Na Xiong Development and selection of gd David H. Raulet Tcells

Authors’ address Summary: Two main lineages of T cells develop in the thymus: those that Na Xiong*, David H. Raulet express the ab T-cell receptor (TCR) and those that express the gd TCR. Department of Molecular and Cell Biology and Whereas the development, selection, and peripheral localization of newly Cancer Research Laboratory, University of differentiated ab T cells are understood in some detail, these processes are California, Berkeley, CA, USA. less well characterized in gd T cells. This review describes research carried out in this laboratory and others, which addresses several key aspects of *Present address: Center of Molecular Immunology and gd T-cell development, including the decision of precursor cells to Infectious Diseases, Department of Veterinary and Biomedical differentiate into the gd versus ab lineage, the ordered differentiation over Sciences, The Pennsylvania State University, 115 Henning the course of ontogeny of functional gd T-cell subsets expressing distinct Building, University Park, PA 16802, USA. TCR structures, programming of ordered Vg gene rearrangement in the thymus, including a molecular switch that ensures appropriate Vg Correspondence to: rearrangements at the appropriate stage of development, positive selection David H. Raulet in the thymus of gd T cells destined for the epidermis, and the acquisition Department of Molecular and Cell Biology and Cancer by developing gd T cells of cues that determine their correct localization Research Laboratory in the periphery. This research suggests a coordination of molecularly 485 LSA programmed events and cellular selection, which enables specialization of University of California, Berkeley the thymus for production of distinct T-cell subsets at different stages of Berkeley, CA 94720-3200, USA development. Tel.: þ1 510 642 9521 Fax: þ1 510 642 1443 Keywords: gdT cells, T-cell development, T-cell receptors, T-cell receptor gene E-mail: [email protected] rearrangement, tissue localization

Introduction

Unlike conventional ab T cells, which reside primarily in secondary lymphoid organs and play a central role in adaptive immune responses, many gd T cells reside in epithelial layers of tissues underlying internal and external surfaces of the body, such as the skin, intestinal epithelium, lung, and tongue, where they function as a first line of defense (1–4). In these locations, the diversity of the repertoire of T-cell receptor (TCR) genes is much more limited than that observed in ab T cells or gd T cells that reside in secondary lymphoid organs (3). Furthermore, the

Immunological Reviews 2007 repertoire of expressed variable region genes differs strikingly Vol. 215: 15–31 in the various anatomical locations. It is believed that the highly Printed in Singapore. All rights reserved restricted TCRs expressed by different subsets of gd T cells ª 2007 The Authors enable them to recognize ligands that are specifically expressed Journal compilation ª 2007 Blackwell Munksgaard Immunological Reviews in infected, diseased, or stressed cells in those anatomical sites 0105-2896 (3, 4). Early T-cell development in the thymus is characterized

15 Xiong & Raulet gd T-cell development and selection

by progressive waves of differentiation of distinct gd subsets Overview of gd T-cell development in mice characterized by expression of specific Vg and Vd gene Role of TCR expression in commitment of thymocytes to segments. The underlying processes presumably evolved to differentiate into ab T cells versus gd T cells produce functionally distinct sets of gd T cells in an organized gd T cells and ab T cells arise from a common progenitor cell fashion. in the thymus. The molecular events leading to the lineage A major focus of our research on gd T cells over the years has decision of developing CD4CD8 thymocytes to differentiate been the molecular and cellular processes that dictate the highly into gd T cells versus ab cells have not been fully resolved. It is organized differentiation of gd cells and gd subsets in mice. clear that the decision does not reflect a specialization in TCR Therefore, we have addressed the regulation of Vg gene gene rearrangements, such that one subset of cells rearranges rearrangements in one of the clusters of murine TCRg genes, the TCR g and d genes while another rearranges TCRb genes, the Cg1 cluster (Fig. 1). All the V genes in this cluster rearrange to latter event being necessary for expression of the pre-TCR that is a single Jg gene segment (Jg1), yet occur in a strikingly ordered important for early stages of ab T-cell development. Indeed, fashion over the course of ontogeny (Fig. 1). Thus, this cluster studies show that complete rearrangements of TCR g, d, and includes Vg genes expressed by cells that arise at the fetal stage b genes are present in both mature ab and gd T cells. (Vg3 and Vg4) and at the adult stage (Vg2andVg5), using the Furthermore, complete rearrangements (V–D–J or V–J) occur at nomenclature described by Garman et al. (5). We chose this about the same time in T-cell development (6, 7). cluster because its complex developmental regulation was Most TCR gene rearrangements result in joining of the V and accompanied by a compact size (approximately 50 kb) and the J segments in the incorrect translational reading frame and relatively simple composition (four Vg genes and one Jg and Cg are therefore non-productive, so the fact that a cell makes gene), suggesting that it would be amenable to molecular a particular type of rearrangement is not an indication that it can analysis. In this review article, we describe findings from our ever express a corresponding TCR chain. Approximately 33% of laboratory and others, which have led to a greater understand- complete rearrangements are productive. The randomness ing of many of these underlying processes. In addition, another underlying the creation of a productive TCR gene rearrange- focus of the laboratory has been the decision made by precursor ment has been envisaged as one plausible determinant of cells at the adult stage of T-cell development to differentiate lineage in this instance. Indeed, we and others reported along the ab versus gd lineage. Overall, the findings show that evidence that developing T cells that express a functional gd programmed differentiation events and cellular selection pro- TCR are generally excluded from the ab lineage (8–10). cesses work together in a systematic fashion to produce Whereas these data are at least partially consistent with the functional gd (and ab) T-cell sets. notion that the expressed TCR dictates the alternative lineage fate decision of the cell (ab versus gd), they do not prove this hypothesis. The data are also consistent with the possibility that the expressed TCR simply enables the survival and further differentiation of cells that already made a lineage decision consistent with the identity of the expressed TCR. For example, progenitor cells may differentiate into ab-orgd-committed populations before TCR gene rearrangement occurs in devel- opment. Further successful differentiation may only occur in the subset of each population that subsequently succeeds in productively rearranging and expressing a TCR that matches its predetermined lineage. A study from our group provided evidence for the notion that lineage determination to some extent is independent Fig. 1. Organization of mouse TCRg genes and cis-acting regulatory elements, the developmental pattern of Vg gene rearrangements, of the type of TCR expressed (11). At the double-negative and the peripheral localization of specific subsets of gd T cells 2 (DN2) stage of thymocyte development (phenotype defined by Vg expression. The locus consists of three functional CD3CD4CD8CD25þCD44hi), at which stage complete clusters (g1, g2, and g4) and a non-functional one (g3). The g1 cluster, explored in this review in detail, is expanded to show specific gene TCR gene rearrangements are not yet made in the great majority segments and regulatory elements (2, 4, 16). of cells, cells that express high or low levels of the interleukin-7

16 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

receptor (IL-7R) on the cell surface can be distinguished. The different subsets of gd T cells arise in the thymus at Analysis of the developmental potential of these cells, based on different stages of ontogeny. Vg3þ T cells are exclusively intrathymic injection experiments, showed that the IL-7Rhi generated in the early fetal thymus, where they are the first T DN2 cells were more likely to differentiate into gd T cells, cells detected in ontogeny, around day 13 of gestation. From the whereas IL-7Rlow cells were more likely to differentiate into ab fetal thymus, these gd T cells migrate to the epidermal T cells (11). These data suggested that some DN2 cells exhibit epithelium, where they expand locally to reach adult numbers a bias in their lineage potential at a stage before TCRs are ever and are normally maintained there for the life of the animal (18, expressed on the cell surface. Recent clonal analysis of the 24). Vg4þ T cells are also produced in the early fetal thymus and potential of DN2 cells in a Notch ligand-dependent cell culture they migrate to their epithelial destinations, such as the system arrived at a similar conclusion (12). reproductive tract, tongue, and lung. Vg3 and Vg4 TCRs Other recent studies have focused on whether lineage display a high frequency of invariant canonical junctional determination might result from differences in the quality or sequences in the g chains due, in large part, to the absence in strength of the signal propagated by the pre-TCR versus the gd fetal thymocytes of terminal deoxynucleotidyl transferase TCR (13–15). These studies provided strong evidence that the (TdT), which is necessary for the addition of N-nucleotides at strength of signal delivered by the TCR in developing T cells has V–J junctions, coupled with a preference for rearrangement at a major impact on the outcome of the process. The gd TCR the sites of microhomologies at the ends of the Vg and Jg gene appears to provide stronger signals that promote gd TCR segments (25, 26). Similar mechanisms have been invoked development, whereas weaker signals associated with pre- to account for canonical junctional sequences in the Vd1 TCR signaling promote the development of ab T cells (14, 15). rearrangements that encode the dominant TCRd chain ex- The effects of TCR signal strength are not necessarily pressed by both Vg3þ and Vg4þ T cells (27). Positive selection incompatible with the aforementioned TCR-independent bias during development in the thymus may also play a role in in lineage potential. Effective rescue of gd-committed pre- allowing only those cells with the canonical junctional cursors may simply require a stronger TCR signal, whereas sequences to mature. rescue of ab-committed precursors may require a weaker TCR Vg3þ and Vg4þ T cells are not generated in the adult thymus. signal. Alternatively, it was suggested that the bias in lineage Instead, the adult thymus switches to the production of gd T potential, which arises early in DN2 cell differentiation may be cells expressing Vg2, Vg1.1, and Vg1.2 gene segments with reversible in the case of a sufficiently strong instructive signal at highly diverse junctional sequences, which localize to the a later stage in the process (11). Success in identifying the secondary lymphoid organs on exiting the thymus (1, 28–30). molecular signals responsible for lineage determination should Adult Vg5þ cells localize to the intestinal epithelium as well as ultimately resolve this issue in a definitive fashion. secondary lymphoid organs (22). There are also reports suggesting that some intestinal epithelial Vg5þ gd T cells Differentiation of subsets of gd T cells develop extrathymically (31, 32). Many tissue-specific gd T cells express different specific subsets of The sequential generation of specific subsets of gd T cells at TCRs with little or no diversity (2, 4, 16, 17). At one extreme, different stages of ontogeny is a fixed developmental program nearly all gd T cells in the epidermal epithelium express an that cannot be easily reordered. For example, disruption of the identical (‘canonical’) gd TCR composed of Vg3–Jg1Cg1and generation of gd T cells in the early fetal thymus by injecting Vd1–Dd2–Jd2Cd chains with identical junctional sequences anti-gd-TCR antibody into pregnant mothers results in the (18). These skin gd T cells are usually called dendritic epidermal T absence of DETCs in adult mice (3). Furthermore, DETCs cannot cells (DETCs). Similarly, gd T cells in the tongue, lung, and be replaced with bone marrow stem cells in adult mice, reproductive tract epithelium express a canonical Vg4–Jg1Cg1/ following full-body irradiation and reconstitution (33). Vd1–Dd2–Jd2Cd TCR (19). In a less exclusive situation, the epithelium of the small intestine contains both gd and ab Tcells Programmed rearrangement of TCR Vg gene segments (20, 21) and the gd T cells predominantly express Vg5þ and Vg1.1þ Vg chains, but with diverse junctional sequences (22). Evidence that ordered Vg gene rearrangement is Finally, in secondary lymphoid organs, where gd T cells represent a programmed process only a small fraction of total T cells, gd TCRs predominantly The appearance of cells with different Vg gene rearrangements include Vg2, Vg1.1, and Vg1.2 V regions, although with at different stages of development could reflect a developmental extensive, even dramatic, junctional diversity (22, 23). program that controls the rearrangement machinery or could be

