The thymus generates T cells that are generally functionally immature and thus require peripheral activation for differentiation into effector lymphocytes. Notable exceptions to this rule are murine γδ T cells, many of which have been shown to acquire their functional potential during thymic development from late embryonic stages. Here, we review the underlying ontogenic processes and molecular differentiation mechanisms of murine γδ T cells, focusing on the transcriptional control of IFN-γ and IL-17 expression. We propose that functional commitment of γδ T cells occurs in “developmental windows” defined by the molecular composition of the thymic microenvironment, such as T-cell receptor (TCR), TCR coreceptor ligands, and cytokines. We further discuss the similarities and particularities of functional development of γδ T cells in mice and humans, while highlighting some key unresolved issues for future investigation.
γδ T cells are conserved throughout evolution and across species. They develop in the thymus together with αβ T cells but rearrange a different T-cell receptor (TCR) consisting of a TCR-γ and TCR-δ chain. These TCRγδ are not MHC restricted, thus their antigen recognition does not rely on CD4 or CD8 co-receptors, although the majority of mouse intestinal intraepithelial γδ T lymphocytes (γδ iIELs) express the CD8αα dimer and various human γδ T-cell subsets express CD8 (but not CD4) . Soon after their discovery more than two decades ago, it became clear that the γδ T-cell population was significantly heterogeneous, and γδ T-cell subsets were subsequently grouped by the variable (V) segments employed during TCRγδ chain rearrangements [1-3]. Alternatively, γδ T-cell subsets may also be grouped according to tissue location. Here, we revisit these issues and suggest that γδ T-cell subsets might better be grouped with respect to γδ T-cell function. We discuss how some γδ effector T-cell subsets arise exclusively before birth, which literally qualifies them as inborn or innate lymphocytes. Further, we review the transcriptional control of mouse and human effector γδ T-cell differentiation and highlight lineage- and species-specific mechanisms.
Some γδ T cells acquire effector functions during thymic development
As illustrated in Figure 1, CD44(hi) early thymic precursors in mice simultaneously initiate transcription and rearrangement at their Trb, Trg, and Trd loci. Productive recombination and pairing of matching TCR-γ and TCR-δ chains facilitate the transition of a “γδ quality control checkpoint” selecting γδ T-cell precursors for their competence to transduce signals via their TCR [4-6]. Successful selection induces sustained high expression of TCRγδ genes . At this point, naive γδ T cells as defined by a CD44(lo)CD27(+)CD62L(+) phenotype can leave the thymus and populate secondary lymphoid organs and blood . However, other γδ T cells undergo further intrathymic differentiation before they are exported to the periphery as CD44(hi) γδ T cells. This differentiation results in the development of multiple γδ T-cell subsets such as dendritic epidermal γδ cells (DETCs) , IL-17A-producing γδ cells (γδ17 cells) , or NK1.1(+) γδ cells (γδ NKT cells) [10, 11]. In the thymus of the adult mouse, DETC precursors are virtually absent while γδ17 and γδ NKT cells together make up the large majority of CD24(+)CD44(hi) γδ thymocytes, also termed “cluster B” γδ thymocytes [6, 9]. Interestingly, development of all three CD44(hi) γδ cell subsets was proposed to be fully or at least partially restricted to the fetal thymus [8, 12, 13]. Thus, the presence of these three subsets in the adult thymus may be due to self-renewal and/or recirculation from the periphery back into the adult thymus. While some peripheral lymphocytes are certainly able to reenter the thymus , experiments based on transplantation of fetal or neonatal thymuses clearly support long-term persistence and local self-renewal of thymic γδ CD44(hi) T-cell subsets [12, 13]. Thus, in conclusion, there is a strong evidence that certain γδ cell subsets differentiate into CD44(hi) cells with committed effector function already in the embryonic thymus.
γδ T cells develop in coordinated waves
Fetal thymic γδ T-cell development progresses in successive waves that associate with variable Vγ segment usage . Adopting the “Tonegawa nomenclature” , a seminal review by Carding and Egan  highlighted that murine γδ T-cell subsets develop in the following order: Vγ5(+) cells after embryonic day 13 (E13) to approximately E17, Vγ6(+) cells from E14 to around birth, and finally Vγ1(+) and Vγ4(+) cells from E16 onward. Note that Vγ7(+) cells supposedly develop extrathymically, as discussed elsewhere . Here, we would like to resist this Vγ-centered view of fetal γδ T-cell development and instead support a concept of sequential development of functional γδ cell waves (Fig. 2). In the first days of fetal T-cell development, exclusive Vγ5 and Vγ6 gene rearrangement is controlled by epigenetic TCR gene activation. However, the differentiation of fetal thymocytes into Vγ5(+) DETCs or invariant Vγ6(+) γδ cells  is not dependent on unique oligoclonal TCR-γ and TCR-δ rearrangements. Transgenic expression of other TCR combinations during that early fetal period still supports DETC development [18, 19], and it appears that preferential early rearrangement of Vγ5 segments is largely determined by genomic location .
