Myeloid Cells and Sphingosine-1-Phosphate Are Required for TCRαβ Intraepithelial Lymphocyte Recruitment to the Colon Epithelium

Intraepithelial lymphocytes (IELs) are a unique, heterogeneous population of T cells that are defined by their presence within the epithelial layer of the small intestine (SI) and colon. The IELs of the SI are composed primarily of TCRαβ⁺ CD8αα⁺ and TCRγδ⁺ T cells, which develop in cryptopatches (1), with a smaller CD4⁺ CD8⁺ double-positive TCRαβ⁺ subset that develops from a CD4⁺ Foxp3⁺ precursor population in response to antigenic stimulation at the mucosa (2). By contrast, the IELs of the human and murine colon are primarily TCRαβ⁺ (3, 4) and composed of CD4⁺, CD8⁺, and CD4⁻ CD8⁻ double-negative (DN) subsets. It is believed that they are largely thymically derived (3, 5), although much less is known about the ontogeny of colon IELs compared with those in the SI. Functionally, the IELs of the SI have been shown to play a key role in the response to enteric pathogens (6) while simultaneously maintaining important regulatory functions (2). Due to the distinct functions and bacterial burdens of the colon versus the SI, colon IELs must be interrogated as a distinct set of cells to elucidate their function and regulation.

Colon IELs are critical for epithelial homeostasis and repair after injury. In the absence of functional IELs, mucus thickness is reduced and epithelial permeability is increased due to the loss of tight junction formation (7). Following injury by wounding, dysregulated inflammation, or infection, colon IELs upregulate IL-6 expression to stimulate epithelial proliferation (7, 8). Intriguingly, IL-6 production by IELs is regulated by bacterial presence, suggesting a high degree of crosstalk between colon IELs and microbial communities of the colon (7). Within ulcerative colitis and Crohn’s disease, colon IELs display signs of increased inflammatory capacity associated with alterations in specific bacterial taxa (9). Therefore, elucidating the mechanisms of colon IEL recruitment, development, and activation is critical for understanding colon tissue homeostasis.

Prior work has demonstrated that both bacteria and host expression of MyD88 are required for IEL presence in the colon. Colon IELs are absent in both germ-free and antibiotic-treated mice. However, monocolonization with a number of bacterial strains within the order Bacteroidales is sufficient to restore IEL presence in the colon (7). Germline deficiency of MyD88 also results in loss of IELs from the colon (7). Therefore, because MyD88 is critical for the signaling competence of extracellular TLRs and the IL-1β receptor, we hypothesized that colon IEL recruitment is dependent upon bacteria-stimulated MyD88 signaling in tissue-resident colon cells.

In this study, we demonstrate that MyD88-dependent sensing of microbial products by myeloid cells is required for the recruitment of colon TCRαβ⁺, but not TCRγδ⁺, IELs. We further identify sphingosine-1-phosphate (S1P) production by bacteria-stimulated myeloid cells as an important means through which recruitment occurs. Finally, we show that ablation of MyD88 signaling in the myeloid cells of TNF^ΔARE/+ mice with intestinal inflammation results in an alleviation of histologic disease, potentially due to a loss of IEL-mediated inflammation. These results demonstrate a mechanism by which colon IELs are recruited and highlight IEL importance in the mediation of inflammatory disease.

C57BL/6 and KikGR [Tg(CAG-KikGR)33Hadj/J] mice were obtained from The Jackson Laboratory and maintained as a local colony in specific pathogen-free (SPF) housing conditions. Villin-Cre, Lck-Cre, LysM-Cre, and MyD88^fl/fl mice were obtained from The Jackson Laboratory and crossed for the desired genotypes. TNF^ΔARE/+ mice were a gift from Sean Colgan (10), and Rag1^−/− mice were obtained from The Jackson Laboratory. Cre⁺ mice were compared with Cre⁻ littermate controls. Germ-free C57BL/6 mice were obtained from the National Gnotobiotic Rodent Resource Center at the University of North Carolina and maintained as a local colony bred and housed in sterile vinyl isolators at the University of Colorado Anschutz Medical Campus Gnotobiotic Core Facility. For all studies, both male and female mice aged 8–10-wk-old were used. TNF^ΔARE/+ mice were used at 14 wk of age or older to allow chronic inflammation to occur (10). All animal experiments were approved by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee.

Depletion of the microbiome was carried out using a solution of 1% grape-flavored Kool-Aid (Kraft), 0.01% ampicillin, 0.01% neomycin, 0.05% metronidazole (Research Products International), and 0.05% vancomycin (Alfa Aesar) delivered ad libitum in the drinking water for 7 d. Bacterial depletion was evaluated by quantitative PCR (qPCR) of fecal samples after 4 d of antibiotic treatment, using primers against bacterial rpoB (forward primer: 5′-AACATCGGTTTGATCAAC-3′; reverse primer: 5′-CGTTGCATGTTGGTACCCAT-3′). Recolonization of antibiotic-treated SPF housed animals was done by introducing mixed dirty bedding from untreated littermates to treated cages. Recolonization was assessed by qPCR as stated above. Colonization of germ-free mice was done by oral gavage of 100 μl of 100 μg cecal contents in PBS. Inhibition of IEL recruitment was carried out using daily i.p. injections of the S1PR antagonist FTY720 (Cayman Chemical) at a concentration of 1 mg/kg. As a control, mice were injected with 10% DMSO in PBS (Life Technologies).

