM signature (Milner et al., 2017) to identify genes broadly relevant to T_RM cell fate. Finally, we incorporated genes of T_PEX CD8⁺ T cells isolated from murine tumors (Miller et al., 2019). Xcl1 is conserved across all three datasets (Fig. 1 A), suggesting the functional relevance of this chemokine in tissue residency and T cell differentiation.

Further analysis of T cell datasets also demonstrated high expression of Xcl1 in the T_PEX population during chronic LCMV-Cl13 (Fig. S1 A). Previous assay for transposase-accessible chromatin sequencing (ATAC-seq) profiling of T_PEX cells demonstrated high chromatin accessibility at the Xcl1 locus compared to exhausted T cell subsets (Jadhav et al., 2019). We also confirmed that Xcl1 transcription was not primarily induced by tissue digestion (Fig. S1 B).

XCL1 is a chemokine produced by activated CD8⁺ T and natural killer (NK) cells (Dorner et al., 2002), and the cognate receptor XCR1 is selectively expressed by cross-presenting cDC1 cells (Dorner et al., 2009). Mice lacking XCR1⁺ DCs exhibit homeostatic defects in intestinal T cell populations, including TCRαβ⁺ CD103⁺ CD8⁺ T cells (Ohta et al., 2016). In tumors, loss of the cDC1 transcription factor Batf3 (Hildner et al., 2008) impairs antitumor immunity, in part due to defective DC trafficking of tumor antigen (Spranger et al., 2017). Additionally, NK cell–derived XCL1 facilitates cDC1 recruitment to the tumor microenvironment (TME), promoting antitumor immunity (Böttcher et al., 2018). However, the role of CD8⁺ T cell–derived XCL1 during infection and tumor challenge has not been thoroughly investigated.

Analysis of P14 T cell scRNA-seq following LCMV-Arm (Kurd et al., 2020) demonstrated that Xcl1 is highly expressed by CD8⁺ T cells isolated from the intraepithelial layer (IEL) of the small intestine (SI) compared with circulating CD8⁺ T cells (Fig. 1 B). Analysis of RNA-seq data profiling T_RM populations (Crowl et al., 2022) showed high Xcl1 expression in all T_RM compared with circulating P14 T cells, with the highest levels in the SI and salivary gland (Fig. S1 C). We confirmed the high expression of XCL1 and its paralog XCL2 in human ileum CD8⁺ T_RM (Boland et al., 2020) compared to peripheral blood mononuclear cells (PBMC) (Fig. 1 C). Human lung CD8⁺ T_RM also express XCL1, in line with Xcl1 being part of the core T_RM signature (Fig. S1 C). Thus, Xcl1 is preferentially expressed by mouse and human CD8⁺ T_RM, with elevated expression in the SI.

Given the selectivity of XCL1 for its cognate receptor XCR1 (Lei and Takahama, 2012), we asked whether SI CD8⁺ T cells can be observed near XCR1⁺ DC during viral infection. We adoptively transferred LCMV GP_33–41-specific P14 T cells into congenically distinct mice and infected them with LCMV-Arm. During the effector phase, T cells infiltrate the intestine, while cDC1 frequency decreases, likely reflecting antigen-dependent migration to lymphoid organs. At day 8 (D8), we observed P14 T cells and cDC1 in proximity in the lamina propria (LP). Following antigen clearance, P14 T cells contract and LP cDC1 frequencies return to baseline (Fig. 1 D). Interestingly, we also observed contacts between both cell types at D35 (Fig. 1 E). Given the high expression of Xcl1 at memory time points, we hypothesized a role for this pathway in T_RM differentiation and sought to perturb this chemokine axis.

scRNA-seq of P14 T cells during LCMV-Arm (Kurd et al., 2020) demonstrates that Xcl1 expression in the SI gradually increases and remains elevated at D90 (Fig. 2 A). We hypothesized that disrupting XCL1 production may influence memory T cell formation. Several studies have demonstrated functional interactions between XCR1⁺ cDC1 and XCL1-producing NK cells in tumors and lymphoid tissues (Barry et al., 2018; Ghilas et al., 2021). As an alternative to germline Xcl1 KO mice, we leveraged the CRISPR-RNP system (Nüssing et al., 2020) to test a role for CD8⁺ T cell–derived XCL1 (Fig. S2 A). P14 T cells were electroporated with sgThy1 or sgXcl1 prior to single transfer and infection with LCMV-Arm. Tissues were harvested at effector (D8) and memory (D25) time points for analysis (Fig. 2 B).

