A dynamic atlas of immunocyte migration from the gut

Abstract

Dysbiosis in the gut microbiota impacts several systemic diseases, possibly by driving the migration of perturbed intestinal immunocytes to extraintestinal tissues. Combining Kaede photoconvertible mice and single-cell genomics, we generated a detailed map of migratory trajectories from the colon, at baseline and in several models of intestinal and extraintestinal inflammation. All lineages emigrated from the colon in an S1P-dependent manner. B lymphocytes represented the largest contingent, with a surprising circulation of non-experienced follicular B cells, which carried a gut-imprinted transcriptomic signature. T cell emigration included distinct groups of RORγ⁺ or IEL-like CD160⁺ cells. Gut inflammation curtailed emigration, except for dendritic cells disseminating to lymph nodes. Colon-emigrating cells distributed differentially to different sites of systemic inflammation (skin, arthritic synovium, tumors). Thus, specific cellular trails originating in the gut, and influenced by microbiota, may shape peripheral immunity in varied ways.

One sentence summary:

The gut is the origin of a diverse circulation of immune cells that carry specific imprints, and differentially fan out to different organs.

INTRODUCTION

The mammalian gut is a complex organ, harboring most of the organism’s microbiota, a constant flux of dietary antigens, and a large number of immunological cell-types. Gut immunocytes face the unique challenge of maintaining tolerance to food and commensal antigens without compromising effective responses against pathogens. In addition, perturbations that occur in the intestines can influence systemic immune responses (1, 2). Indeed, the onset and progression of some extraintestinal inflammatory diseases appear to be influenced by the intestinal microbiota (3). For instance, gut-colonizing segmented filamentous bacteria (SFB) induce interleukin-17–producing T (Th17) cells, which exacerbate disease in autoimmune mouse models of arthritis or multiple sclerosis (4, 5). Several studies have reported perturbations in the intestinal microbiota of rheumatoid arthritis (RA), multiple sclerosis, and type-1 diabetes patients (6, 7), and the composition of the gut microbiome affects the responsiveness to immunotherapy in cancer patients (8–10). Moreover, inflammation at distal sites such as the skin, the joints, and the eyes can occur in up to 50% of inflammatory bowel disease (IBD) cases (11).

The underlying mechanisms by which the gut mediates this interorgan communication remain incompletely understood. Several studies have demonstrated a role for metabolites and other bacterially derived products (12) and others have identified live bacteria as possible conduits (13, 14). The intestines may also exert systemic influences via the migration of gut immune cells. Although the cell migratory mechanisms and dynamics from lymphoid organs to the gut are appreciated to a certain extent (6), our understanding of migration from the gut to lymphoid and other systemic organs is limited. There is evidence that trafficking of several immune cell populations from the intestines to the spleen and lymph nodes (LNs) occurs under steady state (15–18). Th17 cells activated by the gut microbiota play a pathogenic role in multiple tissues in autoimmune mouse models of arthritis, uveitis, renal disease, and encephalomyelitis (4, 5, 19, 20). Moreover, IgA⁺ plasma cells and FoxP3⁺ regulatory T cells (Tregs) of gut origin are protective in the brain and pancreas (21–24), whereas intestinal innate lymphoid cells (ILCs) mediate anti-helminth defense and tissue repair in the lungs (25). In order to therapeutically target systemic effects of the microbiota in the context of disease, a solid understanding of these systemic migration pathways at baseline as well as in inflammatory conditions are urgently needed.

Kaede transgenic mice express a fluorescent protein that converts from green to red upon exposure to violet light, making them an ideal tool to track cellular migration in vivo, in a minimally invasive manner (26). They have been previously used to temporally trace the migration of intestinal cells following photoconversion of the small intestine or ascending colon via surgical exposure (20, 27–30), or in combination with a less invasive endoscopic system (15, 17). Although immune cells have been shown to be capable of exiting the intestines, information on their destination and function has primarily concerned lymphoid tissues (mostly local LNs) or specific cell types in autoimmune models (15, 16, 19, 20, 31, 32). Here we used a systems approach to obtain a global picture of the contribution of colon-derived immunocytes to circulation at homeostasis and how this map is modified by inflammation or tumors in the gut or in destination organs. This work revealed that immunologic cell-types emigrating from the intestines carry along a phenotypic memory of their origin and represent specific subsets in their destination organs, with cell-specific preferences for different inflammatory sites or tumors.

RESULTS

Gut immunocytes have high mobility and turnover

Colonoscopic illumination with violet light temporarily labels cells of the descending colon of photoconvertible Kaede mice in a short time window (15–17, 28), without labeling cells in local draining lymphoid tissues, other intestinal sections or peripheral tissues (fig. S1, A and B). Cells were converted throughout the lamina propria down to the muscularis (fig. S1C), which should thus apply as well to colonic patches and “solitary intestinal lymphoid tissue” (33). To ascertain the fraction of photoconverted cells that were blood cells circulating through the colon at the time of illumination, we labeled intravascular cells by intravenous (i.v.) injection of anti-CD45 antibody at the time of photoconversion (fig. S1D) (34). Only 10–100 cells from the blood were photoconverted in passing (fig. S1E), a minute fraction compared to the 50,000–250,000 cells converted in the colon, and cannot be meaningful contributors to the phenomena described below.

To better understand the turnover of intestinal immunocytes, we tracked the proportion of photoconverted cells remaining in the descending colon at various times after the light pulse (Fig. 1, A to C, and fig. S2A). All lineages photoconverted similarly (fig. S2B), although there was mouse-to-mouse variation. This turnover reflected egress from the colon, but could also encompass cell death and proliferation. Tagged mononuclear phagocytes (MNPs), as well as T cells, disappeared in a bimodal fashion, the fastest drop occurring in the first 12–24 hours, followed by a phase of slower decay, with 20 to 40% of the photoconverted populations still present after 3 days (Fig. 1, A and B). By contrast, the number of phototagged B cells declined more quickly, fitting a monophasic exponential decay, with very few red-labeled cells remaining after 3 days. Colonic intraepithelial lymphocytes (IELs) also showed turnover, albeit to a slower extent, the CD8aa IEL population being especially stable (Fig. 1C).

Figure 1. Turnover of colon immunocytes at steady-state.

