Invariant NKT Cells and Control of the Thymus Medulla

The thymus is a primary lymphoid organ that is specialized in its ability to support T cell development. As the thymus contains no long-term hemopoietic stem cell populations, T cell development depends upon the continuous importation of lymphoid progenitors from the bone marrow via the circulation (1, 2). Although T cell development represents a complex and multistage process, it can be simplified and measured by defined changes in cell surface phenotype that take place in developing thymocytes. Such a developmental program is perhaps most readily evident from analysis of conventional αβ T cell development. For example, early T cell progenitors that lack expression of CD4 and CD8 undergo maturation into CD4⁺CD8⁺ intermediates, which is then followed by the generation of both MHC class I–restricted CD8⁺ and MHC class II–restricted CD4⁺ αβ T cells that represent essential cellular components in immune responses to invading pathogens (3, 4). Importantly, analysis of the stages in conventional αβ T cell development in relation to their positioning within intrathymic microenvironments has uncovered important information about the roles of defined thymic stromal cells in this process. Thus, development of cortex-resident CD4⁻CD8⁻ and CD4⁺CD8⁺ thymocytes involves signals from cortical thymic epithelial cells (cTEC), whereas in the medulla, interactions between CD4⁺ and CD8⁺ single-positive (SP) thymocytes with medullary thymic epithelial cells (mTEC) are important (5, 6).

Collectively, these observations fit well with the idea that anatomical compartmentalization within the thymus exists to support stepwise stages in conventional αβ T cell development, which is further supported by conventional αβ T cells being the dominant lineage produced during thymopoiesis. Interestingly, however, the thymus also supports the development of other αβ T cell lineages that branch off from mainstream conventional thymocytes yet retain the requirement for particular thymic microenvironments for their development. For example, CD4⁺CD8⁺ thymocytes expressing the Vα14⁺ invariant αβ TCR that recognize glycolipid/CD1d complexes represent progenitors of invariant NKT (iNKT) cells (7), with accumulating evidence indicating that these cells require and influence medullary thymic microenvironments (8–10). In this review, we summarize the role of the thymus medulla in αβ T cell development, focusing in particular on emerging evidence that indicates the importance of interplay between innate and adaptive αβ T cells within this site.

Following low-affinity αβ TCR engagement in the cortex, positively selected CD4⁺CD8⁺ thymocytes undergo a program of differentiation and guided migration, resulting in the generation of CD4⁺ and CD8⁺ thymocytes that reside in medullary thymic regions. The migration relies upon chemokine ligand responsiveness, typified by thymocyte upregulation of CCR7 and migration toward CCL21 produced by mTEC (11). Notably, entry of conventional αβ T cells to the medulla drives several key developmental processes, including mechanisms of central tolerance prior to T cell export into peripheral tissues.

In addition to the clonal deletion of potentially autoreactive T cell clones via the combined action of mTEC and dendritic cells, the thymus medulla supports regulatory T cell (T-Reg) development (12). Such intrathymic skewing of CD4⁺ αβ T cells toward the T-Reg lineage is associated with the upregulation of Foxp3 and acquisition of suppressive functions (13). The acquisition of effector function by T-Reg prior to thymic export stands in contrast to the process for conventional αβ T cells. Although conventional αβ T cells undergo a process of progressive maturation during their medullary residency, associated with a gain in proliferative response to TCR triggering and capacity for cytokine secretion (14, 15), they are exported from the thymus in a naive “vanilla” state, only gaining specific effector function following peripheral T cell priming. Although thymic T-Reg are arguably the most well-defined subset of intrathymically generated diverse αβ T cells that acquire functional lineage specification prior to thymic exit, the thymic medulla also represents a critical developmental locale for the formation of additional natural T cell subsets, including thymus-dependent RORγt⁺ CD4⁺ Th17 and eomesodermin⁺ CD8⁺ memory-like T cells (16–18), as discussed below. The significance of preprogramming T cell subsets prior to thymic exit likely corresponds with the ability of such subpopulations to rapidly exert effector functions following peripheral stimulation in an innate-like fashion. However, that the majority of TCR-diverse conventional αβ T cells exit the thymus in a base, naive state presumably highlights the functional importance of possessing flexibility in effector function, which allows an effective T cell response to be tailored toward defined pathogenic challenges. The beneficial nature of this process likely outweighs the negative impact of the time lag associated with peripheral T lineage effector programming, with this drawback at least in part being balanced by the rapid activity of both innate and innate-like systems of immunity, including αβ T cells that undergo naturally acquired effector lineage specification prior to thymus egress.