Immunological Reviews 215/2007 17 Xiong & Raulet gd T-cell development and selection

because of a cellular selection process that promotes the survival Over the years, we and others have identified and of different gd subsets at different stages of development, or characterized several cis-acting elements in the TCRg Cg1 both. Several lines of evidence showed that developmentally cluster, which play unique and overlapping roles in regulating regulated Vg gene segment recombination is a developmentally the Vg gene rearrangement process as well as transcription and programmed process. Initially, scrutiny of Vg gene rearrange- expression of TCRg genes (Fig. 1). The roles and properties of ments showed that non-productive Vg gene rearrangements in some of these elements are described in more detail and in cell lines were usually of the same type as the productive different contexts later in this review, but their role in ordered rearrangements (16). Because non-productive rearrangements rearrangement is addressed first. The elements include the cannot be subject to cellular selection processes, this finding promoter regions of each Vg gene, two enhancer-like elements suggested that developing gd T cells are programmed to choose (3#-ECg1 and HsA) that cooperate to regulate transcription of specific Vg gene segments for rearrangement. Consistent with the locus, and a putative silencer element that prevents TCRg an intrinsic program of Vg gene rearrangement, we later found expression in ab T cells. that germline transcription of Vg genes correlates with their rearrangement pattern. Germline transcription is usually Vg gene promoter regions determine the rearrangement correlated with greater accessibility of genes in chromatin to pattern in the adult thymus recombinases, suggesting a possible link between gene Numerous studies have shown that the accessibility of accessibility and rearrangement in this system (34). immunoglobulin (Ig) and TCR genes is a critical determinant As a more direct test of whether the ordered pattern of TCRg of rearrangement patterns during development. Accessibility gene rearrangement occurs in the absence of cellular selection, refers to whether chromatin is in a sufficiently open we generated several lines of transgenic mice that harbored configuration for recombinase to access the recombination a 39-kb genomic transgene (called gB) that spans most of the signal sequences (RSSs) that flank the gene. Accessibility usually unrearranged TCRg Cg1 cluster. Each of the three Vg genes in correlates with transcription of the unrearranged genes (germ- the transgene (Vg2, Vg4, and Vg3) contained a frameshift line transcription). As already mentioned, early studies found mutation that prevented functional expression of the corre- a correlation between germline transcription and rearrange- sponding V region (35). Despite the fact that the transgene ment of individual Vg genes in the adult thymus. Specifically, could not encode intact TCRg chains that might influence the germline transcription from the unrearranged Vg3 gene was selection of cells in which they occurred, rearrangements of strongly suppressed at this stage, while germline transcription the Vg genes in the reporter transgene, like rearrangements of from the unrearranged Vg2 gene remained high, suggesting the endogenous Vg genes, appeared at the appropriate stage in greater accessibility of the Vg2 gene. Subsequent studies of thymic development. In a different but related approach, it was adult thymocytes substantiated the conclusion that the Vg2 shown that a targeted deletion of the TCRd gene, which gene is more accessible than the Vg3 gene at this stage, in- abrogates the formation of functional gd TCRs (although not cluding a report that showed a much greater degree of histone protein expression of TCRg genes), did not disrupt the normal acetylation in the vicinity of the Vg2 gene than in the vicinity of developmental pattern of Vg or Vd gene rearrangements (27). the Vg3 gene (36). Furthermore, it was shown that an inhi- These data provided compelling evidence that programmed bitor of histone deacetylation, which is expected to increase rearrangement of Vg genes is a major determinant of the the acetylation of the Vg3 gene and therefore its accessibility, ordered generation of different subsets of gd T cells at different stimulated increased levels of Vg3 rearrangements in the stages of ontogeny. adult thymus (36). Differences in accessibility of the different Vg genes Regulation of programmed Vg gene rearrangement suggested that cis-acting regulatory elements associated with The finding that Vg and Vd rearrangement patterns are the Vg genes might be responsible for controlling Vg gene programmed independent of cellular selection led us to rearrangement at the adult stage. Therefore, we designed investigate the underlying molecular mechanisms. As possible experiments to test whether the promoter segments of these determining factors of the rearrangement patterns, we assessed genes play a role in imposing the adult pattern of rearrange- both the role of cis-acting regulatory elements in the genes and ment. For this purpose, we generated a new transgene based on the location of the V genes within the cluster. Both these factors the gB transgene discussed above. The frameshift mutations in have been proposed or implicated as determinants of rearrange- the Vg gene segments of this construct should prevent ment patterns in studies of other rearranging genes. interference in the outcome of an experiment by cellular

18 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

selection events. The new transgene, called gB-Pr-Sw, was the predominance of Vg3 (and Vg4) rearrangements in the fetal identical to gB except that the promoter segments of the Vg3 thymus. We therefore considered other explanations. and Vg2 genes were reciprocally swapped (35) (Fig. 2). Compared with the gB transgene, the rearrangement patterns of The location of Vg genes in the TCRg locus is a major Vg2 versus Vg3 were reversed in the adult thymus of the gB-Pr- determinant of the rearrangement pattern in the fetal Sw transgenic mice. Consistent with the reversed pattern of thymus rearrangement, the pattern of germline transcription of the Vg2 An obvious candidate as a determinant of the fetal rearrange- and Vg3 genes was also reversed in adult thymocytes of gB-Pr- ment pattern is the location of Vg genes within the TCRCg1 Sw mice, as compared with gB mice. These data showed that locus. Previous studies have documented preferential rear- the promoter regions of the individual Vg genes determine rangement at the fetal stage of 3#- (D- or J-proximal) V genes in the preference for Vg2 rearrangements in the adult thymus, the IgH and TCRd loci, in addition to the comparable findings in probably by controlling the accessibility of the corresponding the TCRg locus already discussed above (5, 29, 37–39). In the Vg genes to the recombinase enzymes. IgH locus, an IL-7-dependent shift in the adult stage to rearrangement of 5#-VH genes has been documented (40, 41). A distinct mechanism for establishing the rearrangement However, other studies indicate that even in the adult stage, pattern in the fetal thymus newly generated B cells show a bias for rearrangement of 3#-VH Our initial expectations were that the rearrangement patterns gene segments (42, 43). Based on transfections of cell lines with in the fetal and adult thymus would reflect the same basic plasmids containing various recombination substrates, it was mechanism and that the switch between these two patterns proposed that the preferential rearrangement of 3#-VH genes in would consist of altered regulation of this basic mechanism. the IgH locus was because of the potency of the RSS associated However, the notion that promoter activity is responsible for with these genes (44). In contrast, another study showed that the fetal rearrangement pattern was rapidly refuted by analysis the rate of recombination observed in vivo in the natural locus did of gB-Pr-Sw mice. Despite the reversed rearrangement pattern not correlate with the rate of recombination of corresponding in the adult thymus of these mice, the exchanged promoter substrates in transfected cell lines, but did correlate with the segments had no significant effect on transgene rearrangements position of the gene in the locus (45). The consistency of these in the fetal thymus, where Vg3 rearrangements predominated latter findings with ours suggested that location of a V gene to a similar extent as in the gB transgenic mice. These results within a V gene array might be a major determinant of indicated that differential promoter activity cannot account for rearrangement at the fetal stage. To directly test the role of Vg gene location in rearrangement, we generated a new transgene in which the entire Vg2 and Vg3 gene segments were swapped within the gB transgene (Fig. 2). The exchanged fragments were approximately 1.6–2.3 kb in length and included the promoter regions, coding regions, and some downstream sequence including the RSS. This transgene, called gB-Gn-Sw, showed a reversed pattern of rearrangement in the fetal thymus compared with gB, such that Vg2 rearrangements were several fold more frequent than Vg3 rearrangements (46). These data clearly showed that the relative location of Vg genes is a dominant factor in determining their rearrangement levels in the fetal thymus. In recent unpublished Fig. 2. Configuration of gB, gB-Pr-Sw, and gB-Gn-Sw transgenes. The transgenes include all contiguous sequence as present in the experiments, we generated gene-targeted mice that have endogenous locus. The configuration of gB is identical to that of the corroborated this conclusion (our unpublished data). endogenous locus. In gB-Pr-Sw, the promoter regions and leader exon How does the position of Vg genes in the locus determine (which encodes the signal peptide) of the Vg2 and Vg3 gene segments were exchanged. In gB-Gn-Sw, 1.6- to 2.3-kb fragments containing the rearrangement levels? One model to consider is that the Vg gene entire Vg3 and Vg2 gene segments were exchanged, including the position somehow determines whether the Vg gene is in an promoters and RSS. The asterisks indicate the sites where frameshift open chromatin configuration and is therefore accessible to the mutations were introduced into each Vg gene segment. [Reprinted with permission from Proceedings of the National Academy of Sciences, recombinase machinery and transcription apparatus. This USA (48).] model is reminiscent of a study showing that proximity of

