Intestinal CD4 - CD8αβ - TCRαβ + T cells function as tolerogenic antigen presenting cells

The intestinal immune system is constantly exposed to a plethora of antigens (Ags) from innocuous ingested material and the commensal flora that must be distinguished from pathogen-derived antigens. To this end, a number of anatomic, cellular and molecular mechanisms operate in the intestinal tract to acquire, process, present and interpret Ags from the intestinal lumen. The intestinal mucosa is also populated by a large number of T cells that reside within the epithelial lining (intraepithelial lymphocytes, IELs), in the underlying lamina propria (LPLs) and in gut-associated lymphoid tissues (GALT). Both IELs and LPLs are heterogeneous populations consisting of conventional CD4 + and CD8 + T cells and numerous unconventional T cells, including

Intestinal γδ and DNT cells differ from conventional CD4 + and CD8 + T cells in many aspects. Most, but not all, express the CD8αα homodimer and their TCRs are thought to recognize self-Ags. Phenotypically, these cells often express various NK receptors, but they do not express many common markers of conventional T cells 11 . Unlike conventional intestinal CD4 + and CD8 + T cells, which are absent at birth and increase with age, intestinal γδ and DNT cells are abundant at birth and decrease with age. These cells are thought to exert antiinflammatory effects rather than promoting intestinal immunity and inflammation, but the mechanisms by which they exert their functions remain to be clarified [11][12] .
Here, we hypothesized that intestinal γδ and/or DNT cells may function as tolerogenic Ag presenting cells (APCs). This idea seemed particularly attractive for γδ T cells, which are abundant in the intraepithelial compartment from where they can dynamically move to the lamina propria 14 . Indeed, human peripheral blood γδ T cells have been shown to function as APCs 15, 16 , and murine γδ T cells express modest amounts of MHC-II mRNA, although their role in Ag presentation has been unclear 17 .
To systematically address whether and which murine intestinal T cell subsets may have the potential to function as APCs, we measured MHC-II expression.
The frequency and composition of intestinal MHC-II + cells among both the CD3non-T and CD3 + T cell subset varied with the animals' age ( Fig. 1b,c, Sup. Fig.   2). In juvenile mice (6 weeks old), the abundance of MHC-II + T cells in IELs and LPLs was low (<5%) regardless of subset, but by 10 weeks this frequency had increased significantly, especially for DNT cells among which 9.5 ± 1.9% and 17.5 ± 3.2% expressed MHC-II among IELs and LPLs, respectively. The positive correlation between age and frequency of MHC-II + DNT cells among SI LPLs continued well beyond 10 weeks and was statistically highly significant (Fig. 1c).
Similar frequencies of MHC-II + γδ and DNT cells were also seen in colon LPLs (Sup. Fig. 3). MHC-II was even more prominently expressed on DNT cells in PPs where as many as 38.7% and 56.4% were positive by 6 and 10 weeks, respectively ( Fig. 1d, Sup. Fig. 2). PPs also harbored MHC-II + cells among the γδT > CD4 > CD8 subsets, but at markedly lower frequencies than among DNT cells.
As expected, most γδ and DNT cells in the SI expressed CD8αα (i.e. they were CD8α + CD8β -) while ~half in PPs and less than 20% in MLNs and spleen had this phenotype (Sup. Fig. 4a). Interestingly, among DNT cells in PPs, MHC-II was preferentially expressed in the CD8αsubset (Sup. Fig. 4b). Moreover, intestinal T cells (esp. DNT cells) in MHC-II-eGFP reporter mice expressed GFP with similar frequencies and tissue distribution as observed by immuno-fluorescence (Sup. Fig. 5), indicating that the apparent MHC-II signal was not due to nonspecific staining or cross-reactivity of our monoclonal antibody.