In addition to supporting the maturation of diverse, adaptive αβ T cells, the thymus supports the development of innate-like αβ T cells, including iNKT cells and mucosal-associated invariant T (MAIT) cells. In line with conventional αβ T cells, iNKT and MAIT cells undergo positive selection within cortical compartments of the thymus. However, in contrast to positive selection of conventional αβ T cells by self-peptide/MHC complexes on cTEC, iNKT and MAIT cells undergo positive selection via their respective interactions with CD1d or MR1 MHC class I–like molecules expressed by CD4⁺CD8⁺ thymocytes (7, 19–21). In addition, the selection of CD4⁺CD8⁺ thymocytes toward the iNKT and MAIT cell lineages involves semi-invariant TCR usage, which in mice is characterized by Vα14-Jα18 and Vα19-Jα33, respectively (22, 23). Interestingly, semi-invariant innate-like MAIT and iNKT cells can acquire effector function prior to thymic export. This process is characterized by the expression of defined transcription factors, such as RORγt and Tbet, and associated capacity to secrete effector cytokines, including IL-17 and IFN-γ (24, 25).

The bidirectional dependency of mTEC and SP thymocytes is well studied, with early reports highlighting the loss of mTEC compartments in mice lacking mature thymocytes (26). More recently, the molecular basis for such developmental cross-talk between mTEC and conventional SP αβ T cells was shown to include signaling via TNF receptor superfamily (TNFRSF) members expressed by mTEC, including RANK, CD40, and LTβR, and the provision of TNFSF ligands by CD4⁺ thymocytes (27–31). In parallel with conventional thymocyte/mTEC cross-talk, quantitative loss of mature mTEC in adult CD1d-deficient mice (8) indicates that mTEC development is additionally supplemented by the provision of RANKL by iNKT cells. Whereas these findings suggest both conventional and innate-like αβ T cells act cooperatively to condition medullary microenvironments in the postnatal thymus, in the embryo, the mTEC compartments are regulated by distinct innate lymphoid lineages, including RORγt⁺ lymphoid tissue inducer cells and thymic Vγ5⁺ dendritic epidermal T cells (32, 33). Although such data highlight distinct cellular mechanisms for fetal and adult cross-talk involving mTEC, the molecular basis for mTEC maturation via TNFRSF signaling appears to be a process that is conserved in pre- and postbirth stages. At a functional level, the conditioning of medullary microenvironments at fetal stages by innate-like cells may be a contributing mechanism that helps to pre-establish mTEC compartments, at least to a certain degree, prior to the first cohorts of conventional SP thymocytes transiting through the thymus.

Intact mTEC compartments are critical for the development of thymic Foxp3⁺ T-Reg (12), and similarly, thymic maturation of iNKT cells is also dependent upon the presence of the thymus medulla (8). Although the requirement for mTEC in Foxp3⁺ T-Reg development is at least in part dependent on MHC presentation of self-peptides (34), the importance of iNKT TCR–driven interactions with CD1d molecules in medullary microenvironments remains to be fully determined. Although the role of TCR triggering for mTEC-dependent iNKT cell maturation is uncertain, the requirement for iNKT mTEC dependency can, at least in part, be substituted by supplementation of mTEC-deficient mice with IL-15–IL-15R complexes, suggesting a dominant role for mTEC provision of cytokines for quantitative iNKT cell development (8). However, whether IL-15–IL-15R supplementation in the context of mTEC deficiency leads to the re-establishment of qualitatively normal iNKT cell subpopulations remains an open question. In an additional layer of complexity, recent studies have highlighted that type I IFNs are critical for the development of thymic effector-primed CD8⁺ eomesodermin⁺ T innate memory cells (T_IM) in adult mice (35). Given that mature mTEC provide a constitutive source of type I IFN (36, 37) and mTEC themselves are regulated by both conventional thymocytes and iNKT cells, the development of effector-primed diverse αβ T cells would also appear to form part of this complex medulla-dependent developmental network (Fig. 1).