Immunological Reviews 215/2007 19 Xiong & Raulet gd T-cell development and selection

globin genes to the b globin locus-control region (LCR) determines the extent of globin gene transcription in the early yolk-sac stage of erythropoiesis (47). We have shown, however, that combined deletion of both known enhancers in the Cg1 locus, 3#-ECg1 and HsA did not substantially perturb the fetal or adult patterns of Vg gene rearrangement, nor did it cause a major reduction in the extent of Vg gene rearrangement (48). Therefore, proximity of the V genes to these elements cannot be a necessary determinant of the patterns of rearrange- ment. While it might be proposed that differential accessibility is controlled by proximity to a distinct unidentified regulatory Fig. 3. A model of programmed Vg gene rearrangement. At the fetal element, direct tests showed that gene position did not alter the stage, both the 5#- and the 3#-Vg genes are in an open ‘accessible’ accessibility of the Vg genes at the fetal stage, even as it did alter chromatin configuration. Rearrangement of the 3#-Vg genes (Vg3 and Vg4) is preferred at the fetal stage because of their location in the g the tendency of the V genes to rearrange. Thus, comparing cluster, possibly related to their proximity to Jg1. This location fetal thymocytes from gB-Gn-Sw transgenic mice with those preference may reflect competition of the Vg gene segments for the from gB transgenic mice, we found no significant difference in rearrangement apparatus. At the adult stage, events involving the specific Vg promoters result in suppressed accessibility of Vg3 and Vg4. germline transcription or restriction enzyme accessibility of the Rearrangement of 5#-Vg genes is strongly increased at this stage, Vg2orVg3 genes (46). Consistent with the transgenic studies, possibly because the inaccessible Vg genes can no longer compete with in the endogenous TCRg locus, both genes showed high levels the 5#-gene segments. of germline transcription and histone acetylation in fetal thymocytes (36, our unpublished observations). All these data because of their proximity to the Jg1 gene and, therefore, are are in agreement that both Vg2 and Vg3 genes are ‘open’ and preferentially rearranged over other upstream Vg genes at the accessible in the natural locus at the fetal stage and suggest that fetal stage. In the adult thymus, as a result of the activity of the the regulatory mechanism that dictates differential Vg gene associated promoter elements, the upstream Vg2(andVg5) rearrangement at the fetal stage differs from that in the adult genes are maintained in an open and accessible state, while the stage, where regulated accessibility of the Vg genes appears to accessibility of the downstream Vg3 (and 4) genes is play a dominant role. suppressed. The altered accessibility shifts the balance in favor of rearrangement of the upstream Vg2(andVg5) genes that are Models of regulated Vg gene rearrangement therefore preferentially rearranged in the adult thymus. Two specific mechanisms are under consideration to account for preferential rearrangement of the 3#-Vg gene at the fetal The potential role of transcription factor E2A in regulating stage. One possibility, reminiscent of the one first proposed to Vg gene rearrangement account for preferential rearrangement of 3#-VH genes (37, The role of specific transcription factors in regulating cis-acting 38), is that rearrangement at the Cg1 cluster involves some form elements in the TCRg locus has not been comprehensively of one-dimensional ‘tracking’ mechanism along the DNA by the studied. However, studies have implicated the E2A transcription recombinase, in which downstream Vg genes are encountered factor in the process, specifically in imparting the adult pattern before upstream genes because of their greater proximity to the of Vg gene rearrangement. E2A is a member of the basic helix- Jg gene segment. A second possibility is that the 3#-Vg gene RSS, loop-helix transcription factor family and has been found to being more closely tethered to Jg1 in genomic DNA, is more play various important roles in regulating lymphocyte devel- likely to collide with the Jg1 gene RSS than a 5# Vg gene RSS opment (49). In mice lacking E2A, Vg2 gene rearrangements during three-dimensional diffusion, initiating rearrangement. were decreased in the adult thymus, while Vg3 gene rearrange- Whatever the specific molecular mechanism to account for ments were increased (50). The switch in Vg2 versus Vg3 the fetal pattern of rearrangement is, the available data support rearrangement in E2A knockout mice was associated with the general model depicted in Fig. 3 to account for the a corresponding change in the germline transcription of these ‘developmental switch’ in Vg gene rearrangement that occurs Vg genes (50), suggesting that E2A contributes to the in ontogeny. The model proposes that in the fetal thymus, programmed Vg gene rearrangement pattern at the adult stage the whole TCRg locus is open for rearrangement, but the by differentially regulating the accessibility of the Vg2 versus downstream Vg3(andVg4) genes have a competitive advantage Vg3 genes. Whereas little effect of E2A deficiency on Vg gene

20 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

rearrangement was noted in the fetal thymus at a very late stage those Vg genes in each construct that was flanked by the Vg2 in fetal ontogeny (day 19 of gestation) (50), we observed promoter fragment (Vg2 in the gB-Gn-Sw transgene and Vg3in a substantial defect in rearrangement of both the Vg3and the gB-Pr-Sw transgene) (Table 1, N. Xiong, C. Murre, D. H. the Vg2 genes in the early fetal thymus (day 15), at a time Raulet, unpublished data). Hence, it is plausible that E2A acts when Vg3 rearrangements normally predominate (N. Xiong, directly on the Vg2 promoter to enhance rearrangement at the C. Murre, D. H. Raulet, unpublished data). These findings sug- adult stage, perhaps by maintaining accessibility of this region gest that E2A plays distinct roles in regulating the development of the locus. In contrast, the prediction was not borne out in the of gd T cells at different stages in ontogeny. case of the Vg3 promoter because in both the gB-Gn-Sw and the E2A may interact directly with cis-acting elements in the gB-Pr-Sw transgenes, E2A deficiency caused a decrease rather TCRg locus to regulate rearrangement of Vg genes differentially. than an increase in rearrangement of those Vg genes that were It was reported, for example, that E2A sites are present in the RSS flanked by the Vg3promoter(Table 1,N.Xiong,Murre,D.H.Raulet, of most Vg and Vd gene segments (50), and candidate E2A sites unpublished data). These data suggest that the Vg3 promoter have been identified in the promoter regions of Vg2, Vg3, and region is not a dominant interaction site for E2A regulation of Vg4 genes. The sites in the promoter regions might be more Vg gene rearrangement. In both these transgenes, the Vg3 important, judging from the aforementioned evidence that the promoter was relocalized to the 5# position normally occupied adult pattern of rearrangement is dependent on the promoter by the Vg2 promoter. A possible explanation for the puzzling sequences and not on the Vg gene segments or RSS. However, results with these transgenes is that E2A acts on another site near evidence that E2A acts directly on Vg genes in vivo is lacking. this upstream region to enhance rearrangement in wildtype To help define cis-acting elements through which E2A may cells. A deeper understanding of the role of E2A in regulating act in regulating Vg gene rearrangement in vivo, we crossed E2A this locus will require detailed analysis of the roles of the various knockout mice with gB transgenic mice. In adult thymocytes E2A sites in the context of endogenous genomic DNA. from gB transgenic mice, E2A deficiency resulted in an enhancement in transgene Vg3 rearrangements and a reduction Role of enhancer elements in promoting rearrangement in transgene Vg2 rearrangements, much as it does for the and transcription of TCRg genes endogenous Vg genes (Table 1). These data are consistent with the possibility that E2A interacts with cis-acting elements within Likely redundancy of enhancer elements in regulating TCRg the span of the gB transgene to regulate Vg gene rearrangement. gene rearrangement To address whether E2A regulates rearrangement through Although the local Vg promoters influence the accessibility of interaction with the promoter regions of Vg genes, we crossed the corresponding Vg genes, there is likely to be a higher order the E2A knockout mice with gB-Pr-Sw and gB-Gn-Sw trans- regulation of the chromatin condensation of the TCRg locus by genic mice. The rationale of this analysis was that if the additional cis-acting elements, including the aforementioned promoter regions of Vg genes are the dominant interaction sites enhancer elements HsA and 3#-ECg1. Indeed, deletion of small for E2A to regulate rearrangement of the corresponding Vg DNA segments containing HsA and 3#-ECg1 from a genomic genes, E2A deficiency should alter rearrangement of Vg genes in transgene abrogated rearrangement of the transgene in vivo (51), these transgenes according to the origin of the promoter regions suggesting the roles of HsA and 3#-ECg1 in maintaining the (Vg2orVg3) that flank the Vg gene, not the origin of the Vg chromatin in an open configuration for rearrangement. gene itself. This prediction was borne out in the case of the Vg2 Surprisingly, however, simultaneous deletion of both elements promoter, in which we observed decreased rearrangements of in the endogenous locus by gene targeting only modestly

Table 1. Effect of E2A deficiency on endogenous and transgenic Vg2 and Vg3 gene rearrangements

Endogenous gB gB-Pr-Sw gB-Gn-Sw

Vg2Vg3Vg2Vg3Vg2Vg3Vg2Vg3

Change in E2A/ mice* Y 2–10y [ >3 Y 3–5 [ 3–5 Y 3–5 Y 2–3 Y 3 Y 4–5

*Change in E2A/ mice indicates increased ([ fold change) or decreased (Y fold change) Vg gene rearrangements in E2A/ mice compared with that in E2Aþ/þ littermates. yThe data for endogenous Vg2 rearrangements in E2A/ mice are from Bain et al. (50).