In light of the fact that DNT cells contained the highest frequency of MHC-II + cells among all intestinal T cells, especially in PPs, we focused our further analysis on DNT cells whose function is still largely enigmatic. DNT cells display an unusual pattern of MHC restriction that is non-overlapping with that of conventional CD4 + or CD8αβ + T cells 18 . Moreover, DNT cells are thought to play a protective role in the pathogenesis of colitis, but the mechanism remains to be clarified 19 20 . Indeed, when compared to CD11c + DCs, most MHC-II + PP DNT cells expressed little or no costimulatory molecules, such as CD40, CD80 and CD86, suggesting that DNT-mediated Ag presentation may result in tolerogenic rather than effector responses by T cells (Fig. 1e and Sup. Fig. 6).
To further explore the biology and function of DNT cells and to compare them to other intestinal T cells, we performed a transcriptome analysis of each T cell subset in PPs. Consistent with our flow cytometry results, DN and γδ T cells, but not CD4 or CD8 T cells, expressed high mRNA levels for MHC-II and related molecules (Fig. 1f). DN and γδ T cells in PPs also expressed many NK markers and granzymes A and B (Sup. Fig. 7), which is consistent with previous reports of expression profiles of DN and γδ T cells in IEL 11, 12 . DNT cells were transcriptionally more similar to γδ T cells than to CD4 or CD8 T cells, but DNT cells were distinct from γδ T cells by expressing higher levels of several genes, including Klra15,Pabpc4,Myl1,Klra6,Klra5,Reg3b and trf (Sup. Fig . 8).
Principle component (PCA) analysis and Volcano Plots also show the similarity of DNT cells and γδ T cells and difference from CD4 or CD8 T cells, which is consistent with the previous report about natural IELs (γδ and DN) and induce IELs (CD4 and CD8) 11, 12 (Sup. Fig. 9, 10).
At the protein level, most MHC-II + DNT cells in the intestine and, in particular, in PPs (but not in MLN or spleen) expressed high levels of EpCAM, an adhesion molecule that is also expressed on intestinal epithelial cells (IECs) and mediates interactions between IECs and IELs 21 (Sup. Fig. 11a-c). By contrast, conventional T cells in PPs expressed little or no EpCAM. γδ and DNT cells were also distinct from intestinal CD4 and CD8 T cells on a functional level. They produced little IFN-γ or IL-17A upon activation unlike most conventional T cells and mucosa-associated invariant T (MAIT) cells 22 , which function primarily by orchestrating cytokine dependent effector responses to infectious pathogens (Sup. Fig. 12).
In light of these observations, we asked whether DNT cells possess the capacity to acquire and process exogenous Ags. To this end, T cell subsets in SI and, as a positive control, CD11c + myeloid cells in MLN were purified and incubated with DQ-Ovalbumin (DQ-OVA), a fluorescently tagged protein that becomes brightly fluorescent upon endocytosis and intracellular proteolysis 10 . Remarkably, DNT cells acquired and processed a large amount of DQ-OVA, exceeding even CD11c + cells (Fig. 1g). γδ T cells, but not CD4 or CD8 T cells, also showed a modest uptake. On the other hand, when isolated cells were incubated with labeled bacteria only CD11c + cells acquired fluorescent material, indicating that DNTs acquire Ags via macropinocytosis or other endocytosis pathways, rather than phagocytosis (Sup. Fig. 13).
Next, we asked whether DNT cells can also acquire soluble macromolecules in vivo. For this, Dextran-Alexa Fluor488 (Dex-AF488) was injected into the ligated small intestine and cellular dextran uptake was assessed 5h later. Indeed, DNT cells, but no other T cell subset, in IELs, LPLs and PPs accumulated Dex-AF488 ( Fig. 2a). Negligible numbers of Dex-AF488 + cells were found in MLNs and spleen at this time point (data not shown), indicating that the cells that had acquired fluorescent material in the gut lumen remained, at least initially, confined to the intestine.