Although initial reports suggested that T cells expressing Vα14⁺ TCR transcripts characteristic of iNKT cells were present early in embryonic development in the embryonic body, yolk sac, and fetal liver prior to their detection in the thymus (38), it is now clear that conventional αβ T cells and iNKT cells primarily develop intrathymically from CD4⁺CD8⁺ progenitors (39). However, recent studies have also identified an alternative developmental pathway in the thymus in which iNKT cells stem directly from CD4⁻CD8⁻ thymocytes, providing evidence for developmental heterogeneity during iNKT cell maturation (40). The thymus dependency of iNKT cells is clear from studies demonstrating their absence in the spleen and liver of nude mice and thymectomized mice (41, 42). Previous discrepancies were perhaps due, at least in part, to technical limitations in the tools and mouse strains used to study iNKT cells. More recently, however, the introduction of experimental approaches using tetramer reagents (composed of glycolipids loaded onto CD1d molecules) has facilitated the accurate identification of these cells (43). Indeed, and consistent with their intrathymic origin, use of this approach has shown that CD1d tetramer⁺ iNKT cells can be first identified in the thymus at postnatal day 5 and subsequently appear in peripheral tissues, such as the liver and spleen, by postnatal day 8 (41).

In relation to developmental progression of iNKT cells in the thymus, early studies used the markers NK1.1, CD44, and CD24 to establish a linear model in which distinct developmental stages are defined by differential expression of these markers (22). In this model, CD1d tetramer⁺ stage 0 iNKT cells can display a potential combined phenotype characterized as CD24^highCD44^lowNK1.1⁻CD69⁺CCR7⁺ and Erg-2^high (44, 45). A CD24⁻CD69⁻CD44^lowNK1.1⁻ phenotype was then used to define stage 1 iNKT cells, which could be further divided into IL17RB⁻ and IL17RB⁺ cells (46), whereas stage 2 and stage 3 cells were defined as CD24⁻CD69⁻CD44^hiNK1.1⁻ and CD24⁻CD69⁻CD44^hiNK1.1⁺, respectively (22, 47). Importantly, however, combined analysis of transcription factor and cytokine production capabilities in CD1d tetramer⁺ cells has recently shown that the thymus generates multiple iNKT cell sublineages that cannot be explained by a linear model of development. Thus, within CD1d tetramer⁺ cells, T⁻bet⁺ cells that produce IFN-γ were defined as iNKT1; RORγt⁺ cells producing IL-17 were iNKT17; and GATA3⁺ cells producing IL-4 and IL-13 were iNKT2 (24, 48). Based on this definition, detailed molecular analysis of each of the effector iNKT cell populations has been performed (49, 50), and overlapping linear model stages were highlighted such that NKT-1 cells were most like stage 3, NKT progenitors (NKTp) and NKT17 cells were most like stage 2, and finally, NKT2 cells were most like both stage 1 and 2 (49). Importantly, and in line with the importance of mTEC during iNKT cell development, direct visualization of CD1d tetramer⁺ cells in thymic tissue sections demonstrated that most iNKT1, iNKT2, and iNKT17 cells locate to the thymus medulla (51). However, it is also interesting to note that some iNKT cells are detectable in thymic cortical areas, and at present, the functional significance of this is not known. In relation to medulla-resident iNKT cells, it is also not clear whether this localization, including the potential for interactions with differing medullary stromal cells, dictates their expression of cytokines and/or function.

Although these studies provide a better understanding of intrathymic iNKT cell heterogeneity in the thymus, the nature of iNKT cell progenitors that give rise to such distinct iNKT cell lineages and the signals that drive their continued development and their migration into the medulla remain poorly defined. As mentioned earlier, as iNKT cells derive from CD4⁺CD8⁺ thymocytes, preselection iNKT cells may be defined as CD1d tetramer⁺ CD24^hi cells that reside within the thymic cortex (7, 51). Following CD1d recognition, these cells may upregulate their expression of CCR7, a chemokine receptor known to be important in medullary localization of conventional thymocytes (44), together with PLZF, a member of the BTB/POZ-ZF family of transcription factors described as a master regulator of iNKT cell development (52). Indeed, expression of PLZF coincides with downregulation of CD69 and CD24 (52) and is expressed at high levels in CD24^lowCD44⁻NK1.1⁻ cells, with iNKT cell development in PLZF-deficient mice failing to progress beyond this stage (52, 53). Thus, PLZF and CCR7 expression may be a potential means to define iNKT cell progenitors that give rise to all (iNKT1, iNKT2, iNKT17) intrathymic mature iNKT cell lineages.