For photoconversion, mice were anesthetized using inhaled isoflurane and oxygen. A 7-French (2.3-mm) rigid colonoscope (Storz) was inserted into the rectum to a depth of 25 mm or to the splenic flexure, as visualized via the camera. Photoconversion of epithelial cells was performed by shining 405-nm light through the colonoscope using a 100-mW handheld laser pointer in 15-s pulses. The scope was retracted 5 mm between pulses until it was removed from the colon. Anesthesia was removed, and mice were monitored until recovered.

Isolated colon tissue was minced and shaken at maximum speed on a benchtop vortex for 10 min in 10 ml PBS (Life Technologies) supplemented with 1 mM EDTA and 7.5 mM HEPES to isolate the epithelial fraction as previously published (7). Cells were passed through a 70-μm cell strainer (Corning) and stained for flow cytometry or magnetically enriched.

Colon IELs were isolated for adoptive transfer after epithelial fractionation using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies) following the manufacturer’s recommended protocol with the addition of 0.5 ng/ml biotin-labeled anti-mouse EpCAM (clone G8.8, Invitrogen). For transfer into recipient mice, 1 × 10⁵ isolated IELs were injected via tail vein into each Rag1^−/− recipient mouse. T cell reconstitution was confirmed by weekly cheek bleed and flow cytometry. Tissues were harvested from recipient animals 8 wk after injection.

Lamina propria cells were isolated after epithelial separation. Tissue was finely minced and incubated for 30 min at 37°C in 0.5 mg/ml type IV collagenase (Worthington) and RPMI 1640 (Life Technologies) supplemented with 10% FBS (Thermo), 1 mM EDTA, and 7.5 mM HEPES. Digested tissue was passed through a 70-μm cell strainer (Corning) and washed three times with RPMI 1640. Lamina propria T cells were isolated by negative selection using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies) for microarray analysis, and CD11c⁺ cells were isolated for qPCR with the EasySep Mouse CD11c Positive Selection Kit (STEMCELL Technologies) using the manufacturer’s recommended protocol.

Staining took place at 4°C with 1% of the indicated Abs in PBS supplemented with 10% FBS (Sigma-Aldrich) for 30 min. T cells were stained with the following Abs: CD4 (Tonbo, clone GK1.5), CD8α (Tonbo, clone 53-6.7), CD8β (BioLegend, clone Ly-3), TCRβ (Τonbo, clone H57-597), and TCRγδ (Invitrogen, clone GL-3). CD11c⁺ cells were stained with the following Abs: CD11c (BioLegend, clone N418), I-A/I-E (eBioscience, clone M5/114.15.2), CD103 (BD Biosciences, clone M290), CD64 (eBioscience, clone X54-5/7.1), CX3CR1 (BioLegend, clone SA011F11), and Ly6C (BioLegend, B358688). Viability was assessed using Ghost Dye Violet 510 (Cytek). Screening of TCR Vβ fragments was done using the BD Pharmingen Anti-Mouse TCR Vβ Screening Panel. Cells were fixed using the Foxp3/Transcription Factor Fix/Perm buffers (Tonbo) at 4°C for 20 min. Cells were then washed and analyzed on a Becton Dickinson LSR Fortessa X-20 or Cytek 5-laser Aurora. Data analysis was carried out in FlowJo version 10 (FlowJo LLC). Cells were first gated on lymphocytes or monocytes as determined by forward and side scatter, and doublets were removed using the ratio between the area and height of the peak by forward scatter. Cells that did not stain with viability dye were then quantified by the indicated markers.

Epithelial and lamina propria cells were harvested as above. A total of 10 C57BL/6 mice (5 males and 5 females) at 8–10 wk of age were pooled for each group of recolonized IELs and lamina propria T cells. RNA was isolated using TRIzol (Thermo Fisher), and integrity was confirmed using an Agilent Bioanalyzer (Agilent). RNA was prepared according to the Affymetrix Expression Analysis Technical Manual (Affymetrix). Briefly, first- and second-strand cDNA synthesis was performed followed by cleanup of double-stranded cDNA. Antisense cRNA was biotin labeled for 4 h in an in vitro transcription reaction, and 20 µg biotin-labeled cRNA was fragmented and hybridized. Microarray analysis was performed using Affymetrix gene chips (HuFL-U133 plus2) by the University of Colorado Cancer Center Genomics Shared Resource. Data were analyzed using Ingenuity Pathway Analysis (Qiagen). Full microarray data are deposited in the Gene Expression Omnibus under accession number GSE197729.

