Innate-like lymphocytes are a subset of lymphoid cells that function as a first line of defense against microbial infection. These cells are activated by proinflammatory cytokines or broadly expressed receptors and are able to rapidly perform their effector functions owing to a uniquely primed chromatin state that is acquired as a part of their developmental program. These cells function in many organs to protect against disease, but they release cytokines and cytotoxic mediators that can also lead to severe tissue pathologies. Therefore, harnessing the capabilities of these cells for therapeutic interventions will require a deep understanding of how these cells develop and regulate their effector functions. In this review we discuss recent advances in the identification of the transcription factors and the genomic regions that guide the development and function of invariant NKT cells and we highlight related mechanisms in other innate-like lymphocytes.
Innate-like lymphocytes (ILLs) are populations of lymphoid cells that exist in a primed- or memory-like state and are easily triggered to perform their immune effector functions without prior pathogen exposure. There are many different types of ILLs including NK cells, innate lymphoid cells (ILCs), TCRγδ cells, and subsets of T cells that undergo selection on unconventional MHC molecules such as invariant NKT (iNKT) cells and mucosal-associated invariant T (MAIT) cells (1). ILLs exist as multiple effector cell types that function as a first line of defense against bacteria, viruses, helminths, and fungi (2, 3). These cells also function in antitumor immune responses and tissue repair. Given these properties, as well as their lack of recognition of polymorphic MHC Ags, ILLs may be useful as therapeutics in multiple settings such as vehicles for delivery of chimeric Ag receptors (4). However, although ILLs are generally host protective, their activity can lead to severe pathology (5). Therefore, a better understanding of how these cells are regulated during their development and activation is required to harness their therapeutic potential and to limit disease. Given the large number of ILLs, in this review we focus on the transcriptional and genomic events that control the development of iNKT cells while highlighting unique or overlapping features with other ILLs.
Transcriptional programing of ILL effector fates
Many ILLs, including iNKT cells, ILCs, and subsets of γδTCR+ T cells, differentiate into effector cell types that parallel those of CD4 or CD8 T cells with respect to their transcription factor requirements and the cytokines that they secrete (Fig. 1) (2, 6). Many transcription factor families direct differentiation, survival, and commitment to these effector fates with commonalities but also many differences between distinct ILL programs, and therefore they are discussed only briefly here. iNKT cells and MAIT cells develop from CD4+CD8+ (double-positive [DP]) thymocytes upon rearrangement of the appropriate TCR; Vα14-Jα18 paired with Vβ2, Vβ7, or Vβ8.1–8.2 for iNKT cells in mice and Vα24-Jα18 paired with Vβ11 in humans, or Vα19-Jα33 for MAIT cells in mice and Vα7.2-Jα33 in humans paired with diverse Vβ-Jβ-chains (7, 8). After selection, these cells upregulate the zinc finger and BTB domain containing protein 16 gene (Zbtb16) encoding the promyelocytic zinc finger protein transcription factor (PLZF) (7). For iNKT cells, PLZF induction occurs early in development, as stage (ST)0 cells become ST1 cells, whereas for MAIT cells PLZF expression is delayed and occurs after upregulation of the activation-associated marker CD44 (8–11). In both cell types effector fate specification initiates in these immature PLZF+ cells (3). ILC effector specification similarly occurs in PLZF+ ILC progenitors (ILCPs) that arise from common lymphoid progenitors (12). The development of ST1 iNKT cells requires the Wnt signaling-associated transcription factors TCF1 and LEF1, where LEF1 directly regulates the receptor for IL-7 (Il7ra) and c-Myc to support expansion of these cells (13–15). The development of ILCPs also relies on TCF1/LEF1 family transcription factors and on Il7r, suggesting a pathway related to that of iNKT cells (16).