Interestingly, Lee et al. (49) and Engel et al. (50) produced detailed studies using RNA sequencing of four subsets of iNKT cells. Engel et al. looked at a single-cell level in the four linear subsets at stages 0–3, whereas Lee et al. used Tbet/IL-4 reporter mice to identify and analyze NKTp (PLZF^hiIL17RB⁻IL4⁻ cells), NKT1, NKT2, and NKT17 cells. Both studies highlighted the difference between iNKT cell precursors and iNKT stage 0 and their more mature counterparts. Interestingly, evidence exists for the regulation of iNKT cell precursors by cMyc (49, 50) as well as their inability to produce cytokines (49), unlike their more mature counterparts. Although these studies have provided insight into the properties and developmental requirements of NKTp, further analysis of early stages in iNKT cell development should aid in the identification and characterization of this poorly defined population.

Although iNKT cells have been shown to play important roles in peripheral tissues, emerging evidence also suggests that these cells also display functional properties intrathymically and influence both stromal microenvironments and other innate αβ T cell populations.

T cells are often divided into naive or memory populations based on the phenotypic changes that occur as a result of TCR engagement. For example, naive T cells express the lymph node homing receptors CCR7 and CD62L, and following Ag encounter they express high levels of CXCR3 and CD44 to allow their entry into peripheral tissues as Ag-experienced memory T cells (54, 55). Interestingly, although the majority of T cells in nonimmunized germ-free mice have a naive phenotype, 10–20% of CD8⁺ T cells possess hallmark features of memory cells (56), indicating their presence is not due to exposure to commensal or environmental microbes. Importantly, similar cells have also been described in the thymus and have been termed T_IM, whereas, in peripheral tissues, these cells represent a combination of T_IM and T virtual memory cells (T_vm). Although T_IM and T_vm are phenotypically indistinguishable outside the thymus, making their respective roles a challenge to clarify, thymic CD8⁺ T_IM are characterized phenotypically as CD8⁺CD62L⁺CD44^hiCD122⁺CD24^loCD69⁻CD25⁻eomesodermin⁺Tbet⁻ (57–59).

Intrathymic generation of T_IM was first examined using mice deficient in inducible T cell kinase. These mice exhibited an alteration in the composition of the SP8 population, manifested by increased frequencies of CD44^hiCD25^loCD8⁺ T_IM (57, 58). In addition, mice deficient in several other genes—for example, Kruppel-like factor 2 (KLF2) (60), cAMP-responsive element-binding protein (61), and CD155 and its ligand CD226 (62)—have a disrupted population of CD8⁺ T_IM in the thymus. The involvement of these genes in the generation of PLZF⁺ iNKT cells, or the production of IL-4 by iNKT cells, has helped establish the requirement for NKT2 to generate CD8⁺ T_IM. This was conclusively demonstrated by the absence of CD8⁺ T_IM in iNKT cell–deficient CD1d^−/−, IL-4Rα^−/− (63), and IL-4^−/− mice (64). In addition, the close interplay between iNKT cells and CD8⁺ T_IM is clearly visible in inbred strains of wild type (WT) mice—for example, BALB/c mice, which have a prominent population of PLZF⁺ NKT2 and a corresponding enlarged population of CD8⁺ T_IM (60). PLZF⁺CD4⁺ thymocytes are also generated in the human fetal thymus; however, these cells do not express a restricted αβ TCR, nor do they bind CD1d/α-GalCer tetramers, indicating they are not iNKT cells (65). Within the same gestation period, eomesodermin⁺ CD8⁺ T cells are present in the human thymus (66), and interestingly, the frequency of both cells declines during fetal and early postnatal development. These results indicate a dependency on PLZF⁺CD4⁺ T cells to drive the development of CD8⁺ T_IM in both mouse and human.