No differences in absolute P14 T cell number or phenotype were observed in any tissues at D8 (Fig. 2 C). Recent work demonstrated that T_RM precursors partially upregulate the T_RM gene signature upon mesenteric lymph node (mLN) priming, including expression of Ccr9, Xcl1, and P2rx7 (Qiu et al., 2023). Given the capacity of mLN and Peyer’s patch–derived DCs to produce CCR9-inducing retinoic acid (RA) (Mora et al., 2003; Iwata et al., 2004; Molenaar et al., 2011), we tested whether cells exhibit differences in CCR9 during the effector phase. We observed no differences between sgThy1 and sgXcl1 cells at D8 (Fig. 2 C). Thus, Xcl1 KO does not appear to impair effector migration into tissue. At D25, no significant differences in frequency or number were observed in the circulatory compartment (Fig. 2 D). However, we observed a marked reduction in both frequency and number of intraepithelial lymphocyte CD69⁺ CD103⁺ P14 T cells. This was accompanied by a corresponding increase in the frequency of CD69⁻ CD103⁻ cells (Fig. S2 B). No significant differences were observed in the LP or kidney (Fig. 2, E and F). P2RX7 is an extracellular ATP scavenger receptor and target of RA (Heiss et al., 2008). It has also been shown to promote TGFβ sensing and T_RM homeostasis (Borges da Silva et al., 2020). Control P14 T cells (sgThy1) express P2RX7 upon upregulation of CD69 and CD103 (Fig. 2 G), and we observed a reduced frequency and number of CD69⁺ P2RX7⁺ cells in Xcl1 KO cells (Fig. 2 F). Loss of P2RX7 supports previous work examining defective OT-I responses in Itgax-cre.Irf8^fl/fl mice following oral antigen administration (Luda et al., 2016). Together, these data demonstrate a role for T cell–derived XCL1 in the formation of SI T_RM during acute viral infection.

To further dissect the role of XCL1/XCR1 in cell–cell interactions influencing CD8⁺ T_RM, we utilized the XCR1^DTR/+ mice (Yamazaki et al., 2013), which enable depletion of XCR1⁺ cells with diphtheria toxin (DT). We transferred P14 T cells into WT and XCR1^DTR/+ recipients, followed by LCMV-Arm infection. Although we observed no differences in Xcl1 KO cells at D8, we hypothesized that early depletion of XCR1⁺ cells could influence the accumulation of SI CD8⁺ T cells. To minimize the broad effects of DC depletion on P14 T cell expansion, we administered DT days 4–7 after infection. Tissues were isolated at D8 for flow cytometry and confocal microscopy (Fig. 3 A). We confirmed systemic depletion of XCR1⁺ DC following DT treatment (Fig. 3, B and C). Confocal imaging of the SI (Fig. 3 D) revealed a profound defect in P14 T cell accumulation, with an approximately threefold decrease in the number of P14 T cells (Fig. 3, E and F). Flow cytometric analysis showed a reduced frequency of CD69⁺ CD103⁺ cells (Fig. 3 G), suggesting impaired differentiation of T cells in the absence of cDC1. Analysis of P14 T cell location revealed a twofold increase in P14 occupying the lower half of the villus (Fig. 3 H). Additionally, IEL but not splenic P14 T cells displayed reduced expression of the IL-7 receptor alpha chain, CD127 (Fig. 3 I). Absolute numbers of P14 T cells did not differ in lymphoid tissues or the kidney, suggesting a preferential role for XCR1⁺ DCs in intestinal immunity, in line with work studying this population during homeostasis (Ohta et al., 2016). Circulating cells revealed no differences in the proportions of terminal-effector or memory-precursor P14 T cells (Fig. 3 J).