The iNKT1 and ILC1 effector fates are defined by expression of the T-box transcription factor TBET and production of IFN-γ (Fig. 1) (17, 18). These IFN-γ–producing cells function in response to intracellular pathogens, including some bacteria and viruses (5). NK cells are aligned with iNKT1 and ILC1 in that their terminal maturation and migration requires TBET and they produce IFN-γ, and all of these cells are capable of cytotoxic responses (19, 20). However, the NK cell gene program is more highly related to that of CD8 T cells with a maturation program paralleling a CD8 T memory precursor to effector transition (21, 22). The similarity between NK cell and CD8 T effector programs is further strengthened after murine CMV–induced activation of NK cells (23). NK cells and iNKT1 cells are distinguished from ILC1 by expression of the T-box transcription factor EOMES, which plays a major role in NK cell development and is at least partially required in peripheral iNKT1 cells (19, 24–26).
High expression of GATA3 defines iNKT2 and MAIT2, which make IL-4 and IL-13 (3, 27). ILC2s also rely on GATA3 and retinoic acid–related orphan receptor (ROR)α, and they secrete IL-5, IL-13, and the tissue repair factor amphiregulin (28, 29). These cells function in antihelminth responses but can also contribute to asthma (30). The transcription factor RORγt defines IL-17–producing iNKT17 and MAIT17 cells and IL-17– or IL-22–producing ILC3 (6), which function in antibacterial and antifungal immunity (12). These cells have been implicated in multiple diseases including various autoimmune conditions and inflammatory bowel disease such as Crohn’s disease (31–33). RUNX1 also plays an important role in iNKT17 effector differentiation, controlling expression of several iNKT17 transcription factors, whereas RUNX3 appears to be more essential for ILC3 (34, 35). Most MAIT cells in mice are RORγt+ IL-17 producers (11); however, there are significant populations that produce IFN-γ in humans where MAIT cells are more prevalent (36).
There are transcription factors that influence the development of more than one effector fate. For example, both iNKT17 and iNKT2 cell development are dependent on β-catenin, a coactivator for TCF1 and LEF1 (37). BCL11b and FOXO1 are both required for iNKT1 and iNKT2 development (38, 39). In contrast, in ILC, BCL11b plays a central role in maintaining ILC2 numbers and enforcing lineage commitment but is not required in ILC1, although recent studies implicated BCL11b in NK cell development in humans (40–42). Moreover, in MAIT cells, BCL11b functions upstream of PLZF to promote MAIT17 differentiation (43). In MAIT17 cells BCL11b is bound to many of its iNKT1 cell target genes, indicating that its ability to activate or repress its targets may be highly context-dependent.
iNKT cell effector cell numbers and frequency are influenced by genetic background and tissue type. iNKT1 cells are the major iNKT cell subset in the thymus, but thymic iNKT2 cell numbers can vary in different mouse strains (3). iNKT1 cell are also the major subset in the liver, lung, and spleen in C57BL/6 mice whereas iNKT17 cells dominate in lymph nodes, with the exception of mesenteric lymph nodes (reviewed in Ref. 5). iNKT2 cells are found in most tissues and represent about half of mesenteric lymph node iNKT cells but are rare in the liver (5). Additional iNKT cell effector types have been identified in tissues, including IL-21–producing iNKT follicular helper (iNKTfh) cells that require the transcription factor BCL6 (44, 45). These iNKTfh cells participate in both cognate and noncognate interactions with B lymphocytes and support early germinal center formation and long-lived plasma cells, respectively. Interestingly, cognate interactions between iNKT cells and B lymphocytes promote their production of the immunosuppressive cytokine IL-10 (46). There is also a “regulatory” IL-10–producing iNKT subset (iNKT10) that expresses the transcription factor E4BP4/Nfil3 that was identified in adipose tissue (47, 48). E4BP4 is expressed at the inception of ILC and NK cell development and in many ILC effectors but to date, the identity of IL-10–producing or Tfh-like ILCs has been controversial (12, 49, 50). However, NK cells with Tfh-like features have been identified and these cells can influence B cell responses (51, 52).