After DETC development, the next functional developmental wave consists of γδ17 cells. γδ17 cells are a heterogeneous subset that mainly includes Vγ6(+) and Vγ4(+) cells, but also Vγ1(+) cells. In the periphery, γδ17 cells are found in the peritoneal cavity , in the dermis [22-25], but also in secondary lymphoid organs [9, 26]. We recently demonstrated that the capacity to produce IL-17A is instructed to developing thymocytes before, and thus independent of, TCR rearrangement  and collective evidence suggests that the development of γδ17 cells does not occur after the perinatal period [12, 21, 23, 24]. However, this notion remains controversial as it was later suggested that uncommitted or naïve γδ cells from the periphery of adult mice might initiate a rapid antigen-specific IL-17A response upon recognition of the foreign protein phycoerythrin [27, 28].
The third functional wave of γδ cell development generates γδ NKT cells [10, 11], which, similar to invariant TCRαβ(+) NKT cells, are agonist selected . While γδ NKT cells develop with a very restricted oligoclonal repertoire in the fetal thymus, related but more polyclonal γδ NKT cells continue to be produced at later stages . Although the origin of CD8αα(+) iIELs remains controversial, it appears that the thymus starts to export γδ iIEL precursors during the perinatal period, as CD8αα(+) iIELs were substantially reduced in neonatal mice that had undergone thymectomy [30-32]. However, the thymus continues to generate precursors that are fully capable of reconstituting the γδ CD8αα(+) iIEL compartment well into adult life .
In conclusion, at least five functional subsets of γδ T cells are produced and exported from the thymus at defined periods of fetal, neonatal, and adult development (Fig. 2). Importantly, these subsets do not exactly overlap with Vγ gene usage. Nonetheless, linking γδ T-cell function to Vγ expression has proven practical, and has yielded useful data. Recently, transcriptome analysis of γδ thymocytes by the ImmGen consortium based on surface TCR-γ or TCR-δ expression indicated the existence of three separate subtypes of γδ effector cells in the thymus . These subsets roughly overlap with DETC, γδ17, and γδ NKT cell subsets, and express unique transcription factor modules that likely program effector function.
Tissue-specific γδ cell subsets
We are only beginning to understand what drives the association of single TCR-γ or TCR-δ chains with specific anatomical sites. Most human peripheral blood γδ T cells express Vγ9(+) and Vδ2(+) TCR chains of relatively limited diversity, while human γδ T cells with other Vδ segments are more abundant in other tissues, such as Vδ1(+) cells in the liver . In mice, a combination of TCR-dependent positive selection  and an epigenetic window of opportunity  induces expression of the skin-homing chemokine receptor CCR10 in DETC precursors. This process is thought to require TCRγδ interactions with yet-to-be discovered ligands as fetal Vγ5(+) precursors in mice in which the Skint1 gene (encoding a protein involved in DETC functional maturation) is defective are not properly selected and do not efficiently localize to the skin . iIELs are guided to and are retained within the intestinal epithelium by CCR9/CCL25 interactions [33, 37, 38], while γδ NKT cells are preferentially localized in the spleen and are even more abundant in the liver [39, 40]. However, it is currently unclear whether this results from specific homing or homeostatic factors that support the cells’ survival. γδ17 cells are marked by CCR6 expression . γδ17 cells are present within diverse tissues, such as dermis, cornea, CNS, liver, LNs, tongue, and the female and male reproductive tracts. Gray et al.  recently showed that CXCR6(+) γδ17 cells, in contrast to DETCs in the epidermis, are motile within the dermis. As these dermal CXCR6(+) γδ17 cells were shown only to migrate to skin draining LNs at low rates , it remains to be determined to what extent γδ17 cells from different tissues migrate via the circulation between lymphoid organs, blood, and various tissues.
Transcriptional control of γδ T-cell differentiation
The molecular cues received by developing thymic γδ T cells can include ligands for both TCRγδ and CD27 [42, 43], and also cytokines such as TGF-β  or IL-7 , which may contribute to the balance between IFN-γ-committed versus IL-17A-commited γδ T-cell pools. These signals are integrated at the level of downstream transcription factors and chromatin modifications that ultimately control Ifng and Il17 gene expression.