Despite difficulties in the identification of T_IM and T_VM as two distinct populations of cells within peripheral tissues, a clear difference is evident from their reliance on cytokines for their generation. As previously discussed, CD8⁺ T_IM depend on IL-4 production for their development. However, CD8⁺ T_VM are present, albeit at a reduced frequency, in IL-4–deficient BALB/c mice (67, 68). In addition, IL-15^−/− mice completely lack CD8⁺ T_VM, and in particular, IL-15 trans presentation by CD8α⁺ dendritic cells has been shown to be required for their development in the periphery (69). Furthermore, the functional capabilities of CD8⁺ T_IM and CD8⁺ T_VM are similar in that both populations can produce IFN-γ upon Ag binding (59, 69); however, CD8⁺ T_VM have been shown to respond to additional stimuli that CD8⁺ T_IM have not. Studies using Nur77-GFP reporter mice, in which GFP expression levels indicate TCR signal strength (70), have shown that CD8⁺ T_VM can produce IFN-γ when they are stimulated by IL-12, IL-15, and IL-18, independently of TCR triggering (71). CD8⁺ T_VM can also elicit cytotoxic responses by their release of perforin and granzyme, features that have not been attributed to T_IM (71, 72).

Experiments involving engraftment of thymic lobes under the kidney capsule of congenic mice have been used as a system to show the rapid export of conventional αβ T cells, followed by repopulation with host-derived T cell precursors (73, 74). This is in contrast to iNKT cells, where in such grafting systems, some donor cells can remain resident for 12 wk following transplantation (73), suggesting their presence in the thymus is long lived. CXCR3 has been identified as an example of a chemokine receptor that is required for this retention of iNKT cells within the thymus. For example, direct injection of FITC into the thymus of WT and CXCR3^−/− mice revealed an increase in FITC⁺ iNKT cell recent thymic emigrants (RTE) in the latter. In the same experiments, numbers of FITC⁺ conventional αβ T cell RTE were unaltered, suggesting that CXCR3 may act to specifically retain iNKT cells in the thymus (75). Although this study provides evidence of a distinct mechanism controlling the egress of iNKT and conventional T cells, similarities do exist. One such similarity is the role of TNFRSF member LTβR, which has been proposed to control the egress of both iNKT cells and conventional αβ T cells (9, 27, 76). Despite no alterations in the frequency of thymic iNKT cells in LTβR^−/− mice, reduced iNKT cell numbers have been described in the liver and spleen (9, 76). To study this further, bone marrow chimeras were generated to restrict LTβR expression to stromal cells. These chimeras recapitulated the phenotype seen in LTβR^−/− mice, providing evidence for a requirement of stromal cell expression of LTβR in the regulation of the peripheral iNKT cell pool. Reduced peripheral expansion was ruled out as an explanation for this phenotype, as transfer of CFSE-labeled congenic CD4⁺ T cells showed the same extent of proliferation in WT and LTβR^−/− mice. Instead, intrathymic injection of FITC into LTβR^−/− mice revealed a reduction in FITC⁺ iNKT RTE compared with WT mice. Interestingly, this study also reported normal numbers of FITC⁺ conventional T cell RTE in LTβR^−/− mice, which contrasts with earlier work indicating the accumulation of mature conventional SP thymocytes in the LTβR^−/− thymus occurs as a result of reduced thymic egress (27). Furthermore, both conventional αβ T cells and iNKT cells express S1PR1 and share S1P-mediated dependency for thymic egress, as seen by a thymic emigration defect in both populations in p56Lck^CreS1P1^fl/fl mice. Interestingly however, unlike conventional αβ T cells, the S1P axis is not required to control the distribution of iNKT cells across peripheral sites (77).