Depletion of cDC1 following memory formation led to a frequency reduction in CD103⁺ T_RM in IEL and LP compartments (Fig. 3 K). We tested a functional role for the XCR1 receptor by transferring P14 into XCR1-deficient mice (XCR1^Ven/Ven). At a memory time point, we observed a modest loss of CD103⁺ T cells in the LP of XCR1^Ven/Ven hosts, consistent with the maintained in situ presence of XCR1-deficient cDC1 (Fig. S2 C). Together, these data support a role for XCR1 and the cDC1 population in T_RM formation and maintenance, respectively.

SI CD8⁺ T cell phenotype is linked to intratissue location (Reina-Campos et al., 2024, Preprint). To interrogate the impact of XCR1⁺ cell depletion on the transcriptome and location of P14 T cells, we utilized the 10X Xenium platform to measure RNA transcripts at subcellular resolution. This enables the detection of 350 genes, curated to enable high-quality cell-type annotation, ligand-receptor activities, and immune cell states.

To identify P14 T cells, we transferred Xist⁺ CD45.1 female cells into male CD45.2 recipients prior to DT treatment (Fig. 4 A). We validated our previous findings that depletion of XCR1⁺ cDC1 (Fig. S2 D) leads to a frequency reduction of CD103⁺ P14 cells (Fig. S2 E). The loss of CD103⁺ cells aligns with the observation that CD8⁺ T_RMItgae transcripts are more abundant at the villus tip (Reina-Campos et al., 2024, Preprint). Quantification of P14 T cells using Xist transcripts confirms the reduction in XCR1-DTR mice (Fig. 4 B). We generated Immune Allocation Plots (IMAP) (Reina-Campos et al., 2024, Preprint) to depict the localization and density along two axes: epithelial and crypt-villi (Fig. 4 C). Upon cDC1 depletion, P14 T cells mislocalized toward the base of the crypt (Fig. 4 D). Transcript images showed proximity of Xcl1⁺ and Xcr1⁺ cells in WT mice with comparable Xcl1 transcript localization in WT and XCR1-DTR mice (Fig. 4 E). In mice, Xcl1 expression increases toward the villi tip, mirroring Itgae and opposing Tcf7. In parallel, human CD8αβ⁺ T cell XCL1/XCL2 displayed a similar crypt-villi expression pattern (Fig. 4 F) (Reina-Campos et al., 2024, Preprint). Consistent with previous studies (Kurd et al., 2020; Milner et al., 2020), the detection of P14 Xcl1 expression increased during LCMV-Arm (Fig. S2 F). Differential gene expression of P14 T cells in WT and XCR1⁺ cell-depleted mice revealed numerous changes (Fig. 4 G), several of which have been linked to T_RM development and function. This includes Gzma (Crowl et al., 2022), Srebf2 (Reina-Campos et al., 2023), and Fabp2 (Frizzell et al., 2020). P14 in XCR1-DTR mice also displayed marginally lower Cxcr3 expression, potentially altering CXCL9/10-mediated chemotaxis (Fig. S2 G). In WT mice, proximity to cDC1 influences the expression of memory-associated genes, including Tcf7, Il15ra, Tgfbr2, and Ccr9 (Fig. 4 H). Thus, disrupting CD8-cDC1 crosstalk leads to impaired localization and disrupts T_RM-associated gene expression.

T cells which share features of T_RM cells have been associated with positive prognoses, immunotherapy responsiveness, and CAR-T performance across multiple tumor types (Savas et al., 2018; Ganesan et al., 2017; Edwards et al., 2018; Oliveira et al., 2023; Jung et al., 2023). In parallel, studies have demonstrated the role of TCF1^hi T_PEX and cDC1 cells, particularly in response to checkpoint blockade (Im et al., 2016; Salmon et al., 2016; Miller et al., 2019; Siddiqui et al., 2019; Teijeira et al., 2022; Meiser et al., 2023).