Immunological Reviews 215/2007 21 Xiong & Raulet gd T-cell development and selection

inhibited Vg gene rearrangement, and knocking out either The roles of HsA and 3#-ECg1 in regulating TCRg gene element in isolation had no effect on rearrangement or expression and gd T-cell development were probed with gene- accessibility (48). Considering that rearrangement of other targeted mice lacking one or both of these elements (48). Ig/TCR genes requires enhancer elements (52), it appears likely Although germline deletion of HsA and 3#-ECg1 only impaired that an unidentified cis-acting element(s) necessary for Vg gene rearrangement modestly (by twofold), the numbers of recombination exists in the endogenous Cg1 locus, separate Vg2þ and Vg3þ (and presumably Vg4þ) gd T cells in the from the segment that comprised our transgene. However, the thymus were dramatically decreased by more than 20-fold. putative additional regulatory element(s) cannot, by itself, Impaired development of these subsets was correlated with support transcription of most of these rearranged Vg genes a dramatic reduction in the transcription of the TCRg genes because transcription of rearranged Vg2, Vg3, and Vg4 genes located between HsA and 3#-ECg1, including Vg2, Vg4, and was nearly abolished in the HsA/3#-ECg1 double knockout Vg3 genes, whereas the upstream Vg5 and Vg genes in other mice. Thus, whereas HsA and 3#-ECg1 together possess clusters were not substantially affected. Interestingly, separate properties of an LCR and are capable of maintaining the deletions of HsA or 3#-ECg1 had little or no effect on genomic accessibility of an ectopically integrated TCRg trans- transcription of TCRg genes and development of gd T cells gene, they probably function redundantly with other cis-acting in the thymus. This study, along with a study of enhancer regulatory elements in the endogenous locus to promote elements in the Igk locus, provided the first direct evidence genomic accessibility for gene rearrangement. of redundantly functioning enhancer elements in regulating There is an indication that the putative additional regulatory transcription of an antigen-receptor gene (48). It is noteworthy element(s) that promotes recombination in the Cg1 cluster may that although expression of rearranged Vg2–4 genes was highly reside upstream of HsA. In gene-targeted mice in which HsA impaired in the double knockout mice, a few T cells expressing was replaced by a PGK-neo gene cassette, development of Vg2þ Vg2 and Vg3 at normal levels were detectable in these mice, cells was severely reduced (data not shown), whereas deletion suggesting that while initiation of gene expression is greatly of PGK-neo resulted in restored numbers of Vg2þ cells. In reduced in the absence of these elements, a few cells that contrast, replacement of 3#-ECg1 with PGK-neo did not succeed in initiating gene expression can express adequate significantly affect gd T-cell development (data not shown). levels of the corresponding mRNAs. Because insertion of a PGK-neo gene between a cis-acting element and the gene it regulates often inhibits the function of the A unique function of HsA in upregulating TCRg gene element (53), these data may imply the existence of an expression in peripheral Vg2þ gd T cells unidentified regulatory element(s) upstream of HsA, which Although separate deletions of HsA or 3#-ECg1 did not impair plays a role in Vg gene recombination and/or transcription. expression of TCRg genes in the thymus or thymic development These results together with the results described earlier in this of gd T cells, deleting HsA resulted in a significant defect in the review suggest that V–J rearrangement in the Cg1 cluster is periphery. The number of Vg2þ gd T cells in the spleen or regulated by multiple cis-acting regulatory elements that lymph nodes of HsA knockout mice was reduced by a factor of 4 function in a hierarchical complex fashion, in which some when compared with wildtype mice (48). No such defect was elements regulate larger segments of the cluster while others evident in 3#-ECg1 knockout mice, suggesting a unique role for play a fine tuning role by regulating accessibility of individual HsA at the post-thymic stage of gd T-cell development. Vg genes. Analysis of the cell surface phenotype of gd T cells provided a possible clue as to the cause of the depressed Vg2þ gd T-cell Redundant roles of HsA and 3#-ECg1 in regulating TCRg compartment in the periphery of HsA knockout mice. Most gene transcription and gd T-celldevelopment in the thymus peripheral Vg2þ gd T cells in wildtype mice have probably been As expected, in addition to their roles in regulating Vg gene previously exposed to antigen, as they exhibit a memory rearrangement, cis-acting elements in the Cg1 cluster are also phenotype characterized by expression of CD44 but not CD62L important in controlling expression of the rearranged TCRg (54) (Fig. 4). In contrast, the majority of residual Vg2þ T cells in genes and development of corresponding gd T cells. The various HsA knockout mice were CD44low cells, suggesting that elements contribute differentially to rearrangement versus memory Vg2þ T cells either fail to form or fail to survive in transcription. In particular, although the HsA and 3#-ECg1 the absence of HsA. As a possible explanation for this defect, it enhancer elements are not required for rearrangement in the may be significant that in wildtype mice, the TCR levels were endogenous locus, they play critical roles in transcription. elevated by approximately threefold on CD44þVg2þ T cells as

22 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

Selection of gd T cells in the adult thymus The necessity for class I and class II major histocompatibility complex (MHC) molecules in the positive selection of ab T cells in the thymus was clearly shown by the absence of these cells in mice lacking MHC class I and/or class II molecules (55, 56). In contrast, the absence of MHC class I and class II molecules had no observable effect on the development of gd T cells in the thymus (56–58). These data suggested that most gd T cells are not selected by classical MHC molecules or by non-classical class I molecules that depend on b2-microglobulin (b2m) for expression on the cell surface. However, it remained possible

þ gd Fig. 4. Role of HsA in peripheral development of Vg2 cells. (A) that MHC does play a role in selection of a subset of T cells. Reduced percentage of CD44þ cells or CD62L cells among Vg2þ Furthermore, the natural ligands for gd T cells are in most cases / þ lymph node cells in HsA mice. Gated Vg2 cells are shown. (B) not known and it remains possible that many or all the cells are Mean percentages ( standard deviation) of CD44þ cells among Vg2þ cells in lymph nodes of wildtype versus HsA/ mice (n > 5). The total positively selected by interactions with unidentified ligands in number of lymph node Vg2þ cells was also reduced (by 4- to 5-fold) in the thymus. / the HsA mice. Because of limited knowledge of the natural ligands for gd TCRs, many studies designed to examine whether developing þ compared with CD44 Vg2 T cells, suggesting that formation gd T cells undergo selection have used the G8 and KN6 þ of Vg2 memory cells is associated with increased levels of transgenic mouse strains. The two transgenic strains harbor þ Vg2 TCRs. Memory ab T cells usually show depressed TCR different gd TCRs, although both include a Vg2 chain. The two levels. In HsA knockout mice, gd TCR levels (and transcripts) TCRs are specific for the same ligands, the T10 and T22 non- þ were relatively low on residual Vg2 T cells, consistent with the classical MHC class I molecules (59–62). T10 and T22 are þ possibility that upregulation of the Vg2 TCR on memory cells encoded in the TL region of the MHC and are expressed highly requires the action of HsA (48). We propose that TCR in B6 mice (H-2b haplotype) but only weakly in BALB/c mice upregulation and consequent increased TCR signaling may be (H-2d haplotype) because of defective T22 gene expression in þ necessary for differentiation and/or maintenance of Vg2 the latter strain (59). Studies with T22 tetramers show that memory cells. a small subset of naturally arising mature peripheral gd T cells in normal mice is specific for these same ligands (63). Thymic selection and development of tissue-specific Early studies with G8 and KN6 transgenic mice reported that gd T cells the transgenic gd T cells underwent negative selection in mice The findings discussed above have documented the role of that express strong ligands for the receptors, such as the B6 programmed Vg gene rearrangement in the sequential develop- strain, but not in mice that express a weak ligand, such as the ment of gd subtypes during ontogeny. In principle, programmed BALB/c strain (64, 65). It has been much more controversial Vg (and Vd) rearrangement might be a sufficient explanation for whether interactions of these transgenic gd T cells with ‘weak’ the sequential appearance of gd T-cell subsets, but it has long been ligands are necessary for positive selection of the cells. Several proposed that cellular selection also plays a role in the process. studies suggested that positive selection was necessary, based on Findings have been somewhat contradictory on this point. A the finding that transgenic gd T cells developed much less related issue concerns how the appearance of gd T-cell subsets in efficiently in b2m-deficient mice, which lack surface expres- the thymus is coordinated with the localization of these cells to sion of T10 and T22, than in BALB/c mice, which express distinct peripheral sites and, potentially, to the acquisition of a weak ligand (14, 66, 67). Another study attributed much of different functional activities. Recent studies provide compelling this difference to negative selection, because of variations in the evidence that thymic selection plays a key role in the development genetic backgrounds of the mice under study (68). On top of of at least some gd T-cell subsets in the periphery, especially the this contradiction, a study that used T22 tetramers to directly invariant gd T-cell subsets that arise in the fetal thymus and in assay the development of natural T10/T22-specific gd T cells in localization of these gd T cells to their preferred home in the non-transgenic mice concluded that development of these cells periphery. Studies examining selection in the adult versus fetal is not impaired in b2m-deficient mice (K. Jensen & Y. Chien, thymus are reviewed separately. personal communication). In lieu of positive selection, these

