Critical Role for SLAM/SAP Signaling in the Thymic Developmental Programming of IL-17– and IFN-γ–Producing γδ T Cells

Innate-like γδ T cells are unusual T cells that are highly enriched in skin and mucosal tissues in which they constitute a prominent source of cytokines and chemokines (1). In mice, most γδ T cells produce either IL-17 or IFN-γ, and IL-17 production is primarily restricted to the CD27⁻CCR6⁺ Vγ4 and Vγ6 subsets [Heilig and Tonegawa nomenclature (2)], whereas IFN-γ production is observed in CD27⁺CCR6⁻ Vγ1, Vγ4, Vγ5, and Vγ7 subsets (3–6). Accumulating data suggest that the relative balance of these IL-17– or IFN-γ–producing subsets can have significant consequences on disease outcome. IFN-γ–producing γδ T cells, for example, contribute to antitumor immunity (7) and inflammation in a model of cerebral malaria (3, 8). IL-17–producing γδ T cells (γδT17), in contrast, have been associated with protection against bacterial and fungal infections (9–11), promotion of tumor metastasis (12, 13), and immunopathologies in settings of autoimmunity (14–19).

Like their innate-like T cell counterparts, invariant NKT (iNKT) cells and MAIT cells, the functional programming of innate-like γδ T cells occurs during thymic development. Unlike iNKT and MAIT cells, however, a significant portion of γδ T cell developmental programming is restricted to a discrete developmental window (20), as it has been demonstrated that most natural γδT17 are generated only during the embryonic/early postnatal stage (21). Although the mechanisms that regulate the differentiation of γδ T cells into IFN-γ– or IL-17–producing subsets remain unclear, a significant body of data supports a model in which strong or weak TCR signals skew the developing γδ T cell toward IFN-γ (strong signal) or IL-17 (weak signal) production (22–26). In contrast, there is also evidence supporting a TCR-independent model in which the development of some γδT17 subsets is dependent on pre-existing hardwired transcriptional programs (27–29). Exactly how these seemingly independent mechanisms converge to regulate the developmental programming of innate-like γδ T cell subsets remains an open question.

SLAM family receptors compose a family of nine cell surface receptors (SLAMF1 [SLAM; CD150], SLAMF2 [CD48], SLAMF3 [Ly9; CD229], SLAMF4 [2B4; CD244], SLAMF5 [CD84], SLAMF6 [Ly108; CD352], SLAMF7 [CRACC; CD319], SLAMF8 [BLAME], and SLAMF9 [SF2001; CD84H]) that are expressed primarily on hematopoietic cells and play diverse roles in immune development and function (30). Most SLAM family receptors interact in a homophilic manner and, therefore, serve as self-ligands, the exception to this rule being the heterophilic interaction between SLAMF4 and SLAMF2. SLAM receptor signals can be both activating or inhibitory depending on the balanced recruitment of the cytosolic signaling adapter proteins SAP (31) and EAT-2 (32) and the inhibitory phosphatases SHP-1 or SHIP to immunoreceptor tyrosine switch motifs in the SLAM family receptor cytoplasmic domain.

Inactivation of the gene encoding SAP (Sh2d1a) disrupts the signaling pathway of the majority of SLAM family receptors and results in the development of the primary immunodeficiency X-linked lymphoproliferative disease that is characterized by broad immunological defects, among which is a failure in iNKT (33, 34) and MAIT (35) cell development. Investigations into the identity of the relevant SLAM family receptors involved in NKT cell development have revealed a complex interplay of individual receptors, including SLAMF1, SLAMF3, and SLAMF6 (34, 36–40). Interestingly, within the γδ T cell compartment, it was previously demonstrated that development of the Vγ1Vδ6.3 (γδNKT) T cell subset is dependent on SAP (41), indicating that the requirement for SLAM/SAP signaling during development is shared by NKT cells, MAIT cells, and at least one γδ T cell subset. Yet, very little is known regarding SLAM family receptor expression on γδ T cells in either the periphery or the thymus or the extent to which this receptor system is involved in γδ T cell development. In this study, therefore, we explored the extent to which the SLAM/SAP signaling pathway regulates the developmental programming of thymic γδ T cells that determines their effector function.

C57BL/6J, BALB/cJ, and B6.Sh2d1a^−/− (SAP^−/−) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Mice were housed and bred in a specific pathogen-free barrier room at the American Association of the Accreditation of Laboratory Animal Care–approved animal facility of the University of Vermont. Age- and sex-matched mice between 8 and 12 wk of age were used for experiments involving adult naive mice as indicated in the figure legends. To analyze γδ T cells in the embryonic thymus, timed-pregnant matings were used. The presence of a vaginal plug indicated day 0.5 of pregnancy. No attempt of sex matching was done for fetuses and newborn mice. All procedures involving animals were approved by the University of Vermont Institutional Animal Care and Use Committee.

Single-cell suspensions of splenocytes, lymph nodes, and thymocytes were obtained by gently pressing tissue through nylon mesh. RBCs were lysed using Gey's solution. Lung cell preparations were conducted as described (42). Lungs were dissected away from the trachea and finely minced in enzymatic digestion buffer (DMEM, 1 mg/ml collagenase type IV [Invitrogen], 0.2 mg/ml DNase [Sigma-Aldrich]). Lung tissue was resuspended in additional enzymatic digestion buffer and shaken at 200 rpm at 37°C for 20 min. The samples were triturated with a 16-gauge needle and allowed to digest for an additional 20 min, after which they were triturated one last time. After RBC lysis using Gey's solution, the digested tissue was passed through a 70-μm filter and washed in PBS/2% FCS.

