GPR18 is required for a normal CD8αα intestinal intraepithelial lymphocyte compartment

GPR18 is highly expressed in lymphoid tissues and is widely expressed in immune cells, being abundant in splenic T and B cells (Fig. 1 A). Public domain gene expression data indicated two- to threefold higher expression of GPR18 in intestinal IELs compared with splenic T cells (http://www.immgen.org; http://refdic.rcai.riken.jp), and this was confirmed in quantitative PCR (QPCR) analysis of sorted cells, with expression being highest on CD8αα γδT IELs and CD8αα αβT IELs, followed by CD8αβ αβT IELs (Fig. 1 A). GPR18 transcript abundance in intestinal CD4 αβT cells, which are rare within IEL preparations (∼5%) and are much more abundant in the lamina propria (∼30%), was similar to the expression level in splenic lymphocytes (Fig. 1 A). To examine the role of GPR18 in lymphocytes, we generated GPR18-deficient mice by gene targeting. GPR18-deficient mice had normal numbers of splenic and LN αβ and γδT cells and B cells, and lymphocyte development in thymus and BM appeared normal (not depicted). Analysis of IELs showed there were similar numbers of total cells in matched adult (2–3 mo old) control and Gpr18^−/− mice (Fig. 1 B). However, when the IEL subsets were examined, Gpr18^−/− mice exhibited a selective deficiency in CD8αα γδT IELs (Fig. 1 C). The properties of the IEL compartment are known to change with age (Laky et al., 1997; Cheroutre et al., 2011), and analysis of a group of aged (∼7 mo old) Gpr18^−/− and littermate control mice showed that there was now a significant reduction in total IEL numbers (Fig. 1 D). The CD8αα γδT IEL compartment continued to be the most affected (approximately fourfold reduced, P < 0.01), but CD8αα and CD8αβ αβT cell numbers were also reduced (approximately twofold, P < 0.05, n = 6; not depicted).

GPR18 has been suggested to promote cell chemotaxis in a Gαi protein–coupled manner to the endocannabinoid N-arachidonoyl-glycine (Kohno et al., 2006; McHugh et al., 2012). However, when WEHI231 B cells transduced with GPR18 were tested for responses to this ligand, no migration was observed, whereas the same cells transduced with several other receptors, including the endocannabinoid receptor CB2, the oxysterol receptor EBI2, the sphingosine-1-phosphate receptor S1PR1, and the chemokine receptor CXCR5, showed robust migratory responses to the appropriate ligands (Fig. 1 E and not depicted). The inability of GPR18 to respond to N-arachidonoyl-glycine was also noted in a study using β-arrestin recruitment to assess GPCR activation (Yin et al., 2009). Although we could not confirm N-arachidonoyl-glycine as a ligand for GPR18, we did observe that transduction of WEHI231 cells with GPR18 led to an approximately fivefold reduction in the CXCR4-mediated chemotactic response to CXCL12 (Fig. 1 E). This type of inhibitory effect is seen after transduction of WEHI231 cells with a range of established Gαi-coupled receptors, including CB2, EBI2, and S1PR1 (Fig. 1 E and not depicted), but does not occur when the cells are transduced with GPCRs that are established not to be Gαi coupled, including the thromboxane A2 receptor (TBXAR2) and S1PR2 (Fig. 1 E and not depicted). Based on these data, we suggest that GPR18 is effective in competing for Gαi binding and that it may therefore be able to couple to this G protein and have a pro-migratory role.

Analysis of duodenum sections for the distribution of γδT IELs showed that although reduced in frequency in the GPR18-deficient tissue, the CD8α γδT double-positive cells were distributed similarly in both types of mice, with most being in an intraepithelial location (Fig. 1 F). The migration dynamics of γδT IELs has recently been studied using real-time intravital imaging with discrepant findings regarding whether the cells are largely immobile (Chennupati et al., 2010), like dendritic epidermal T cells in the skin (Gray et al., 2011; Sumaria et al., 2011), or whether they are motile (Edelblum et al., 2012). To further examine whether IELs are motile and test whether GPR18 deficiency affected their movement dynamics, we developed a procedure to intravitally label CD8α⁺ IELs. Systemic injection of anti–CD8α-PE 5 h before analysis led to uniform labeling of CD8α⁺ cells in the epithelial and lamina propria compartments (Fig. 1 G). Using two-photon microscopy to image CD8α⁺ IELs from the luminal side of duodenal villi, we observed WT cells moving up and down the lateral intercellular junctions of epithelial cells, migrating between epithelial cells via the subepithelial space, and occasionally accessing the lamina propria (Fig. 1 H and Video 1). These behaviors for total CD8α⁺ IELs are similar to the findings of Edelblum et al. (2012) for γδT IELs using TcrdEGFP reporter mice, supporting the conclusion that IELs are motile and suggesting that the major IEL populations (CD8αα γδT and αβT) have similar movement dynamics. Analysis of GPR18-deficient mice using the same intravital CD8α-labeling procedure revealed that, at the level that could be resolved with our current imaging platform, their CD8α⁺ IEL migration dynamics were indistinguishable from those of the controls (Fig. 1 H and Video 2).

