Immune Modules Shared by Innate Lymphoid Cells and T Cells

Abstract

In recent years, innate lymphoid cells (ILCs) have emerged as innate correlates to T cells. The similarities between ILCs and T cells indicate that lymphocytes of fundamentally distinct lineages can share core “immune modules” that encompass transcriptional circuitry and effector functions, while utilizing non-redundant, complementary mechanisms of pattern recognition to enact these functions. We review modules currently recognized to be shared between ILCs and T cells.

Introduction

The hematopoietic universe has rapidly expanded in recent years with the identification of a new major class of cells found in both human and mouse, called innate lymphoid cells (1–7). ILCs belong to the lymphoid lineage as they derive from the common lymphoid progenitor (CLP) and depend on the master lymphocyte cytokine receptor Interleukin-2 receptor common gamma chain (IL-2rg, γ_c) (2, 4). Yet, their lack of recombined antigen-specific receptors and other lineage markers distinguish them from conventional T and B cells as well as innate-like lymphocytes (ILLs) such as Natural Killer T (NKT) cells (1). Remarkably, the functional specializations of ILCs as well as their developmental programs resemble those previously recognized in CD4⁺ T helper cell subsets (2); therefore ILC1, ILC2 and ILC3 are now considered the innate counterparts of Th1, Th2 and Th17, while natural killer (NK) cells are considered the innate counterpart to CD8⁺ cytotoxic T cells (2, 8). To date, there is no recognized ILC counterpart to regulatory T cells (Treg) or T follicular helper cells (Tfh). Furthermore, unlike CD4⁺ T cell polarization from a multipotent naïve cell in the periphery, ILC fate-commitment occurs early in development (9–11) (Figure 1).

Figure 1. Development of T cells and ILCs.

T cells and ILCs both develop downstream of the CLP. T cell progenitors (left) undergo progressive developmental stages in the thymus, which ultimately result in the export of naïve CD8⁺ and CD4⁺ T cells to the periphery. In contrast, ILC generation (right) occurs in the bone marrow through a series of progenitors with increasingly restricted developmental potential, resulting in the export of lineage-specified ILCs directly to the periphery. The sequence of recognized ILC progenitors are indicated in black in the bone marrow, while presumed progenitors are denoted by question marks.

ILC subtypes are grouped together based on shared developmental requirements and effector functions. Eomesodermin⁺ (Eomes) T-bet⁺ NK cells are cytotoxic and produce IFN-γ and TNF-α; T-bet⁺ ILC1 produce IFN-γ and TNF-α, like NK cells; GATA3^hi ILC2 produce IL-5, IL-13, IL-9 and amphiregulin; and RORγt⁺ ILC3 produce IL-22, GM-CSF and/or IL-17.

ILC1 and NK cell distinctions are controversial because, as developmental studies in mouse separate these lineages, but their defining cell-surface markers vary between tissues and often overlap (12–15). Thus, mouse ILC1 are currently best defined by their lack of Eomes expression within NKp46⁺ NK1.1⁺ cells (15). In human, similar phenotypic heterogeneity in the absence of lineage tracing makes distinctions even more difficult, but two IFN-γ-producing subsets have currently been described: Intraepithelial ILC1 express a CD103⁺CD56⁺CD94⁺NKp44⁺ phenotype, while CD127⁺ILC1 express a CD127⁺CD161⁺ phenotype and do not express intraepithelial ILC1 markers (16–18).

ILC2 have been identified in many organs in mice and human during homeostasis. Cells from both species express CD25, CD127, the IL-33 receptor ST2, and the inhibitory receptor KLRG1. Human ILC2s along express the prostaglandin D2 receptor-homologous molecule expressed on Th2 cells (CRTH2) (7, 16, 19, 20).

ILC3s are a particularly diverse class of cells in both species. In adult mice these cells include two developmentally distinct lineages, one of which has many subsets. The first lineage, LTi-like cells or NKp46⁻CCR6⁺ ILC3 cells, expresses MHC-II and produces IL-22, IL-17a, IL-17f, and LT-α (21–26). The second lineage includes a sequence of cells that progressively differentiate and express different cytokines, cell-surface markers, and variable levels of T-bet: T-bet^lo NKp46⁻CCR6⁻ ILC3 express IL-22 and IL-17a (27, 28); T-bet⁺ NKp46⁺CCR6⁻ ILC3 express IL-22 and GM-CSF (24, 27–32); and T-bet⁺NK1.1⁺NKp46⁺ “ex-RORγt” ILC3 produce IFN-γ and cannot be distinguished from ILC1 without lineage tracing (13). Recently, single-cell sequencing of human ILC3 has similarly identified 3 cellular clusters based on shared and distinct transcriptional programs, including NKp44⁺ ILC3, a CD62L⁺NKp44⁻ ILC3, and HLA-DR⁺ LTi-like ILC3, reminiscent of diversity found among mouse ILC3 (33). For additional phenotypic information, the cell-surface expression profiles of human (5, 16) and mouse (5) ILCs have recently been reviewed.