Single-cell suspensions from skin were prepared as described (43). Briefly, depilated skin was digested in trypsin/GNK buffer (0.3% bovine pancreatic trypsin [type XI] [Sigma-Aldrich], 5 mM KCl, 150 mM NaCl, and 0.1% glucose [pH 7.6]) for 2 h at 37°C, after which the epidermis was separated from the dermis. The epidermis was then placed in trypsin/GNK buffer containing 0.5% DNase I and incubated for 15 min at 37°C. After incubation, samples were passed through 70-μm mesh, washed with DMEM + 10% FBS, resuspended in RPMI 1640 supplemented with 2 mM l-glutamine, 50 μM 2-ME, 50 U/ml penicillin, 50 μg/ml streptomycin + 10 U/ml IL-2, and incubated overnight at 37°C in a 5% CO₂ incubator prior to staining.

Liver intrahepatic leukocytes were isolated as previously described (44). Briefly, livers were perfused through the portal vein with PBS, subsequently removed, minced, and gently pressed through nylon mesh. The resulting cell suspension was washed twice and then resuspended in isotonic 33.8% Percoll (GE Healthcare). After centrifugation, the intrahepatic leukocyte cell pellet was resuspended and washed with PBS, and RBCs were lysed using Gey's solution. Single-cell suspensions were resuspended in PBS/2% FCS.

Intestinal intraepithelial lymphocytes (IELs) were isolated as previously described (45). Briefly, mouse small intestines were flushed with cold CMF buffer (1 mM HEPES, 2.5 mM NaHCO₃, 2% FBS in Ca²⁺/Mg²⁺-free HBSS [pH 7.2]). Surrounding fat tissue was removed, intestines were opened with a longitudinal incision, and excess mucous was removed by gentle scraping. Tissue was cut into small pieces and washed three times with CMF followed by shaking in CMF + 10% FBS/1 mM dithioerythritol (DTE) for 20 min at 37°C. This step was repeated, and the cells in the supernatants of both extractions were combined. Cells were incubated for 20 min at 4°C, before transferring cell supernatants into a fresh tube. Cells were centrifuged and washed once with RPMI 1640 before resuspending in 44% Percoll solution. Cells were overlaid on a 67% Percoll solution and spun for 20 min at room temperature. Cells at the interface were collected and washed twice with RPMI prior to counting and further analysis.

For all tissues, cell numbers were determined on an MACSQuant VYB (Miltenyi) using propidium iodide (1 μg/ml in PBS) for dead-cell exclusion. Cell stimulations were conducted in Click's medium (Sigma-Aldrich) supplemented with 2 mM l-glutamine, 50 μM 2-ME, 50 μg/ml gentamicin, 50 U/ml penicillin, 50 ug/ml streptomycin, and 5% FBS with 25 ng/ml PMA and 500 ng/ml ionomycin for 4 h at 37°C and 5% CO₂. After 2 h, 2 μM monensin was added to the culture.

Cells were stained with Live Dead Fixable Blue Staining Reagent (Thermo Fisher Scientific, Grand Island, NY) at 4°C in PBS for 30 min, after which, they were washed and resuspended in PBS/2% FCS containing 0.1% sodium azide and Fc-Block (BioLegend) to block nonspecific Ab binding prior to the addition of Abs. Brilliant Violet Staining Buffer (BD Biosciences) was used in experiments containing more than one Brilliant Violet or Brilliant UV fluorophore–conjugated Ab. To identify the Vγ6 subset, the staining procedure described in Roark et al. (46) was followed.

Cells were washed twice with PBS/2% FCS/0.1% sodium azide solution and fixed in PBS/1% PFA buffer. For intracellular cytokine staining, cells were permeabilized with PBS/1% FCS/0.1% sodium azide/0.1% saponin after surface staining and fixation. Twenty-five micrograms per milliliter total rat IgG were added to reduce unspecific binding prior to cytokine staining with fluorescent-labeled Abs. Nuclear transcription factors were analyzed using the Foxp3 transcription factor fixation/permeabilization reagent (Thermo Fisher Scientific). Briefly, after staining for surface markers, cells were resuspended in the fixation/permeabilization reagent and incubated overnight at 4°C. Cells were washed, incubated with 25 μg/ml total rat IgG prior to adding Abs against transcription factors. After staining at 4 ^οC, cells were washed and resuspended in PBS/2% FCS/0.1% sodium azide. Flow cytometry data were collected on an LSRII Flow Cytometer (BD Bioscience) and analyzed with FlowJo Software (Tree Star). Representative gating schemes used to identify γδ T cell subsets in thymus and lung are shown in the supplement (Supplemental Fig. 1).

Abs used in these experiments were as follows: CD11b (M1/70), CD19 (1D3), CD24 (M1/69), CD27 (LG.3A10), CD44 (IM7), CD45 (30-F11), CD48 (HM48-1), CD73 (TY/11.8), CD84 (mCD84.7), CD150 (TC15-12F12.2), CD229 (Ly9ab3), CD352 (330-AJ), IFN-γ (XMG1.2), IL-17A (TC11-18H10.1), TCRβ (H57-597), TCRδ (GL3), Vγ1 (2.11), Vγ5 (536), and Vδ6.3 (C504.17c) from BioLegend; CD196 (140706), CD244 (2B4), and RORγt (Q31-378) from BD Biosciences; and CD244.1 (C9.1), CD352 (13G3-19D), Eomes (Dan11mag), and Vγ4 (UC3-10A6) from eBiosciences.