To further probe the stage at which GPR18 was influencing the accumulation of CD8αα γδT IELs, and to test whether the requirement was cell intrinsic, we generated mixed BM chimeras. This analysis revealed that even under conditions of competition there was no effect of GPR18 deficiency on CD4 or CD8 T cell numbers in the spleen or LNs (not depicted). However, within the IEL compartment there was a marked (∼10-fold) defect in CD8αα γδT cells and there was now also a reduction in CD8αα αβT cells (∼5-fold) and CD8αβ αβT cells (∼3-fold), whereas CD4 αβT cells were unaffected (Fig. 2 A). A comparison of the ratio of GPR18-deficient and WT γδT IELs along the length of the small intestine revealed a graded effect with the strongest defect being in the duodenum (Fig. 2 B). The basis for the greater impact of GPR18 deficiency under competitive compared with noncompetitive conditions is not known, but might indicate a reduced ability compared with WT cells to access or use a maturation signal or trophic factor that is in limited supply.

Surface phenotyping revealed that the Gpr18^−/− γδT IELs in the mixed BM chimeras were enriched for Thy1^hi cells and were depleted for the major γδTCR typical of intestinal IELs, Vγ7 (Fig. 2 C). The Gpr18^−/− IELs were also depleted for granzyme B–high cells (Fig. 2 C). The high Thy1 and low granzyme B expression is consistent with a less mature effector state (Ma et al., 2009; Cheroutre et al., 2011). CD103 (αE integrin) expression was comparable on the Gpr18^−/− and control cells, whereas CCR9 and CXCR3 surface levels were slightly elevated (Fig. 2 C and not depicted). Contrary to this elevated CCR9 and CXCR3 expression, GPR18-deficient γδT IELs in mixed BM chimeras showed reduced migration to the CCR9 ligand, CCL25, and the CXCR3 ligand, CXCL10 (Fig. 2 D). However, they also had a significantly reduced baseline migration in the absence of chemokine (Fig. 2 D). When the fold change in migration compared with “nil” was determined, there was an increase in the relative CCL25 response of the Gpr18^−/− cells (Fig. 2 D). The alterations in CD8αα γδT cell surface phenotype and migration behavior observed for Gpr18^−/− cells from mixed BM chimeras (Fig. 2, C and D) were not seen for CD8αα γδT cells from Gpr18^−/− mice (Fig. 2 D and not depicted). This may indicate that these phenotypes are secondary consequences of a role played by GPR18 in helping cells to compete for a maturation signal.

CCR9-deficient mice suffer a deficiency in CD8αα γδT IEL numbers (Wurbel et al., 2001, 2007; Uehara et al., 2002) that is more severe than observed in GPR18-deficient mice (Fig. 2 E and Fig. 1 C). Analysis of CCR9 GPR18 double-deficient mice showed a similar magnitude of CD8αα γδT IEL deficiency to that observed in Ccr9^−/− mice (Fig. 2 E). These data suggest the function of CCR9 is dominant over the role of GPR18 in establishing an IEL compartment of normal size. However, in contrast to Gpr18^−/− γδT IELs in mixed BM chimeras (Fig. 2 C), Ccr9^−/− γδT IELs in mixed BM chimeras contained normal Vγ7⁺ cell frequencies (Fig. 2 F). These findings indicate that GPR18 has a function distinct from or in addition to any role it might have in supporting CCR9 function.

Analysis of cytokine production after in vitro activation of cells from mixed BM chimeras revealed similar IFN-γ expression by Gpr18^−/− and internal control cells but elevated TNF production by the Gpr18^−/− cells (Fig. 2 G). The TNF expression level was similar to that observed in Thy1^hi WT cells (not depicted), suggesting that it was secondary to the effect of GPR18 deficiency on cell maturation state in competitive chimeras. To examine the IEL turnover rate, mixed BM chimeras were maintained on drinking water containing BrdU for 1 wk and then examined by flow cytometry. Similar fractions of control and Gpr18^−/− cells were BrdU labeled (Fig. 2 H), suggesting that the reduced Gpr18^−/− cell number was not a consequence of a markedly accelerated loss of IELs. Consistent with this conclusion, the fraction of apoptotic γδT IELs detected by annexin V staining was similar for the mutant and WT cells (Fig. 2 I).