For qPCR, RNA was isolated from live cells using the RNeasy Mini Kit (Qiagen) and processed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories). qPCR was run using lo-ROX qPCR master mix (Tonbo) on an ABI 7500 real-time PCR system. The following genes were assayed by qPCR: β-actin (forward primer: 5′-TACGGATGTCAACGTCACAC-3′; reverse primer: 5′-AAGAGCTATGAGCTGCCTGA-3′), Sphk1 (forward primer: 5′-AAAATACTGAGAAACTCGGTCGG-3′; reverse primer: 5′-GCATCGCTTCTTAAAGTCCAGA-3′), Sphk2 (forward primer: 5′-CACGGCGAGTTTGGTTCCTA-3′; reverse primer: 5′-CTTCTGGCTTTGGGCGTAGT-3′), Spns2 (forward primer: 5′-GATCTTCTAGCCCTGACCTGC-3′; reverse primer: 5′-CAGATGGGAGGTGAAGCTCTG-3′), S1pr1 (forward primer: 5′-ATGGTGTCCACTAGCATCCC-3′; reverse primer: 5′-CGATGTTCAACTTGCCTGTGTAG-3′), Ahr (forward primer: 5′-AGCCGGTGCAGAAAACAGTAA-3′; reverse primer: 5′-AGGCGGTCTAACTCTGTGTTC-3′), Ccr7 (forward primer: 5′-TGTACGAGTCGGTGTGCTTC-3′; reverse primer: 5′-GGTAGGTATCCGTCATGGTCTTG-3′), Cd37 (forward primer: 5′-GCCCAAGAGAGTTGCCTCAG-3′; reverse primer: 5′-GGCCGCCTAGTACAAAGAAGAA-3′), Cd96 (forward primer: 5′-TGGGAAGAGCTATTCAATGTTGG-3′; reverse primer: 5′-AGAGGCCATATTGGGGATGATAA-3′), Emb (forward primer: 5′-TGAGGGCGATCCCACAGAT-3′; reverse primer: 5′-CCGTCACTGAGATATTACAGCTC-3′), Fmnl1 (forward primer: 5′-CTGCTGAGCCAGTATGACAATG-3′; reverse primer: 5′-CGGTATCCAGGTAGCTCTTCA-3′), Fn1 (forward primer: 5′-ATGTGGACCCCTCCTGATAGT-3′; reverse primer: 5′-GCCCAGTGATTTCAGCAAAGG-3′), Fxyd5 (forward primer: 5′-CAGGGCCACATACAAGCAG-3′; reverse primer: 5′-GCATGAAGTTTTTGGATGGGC-3′), Gpr18 (forward primer: 5′-CACCCTGAGCAATCACAACCA-3′; reverse primer: 5′-AGTGACATTAACAAACAGCCCA-3′), Il7r (forward primer: 5′-GCGGACGATCACTCCTTCTG-3′; reverse primer: 5′-AGCCCCACATATTTGAAATTCCA-3′), Il21r (forward primer: 5′-GGCTGCCTTACTCCTGCTG-3′; reverse primer: 5′-TCATCTTGCCAGGTGAGACTG-3′), Itga4 (forward primer: 5′-GATGCTGTTGTTGTACTTCGGG-3′; reverse primer: 5′-ACCACTGAGGCATTAGAGAGC-3′), Itgb2 (forward primer: 5′-CAGGAATGCACCAAGTACAAAGT-3′; reverse primer: 5′-CCTGGTCCAGTGAAGTTCAGC-3′), Ms4a4b (forward primer: 5′-TGACACTTCAACCATTGCTACC-3′; reverse primer: 5′-ACACATTTCCTGGAACATTGGTC-3′), Msn (forward primer: 5′-TCTTATGCCGTCCAGTCTAAGT-3′; reverse primer: 5′-GGTCCTTGTTGAGTTTGTGCT-3′), Myo1g (forward primer: 5′-GGCCCTGAGTATGGGAAACC-3′; reverse primer: 5′-GATACGAGCACCTCACCAATG-3′), Sell (forward primer: 5′-TACATTGCCCAAAAGCCCTTAT-3′; reverse primer: 5′-CATCGTTCCATTTCCCAGAGTC-3′), Selplg (forward primer: 5′-GAAAGGGCTGATTGTGACCCC-3′; reverse primer: 5′-AGTAGTTCCGCACTGGGTACA-3′), Timp2 (forward primer: 5′-TCAGAGCCAAAGCAGTGAGC-3′; reverse primer: 5′-GCCGTGTAGATAAACTCGATGTC-3′).