Immunological Reviews 215/2007 23 Xiong & Raulet gd T-cell development and selection

authors attribute the relatively high frequency of T10/T22- ‘idiotypic’ marker detected with the 17D1 monoclonal reactive gd T cells in normal mice to programmed rearrange- antibody, a marker that is also found on Vg3Vd1 TCRs (72). ment mechanisms that assemble a reactive TCR in a small but However, this marker was absent on the DETCs that arose in Vd1 significant population of developing gd T cells. Given the knockout mice, suggesting that 17D1 reactivity might be variability in experimental outcomes and the possibility that associated with Vd1-containing TCRs rather than with speci- these TCRs may be positively selected by an unknown b2m- ficity of the TCR (73). A more compelling suggestion that the independent ligand, it remains uncertain whether positive DETCs in Vg3 knockout mice had undergone selection was the selection is required for the development of T10/T22-reactive finding that these cells produced IL-2 when stimulated with gd T cells in the adult thymus. keratinocyte cell lines, a specificity previously shown to be associated with DETCs, but not other subsets of gd T cells (72). Selection of gd T cells in the fetal thymus As an approach to investigate the requirement for thymic Evidence for thymic positive selection of DETC precursors selection of gd T cells in the early fetal thymus, especially the The previous studies were consistent with the selection of DETC Vg3þ DETC population, several studies have investigated mice precursors, but did not distinguish whether the selection equipped with various gd TCR transgenes non-native to the occurred in the skin itself or was the result of selection in the skin. The underlying assumption of most of these studies was thymus. We recently used gene targeting to reinvestigate the that the transgenic TCR would exclude expression of TCRs requirement for specific Vg chains in the development of normally found in the fetal thymus, especially the canonical DETCs. The results provided compelling evidence for selection Vg3Vd1 TCR found on most DETCs. Positive selection of DETCs, of DETCs and clear evidence that it occurs in the thymus, and at if necessary, should be aborted if the transgenic TCR had the the same time these studies showed a possible mechanism to inappropriate specificity. It was shown that gd T cells expressing couple thymic selection with emigration of the T cells to the ‘inappropriate’ transgenic TCRs were able to develop normally epidermis (74). into DETCs in the skin, suggesting that DETC development is not We generated mice with a large deletion of the Cg1 cluster dependent on expression of Vg3orVd1 and therefore is not that encompassed HsA, Vg2, Vg4, Vg3, Jg1, Cg1, and 3#-ECg1 dependent on positive thymic selection by ligands that are (234JCg1/ mice). The remaining Cg clusters remained intact, specifically recognized only by the Vg3Vd1 DETC TCR (69, 70). and Vg1.1þ and probably Vg1.2þ T cells developed normally in The assumption that the transgenic TCR necessarily excludes the thymus of the knockout mice, where they were first expression of endogenous TCRs was later proven incorrect, detected at days 15 and 16 of gestation. Normal numbers of gd T however, raising doubts about the earlier conclusions. For cells, most of them Vg1.1þ cells, were detected in the spleen example, whereas plentiful transgene-expressing skin-resident and intestines of these mice, but Vg2þ,Vg3þ, and Vg5þ T cells gd T cells developed in Vd6.3 transgenic mice, none developed were absent, as expected. Remarkably, despite the abundance of when the transgene was crossed onto a TCRd/ background in gd T cells in other peripheral sites and in the fetal thymus, there which endogenous TCRd chains cannot be expressed (71). were virtually no TCRgdþ DETCs in the epidermis of these mice These data suggested that development of the skin-resident gd T (74). Therefore, DETC development was dependent on ex- cells depends on endogenously encoded TCRd chains in the pression of specific g chains, consistent with a requirement transgenic mice, such as the Vd1 chain. This finding resurrected for positive selection of the cells. These data were in contrast to the possibility that DETC development requires positive the findings using Vg3 knockout mice reported above, where selection on specific ligands in the thymus. Vg1.1 and other gd T cells were detected in the epidermis in lieu As a related approach to this question, the requirement for of Vg3þ T cells. The basis for this discrepancy in results is the Vg3Vd1 TCR in development of DETCs was assessed by currently under investigation. knocking out the Vg3orVd1 genes. Disabling either of these Additional evidence that the DETC precursors undergo genes did not prevent development of DETCs, proving that TCRs specific positive selection and that selection occurs in the fetal other than Vg3Vd1 are compatible with DETC development thymus came from analysis of marker expression on fetal (72, 73). Analysis of gd TCRþ DETCs in the Vg3/ mice thymocytes. Consistent with earlier studies (75, 76), we found detected Vg1.1, Vg2, and Vg5 paired Vd1 and other Vd chains. that approximately 70% of Vg3þ gd T cells in the fetal thymus The possibility that the DETCs in Vg3 knockout mice had express various activation markers, including CD122, which is undergone specific intrathymic positive selection was inferred the b chain of the receptors for IL-2 and IL-5. We showed that from the finding that the TCRs on these DETCs had a specific CD122 expression on Vg3þ fetal thymocytes was correlated

24 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

with expression of Vd1, the chain characteristic of DETCs (74). A recent report provides independent support for positive Vg3þ fetal thymates lacking CD122 showed a much more selection of DETC precursors in the thymus. A comparison of diverse expression of TCR Vd chains, indicating that CD122 mouse strains showed normal development of Vg3þ DETCs in expression occurred specifically on Vg3þVd1þ fetal thymocytes FVB-Jax strain mice but none in the related FVB/N-Taconic and not on thymocytes with Vg3 paired to other TCRd chains. strain, and chimera experiments showed that the defect in FVB/ Remarkably, CD122 was completely absent on the gd T cells in N-Taconic was consistent with the absence of a selecting ligand the fetal thymus of 234JCg1/ mice, which lack DETCs, for DETCs in the fetal thymus (77). The identity of the putative showing a correlation between the expression of activation selecting ligand has not yet been reported. markers in the thymus and development of DETCs. Thus, expression of DETC-specific Vd1 was associated with CD122 Preference for the Vg3Vd1 pair among DETCs in wildtype expression in the thymus, which was in turn correlated with mice DETC development. These data suggested that among fetal If different VgVd pairs are compatible with DETC development, thymic precursors expressing various gd TCRs, those expressing why does the Vg3Vd1 pair predominate in wildtype mice? A Vg3 and Vd1 engaged a thymic ligand, leading to expression of likely possibility, already discussed above, is that the Vg3Vd1 activation markers and ultimately development of DETCs. pair is more likely to arise in the fetal thymus than other VgVd While highly suggestive of selection, the correlations described pairs, as a result of favored rearrangement of these genes in the above could theoretically arise by other mechanisms. However, as fetal period. However, this preference cannot fully account for a result of an unexpected observation, we were able to provide the dominance of the Vg3Vd1 pair. In wildtype mice, a small strong additional evidence that these events reflected positive but significant percentage of DETCs in the epidermis are Vg3 selection. As one approach to test whether the absence of DETC in the neonatal period, whereas such cells are nearly undetect- development in 234JCg1/ mice was specifically because of the able in the adult stage (73, 78). Thymic production of Vg3þ absence of Vg3, we attempted to complement the 234JCg1/ cells ceases by the time of birth, so the change cannot reflect mice with a transgene encoding a rearranged functional Vg2– greater thymic production of Vg3þ cells at the adult stage. Jg1Cg1 chain. Surprisingly, the transgene restored DETC Other evidence suggests that the Vg3Vd1 TCR is favored development and all the resulting DETCs expressed the transgenic among DETCs, even when it is not the most commonly Vg2 chain (74). While this finding could have been seen to expressed gd TCR in the early fetal thymus. In the previously resurrect the notion that DETC development does not depend on discussed gene-targeted mice in which the 3#-ECg1 and HsA specific TCR expression, analysis of the DETCs showed instead enhancer elements were simultaneously deleted in the Cg1 a very specific TCR configuration: all the cells expressed Vg2, as cluster, the number of Vg3þ T cells in the fetal thymus was expected, and showed great enrichment for Vd7 expression and reduced by fivefold to 20-fold at all time points, but Vg1.1þ and little or no expression of other TCRd chains, including Vd1. Vg5þ gd T cells developed normally. Therefore, Vg3þ T cells Furthermore, the Vg2þVd7þ DETCs, but not peripheral gd Tcells, represented only a small percentage of gd T cells in the fetal produced IL-2 when stimulated with keratinocyte cell lines, thymus of these mice (47) (Fig. 5). Nevertheless, approxi- suggesting that the Vg2Vd7 receptor combination, like the mately 75% of DETCs in the skin of adult knockout mice Vg3Vd1 combination, confers specificity for a keratinocyte expressed Vg3, suggesting selection for the canonical Vg3Vd1 ligand. The story was brought full circle when the analysis of TCR among DETCs (47) (Fig. 5). the Vg2 transgenic/knockout mice showed that the Vg2 transgene Between the neonatal and adult period, DETCs are believed to resulted in restored expression of CD122 on a subset of Vg2þ fetal expand greatly in the epidermis, and it is likely that the local thymocytes. Furthermore, the Vg2þ fetal thymocytes that ex- environment in the skin is important for selective expansion of pressed CD122 predominantly expressed Vd7, whereas Vg2þ fetal the Vg3þ gd T cells. The expansion could result from local thymocytes that did not express CD122 expressed a diversity of Vd engagement of the DETC TCR with epidermal ligands, although genes. These data strongly supported the conclusion that only it is believed that the putative ligand is only expressed well in specific subsets of developing fetal thymic gd T cells, including damaged or stressed epidermis (79). Alternatively, one could cells expressing the Vg3Vd1 pair and the Vg2Vd7pair,exhibit speculate that the Vg3Vd1þ DETCs may receive more potent reactivity with an unknown ligand that is expressed by fetal thymic positive selection signals in the fetal thymus, perhaps because of stromal cells and probably shared with keratinocytes. Moreover, a higher affinity of the Vg3Vd1 TCR for the selecting ligand, such reactivity leads to expression of CD122 on developing fetal which may enhance the ability of the cells to survive and expand thymocytes and is necessary for development of DETCs (74). later in the epidermis.

Immunological Reviews 215/2007 25 Xiong & Raulet gd T-cell development and selection

erations may aid in understanding why and how the thymus specializes in producing different types of T cells at different stages of development.