Here, we employed an adoptive transfer model of P14 T cells into tumor-bearing mice (Fig. S3 A). P14 T cells profiled D7 after transfer consist of three differentiation states. T_PEX cells are marked by high expression of TCF1 and SLAMF6, while T_INT and T_TERM downregulate these markers and upregulate PD1, CD39, and Granzyme B (Fig. S3, B and C). During an endogenous response to B16 melanoma, we observed a correlated decay of intratumoral T_PEX and cDC1 frequencies, while these populations remained stable in the tumor-draining lymph node. Given the high expression of Xcl1 by T_PEX (Fig. 5 A), we hypothesized that enhancing XCL1 production could rescue cDC1 abundance and modulate differentiation. We transduced P14 T cells with an Xcl1 overexpression (CD8^Xcl1) or empty vector construct (CD8^EV) (Fig. S3 D) and transferred cells into tumor-bearing mice. At D4–6 after transfer, we observed a tumor-specific accumulation of XCR1⁺ DC in mice receiving CD8^Xcl1 (Fig. 5 B). At D6 after transfer, CD8^Xcl1 mice had a higher frequency of intratumoral P14 T cells. TIL analysis revealed a significant increase in TCF1 and KI67, reflecting enhanced persistence and proliferation (Fig. 5 C). Transfer of CD8^Xcl1 led to delayed tumor growth and prolonged survival during B16-GP_33–41 melanoma challenge. Similar results were also observed using the MC38-GP_33–41 colorectal carcinoma model (Fig. 5 D and Fig. S3 E).

XCL1 has not been extensively studied in human T cells. Analysis of a T cell atlas (Zheng et al., 2021) provides single-cell expression data across 21 tumor types. Cluster analysis shows an enriched expression of XCL1 and XCL2 in four populations: NK-like (Tk.KIR2DL4, Tk.TYROBP), TCF7^hi (Tex.TCF7), and T_RM-like (Trm.ZNF683) (Fig. 5 E and Fig. S3 F), matching the murine expression patterns. Further, analysis of TCGA data of 457 skin cutaneous melanoma (SKCM) patients demonstrates survival stratification based on XCL1/XCL2 expression. XCL1/XCL2 also correlated with a functional cDC1 signature (Meiser et al., 2023) (Fig. 5 F). The top 20 genes with similar expression pattern as XCL1 include immune activation and T_RM genes, such as GZMA, TIGIT, CCL4, and CXCR6 (Hombrink et al., 2016; Kumar et al., 2017) (Fig. S3 G). Thus, the human expression profile supports exploration of the XCL1–XCR1 axis, for example as a target for TME remodeling from “cold” to “hot” prior to immunotherapy initiation.

In summary, we demonstrate that CD8⁺ T_RM highly express the cDC1 chemoattractant, Xcl1. Intestinal CD8⁺ T cells upregulate Xcl1 during infection and expression remains elevated following clearance. Depletion of cDC1 impaired the localization and differentiation of SI CD8⁺ T cells during LCMV-Arm. Further, T cell–specific KO of Xcl1 leads to defective T_RM formation. The selective expression of TGFβ activating integrin Itgb8 and RA enzyme Aldh1a2 in XCR1⁺ DCs (Shiokawa et al., 2017) supports a role for these cells in T_RM differentiation. Additionally, the expression of Tgfb1 in LP cDC1 suggests a role in maintaining T cell Itgae expression (Reina-Campos et al., 2024, Preprint). While the bidirectional signal exchange between cDC1 and CD8⁺ T cells remains to be investigated, our data suggest a role for interactions at effector and memory time points. In tumors, the proinflammatory functions of cDC1, such as IL-12 production, may favorably remodel the TME to support T cell infiltration and survival. Further studies are needed to understand the regulation of intratumoral niches, given their presence in human cancers (Jansen et al., 2019; Wang et al., 2023).

These findings highlight the importance of studying cell–cell interactions which shape tissue immunity (Nakandakari-Higa et al., 2024). These efforts will be strengthened by advances in spatial transcriptomics, enabling high-resolution exploration of cellular states in healthy and diseased tissues. Defining regulators of Xcl1 (Lei et al., 2011; Russ et al., 2014) and Xcr1 (Bayerl et al., 2023), particularly across tissue environments will inform strategies to harness the T_RM interactome for solid tumor immunotherapy.