To test more rigorously whether DNT cells traffic after Ag uptake from the gut to MLNs, we used Kaede transgenic mice, which express a photoconvertible fluorescent protein 23 . The serosal side of the small intestine of Kaede mice was illuminated with a blue laser and 48h later, photo-converted T cells in each intestinal compartment were enumerated by flow cytometry. There were comparable numbers of photo-converted γδ, DNT cells and TCRleukocytes in MLNs, while almost no photo-converted CD4 or CD8 T cells were detected (Fig.   2b,c).
In aggregate, the above results indicate that many intestinal DNT cells express MHC-II and acquire soluble material from the intestinal lumen, which they can then carry to MLN. Thus, we asked whether DNT cells can present exogenous Ags to CD4 T cells. To this end, T cell subsets were purified from PPs and cocultured with OVA and CFSE-labeled naive OT-II-CD4 T cells. On day 5, 6 and 8, the recovered number and CFSE fluorescence of OT-II cells were analyzed.
Indeed, DNT cells and, to some degree, also γδ T cells, but not CD4 or CD8 T cells, induced OT-II cell proliferation (Fig. 2d). However, OT-II cells that were exposed to Ag presenting DNT cells produced only modest amounts of cytokine and had undergone fewer divisions and expressed fewer phenotypic markers of activated effector cells when compared to OT-II cells that had been co-cultured with OVA pulsed CD11c + DCs ( Fig. 2d-g and Sup. Figs. 14, 15). These results are consistent with the idea that incomplete Ag presentation (i.e. TCR stimulation with little or no co-stimulation) by γδ and DNT cells may have induced an anergic state in responsive T cells. Indeed, the number of viable OTII-CD4 T cells cocultured with γδ or DNT cells was dramatically reduced on day 6 and 8 ( Fig. 2ef). Similar to PPs, γδ and DNT cells harvested from MLNs also had the capacity for Ag presentation, whereas IELs and LPLs were inactive (Sup. Fig. 16a-c), presumably due to the low expression of MHC-II in this compartment (Fig. 1d).
Next we sought to clarify the physiological importance of DNT cells mediated uptake of intestinal Ag in vivo. Thus, we performed and ex vivo Ag presentation experiments 6, 24 by feeding mice with a diet containing 1% OVA for 3 days and then coculturing sorted γδ, DN, CD4, CD8 and TCRβ -TCRγδcells from PPs and CD11c + cells from MLN with CFSE-labeled naïve OT-II CD4 T cells. In these experiments, the percentage and number of divided OT-II cells in co-cultures with DNT cells was higher than in any other coculture setting ( Fig. 2h-j,).
Having determined that certain unconventional intestinal T cells, most prominently the DNT cells subset, can acquire antigenic material from the gut lumen and then process and present this material to cognate naive CD4 T cells, we set out to assess the patho-physiological significance of these observations.
Next, we set out to compare the ability of DNT cells from conditional mutant donors and WT littermates to elicit a tolerogenic response in naive T cells . To this end, naive OT-II CD4 T cells co-cultured with purified γδ, DN, CD4 or CD8 T cells from PP in the presence of antigenic peptide (OVA323-339). Two hours later, MLN CD11c + cells were added to the cultures. In the presence of fully MHC-II sufficient DNT cells significantly fewer OT-II cells were recovered than in cocultures with any other intestinal cell subset, whereas conditional deletion of MHC-II in T cells abrogated this apparent tolerogenic effect of DNT cells (Fig. 2k and Sup. Fig. 18a-c).