ILL cells acquire an activated genomic context during their development
The mechanisms and kinetics of ILL activation differ from their adaptive counterparts. Adaptive lymphocytes detect pathogens through highly variable Ag-specific receptors that are rare and require multiple costimulatory signals to initiate their expansion and activation. In contrast, ILLs can be triggered directly through proinflammatory cytokines or receptors that are broadly expressed on most cells. All iNKT cells express a similar semi-invariant TCR that recognizes a limited range of glycolipid Ags in the binding pocket of CD1d (53). Similarly, MAIT cells recognize vitamin B–derived metabolites in the context of MR1 using a conserved TCRα-chain (8). ILCs and NK cells do not express a TCR but they can be directly activated by proinflammatory cytokines, and NK cells express a battery of activating and inhibitory receptors that control their activation (54).
How can ILLs respond so rapidly when conventional T cells require a longer period of time? The answer to this question lies primarily in the genomic context of the effector genes, a context that arises during the development of ILLs but only after activation of adaptive lymphocytes. Identification of ILC “regulomes” through analysis of RNA sequencing (RNA-seq), ATAC-seq (assay for transposase-accessible chromatin with high-throughput sequencing), and chromatin immunoprecipitation sequencing for the transcriptional coactivator p300 and the activating histone modification H3K27Ac revealed that ILC effector subsets are unique from unstimulated CD4 T cells in having pre-existing genomic accessibility at their effector genes (cytokine loci and cytotoxic genes) (55, 56). However, the regulomes of CD4 Th cells converge with ILCs after their activation (56). Similar observations have been made for mouse and human NK cells, where mature NK cells acquire a gene expression and chromatin accessibility program aligned with effector CD8 T cells, although this is enhanced by viral activation of NK cells (21–23). These data indicate that ILCs acquire a chromatin state during their development that is mirrored by their T cell counterparts only after T cell activation.
Many ILLs live much of their life in one tissue, and therefore similar effector subsets residing in different tissues could have gene programs impacted by local tissue signals. However, iNKT cells with similar effector fates that were isolated from different tissues had highly similar gene programs and chromatin accessibility profiles, suggesting that their gene programs are determined by conserved mechanisms (57). The exception to this homogeneity is in the lung, where iNKT1 cells may be more activated and therefore show a more divergent gene program. In contrast to iNKT cells, ILCs appear to be more strongly influenced by their environment. For example, ILC2s in different tissues do have a core “ILC2” transcriptional program but they show substantial tissue-associated gene programing (58). Indeed, in bulk and single-cell RNA-seq studies, ILC2s were grouped by tissue rather than by cell type (58, 59). Some of these differences may relate to the frequency of progenitors versus more mature cells in different tissues. For example, IL-18R1, which has high levels on lung and skin ILC2s, was associated with a progenitor gene program comparable to ILCPs (59). However, these IL-18R1+ cells are ILC2 restricted and produce IL-13 when stimulated, indicating that they are not immature ILCPs. Another reason for transcriptomic differences could be the developmental origin of the progenitors in different tissues. ILCs can be generated from ILCPs in the fetal liver or postnatal bone marrow, and a subset of lung ILC2s derives from thymic progenitors (60, 61). Thymus-derived ILC2s may have a unique gene program imprinted by the E protein transcription factors that are expressed in these thymic progenitors (62). Similarly, ILC1s in the liver may derive from fetal precursors under the influence of IFN-γ and are distinct from ILC1s in other tissues of adult mice (63). Thus, our understanding of the contribution of intrinsic, developmental, and tissue-derived factors to the transcriptome of ILCs is far from complete. Moreover, how chromatin states and gene expression programs are established requires a closer look at ILL development. As revealed above, for some ILLs this may require a tissue-by-tissue approach that has not yet been achieved, whereas for iNKT cells this can be examined in the thymus.