To confirm that the deficiency in IELs was solely caused by a loss of GPR18 and not caused by altered expression of a linked gene or other off-target effects, GPR18 expression was rescued in Gpr18^−/− BM by retroviral transduction. Analysis of spleen cells from mice reconstituted with GPR18-encoding versus control retrovirus–transduced BM showed that reporter-positive cells were present at similar frequencies in both groups, confirming that GPR18 does not influence the development of recirculating T cell populations (not depicted). However, within the intestine, each of the IEL subsets that were deficient in the absence of GPR18 was markedly rescued by GPR18 transduction (Fig. 3 A). This was evident both as a higher frequency of transduced (reporter positive) cells and an increase in their number to within twofold of the normal range (Fig. 3 B and Fig. 1, B and C). Moreover, the rescued CD8αα γδT cells down-regulated Thy1 to a similar extent as control cells (Fig. 3 C).

Vγ7⁺ IELs are generated in the thymus and may undergo a final maturation process in mesenteric LNs (mLNs; Guy-Grand et al., 2013). Production of Vγ7⁺ γδT cells in the thymus of mixed BM chimeras was not affected by GPR18 deficiency (Fig. 4 A and not depicted). The cells were also present in frequencies comparable with those of Gpr18^+/− controls in the mLNs (Fig. 4 A and not depicted). To test whether homing of IELs from blood to intestine might require GPR18, a mixture of total IELs from GPR18-deficient (CD45.1/2) and WT (CD45.1) donors were transferred to γδT IEL-deficient (Ccr9^−/− CD45.2) recipients. Although CD8αα IELs are nonrecirculatory and home only inefficiently back to the small intestine after i.v. transfer (Hayday et al., 2001; Chennupati et al., 2010; Cheroutre et al., 2011), analysis at 1 d showed that Gpr18^−/− cells were present among the transferred cells within the IEL compartment at a similar frequency to their representation in blood (Fig. 4 B). These data disfavor a role for GPR18 in promoting CD8αα IEL recruitment from blood into tissue. Given that GPR18 is thought to most likely respond to a small molecule ligand (Inoue et al., 2012), these findings are not unexpected as the step of lymphocyte recruitment from blood into tissue usually involves a protein (chemokine) ligand that can be displayed on the endothelium (Islam and Luster, 2012). CCR9 and CCL25 are established to play an important role in this recruitment step for CD8 T cells and plasma cells (Kunkel et al., 2003; Pabst et al., 2004; Stenstad et al., 2007; Wurbel et al., 2007), a requirement we could confirm for CD8 T cells that had been activated under CCR9-inducing conditions (Fig. 4 C).

CD8αα IELs are considered to be restricted to the intraepithelial compartment, and consistent with this view, the cells make up only a very minor component of standard lamina propria lymphocyte (LPL) preparations (Fig. 5 A; Cheroutre et al., 2011), and it has so far not been possible to determine their distribution in the lamina propria by immunofluorescence analysis. However, because our intravital imaging and that of Edelblum et al. (2012) suggested that CD8α⁺ IELs occasionally access the lamina propria (Videos 1 and 2), we asked whether GPR18 differentially affected the frequency of CD8αα γδT cells in these compartments. Analysis of IEL and LPL preparations from mixed BM chimeras revealed that the fraction of CD8αα γδT and αβT IELs was more strongly reduced in the IEL compartment than within the LPL compartment (Fig. 5 B). Reciprocally, in mice reconstituted with GPR18-tranduced Gpr18^−/− BM, the reporter-positive CD8αα γδT cells were enriched in the IEL compared with the LPL compartment (Fig. 5 C). These observations suggest that GPR18 augments access of CD8αα γδT cells to, or retention within, the IEL compartment.