Next, we set out to directly observe intestinal T cells in situ using multi-photon intravital microscopy (MP-IVM) in anesthetized mice. As shown in Fig. 3a and Sup. Movie 1, T cells in DPE-GFP mice, in which T cells express GFP 26 , numerous GFP + cells moved dynamically between the LP and the intra-epithelial region. GFP + T cells in Tcrd-GFP mice showed a similar behavior ( Fig. 3b and Sup. Movie 2), consistent with a previous report 14 . Since some non-T cells, such as plasmacytoid dendritic cells (pDC) and some macrophages, are known to express GFP in DPE-GFP mice 27, 28 , we also performed MP-IVM in the SI of DPE-GFPxRAG-1 -/mice. As shown in Fig. 3c and Sup. Movie 3, the motility of the sparse GFP + non-T cells in these mice was low, and they rarely localized to the intra-epithelial region. Although some GFP + cells in CX3CR1-GFPmice were in contact with IEC, most of them located in LP lesion ( Fig. 3d and Sup. Movie 4). The average speed of GFP + cells in DPE-GFP and Tcrd-GFP mice was higher than those of DPE-GFPxRAG-1 -/or CX3CR1-GFP, but there was no difference in the track length between DPE-GFP/Tcrd-GFP and CX3CR-1GFP mice ( Fig.   3e, f). The number of GFP ＋ cells that were positioned in the intra-epithelial compartment and/or were exposed to the intestinal lumen was higher in DPE-GFP and Tcrd-GFP mice than in DPE-GFP x RAG -/or CX3CR1-GFP mice (Fig.   3g, h). Some GFP + cells in DPE mice were observed to extend protrusions across the epithelial layer ( Fig. 3i and Sup. Movie 5), a phenomenon that was not observed in Tcrd-GFP or DPE-GFPxRAG1 -/mice. Transepithelial protrusions of GFP + cells, presumably reflecting macrophages, were also occasionally detected in CX3CR1-GFP mice, consistent with previous studies 4, 5 , but the frequency of these occurrences was much lower than in DPE-GFP mice. When DPE-GFP mice were injected with OVA-AF594 into the lumen of the small intestine, GFP + (presumably DNT) cells could be seen ~3h later that had acquired the fluorescent Ag in the lumen and then moved it across the epithelial barrier ( Fig. 3j and Sup.

Movie 6).
To more rigorously compare each T cell subset in vivo, we resorted to an adoptive transfer strategy. Purified γδ, DNT, CD4 or CD8-IELs from EGFP-Tg mice were separately transferred into individual NOD.scid.IL-2Rγ -/-(NSG) mice. Seven to eight weeks later, intravital imaging was performed in the small intestine. We confirmed more than 95% purity before the transfer. We also performed FACS analysis after the intravital imaging and confirmed that each subset did not change even 7-8 weeks after the transfer (Sup . Fig 19a-

b).
These experiments revealed distinct differences in the distribution and motility of each transferred T cell subset (Fig. 3k-n and Sup. Movie7-10). For example, CD4 T cells migrated somewhat more slowly and with shorter track length than other T cells (Fig. 3o, p). Some cells in the LP lesion of CD4→NSG mice did not move ( Fig. 3m and Sup. Movie9). The track length of DNT cells was the shortest, even (Fig. 3o, p), which suggests that the movement of DNT cells was restricted.
Of note, GFP + cells in DN→NSG mice were frequently observed to attached to IECs and lumen (Fig. 3l, q, r and Sup. Movie8). As shown in Supplemental movie 11, multiple GFP + cells in recipients of DNT cells meandered between the LP and IE compartment and extended dendrites toward the intestinal lumen, a behavior that was not seen in DPE-GFP mice.
Having characterized intestinal MHC-II + T cells, particularly the DNT cells subset, at steady-state, we set out to assess their role in intestinal inflammation using CD4-Cre + x H2-Ab1 flox/flox mice. In our animal colony, CD4-Cre + x H2-Ab1 flox/flox mice did not develop spontaneous colitis, however, upon challenge with 3% DSS for 5 days the mutant mice developed much more severe colitis than WT littermates, as evidenced by accelerated loss of body weight, shortening of the colon and clinical score ( Fig. 4a-d). Accordingly, histologic evaluation of colons revealed a much more pronounced trans-mural inflammation in CD4-Cre + x H2-Ab1 flox/flox mice ( Fig. 4e-f).