Induction of PLZF during ILL development
In contrast to ILCs and NK cells, iNKT and MAIT cells arise from DP thymocytes after they have rearranged their Tcrb and Tcra loci. Successful rearrangement of the iNKT cell receptor is dependent on transcription factors that control the survival of DP thymocytes because Vα14 and Jα18 are encoded far apart in the Tcra locus, thus requiring secondary rearrangement, as well as factors that mediate recombination, such as RORγt, RUNX1, E proteins, and ETS1 (64–66). iNKT cell selection occurs on neighboring thymocytes that present glycolipid Ags on CD1d. Therefore, transcription factors that impact genes involved in processing or presentation of glycolipids on DP thymocytes, such as BCL11b and MYB, can severely impact iNKT cell development (67, 68). iNKT cell selection requires coactivating receptors from the SLAM family (CD150/SLAMF1 and Ly108/SLAM6) that signal through SAP to activate the transcription factors EGR2/3 (69–71). NKT receptor signals at this stage are selectively enhanced by SOX4, a high mobility group transcription factor that regulates these signals through induction of Mir181 (72). MAZR, a PLZF-related zinc finger BTB transcription factor, also plays a role at this stage to augment the expression EGR2 and Th-POK, which represses RUNX3 to maintain CD4 and prevent CD8 expression on iNKT cells and it also limits iNKT17 differentiation (73, 74). The outcome of these induced transcriptional programs is downregulation of CD8 and CD24, which identify ST0 iNKT cells, and induction of PLZF, which supports further iNKT cell development (9, 10). The induction of PLZF in iNKT cells is dependent on multiple transcription factors, including EGR2/3, YY1, RUNX1, E proteins, BCL6, and ETS1 (Fig. 2) (75–80). PLZF is expressed most highly in ST1 and iNKT2 cells and is lowest in iNKT1 cells, where Let-7 microRNAs dampen Zbtb16 mRNA (6, 7, 81). The zinc finger transcription factor and adaptor protein HIVEP3 also controls PLZF expression, potentially indirectly through regulation of microRNA processing enzymes (78). MAIT cells and murine NKT-like Vγ1.1Vδ6.4 T cells also express PLZF and require SLAM/SAP signaling for their development (11, 82, 83). Whether the same transcription factors control PLZF expression in ILCPs, where TCR signals are not required, has not been fully explored. In mice, PLZF is not expressed in NK cells, although a small fraction of NK cells are fate mapped by Zbtb16Cre (84). In contrast, human NK cells express PLZF, suggesting that this transcription factor, or related factors, plays a key role across multiple ILL subsets (85).
Genomic regulation of PLZF
Given the central role of PLZF in iNKT cells, MAIT cells, and ILCPs, much focus has been placed on how PLZF is regulated, but little is known about the genomic context that drives PLZF expression. In an elegant study by Mao et al. (77), a combination of ATAC-seq and CRISPR/Cas9-mediated deletion was used to interrogate genomic requirements at the Zbtb16 gene in ILCPs and iNKT cells. This analysis revealed a region (+18–32 kb downstream of the transcription start site) that is required for PLZF expression in both cell types, but does not lead to the skeletal abnormalities observed in mice with germline deletion of Zbtb16 (86). The +18–32 region also impacts PLZF expression in γδ T cells and MAIT cells indicating that this is a pan-hematopoietic, or at least ILL-specific, enhancer region (77). Further interrogation demonstrated that the +21–24 region contained the sequences essential for PLZF expression in ILCPs and was at least partially required in iNKT cells. Interestingly, the +21–24 mutant, and another mutant lacking +40–43, had a reduced number of iNKT cells due to reduced numbers of ST1 cells but had essentially normal PLZF expression in differentiated effector subsets indicating a strong selection for cells expressing PLZF. The +21–23 region contains two RUNX binding motifs that contribute to PLZF expression in ILCPs and iNKT cells. RUNX1 bound at these regions, and multiple other regions, across the Zbtb16 gene in iNKT cells. To examine the necessity for RUNX1, the authors examined mice lacking RUNX1 due to conditional deletion in DP thymocytes and found a loss of PLZF expression in ST0 and ST1 iNKT cells and arrested iNKT development. This block was partially overcome by transgenic expression of PLZF, corroborating that Zbtb16 is an important target of RUNX1. In contrast, deletion of Runx1 in multipotent lymphoid progenitors did not affect ILCP expression of PLZF or development. However, deletion of Cbfb, encoding the essential binding partner for all RUNX proteins, did impact ILCP numbers and expression of PLZF. This study revealed that similar genomic elements control PLZF expression in multiple ILLs but that different members of transcription factor families can regulate these elements in different cell types.