The IL-17A-producing differentiation pathway appears to be regulated both by a set of γδ-specific mechanisms, and by others shared with CD4(+) T helper 17 (Th17) cells. A notable example of the former is the Src family kinase Blk, mostly known for its role in (pre-)BCR signaling, which was shown to be selectively required for the differentiation of γδ17 cells in the thymus . Blk, which negatively regulates TCRγδ signaling, is essential for the development of Vγ6(+) (and, to lesser extent, Vγ4(+)) IL-17A-secreting γδ T cells (but not Th17 cells). Another γδ-specific determinant is the transcription factor RelB, a component of the noncanonical NF-κB pathway, which appears to be activated by LTβR signaling during the development of γδ17 cells . This regulates the expression of the “master” transcription factor RORγt, which in Th17 cells is instead induced by the canonical c-Rel-dependent NF-κB pathway . More recently, Kang and colleagues  have shown that RORγt expression in Vγ4(+) IL-17A(+) γδ T cells is also controlled by the high-mobility group box transcription factor Sox4, whereas the related Sox13 factor induces Blk expression. The importance of Sox13 was further illustrated by the selective deficiency in Vγ4(+) IL-17A(+) γδ cells in two mouse strains harboring a Sox13 frameshift mutation, which blocked the neonatal development of this specific population . Of note, this protected the mice from developing imiquimod-induced psoriasis-like dermatitis .
A genetic deficiency in RORγt (Rorc) was shown to deplete IL-17A-producing γδ T cells from peripheral lymphoid organs and mucosal tissues [50-52]. However, Shibata et al.  showed that Rorc-deficient thymuses retain approximately 30% (in the fetus) to approximately 50% (in the adult) of IL-17A(+)-secreting γδ T cells, and identified Hes1 as a more stringent determinant of γδ17 cell development. Thus, IL-17A-producing γδ thymocytes are essentially absent in Hes1-deficient animals, indicating a dependence on Notch1 signaling that is common to peripherally induced Th17 cells .
In contrast, the transcription factor Irf4, which (when complexed with Batf) sets the developmental program of Th17 cells , is largely dispensable for the differentiation of IL-17A-producing γδ T cells [47, 55]. In fact, our analyses showed that Irf4 is very poorly expressed in CD27(−)CCR6(+) γδ thymocytes, which are selectively enriched in transcripts for RORγt and ROR-α .
Regarding the IFN-γ (type 1) differentiation pathway, it has been known for a long time that, similar to Th1 cells, γδ cells express high levels of “master regulator” T-bet (Tbx21) [57, 58]. The balance between T-bet and RORγt expression in γδ cells has been recently shown to depend on the transcription factor Egr3. Egr3 expression is upregulated in embryonic γδ thymocytes upon interactions with Skint1(+) medullary thymic epithelial cells . Skint1 acts as a selecting determinant for thymic Vγ5(+)Vδ1(+) progenitors of DETCs that populate the mouse skin . TCR stimulation of adult γδ thymocytes also induced Egr3 that promoted Tbx21 (and thus Ifng) at the expense of Rorc, Sox13, and Il17a expression . Although these data associate TCR signals with the expression of type 1 transcription factors, which are consistent with the model initially proposed by Chien and colleagues [28, 60], the role of the TCR in the regulation of functional γδ thymocyte differentiation is complex  and warrants further investigation.
Recent data also seem to point at a role for TCR signaling in the type 17 differentiation pathway . Consistent with this, IL-17A-producing γδ thymocytes are enriched in factors shown to be induced by TCR signals, such as ThPOK  or Blk . Similar to Th17 cells, some IL-17A-producing γδ T cells can also produce IL-22 in the periphery . However, IL-22 protein is not detected in γδ thymocytes (K. Serre and B. Silva-Santos, unpublished results); instead, it is induced by inflammatory IL-1β/ IL-23 signals and further enhanced by retinoic acid . As with CD4 T cells and innate lymphoid cells, IL-22 production by γδ cells is strictly dependent on the transcription factor AHR, which serves as a receptor for toxins such as dioxin [26, 64]. Interestingly, the environmental sensing of such toxins is critical for the expansion and maintenance of intraepithelial γδ cell populations in the murine skin and gut .
For the third functional wave of γδ cell development, namely γδ NKT cells, development and acquisition of effector functions rely on similar factors as for αβ NKT cells, in particular transcriptional regulator promyelocytic leukemia zinc finger  and the Tec family tyrosine kinase Itk . It is thus not surprising that these innate γδ NKT and αβ NKT cells exert similar functions and compete for a thymic niche . As recently reviewed elsewhere , both these NKT cell subsets are capable of producing IL-4 or IFN-γ depending on their stage of differentiation.