In summary, we establish that the orphan receptor, GPR18, is required for the normal homeostasis of CD8αα γδT and αβT IELs and CD8αβ IELs. Our intravital imaging helps resolve a discrepancy between two recent studies (Chennupati et al., 2010; Edelblum et al., 2012) by providing additional evidence that IELs are motile and that they do not remain fixed between one set of epithelial cells but travel between intraepithelial niches through a path that sometimes takes them into the lamina propria. Studies on other cell types, such as germinal center and marginal zone B cells (Allen et al., 2004; Arnon et al., 2011), indicate that for cells to move repeatedly between adjacent compartments, they need to respond to at least two GPCR ligands that are distributed unequally between the compartments. The only directional cue previously established to promote IEL positioning in the small intestine is CCL25 and its receptor CCR9, and at steady-state these studies only found effects on γδT IELs (Abadie et al., 2012; Islam and Luster, 2012). Based on our findings here, we suggest that GPR18 may respond to a ligand that is distributed in areas overlapping with but not identical to the distribution of CCL25 and that the receptor helps achieve fine control of IEL distribution within the small intestine. We suggest that by affecting the efficiency of access to currently imprecisely defined intestinal niches, GPR18 influences exposure of IELs to factors necessary for their maturation or trophic support and that the reduced ability to access these factors most strongly affects their numbers under conditions of competition. Strong competitive effects on cell numbers have been observed in other instances where trophic support is limited (e.g., Lesley et al., 2004; Thien et al., 2004). The altered maturation state may in turn be responsible for the reduced in vitro motility of the cells, an alteration which might amplify effects of GPR18 deficiency on cell positioning and homeostasis. It is notable that mice lacking IL15Rα or AHR have decreased CD8αα IELs but retain normal numbers of CD8αβ IELs (Lodolce et al., 1998; Ma et al., 2009; Li et al., 2011). The reduced competitiveness of GPR18-deficient CD8αβ as well as CD8αα IELs indicates GPR18 may be required for accessing factors in addition to or other than IL15 or AHR ligands or that it is instead required for retention of IELs in the intestine. We do not exclude the possibility that GPR18 transmits signals that have their own trophic or maturational influence on IELs. It will be important in future studies to define the nature of the GPR18 ligand and to study whether GPR18 and GPR18 ligand abundance are altered in disease states where γδT IEL numbers and distribution are affected, such as in inflammatory bowel disease and celiac disease (Cheroutre et al., 2011; Abadie et al., 2012).

C57BL/6J (B6, CD45.2) and congenic B6 CD45.1⁺ mice were obtained from the Jackson Laboratory, and these strains were intercrossed to generate B6 CD45.1/2 F1 mice. CCR9-deficient mice (Uehara et al., 2002) were obtained from P. Love (National Institutes of Health, Bethesda, MD) and provided by E. Verdin (University of California, San Francisco [UCSF], San Francisco, CA). GPR18-deficient mice were generated by Ozgene Pty Ltd. using the Cre/loxP system. In brief, three DNA fragments (a 4-kb 5′ homology arm, a 1.4-kb floxed arm containing the single Gpr18 exon, and a 4-kb 3′ homology arm) were generated by nested PCR from C57BL/6 mouse genomic DNA and cloned into the pOzIII vector (Ozgene Pty Ltd.). The targeting vector was linearized and electroporated into B6-derived embryonic stem cells. Positive clones were screened and confirmed with Southern blot. Chimeras were bred to B6 mice, and germline transmission was confirmed by allele-specific PCR. Gpr18 floxed mice were intercrossed to actin-Cre transgenic mice to obtain Gpr18 germline deleted mice. To generate BM chimeras, CD45.1⁺ B6 mice were irradiated by exposure to 1,100 rad of γ-irradiation in two doses 3 h apart and i.v. injected with 2 × 10⁶ total BM cells from CD45.1/2 WT plus 2 × 10⁶ total BM cells from CD45.2 mice of each genotype as indicated (Gpr18^+/−, Gpr18^−/−, or Ccr9^−/−) and analyzed after 2–3 mo. All chimeras appeared healthy at the time of analysis. Although the input cells were mixed at a 1:1 ratio, the reconstitution efficiencies of the different donor BMs were not always identical, a common occurrence in mixed BM chimera studies. Animals were housed in a specific pathogen–free environment in the Laboratory Animal Research Center at the UCSF, and all experiments conformed to ethical principles and guidelines approved by the UCSF Institutional Animal Care and Use Committee.

