|Home | About | Journals | Submit | Contact Us | Français|
γδ T cells function in the early phase of immune responses. Although innate γδ T cells have primarily been studied as one homogenous population, they can be functionally classified into effector subsets based on the production of signature cytokines, analogous to adaptive T helper subsets. Unlike adaptive T cells, however, γδ T effector function correlates with genomically encoded TCR chains, suggesting that clonal TCR selection is not the primary determinant of γδ effector differentiation. A high resolution transcriptome analysis of all emergent γδ thymocyte subsets segregated based on TCRγ/δ chain usage indicates the existence of three separate subtypes of γδ effectors in the thymus. The immature γδ subsets are distinguished by unique transcription factor modules that program effector function.
The Immunological Genome Project (ImmGen) is a consortium of immunologists and computational biologists constituted to comprehensively define gene regulatory networks in the immune system of the mouse 1. In this context, we determined global gene expression profiles of intrathymic γδTCR+ cell subsets to ascertain the heterogeneity of the T cell lineage and to identify gene networks controlling innate effector subset production distinct from those that operate in conventional adaptive T cells. T cells in vertebrates are separated into two lineages based on the expression of either γδTCRs or αβTCRs on their cell surface. The adaptive αβ T cell lineage is subdivided into T helper 1 (TH1), TH2, TH17, TH follicular, regulatory T and cytotoxic effectors that are often considered distinct cell lineages. There are indications that the innate γδ T cell lineage is also composed of distinct subsets that are programmed to secrete a discrete cluster of effector cytokines 2, 3, but the mechanism of innate effector specialization is unclear. γδ T cells are sentinels of the immune system, mostly localized to the mucosal epithelia where pathogens are first encountered 4 and as such, the rapidity of their response to infection is paramount. Similar to other innate lymphocytes 5, 6, γδ T cells are exported from the thymus as “pre-made” memory-like cells, displaying cell surface markers associated with cellular activation. Upon infection, γδ T cells rapidly produce effector cytokines and growth factors similar to memory αβ T cells 7–9.
While the genes encoding the γδTCR and αβTCR were identified contemporaneously, studies delving into the distinct development and function of γδ T cells have been greatly hampered by the scarcity of known molecules distinguishing them, other than the TCR and the transcription factor Sox13 10, that can be classified as bona-fide γδ T cell lineage-specific markers.. Identification of the genetic circuits underpinning two unique features of γδ T cell development is particularly critical. First, the segregation of adult γδ T cell effector functions is based on Vγ and Vδ gene usage 11: IL-17, IFN-γ, and IL-4 production are associated with Vγ2+ (designated as V2, Tγδ17), Vγ1.1+ Vδ6.3− (V1) and Vγ1.1+Vδ6.3+ (V6) γδ T cell subsets, respectively. In comparison, conventional αβ T cells are classified into functional subsets based on the repertoire of effector cytokines produced, not by TCR repertoire, which is vastly diverse within each subset. Second, γδ T cell subset function appears to be programmed in the thymus 12–14. In contrast, conventional αβ T cells differentiate into effector subsets upon encounter with pathogens in peripheral tissues. How and when γδ subsets are programmed and/or selected toward distinct effectors in the thymus is not well understood. To date, γδ T cell development has been studied mainly by analysis of total γδTCR+ cells as one uniform population, not as distinct subsets defined by Vγ/Vδ usage. We compared the gene expression profiles of emergent adult γδTCR+ thymocytes segregated based on Vγ/Vδ gene usage and show that the major subsets are as distinct from each other as they are from αβTCR+ thymocyte subsets. Most important, the profiles of the emergent immature γδ T cell subsets were already embedded with unique gene programs directing subset-specific effector function, indicating that γδ function is molecularly programmed in the thymus prior to, or concurrent with, TCR expression.
A correlation between effector function and TCR Vγ/Vδ chain usage 11 raised the possibility that the earliest identifiable γδ T cells in the thymus are composed of molecularly heterogeneous cell subtypes distinguishable by the expression of uniquegermline-encoded γδ TCRs. To test this possibility, we determined gene expression signatures of four adult γδ T cell subsets based on TCRγ and/or δ expression and maturational stages based on CD24 (HSA) expression: V2 (Vγ2+, see On-line methods for partner δ chain repertoire and frequency among total γδTCR+ thymocytes), V1 (Vγ1.1+), V6 (Vγ1.1+Vδ6.3+) and V5 (Vγ5+, intraepithelial lymphocytes, IELs) thymocyte subsets were compared in relation to other thymic subsets of the T cell lineage profiled within ImmGen. The paucity of mature (mat, CD24lo) V5 thymocytes prohibited their analysis. In addition, we established gene expression profiles of three immature (imm, CD24hi) fetal γδ cell subsets: V3 (Vγ3+, dendritic epidermal T cells, DETCs), V4 (Vγ4+) and V2 (Vγ2+).