A) Representative flow cytometry plots of photoconversion proportions of B, T and MNPs in the descending colon in unconverted (control) mice versus 10min, 12hr, 24hr, 72hr post colonic photoconversion. B) Egress rates of cells photo-tagged in the descending colon, measured by flow cytometry (n=4–6 mice, from 3 or more independent experiments). Normalized to the average of time 0hr. C) Egress rates of cells within the IEL fraction photo-tagged in the descending colon, measured by flow cytometry (n=4–6, from three or more independent experiments, each datapoint represents a mouse) D) Proportions of photo-tagged B, αβ T and CD11c⁺ DCs left in the colon 24hr post-photoconversion, with and without pre-treatment with FTY720 (1mg/kg). Normalized to the average of time 0hr timepoints. Each datapoint represents a mouse, from three independent experiments E) Schematic of gut draining LNs. F) Percentage of migratory Kaede-red populations across individual mLNs, caLN and iLN, 24hr post-photoconversion of the descending colon, measured by flow cytometry (n=2–3 mice from two independent experiments). Error bars indicate mean ± SD; *p < 0.05, ****p < 0.0001, unpaired t test.

We next queried the mechanism(s) underlying this fast turnover. Egress of T cells from the colon depends on Sphingosine-1-posphate (S1P) (16, 20). Pretreatment of Kaede mice with the S1P receptor (S1PR) functional antagonist FTY720 prior to photoconversion of the descending colon completely inhibited immune cell turnover measured 24 hours later, not just for T cells but also CD11c⁺ dendritic cells (DCs) and B cells (Fig. 1D). Thus, the turnover of phototagged cells resulted from egress (as opposed to cell death or division) and S1P was the main mechanistic controller of this phenomenon.

Given this extensive egress from the colon, we next investigated migration to local draining lymph nodes (Fig. 1E). Mesenteric LNs (mLNs) and caudal LNs (caLNs) are sites of direct lymphatic drainage for intestinal tissue and the primary location where DCs present microbial and food antigens to T cells by ferrying them from the gut (35). Drainage of the gut is compartmentalized, with the distal colon draining predominantly into the caLN and a minority of the mLNs (27, 36, 37) (Fig. 1E). When individual LNs in the mLN chain were studied 1 day after photoconversion of the colon in Kaede mice, we found MNPs of recent colonic origin—CD11b⁺CD11c⁺ dendritic cells (DCs) preferentially accumulating in the caLN (Fig. 1F). Immigrant DCs were also present in the most distal members of the mLN chain (#1 and #7), albeit to a lesser extent, but not in other mLNs, in agreement with prior studies (27, 36, 37). Phototagged B and T cells of colon origin followed a similar pattern of accumulation. The percentage of phototagged immunocytes in the caLN (DCs but also lymphocytes) was higher than in the spleen or subcutaneous LNs, consistent with classic models of lymph drainage and suggesting that mLN is a conduit for immunocyte egress from the colon, although we cannot be sure that it is the only one.

Migratory immunocytes connect the colon with system-wide locations at steady-state

We examined both lymphoid and non-lymphoid tissues for the destinations of this migration. Twenty-four hours after photoconversion in the colon, phototagged red CD45⁺ cells were detected in almost all tissues examined at higher frequencies than in the blood and, not surprisingly, at much lower frequencies than in the colon, since colon-derived cells were diluted by local tissue-resident cells (Fig. 2, A and B). Mice were perfused to exclude most circulating cells from the tissues analyzed, and intravascular anti-CD45 antibody stain confirmed that colon-derived immunocytes present in spleen and lymph nodes 24 hours post-photoconversion are not located within the vasculature (fig. S3, A and B). The two notable exceptions were the thymus and the skin, where migrant cells were not detected (Fig. 2A). The absence of phototagged CD45⁺ cells in the skin supports the concept that the gut and skin immune systems are compartmentalized under normal conditions, even though both are microbiota-carrying barrier tissues (38). In the case of the thymus, which has been linked to the intestines via migratory DCs (17, 18), any colon-derived population may have been diluted by the large numbers of thymocytes, and may therefore be present but below the level of detection. By contrast, migration of colon-derived immunocytes to the small intestine, lungs, liver and kidneys was comparable with migration to lymphoid organs in terms of ultimate seeding frequency.

Figure 2. Immunocytes from the colon migrate into both lymphoid and non-lymphoid tissues.

A) Percentage of CD45⁺ migratory Kaede-red cells across tissues, measured by flow cytometry 24hr following colon photoconversion (n=4–8). B) Percentage of photo-tagged MNPs, B and T cells in the blood 12hr, 24hr, 48hr and 72hr post colonic photoconversion (n=4–9). C and D) Local (Kaede-green) immunologic populations compared to colon-origin (Kaede-red) cells in (C) BM, spleen, iLN, (D) lung, liver and kidneys post colonic photoconversion by high dimensional flow cytometry. tSNE of CD45⁺ cells built on CD11c, CD11b, F4/80, γδ TCR, αβ TCR, CD8, CD4, CD25, CD19 (pool of n=6). E) Chord diagrams depicting the numeric distribution of each Kaede-red population across all tissues examined (n=6 mice). All results are pooled from three independent experiments; except in the tSNE, each datapoint represents a mouse; error bars indicate mean ± SD.

To characterize the incoming colon-derived populations and compare them with their local tissue counterparts, we used high-parameter flow cytometry represented on a 2D tSNE projection (Fig. 2C). In the spleen and bone marrow, B cells were the most dominant immigrant population. This dominance was less pronounced in the LNs, where colon-derived T cells were better represented. When represented as a fraction of whole lineages, migrated cells represented between 0.2 and 0.8% of the total populations (fig. S3C), indicating that colon-derived cells contribute to every lineage in lymphoid organs. As the exception, although they represent a major fraction of the bone marrow cells, very few CD11c⁻CD11b⁺ cells were colon-derived. This is somewhat expected if the bone marrow, as a site of myeloid differentiation, is considered to be partially shielded from environmental influences.

In non-immunological organs, B cells were again the dominant CD45⁺ immigrant population, much as in lymphoid organs (Fig. 2D). T cells were also present, with colon-derived Tregs making a larger contribution to the local niche than in the spleen (fig. S3C). Independently of proportional representation, compiling the tissues in which phototagged immunocytes that emigrate from the gut were found, revealed subtle but reproducible tissue preferences. MNPs in general migrated in greater numbers to non-immunological tissues than to LNs and spleen, in contrast to lymphocytes, which did the opposite (Fig. 2E). Within lymphocytes, γδ T cells and Tregs more closely resembled MNPs in their migratory preferences. Thus, immunocytes migrate widely from the colon to both lymphoid and non-lymphoid tissues.

B cells of colonic origin bring a distinct transcriptome in the spleen

This widespread migration of gut immunocytes raised several questions relevant to their impact on systemic immunologic function. For example, how do these gut migrants differ from their counterparts in destination tissues? Do they express gut-imprinted phenotypic and functional programs? To compare colon-derived versus local populations in a broad and unbiased manner, we performed single-cell RNA sequencing (scRNA-seq) on total CD45⁺ splenocytes of Kaede mice at different time points after colonoscopic photoconversion (Fig. 3A; sorting equivalent numbers of Kaede-green and -red cells, every timepoint from 2 individual mice). The use of DNA-coded antibody hashtags (39) enabled us to profile 12 samples in the same run, providing an ideal comparison between immigrant and resident cells. Dimensionality reduction and projection on a uniform manifold approximation and projection (UMAP) revealed all the immunological lineages expected in the spleen (Fig. 3B), which were identifiable with prototypic marker genes (fig. S4A) and represented across both resident and migrant populations (Fig. 3B and fig. S4B).