All mice were bred on the C57BL6/J background and housed in specific pathogen–free conditions in accordance with the Institutional Animal Care and Use Committees (IACUC) of the University of California, San Diego (UCSD). Both male and female mice were used throughout the study, with sex-matched T cell donors and recipients (or female donor cells transferred into male recipients) and between 1.5- and 4-mo-old. C57BL/6J mice (stock #000664; The Jackson Laboratory), XCR1^DTR/Venus (Gutkind Laboratory, UCSD), P14 mice (with transgenic expression of H-2D^b-restricted TCR specific for LCMV glycoprotein GP 33–41; stock #037394-JAX; The Jackson Laboratory), CD45.1⁺ and CD45.1.2⁺ congenic mice were bred in house.

For P14 CD8⁺ T cell activation, naive CD8⁺ T cells from spleens and lymph nodes were negatively enriched with LS MACS columns (Miltenyi Biotec) using biotin anti-CD4 (GK1.5), anti-Ter119 (TER-119), anti-GR-1 (RB6-8C5), anti-MHCII (M5/114.15.2), anti B220 (RA3-6B2), and anti-NK 1.1 (PK136). P14 cells (2 × 10⁶) were plated in a well of a 6-well plate that was precoated with 100 μg/ml goat anti-hamster IgG (H+L; Thermo Fisher Scientific). The activation medium contained 1 μg/ml anti-CD3 (145-2C11) and 1 μg/ml anti-CD28 (37.51) (eBioscience). Culture medium was replaced after 18 h of activation with retroviral supernatant mixed with 50 μM BME and 8 μg/ml polybrene (Millipore) followed by spin-infection (1-h centrifugation at 2,000 RPM, 37°C). The plate was incubated at 37°C for 3 h after spin-infection, and then the retroviral supernatant was replaced by T cell medium supplemented with IL-2 (Peprotech) and incubated for 24–72 h prior to use.

For enforced Xcl1 expression, retroviral particles were generated using platinum E cells grown in 10-cm plates with full selection media (DMEM, 10% FBS [vol/vol], 2 mM L-glutamine, 100 U/ml penicillin-streptomycin, 1 µg/ml puromycin, and 10 µg/ml blasticidin). 18 h before transfection selection media was replaced with antibiotic-free media (DMEM, 10% FBS, 2 mM L-glutamine). For each 10-cm plate, 10 µg of each construct and 5 µg of pCL-Eco helper plasmids were mixed in Opti-MEM (Thermo Fisher Scientific) to a volume of 700 µl. This was combined with 45 µl of TransIT-LT1 reagent (Mirus Bio) and 655 µl of Opti-MEM for 20 min at room temperature. This mixture was added dropwise to each 10-cm plate. 12 h later media was replaced, and the viral supernatant was subsequently harvested at 48 and 72 h.

Male B16 melanoma cells and MC38 colorectal tumor cells expressing the LCMV glycoprotein epitope amino acid 33–41 (B16-GP_33–41) and female PLAT-E cells were maintained in DMEM containing 5% bovine growth serum, 1% HEPES, and 0.1% 2-mercaptoethanol. Both cell lines have been confirmed to be free of mycoplasma through quantitative PCR (qPCR).

B16-GP_33–41 cells and MC38-GP_33–41 (5 × 10⁵) were transplanted subcutaneously into the right flank of wild-type mice. After tumors became palpable, 7–9 days after transplantation, 1–2.5 × 10⁶ P14 cells transduced with Xcl1 or EV retroviral constructs and expanded in vitro with 50 U/ml IL-2 (Peprotech) for 2–3 days were transferred intravenously. Tumors were monitored daily and mice with ulcerated tumors or tumors exceeding 1,500 mm³ in size were euthanized in accordance with UCSD IACUC. TIL were isolated as previously described (Milner et al., 2017) 4–8 days following adoptive transfer.

Mice were infected with 2 × 10⁵ pfu lymphocytic choriomeningitis virus-Armstrong (LCMV) by intraperitoneal injection. For naive P14 transfers, 5 × 10⁴ congenically distinct cells per recipient mouse were adoptively transferred prior to infection. To distinguish vascular-associated CD8⁺ T cells in certain NLT (salivary gland, kidney, IEL, and mLN), 3 μg of CD8α conjugated to APC eFlour780 was injected i.v. into mice 3 min prior to sacrifice and organ excision. Cells lacking CD8α labeling were considered localized within NLT.