Pair-wise comparisons showed that the earliest identifiable adult immV2 subset, the largest population within the γδ T cell lineage, was distinct in its global gene expression profile compared to all other adult γδ T cell subsets (Fig. 1). At an arbitrary threshold of 2-fold change in expression value, the number of genes distinguishing the immV2 subset from other immature γδ subsets was roughly equivalent to the difference observed between the immV2 subset and αβ lineage CD8+ thymocytes (Fig. 1). To contextualize this finding, differentially expressed genes between immature αβ lineage DP (CD4+CD8+CD69−) and MHC-selected DP69+ cells relative to total γδTCR+ thymocytes numbered ~1600 and ~900, respectively. Within theαβ T cell lineage, FOXP3+ regulatory versus conventional CD4+ T cells in the spleen differed in the expression of ~400genes (Fig. 1). The distinct gene expression signature of immV2 cells was in sharp contrast to the similarity between V1, V5 and V6 cells, which differed at most in the expression of 54 genes. However, the higher expression of the NKT lineage marker PLZF (encoded by Zbtb16) and the IEL-enriched Granzyme A 15 in immV6 and immV5 cells, respectively, suggests that despite their similarity, they were already marked with the molecular features of their fully differentiated peripheral counterparts (Fig. 2a, data not shown). Confirmation of the relative amounts of protein products of select differentially expressed transcription factors (TFs) in adult γδ T cell subsets is shown (Fig. 2a) with confirmatory data for additional differentially expressed genes (Supplementary Fig. 1), which captures the impressive heterogeneity in the phenotype of γδ subtypes. Fetal immV2 and ImmV4 cells (the fetal subset programmed for IL-17 production 11, 16) were highly similar in global gene expression profiles, correlating with their common Tγδ17 effector phenotype. In contrast, fetal immV3 cells, the DETC precursors, were distinct from other fetal γδ T cell subsets, consistent with their known distinct developmental origin and requirements 17, 18. Thus, the γδ T cell lineage can be divided into three distinct subtypes based on unique gene expression profiles at the immature developmental stage: V3, V4/V2 and V1/V6/V5.
Lineage relatedness among diverse subsets can be visualized by principal component analysis (PCA), a mathematical transformation that reduces the dimensionality of gene expression data to reveal the main components of variability in comparators. We focused on adult γδ subsets as the data set is complete. We first performed PCA to compare the gene expression profiles of non-segregated total γδTCR+ thymocytes to three αβ thymocyte subsets ranging in maturity and lineages using the 15% most differentially expressed genes in the comparators, based on expression levels and variance across the populations (Fig. 2b, see On-line methods). Matureαβ thymocyte subsets were positioned together, increasingly displaced from precursor subsets (ISP and DP) as a function of their relative maturity. Innate-like invariant αβNKT cells (iNKT) at three stages of maturation clustered separately from other αβ cell subsets and total γδ T cells were positioned away from the four distinct clusters of αβ thymocyte subsets (Fig. 2b). Similar results were obtained when subset relatedness was measured by Euclidean distance and Pearson’s correlation coefficients on the 15% most differentially regulated genes used for PCA (Supplementary Fig. 2a, b). When the analysis was performed using the segregatedγδ T cell subsets, however, it was clear that the immV2 subset was distinct from all other γδ thymocyte subsets, positioned uniquely in relation to other γδ and αβ subsets. (Fig. 2c, Supplementary Fig. 2c, d). Second, using genes differentially expressed in immV2 subsets compared to other immature γδ cell subsets (1006 genes, Supplementary Fig. 3a, Supplementary Tables 1, 2), PCA, Euclidian distances, and correlation scores supported a close similarity between immV2 and αβ DP thymocytes, which shared >50% of the variance and segregated distantly from T precursor subsets (Fig. 3a, Supplementary Fig. 3b, c). When DP populations were segregated, PCA indicated that immV2 cells were most similar to CD69+DP cells (Supplementary Fig. 3d). These results revealed that immV2 cells distinguished by germline-encoded TCR domains are radically different from other immature γδ T cells, and conversely, that γδ T cells expressing the Vγ2-Cγ1 TCR, despite the extensive TCR clonal diversity generated from their TCRδ partners and VJ/VDJ junctional diversity, nevertheless represent a unique cell type.