Figure 3. Transcriptome of colon-derived B cells.

A) Single cell experiment schematic. Splenic cells were isolated from mice photoconverted at various points and sorted as CD45⁺ Kaede-green⁺ or CD45⁺ Kaede-red⁺, labelled with hashtag antibodies and sequenced in a single lane. Sample demultiplexing was performed computationally. B) scRNAseq analysis of total CD45⁺ cells in spleen. UMAP representation, color-coded by cell identity. C) UMAP representation of B cells selected from (B). Color coded by k-nearest neighbour (knn) cell clusters (total 8440 cells). D) Distribution of 24hr post-photoconversion Kaede-green vs Kaede-red cells across the UMAP in (C) (1194 Kaede-green cells, 2002 Kaede-red cells). E) Clustered heatmap of differentially expressed genes between Kaede-red and Kaede-green cells within the top four most migrant-enriched clusters from (D), across all three timepoints (355 genes, p < 0.01). F) Gene expression of selected markers across Kaede-green and Kaede-red cells (same plots as in D).

In agreement with our flow cytometry data, B cells constituted a large portion of the migratory compartment. To more finely interrogate the phenotype of these migratory cells, we reclustered them on their own, which resolved seven different B cell-types (Fig. 3C, fig. S4, C and D). These subsets included distinctive marginal zone (MZ) B cells, mature IgM^hi B cells, a small B cell population with high expression of interferon signature genes (ISG) (hereafter called “ISG-B cells”, by analogy with comparable T cells (40–42)), and a large contingent of naive follicular B cells. Newly arrived colon-derived B cells (24 hours after phototagging) were notably missing from the IgM^hi mature group and distinct from the main MZ subset (Fig. 3D). Migrants mostly belonged to the follicular clusters, with a dominance of IgD⁺CD23⁺CD21^lo B follicular-like cells. A differential density plot between the colon-derived and resident cells also reflected these differences, which remained largely unchanged at a population level for 72 hours (fig. S4E), although the distribution within the follicular clusters did evolve somewhat. These particularities were confirmed by flow cytometry (fig. S4F) and in an independent scRNA-seq experiment (fig. S4G). This dominance of follicular-like B cells among colon-derived cells was intriguing, since they are not thought to be a dominant population in the colon. However, flow cytometric analysis of B cells in the descending colon showed that most shared the IgD⁺IgM^lo characteristics of spleen follicular B cells (fig. S4F) indicating phenotypic consistency. Colon-derived B cells also included a surfeit of ISG-B cells, which are of interest due to recent reports highlighting the role of microbiota-induced type-I IFN (43, 44).

To identify gut-imprinted gene-expression programs, we performed differentially expressed gene (DEG) analysis between the migratory (red) and local (green) splenic B cells. To avoid differences merely stemming from variable cluster representation, DEGs were determined within each of the four most represented clusters, selecting genes with significantly different expression between migrant and resident B cells in every cluster (at P<0.01). The resulting group of 355 DEGs distinguished colon-derived B cells (Fig. 3E). Some remained differential at all timepoints (module #1) (Fig. 3E), including the genes encoding the major gut-homing integrin Itgb7, and perhaps more surprisingly transcripts encoding the chylomicron apolipoprotein ApoE or the lysophosphatidylserine receptor Gpr174 (Fig. 3F), whose functions in B cells are not clear (45). The expression of some transcripts normalized after a day or two (module #2), whereas other gene-sets remained underexpressed in immigrant cells (modules #3 and #4). We confirmed by flow cytometry that there was a small population of integrin beta-7-expressing B cells in the spleen, enriched in colon-derived cells (fig. S4H). Thus, there is a circulatory population of gut-derived B cells with gut-imprinted characteristics that are still distinguishable 3 days after exit from the colon.

Multiple populations of colon-derived T cells are present outside the gut

The results were different when we focused on migratory T cells. Reclustering resolved the expected populations of CD4⁺ and CD8⁺ T cells (naïve and effector) as well as more distinctive subsets (e.g. FoxP3⁺ Tregs, ISG-T, etc. (40–42); Fig. 4A and fig. S5, A and B). Although many of the colon-derived cells corresponded to naïve CD4⁺ and CD8⁺ T cell groups, a comparison of Kaede-green and red T cells revealed several activated/memory populations that clearly distinguished spleen-resident and colon-derived cells (Fig. 4, A to C). iNKT cells, identified by their invariant TRAV11-TRAJ18 TCRα chains and by a distinct set of transcripts (fig. S5C), were essentially absent from migrant cells (in keeping with the observation that iNKT cells are rare in the colonic lamina propia (LP)). By contrast, two distinct clusters were almost exclusively represented by recent immigrant Kaede-red cells. One of these (“Rorγ T”) (Fig. 4A) expressed Rorc transcripts and included cells expressing either αβ or γδ TCRs (Fig. 4C). These cells also expressed Cd44, Cxcr6, Il2ra, Tmem176 (but not Foxp3) and lacked Cd27, characteristics of Th17 cells (fig. S5, A and B). Although transcripts encoding IL17 were scarce, these characteristics pointed to colon-derived cells that could mediate the influence of Th17-inducing microbes on extraintestinal immunity (4, 5, 20).

Figure 4. Distinct colon-derived migratory T cell populations are found in the spleen.

A) UMAP representation of T cells selected from 3B. Color-coded by knn cell clusters (total 3500 cells). B) Same plot as in (A), with Kaede-green vs Kaede-red distribution 24hr post-photoconversion (782 Kaede-green cells, 852 Kaede-red cells). C) Gene expression of selected markers across the ‘Rorγ T’ and ‘CD160⁺ T’ clusters from (A). D) Differential marker genes for the CD160⁺ T cluster. E) Percentage quantification of the distribution of the Kaede-green and Kaede-red cells across clusters per timepoint. F) scRNAseq analysis of total CD45⁺ cells from cecum. UMAP representation, color-coded by cell identity (n=2) (left); Kaede red T cells projected onto the dataset from (F) UMAP (right).