The BTB-POZ transcription factor BCL6 plays a role in setting up the genomic context for iNKT cell development and for the initial expression of PLZF (76). DP thymocytes express BCL6, but it is rapidly extinguished as a consequence of conventional positive selection. In contrast, in iNKT cells undergoing positive selection, BCL6 increases and is not extinguished until PLZF is expressed. Despite its restricted expression, deletion of Bcl6 in CD4+CD8+ thymocytes dramatically affects iNKT cell numbers. Mechanistically, in the absence of BCL6 the boundary between ST0 and ST1 iNKT cells is blurred and ST1 cells continue to express genes that should only be expressed in ST0 and fail to express some genes associated with ST1. Analysis of chromatin accessibility in ST0 and ST1 iNKT cells was consistent with the mixed ST0/ST1 phenotype. Notably, regions of accessibility near some genes that are expressed in ST1 or later in iNKT cell development failed to become accessible in the absence of BCL6, including Zbtb16. Indeed, PLZF target genes are a subset of the genes that fail to be repressed in BCL6-deficient ST1 cells. Interestingly, PLZF expression was reduced and delayed in CD4Cre Bcl6F/F iNKT cells, but some mature iNKT cells emerge that express PLZF, similar to what was seen in mice with deletion of the +21–24 region in Zbtb16 (77). Thus, BCL6 is required to initiate accessibility at key iNKT cell genes, including Zbtb16, and for the repression of PLZF target genes in ST1 iNKT cells. How BCL6 mediates chromatin remodeling requires more investigation, but BCL6 can interact with multiple corepressors that may be relevant to iNKT cells including SMRT/NCOR-HDAC3, EZH2-BCOR-RING3, and the LSD1 histone demethylase (87). Interestingly, HDAC3, similar to BCL6, is required at the inception of iNKT cell development, and although it can be recruited to DNA by the transcriptional repressor NKAP, it is possible that it is also directed by BCL6 (88).
In DP thymocytes, Zbtb16 has both H3K27me3 and H3K4me3 at its promoter, indicating that it is in a repressed, but poised, chromatin state (89). Deletion of EZH2, the histone methyltransferase component of the polycomb repressive complex 2 (PRC2), in DP thymocytes leads to reduced H3K27me3 in iNKT cells and a significant increase in PLZF+ thymocytes (89–91). A similar phenotype is observed in mice lacking JARID2, a component of PRC2 and other histone methyltransferase complexes (92). Many of the PLZF+ cells arising in the absence of EZH2 fail to be detected by CD1d tetramers, indicating that they have an altered TCR repertoire and suggesting that PLZF induction occurs in cells undergoing conventional positive selection. Notably, deletion of the histone lysine demethylases UTX and JMJD3 in DP thymocytes had a negative impact on iNKT cell development and PLZF expression, suggesting that demethylation of H3K27me3 is required (93, 94). However, deletion of EED or SUZ12, two essential components of the PRC2 complex, resulted in a loss of iNKT cells and a more severe loss of H3K27me3 in iNKT cells than observed with EZH2 deletion (95). Thus, EZH2 may have a function distinct from PRC2 in iNKT cells. EZH2 is capable of methylating nonhistone proteins and PLZF contains a functional degron that can be methylated by EZH2 (95, 96). A role for EZH2 in PLZF degradation is consistent with the increased DNA binding by PLZF observed after deletion of EZH2 (97). Although degradation of PLZF may be an important role of EZH2 in iNKT cells, it would not explain the use of alternative TCRs in PLZF+ T cells after EZH2 deletion unless there are cells that initiate and then extinguish PLZF via this mechanism. Thus, the early and/or heightened expression of PLZF must be regulated to ensure appropriate iNKT cell numbers.