About one-third of the genes expressed at lower levels in immV2 cells versus V1/V6 cells encode for proteins involved in DNA synthesis, RNA biogenesis, oxidative phosphorylation and protein translation (~270 genes), suggestive of decreased metabolism and energy production in immV2 cells (Supplementary Tables 1, 3, 4) compared to other γδ subsets. The expression of metabolic genes in immV2 cells corresponded to the pattern seen in immature αβ lineage DP cells, providing the primary element of similarity between these two populations (Fig. 3b, Supplementary Table 4, and data not shown). The distinct gene signatures of γδ subsets were not a consequence of different cell cycle properties or susceptibility to death as the immature subsets incorporated BrdU similarly, had comparable frequencies of cells expressing Ki-67, a marker of non-G0 state of cell cycle (Supplementary Fig. 4a, b), and Annexin V, a marker of apoptotic cells (data not shown). These results suggest that immV2 cells develop distinctly with unique maturational transitions.
Lineage specific TFs propel the lineage commitment process by turning on lineage-associated genes and turning off lineage-mismatched genes. Distinct programming of the dominant TFs in subset-specific effector function was already evident in emerging γδ T cell subsets. Hierarchical clustering analysis of TFs differentially expressed in immV2 and immV1/V5/V6 cells indicated that most TFs were expressed at strikingly higher levels in the immV2 subset, the most prominent belonging to the SOX-TCF1-TOX High Mobility Group (HMG) box TF family (Fig. 3c and Supplementary Tables 2, 5, 6). Among immune cell types, Sox13 expression is restricted to developing γδ T cells. It interacts with the nuclear effectors of WNT signaling, TCF1 (encoded by Tcf7) and LEF1, to establish, in part, the pan-γδ lineage gene expression signature 10. Conversely, expression of Id2 and Id3, the negative regulators of E Helix-Loop-Helix TFs (E47 (encoded by Tcfe2a) and HEB (encoded by Tcf12)), along with their upstream regulator Egr2 were elevated in immV1/V5/V6 relative to immV2 cells 19, 20. Further, a negative regulator of NK cell development Bcl11b 21 was decreased in expression in immV1/V5/V6 relative to immV2 cells (Fig. 3c, Supplementary Table 6).
In the periphery, V2 CCR6+CD27− γδ T cells are biased to produce IL-17 while V1 CD27+ T cells secrete IFN-γ 8, 14. We further verified this effector pattern during the γδ T cell response to Mycobacterium tuberculosis infection 7. Fourteen days post infection in the lung, all γδ subsets synthesized IFN-γ, but V2 and V4 cells were the predominant IL-17 producers (Supplementary Fig. 4c–f). The TFs Rorc, Eomes and Zbtb16 regulate the expression of IL-17, IFN-γ and dual IL-4-IFN-γ, respectively, in αβ T cells 22–25. Emergent immature V2, V1/V5 and V6 thymic subsets were precociously distinguished by selective expression of these three TF genes, correlating with their peripheral γδ T cell subset-specific functions (Fig. 2a, ,3c).3c). As further support for the thymic programming of immV2 cells, the genes encoding the three markers of Tγδ17 cells, B lymphocyte kinase (BLK 26 and Supplementary Fig. 4f) and the scavenger receptors SCART1 and SCART2 27, were elevated in expression in immV2 cells compared to other subsets (Supplementary Table 2, 5, 6). Other factors required for αβ TH effector differentiation, including Maf 28, Gata3 29, 30 and Runx3 31, were also expressed more highly in immV2 cells relative to other γδ subsets (Fig. 2, ,3c).3c). These results demonstrate that prior to or upon expression of different TCRs, γδ thymocyte subsets are regulated by divergent TF networks that are predicted to program distinct effector functions.