The other population dominated by cells of recent colonic origin (“Cd160⁺ T” in Fig. 4A) was a Cd160-expressing cluster that also included both αβ and γδ T cells. They expressed Cd8a more than Cd8b1, which is reminiscent of CD8αα IELs in the gut (46). The possibility of such a relationship was strengthened by the expression of several other markers typical of IELs such as Fcer1g, Il2rb, and several Klr family members (but not Klrb1c, which encodes NK1.1) (Fig. 4D and fig. S5A). The expression of a broader IEL gene signature (47) was also enriched in this population (fig. S5D). We confirmed by flow cytometry the existence in the spleen of a distinct population of CD160⁺ T cells, as well as a small population of CD8αα T cells, which are enriched among Kaede-red cells (fig. S5E). Over time, while the main naïve CD4⁺ and CD8⁺ subsets faded—likely because they represented circulatory populations that just happened to be in the colon at the time of phototagging—this IEL-like population as well as other populations enriched in colon-derived cells remained stable in the spleen for several days (Fig. 4E).

We then asked the inverse question: could we identify, in a scRNA-seq dataset of intestinal T cells, those cells that gave rise to migratory populations in the spleen? We generated a scRNA-seq dataset of cecum LP T cells (Fig. 4F) and used a transfer-learning algorithm (48) to project onto these coordinates the Kaede-red immigrant cells from the spleen data. The cells that could be confidently assigned (fig. S5F) mapped to specific T cell subsets (Fig. 4F). Other than the expected naïve CD4⁺ and CD8⁺ populations and ISG-T cells, the majority of cells mapped to two clusters: effector CD4⁺ T cells (“T4eff”) and an IEL-like effector population (“IEL-like”), marked by expression of Cd160, Eomes, and multiple Klr genes (Klrc1, Klrb1c, Klrc2, Klra7, Klre1) (Fig. 4D). Other populations, like T follicular helper (“Tfh”) cells, were not represented in colon-derived cells in the spleen, despite being a sizeable subset in the colon. Thus, the T cell that emigrate from the colon include several intriguing and unexpected subsets.

Innate-like lymphocytes also emigrate from the colon

ILCs are generally considered to be mostly tissue resident (49). Although we had not tracked them systematically in our mapping experiments (Fig. 2), the scRNA-seq data revealed a small but distinct group of colon-derived cells within the ILC/NK cell cluster (Fig. 3B and fig. S4B). Indeed, a few instances of ILCs migrating out of the intestines via lymphatics have been previously reported (25, 50, 51). To examine more closely the identity of these colon-origin cells, we reclustered them on their own. This approach uncovered two principal groups distinguished by CD27 expression (Fig. 5A). Both clusters expressed Klrb1c (NK1.1), Ncr1 (NKp46), and Tbx21 (T-bet), characteristics of NK cell and type 1 ILC (ILC1) subsets (Fig. 5B) (52). The CD27^lo cluster expressed higher levels of Gzmb and Klrg1, thereby more closely resembling NK cells, whereas the CD27^hi cluster had higher expression of Ccr2 and Ctla2a. Colon-derived cells mapped to both clusters, albeit with a higher concentration in the CD27^hi cluster, fitting with its Cd160 expression (Fig. 5C). No expression was detected however for genes typical of gut type 2 (ILC2) and type 3 ILC (ILC3) populations (Fig. 5B). Flow cytometry analysis confirmed the presence of cells of recent colon origin, with the highest proportion in the NK1.1⁺NKp46⁺ subset, which would include ILC1 and NK cells, in agreement with the scRNA-seq data. Thus, NK/ILC1 populations do emigrate from the colon although not to the level of some other migrant populations.

Figure 5. Migratory ILC/NK populations found in the spleen.

A) UMAP representation of the ILC/NK cells selected from 3B. Color-coded by knn cell clusters (total 426 cells). B) Marker and differential gene expression for the CD27^hi and CD27^lo clusters from (A). C) Same plot as in (A), with kaede-green vs kaede-red distribution (309 Kaede-green cells, 117 Kaede-red cells). D) Representative flow cytometry plots and gating strategy of the ILC/NK populations in the spleen, 24hr post-photoconversion of the descending colon (left), together with quantification of the percentage of Kaede-red cells within each population (right) (Lin (lineage): CD11c, CD11b, CD3, CD19, γδ TCR, αβ TCR). Quantification pooled from three independent experiments (n=6 mice).

Intestinal perturbations modify the systemic flux of colon-derived populations

Having established in fine detail the landscape of cells migrating from the colon at baseline, we then investigated how perturbations in the gut affected this traffic. In mammalian organisms, the colon harbors the largest population and diversity of microbes, which led us to question their influence on migratory populations. We treated Kaede mice with a broad-spectrum antibiotic cocktail for 10 days (Fig. 6A), which was long enough to deeply reduce bacterial density (>10⁴-fold reduction in cultivatable bacteria in feces, both aerobic and anaerobic) but not completely reshuffle the local immunological ecosystem (fig. S6A). The migration of immunocytes phototagged in the colon to the spleen and lymph nodes was significantly reduced in antibiotic-treated mice when compared with controls (Fig. 6B and C and fig. S6, B and C). Although this reduction was only statistically significant for B cells and γδ T cells, there were clear trends for the other lineages as well, thus indicating that migration from the colon was influenced by the microbiota.

Figure 6. Intestinal perturbations significantly alter systemic migration.

A) Experimental set-up for antibiotic-treated mice. B) Percentage of colon-origin Kaede red populations in the spleen and iLN, 24hr post-photoconversion of the descending colon in mice treated with an antibiotic cocktail for 10 days (VGCA, vancomycin, gentamycin, clindamycin and ampicillin). C) Proportions of immunocytes left in the colon in the same experimental setup as in (B) D) Experimental set-up for DSS-treated mice. E) Proportions of immunocytes left in the colon 24hr post-photoconversion in control mice (Ctl) as well as different stages of DSS-induced colitis. F) Percentage of colon-origin Kaede-red populations in the spleen in the same experimental setup as in (E). G) Representative flow cytometry plots of the CD11c⁺ population in the spleen and iLN of control mice (steady-state) vs mice on day 20 post initial DSS administration, 24hr post-photoconversion of the descending colon. H) Correlation between Kaede-red CD11c⁺ in iLN (y-axis) and Kaede-red CD11c⁺ in spleen (x-axis) (left) or Kaede-red CD11c⁺ in colon-draining lymph nodes (right) in control vs mice on day 20 post initial DSS administration, 24hr post-photoconversion of the descending colon. All results from two to four independent experiments. Each dot represents a mouse, error bars: mean ± SD *p<0.05; **p<0.01, ***p<0.001, unpaired t-test.