In contrast to iNKT cells, early ILC development is not impacted by deletion of EZH2, although there is an effect on NK cell maturation (98–100). Nonetheless, EZH2 is required for repression of Zbtb16 mRNA and multiple ILC genes, including Bcl11b and Gata3, in pro–B lymphocytes (101). Therefore, EZH2 actively restricts the expression of PLZF to appropriate ILL cell types.
PLZF target genes in ILLs
PLZF is sufficient to induce the innate properties of iNKT cells, as revealed in mice with ectopic expression of PLZF starting in CD4+CD8+ T cells (90, 91). In these mice, PLZF promotes a memory-like phenotype on conventional CD4 T cells and the ability to coproduce IFN-γ and IL-4 rapidly after activation. PLZF concentration could affect target site choice or its ability to interact with cooperating factors. PLZF binding sites have been examined in PLZF transgenic mice and revealed that PLZF binds DNA enriched for ETS, E protein, and RUNX family binding motifs, implicating these transcription factors not only in induction of PLZF but also in the recruitment of PLZF to DNA (102). RNA-seq analysis of wild-type and PLZF-deficient iNKT cells, as well as wild-type and PLZF-transgenic CD4 T cells, revealed a small number of PLZF-dependent genes with approximately half being direct targets based on chromatin immunoprecipitation sequencing (102, 103). These targets include genes that are downregulated early in iNKT cell development and encode proteins associated with migration such as Sell and Klf2, as well as Bach2, which codes for a repressor of Th effector fates. Multiple chemokine and cytokine receptors involved in iNKT cell activation are targets of PLZF, including Il12rb, Il18r1, Il4ra, Il21r, and Ifngr1, as are transcription factors involved in iNKT cell effector fate with the exception of Tbx21. PLZF also directly regulates c-Maf, which encodes a transcription factor that is essential for IL-17 production in iNKT17 cells and directly regulates expression of Il4 (103, 104). Therefore, although PLZF has a limited number of targets in iNKT cells, these targets function in the most fundamental aspects of the primed state and effector fate of iNKT cells. The targets of PLZF in ILCPs are currently not known but are likely to overlap with those in ST1 iNKT cells.
PLZF is involved in acute myeloid leukemia where it interacts with multiple corepressors, including N-COR and SMRT (105). In thymocytes, PLZF interacts with Cullin3 (CUL3) and a chromatin-remodeling complex with key components that are ubiquitinylated by CUL3 (106). This complex includes the histone deacetylase HDAC1 and the DNA methylase DNMT1, both of which interact with PLZF in myeloid cells (107). HDAC1 and DNMT1 have not been demonstrated to be essential for iNKT cell development, but it is likely that they contribute to PLZF-dependent gene repression. Another CUL3 target in this complex is SATB1, a matrix attachment region binding protein that functions in chromatin organization (108). Many of the proteins in this complex are ubiquitinylated in PLZF-transgenic mice, indicating that they are directly under the control of PLZF (106). Importantly, deletion of CUL3 in CD4+CD8+ thymocytes limits the development of iNKT cells, indicating that at some point in their development CUL3 plays a central role. Therefore, PLZF may cooperate with many critical iNKT transcriptional regulators to recruit histone and DNA-modifying proteins and promote their ubiquitination. Further investigation is required to determine the consequence of CUL3-mediated ubiquitination on the function of these transcriptional regulators and on PLZF itself.