Functional differentiation of human γδ cells
While the studies cited above have been performed with murine γδ cells, various groups have addressed the functional differentiation of human γδ cells, mostly focusing on the dominant Vγ9(+)Vδ2(+) γδ T-cell subset in the peripheral blood. As in mice, γδ T cells are the first functional population of circulating T cells in humans [70, 71]. For example, preterm babies harbur significant proportions of IFN-γ(+) γδ (but not αβ) T cells in the blood  and CMV infection in utero promotes the differentiation of IFN-γ(+) and perforin(+) γδ cells . Moreover, IFN-γ(+)/perforin(+) cord blood γδ cells are restricted to CMV-positive neonates . Thus, the history of infections seems to be tightly linked to the differentiation (and expansion) of cytotoxic and IFN-γ-producing γδ cells that are abundant in the human peripheral blood throughout life [42, 70].
By contrast, CD161 identifies IL-17A-producing RORγt(+) T-cell subsets already in umbilical cord blood . Although IL-17A(+) cells are very rare (<1%) among γδ cells in the blood of “healthy” individuals [42, 73, 74], they can accumulate following severe bacterial  or viral  infection. Moreover, IL-17A-producing γδ cells are selectively abundant in the skin of psoriatic patients  and in the cerebrospinal fluid of patients with multiple sclerosis .
Morita and colleagues  have assessed the cytokine requirements for the differentiation of human IL-17A(+) Vγ9(+)Vδ2(+) cells activated in vitro with their specific TCR agonist, HMB-PP. In addition to the expected role of TGF-β and IL-1β, they observed an intriguing dichotomy concerning IL-6 and IL-23: the first was required for differentiation of neonatal γδ17 cells (from mostly naive precursors in umbilical cord blood), whereas the latter promoted the generation of adult IL-17A-producing γδ T cells (from cells with a memory-like phenotype in the peripheral blood) . However, Vermijlen and co-workers  showed that IL-23 promoted the induction of IL-17A in neonatal (but not adult) γδ cells stimulated with zoledronate (which was more efficient than HMB-PP at stimulating γδ cells). Finally, in the study that reported the highest (∼35%) degree of polarization of adult Vγ9Vδ2 cells toward an IL-17A(+) phenotype, Caccamo et al.  demonstrated synergistic effects of combining TGF-β, IL-1β, IL-6, and IL-23 (on naive cells stimulated with isopentenyl pyrophosphate). Future studies should clarify if such discrepancies were due to the distinct activation regimens, or due to the intrinsic differences in the starting populations used in these assays.
More in-depth discussions on the differentiation of human peripheral Vγ9Vδ2 cells into additional cytolytic, regulatory, TFH, Treg, Th1, and Th2 functional subsets are discussed elsewhere (reviewed in [79, 80]).
It is clear that γδ cells comprise diverse functional subsets. The current view in the field suggests that functional potential in murine γδ cells associates with Vγ usage, whereas that in humans correlates with Vδ usage. Here, after reviewing the underlying literature, we advocate that selection of TCRγδ with specific V-elements may be less influential for the differentiation of functional subsets than other parameters such as the presence of TCR ligands, the thymic microenvironment, cytokines, and epigenetic factors. Rather, recent evidence suggests that distinct functional subsets each comprising diverse TCRγδ are generated during successive waves of γδ T-cell development defined by certain windows of opportunity. Within each window, a combination of progenitor-specific precommitment, TCR rearrangement, exposure to TCR ligands, and thymic-specific factors dictates developmental fate. Understanding how these windows function and how they relate to one another is the next major challenge for the field.
Finally, it will be important to investigate to what extent the developmental rules dissected in mouse models apply to human γδ T cells. This is particularly relevant given the divergence between the TCR-γ and TCR-δ repertoires of rodents and primates, and the potential (unknown) differences in the respective thymic microenvironments. Thus, future studies with human γδ thymocytes will be instrumental to identify molecular determinants of functional differentiation and to translate this knowledge into novel γδ T-cell-based immunotherapies.
We thank Karine Serre, Julie Ribot, Nina Schmolka, Jan Haas, Andreas Krueger, Gleb Turchinovich, and Adrian Hayday for helpful discussions. Many findings discussed here were only possible through the generous distribution of mAb for specific mouse γδ TCR segments, in particular through Pablo Pereira and Bob Tigelaar. This work was supported by The Wellcome Trust (DJP), EMBO Young Investigator Program and European Research Council StG_260352 (BS-S), and by the Deutsche Forschungsgemeinschaft DFG PR727 4–1 and DFG PR727 5–1 (IP).
Conflict of interest
The authors declare no financial or commercial conflict of interest.