How does positive selection in the fetal thymus enable development of gd T cells in the epidermis? Having provided evidence for positive selection of DETCs, we sought to understand how selection in the thymus is coordinated with T-cell development in the skin. In theory, thymic selection could influence migration from the thymus, Fig. 5. The composition of different gd subsets in the fetal thymus (day 15 of gestation) and adult skin (DETC), comparing wildtype migration to the epidermis, and survival and/or expansion in littermates with H/E/ mice lacking both known Cg1 enhancer the epidermal environment. elements (HsA and 3#-ECg1). Most DETCs express Vg3 in mice of þ Consistent with a role in lymphocyte migration, we found both genotypes, but Vg3 cells are quite rare in the fetal thymus of mice. These data suggest a strong preference for Vg3þ DETCs, even that positive selection in the fetal thymus resulted in a when these cells are produced in limiting numbers. Data are based on coordinated switch in expression of several chemokine and analysis of cells from three or more mice of each type on a mixed homing receptors in the selected cells (74). Chemokine (C57BL/6/129Sv) genetic background. receptor CCR10 and sphingosine 1-phosphate receptor-1 (S1P1) were upregulated, while CCR6 was downregulated in Positive versus negative selection in the thymus may vary at the positively selected fetal thymic gd T cells. Together, these different developmental stages changes are expected to enhance emigration from the thymus Differences have been noted in comparing negative selection in and may direct cells specifically to the epidermis. Recent studies the thymus of adult and fetal (or neonatal) mice, which may be show, for example, that S1P1 is essential for the exiting of instructive in understanding the different roles of the gd T-cell mature ab T cells from the thymus in adult mice (83). Also, the subsets that arise at these stages. In the adult thymus, T cells ligand for CCR6, CCL20, is highly expressed in the thymus (84– expressing the G8 gd TCR underwent negative selection in mice 87). Upregulation of SIP1 and loss of CCR6 are therefore b that express the strong TL ligand, but little or no negative predicted to promote thymic export. Furthermore, the ligand selection occurred in neonatal mice of the same genotype, for CCR10, the chemokine CCL27, is constitutively expressed in despite indications that the G8 TCR was engaged with its ligand the skin (88–90), including the fetal skin (74). Hence, CCR10 at that stage (80). Another report indicated that transgenic upregulation may promote skin homing of positively selected þ Vg3Vd1 gd T cells underwent negative selection in adult but gd T cells after they exit the fetal thymus. These predictions not fetal thymi in certain genetic backgrounds (81). Finally, remain to be tested experimentally. treatment of fetal thymic organ cultures with anti-gd-TCR Positive selection in the thymus may also influence antibodies did not delete gd T cells, whereas the same treatment propagation of DETCs in the epidermis in other ways. A likely causes deletion of adult thymic gd T cells (82). These possibility is that the survival and expansion of DETCs in the indications suggest that developing fetal thymic gd T cells are skin are enabled by the upregulation of CD122 that occurs as highly resistant to negative selection. Perhaps this resistance a result of positive selection of DETC precursors in the fetal reflects the specialization of the fetal thymus to select cells with thymus. CD122 is the b chain of the IL-2/IL-15 receptor, and highly defined, invariant specificities. The highly constrained IL-15 is known to be essential for DETC development (91), repertoire of rearrangements and junctional diversification that whereas IL-2 is not essential (92). In both CD122 and IL-15 occur in the fetal thymus is predicted to produce a large number knockout mice, Vg3þ gd T cells were generated in the fetal of cells with receptors of the optimal affinity for the correct thymus nearly as efficiently as in wildtype mice, but DETCs did ligand. If the selecting ligands in the thymus are identical to or not develop (91, 93). It has been proposed that IL-15 have the same affinity for the receptor as the ligands recognized expressed by keratinocytes maintains survival/expansion of by the cells in the periphery, negative selection, if it occurred, DETCs in the skin, apparently through interaction with IL-15 would presumably result in deletion of the cells. Because gd T receptors expressed on skin-specific gd T cells (91). Therefore, cells that arise at the adult stage are highly diverse, effective the survival and expansion of DETCs in the skin are probably negative selection is probably necessary to reduce the risk of promoted by the upregulation of cytokine receptors, which generating pathogenic autoreactive gd T cells. These consid- occurs during positive selection in the thymus. It remains

26 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

possible that additional stimulation of the cells in the enabling them to exit the thymus and home to secondary epidermis is also important in maintaining expression of lymphoid organs. these cytokine receptors. Presumably, the directed differential homing of specific subsets of gd T cells is ultimately important for positioning these A proposed model for development of tissue-localized gd different T cells appropriately for their specialized roles. In the T-cell subsets case of invariant gd subsets, such as the Vg3Vd1þ DETCs, the Programmed Vg gene rearrangement and cellular selection cells are localized to the site where they can recognize their cooperate to regulate development of different tissue-specific specialized ligands, which are induced in the skin by the gd T cells (Fig. 6). In the early fetal thymus, a programmed pathophysiological changes (79). The much more variable rearrangement process based on proximity of the gene to Jg1 adult gd T-cell subsets, such as Vg2þ T cells, are believed to be favors the generation of Vg3Vd1þ gd T cells, the natural skin- involved in surveillance for more diverse sets of ligands, specific gd T-cell precursors, and Vg4Vd1þ T cells, destined for appropriate to their localization to the secondary lymphoid other epithelial locations. Subsequently, the Vg3Vd1þ gd T cells organs. are positively selected by engagement of the Vg3Vd1 TCR with a ligand expressed in the fetal thymus, resulting in a coordinate Concluding remarks switch in expression of chemokine receptors and cytokine receptors, which in turn directs the specific migration of these Our research has addressed all stages of the development of gd cells to the skin and enhances their subsequent expansion and T cells, including the decision to differentiate along the ab survival there. versus gd lineage, ordered production of Vg-defined subsets of In the adult thymus, Vg3 and Vg4 (and Vd1) rearrangements gd T cells over the course of ontogeny and the acquisition are suppressed by the programmed rearrangement process, during development of guidance information that directs the favoring production of Vg2þ and other adult-type gd T cells. cell to its appropriate destination. Our findings together with Although it remains somewhat murky whether positive those of others show that programmed events, such as Vg gene selection acts on adult-type gd T cells, it is likely that rearrangement and junctional diversification mechanisms, a maturation process, whether selection or another process, cooperate with subsequent cellular selection processes, confers these cells with their specific homing properties, ultimately generating distinct sets of specific tissue-localized gd T cells. For example, in the absence of selection, most of the productive Vg rearrangements in the early fetal period are canonical Vg3 rearrangements (9). Subsequent development of DETCs nevertheless requires a cellular selection process. Hence, with respect to gd T-cell specificity, the molecular program greatly restricts the possible repertoire at the fetal stage, and selection reinforces this restriction. In the normal course of events, many of the cells subjected to selection will already express the canonical Vg3Vd1 receptor, so it might be surmised that shaping the repertoire is not the sole or even main role of positive selection. But under exceptional circumstances, the selection process is capable of selecting alternative TCRs, including the Vg2Vd7 receptor, which apparently provides specificity similar to that of the canonical

Fig. 6. Proposed overall scheme of developmental events leading to receptor. development of gd T cells in the skin (DETCs). In the first phase, The role of selection extends beyond shaping the repertoire, programmed rearrangement events lead to a striking bias in TCR gene however, by inducing expression of specific chemokine rearrangements, favoring production of the Vg3Vd1 canonical TCR, and may direct expression on thymic stromal cells of ligands necessary receptors and cytokine receptors that are likely to determine for positive selection of DETCs. Positive selection ensures that only cells the subsequent localization, survival, and proliferation of the with the correct specificity differentiate further and bestow the selected cells. This outcome of positive selection at the fetal stage may be cells with chemokine receptors, which enable the cells to exit the thymus and migrate to the epidermis, and with cytokine receptors, ‘hardwired’ in the sense that all gd lineage cells at this stage will which aid in subsequent survival and expansion of the cells. make this response to TCR ligation. We found, for example, that

Immunological Reviews 215/2007 27 Xiong & Raulet gd T-cell development and selection

fetal thymocytes stimulated with anti-gd-TCR antibodies negative selection of conventional T cells, were found to upregulated the same chemokine and cytokine receptors as promote development of CD8aaþ T cells (98). It remains to be DETC precursors, even if the thymocytes lacked a known DETC- seen whether positive selection of these and other T-cell subsets compatible receptor (74). These considerations suggest that the also induces chemokine and cytokine receptors that aid in specialization of the fetal thymus to produce DETCs is imposed peripheral localization and survival of the cells. at numerous levels of development and prompt the question of gd T cells in the human also display unique tissue distribution why such stage-specific specialization evolved. We suggest that patterns. For example, Vd1þ gd T cells are predominant in it is necessary because the repertoires, diversity, and functions intestines and skin (99, 100), where they are believed to play an of the cells produced at the fetal versus adult stages differ so important role in tumor surveillance and maintenance of tissue greatly. Specialization of the thymus for production of different integrity, while Vd2þ gd T cells are the dominant population of subsets at different stages is one way to vary developmental secondary lymphoid organs and blood (101). The findings in processes at different stages to impart different repertoires the mouse system will ultimately provide a guide for and functionality on the corresponding cells. As examples, the understanding the development and localization of gd T cells absence of TdT at the fetal stage is critical to producing a high in humans. frequency of canonical Vg3 rearrangements appropriate for Detailed molecular events that dictate programmed Vg gene DETC production (24, 25), and we have proposed in this review rearrangement and generation of corresponding gd T cells that the reduced susceptibility of fetal thymocytes to negative remain to be defined. In understanding the adult pattern, selection may be important for producing cells like DETCs, with identification of transcription factors that interact with the cis- a fixed specificity for a self-ligand. acting elements in the TCRg locus, especially the promoters, is Despite its role in the production of DETCs and Vg4Vd1 one of the most prominent issues to be resolved. Understanding invariant T cells at the fetal stage, ontogenic specialization is the role of E2A in this process will be an important step. clearly not required to produce other specialized T-cell subsets. However, E2A cannot be the sole determinant of the adult Aside from the obvious CD4/CD8 subsets, some of the T-cell developmental pattern because E2A-deficient mice still have subsets that arise in the adult thymus have properties similar to plentiful Vg2 gd T cells, but no Vg3þ gd T cells in the adult those of DETCs. Examples are natural killer (NK) T cells, thymus (N. Xiong, C. Murre, D. H. Raulet, unpublished data). CD8aaþ T cells, mucosal-associated invariant T cells (MALT), In understanding the fetal pattern, it will be necessary to dissect which exhibit restricted TCR repertoires, are localized to how the rearrangement pattern is controlled by gene location in specific peripheral tissues, and exhibit distinctive immune the cluster, independent of differences in Vg gene accessibility. functions. The development of these cells is not separated in With respect to gd T-cell selection, it is critical to identify the ontogeny from the production of conventional ab T cells and ligands that mediate positive selection of DETCs and potentially adult-type gd T cells, but the differentiation process they other gd T-cell subsets and to determine if and how their undergo may share one or more of the features of DETC expression is restricted to specific developmental stages or cell development outlined in this review. Developmentally, many types in the thymus. To understand tissue localization, it will be of these cells undergo selection processes that are unlike those of essential to test directly the roles of specific chemokine conventional ab T cells but similar in some respects to those receptors and other molecules that influence cell localization. of gd T cells. For example, NK T cells are positively selected by Finally, an essential issue to explore is whether and how a self-glycolipid antigen associated with CD1 (94, 95), while positively selected gd T cells at different stages of ontogeny (and MALT cells are positively selected by the self-antigen MR1 (96, therefore selecting environments) are bestowed with different 97). Strong TCR/antigen interactions, which usually result in homing properties.