In relation to intrathymic functions of iNKT cells, recent work has demonstrated their importance in the regulation of egress of conventional αβ T cells from the thymus (10). Thus, in CD1d^−/− mice, accumulations of mature CD4⁺ thymocytes within the thymic perivascular space (PVS) were observed. Moreover, mTEC were shown to express the type 2 IL-4R and respond to iNKT cell–derived cytokines IL-4 and IL-13. Indeed, analysis of thymus emigration in IL-4Rα^−/− mice provided evidence for a mechanism of thymic egress in which IL-4/IL-13 production by long-term thymus-resident iNKT cells triggers type 2 IL-4R signaling in thymic microenvironments to control the effective emigration of mature conventional thymocytes (Fig. 2). Interestingly, mature thymocytes from IL-4Rα^−/− mice had an intact thymus exit phenotype, including normal levels of KLF2, CD69, and S1PR1. The latter finding was of particular significance, as it indicated the requirement for type 2 IL-R4 signaling is distinct from that of the S1P–S1PR1 axis. Consistent with this, treatment with the S1PR1 agonist FTY720 promoted further intrathymic accumulation in IL-4Rα^−/− mice. Thus, intrathymic microenvironments appear to regulate thymocyte egress from the adult thymus via at least two separate pathways. First, intrathymic S1P levels are kept low to create an S1P gradient that directs mature thymocytes to blood vessel exit points in the medulla, whereas IL-4/IL-13 from iNKT cells aids in the trafficking of thymocytes from the PVS into the circulation. As mentioned earlier, in addition to their specialized cytokine-producing properties, IL-4/IL-13–producing iNKT cells can remain in the thymus for long periods (10, 73), and this intrathymic retention may provide some insight into their involvement in conventional αβ T cell migration from the thymus. For example, as current evidence suggests that SP thymocytes are 4–5 d old prior to emigration, with an intrathymic conveyor belt ensuring oldest thymocytes exit first (78), the availability of tissue-resident iNKT cells may maintain this process by ensuring ordered access to the PVS around thymic exit points. Although the lower Rag2-GFP levels on IL-4Rα^−/− thymocytes are indicative of prolonged medulla dwell time, further experiments are required to determine whether the ordered program of emigration is altered in iNKT cell–deficient CD1d^−/− mice and IL-4Rα^−/− mice. Significantly, although this study demonstrated a role for IL-4Rα signaling in CD4⁺ thymocyte egress, the defect in eomesodermin⁺ CD8⁺ T_IM development means that the requirement for stromal cell–expressed IL-4Rα in conventional CD8⁺ thymocyte emigration was not assessed. Furthermore, although the downstream regulators that are triggered by IL-4/IL-13 signaling in thymic stroma to regulate thymocyte emigration are not clear, it is interesting to note that type 2 IL-4R signaling in mTEC triggers expression of the chemokines CXCL10 and CCL21 (10). Given that CCL21 production in the thymus occurs within mTEC^low (79) and the localization of IL-4–secreting NKT2 in the thymic medulla (51), it is possible that CCL21 influences the intrathymic positioning of iNKT cells. Finally, as CXCL10 and CCL21 have been shown previously to influence long-term resident iNKT cells and mature conventional thymocytes in mice (75, 80), this may also suggest that type 2 IL-4R signaling controls chemokine availability for innate and adaptive αβ T cells, which influences their respective thymus retention and emigration. In comparison with murine studies, the development and functional significance of iNKT cells in the human thymus is less well defined. Although initial studies suggested that numbers of thymic iNKT cells decline rapidly during gestation (81), subsequent studies using more stringent gating to identify iNKT cells, in conjunction with matched thymus and blood samples, showed the presence of iNKT cells in the postnatal thymus, with their numbers stable in the thymus up to 9 y postbirth (82, 83). Although human iNKT cells undergo maturation in the periphery (82), the functionality of iNKT cells in the human thymus, including their potential for regulation of conventional αβ TCR⁺ thymocyte emigration, has yet to be addressed.

The thymus is well known as an important site for the production of conventional αβ T cells. In this review, we have summarized how the thymus medulla, a key site for central tolerance during conventional αβ T cell development, represents an important feature in the intrathymic development of CD1d-restricted iNKT cells. It is now clear that, as with conventional thymocytes, iNKT cells influence the development and function of mTEC, demonstrating that thymic cross-talk in the adult thymus involves both innate and adaptive αβ T cells. In addition, a clearer definition of intrathymic iNKT cell heterogeneity has helped to show their involvement in the regulation of thymic egress and the intrathymic production of eomesodermin⁺ innate memory CD8⁺ T cells. Further examination of the links between iNKT cells and the thymus medulla will be an important step in understanding the mechanisms that control the development and function of this site.

We thank all laboratory members for helpful discussions during the preparation of this manuscript.

The authors have no financial conflicts of interest.