Dimensionality reduction of flow cytometry data were conducted using t-stochastic neighbor embedding (t-SNE) in FlowJo. First, traditional gating was used to identify γδ T cells, after which the γδ T cell node data from five B6 and five BALB/c mice were concatenated for analysis. t-SNE was conducted on the SLAMF1, SLAMF4, SLAMF6, Vγ1, Vγ4, Vγ6, CD44, and CD27 parameters on the concatenated B6/BALB/c data using 1000 iterations (perplexity = 30, learning rate = 957), after which the B6 and BALB/c data were extracted for independent visualization.

Leukocytes from lung cell preparations were isolated using anti-CD45 Microbeads and LS columns according to the manufacturer’s instruction (Miltenyi). Cells were stained with Fixable Viability Dye eFluor780 (eBioscience), washed, and surface stained with Abs against SLAMF1, SLAMF6, NK1.1, TCRδ, TCR-Vγ1, TCR-Vγ4, TCRβ, CD11b, CD19, and CD45 in sterile PBS/2% FBS. Cells were sorted on an FACS Aria III (BD Bioscience) into culture medium, pelleted, washed, and snap frozen until later use. RNA was isolated using an RNeasy Micro Kit (Qiagen). RNA quality was assessed using an Agilent Bioanalyzer 2100, and RNA concentration was calculated using a Qubit Fluorometer (Thermo Fisher Scientific). Amplified cDNA was prepared using the Ovation Pico WTA System V2 (Nugen Technologies), and biotinylated using the Encore biotin module (Nugen Technologies). Hybridization with GeneChip Mouse Gene 2.0 ST arrays (Affymetrix) was conducted overnight at 45°C at 60 rpm. CEL files were analyzed using Transcriptome Analysis Console (Thermo Fisher Scientific) Software. Genes that demonstrated a greater than or less than 2-fold difference in expression and a false-discovery rate of < 0.2 were identified. This resulted in 316 SLAMF1-enriched genes and 229 SLAMF6-enriched genes. These gene lists were filtered by removing unannotated transcripts, predicted transcripts, and all but one of the 28 multiple transcript isoforms of Syne1, yielding 244 SLAMF1-enriched and 181 SLAMF6-enriched transcripts. All differentially expressed genes in this list exhibited a p value < 0.004. The raw data files for this analysis were deposited in National Center for Biotechnology Information Gene Expression Omnibus (GSE128544), https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE128544.

Data were analyzed for gene-ontological and pathway enrichment using National Institutes of Health DAVID analysis and Ingenuity Pathway Analysis (IPA) (filtered to include only those with a p value and FDR < 0.05). Heatmaps were generated using pretty heatmap with row-scaling. Upstream regulator prediction using IPA was conducted on the 425 differentially expressed genes based on fold-change (SLAMF1⁺SLAMF6⁻/SLAMF1⁻SLAMF6⁺). The result was filtered to include only those predicted regulators with p value of overlap <1 × 10⁻⁵, ranked according to activation z-score, after which the 10 highest (predicted to be active in SLAMF1⁺SLAMF6⁻) and lowest (predicted to be active in SLAMF1⁻SLAMF6⁺) were chosen for display.

Statistical analysis was performed using Prism (GraphPad Software, San Diego, CA). Unpaired Student t tests, one-way ANOVA, or two-way ANOVA was used where appropriate. For ANOVA posttest comparisons, the Dunnett, Tukey, or Sidak multiple-comparison tests were used where appropriate. In all cases, tests were considered significant when p < 0.05. Unless otherwise indicated, error bars represent SD.

γδ T cell effector functions are largely programmed during thymic development, and IL-17 functional programming is thought to occur primarily during embryonic and neonatal thymic development (21). Therefore, we examined the coexpression of SLAMF1 through SLAMF6 receptors on γδ T cell subsets in C57BL/6 embryonic day 17 (E.17) thymus. This analysis revealed that SLAMF5 was undetectable on γδ T cells and that the expression of SLAMF2 and SLAMF3 was relatively uniform among γδ T cell subsets (Supplemental Fig. 2). In contrast, SLAMF1, SLAMF4, and SLAMF6 exhibited a surprising degree of heterogeneity in their coexpression (Fig. 1A). For example, a significant fraction of E.17 thymic γδ T cells were SLAMF1⁻SLAM4⁺SLAMF6⁻, and this population consisted primarily of Vγ5 T cells (Fig. 1B). In the skin, mature Vγ5 T cells exhibit the same SLAMF1⁻SLAMF4⁺SLAMF6⁻ phenotype (Supplemental Fig. 2), suggesting that the Vγ5 SLAM family receptor expression profile is established during embryonic thymic development (47). Similarly, we observed that Vγ6 T cells were predominantly SLAMF1⁺SLAMF4⁻SLAMF6⁻ in the E.17 thymus (Fig. 1A), similar to the SLAM receptor expression profile that they exhibit in the adult lung (see below).

In contrast to the Vγ5 and Vγ6 populations, E.17 Vγ1 and Vγ4 T cells showed broad expression of SLAMF1 and SLAMF6 and they rarely expressed SLAMF4 (Fig. 1A). We identified prominent SLAMF1 and SLAMF6 single positive (SP) Vγ1 and Vγ4 T cell populations that represented the predominant phenotypes detectable in peripheral tissues (Figs. 1A, and below). Consistent with a previous report that γδT17 preferentially express SLAMF1 (48), RORγt⁺ E.17 γδ T cells were largely found within a SLAMF1^hiSLAMF6^low γδ T cell subset (Fig. 1C) and the majority of E.17 γδT17 were SLAMF1⁺, whereas the majority of IFN-γ–producing γδ T cells were SLAMF1⁻ (Fig. 1D).