Bones were harvested from euthanized animals, and bone marrow was isolated. In each well of a 6-well culture plate (Corning), 1 × 10⁷ isolated bone marrow cells per well were cultured in 4 ml RPMI (Life Technologies) supplemented with 20 ng/ml recombinant GM-CSF (BioLegend), 0.1% 2-ME, 2 mM penicillin and streptomycin, 2 mM sodium pyruvate, and 2 mM l-glutamine. Cells were cultured at 37°C and 5% CO₂. On days 2 and 3 of culture, half of the spent media was aspirated and replaced with fresh media. Cells were harvested on day 6 and plated at 5 × 10⁵ cells per well in 96-well cell culture plates (Corning). Cells rested for 24 h before use. Cecal contents were harvested and weighed before being resuspended in 1 ml PBS and heat killed at 60°C for 10 min. Contents were diluted to 1 ng/μl, and 5 μl was added to each well.

Bone marrow–derived myeloid cells (BMDMs) were cultured as above, and 1.5 × 10⁶ cells per well were plated on the bottom of 12-well cell culture plates with a 5-µm Transwell setup (Corning) and allowed to rest for 24 h in complete media. Cells were stimulated for 8 h with 5 ng/ml of isolated cecal contents from SPF-housed C57BL/6 donors. After 8 h, media were replaced with lipid- and serum-free media. T cells were magnetically enriched from whole spleens using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies), plated in the upper chamber of the Transwell, and allowed to migrate for 4 h. After 4 h, cells were harvested and enumerated by flow cytometry, with migration expressed as the percentage recovered from the bottom well. For FTY720 treatment, T cells were incubated in 100 nM FTY720 for 1 h prior to addition to the upper Transwell chamber at 100 nM.

S1P analysis was performed at the University of Colorado School of Medicine Metabolomics Core. S1P was extracted from frozen cell pellets at 2 × 10⁷ cells/ml of 5:3:2 methanol:acetonitrile:water containing 0.5 μM D7 S1P (Avanti Lipids) (11). Extraction and centrifugation were performed as previously described (12), then supernatants were diluted 10-fold with additional extraction buffer and analyzed on a Thermo Fisher Scientific Vanquish UHPLC coupled to a Thermo Fisher Scientific Q Exactive Mass Spectrometer. Chromatographic separation was achieved using a Kinetex C18 column (2.1 × 150 cm, 1.7 μm) and a 4-min gradient with A phase of 0.1% formic acid in water and B phase of 0.1% formic acid in acetonitrile. The column was held at 45°C, with a flow rate of 400 μl/min and gradient as follows: 0–2-min increase from 50% to 95% B, 2–2.5-min decrease to 50% B, hold until 4 min at 50% B. Eluate was introduced to the mass spectrometer using positive electrospray ionization. The mass spectrometry (MS) scanned in full MS mode over the range 100–1500 m/z at a resolution of 70,000. Signals for light and heavy S1P were annotated and integrated using Maven (Princeton University), and absolute levels of S1P were determined.

Freshly harvested colon and ileum were washed with PBS (Life Technologies) and longitudinally bisected before being pinned out flat and fixed in 4% paraformaldehyde for 24 h. Fixed tissues were rolled into Swiss rolls and stored in 70% ethanol. Paraffin embedding, sectioning, and H&E staining were performed by the Research Histology Services at the University of Colorado Anschutz Medical Campus. A minimum of 5 inframe images were taken of each slide, and all inframe anatomical structures were measured on each slide in Photoshop (Adobe). All measurements taken for a given animal were averaged and represent one displayed biological replicate. Ileum damage was scored as described previously (13). Active inflammation (polymorphonuclear cell infiltration and mucosal erosion), chronic inflammation (mononuclear cell infiltrate and crypt distortion), and villus architecture (decrease in height and/or increase in width) were evaluated on a scale of 0.5 to 3. The amount of cross-sectional area that was involved was scored on a scale of 0.5 to 4. Each inflammatory marker (active and chronic inflammation, villus architecture) was multiplied by the cross-sectional score to yield the final score. Colon damage was scored using a well-established system (14). Severity and extent of inflammation, tissue repair, crypt damage, and percentage of tissue involvement were scored on a scale of 0 to 4. Each histologic feature score was multiplied by the percentage involvement score to give a final score. Statistical analysis was performed in Prism version 9 (GraphPad Software). Histological analysis was performed on blinded samples.