We then asked what happens to traffic during intestinal inflammation, a question related to extraintestinal manifestations of IBD. We used the dextran sodium sulphate (DSS)–induced model of colitis, and photoconverted the distal colon at various stages of disease, from very early (48 hours after initiation of DSS) to well into the resolution phase (2 weeks after cessation of treatment) (Fig. 6D). The turnover of phototagged cells in the colon (assessed 24 hours after photoconversion) was greatly accelerated during the peak of inflammation (day 7), mostly returning to baseline values after recovery (day 20) (Fig. 6E). This higher turnover was neither a reflection of photoconversion efficiency (fig. S6D) nor of increased systemic migration, which was also seen reduced in the spleen (Fig. 6F and fig. S6E). Migration returned to mostly normal levels after DSS administration was stopped (day 20). The combination of faster turnover in the colon and reduced systemic emigration suggests that the inflammation induced by DSS leads to increased cell death and/or proliferation or luminal shedding (53). The net result is a decreased seeding of extraintestinal tissues during gut inflammation.

Unexpectedly, migration of CD11c⁺MHC-II⁺ DCs to the iLNs was markedly increased after DSS, constituting up to 1% of the total DC pool in this subcutaneous lymph node on day 20 (Fig. 6G and fig. S6F). This biased migration was not observed in the spleen or colon-draining LNs where the proportion of immigrant DCs remained similar to control mice (Fig. 6, G and H). When compared on a per-mouse basis, DC migration to the iLNs was biased relative to the spleen and colon-draining LNs, but only in DSS-treated mice (Fig. 6H; fig. S6G; and data S1). No other populations showed such a bias, with perhaps the exception of γδT cells, although to a much lesser extent (fig. S6G), suggesting that this uniquely enhanced dissemination of DCs during gut inflammation may follow different cues.

Immune cell migration from the colon to systemic sites of inflammation

Finally, we interrogated how inflammation in destination organs affects the patterns of immune cell migration using disease models that, while not affecting the gut, have previously been reported to have a connection to the gut microbiota: K/BxN arthritis (4, 6, 54), the MC38 tumor (10, 55–58) and delayed-type hypersensitivity (DTH) to keyhole limpet hemocyanin (KLH) in the ear skin (59) (Fig. 7A). After the onset of inflammation or once subcutaneous tumors were established, cells in the distal colon were phototagged by colonoscopy as above and migration to inflamed sites was assessed. Colonic egress, as measured by percentage of remaining Kaede-red immunocytes, was not affected by any of these models (Fig. 7B, fig. S7A), nor was immunocyte migration from the colon to the spleen (fig. S7B).

Figure 7. Immunocyte migration from the colon to distal diseased sites.

A) Experimental set-up for the KxB/N serum-transfer arthritis model, MC38 subcutaneous tumor model and the skin hypersensitivity DTH model. B) Proportions of immunocytes left in the colon 24hr post-photoconversion in control mice and mouse models of arthritis (KxB/N, day 10), MC38 tumors (day 11), and DTH (day 8) (n=6–12). C) Percentage of CD45⁺ cells 24hr post-photoconversion of the descending colon in the synovial fluid of KxB/N mice, MC38 tumors and the inflamed ear of DTH mice (n=6–12). D) Identity of the colon-derived immunocytes, as measured in (C). Each column represents a mouse. E, F, G) Mice examined 24hr post-photoconversion of the descending colon. Correlation between Kaede-red CD19⁺ B cells in the synovial fluid of KxB/N mice (y axis) and the spleen from the same mice (x axis) (n=11) (E); correlation between Kaede-red MNPs in MC38 tumors (y axis) and the spleen from the same mice (x axis) (n=12) (F); correlation between Kaede-red Tconv (CD4⁺ CD25⁻) in the inflamed ear of DTH mice (y axis) and the spleen from the same mice (x axis) (n=6) (G). H) Percentage of colon-derived Kaede-red cells in MC38 tumors of antibiotic-treated mice (11 days, VGCA), 24hr post-photoconversion of the descending colon. Data normalized to the average of each control (n=3–7). I) Percentage of colon-derived Kaede-red cells in MC38 tumors of mice pre-treated with anti-CXCR3 (250μg/mouse) and anti-CCL2 (250μg/mouse) antibodies, 24hr post-photoconversion of the descending colon. Data normalized to the average of each control (n=5–7). All results from two to four independent experiments. Each dot represents a mouse, mean is marked. *p<0.05; **p<0.01, unpaired t-test.

Although inflammation in extraintestinal tissues did not change the rate of migration out of the colon, entry into these various sites of inflammation did show distinct characteristics. The overall numbers of recent emigrants from the colon was roughly equivalent in all three affected tissues (Fig. 7C), but the identity of the cells infiltrating each tissue was quite different (Fig. 7D, S7C, data S2), with over-representation of those cell-types in these locations relative to the spleen in the same mice (Fig. 7E–G). In the synovial fluid, colon-derived Kaede-red cells mostly comprised B cells (Fig. 7, D and E).

By contrast, MC38 tumors were predominantly infiltrated by myeloid cells from the colon, with CD11c⁻CD11b⁺F480⁺ macrophages the dominant colon-derived population. This over-representation was notable because macrophages were not a dominant migratory population elsewhere, and indeed biased relative to the spleen of the same mice (Fig. 7F). This increase was not mirrored in the tumor-draining iLN, where number of macrophages if anything decreased slightly (fig. S7D). We confirmed by intravascular labeling that the total pool of colon-derived immunocytes, as well as the dominant MNP populations, were not simply intravascular cells (fig. S7, E and F). Given the influence of macrophage subtypes on tumor progression, this influx suggests a possible explanation for the influence of the microbiome on tumor immunotherapy (8–10). Moreover, these data argue that tumor-infiltrating macrophages can come from other locations than simply differentiating from blood monocytes (60). Much as migration of gut populations could be curtailed by reducing bacterial loads (Fig. 6B), antibiotic treatment during tumor growth also diminished the entry of colon-derived immunocytes in the tumors (Fig. 7H). For mechanistic insight, we treated tumor-bearing mice at the time of photoconversion with blocking antibodies against CXCR3 and CCL2, which are both used by immunologic populations to infiltrate MC38 tumors (61). This blockade reduced entry of colon-derived T cell populations into the tumors, but had no effect on their proportions within the spleen (Fig. 7I). MNPs were not affected however, perhaps as a result of redundancy in their chemokine receptors. These data indicate that once these immunocytes exit the intestines, they use the same mechanisms as other circulatory populations to traffic to and enter distal tissues.

Finally, although a mix of immune cells were recruited to ears with DTH, CD4⁺ Tconv and γδ T cells were over-represented relative to other locations, with hardly any B cells migrating to the inflamed skin (Fig. 7, D and G, and fig. S7C). This homing contrasts with the absence of migration to the skin at baseline (Fig. 2), indicating that inflammation creates new destinations for colon-derived immunocytes, with distinct cell-type specificity according to the inflammatory lesion.