In addition to the SLAMF1 and SLAMF6 SP Vγ1 and Vγ4 T cell populations, we detected prominent SLAMF1SLAMF6 double positive (DP) Vγ1 and Vγ4 T cell populations (Fig. 1A) that were rarely observed in the periphery (see below). To characterize this SLAMF1SLAMF6 DP population in more detail, we evaluated neonate thymus Vγ1 and Vγ4 SLAMF1 and SLAMF6 expression in the context of CD24 and CD73 expression, which can be used to track γδ T cell development (49, 50). Recently, two distinct pathways of Vγ1 and Vγ4 maturation have been defined using these markers (51). In pathway 1, used by both Vγ1 and Vγ4 T cells, immature CD24⁺CD73⁻ γδ T cells upregulate CD73 (CD24⁺CD73⁺) prior to CD24 downregulation and maturation (CD24⁻CD73⁺). In pathway 2, used by Vγ4 and Vγ6 T cells, immature CD24⁺CD73⁻ γδ T cells downregulate CD24 (CD24⁻CD73⁻) prior to CD73 upregulation. Consistent with a previous report (51), we found that day 7 thymus Vγ1 and Vγ4 T cells exhibited distinct patterns of CD24 and CD73 expression and that Vγ1 cells exhibited prominent immature CD24⁺CD73⁻ and mature CD24⁻CD73⁺ populations, whereas Vγ4 cells were primarily CD24⁺CD73⁻ and CD24⁻CD73⁻ (Fig. 1E).

Examination of SLAMF1 and SLAMF6 expression on these subsets revealed that in both Vγ1 and Vγ4 T cells, the SLAMF1SLAMF6 DP populations largely resided within the immature CD24⁺CD73⁻ population, whereas SLAMF1 SP and SLAMF6 SP and SLAMF1SLAMF6 double negative (DN) cells predominated in the mature CD24⁻CD73⁺ and CD24⁻CD73⁻ populations (Fig. 1F). In Vγ1 cells, which develop using pathway 1, there was a significant decrease in the frequency of SLAMF1SLAMF6 DP cells as cells progressed through the CD24⁺CD73⁻ to CD24⁺CD73⁺ to CD24⁻CD73⁺ stages (Fig. 1G), and the prominent mature CD24⁻CD73⁺ Vγ1 population was largely dominated by SLAMF6 SP Vγ1 cells (Fig. 1F). In Vγ4 cells, which develop primarily using pathway 2 during neonatal development (51), there was a significant decrease in the frequency of SLAMF1SLAMF6 DP cells as they progressed from CD24⁺CD73⁻ to CD24⁻CD73⁻ (Fig. 1G). Interestingly, mature CD24⁻CD73⁻ Vγ4 cells were predominantly SLAMF1 SP, whereas the small number of mature CD24⁻CD73⁺ Vγ4 cells were predominantly SLAMF6 SP and SLAMF1SLAMF6 DN (Fig. 1F). Similar results were observed in the adult thymus, with the exception that in Vγ4 cells, which develop primarily using pathway 1 during adult development (51), there was a decrease in the frequency of SLAMF1SLAMF6 DP cells as they progressed from CD24⁺CD73⁻ to CD24⁺CD73⁺ (Supplemental Fig. 2). In both Vγ1 and Vγ4 subsets, we noted the presence of SLAMF1 SP and/or SLAMF6 SP γδ T cells in the immature CD24⁺CD73⁻ populations (Fig. 1F), suggesting that the alterations in SLAM family receptor expression are already occurring at this developmental stage. Taken together, these data indicated that the coexpression of SLAMF1, SLAMF4, and SLAMF6 distinguished three groups of γδ T cell subsets in the thymus: Vγ1/Vγ4, Vγ5, and Vγ6. Our findings were consistent with a developmental scheme in which immature CD24⁺CD73⁻ thymic Vγ1 and Vγ4 T cells initially possess a SLAMF1SLAMF6 DP profile that resolves into either mature SLAMF1, SLAMF6 SP, or SLAMF1SLAMF6 DN phenotypes upon maturation and acquisition of function.

SLAMF1, SLAMF4, and SLAMF6 transduce signals by using the signaling adapter proteins SAP and EAT-2, which bind to immunoreceptor tyrosine switch motifs located in the cytoplasmic tails of SLAM family receptors (30). SAP is the most widely expressed of these adapter proteins in γδ T cells, and SAP-deficient mice have previously been reported to be deficient in the IFN-γ–producing Vγ1Vδ6.3 (γδNKT) subset (52). To determine whether the SLAM/SAP signaling pathway played any additional roles in the functional programming that γδ T cells undergo during thymic development, we examined γδ T cell development in SAP-deficient mice during embryonic and neonatal stages of development.

A comparison of thymic γδ T cells between B6- and B6.SAP-deficient mice revealed no significant change in the overall frequencies of Vγ4, Vγ5, or Vγ6 T cell subsets, although we did confirm the critical role of SAP in the development of the Vγ1Vδ6.3 subset on day 5 (52) (Supplemental Fig. 3). An examination of the SLAM family receptor expression profiles on E.17 γδ T cells revealed a significant decrease in the frequency of SLAMF1 SP Vγ1 and Vγ4 T cells (Fig. 2A). The reduction in the SLAMF1 SP subset coincided with a significant increase in the frequency of SLAMF1SLAMF6 DP subsets in both E.17 (Fig. 2A) and neonate (Supplemental Fig. 3) B6.SAP^−/− thymus. In accordance with our observations that SLAMF1SLAMF6 DP largely resided within the CD24⁺CD73⁻ population, whereas SLAMF1 and SLAMF6 SP cells largely resided within the CD24⁻CD73⁻ and CD24⁻CD73⁺ populations (Fig. 1F), we observed a significant increase in the frequency of CD24⁺CD73⁻ thymic γδ T cells in SAP-deficient mice (Fig. 2B). Concomitant with this increase was a significant decrease in CD24⁺CD73⁺, CD24⁻CD73⁺, and CD24⁻CD73⁻ γδ T cells (Fig. 2B).