The biological processes associated with thymic γδ T cell maturation have not been systematically studied. The gene expression profiles of mature (CD24lo) thymic γδ T cell subsets (V2, V1 and V6) were determined to examine the duration and impact of the immature subset-specific gene networks (Fig. 1). Despite the substantial subset specific differences among γδ T cells, there was a shared set of 495 genes that were coordinately regulatedupon maturation of all γδ T cells (Supplementary Fig. 5a, b, Supplementary Table 7). PCA, hierarchical clustering, Euclidian distances and correlation scores of these γδ maturation-dependent genes in other thymocyte subsets clearly showed that the γδ maturation signature was conserved during αβ T cell maturation as well, as precursors and immature subsets formed a distinct cluster, separate from the cluster of mature subsets, irrespective of T cell lineage (Fig. 4a, b, Supplementary Fig. 5c, d). Of the 495 γδ maturation genes identified, ~78% were coordinately regulated upon maturation of αβ T cells subsets (matCD4, matCD8, or iNKT vs. DPs), while ~22% were uniquely modulated during γδ T cell maturation only (data not shown). These results clearly establish the loss of CD24 as a demarcation of the transition to maturity in the γδ T cell lineage.
There were four overriding themes that characterized γδ subset-specific maturation. First, the expression pattern of metabolic/RNA-DNA biogenesis genes in matV2 cells and matV1 cells converged, while matV6 cells remained distinct (Fig. 4c and Supplementary Table 8). Second, upon maturation, V2 cells down-regulated subset-specific TFs to the level associated with immV1 cells (Fig. 4d, Supplementary Table 9). Third, mature γδ T cell subsets modulated expression of proteins directly involved in the execution of their peripheral subset-specific effector functions, including chemokine and cytokine receptors. Fourth, V1 and V6 cells diverged upon transit to the mature state, with the latter acquiring gene expression profile of PLZF+αβ iNKT cells. The latter two features of γδ T cell maturation are discussed in greater detail.
The transition to the mature stage endows adult γδ subsets with unique homing properties as evidenced by the acquisition and loss of chemokine and integrin receptors. There were two patterns of chemokine receptor expression: those expressed in immature subsets and then extinguished or downregulated as γδ T cells mature (highlighted in red), and those induced precipitously during maturation and maintained in fully differentiated subsets (blue) (Fig. 5a, Supplementary Table 10). The expression pattern of CCR10 was confirmed by analysis of Ccr10-Gfp reporter mice, with the highest frequency of CCR10+ immature thymocytes found in the V6 subset (Supplementary Fig. 1). CCR6 was strongly and uniquely upregulated in the V2 subset upon transit to the mature stage, whereas CXCR3 induction was associated with V6 and V1 subset maturation (Fig. 5a, Supplementary Fig. 1, Supplementary Table 10). CCR6 expression is tightly correlated with IL-17 producing lymphocytes 32 whereas CXCR3 expression is controlled in part by EOMES 33 and permits trafficking to non-lymphoid tissues. Interestingly, matV2 cells are Itgb7hiCxcr6hiCCR6+RORγt+IL-7Rhi, reminiscent of fetal liver-derived RORγt+ CD3− innate lymphoid cells (ILCs), some with the capacity to secrete IL-17 or IL-22 34. The dynamic subset-specific alterations in the tissue tropism upon maturation further underscore the separation of V2 cells from other γδ T cells and suggest that the Tγδ17 differentiation program may overlap with that of TCR negative ILCs.