Our prior work demonstrated that colon IEL numbers were reduced ∼1000-fold in MyD88^−/− mice compared with MyD88^+/+ mice (7). Therefore, we first aimed to identify the cell type in which MyD88 signaling was critical. We used lineage-specific Cre-expressing mice [Lck-Cre for T cells (15), villin-Cre for intestinal epithelial cells (16), and LysM-Cre for myeloid cells (17)] crossbred to MyD88^fl/fl mice. Due to known differences in IEL recruitment under colonization by distinct microbes (7), all experiments were performed relative to littermate Cre⁻ controls to minimize confounding effects of microbial differences between cages. Colon IEL numbers were evaluated by flow cytometry (Supplemental Fig. 1A). In T cell–specific Lck-Cre⁺ MyD88^fl/fl mice, colon IEL numbers did not differ significantly from Lck-Cre⁻ MyD88^fl/fl littermates (Fig. 1A), suggesting that the MyD88 signaling needed for colon IEL recruitment was T cell extrinsic. In the epithelial cell–specific villin-Cre⁺ MyD88^fl/fl mice, we observed a significant increase in IEL numbers in the colon compared with Cre⁻ MyD88^fl/fl littermates (Fig. 1B), likely owing to the previously described decreased barrier function in villin-Cre⁺ MyD88^fl/fl mice (18). Finally, we observed a significant decrease in TCRαβ⁺ IELs (Fig. 1C) without effect on TCRγδ⁺ IELs (Fig. 1D) or lamina propria T cells (Supplemental Fig. 1B) in LysM-Cre⁺ MyD88^fl/fl mice compared with Cre⁻ MyD88^fl/fl littermates. Within the TCRαβ⁺ fraction in LysM-Cre⁺ MyD88^fl/fl mice, we observed that the reduction was consistent across the CD4⁺ (Fig. 1E), CD8⁺ (Fig. 1F), and DN subsets (Fig. 1G). The specificity of the LysM-Cre⁺ MyD88^fl/fl knockout was confirmed by qPCR on magnetically enriched lamina propria CD11c⁺ monocytes (Supplemental Fig. 1C), epithelial cells, and magnetically enriched splenic T cells (Supplemental Fig. 1D). We confirmed that the LysM-Cre⁺ MyD88^fl/fl knockout did not influence overall numbers or phenotypes of CD11c⁺ MHCII⁺ mononuclear phagocytes (Supplemental Fig. 1E, 1F) or expression of tissue-resident markers in colon IELs (Supplemental Fig. 1G). These results indicate the importance of MyD88 expression specifically in myeloid cells for the recruitment of multiple subsets of TCRαβ⁺ colon IELs, but not TCRγδ⁺ colon IELs.

We previously identified trafficking of colon IELs to the joint under homeostatic conditions using photoconvertible transgenic KikGR mice in which exposure of the distal colon with 405-nm light induced a green-to-red change in fluorescence emission (19, 20). Using the same KikGR transgenic mice, colon IELs can be found in multiple peripheral organs, including spleen, liver, and lungs, for up to 2 wk following colonoscopy-induced photoconversion (Supplemental Fig. 2A), suggesting constitutive circulation and recruitment of IELs. We then treated colonoscopy-photoconverted KikGR mice with broad-spectrum antibiotics for 7 d to deplete the microbiome and reduce colon IELs as previously described (7). Recolonization of the same mice with bacteria by exposure to dirty bedding resulted in the return of the previously photoconverted IELs to the colon epithelium (Supplemental Fig. 2B). Interestingly, antibiotic treatment and recolonization did not alter TCR Vβ usage (Supplemental Fig. 2C), indicating that IELs were actively recruited to the epithelium by Ag-independent bacterial signals, likely through MyD88 sensing in myeloid cells, based on the results in our conditionally deficient mice.

To determine the specific mechanisms by which myeloid cells might recruit colon IELs, we performed a microarray comparing gene expression between colon IELs at baseline and following recolonization, reasoning that the normalization of IELs following bacterial recolonization of antibiotic-treated mice (7) was a phase of active IEL recruitment. Examination of cell surface receptors known to lead to cellular migration did not reveal noteworthy fold changes between the two groups (Supplemental Fig. 3A, 3B), possibly because homeostatic IEL trafficking (Supplemental Fig. 2A) may obscure the signal in the baseline IEL group. Therefore, we compared IELs from recolonized mice to lamina propria T cells, isolated by negative selection magnetic sorting. Several candidate cell surface receptors and cell migration factors were greatly impacted (Fig. 2A and Supplemental Fig. 3B), although most could not be validated by qPCR performed on a larger cohort (Supplemental Fig. 3C). Confirmatory qPCR did identify S1P receptor 1 (S1pr1) as upregulated on colon IELs compared with lamina propria T cells (Fig. 2B), suggesting that S1PR1 was a factor unique to IELs that supported recruitment to the epithelium. S1pr1 expression was unchanged on IELs from LysM-Cre⁺ MyD88^fl/fl mice compared with Cre⁻ littermate controls (Supplemental Fig. 3D). We then queried if the ligand for S1PR1, S1P, fluctuated with colonization status. By liquid chromatography MS, we observed that S1P was present in significantly higher levels in the colons of conventionally housed mice compared with those treated with broad-spectrum antibiotics or germ-free mice (Fig. 2C). Together, these data suggested that S1P may be a factor produced in the colon in response to microbial colonization that mediates T cell localization to the epithelium.