DISCUSSION

Half a century ago, Hall and Smith postulated that any immune response must incorporate cells from the intestines, arguably the largest immunologic organ (62). Although cell trafficking between lymphoid organs has been studied for decades (63, 64), it is only recently that that cell migration from the intestines has been queried (15, 18, 21, 25, 29). Building on our previous work with the Kaede system (15), we set out to map in more detail these migratory paths from the gut to other organismal locations, their cellular and molecular specificity, and how these patterns are influenced by immunologic challenges. We uncovered clear cell specificity in terms of the emigrating subsets and their preferential destinations, ferrying antigen-loaded cells and specific immune cell phenotypes, several of which were unexpected. It may be worth emphasizing that observations result from a single pulse-label of what is really a constant conveyor belt emanating from the gut, and the factions observed underrepresent the cumulative numbers of colon-derived immunocytes in tissues.

Traffic to other intestinal sites did not appear to be privileged, although we could not evaluate how much re-entry into the colon was occurring. Egress from the colon appeared to be principally controlled by S1P-related phenomena, as the S1PR antagonist FTY720 caused complete retention of immunocytes in the colon. Although most sphingolipids are derived from diet or de novo synthesis in tissues, gut bacteria can also produce them (65), including analogs with higher specificity than S1P itself for some of the S1PR family members (66). One might speculate that particular microbes control immune cell transfer from the gut to systemic locations by affecting, accounting for the modifications we observed after antibiotic depletion or during the dysbiosis resulting from DSS colitis. Microbe-related variations in adhesion molecules, and specifically the MadCAM–α₄β₇ integrin molecular pair, have recently been shown to control the emigration of RORγ⁺ Treg cells into tumors (10). The two mechanisms are not mutually exclusive, however, as MadCAM–α₄β₇ adhesion and S1p/S1Pr signalling are likely complementary controllers of cell mobilization.

B cells numerically dominated traffic from the colon and were found at every destination. Surprisingly, the majority of these immigrant B cells were of the common follicular type and although they carried a genomic signature of their gut residence it is not obvious what these cells might bring to the systemic tissues. B cell emigration may only represent the exit phase of a strong flow of B cells through the gut, whose function would be to present the widest fraction of the B cell repertoire in order to exploit all microbe-specific immunoglobulins it may contain (in a sense, an evolutionary equivalent in mammals of birds’ bursa of Fabricius?).

By contrast, migrant T cells included more striking differentiated phenotypes, suggestive of microbial imprints. The combined single-cell profiling, with oversampling of the Kaede-red populations, resolved distinct subsets that would otherwise have been missed if aggregated with other Teff cells by the clustering process. One such population of gut origin was the CD160⁺ T cell group, which included both αβ and γδ T cells. These cells exhibited a pattern of gene expression reminiscent of IEL populations (e.g., Cd8a, Itgb7, Klr genes), which is intriguing given that IELs are generally considered to be tissue-resident. IELs did show turnover in the colon, albeit more slowly than cells from the LP. These results concur with a recent study documenting the emigration of several tissue-resident, non-classical T cells to local draining lymph nodes (31). Thus, while many IELs may be tissue-resident, some can enter the general circulation and seed systemic organs. Their impact there remains unclear, but their unique access to both epithelia and microbial antigens, and their innate-like responsiveness, suggests they may serve as unique modifiers of immunologic balances outside of the gut (67).

Another distinct population of colon origin uncovered in the spleen was the RORγ⁺ T cell, also composed of both αβ and γδ T cells. IL-17–producing T cells in the gut, which depend on the microbiota (68–71), help maintain local homeostasis (72–74). However, they also exhibit pathogenic roles in extraintestinal autoimmune disorders (4, 5, 19, 20, 32), with previously reported migration of gut Th17 in these pathological settings. It has been proposed that gut Th17 cells can be variably protective or pathogenic (75, 76), and it is interesting that a robust emigration of RORγ⁺ T cells from the gut occurs at steady-state in unperturbed mice.

The present study showed that ISG^hi phenotypes, previously described in “ISG-T” cells (40–42), also exist in B cells, as a small subset characterized by unusually high ISG expression. They may result from either a particular sensitivity in some cells to normal levels of IFN, or from high exposure to IFN in a particular microenvironment. The over-representation of ISG-T and ISG-B cells among gut emigrants in systemic locations prompts the speculation that they may generally derive from the gut, noteworthy given that the mucosa is a frequent point of entry for pathogens.

DCs moving from the colon to the draining caLN represented the highest migration rate, in terms of proportional representation, as 10% of the caLN DC pool came from the colon over a 24-hour period, dwarfing any other frequencies. This corresponds to the classic ferrying of antigen by DCs to local draining LNs. DCs also trafficked to other non-intestinal LNs, however, especially during the resolution phase of gut inflammation, when they ended up comparatively more abundant in subcutaneous LNs than in the spleen or even in the colon-draining LNs (Fig. 6H). Accordingly, intestinal DCs have been shown to reroute through lymphatics to peripheral sites following infection (28), although other studies found only extremely rare DC traffic in lymph (77). It is possible that these DCs directly exit from the LP to blood after tissue damage (78). Whatever the route these gut DCs take, their over-representation in many systemic LNs may facilitate a spreading of responses to antigens exposed by gut damage.

Our results show that gut-educated populations also homed to sites of inflammation, with a cell specificity that varied with the type of lesion. It was in DTH lesions that T cells were most present, perhaps not coincidentally since DTH is a T cell-mediated disease. The relative dominance of myeloid cells in MC38 tumors was intriguing. Tumor-infiltrating MNPs are thought to be largely of monocyte origin, but our results, in agreement with a recent report (10) indicate a sizeable flow of MNPs from the gut, which may relay the documented effects of the gut microbiota on tumor sensitivity to checkpoint blockade (8).

In conclusion, one can consider the gut as the origin of a conveyor belt of immunocytes than fan out from the intestines to many organs and tissues, ferrying antigens and effector functions in a cell- and tissue-preferential manner. This traffic is modified by inflammation at either of its origin or destination locales, which may provide explanations for the distant effects of perturbations of the gut microbiota, and new approaches to the modulation of immunologic disorders.

MATERIALS AND METHODS

Mice

Kaede transgenic mice were originally obtained from O.Kanagawa (RIKEN, Wako, Japan), and bred on a C57Bl/6 background in our specific-pathogen-free facility. Unless otherwise stated, both male and female mice, 6–8 weeks of age were used. KRN transgenic mice were crossed to NOD/LtJ to generate arthritogenic serum as described (79). All experimentation performed following the animal protocol guidelines of Harvard Medical School (protocol IS00001257; IS00000196)

In vivo treatments and disease models

Intravascular labeling was performed as previously described (34). Mice were injected via the tail vein with 2 μg of anti-CD45 antibody (clone 30F-11, BioLegend) in sterile PBS. Tissues were then harvested 3 min later. To detect CD45 by flow cytometry, a different antibody clone was used (anti-CD45.2, clone 104, BioLegend).