Maturation stage-dependent alterations in cytokine receptor expression were also γδ T cell subset-specific and endow γδ subsets with the capacity to sense environmental cues that fully engage their programmed effector functions. Upon thymic maturation, V2 cells entered the effector-poised phase by abrupt super-induction of cytokine receptor genes that are dedicated for IL-17 production and responsiveness: Il23r (the most highly induced gene, ~100 fold increase in CD24lo relative to CD24hi), Il17re, Il1r1, IL-18r35 and Tnfrsf25 36 were the strongest induced genes in matV2 cells (Fig. 5b, Supplementary Table 7, 11). This super-induction was accompanied by the enhanced expression of RORγt and the signaling molecule BLK that is dedicated to IL-17 production (Fig. 5c). Receptors for four counter-regulators or modulators of IL-17 production were divergently expressed upon maturation, with those encoding IL-2 and IL-27 expressed higher in matV2 cells, whereas Il17rb (part of the IL-25R) and Il12rb2 (part of the IL-12R) were expressed at 1/10 and the amounts observed, respectively, in immature cells (Fig. 5b, Supplementary Table 11). Thus, matV2 cells are programmed to respond to autocrine IL-17 turned on by paracrine IL-23, with multiple additional receptors for fine-tuning the response. V1 cells mainly upregulated Il18r upon maturation and this pattern was shared among the mature γδ T cell subsets. Il17rb was selectively increased in expression in V6 cells, while depressed in V1 and V2 cells, upon maturation (Fig. 5b, Supplementary Table 11). IL-25 drives the expansion of Th2 cytokine (IL-5 and IL-13) producing innate lymphocytes, includingαβNKT and TCR− Type II cells 34. The selective induction of Il17rb, which encodes part of the IL-25R, on IL-4 producing V6 cells indicates that, like Il23r acquisition by matV2 cells, the quintessential cytokine responsiveness of γδ effector subsets is intrathymically programmed. Coincident with the overall modulation of cytokine receptor expression as the subsets mature, effector cytokine gene expression was also specifically initiated at the mature stage. The best illustration of this pattern was observed in matV6 cells expressing high levels of PLZF that correlateed with increased Il4 transcription (Supplementary Fig. 5e) and matV2 cells which produced IL-17A (Fig. 5c).
In sum, the unfolding gene programs that dictate γδ subset-specific effector functions occurs in two steps: the establishment phase at the immature stage and the primed phase at the mature stage that equips the thymus-exiting γδ thymocytes with tissue migratory cues as well as sensors that once engaged at tissues sites will rapidly induce effector molecules (see summary model, Supplementary Fig. 6a). The phase transition is accompanied by an overall dampening in the activities of TFs that established the effector gene circuits as γδ thymocytes first arose. This overall pattern of dynamic TF expression program fits well with the established gene regulatory networks controlling cell differentiation in other systems 37.
MatV6 cells were predicted to overlap with αβ iNKT cells at the gene expression level given their shared expression of PLZF and similar functional properties. To determine the extent of similarity, we first derived the αβNKT cell lineage gene signature by identifying genes differentially expressed in thymic iNKT cells versus other mature conventional αβ thymocyte subsets (Fig. 6a, b, Supplementary Table 12). PCA, Euclidian distances, and correlation scores of the 540 iNKT signature genes showed that matV6 cells shared the expression of this signature, as matV6 and iNKT populations were nearly identical intheir placement along PC1 and PC2 (Fig. 6c, Supplementary Fig. 6b, c). This indicates that the acquisition of the iNKT gene signature is independent of TCR type (αβ versus γδ). Genetic studies also supported the relatedness of V6 and αβ iNKT cells: in vivo, the loss of Zbtb16, Sh2d1a (encodes for SAP), Itk, Id3, Ctsl (Cathepsin L), and Cd74 (the invariant chain) negatively impact the development of αβ iNKT cells 38. The same mutations also specifically impacted V6 cells, but not all led to decreased production of V6 cells (39, 40 and Fig. 6d, e). Together, these results indicate that common gene networks and cellular processes regulate the generation of γδ V6 and αβ iNKT cells. Hence, it is either the molecular programming by similar TF networks or a similarity in TCR signaling features independent of TCR type that primarily specifies effector cell fate.
Systematic global gene expression profiling of ex vivoγδTCR+ thymocyte subsets distinguished by TCRγ/δ repertoire demonstrated that the precocious divergence in gene expression programs of emergent γδ cell subsets underpins innate T cell effector diversification. We have identified three distinct cell types in the γδ T cell lineage based on TCR repertoire and shared gene expression profiles: Tγδ17 cells consisting of V2 and fetal V4 cells; fetal V3 cells 17, 18; and adult V1/V5/V6 cells. The only two remaining γδ T cell subsets based on TCR repertoire in adult thymus, Vγ1.2+ and Vγ4+ cells that make up the residual ~5% of total γδTCR+ thymocytes, could not be directly examined as antibodies to sort these cells are lacking. However, analysis of adult immature Vγ1.1, Vγ2 and Vγ5 negative γδ thymocytes (consisting of ~85% Vγ1.2+ and 15% Vγ4+ cells, estimates based on the observed functional Tcrg gene rearrangement frequencies 41) suggests that immVγ1.2+ cells resemble immV1/V5/V6 thymocytes in gene expression (data not shown).