Given the specific importance of MyD88 expression by myeloid cells for colon TCRαβ⁺ IEL recruitment (Fig. 1A) and indications that tissue S1P levels might be an important factor for that recruitment (Fig. 2), we sought to show that myeloid cells could secrete S1P that would cause T cell migration in response to bacteria. To do this, we generated BMDMs (21) through culture with GM-CSF for 6 d, then stimulated the cells with 5 ng/ml cecal contents for 24 h. Due to the technical challenges of detecting S1P by liquid chromatography MS, cellular expression of sphingosine kinase 1 [Sphk1, the enzyme primarily responsible for exported S1P (22, 23)] was assayed by qPCR as a proxy for S1P production. We observed a significant increase in Sphk1 expression beginning 8 h following stimulation and decreasing to below baseline by 24 h (Fig. 3A). We next tested if this increase in Sphk1 expression was dependent on MyD88 using BMDMs from LysM-Cre⁺ MyD88^fl/fl mice in comparison with their Cre⁻ littermate controls. The increase in Sphk1 expression upon stimulation with cecal contents was lost in LysM-Cre⁺ MyD88^fl/fl BMDMs (Fig. 3B).

Because S1P may be secreted by transporters other than the classically described SPNS2 (sphingolipid transporter 2) (23–25), and because the expression level of an enzyme may not directly correlate with the presence of the downstream product, we decided instead to directly test if stimulation of myeloid cells with bacterial products was sufficient to induce T cell migration in a Transwell assay. We generated BMDMs as before and stimulated them with cecal contents for 8 h before washing the cells and transferring them to culture in serum-free media. We then isolated splenic T cells by negative selection magnetic enrichment and plated them in the top well of a 5-μm Transwell system. Because colon IELs are present in the splenocyte population (Supplemental Fig. 2A), we reasoned that a portion of migrating T cells may be biologically relevant IELs. Significantly more T cells migrated into the lower chamber of cecal content–stimulated BMDMs than unstimulated counterparts (Fig. 3C), although in both cases we observed a substantial amount of migration. This was likely due to the baseline activation status of the GM-CSF–derived cells, which produce substantial amounts of various chemokines (26). Finally, to determine if the observed increase in T cell migration was due to MyD88-dependent production of S1P, we repeated this experiment using BMDMs derived from LysM-Cre⁺ MyD88^fl/fl mice versus Cre⁻ littermate controls. As an additional control to demonstrate the importance of S1P receptor signaling, T cells were pretreated with the pan-S1PR inhibitor FTY720 (27). We observed similar reductions in T cell recovery in the lower wells of both the Cre⁺ BMDMS and FTY720-treated BMDMs (Fig. 3D), suggesting that MyD88-mediated upregulation of Sphk1 is important for the induction of T cell migration in this system.

Because BMDMs are not fully representative of mononuclear phagocytes in the gut, we performed in vivo studies to examine the intestinal monocyte response. Therefore, we tested if myeloid cells in the colon upregulated Sphk1 in vivo in response to bacterial colonization of germ-free mice, which have significantly reduced IELs compared with conventionally housed mice (7). Colonization of germ-free mice by oral gavage of cecal contents from conventionally housed mice resulted in an influx of IELs 6 d following gavage (Fig. 3E). Prior to this influx, 3 d after gavage, we observed an increase in Sphk1 expression in CD11c⁺ lamina propria monocytes (Fig. 3F). We did not observe significant increases in Sphk1 expression within epithelial cells (Supplemental Fig. 3E), suggesting that myeloid cells may be the primary producers of S1P in the colon. Overall, these findings support our in vitro results that myeloid cells produce S1P in response to stimulation with bacterial products and suggest that the increase in S1P precipitates IEL recruitment.

To directly determine the importance of the S1P-S1PR signaling axis in colon TCRαβ⁺ IEL recruitment, we used FTY720 in vivo at 1 mg/kg, a dose that has been shown to alter T cell trafficking and proliferation (28) and reduce T cell tissue infiltration in experimental autoimmune encephalitis (29). To assay IEL recruitment versus retention and survival, mice were treated with broad-spectrum antibiotics and recolonized as described previously. Following antibiotic treatment, they were randomly assigned to daily i.p. injections of 1 mg/kg FTY720 or vehicle (DMSO) control for 7 d, followed by euthanasia and IEL harvest. FTY720 treatment, compared with vehicle, resulted in a significant reduction of CD4⁺ and DN, but not CD8⁺, TCRαβ⁺ IELs (Fig. 4A–4D), similar to our observations in LysM-Cre⁺ mice. We observed no difference in TCRγδ⁺ IELs between FTY720 and control treatments (Fig. 4E), indicating that localization of TCRγδ⁺ IELs to the colon epithelium is not dependent upon S1P signaling as it is for TCRαβ IELs.

The direct impact of colon IELs in inflammatory bowel disease (IBD) is at present unclear. Prior work in human participants demonstrates that in Crohn’s disease, but not ulcerative colitis, colon IELs have increased capacity for the production of IL-17, IFN-γ, and TNF, suggesting that IELs may play different roles during distinct disease pathologies (9). Because our data in this study demonstrate that myeloid cell deficiency of MyD88 reduces colon IELs, we crossed LysM-Cre⁺ MyD88^fl/fl mice to TNF^ΔARE/+ mice to probe the role of colon IELs in a murine model of IBD. TNF^ΔARE/+ transgenic mice contain a 69-bp polymorphism in the 3′-untranslated region of the TNF gene that stabilizes the mRNA transcript (10). This stabilization results in greater TNF levels and ultimately a state of microbiome-dependent colitis and ileitis (30).