Examination of the individual γδ T cell subsets revealed that among Vγ1 T cells, SAP deficiency resulted in increased CD24⁺CD73⁻ and CD24⁺CD73⁺ populations and a decreased frequency of CD24⁻CD73⁺ cells (Fig. 2C). This was due in large part to the near-complete absence of the Vγ1Vδ6.3⁺ subset, which was primarily CD24⁺CD73⁺ and CD24⁻CD73⁺ (Supplemental Fig. 3), as we observed no significant SAP-dependent changes in maturation among the Vγ1Vδ6.3⁻ T cells (Fig. 2C). Therefore, among neonate Vγ1 T cells, the major effect of SAP deletion was on the Vγ1Vδ6.3⁺ subset. Surprisingly, however, we also observed that among Vγ4 T cells, SAP deficiency resulted in significantly more CD24⁺CD73⁻, and significantly fewer CD24⁻CD73⁻ Vγ4 cells, suggesting that SAP was critical in the CD24⁺CD73⁻ to CD24⁻CD73⁻ (i.e., pathway 2) transition of a Vγ4 subset (Fig. 2D). No effect of SAP deficiency on the CD24 expression of the Vγ6 subset was observed (data not shown). Together, these data suggest that SAP-mediated signaling is integral to the developmental progression of Vγ4 as well as Vγ1 γδ T cells during thymic development and that it functions to promote γδ T cell maturation.

Next, we assessed whether SAP-deficient thymic γδ T cells exhibited any functional defects. Examination of the E.17 CD44⁺RORγt⁺ γδ T cell population revealed a significant decrease in the frequency of CD44⁺RORγt⁺ γδ T cell subsets in SAP-deficient mice (Fig. 3A). Significant decreases were observed in the Vγ1, Vγ4, and Vγ1⁻4⁻5⁻ subsets, but not in Vγ5, very few of which were CD44⁺RORγt⁺ (Fig. 3A). Examination of neonatal day 5 thymocytes revealed that the SAP-dependent decrease in CD44⁺RORγt⁺ γδ T cells was not restricted to embryonic thymic development, as there was a significant decrease in the frequency of CD44⁺RORγt⁺ Vγ4 T cells, which comprise the majority of thymic γδ T cells at this stage of development (Fig. 3B). However, the SAP-dependent decrease in CD44⁺RORγt⁺ Vγ1 and Vγ1⁻4⁻5⁻ populations was not observed in the neonatal thymus (Fig. 3B).

Next, we investigated whether SAP played a role in the acquisition of γδ T cell function during thymic development. A comparison of E.17 γδ T cell cytokine production between B6 and B6.SAP^−/− mice revealed a significant decrease in both IL-17 and IFN-γ production in SAP-deficient mice (Fig. 3C). Subset analysis revealed that the reduction in IFN-γ production was due primarily to an effect on Vγ5 γδ T cells (Fig. 3D), whereas the reduction in IL-17 production was instead observed in both Vγ1 and Vγ4 γδ T cell subsets (Fig. 3E). No significant decrease in the IL-17–producing Vγ1⁻Vγ4⁻Vγ5⁻ population was observed at this time, consistent with SAP-independent development of the Vγ6 population (data not shown).

Analysis of later stages of development revealed that the effect of SAP deficiency was dependent on both the subset and developmental stage. A significant effect of SAP deficiency was observed on thymic Vγ4 IL-17 production at all stages examined except for day 0. Conversely, we found that Vγ4 IFN-γ production was significantly decreased in SAP-deficient mice on day 0, but not any of the other timepoints examined. Similarly, we noted decreased thymic Vγ1 IL-17 in B6.SAP^−/− mice only at E.17 and decreased thymic Vγ1 IFN-γ only on days 5 and 7 (Fig. 3E). Together, these data suggested that the SLAM/SAP signaling pathway played a significant role in the developmental programming of thymic Vγ1, Vγ4, and Vγ5 IL-17 and IFN-γ production during embryonic and neonatal thymic development and that SAP coordinately regulated γδ T cell maturation and the acquisition of γδ T cell function.

Next, we examined the extent to which SAP-dependent thymic γδ T cell development regulated γδ T cell function in peripheral tissues. We first assessed the coexpression of SLAMF1 through SLAMF6 on γδ T cell subsets from different tissues. Similar to our results in the thymus, we found little evidence of SLAMF5 expression, and we found uniform expression of SLAMF2 and SLAMF3 on γδ T cells and from the lung, spleen, and lymph node (Supplemental Fig. 4). In contrast, our analysis revealed significant heterogeneity in SLAMF1, SLAMF4, and SLAMF6 coexpression on γδ T cell subsets obtained from different tissues (Fig. 4A). In the lung, we found two distinct populations of Vγ1 and Vγ4 γδ T cells that were either SLAMF1 SP or SLAMF6 SP. In contrast, Vγ6 γδ T cells were uniformly SLAMF1 SP with no detectable SLAMF6 expression (Fig. 4A). A similar SLAM receptor expression pattern was observed in the spleen, although the frequency of the SLAMF1 SP Vγ4 T cell population was markedly reduced compared with the lung and essentially absent in the Vγ1 subset (Fig. 4A). In contrast to lung and spleen, a significant proportion of SLAMF4-expressing γδ T cell subsets were detected in the liver, whereas gut IEL γδ T cells only expressed SLAMF4, and not SLAMF1 or SLAMF6 (Fig. 4A). Heterogeneity in SLAM family receptor coexpression was also observed in BALB/cJ γδ T cells, which possess a different Slam haplotype (53) (Fig. 4B). We did note significant differences between the strains indicating that SLAM receptor coexpression profiles were influenced by genetic background. Notably, we observed a significantly greater proportion of SLAMF1⁺SLAMF6⁻ lung and spleen Vγ4 T cells in BALB/c compared with their B6 counterparts (Fig. 4B).