PLZF activity is also regulated by acetylation (109, 110). Both p300 and HAT1 acetylate lysine residues in the C-terminal zinc finger domain of PLZF to promote DNA binding and gene repression. In myeloid cells, PLZF is acetylated in response to TLR or TNFR signaling, resulting in recruitment of HDAC3 and NF-κB p50 to DNA-bound PLZF to repress the NF-κB response (109). Mutations that inhibit acetylation of these lysines abrogate DNA binding by PLZF. The role of PLZF acetylation has recently been investigated in iNKT cells using a mutant form of PLZF that mimics constitutive acetylation (called PLZFON) (111). In these mice, iNKT1 and iNKT17 cell numbers are decreased and a novel TBET−RORγt− immature cell emerges. These immature cells show reduced proliferation, increased apoptosis, and a heightened capacity to produce IL-4 and IL-13. At the transcriptome level, these immature cells show reduced expression of transcription factors and cytokine receptors associated with iNKT1 and iNKT17 cells and increased expression of known PLZF target genes, including Sell, Bach2, and Bcl6. This gene signature suggests that acetylation of PLZF in iNKT cells interferes with its repressive ability. However, PLZFON iNKT cells showed greatly increased overall CpG methylation compared with controls, which was postulated to be a mechanism to prevent premature differentiation (111). Acetylation is an important mechanism controlling transcription factor function and gene regulation in iNKT cell development beyond PLZF. The histone lysine acetyltransferase GCN5 positively controls the function of EGR2 to promote expression of Zbtb16, Runx1, and Tbx21 while limiting iNKT17 cell differentiation (112). Further studies are needed to clarify the role of acetylation on PLZF or other transcription factors in iNKT cells and other ILLs.
The E protein–ID protein axis
A major distinction between ILLs and their adaptive counterparts is the constitutive versus inducible expression of the E protein transcription factor inhibitors ID2 and ID3 (113). During conventional TCR selection, ID proteins are induced and promote differentiation while enforcing the cessation of rearrangement at Tcra through repression of E proteins (114). When Id2 and Id3 are deleted in DP thymocytes, conventional TCR-mediated selection cannot occur and thymocytes continue to rearrange Tcra until they create a TCR that can undergo unconventional selection, for example the iNKT cell receptor or the MAIT cell receptor (115). iNKT cell selection does not require ID2 or ID3; rather, E proteins participate in the induction of Zbtb16 (79). However, ID2 and ID3 are required for proper iNKT cell differentiation. Failure to express both ID2 and ID3 results in an accumulation of Vα14-Jα18 TCR+ cells with low PLZF and a phenotype consistent with ST0 or ST1 cells, although they produce IL-4 (79, 115–117). ID2 is globally expressed in ILCs and NK cells and its expression marks the most immature progenitors destined for these lineages (118). Moreover, ID2 is required for the development of ILCs and NK cells but it is not required for the initial induction of lineage specification, possibly due to redundancy with ID3 (16, 21, 119).