The biological processes governing the generation of these three distinct immature γδ cell subsets are unknown, but the list of possibilities can now be narrowed down. αβ TCR+ thymocytes do not appreciably impact γδ T effector subset differentiation in trans (Narayan et al., manuscript submitted). The most striking gene modules distinguishing the emergent immature γδ thymocyte subsets consist of TFs. Distinct responsiveness of precursors to cytokines and growth factors that control the expression of the TFs defining effector subtypesis one possible way to generate programmed effectors. The HMG TF SOX13, its interacting partners TCF1-LEF1, and their gene targets are one central gene circuit involved in adult Tγδ17 cell differentiation, as Sox13−/− mice lack V2 Tγδ17 cells (Malhotra et al., manuscript in preparation). Given the modulatory effect of WNT signaling on TCF1-LEF1, canonical WNT ligands are another potential regulator of γδ effector diversification. Alternatively, genomically encoded TCR Vγ chains may transmit distinct signals, either by unique interactions with selecting ligand(s) 13 or based on intrinsic differences in TCR conformation and/or membrane localization, independent of selecting ligands presented in trans 42. For αβ T cells, the preTCR-mediated transition to the immDP stage occurs in a ligand-independent manner. Similarly, there is an in vitro precedent that the Vγ2Vδ5TCR can also signal in a ligand-independent, TCR dimerization-dependent manner 13. Co-segregation of V2 cells with αβ lineage DP cells for the gene cluster involved in cellular metabolism support unique TCR signaling property as one potential discriminator of effector subset diversification.
Lastly, it is possible that the distinct timing of generation different VγTCRs during precursor maturation leads to alternate cell fates by inheritance and fixation of distinct precursor gene expression programs. The most clear-cut example of this concept is the programmed Vγ3 gene rearrangement exclusively in early fetal T cell progenitors that generates DETCs 18. It has also been well documented that the timing of TCR expression during precursor maturation, rather than TCR type, is critical for defining resultant T cell properties 43, 44. That γδ T cell subsets may originate at different stages of precursor maturation is supported by several studies indicating that the three functional Cγ loci are regulated distinctly and independently 45, 46. Most directly, we have found that the earliest c-Kit+ T cell progenitors were skewed toward the generation of CCR6+ V2 cells in OP9-DL1 cultures (Malhotra et al., manuscript in preparation), suggesting that the differing onset of functional TCRγ chain expression with the attendant inheritance of distinct collective gene activities of the precursors may underpin the molecular heterogeneity of γδ subsets.
Identification of gene networks specifying effector functions of innate γδ T cell subsets established at the earliest stage of γδ thymocyte differentiation and fixed during thymic maturation suggests that γδ T cells are unlikely to respond homogeneously to external cues. Discovery of the upstream regulators of these gene networks will identify the mechanism of thymic programming of innate γδ effector subset diversification. Moreover, developmentally hard-wired programs to produce IL-17 family cytokines by other recently discovered innate lymphocyte subsets in the gut are predicted to utilize genetic networks similar to those operational in Tγδ17 cells 34, 47, providing a possible route to the identification of a unified origin of innate lymphoid effector subsets.
5 week-old male C57BL/6J mice (Jackson Labs) were used for microarray analysis within 1 week of arrival. Ccr10-Gfp (N. Xiong, Penn State) and Il4-Gfp reporter (M. Mohrs, Trudeau Institute) mice have been described previously. Axin2 (H. Birchmeier, MDC Berlin), Cenpk (SOLT, IGTC), and Sox5 (V. Lefebvre, Cleveland Clinic) LacZ reporter mice have been published. Ctsl−/− mice were provided by Kenneth Rock (UMMS, Worcester, MA) and Cd74−/ − mice were provided by Eric Huseby (UMMS, Worcester, MA). Mice were housed in a specific pathogen free rodent barrier facility. Experiments were approved by the UMMS IACUC (Worcester, MA).