Mice were pretreated via intraperitoneal (ip) injections with either S1P receptor antagonist FTY720 (1 mg per kilogram of body weight, Cayman Chemical), anti-CXCR3 antibody (BioXcell, clone CXCR3–173), anti-CCL2 antibody (BioXcell, clone 2H5) or anti-Armenian hamster IgG (BioXcell, polyclonal), all 250 μg per mouse, at the time of endoscopic photoconversion.

For treatment with antibiotics, vancomycin (0.5 mg/ml, VWR Life Science), gentamycin (0.5 mg/ml, VWR Life Science), clindamycin (0.5 mg/ml, Alfa Aesar) and ampicillin (1 mg/ml, Sigma) were dissolved in drinking water with sweetener (‘Equal’ 1mg/1mL, Target) and given to the mice for 10 days. This regimen routinely yielded a 10⁴-fold reduction in intestinal bacteria. For DSS treatment, 2.5% DSS (Thermo Scientific) was provided in the drinking water for 6 days.

For tumor induction, MC38 cells were first cultured in vitro in DMEM supplemented with 10% FCS and 1% L-glutamine for a week, prior to subcutaneous injection of anesthetized mice with 10⁶ cells in 100 μl of PBS.

K/BxN serum-transferred arthritis was induced by ip injection of 150 μl of pooled serum from 8-week-old K/BxN mice on days 0 and 2. Arthritis development was assessed by visual inspection of the paws and caliper (Kafer, 10 mm range with flat anvils) measurement of ankle thickness.

For the DTH model, mice were anesthetized, backs shaved and injected subcutaneously with KLH (Sigma, 1 mg/ml) emulsified in complete Freund’s adjuvant (CFA) on four different back sites (50 μl altogether). Eight days later, the mice were intradermally injected with 10 μl of KLH (dissolved in sterile PBS, 1 mg/ml) into one of the ears. The other ear was injected with the same volume of PBS as a control.

Photoconversion procedures

The colon was photoconverted as previously described (15). Briefly, a custom-built fiberoptic endoscope (ZIBRA Corporation) was connected to a handheld 405 nm blue purple laser (≤5 mW) via an in-house custom-made connection device (fixed mounts from ThorLabs). After cleaning the colon of anesthetized mice with PBS, the endoscope was inserted through the anus to a depth of 3 cm. The laser was switched on, exposing the inner colon to violet light (3.5-mm beam diameter). The endoscope was then gently retracted, pausing at 2-mm increments for 30-s light pulses at each interval (for a total of up to 10 min).

Immune cell isolation from tissues

Spleen, bone marrow and lymph nodes

Immunocytes were released by mechanical disruption followed by filtration (70 μm filter) and washed in RPMI containing 5% fetal calf serum (FCS). Red blood cells in the spleen and bone marrow were lysed using ACK lysing buffer (Gibco, ref A10492–01).

Colon and small intestine

Immune cells were isolated as previously described (29). Briefly, intestines were cleaned (Peyer’s patches removed in the case of the small intestine), and treated with RPMI containing 1 mM DTT, 20 mM EDTA, and 2% FCS at 37°C for 15 min to remove epithelial cells. They were then minced and dissociated in collagenase solution (1.5 mg/ml collagenase II (Gibco), 0.5 mg/ml dispase (Gibco), and 1% FCS in RPMI) with constant stirring at 37°C for 40 min. Single-cell suspensions were filtered (70 μm filter), washed with RPMI containing 5% FCS and centrifuged (520g, 10 min) to pellet the cells.

Liver, lungs, and kidneys

Mice were first perfused with 5 ml of ice-cold PBS through the left ventricle. Tissues were minced and dissociated in collagenase solution (0.5 mg/ml collagenase IV (Gibco), 150 μg/ml DNase I (Sigma), and 1% FCS in DMEM) and incubated in a water bath at 37°C with constant shaking for 40 min. Digested tissues were filtered (70 μm filter) and washed in 2% FCS. For the liver and kidneys, immune cells were enriched by Percoll (GE Healthcare) density centrifugation (36%, 800g for 10 min). For lungs, red blood cells were lysed using ACK lysis buffer.

Thymus

Tissue was chopped into small pieces in RPMI with 25 mM HEPES and 2% FCS. Following centrifugation (520g for 5 min), supernatant was removed and cells digested in collagenase solution (0.5 mg/ml collagenase type D (Sigma), 0.2 mg/ml DNase I (Sigma), and 2% FCS in RPMI) at room temperature for 30 min with constant shaking. Digested tissue was filtered (70 μm filter) and then washed in 2% FCS.

Skin

The dorsal and ventral halves of each ear were separated, minced, and incubated in collagenase solution (3 mg/ml of collagenase IV (Gibco), 0.1 mg/ml of DNAse I (Sigma), 0.5 mg/ml of hyaluronidase (Sigma)) for 30 min at 37°C with constant shaking. Digested tissue was filtered (70 μm filter) and washed in 2% FCS.

Tumors

Subcutaneous tumors were excised from the abdomen area with skin and fat carefully removed. They were minced in collagenase solution (1 mg/ml of collagenase IV (Gibco), 20 μg/ml of DNAse I (Sigma), and 2% FCS) and incubated at 37°C with constant shaking for 20 min. Digested tissues were filtered (70 μm filter) and washed in 2% FCS.

Synovial fluid:

synovial fluid was obtained by puncturing the inflamed synovium with a 25G needle and collecting the extracted fluid into DMEM with 5% FBS and 20 mM EDTA.

Flow cytometry

Cells were stained with surface marker antibodies (details in table S1). To identify NK cells/ILCs, lineage (Lin) negative gating was determined based on CD11c, CD11b, CD3, CD19, γδ TCR, and αβ TCR expression. Cells were stained for 25 min at 4°C, washed, and then acquired on the BD FACSymphony. Analysis was performed with FlowJo software. Flow t-SNEs were generated with the t-SNE FlowJo package. Chord diagrams of flow quantifications were generated using the Circlize package in R (80)

Confocal imaging

The descending colon was collected and fixed in 2% PFA overnight, followed by incubation in 30% sucrose for 12–24 hours. Tissues were then embedded in OCT and frozen at −80°C. Cryostat sections were kept at −20°C until imaging. Sections were mounted on 100% glycerol and covered with a 1.5 coverslip. Tile images of the section were acquired in confocal mode on a DM6 microscope stand, equipped with a second-generation supercontinuum While Light laser (440 – 790 nm) and motorized Scientifica stage with a single slide insert. Images were acquired with a multi-immersion Plan Apo 20x/0.8 objective lens, using glycerol as immersion media and adjusting the correction collar to minimize spherical aberration. Non-photoconverted Kaede green and photoconverted Kaede red were excited sequentially with 487 nm or 561 nm, setting the transmittance to 0.5% and 3% respectively. The emission range for Kaede green was set to 492–543 nm and collected with HyDS GaASP detector on analog mode with 15 amplification gain, whereas the range for Kaede red was set to 562–660 and collected with a HyDX detector in standard mode with gain set to 20. Images were acquired as 12-bit 400 Hz scan rate, zoom 1 and line average of 4. The pinhole size was set to 1AU, using 580 nm as a reference wavelength (59.4 μm diameter). Tiled images were collected using the Navigator application in LASX 4.4.0.2486, and analysed using ImageJ.