References

1. Raulet DH. The structure, function, and 3. Allison JP, Havran WL. The immunobiology 5. Garman RD, Doherty PJ, Raulet DH. Diversity, molecular genetics of the g/d T cell receptor. of T cells with invariant gd antigen receptors. rearrangement and expression of murine T cell Annu Rev Immunol 1989;7:175–207. Annu Rev Immunol 1991;9:679–705. gamma genes. Cell 1986;45:733–742. 2. Haas W, Pereira P, Tonegawa S. Gamma/ 4. Hayday AC. [gamma][delta] cells: a right 6. Capone M, Hockett RD, Zlotnik A. Kinetics of delta cells. Annu Rev Immunol time and a right place for a conserved third T cell receptor b, g, and d rearrangements 1993;11:637–685. way of protection. Annu Rev Immunol during adult thymic development: T cell 2000;18:975–1026. receptor rearrangemnts are present in CD44þCD25þ Pro-T thymocytes. Proc Natl Acad Sci U S A 1998;95:12522–12527.

28 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

7. Liva´k F, Tourigny M, Schatz DG, Petrie HT. 23. Davis MM, Bjorkman PJ. T cell antigen genes 37. Yancopoulos GD, Desiderio SV, Paskind M, Characterization of TCR gene rearrange- and T cell recognition. Nature Kearney JF, Baltimore D, Alt FW. Preferential

ments during adult murine T cell develop- 1988;334:395–402. utilization of the most JH-proximal VH gene ment. J Immunol 1999;162:2575–2580. 24. Tigelaar RE, Lewis JM. Immunobiology of segements in pre-B-cell lines. Nature 8. Livak F, Petrie HT, Crispe IN, Schatz DG. In- mouse dendritic epidermal T cells: a decade 1984;311:727–733. frame TCR d rearrangements play a critical later, some answers, but still more questions. 38. Yancopoulos GD, Malynn BA, Alt FW. role in the ab/gd lineage decision. Immu- J Invest Dermatol 1995;105:43S–49S. Developmentally regulated and strain-spe-

nity 1995;2:617–627. 25. Asarnow DM, Cado D, Raulet DH. Selection cific expression of murine VH gene families. 9. Dudley EC, Girardi M, Owen MJ, Hayday AC. is not required to produce invariant T cell J Exp Med 1988;168:417–435. ab and gd T cells can share a late common receptor g-gene junctional sequences. 39. Chien Y-H, et al. T cell receptor d gene precursor. Curr Biol 1995;5:659–669. Nature 1993;362:158–160. rearrangements in early thymocytes. Nature 10. Kang J, Baker J, Raulet D. Evidence that 26. Zhang Y, et al. The role of short homology 1987;330:722–727. productive rearrangements of TCRg genes repeats and TdT in generation of the invari- 40. Peschon JJ, et al. Early lymphocyte expansion influence the fate of developing T cells. ant gd antigen receptor repertoire in the fetal is severely impaired in interleukin 7 recep- Eur J Immunol 1995;25:2706–2709. thymus. Immunity 1995;3:439–447. tor-deficient mice. J Exp Med 11. Kang J, Volkmann A, Raulet DH. Evidence 27. Itohara S, et al. T cell receptor d gene mutant 1994;180:1955–1960. that gammadelta versus alphabeta T cell fate mice: independent generation of ab T cells 41. Corcoran AE, Riddell A, Krooshoop D, determination is initiated independently of T and programmed rearrangements of gd TCR Venkitaraman AR. Impaired immunoglobu- cell receptor signaling. J Exp Med genes. Cell 1993;72:337–348. lin gene rearrangement in mice lacking the 2001;193:689–698. 28. Korman A, Marusic-Galesic S, Spencer D, IL-7 receptor. Nature 1998;391:904–907. 12. Ciofani M, Knowles GC, Wiest DL, von Kruisbeek A, Raulet DH. Predominant vari- 42. Malynn BA, Yancopoulos GD, Barth JE, Bona

Boehmer H, Zuniga-Pflucker JC. Stage- able region gene usage by g/d T cell recep- CA, Alt FW. Biased expression of JH- specific and differential notch dependency at tor-bearing cells in the adult thymus. J Exp proximal VH genes occurs in the newly the alphabeta and gammadelta T lineage Med 1988;168:1021–1040. generated repertoire of neonatal and adult bifurcation. Immunity 2006;25:105–116. 29. Heilig JS, Tonegawa S. Diversity of murine mice. J Exp Med 1990;171:843–859. 13. Saint-Ruf C, Panigada M, Azogui O, Debey P, gamma genes and expression in fetal and 43. Jung D, Giallourakis C, Mostoslavsky R, Alt von Boehmer H, Grassi F. Different initiation adult T lymphocytes. Nature FW. Mechanism and control of V(D)J of pre-TCR and gammadeltaTCR signalling. 1986;322:836–840. recombination at the immunoglobulin Nature 2000;406:524–527. 30. Takagaki Y, Nakanishi N, Ishida I, Kanagawa heavy chain locus. Annu Rev Immunol 14. Haks MC, et al. Attenuation of gammadel- O, Tonegawa S. T-cell receptor-g and -d 2006;24:541–570. taTCR signaling efficiently diverts thymo- genes preferentially utilized by adult thy- 44. Connor AM, Fanning LJ, Celler JW, Hicks LK, cytes to the alphabeta lineage. Immunity mocytes for the surface expression. Ramsden DA, Wu GE. Mouse VH7183 2005;22:595–606. J Immunol 1989;142:2112–2121. recombination signal sequences mediate 15. Hayes SM, Li L, Love PE. TCR signal strength 31. Guy-Grand D, Cerf-Bensussan N, Malissen B, recombination more frequently than those influences alphabeta/gammadelta lineage Malassis-Seris M, Briottet C, Vassalli P. Two of VHJ558. J Immunol 1995;155: fate. Immunity 2005;22:583–593. gut intraepithelial CD8þ lymphocyte popu- 5268–5272. 16. Raulet D, et al. Control of gd T cell devel- lations with different T cell receptors: a role 45. Williams GS, et al. Unequal VH gene rear- opment. Immunol Rev 1991;120:185–204. for gut epithelium in T cell differentiation. rangement frequency within the large 17. Allison JP. Gamma delta T-cell development. J Exp Med 1991;173:471–481. VH7183 gene family is not due to recom- Curr Opin Immunol 1993;5:241–246. 32. Lefrancois L, Puddington L. Extrathymic bination signal sequence variation, and 18. Havran WL, Allison JP. Thy-1þ dendritic intestinal T-cell development: virtual reality? mapping of the genes shows a bias of rear- epidermal cells of adult mice arise from fetal Immunol Today 1995;16:16–21. rangement based on chromosomal location. thymic precursors. Nature 1990;344:68–70. 33. Ikuta K, et al. A developmental switch in J Immunol 2001;167:257–263. 19. Itohara S, et al. Homing of a gd thymocyte thymic lymphocyte maturation potential 46. Xiong N, Baker JE, Kang C, Raulet DH. The subset with homogenous T-cell receptors to occurs at the level of hematopoietic stem genomic arrangement of T cell receptor mucosal epithelia. Nature 1990;343: cells. Cell 1990;62:863–874. variable genes is a determinant of the 754–757. 34. Goldman J, Spencer D, Raulet D. Ordered developmental rearrangement pattern. Proc 20. Viney JL, MacDonald TT, Kilshaw PJ. T-cell rearrangement of variable region genes of Natl Acad Sci U S A 2004;101:260–265. receptor expression in intestinal intra- the T cell receptor gamma locus correlates 47. Tanimoto K, Liu Q, Bungert J, Engel JD. epithelial lymphocyte subpopulations of with transcription of the unrearranged Effects of altered gene order or orientation of normal and athymic mice. Immunology genes. J Exp Med 1993;177: the locus control region on human beta- 1989;66:583–587. 729–739. globin gene expression in mice. Nature 21. Rocha B, Vassalli P, Guy-Grand D. Thymic 35. Baker JE, Cado D, Raulet DH. Developmen- 1999;398:344–348. and extrathymic origins of gut intraepithelial tally programmed rearrangement of T cell 48. Xiong N, Kang CH, Raulet DH. Redundant lymphocyte populations in mice. J Exp Med receptor Vg genes is controlled by sequences and unique roles of two enhancer elements 1994;180:681–686. immediately upstream of the Vg genes. in the TCR gamma locus in gene regulation 22. Pereira P, Hermitte V, Lembezat MP, Immunity 1998;9:159–168. and gamma delta T cell development. Boucontet L, Azuara V, Grigoriadou K. 36. Agata Y, et al. Histone acetylation Immunity 2002;16:453–463. Developmentally regulated and lineage- determines the developmentally regulated 49. Greenbaum S, Zhuang Y. Regulation of early specific rearrangement of T cell receptor accessibility for T cell receptor gamma lymphocyte development by E2A family Valpha/delta gene segments. Eur J Immunol gene recombination. J Exp Med proteins. Semin Immunol 2002;14: 2000;30:1988–1997. 2001;193:873–879. 405–414.