In aggregate, the above experiments in murine models indicate that many unconventional intestinal T cells, most notably the DNT cells subset, express MHC-II and have the capacity to acquire and present soluble antigenic material both in vitro and in vivo. Intravital imaging experiments suggest that DNT cells are highly motile and rapidly shuttle between the LP and the IE compartment from where they may directly access Ag in the intestinal lumen by extending transepithelial processes. However, DNT cells are distinct from the majority of T cell subsets among IEL, which are a tissue resident population 11-13 , whereas DNT cells can exit the gut and migrate to MLN. In addition, a sizeable population of MHC-II + DNT cells also resides in PPs. When isolated from PPs, DNT cells as well as γδ T cells could present Ag to naive CD4 T cells, which resulted in anergy rather than effector activity, presumably because DNT cells do not express costimulatory molecules. Indeed, the presence of DNT cells suppressed in an MHC-II dependent manner the proliferation of CD4 T cells that were simultaneously exposed to Ag presenting CD11c + cells. Physiological relevance of these findings is implied by the observation that the conditional deletion of MHC-II in T cells markedly exacerbated DSS colitis, which is consistent with the idea that DNT cells function as tolerogenic APCs.
Finally, we asked whether these findings have a correlate in humans, particularly in the context of inflammatory bowel disease (IBD). To this end, biopsy samples of SI mucosa from patients with either Crohn's disease (CD) or other, noninflammatory bowel diseases (non-IBD) were collected after obtaining informed consent (Sup. Table 1-2) and mononuclear cells were analyzed by flow cytometry.
Although the frequency of γδ and DNT cells was generally lower in humans than in mice, both subsets were readily detectable in both SI-IEL and LPL (Fig. 4g).
In non-IBD samples, most (~55%-75%) human T cells expressed MHC-II regardless of subset, which is in contrast to intestinal T cells in mice where MHC-II was only found on γδ and DNT cells (Fig. 4h). However, similar to our findings in mice, only human DN and γδ T cells, but not CD4 or CD8 T cells in SI-LP took up OVA-DQ (Figure 4i).
Interestingly, in intestines of CD patients the expression of HLA-DR and antigen uptake ability of DN and γδ T cells were significantly reduced as compared to non-IBD patients (Figure 4h, i). There was no statistically significant difference between the ages of patients CD (38.5±4.78 years) and non-IBD patients (52.4±6.59 years), however, our CD patients cohort included more male (8) and fewer female (3) patients than our non-IBD control group (6 male and 5 female).
However, as shown in supplemental figure 20, there was no difference between male and female control patients in the expression of HLA-DR on LP DNTs.
Interestingly, the expression of HLA-DR on DNTs in active CD patients was higher than in patients in remission (Sup. Fig. 21), suggesting that the low expression of HLA-DR on DNTs in CD patients is not due to inflammation.
γδ and DNT cells are uniquely abundant in the intestine, but rare in other organs.
Although they differ from conventional CD4 or CD8 T cells in many aspects, their function is not completely clear. Results of in vivo experiment using knockout mice (Tcrd -/-) or reporter mice (Tcrd-eGFP) suggest that γδ T cells combine features of adaptive and innate immune cells [11][12][13]17 . However, much less information is available on DNT cells, which have remained largely enigmatic. We now show that both human and murine intestinal DNT cells express MHC-II and have the ability to uptake, process and present soluble Ags resulting in an anergic response by naïve CD4 T cells.
In some previous papers, this subset was sometimes called as CD8αα + T cells 29, 30 , because most of them express CD8αα homodimer. In the current study, we defined CD3 + CD4 -CD8β -TCRαβ + T cells as DNT cells, because considerable part of them outside the intestine doesn't express CD8αα (Sup. Fig. 4). In addition, some papers suggest that CD8αα can be transiently expressed 31, 32 , so it is possible that the CD8αα cells are in constant exchange with the CD8αα-neg subset. That's why we selected "DNT cells" in ref 11 and 12 as more stable and reliable definition.