Id2 is highly expressed in iNKT1 cells whereas Id3 is expressed at much lower levels in these cells (79). Despite this, deletion of Id2 did not affect development of thymic iNKT1 cells (120). In contrast, deletion of Id3 had a major impact on thymic iNKT1 cell numbers (116). Although not formally demonstrated, this observation of high Id2 without a requirement for Id2 could be explained by induction of Id3 by E proteins and functional redundancy between ID2 and ID3. Alternatively, ID3 may be critical for the emergence of iNKT1 cells in ST1. Indeed, Id3 is highly expressed in ST1 cells and required for the emergence of TBET-expressing cells in this population (79, 115). In NK cells, Id2 is required for acquisition of the memory CD8-like fate in immature NK cells, NK cell expansion, and differentiation of cells into cytotoxic effectors, a function that it shares with CD8 T cells (21, 121, 122). However, ID2 mediates these different functions through distinct mechanisms, with effector maturation being dependent on ID2’s ability to dampen E protein function at an intronic enhancer for Tcf7 (encoding TCF1), and expansion being dependent on repression of the suppressor of cytokine signaling gene Socs3 (123, 124). Whether ID2 and ID3 limit Tcf7 transcription early in iNKT cell development remains to be determined, but they do play a similar role in γδ T cells (125). Of note, TCF1 and LEF1 are both required for early iNKT cell expansion and iNKT2 differentiation, and thus overexpression of TCF1 or LEF1, which is also regulated by E proteins, might derail the iNKT1 cell program (13, 14). In mice lacking ID3, iNKT2 cells are expanded in number and produce sufficient IL-4 to induce innate-like CD8 T cells, a phenotype that could be consistent with increased TCF1 or LEF1 (116, 117). Similar phenotypes arise in iNKT cells lacking KLF2 and effectors of TCR signaling such as ITK and RLK, which could impact Id3 expression and therefore E protein function (126–128). Thus, the differential regulation of the ID protein–E protein axis is a key feature of all ILLs and is associated with their acquisition of a primed-effector state. Interestingly, in NK cells and ILC1s, Id2 transcription requires a dedicated tissue-specific cis-regulatory element called Rroid that produces a long noncoding RNA (129). Whether Rroid regulates iNKT1-specific Id2 transcription has not been determined.
DNA modifications in ILL cell development
The ten-eleven translocation (Tet) proteins are essential regulators of iNKT cell development (130). These enzymes remove DNA methylation by mediating the hydroxylation of 5-methylcytosine to 5-hydroxymethylcytosine in DNA, a modification that can be passively diluted during replication, function as a precursor for further replication-independent demethylation, or recruit genomic regulators that function in DNA repair and transcription (131). Deletion of Tet2 and Tet3 in DP thymocytes, and to a lesser degree Tet3 alone, results in more iNKT cells that are skewed toward iNKT17 cells. Specifically, in immature iNKT cells, Tbx21 and Zbtb7, encoding Th-POK, which are associated with iNKT1 and iNKT2, respectively, were not expressed, whereas Rorgt and Lef1 were elevated. These Tet2/Tet3-deficient iNKT cells eventually became transformed (130). iNKT cell lymphomas are also observed with age in iNKT cells lacking ID2 and ID3, underscoring the necessity for transcriptional homeostasis in suppressing transformation (132).
The development of ILLs requires the coordinated activity of multiple genome-modifying proteins and transcription factors to ensure their homeostasis and function in multiple tissues. Using iNKT cells as prototypical ILLs, we have examined the genomic and transcriptional programs controlling the selection and effector differentiation of iNKT cells, as well as where these programs are conserved across innate-like T cells, ILCs, and NK cells. With the application of single-cell and spatial technologies to developing ILLs, we anticipate that our understanding of how these cells develop and interact with their environment will expand dramatically during the next few years. Undoubtedly, these insights will contribute to the utility of these cells as therapeutics against cancer and pathogen infections. Application of these technologies to developing ILLs in the face of infection will also be important to understand how lineage plasticity contributes to the protective or pathogenic functions of ILLs. Recent studies have demonstrated plasticity between ILC3s and ILC2s with ILC1s, as well as between ILC3s with ILC2s, but whether this is a feature of all ILLs, or even adaptive lymphocytes, requires more investigation (133). Activation of iNKT1 cells allows these cells to access the iNKTfh cell program, suggesting that at least some degree of plasticity exists in innate-like T cells (57). A better understanding of the transcriptional programs that control this plasticity may allow us to modulate the effector fates of ILLs in the context of tumors and inflammation.
We thank the Kee Lab for helpful discussions.
This work was supported in part by National Institutes of Health/National Institute of Allergy and Infectious Diseases Grants R01 AI123396 and R01 AI106352.
Abbreviations used in this article:
innate lymphoid cell
iNKT follicular helper
mucosal-associated invariant T
promyelocytic zinc finger protein transcription factor
polycomb repressive complex 2
retinoic acid–related orphan receptor
zinc finger and BTB domain containing protein 16 gene
The authors have no financial conflicts of interest.