Given the association of SLAMF1 expression with γδ T cell IL-17 production in the thymus (Fig. 1D), we assessed whether distinct SLAM family receptor expression profiles coincided with CD44 and CD27 expression, which are canonical markers of γδT17 and IFN-γ–producing γδ T cells, respectively. We focused our attention on lung γδ T cells because they exhibited a diversity of SLAM family receptor expression profiles. This analysis revealed that in both B6 and BALB/c mice, SLAMF1⁺SLAMF6⁻ γδ T cells were primarily CD44⁺CD27⁻, consistent with an IL-17–producing phenotype (48), whereas SLAMF1⁻SLAMF6⁺ γδ T cells were primarily CD44⁻CD27⁺ suggestive of an IFN-γ–producing phenotype (Fig. 4C). SLAMF4-expressing γδ T cells, in contrast, were distributed among both the CD44⁺CD27⁻ and CD44⁻CD27⁺ populations (Fig. 4C). Together, these data indicated that the coexpression of SLAM family receptors on γδ T cells constituted unique SLAM receptor expression profiles that varied in both a tissue-dependent and Vγ subset–dependent manner. The data also suggested that the SLAM receptor expression profiles of γδ T cell subsets in the periphery largely reflected those established upon maturation during thymic development, the exception being SLAMF4 whose expression on some γδ T cell subsets appeared to be modulated in the periphery. Finally, the data indicated that these SLAM receptor expression profiles were present in strains representative of Slam haplotypes 1 and 2 and that in each haplotype SLAMF1 and SLAMF6 expression exhibited a strong association with canonical markers of γδ T cell function.

To determine whether SLAM receptor expression profiles marked functional subsets of γδ T cells, we assessed lung γδ T cell cytokine production after stimulation with PMA/ionomycin. These data revealed that lung γδ T cell IFN-γ production was associated with SLAMF1⁻SLAMF6⁺ γδ T cells, whereas IL-17 production was associated with SLAMF1⁺SLAMF6⁻ γδ T cells (Fig. 5A). Similarly, stimulation of both spleen (Fig. 5B) and lymph node (Fig. 5C) γδ T cells revealed a strong association of SLAMF1 and SLAMF6 expression with γδ T cell IL-17 and IFN-γ production, respectively. Together, these data suggested that the mutually exclusive expression of SLAMF1 and SLAMF6 marked IL-17– and IFN-γ–producing Vγ1, Vγ4, and Vγ6 γδ T cell subsets in the periphery.

To explore this association further, we compared the ex vivo transcriptional profiles of sorted lung SLAMF1⁺SLAMF6⁻ and SLAMF1⁻SLAMF6⁺ lung Vγ4 from naive C57BL/6J mice. This analysis revealed the differential expression of 244 genes in SLAMF1⁺SLAMF6⁻ and 181 genes in SLAMF1⁻SLAMF6⁺ lung Vγ4 (Fig. 6A). In accordance with our functional analysis, we found that SLAMF1⁺SLAMF6⁻ lung Vγ4 γδ T cells exhibited a transcriptional profile highly suggestive of IL-17 production. We observed differential expression of Il23r, Il17a, Il17f, Il1r1, Blk, Rorc, Sox13, Maf, Ccr6, Ccr2, and Scart2, all of which have been previously associated with a γδT17 phenotype (27, 28, 54) (Fig. 6A). In contrast, Ifng, Eomes, IL12rb, and Crtam were preferentially expressed in SLAMF1⁻SLAMF6⁺ Vγ4 γδ T cells, suggestive of an IFN-γ–producing phenotype (Fig. 6A). These findings were consistent with a recent report demonstrating that EOMES expression marks IFN-γ–producing γδ T cells (55).

IPA of differentially expressed genes revealed that transcripts enriched in SLAMF1⁺SLAMF6⁻ lung Vγ4 cells exhibited significant overlap with genes regulated by ZBTB16 (PLZF), IL23R, RORγT, and IL-1B. In contrast, transcripts enriched in SLAMF1⁻SLAMF6⁺ lung Vγ4 exhibited significant overlap with genes regulated by IL-12, TBX21 (T-bet), IFN-γ, IL-27, and TCF3 (E2A) (Fig. 6B). These transcriptional profiles were confirmed using flow cytometry, which revealed that RORγt expression was largely confined to the SLAMF1⁺SLAMF6⁻ γδ T cells in both lung and lymph node, whereas EOMES expression was found in the SLAMF1⁻SLAMF6⁺ γδ T cell subsets (Fig. 6C). Together, these data suggested that lung SLAMF1⁺SLAMF6⁻ and SLAMF1⁻SLAMF6⁺ Vγ4 γδ T cells possessed transcriptional profiles consistent with IL-17 and IFN-γ production, respectively.