In some papers, CD4 -CD8α -TCRαβ + T cells are defined as DNT cells 33 . This fraction is not same as DNT cells in the current study, because considerable part of DNTs expresses CD8αα.
Our findings shed new light on previous studies on the role of unconventional T cells in IBD. For example, DNT cells were shown to suppress colitis in RAG deficient recipients of adoptively transferred CD4 + CD45RB high T cells, but the mechanism has been unclear 19 . More recently, it was reported that Lck-Cre x TAK1 flox/flox mice are nearly devoid of CD8α + TCRβ + IELs and develop CD4 + T celldependent colitis 20 The idea that DNT cells promote immunologic tolerance by presenting intestinal Ags offers a potential mechanistic explanation for these observations.
Recently, it was reported that ILC3 cells also express MHC-II and induce anergy in CD4 + T cells 10 . However, in our hands, ILC3 are quite rare in the small intestine, whereas DNTs are encountered at higher frequencies (more than 10 times in SI-LP and PP, 6 times in MLN, data not shown). DNT cells were distinct from ILC3 or other innate lymphoid cells that have previously reported to work as APCs, because they express CD3 and TCRαβ, but not IL-17 or RORγt.
In the current study, we proved that DNT-IEL protruded into the small intestinal lumen, efficiently take up luminal antigens and migrate to MLN. In the PP, DNT cells expressed substantial level of MHC class II and low level of CD40 or CD80.
T cell specific deletion of MHC class II leads to exacerbation of DSS colitis, which suggests that this incomplete antigen presentation by MHC class II + DNT cells induces anergy of gut trophic CD4 + T cells in the GALT (Supplemental Figure   22). Intriguingly, intestinal DNT cells in Crohn's disease patients expressed lower levels of HLA-DR and lower antigen uptake ability than control patients. These findings suggest that MHC-II + DNTs play a key role in intestinal immune homeostasis and may contribute to the pathogenesis of inflammatory bowel disease.

METHODS SUMMARY
Mouse strains used in this study are detailed in online version of Materials and Methods. For in vitro antigen uptake and process experiments, 1 x 10 5 of sort-purified γδ, DN, CD4, CD8 IELs or CD11c + MLN cells from wild type (WT) C57BL/6 mice were co-incubated with or without 20μg/ml DQ-Ovalbumin (Thermo Fisher Scientific) in the condition of 37℃ and 5%CO2 for five hours. For in vivo antigen uptake experiments, a small incision was made on the abdomen of anesthetized C57BL/6 mice, and then small intestine was exposed and ligated for 3cm long.  345-348 (1996  Author Information Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of the paper.
Correspondence and requests for materials should be addressed to Y. N.
(ynemoto.gast@tmd.ac.jp).     In vivo uptake of Dextran-AF488. A small incision was made on the abdomen of anesthetized C57BL/6 mice, and then small intestine was exposed and ligated for 5cm long. and 1 x 10 5 of CFSE labeled OTII CD4 + T cells. Two hours later, 1 x 10 4 CD11c + cells were added into each well. Cells were incubated in PRMI medium with 10%
In vivo live imaging. Mice were anesthetized and injected intravenously with Hoechst 33342 dye (sigma). A small incision was made on the abdomen and small intestine was exposed. Then mice were fixed on the customized stage for in vivo live imaging and temperature of core and exposed small intestine was maintained at 37℃ with the heating plate. Small intestine was gently opened along the antimesenteric border. 10K Dalton Dextran Alexa Flour 594 (Molecular Probes) was added on the luminal side and coverslip was put on it. Cell behavior was recorded for 10-60 min. In HMS, in vivo imaging was performed using the two-photon microscope with the x20 water-immersion objective of an upright microscope (Prairie Technologies). A MaiTai Ti:sapphire laser (Spectra-Physics) was tuned between 870 nm and 900 nm for multiphoton excitation. In TMDU, in vivo imaging was performed using the FV1200MPE (Olympus) in Dr. Okazawa