We confirmed that LysM-Cre⁺ MyD88^fl/fl TNF^ΔARE/+ mice had significantly reduced TCRαβ⁺ colon IELs compared with Cre⁻ TNF^ΔARE/+ littermates and that TNF production by CD11c⁺ cells was unchanged by loss of MyD88 (Supplemental Fig. 4A, 4B). Compared with LysM-Cre⁻ MyD88^fl/fl TNF^ΔARE/+ littermate controls, Cre⁺ TNF^ΔARE/+ mice displayed significantly reduced ileitis and colitis, as measured by pathologic features of active and chronic inflammation and villous architecture in the ileum (Fig. 5A–5F) and crypt damage in the colon (Fig. 5G–5L). These data indicate that loss of MyD88 signaling in the myeloid compartment ameliorates the disease pathology in the TNF^ΔARE/+ model. This result is in accordance with previously published results using the same LysM-Cre conditional MyD88 knockout in the IL-10^−/− model of colitis (31).

To test whether our findings were more likely to be due to the absence of colon IELs or potential changes in myeloid cell function due to the loss of MyD88 signaling, we transferred magnetically sorted colon IELs (Supplemental Fig. 4C) from 16 wk-old TNF^ΔARE/+ and littermate TNF^+/+ donors into Rag1^−/− hosts (Fig. 6). Homeostatic proliferation resulted in circulating T cell levels similar to unmanipulated age- and sex-matched C57BL/6 mice by 3 wk after transfer (Supplemental Fig. 4D). Following an additional 5-wk period to allow establishment of any disease phenotype, tissues were harvested. We confirmed that, unlike bulk CD4⁺ T cells (32), IELs were not sufficient to drive inflammatory disease following transfer. Overall tissue injury score, as well as numbers of inflammatory aggregates, were not significantly different across the three experimental groups (Fig. 6A–6J). These data indicate that in the setting of TNF-driven inflammation, colon IELs may be pathogenic, in accordance with reported results in the IL-10–deficient model (31). However, colon IELs by themselves cannot trigger intestinal inflammation.

IELs are important members of the immunological community of the colon. They contribute to mucosal defense in the setting of infection (33) and barrier health through the production of key mediators such as IL-6 (8). However, the mechanisms of their recruitment to the epithelium in response to changing conditions and the details of their activation remain poorly defined. In this study, we present evidence that the recruitment of colon TCRαβ⁺ IELs to the colon epithelium relies on S1PR1 signaling from myeloid cell–produced S1P following MyD88-mediated detection of bacteria. In these experiments, we did not observe changes in the TCRγδ⁺ IEL population, suggesting that TCRγδ⁺ IELs in the colon may be true tissue-resident T cells or may be recruited and maintained by a separate mechanism from the TCRαβ⁺ fraction.

MyD88 has dual roles as a critical transducer of signals from the extracellular TLRs (34) and IL-1β (35), suggesting that these receptors may act as mediators of the interactions between myeloid cells and luminal microbes in the colon. Our in vitro results demonstrating Sphk1 upregulation on BMDMs following stimulation with cecal contents suggest that the presence of exogenous IL-1β is not strictly required for the activation of myeloid cells and subsequent induction of T cell migration. However, we acknowledge that the use of BMDMs in these experiments presents a caveat in the interpretation of the data due to differences between BMDM and primary intestinal macrophage phenotype and function. Nevertheless, MyD88-dependent signaling through the TLRs may be assumed to be the primary driver of IEL recruitment in the colon, with IL-1β potentially acting as an amplifier of such signals or having no direct role. Signaling through most of the extracellular TLRs is strictly dependent on MyD88 (36), whereas the endosomal TLRs rely on TRIF (36). TLR4, the receptor for LPS and other ligands (37), may signal through either TRIF or MyD88 (38). Because we did not observe a complete loss of colon IELs in LysM-Cre⁺ MyD88^fl/fl mice, TRIF-mediated TLR4 signaling may be responsible for recruiting the remaining IELs we observed. Interestingly, prior work has shown that the survival of TCRγδ⁺ as well as the CD8⁺ subset of TCRαβ⁺ IELs in both the colon and SI requires RIG-I–dependent interactions between the virome and myeloid cells, which express IL-15 in response to RIG-I–dependent detection of commensal viral components (39). Together, these results exemplify the multistep regulation of mucosal immune responses following detection of putative danger signals, because distinct signals are responsible for the recruitment and retention of specific cellular populations.