Finally, we examined the extent to which impaired thymic γδ T cell developmental programming in SAP-deficient mice affected γδ T cell function in peripheral lung and lymphoid organs. A comparison of neonate lung γδ T cells between B6 and B6.SAP^−/− mice revealed that there was a significant decrease in lung Vγ4 IL-17, but not IFN-γ, production (Fig. 7A), and this decrease corresponded to a significant decrease in SLAMF1⁺IL-17⁺ Vγ4 cells (Fig. 7B). No difference was observed in either lung Vγ1 IL-17 or IFN-γ production (Fig. 7A), nor did we observe any SAP-dependent effects on Vγ6 IL-17 production (data not shown). Interestingly, the SAP-dependent decrease in Vγ4 T cell IL-17 production did not persist in the adult lung. Instead, a comparison of adult lung γδ T cell cytokine production between B6 and B6.SAP^−/− mice revealed a significant decrease in both lung Vγ1 and Vγ4 IFN-γ production (Fig. 7C). We detected no SAP-dependent effect on Vγ6 IL-17 production (data not shown). A comparison of adult spleen Vγ1 and Vγ4 cytokine production between B6 and B6.SAP^−/− mice also revealed significantly less IFN-γ in SAP-deficient Vγ1 and Vγ4 T cells than in their B6 counterparts, indicating that the SAP-dependent decrease was not tissue-specific (Fig. 7D). Together, these data demonstrated that γδ T cell IL-17 and IFN-γ production in the periphery was significantly impaired in SAP-deficient mice. In addition, the data indicated that the SAP-dependent impairment in γδ T cell function in the periphery was both age and subset-dependent.

In this study, we identified the SLAM/SAP signaling pathway as a critical regulator of the functional programming that generates both γδT17 and IFN-γ–producing γδ T cells during thymic development. We found that maturation of thymic Vγ1 and Vγ4 was associated with the transition from a SLAMF1SLAMF6 DP phenotype to either SLAMF1 or SLAMF6 SP or SLAMF1SLAMF6 DN phenotypes and that this transition was concomitant with the acquisition of γδ T cell effector function in the thymus. In the periphery, SLAMF1 and SLAMF6 marked γδT17 and IFN-γ–producing γδ T cells, respectively. Importantly, our findings indicated that in addition to its role in Vγ1Vδ6.3 T cell development, SAP also regulated Vγ4 T cell development and that SAP-deficient Vγ4 T cells exhibited impaired developmental progression in the thymus and impaired effector function in both the thymus and the periphery. Taken together, these findings suggest that the SAP-dependent signaling through SLAM receptors has a greater influence on γδ T cell development than previously appreciated and that this signaling pathway is a major contributor to the developmental programming that governs γδ T cell effector function.

The exact developmental stage(s) at which thymic γδ T cell functional programming occurs remains an open question. It has previously been shown that the CD24⁻CD44⁻CD45RB⁻ γδ T cell population can give rise to both IFN-γ and IL-17 producers (24), but it is unclear whether this population is composed of uncommitted precursors or a mixture of already committed γδ T cells consisting of several functional subsets. Our findings that immature CD24⁺ thymic γδ T cells are SLAMF1SLAMF6 DP and that mature CD24⁻ γδ T cells are enriched in SLAMF1 and SLAMF6 SP cells suggest that the CD24⁻CD44⁻CD45RB⁻ γδ T cell population could be a heterogeneous mixture of SLAMF1 and SLAMF6 SP γδ T cells that have already committed to either IL-17 or IFN-γ production. Our results are consistent with a model in which γδ T cell functional programming occurs within the immature CD24⁺ population because both RORγt⁺ and RORγt⁻ γδ T cells are found within the CD24⁺CD73⁻ population (51), as are SLAMF1 SP γδ T cells, which constitute the major subset of γδT17.

A number of studies support a model in which high-affinity binding of γδ TCR during development promotes differentiation toward IFN-γ–producing cells, whereas low-affinity interactions favor programming for IL-17 production (22–26, 56). In light of our findings indicating that both IL-17 and IFN-γ production are altered in SAP-deficient mice, it is interesting to note that SLAM receptor signaling has previously been shown to modulate TCR signaling thresholds in conventional αβ T cells (57). In addition, recent reports indicate that NKT cell differentiation is regulated by TCR signal strength (58, 59) and that the net effect of SLAM receptor signaling is to reduce NKT TCR signal strength (36). Indeed, the coexpression of both SLAMF1 and SLAMF6 on immature CD24⁺CD73⁻ γδ T cells puts them in an optimal position to influence TCR signal strength and differentiation. Because CD73 expression has been reported to increase as a result of TCR signaling (49), SLAM/SAP-mediated regulation of TCR signal strength might be expected to influence the CD24⁺CD73⁻ to CD24⁺CD73⁺ transition. However, some of our findings are not entirely consistent with such a model. First, SAP deletion did not affect neonatal Vγ4 CD73 expression. Instead, it appeared to block the transition from CD24⁺CD73⁻ to CD24⁻CD73⁻ (pathway 2). Second, we found that the frequency and function of Vγ6 T cells, whose development is regulated by TCR signal strength (25), did not appear to be dependent on SAP. Additional investigation will be required to determine whether the SAP regulates γδ TCR signaling or a TCR-independent process.