Immunological Reviews 215/2007 29 Xiong & Raulet gd T-cell development and selection

50. Bain G, Romanow WJ, Albers K, Havran WL, 65. Bonneville M, et al. Blockage of ab T-cell 79. Havran WL, Chien Y, Allison JP. Recognition Murre C. Positive and negative regulation of development by TCR gd transgenes. Nature of self antigens by skin-derived T cells with V(D)J recombination by the E2A proteins. 1989;342:931–934. invariant gamma delta antigen receptors. J Exp Med 1999;189:289–300. 66. Pereira P, Zijlstra M, McMmaster J, Loring Science 1991;252:1430–1432. 51. Baker JE, Kang J, Xiong N, Chen T, Cado D, JM, Jaenisch R, Tonegawa S. Blockade of 80. Lin T, et al. Autospecific gammadelta thy- Raulet DH. A novel element upstream of the transgenic gamma delta T cell development mocytes that escape negative selection find Vg2 gene in the murine T cell receptor g in beta-2-microglobulin deficient mice. sanctuary in the intestine. J Clin Invest locus cooperates with the 3# enhancer to act EMBO J 1992;11:25–31. 1999;104:1297–1305. as a locus control region. J Exp Med 67. Wells F, Tatsumi Y, Bluestone J, Hedrick S, 81. Iwashima M, Green A, Bonyhadi M, Davis 1999;190:669–679. Allison J, Matis L. Phenotypic and functional MM, Allison JP, Chien YH. Expression of 52. Sleckman BP, et al. Accessibility control of analysis of positive selection in the g/d T cell a fetal gamma delta T-cell receptor in adult variable region gene assembly during T-cell lineage. J Exp Med 1993;177:1061–1070. mice triggers a non-MHC-linked form of development. Immunol Rev 1998;165: 68. Schweighoffer E, Fowlkes BJ. Positive selec- selective depletion. Int Immunol 121–130. tion is not required for thymic maturation of 1991;3:385–393. 53. Seidl KJ, et al. Position-dependent inhibition transgenic gd T cells. J Exp Med 82. Itohara S, Tonegawa S. Selection of gd T cells of class-switch recombination by PGK-neor 1996;183:2033–2041. with canonical T-cell antigen receptors in cassettes inserted into the immunoglobulin 69. Ferrick DA, et al. T cell function and fetal thymus. Proc Natl Acad Sci U S A heavy chain constant region locus. Proc Natl expression are dramatically altered in T cell 1990;87:7935–7938. Acad Sci U S A 1999;96:3000–3005. receptor Vg1.1Jg4Cg4 transgenic mice. Cell 83. Matloubian M, et al. Lymphocyte egress 54. Tough DF, Sprent J. Lifespan of gamma/ 1989;57:483–492. from thymus and peripheral lymphoid delta T cells. J Exp Med 1998;187:357–365. 70. Bonneville M, et al. Transgenic mice dem- organs is dependent on S1P receptor 1. 55. Zijlstra M, Bix M, Simister NE, Loring JM, onstrate that epithelial homing of g/d T cells Nature 2004;427:355–360. Raulet DH, Jaenisch R. b2-Microglobulin is determined by cell lineages independent 84. Greaves DR, et al. CCR6, a CC chemokine deficient mice lack CD48þ cytolytic T cells. of T cell receptor specificity. J Exp Med receptor that interacts with macrophage Nature 1990;344:742–746. 1990;171:1015–1026. inflammatory protein 3alpha and is highly 56. Grusby M, et al. Mice lacking major histo- 71. Ferrero I, Wilson A, Beermann F, Held W, expressed in human dendritic cells. J Exp compatibility complex class I and class II MacDonald HR. T cell receptor specificity is Med 1997;186:837–844. molecules. Proc Natl Acad Sci U S A critical for the development of epidermal 85. Baba M, et al. Identification of CCR6, the 1993;90:3913–3917. gammadelta T cells. J Exp Med specific receptor for a novel lymphocyte- 57. Correa I, Bix M, Liao N-S, Zijlstra M, Jaenisch 2001;194:1473–1483. directed CC chemokine LARC. J Biol Chem R, Raulet D. Most gd T cells develop nor- 72. Mallick-Wood C, Lewis JM, Richie LI, Owen 1997;272:14893–14898. mally in b2-microglobulin-deficient mice. MJ, Tigelaar RE, Hayday AC. Conservation of 86. Liao F, Alderson R, Su J, Ullrich SJ, Proc Natl Acad Sci U S A 1992;89:653–657. T cell receptor conformation in epidermal gd Kreider BL, Farber JM. STRL22 is a receptor

58. Bigby M, et al. Most gd T cells develop cells with disrupted primary Vg gene usage. for the CC chemokine MIP-3alpha. normally in the absence of MHC class II Science 1998;279:1729–1733. Biochem Biophys Res Commun molecules. J Immunol 1993;151: 73. Hara H, et al. Development of dendritic 1997;236:212–217. 4465–4475. epidermal T cells with a skewed diversity of 87. Varona R, et al. Molecular cloning, 59. Ito K, Van KL, Bonneville M, Hsu S, Murphy gamma delta TCRs in V delta 1-deficient functional characterization and mRNA DB, Tonegawa S. Recognition of the product mice. J Immunol 2000;165:3695–3705. expression analysis of the murine of a novel MHC TL region gene (27b) by 74. Xiong N, Kang C, Raulet DH. Positive selec- chemokine receptor CCR6 and its specific a mouse gamma delta T cell receptor. Cell tion of dendritic epidermal gd T cell pre- ligand MIP-3alpha. FEBS Lett 1990;62:549–561. cursors in the fetal thymus determines 1998;440:188–194. 60. Schild H, et al. The nature of MHC recogni- expression of skin-homing receptors. 88. Morales J, et al. CTACK, a skin-associated tion by gd T cells. Cell 1994;76:29–37. Immunity 2004;21:121–131. chemokine that preferentially attracts skin- 61. Bonneville M, et al. Recognition of a self 75. Leclercq G, Plum J, Nandi D, De Smedt M, homing memory T cells. Proc Natl Acad Sci major histocompatibility complex TL region Allison J. Intrathymic differentiation of Vg3 U S A 1999;96:14470–14475. product by gd T-cell receptors. Proc Natl T cells. J Exp Med 1993;178:309–315. 89. Homey B, et al. Cutting edge: the orphan Acad Sci U S A 1989;86:5928–5932. 76. Van Beneden K, De Creus A, Stevenaert F, chemokine receptor G protein-coupled 62. Bluestone JA, Cron RQ, Cotterman M, Debacker V, Plum J, Leclercq G. Expression receptor-2 (GPR-2, CCR10) binds the ski- Houlden BA, Matis LA. Structure and speci- of inhibitory receptors Ly49E and CD94/ n-associated chemokine CCL27 (CTACK/ ficity of T cell receptor g/d on major histo- NKG2 on fetal thymic and adult epidermal ALP/ILC). J Immunol 2000;164: compatibility complex antigen-specific TCR V gamma 3 lymphocytes. J Immunol 3465–3470. CD3þ,CD4,CD8 T lymphocytes. J Exp 2002;168:3295–3302. 90. Jarmin DI, et al. Cutting edge: identification Med 1988;168:1899–1916. 77. Lewis JM, Girardi M, Roberts SJ, Barbee SD, of the orphan receptor G-protein-coupled 63. Crowley MP, et al. A population of murine Hayday AC, Tigelaar RE. Selection of the receptor 2 as CCR10, a specific receptor for gammadelta T cells that recognize an cutaneous intraepithelial gammadelta(þ)T the chemokine ESkine. J Immunol inducible MHC class Ib molecule. Science cell repertoire by a thymic stromal determi- 2000;164:3460–3464. 2000;287:314–316. nant. Nat Immunol 2006;7:843–850. 91. De Creus A, Van Beneden K, Stevenaert F, 64. Dent AL, Matis LA, Hooshmand F, Widacki 78. Minagawa M, Ito A, Shimura H, Tomiyama Debacker V, Plum J, Leclercq G. Develop- SM, Bluestone JA, Hedrick SM. Self-reactive K, Ito M, Kawai K. Homogeneous epithelial mental and functional defects of thymic and gd T cells are eliminated in the thymus. gamma delta T cell repertoire of the skin is epidermal V gamma 3 cells in IL-15-deficient Nature 1990;343:714–719. shaped through peripheral selection. and IFN regulatory factor-1-deficient mice. J Dermatol Sci 2001;25:150–155. J Immunol 2002;168:6486–6493.

30 Immunological Reviews 215/2007 Xiong & Raulet gd T-cell development and selection

92. Schimpl A, Hu¨nig T, Elbe A. Development 96. Tilloy F, et al. An invariant T cell receptor alpha 100. Ebert LM, Meuter S, Moser B. Homing and and Function of the Immune System in chain defines a novel TAP-independent major function of human skin gammadelta T cells Mice with Targeted Disruption of the Inter- histocompatibility complex class Ib-restricted and NK cells: relevance for tumor surveil- leukin 2 Gene. San Diego: Academic Press, alpha/beta T cell subpopulation in mammals. lance. J Immunol 2006;176:4331–4336. 1994. J Exp Med 1999;189:1907–1921. 101. Parker CM, et al. Evidence for extrathymic 93. Kawai K, Suzuki H, Tomiyama K, Minagawa 97. Treiner E, et al. Selection of evolutionarily changes in the T cell receptor gamma/delta M, Mak TW, Ohashi PS. Requirement of the conserved mucosal-associated invariant repertoire. J Exp Med 1990;171: IL-2 receptor beta chain for the development T ecells by MR1. Nature 2003;422:164–169. 1597–1612. of Vgamma3 dendritic epidermal T cells. 98. Yamagata T, Mathis D, Benoist C. Self- J Invest Dermatol 1998;110:961–965. reactivity in thymic double-positive cells 94. Zhou D, et al. Lysosomal glycosphingolipid commits cells to a CD8 alpha alpha lineage recognition by NKT cells. Science with characteristics of innate immune cells. 2004;306:1786–1789. Nat Immunol 2004;5:597–605. 95. Kronenberg M. Toward an understanding of 99. Groh V, Steinle A, Bauer S, Spies T. Recog- NKT cell biology: progress and paradoxes. nition of stress-induced MHC molecules by Annu Rev Immunol 2005;23:877–900. intestinal epithelial gammadelta T cells. Sci- ence 1998;279:1737–1740.

Immunological Reviews 215/2007 31