Thymocyte, splenocyte, and LN cell suspensions were prepared by mashing the organs through 70-µm cell strainers. IELs and LPLs were isolated as described previously (Jiang et al., 2013), with modifications. In brief, Peyer’s patches were removed, and then the small intestine was opened longitudinally and washed with PBS containing 0.1% BSA, 100 U/ml penicillin, and 100 µg/ml streptomycin three times. The intestines were then shaken with prewarmed DMEM containing penicillin, streptomycin, and 5% FCS for 30 min at 225 rpm, 37°C. Supernatants were separated on a 30–40–80% Percoll density gradient (GE Healthcare), and the cells that layered between the 40–80% fractions were collected as IELs. After IEL isolation, tissues were shaken in RPMI 1640 containing 5 mM EDTA and 5% FCS at 175 rpm, 37°C for 15 min. This step was repeated four times and the supernatants were discarded. LPLs were then isolated after digestion in RPMI-1640 supplemented with 0.5 mg/ml collagenase type II (Worthington Biochemical Corporation), 0.1 mg/ml DNase I (Sigma-Aldrich), and 10% FCS at 200 rpm, 37°C for 30 min. Released cells were then subjected to Percoll fractionation as described above for isolation of IELs. For IEL adoptive transfer, freshly isolated 3–5 million IELs of each genotype were i.v. co-transferred into recipients. For transfers of T cells activated to induce CCR9 expression, T cells from LNs of each genotype were stimulated with plate-bound CD3 and CD28 plus 100 ng/ml retinoic acid for 3 d and then resuspended in complete RPMI-1640 media with 100 U/ml IL-2 plus 100 ng/ml retinoic acid. At day 5, cells were harvested and transferred into WT recipients. Transwell migration assays were with 5-µm transwells and were performed as described previously (Ngo et al., 1998) using 10⁶ IELs prepared by the Percoll density gradient method above. CCL25 and CXCL10 were obtained from BioLegend; CXCL12 was obtained from PeproTech; N-arachidonoyl-glycine and 2-arachidonyl-glycerol were obtained from Enzo Life Science and Sigma-Aldrich, respectively.

Cells were stained using standard procedures for surface markers. The following monoclonal antibodies were used for flow cytometry: TCRγδ (GL3; BD or BioLegend), TCRβ (H57; BioLegend), CD4 (GK1.5; BioLegend), CD8α (53.6.7; Tonbo Bio), CD8β (H35; eBioscience), Thy1.2 (30-H12; BD), Granzyme B (GB11; Invitrogen), CD45.1 (A20; BioLegend), and CD45.2 (104; BioLegend). Vγ7 antibody was provided by P. Pereira (Institut Pasteur, Paris, France). CD8αα cells were gated as CD8α⁺CD8β⁻. For cytokine analysis, freshly isolated IELs were stimulated with 40 ng/ml PMA (Sigma-Aldrich) and 4 µg/ml ionomycin (Sigma-Aldrich) for 4.5 h in the presence of GolgiPlug (BD) in complete RPMI-1640 media. Cells were then intracellularly stained for IFN-γ (XMG1.2; BioLegend) and TNF (MP6-XT22; eBioscience) with Cytoperm/Cytofix reagents (BD) according to the manufacturer’s instructions. For BrdU incorporation analysis, mice were given water containing 0.5 mg/ml BrdU for 7 d. Staining was performed with a BrdU flow kit (BD) according to the manufacturer’s instructions. Annexin V staining of fresh IELs was performed with the Annexin V staining kit (BD).

Cryosections of 7 µm were fixed and stained as previously described (Wang et al., 2011) with the following first antibodies: Laminin (Invitrogen), CD8α (53.6.7; Tonbo Bio), TCRγδ (GL3; BD or BioLegend), and E-cadherin (GG3A; BD). Images were all acquired with AxioVision using an Axio Observer Z1 microscope (Carl Zeiss).

Mice were i.v. injected with 10 µg anti–CD8α-PE (BioLegend) 5 h before imaging. After anesthetization, mice were injected i.v. with Hoechst 33342 dye (Invitrogen), and the distal duodenum was exposed and opened along the antimesenteric border, as described previously (Edelblum et al., 2012). The mucosal surface was placed against a coverslip bottom of a 3D-printed cassette containing PBS. Images were acquired with ZEN2012 (Carl Zeiss) using a 7MP two-photon microscope (Carl Zeiss) equipped with a Chameleon laser (Coherent). Excitation wavelength was 880 nm. Images were acquired by taking 21-µm Z-stacks at 3-µm steps every 20 s. Each XY plane spans 512 × 512 µm². Videos were made and analyzed with Imaris 7.4 ×64 (Bitplane).

Videos 1 and 2 show the real-time imaging of CD8α⁺ IEL migration in Gpr18^+/− and Gpr18^−/− duodenum, respectively.

Thanks to Jinping An for expert help with the mouse colony, Ying Xu for QPCR analysis, Jagan Muppidi for help with epitope tagging, Hsin Chen for imaging help, and Steven Rosen, Tangsheng Yi, Hayakazu Sumida and Michael Barnes for helpful discussions.

X. Wang is an Associate and J.G. Cyster is an Investigator of the Howard Hughes Medical Institute. This work was supported in part by National Institutes of Health grant AI40098.

The authors declare no competing financial interests.