Accumulating evidence suggests that the development of at least some γδT17 can be regulated by TCR-independent mechanisms. γδ T cell development is dependent on a characteristic subset of transcription factors including SOX4, SOX13, Blk, c-maf, HEB, and PLZF, among others (27, 28, 51, 54, 60, 61). HEB was recently demonstrated to be critical for the progression of neonatal IL-17–producing Vγ4 T cells through pathway 2 (51). Therefore, our finding that SAP also regulates the progression of Vγ4 cells through pathway 2 provides a point of intersection between HEB and SLAM/SAP signaling that warrants future examination. In addition, it was recently reported that at least some γδT17 programming results from a SOX13-dominated transcriptional hard-wiring of early thymocyte precursors (29). These SOX13 progenitors appear to give rise to IL-17–producing Vγ4 cells but not Vγ6 cells. Our observation that Vγ4, but not Vγ6, subsets are SAP-dependent are, therefore, consistent with a role for SLAM receptors and SAP in the initiation of a transcriptional program critical for the acquisition of function. Indeed, because SLAM receptors are expressed on hematopoietic stem cells (62–64), they are in a position to influence γδ T cell programming at the earliest stages of development. Still, we note that because littermate controls were not used in these studies, we cannot exclude the possibility that the SAP-dependent effects on γδ T cell development and function observed in this study are due to strain-specific differences in microbiota or to developmental differences among the different litters. A future assessment of the role that SLAM/SAP signaling plays in γδ T cell precursor-product relationships using fetal thymic organ culture will be helpful in confirming our observations in this study and in elucidating the mechanism(s) through which this pathway regulates γδ development.

Our findings demonstrate the presence of distinct SLAM family receptor expression profiles on γδ T cell subsets in the periphery that are tightly associated with function. These findings not only confirm a previous report that SLAMF1 marks γδT17 (48), but they reveal the presence of a mutually exclusive expression pattern of SLAMF1 and SLAMF6 on γδ T cells that is tightly associated with distinct transcriptional programs and with IL-17 and IFN-γ production, respectively. Heterogeneous SLAM family receptor coexpression was observed among both B6 and BALB/c γδ T cells that possess different Slam haplotypes (53), although we did note significant strain-dependent differences. Specifically, there was a greater proportion of SLAMF1SP Vγ4 in BALB/c lung, liver, and spleen, which would suggest an increased frequency of IL-17–producing Vγ4. Although we have yet to confirm this prediction, it is interesting to note that NKT cell subsets are also differentially distributed in a strain-dependent manner and that IL-17–producing NKT cells are also especially prevalent in BALB/c mice (65).

The finding that certain SLAM family receptor expression patterns are established during thymic development and that they are already associated with function as early as E.17 suggests that they result from a developmental program. In support of this notion, significant changes in SLAMF1, SLAMF4, SLAMF6, and SLAMF7 expression have been described on B cells as they progress through developmental stages (66), and SLAM receptor expression profiles on hematopoietic stem cell and multipotent progenitor cell populations (62) have recently been demonstrated to mark functionally distinct progenitor populations (64). Together, our findings suggest the possibility that the coordinate signaling of distinct arrays of SLAM family receptors plays a role in the development of γδ T cells in the thymus. This would be consistent with reports indicating that multiple SLAM receptors are involved in NKT cell development (34, 36, 38, 39) and may explain the inability to observe γδT17 defects in mice lacking only a single SLAM family receptor (48).

Our observations indicated that SAP deletion did not completely abolish the development of IL-17–producing Vγ4 but, instead, reduced their numbers in the thymus and neonatal lung by ∼40%. This finding likely suggests the existence of one or more specific SAP-dependent Vγ4 subsets, as is the case with the Vγ1 T cell population that is composed of the SAP-dependent Vγ1Vδ6.3 subset and other SAP-independent subsets. Recent data demonstrates that the TCR repertoire of IL-17–producing Vγ4 T cells is highly restricted (67, 68), which suggests the possibility that the partial decrease in Vγ4 IL-17 production observed in this study may be due to a complete failure in the development of a specific IL-17–producing Vγ4 subset. Such a model is also in line with the observation that the SAP-dependent decrease in lung γδ T cell IL-17 production observed in neonates was not consistently observed in adult mice. We speculate that the restricted development of IL-17–producing Vγ4 T subsets in embryonic and neonatal thymus (21) results in decreased SAP-dependent Vγ4 T subsets in the neonatal lung and that over time, homeostatic proliferation of SAP-independent Vγ4 T subsets results in a repopulation of the adult lung. A future analysis of the γδ TCR repertoire in SAP-deficient mice should shed light on this issue.

In summary, our findings demonstrate that the SLAM/SAP signaling pathway plays an important role in shaping the balance of γδT17 and IFN-γ–producing γδ T cell commitment during thymic development. The findings in this study not only extend the influence of this signaling pathway to γδ T cell subsets other than the Vγ1Vδ6.3 γδNKT cell subset, they also establish a framework of SLAM family receptor coexpression that might be useful in dissecting and understanding γδ T cell development and function. Future studies aimed at understanding the specific mechanisms of SLAM family receptor signaling in this process will be required to integrate this pathway into existing models and should provide insight into the central mechanisms that regulate γδ T cell functional programming.

We thank Roxana del Rio Guerra and the Flow Cytometry Cell Sorting Facility (Larner College of Medicine, University of Vermont) for help in cell sorting, Jessica Hoffman and Scott Tighe and the Vermont Integrated Genome Resource (Larner College of Medicine, University of Vermont) for help in microarray, Wendy Havran for providing a protocol for skin γδ T cell isolation, Cynthia Baldwin and Janice Telfer for helpful discussion, and Willi Born and Rebecca O’Brien for the 17D1 hybridoma.

The authors have no financial conflicts of interest.