Bioinformatic analyses of RNA-seq data derived from the following studies: GSE122713, PRJNA497086, GSE149879, GSE182276, GSE94964, GSE125527, GSE147502, GSE131847, GSE156728, GSE107395, GSE125471. Spatial transcriptomics (WT and XCR1-DTR) GSE accession: GSE280895 and GSE284741.

1 day after activation, T cells were electroporated with complexed tracrRNA (IDT), Cas9 (UC Berkeley), and crRNA (IDT). crRNA targeting Thy1 (control) was used alone as a control and was mixed with the conditional crRNA to be used as a marker of electroporated cells (Nüssing et al., 2020). Electroporated cells were cultured in 50 U/ml of IL-2 for 24–48 h. Control and experimental cells (500K/mouse) were transferred into congenically distinct recipients prior to LCMV infection. DNA modifications were quantified using TIDE (Brinkman et al., 2018) and gene expression changes by qPCR. The sgRNA sequence for Xcl1 targeting is the following: 5′-AGU​CUG​CUG​CCU​CAC​CCC​AU-3′. The sgRNA sequence for Thy1 targeting is the following: 5′-CCG​CCA​UGA​GAA​UAA​CAC​CAG​UUU​UAG​AGC​UAU​GCU-3′.

50,000 cells were sorted directly into Trizol, and RNA was extracted by chloroform and isopropanol precipitation. cDNA was synthesized using Superscript IV (Life Technologies) following the manufacturer’s instructions and qPCR was performed using the Stratagene Brilliant II Syber Green master mix (Agilent Technologies). RT-PCR Xcl1 (F) 5′-CTT​TCC​TGG​GAG​TCT​GCT​GC-3′ Xcl1 (R) 5′-CAG​CCG​CTG​GGT​TTG​TAA​GT-3′, Hprt1 (F) 5′-TGA​AGA​GCT​ACT​GTA​ATG​ATC​AGT​CAA​C-3′, Hprt1 (R) 5′-AGC​AAG​CTT​GCA​ACC​TTA​ACC​A-3′.

Mice were sacrificed by CO₂ asphyxiation. Following removal of the SI, duodenums were resected, cut open longitudinally, washed with ice-cold PBS, and shaped into a swiss roll before OCT embedding (Sakura). Samples were snap-frozen on dry ice and stored at −80°C until sectioning. 10 μm cryo-sections were prepared on Superfrost plus glass slides (Fisher Scientific) using the CryoStar NX50 cryostat (Thermo Fisher Scientific) at a cutting temperature of −19°C. Following sectioning, slides were gently washed once in prechilled 1× PBS (Gibco) for 3 min, then fixed immediately in prechilled acetone at −20°C for 20 min. Next, slides were washed for 3 min with 1× PBS to remove residual acetone and blocked using Dako Protein Block (Agilent) for 3 h at RT in the dark. After blocking, primary conjugated antibodies were diluted in Dako antibody diluent with background reducing components (Agilent) at 1:50. Antibodies included CD8a (53-6.7), CD45.1 (A20), CD324 (DECMA-1), XCR1 (ZET), and CD11c (N418). Slides were left to stain overnight at 4°C in a dark humidified chamber. Afterward, slides were moved to RT and left for 1 additional hour to increase staining signal on weakly expressed epitopes. Slides were washed 2× for 10 min with 1× PBS-T (0.1% Triton X-100 [Sigma-Aldrich]) and 2× for 10 min with 1× PBS with gentle agitation. Following the washing step, a nuclear counterstain (DAPI; BioLegend) was added at 1:1,000 diluted in 1× PBS for 5 min at RT in the dark. To remove excess counterstain, slides were washed one last time (5 min in 1× PBS) and air-dried for 1 min in the dark before being mounted in Vectashield Vibrance Mounting Medium (Vector Laboratories). Slides were left to cure for 1 h at room temperature in the dark before imaging on a Zeiss LSM700 confocal microscope using the PlanApo 20×/0.8 objective. Quantification of the average distance of P14 cells in small intestinal villi was done in ImageJ. P14 T cells located in villi were manually counted. To account for phenotypic differences in villi length, P14 T cell distance measurements were normalized and displayed as a percentage of villi distance.