Pooled thymocytes from 4–30 mice were enriched for γδ T cells by depletion of CD8+ cells using magnetic beads and an autoMACS, stained, and sorted directly into Trizol (Invitorgen) using a FACSAria (~2–3 × 104 cells, >99% pure). Independent triplicates were sorted unless noted otherwise (complete sorting details at Immgen.org). As antibodies to Vγ4 are not available, fetal immV4 cells were sorted by gating on TCRδ+ cells that were Vγ1.1−Vγ2−Vγ3−Vγ5− (estimated purity ~95%). Approximate frequencies of TCRδ chains associated with sorted TCRγ subsets are as follows: V2 (Vγ2+, 50% Vδ4+, 40% Vδ5+, all Vδ6.3−, ~45% of total γδ cells); V1 (Vγ1.1+, diverse Vδs including 25% Vδ4+, 15% Vδ5+, others at lower frequencies, all Vδ6.3−, 30% of total γδ cells); V6 (Vγ1.1+, 100% Vδ6.3+, ~8% of total γδ cells); V5 (Vγ5+, 40% Vδ5+ and various others at lower frequencies, ~5% of total γδ cells); and fetal V3 and V4 thymocytes co-express Vδ1.
RNA processing and microarray analysis using the Affymetrix MoGene 1.0 ST array was performed at the ImmGen processing center (SOP at Immgen.org). Data analysis was performed using GenePattern (Genepattern.org) analysis modules. Raw data were RMA normalized with quantile normalization and background correction (ExpressionFileCreator). The ConsolidateProbeSets module (Scott Davis, Harvard Medical School, Boston, MA) was used to consolidate multiple probe sets into a single mean probe set value for each gene. Identification of differentially regulated genes was performed using Multiplot. Unless otherwise indicated, genes were considered differentially regulated if they differed in expression by more than 2-fold, had a coefficient of variation (cv) among replicates of less than 0.5, had a t-test P value of less than 0.05, and had a mean expression value (MEV) of greater than 120 in at least one subset in the comparison. Heatmaps were generated by hierarchical clustering (HierarchicalClustering module) of data based on gene (row) and subset (column) using the Pearson correlation for distance measurement. Data were log transformed and clustered using pair-wise complete linkage. Data were row centered prior to visualization using the HeatmapViewer module. Principal component analysis was performed using the PopulationDistances PCA program (Scott Davis, Harvard Medical School, Boston, MA). Where indicated, the PCA program was used to identify the 15% most differentially expressed genes among subsets by filtering based on a variation of ANOVA analysis using the geometric standard deviation of populations to weight genes that vary in multiple populations. Data were log transformed, gene and subset normalized, and filtered for genes that had a MEV>120 prior to visualization. Euclidian distance and Pearson’s correlation coefficients were calculated using the “dist” and “cor” commands in R to generate distance and correlation matrixes for a given set of genes and subsets. Heatmaps of Euclidian distance and Pearson’s correlation coefficients were generated by hierarchical clustering (HierarchicalClustering module, Pearson’s correlation for row and column, complete-linkage). Data were visualized using a global color scheme in the HeatmapViewer module. Pathway analysis was performed using Ingenuity software (Ingenuity.com) and by manual inspection. Some functional classifications were performed using AmiGO (Amigo.geneontology.org) and KEGG pathways (Genome.kp.kegg).
The following cell surface antibodies were purchased from BD Biosciences: CD25 (PC61), CD27 (LG.3A10), Vγ2 (UC3-10A6), Vγ3 (536), Vδ6.3 (8F4H7B7), Ly49C/I (5E6), CCR6 (140706), Thy1.2 (53-2.1), CD9 (KMC8), streptavidin APC; eBioscience: CD3 (145-2C11), CD4 (RM4-5), CD8 (53-6.7), CD44 (IM7), CD122 (5H4), CD127 (IL-7Rα A7R34), CD24 (HSA, M1/69), c-Kit (2B8), TCRδ (GL3), NKG2D (CX5), CD62L (MEL-15), PD1 (J43), CD73 (TY/11.8), CD199 (CCR9, CW-1.2), CD197 (CCR7, 4B12), CD183 (CXCR3, 174), CD184 (CXCR4, 2B11), CD244.2 (2B4, 244F4), NKG2ACE (20D5), GITR (DTA-1), CD48 (HM48-1), CD28 (37.51), streptavidin PE-Cy7; or other vendors: IL-6R (D7715A7, BioLegend), CD119 (IFNγrα, RDI/Fitzgerald Industries). Vγ1.1 Ab was purified by Bio-XCell and biotinylated using the FluoReporter Mini-Biotin-XX Labeling Kit (Invitrogen). Vδ1 (17D1) antibody was provided by A.Hayday (King’s College, London). Intracellular proteins (IL-17, 17B7, eBioscience; IFN-γ, XMG1.2, BD Biosciences) were detected using the Cytofix/Cytoperm kit (BD Biosciences). Intranuclear staining was performed using the FOXP3 staining kit (eBioscience) for the following antibodies: TOX (TXRX10), EOMES (Dan11mag), RORγt (AFKJS-9), GATA-3 (L50-823), from eBioscience; BLK (cat# 3262), LEF-1 (C12A5), from Cell Signaling; Ki-67 from BD Biosciences; PLZF (D-9) and SMO (N-19), from Santa Cruz. FDG (Invitrogen) staining was performed according to standard protocols. Data were acquired on a BD LSRII cytometer and analyzed using FlowJo (Treestar). When indicated, data were concatenated from multiple independent samples in FlowJo for visualization in histograms.