Single-cell RNA sequencing

Sample preparation

scRNA-seq experiments were performed and analysed as previously described (40). Cecum cell suspensions were stained on ice for 20 min with anti-CD45, anti-CD4, anti-CD19, anti-CD8, anti-CD11b, anti-CD11c, anti-γδ TCR, anti-NK1.1, and DAPI as a viability dye. Cells were sorted as CD45⁺DAPI⁻. Additional sorting was performed for CD19⁻CD45⁺, NK1.1⁺CD45⁺, and TCRγδ⁺CD45⁺ cells to enrich for these populations. Splenic cell suspensions were stained with anti-CD45 and DAPI as a viability dye (BioLegend). Kaede green and Kaede red cells from each mouse were sorted as CD45⁺DAPI⁻ cells into 0.04% BSA. TotalSeq-A hashtag antibodies (0.5 μl per sample for 20 min, BioLegend; hashtag 1, 24 hour Kaede green mouse #1; hashtag 2, 24 hour Kaede green mouse #2; hashtag 3, 48 hour Kaede green mouse #3; hashtag 4, 48 hour Kaede green mouse #4; hashtag 5, 72 hour Kaede green mouse #5; hashtag 6, 72 hour Kaede green mouse #6; hashtag 7, 24 hour Kaede red mouse #1; hashtag 8, 24 hour Kaede red mouse #2; hashtag 9, 48 hour Kaede red mouse #3; hashtag 10, 48 hour Kaede red mouse #4; hashtag 11, 72 hour Kaede red mouse #5; hashtag 12, 72 hour Kaede red mouse #6) were added to each sample after sorting. All samples were washed and then pooled together, centrifuged at 520g for 5 min, and resuspended in 0.04% BSA. Encapsulation was done on the 10X Chromium microfluidic instrument (10X Genomics). Libraries were prepared using Chromium Single Cell 3′ reagents kit v2 according to manufacturer’s protocol. Hashtag oligonucleotide (HTO) libraries were prepared as described in (39). Libraries were sequenced together on the Illumina HiSeq X. Cecum scRNA-seq was performed independently from the spleen.

Data analysis

scRNA-seq data were processed using the standard CellRanger pipeline (10X Genomics). HTO counts were obtained using the CITE-seq-Count package (81). Data was analysed in R using the Seurat package (82). HTOs were assigned to cells using the HTODemux function, and doublets were eliminated from analysis. Cells with fewer than 700 UMIs or 500 genes and more than 2500 UMIs, 10,000 genes, and 5% of reads mapped to mitochondrial genes were also excluded from the analysis. Dimensionality reduction, visualization and clustering analysis were performed in Seurat using the NormalizeData, ScaleData, FindVariableGenes, RunPCA, FindNeighbours (dims=1:30), RunUMAP (dims=1:30) and FindClusters functions. Cluster identity was determined based on expression of key marker genes (fig. S3A). The SubsetData function was used to remove individual B and T cell clusters for further analysis. Differentially expressed genes between Kaede red and Kaede green were obtained using the FindMarkers function on each of the top four Kaede red-enriched clusters and non-redundantly collated using a for-loop for each timepoint. The top 200 upregulated and downregulated genes across each of the three lists were selected and then also collated resulting in a final differentially list of genes generated based on statistical analysis (Student’s t test, P<0.01). Heatmaps of differentially expressed genes were generated using Morpheus (Broad Institute). Differential cell expression between Kaede red and Kaede green cells was visualized using the Buencolors package. Reference mapping of the Kaede red cells onto a colonic reference dataset was performed following Seurat’s data integration tool (48), using the FindTransferAnchors and MapQuery functions.

To identify iNKT cells expressing the canonical TCR, TCR V regions expressed in each cell were assembled using the TRUST4 algorithm (83) and searched for cells utilizing the canonical TRAV11/TRAJ18 combination. Twenty-seven such cells were found in the dataset, all of which shared the typical VVGDRGSAL CDR3a sequence and mapped to the NKT cluster of Fig. 4A.

Gene signatures

The IEL gene signature was based on the expression of the following marker genes: Klra1, Klre1, Klra7, Itgae, Cd160, Klrk1, Fasl, Itgb7, Ccr9, Cd8a. The NKT gene signature was on the expression of the following marker genes: Il4, Atp8a2, Trpm6, Garnl3, Hpn, Apt6v0d2, Klrb1c, B3galt5, Gpnmb, Rnf144b, Serpina3f, Zbtb16, Klra1, Sulf2, Rin2, Art2a-ps Arg1, Gm9195, Pls1, Wnt10a, Fam84a, Csf2, Zfp683, Mgll, Apol7b, Dgki, Cd93, Fbxl21, Fcrls, Chad, Flt4, Rmdn2, Klrb1b, Ckb.

Statistical analysis

Statistical analyses outside of single-cell data were performed using GraphPad Prism software. Unless stated otherwise, data is presented as mean ± SD. Statistical significance was calculated by unpaired Student’s t test. P<0.05 was considered significant.

Supplementary Material

Suppl Figures-Table-Legends

fig. S1. Specific photoconversion of the descending colon

fig. S2. Flow cytometry gating strategy and photoconversion percentages

fig. S3. Representation of colon-derived immune cells across peripheral tissues

fig. S4. scRNA-seq analysis of local and colon-derived CD45+ cells in spleen

fig. S5. scRNA-seq analysis of local and colon-derived T cells in spleen

fig. S6. Immune cell population dynamics and photoconversion in antibiotic or DSS-treated mice

fig. S7. Colon egress and spleen migration of T cell populations in arthritis, tumor and skin hypersensitivity

Table S1. Antibody information

Data S2

Data S2. Number of colon-origin migratory cells in mouse models of disease

Data S3

Data S3. Datapoints for figures

Data S1

Data S1. Number of colon-origin migratory cells in the spleen and iLN in a mouse model of DSS-induced colitis

Data and materials availability:

Single-cell data originally reported in this paper were deposited in the Gene Expression Omnibus (GEO) database under accession number GSE216928. Tabulated underlying data for Figs. 1, 2, 5, 6, and 7 and figs. S1, S2, S3, S6, and S7 are archived in data S3. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.