For DT-mediated ablation of XCR1-expressing cells, 500 ng/mouse was administered by intraperitoneal injection for four consecutive days. Both WT and XCR1-DTR mice received DT treatment. For longer-term depletion, mice received 500 ng every 3 days for the course of treatment.

Single-cell suspensions were prepared from the spleen or lymph node by mechanical disruption. For SI preparations, Peyer’s patches were excised, luminal contents were removed, and tissue was cut longitudinally then into 1 cm pieces. The SI pieces were incubated while shaking in 10% HBSS/HEPES bicarbonate solution containing 15.4 mg/100 μl of dithioerthritol (EMD Millipore) at 37°C for 30 min to extract SIIEL. For LP lymphocyte isolation, SI pieces were further treated with 100 U/ml type I collagenase (Worthington Biochemical) in RPMI-1640 containing 5% bovine growth serum, 2 mM MgCl₂, and 2 mM CaCl₂ at 37°C for 30 min. The salivary gland, kidney, and tumor were cut with scissors into fine pieces and then incubated while shaking with 100 U/ml type I collagenase as above. Lymphocytes from all tissue but spleen and lymph node were purified on a 44%/67% Percoll density gradient.

Formalin fixed parrafin embedded (FFPE) tissues were sectioned to a thickness of 5 μm onto a Xenium slide, followed by deparaffinization and permeabilization following the 10X user guides CG000578 and CG000580. Probe hybridization, ligation, and amplification were done following the 10X user guide CG000582. In short, probe hybridization occurred at 50°C overnight with a probe concentration of 10 nM using a custom gene panel designed to detect 350 different mRNAs. After stringency washing to remove unhybridized probes, probes were ligated at 37°C for 2 h. During this step, a rolling circle amplification primer was also annealed. The circularized probes were then enzymatically amplified (2 h at 37°C), generating multiple copies of the gene-specific barcode for each RNA binding event. After washing, background fluorescence was quenched chemically. Sections were placed into an imaging cassette to be loaded onto the Xenium Analyzer instrument following the 10X user guide CG000584.

For 10X Xenium spatial transcriptomics data, nuclei were segmented using a fine-tuned Cellpose model on max-projected DAPI staining images. Baysor was used to predict cell boundary segmentations using the Cellpose nuclei segmentations as a prior segmentation for 10X Xenium and transcripts as input, with parameters prior-segmentation-confidence = 0.95 and 0.9, respectively, and min-molecules-per cell set to the median nucleus transcript count (see https://github.com/Goldrathlab/Spatial-TRMpaper). Baysor segmentations containing no nuclei were filtered out, and segmentations containing multiple nuclei were split by assigning transcripts to the nearest nucleus centroid within the segmentation boundary. All cells with n < 8 nuclear transcripts, n < 20 total transcripts, or n > 800 total transcripts were filtered out before downstream processing. To integrate spatial replicates into a joint embedding, scVI was used with n_layers = 2, and n_latent = 30. The joint embedding was projected into 2D space using scVI.model.utils.mde. Leiden clustering was performed on the scVI learned embeddings using scanpy.tl.leiden with resolution = 1, and every Leiden cluster was further subclustered at resolution = 1.2. Celltypist and GeneFormer were used for a first-pass cell type assignment, with further manual refinement based on the expression of cell type marker genes to define cell types in a Class > Type > Subtype hierarchy. Anndata format was used for all further processing.

For the Xenium mouse 350 gene panel, 79 SI canonical cell type marker genes were compiled from existing literature, and Xist, a marker for transferred female P14 T cells. An additional 158 genes from a Nichenet database of ligand–receptor pairs were included. Next, a supervised PERSIST was run on an immune-enriched intestine scRNA-seq dataset with the previously compiled set of genes as prior information, adding an additional 70 genes. Lastly, a supervised PERSIST was run on a SI scRNA-seq dataset to capture 59 cell-type marker genes for 350 total targets. To create the Xenium human 422 gene panel, we created a base set of canonical immune marker genes, ligand–receptor pairs, spatially differentially expressed genes within mouse P14 T cells, and IEL the 10X Genomics base human colon panel totaling 343 genes. Using this set as prior information to PERSIST, and a reference human immune cell scRNA-seq dataset, unsupervised PERSIST filled in the remaining 79 genes.