S1P is an important bioactive phospholipid involved in numerous physiological functions, such as vascular development and cell survival (40). Immunologically, S1P functions as a critical regulator of T cell migration (41). S1P gradients are important for both lymphocyte trafficking into the circulation from secondary lymphoid organs (41, 42) and extravasation events into inflamed tissues where local S1P concentrations are high (43). Because S1P export has been shown to take place through ubiquitously expressed ABC family transporters (44), we chose to focus on cytosolic S1P production using expression of the enzyme Sphk1 as a proxy and migration assays to confirm export. Although an ideal system to interrogate S1P production in the colon would involve genetic deficiency of both Sphk1 and Sphk2, the importance of S1P in regulating both phagocytosis (45) and TLR signaling (46) would convolute the interpretation of results gained from such a model. Therefore, we chose to use FTY720, a systemic inhibitor of the S1P signaling axis, to confirm in vivo that this axis was critical for the recruitment of TCRαβ⁺ IELs. Our findings that tissue S1P levels correlate with colonization status and that myeloid cells produce S1P in response to MyD88-dependent stimulation with bacterial products suggest, but do not conclusively show, that S1P is involved in drawing cells to sites of increased bacterial presence. Regardless, these data demonstrate that TCRαβ⁺ colon IELs are indeed recruited from circulation rather than proliferating in situ from a founder population.

Host factors may also lead to barrier dysfunction and downstream recruitment and activation of colon IELs. Colon biopsies taken from patients with Crohn’s disease and ulcerative colitis showed increased Sphk1 expression in bulk tissue relative to healthy controls, although these data were less consistent in the tested murine colitis models (47). Targeting of the S1P-S1PR signaling axis IBD has been under active investigation, leading to FDA approval of the S1PR antagonist ozanimod for the treatment of ulcerative colitis (48). Therapeutic modalities targeting S1PR also show promise in the treatment of Crohn’s disease (49). Our data suggest that one effect of S1PR blockade is significant reduction in TCRαβ⁺ IELs in the colon.

Currently, the role of colon IELs in human colitis remains unclear, although there is evidence that they may have proinflammatory function in Crohn’s disease (9). Our finding of ameliorated colitis with impaired IEL recruitment in the TNF^ΔARE/+ mouse model is in accordance with prior findings in the IL-10 deficiency colitis model using the same conditional deletion of MyD88 in myeloid cells (31). The IL-10–knockout model has no direct effect on the epithelial cells themselves and is instead driven primarily by IFN-γ produced by CD4⁺ TH1-polarized T cells (50). In that model, conditional MyD88 knockout under both the LysM and CD11c promoters resulted in reversion to healthy levels of IL-4, IL-17, and IFN-γ. This result suggests that the presumptive loss of TCRαβ⁺ IEL recruitment to the colon in those animals was sufficient to ameliorate disease through the reduction of a critical source of proinflammatory mediators.

Our data raise an intriguing concept regarding colon IELs: unlike those of the SI (2, 3, 6, 51–55), colon IELs, at least in the TCRαβ⁺ subset, do not appear to be tissue-resident cells. In support of this conclusion, we observe a low fraction of colon IELs that express the tissue integrins α₄β₇ and CD103 (Supplemental Fig. 1G). We also demonstrate homeostatic circulation of colon IELs into extraintestinal tissues (Supplemental Fig. 2) that is consistent with our recent publication demonstrating gut-joint trafficking of colon IELs (20). Using transgenic mice expressing the KikGR photoconvertible protein, we demonstrated that colon IELs traffic constitutively to various tissues, including the skin, lung, and joints, and that they return to the colon epithelium. Although we did not identify broad changes in TCR repertoire as a result of IEL trafficking, our analysis was superficial and will require CDR sequencing to better address in the future.

Gut-derived T cells maintain cytokine competence in the joint, producing IL-17A and TNF-α. Transfer of colon IELs from TNF^ΔARE/+ mice into Rag1^−/− mice significantly enhances joint destruction initiated by a hock injection with CFA. In this study, we do not observe bowel inflammation following transfer of colon IELs into Rag1^−/− mice (Fig. 6). However, as in the arthritis model, this may require a local inflammatory stimulus, such as in the TNF^ΔARE/+ environment, where they enhance disease (Fig. 5) that the IELs are unable to regulate. Regardless, many questions remain, including the following: (1) What are the signals informing colon IEL function in the epithelium? (2) What are the signals that mitigate IEL efflux from the colon and into other tissues? (3) Does circulation of colon IELs change their function, and if so, how? (4) How does TCR specificity of colon IELs influence their trafficking patterns and cytokine competency in other tissues?

In totality, this work demonstrates the requirement for both bacterial presence and an intact S1P signaling axis for the recruitment of TCRαβ⁺ colon IELs. These results, coupled with those of others, showcase both the potential power and pitfalls of therapeutic interventions in IBD that prevent IEL recruitment and indicate the need for a more detailed understanding of the determinants of IEL trafficking at the level of functionally defined subsets.

The authors have no financial conflicts of interest.