Ex vivo stimulations were performed by culturing total cells (2×106/well) with PMA (10ng/ml) and Ionomycin (1ug/ml) for 4 hours at 37°C, with Golgi Stop and Golgi Plug (BD Biosciences) added after 1 hour, according to the manufacturers’ protocols.
WT Mtb bacterial strains were used to infect 4–10 male 6–8 week old C57BL/6 mice (Jackson Labs) with 2×103 CFU of Mtb by the aerosol route. Two weeks post infection, mice were sacrificed and cells were isolated from the spleen and lungs (enzymatic digestion) and stained immediately, or after culture with PMA and Ionomycin.
For the identification of differentially expressed genes, t-test P values were generated using Multiplot (Genepattern). For statistical analysis of flow cytometry data, Prism (GraphPad Software) was used. Data were tested for normality using F tests and then analyzed using unpaired two-tailed t-tests. Pathway analysis was performed using Ingenuity Pathway Analysis and statistical significance was determined using the program’s built-in Fisher’s exact test.
We thank the members of the ImmGen Consortium for discussions, the ImmGen core team (M. Painter, J. Ericson, S. Davis) for help with data generation and processing, and eBioscience, Affymetrix, and Expression Analysis for support of the ImmGen Project. Core resources supported by the Diabetes Endocrinology Research Center grant DK32520 were used. Supportedby R24 AI072073 from NIH/NIAID to the ImmGen group, and CA100382 to J.K.
Natasha Asinovski, Angelique Bellemare-Pelletier, Christophe Benoist, Adam Best, Michael Bevan, Natalie Bezman, David Blair, Harald von Boehmer, Milena Bogunovic, Patrick Brennan, Michael Brenner, Andrew Chow, Nadia Cohen, Jim Collins, James Costello, Scott Davis, Michael Dustin, Kutlu Elpek, Ayla Ergun, Jeff Ericson, Anne Fletcher, Emmanuel Gautier, Roi Gazit, Ananda Goldrath, Daniel Gray, Melanie Greter, Richard Hardy, Daigo Hashimoto, Kimie Hattori, Julie Helft, Tracy Heng, Jonathan Hill, Gordon Hyatt, Claudia Jakubzick, Daniel Jepson, Radu Jianu, Vladimir Jojic, Joonsoo Kang, Charlie Kim, Jamie Knell, Francis Kim, Daphne Koller, Taras Kreslavsky, David Laidlaw, Lewis Lanier, Catherine Laplace, Hu Li, Deepali Malhotra, Diane Mathis, Miriam Merad, Jennifer Miller, Paul Monach, Kavitha Narayan, Adriana Ortiz-Lopez, Henry Paik, Michio Painter, Jeremy Price, Gwendalyn Randolph, Aviv Regev, Derrick Rossi, Ravi Sachidanandam, Tal Shay, Susan Shinton, Joseph Sun, Katelyn Sylvia, Nageswara Tata, Shannon Turley, Amy Wagers, Ei Wakamatsu, Linda Wakim, Yan Zhou.
Author contributionsK.S. sorted cell subsets; K.N., N.M., K.S., C.Y. performed follow-up experiments and analyzed data; G.M., T.V., K.N. carried out M.Tb studies; H.K. supervised M.Tb studies; N.X. provided reagents and cells from a mutant strain; N.C. and M.B. generated NKT subset gene expression profiles; L.B. provided reagents and shared data used in interpretation of some results; K.N. analyzed gene expression data; J.K., K.N. designed studies and wrote the paper.
GEO: microarray data GSE 15907.