The epithelial IMAP axis values are computed through a biexponential transformation applied to the clipped epithelial axis values across all cells. Each cell is positioned on the IMAP by its corresponding crypt–villus axis and transformed epithelial axis values. Density within the scattered point cloud is visualized using color-mapped scipy.stats.gaussian_kde values, with density lines overlaid using seaborn.kdeplot for enhanced clarity and interpretation. Gate boundaries were manually drawn to distinguish the muscularis, villus crypt, and villus top by observing the IMAP locations of cell types known to localize to each region. Transcriptional IMAPs were colored by adding an array of gene expression counts as a point weight parameter to the scipy.stats.gaussian_kde function.

Cells were incubated for 30 min at 4°C in PBS supplemented with 2% bovine growth serum and 0.1% sodium azide. Surface antibodies used include CD45.1 (A20), CD45.2 (104), KLRG1 (2F1), CD8a (53-6.7), CD127 (A7R34), CD62L (MEL-14), CD69 (H1.2F3), CD103 (2E7), CD90.2 (53-2.1), P2RX7 (1F11), MHCII (M5/114.15.2), CD11c (N418), CD11b (M1/70), and XCR1 (ZET). Intracellular staining of Tcf1 (C63D9), GzB (MHGB05), and Ki67 (11F6) was performed using the Foxp3/Transcription Factor Staining Kit according to the manufacturer’s protocol (Tonbo). BD Cytofix/Cytoperm Solution Kit (BD Biosciences). Stained cells were analyzed using LSRFortessa or LSRFortessa X-20 cytometers (BD) and FlowJo software (TreeStar). All sorting was performed on BD FACSAria or BD FACSAria Fusion instruments.

Statistical parameters are reported in the figures and figure legends. A two-tailed Student’s t test (unpaired for single transfer studies) or one-way ANOVA for studies with more than two groups was done on all reported data using Prism software (GraphPad). For survival data, a log-rank Mantel-Cox test was performed. All experiments were executed independently at least two times to ensure reproducibility. All P values are shown in the figures. Human TCGA data were analyzed using the GEPIA 2.0 software tool (Tang et al., 2019).

Fig. S1 shows Xcl1 RNA-seq expression in CD8⁺ T cell subsets across published datasets. Fig. S2 shows CRISPR KO efficiency, experiments using XCR1^Ven/Ven mice, confocal microscopy validation of XCR1⁺ cell depletion, and additional data from the spatial transcriptomics experiment. Fig. S3 shows the experimental setup for TIL analysis, flow cytometry data, Xcl1 overexpression validation, tumor growth curves, and human TIL scRNA-seq analysis.

The data are available from the corresponding author upon reasonable request.

We thank the Goldrath and Gutkind laboratories for advice and manuscript review. We thank Dr. John Chang for a critical reading of the manuscript.

This work was supported by National Institutes of Health grants R01AI179952 (A.W. Goldrath), R37AI067545 (A.W. Goldrath), R01AI072117 (A.W. Goldrath), and R01AI150282 (A.W. Goldrath).

Author contributions: A. Ferry: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing, K.M. Mempel: Formal analysis, Investigation, A. Monell: Data curation, Formal analysis, Software, Visualization, M. Reina-Campos: Conceptualization, Formal analysis, Investigation, Methodology, Resources, Supervision, Writing - review & editing, N.E. Scharping: Investigation, Methodology, M. Heeg: Data curation, Formal analysis, Investigation, Methodology, Software, K.K. Takehara: Investigation, S. Schokrpur: Investigation, Methodology, N. Kuo: Resources, Validation, R. Saddawi-Konefka: Investigation, Methodology, J.S. Gutkind: Conceptualization, Investigation, Resources, Supervision, A.W. Goldrath: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing - original draft, Writing - review & editing.