Unlike other lymphocytes, ILCs respond directly to cytokine signals in the microenvironment without a need for antigen-specific receptor signaling or preactivation. Signaling specificity is regulated by the cytokine receptors that ILCs express, allowing distinct combinations of myeloid and/or tissue-produced cytokines to activate each ILC class. Although ILCs are a relatively small population of cells compared to adaptive lymphocytes, they appear to play major homeostatic roles including through crosstalk with the epithelium. In mouse, ILCs are now recognized to constitutively produce “adaptive” cytokines including IL-22, GM-CSF, and IL-5 at greater levels during steady-state than conventional T cells (7, 24, 34–38). The similarities between ILCs and T cells indicate that lymphocytes of fundamentally distinct lineages can share core “immune modules” that encompass transcriptional circuitry and effector functions, while utilizing non-redundant, complementary mechanisms of pattern recognition to enact these functions. We review modules currently recognized to be shared between T cells and ILCs.

Cytotoxicity

NK cells and CD8⁺ T cells are grouped together in the same immune module based on their shared function as professional killers. A core feature of this module is the presence of granular cytotoxicity, characterized in mouse and human by lytic granules that contain perforin and granzmes, which act together on target cells to induce cellular apoptosis (Figure 2). In human, cytotoxic granules also contain an additional anti-microbial factor, granulysin (39). Cells in this module can also kill and induce other effector functions through shared expression of Fas ligand (FasL), TRAIL, TNF-α, and IFN-γ (40). Upstream, cytotoxic cells are activated by IL-12, IL-18, and IL-15.

Figure 2. Effector Modules of T cells and ILCs.

In the periphery, naïve T cells (left) that recognize their specific antigen and are costimulated by antigen presenting cells (APCs) become activated and mature to effector cells. For CD4⁺ T cells, APC-derived cytokines drive STAT activation and TF induction, leading to Th polarization and the indicated effector functions. Meanwhile, ILCs in the periphery are already mature cells that do not require STAT activation to develop, though STATs are important for some ILC effector functions. The TFs required for Th polarization and ILC development and the effector functions of cells in the same module are strikingly similar. ILC counterparts to Tfh and Treg are not currently recognized.

The TFs Eomes and T-bet are essential regulators of the Cytotoxicity Module in mouse and likely also in human based on their conserved pattern of expression (41) (Figure 2). As members of the same T-box family of TFs, Eomes and T-bet have somewhat redundant functions and can partially compensate for each other (42, 43). Therefore, the cytotoxic program fails to develop only in the absence of both molecules, leading to the selective ablation of CD122^hi (marking IL-2/IL-15 receptor beta, IL-15rβ) memory CD8⁺ T cells and NK cells, and the aberrant expression of IL-17 by remaining CD8⁺ T cells upon their activation (44, 45).

The regulation and specific functions of Eomes and T-bet appear to differ. In CD8⁺ T cell effector differentiation, T-bet expression temporally precedes that of Eomes (46), with the highest T-bet to Eomes ratio in early effector cells and the lowest ratio in memory cells. While T-bet better induces IFN-γ and the IL-12 co-receptor IL-12Rβ2, Eomes superiorly induces perforin and granzyme B expression (43, 46, 47). Initially, this developmental sequence appeared to be conserved in NK cell differentiation, with T-bet sustaining immature NK cells and Eomes sustaining mature cells (45). Yet, evaluation of mice deficient for T-bet and Eomes in addition to transfer studies have demonstrated that Eomes is uniquely required for NK cell development, while T-bet is in fact required for the development of ILC1 (13, 48, 49). Moreover, when ectopically expressed, Eomes skews ILC1 to a cytotoxic NK cell profile with enhanced Ly49 expression, while T-bet overexpression similarly results in the increased production of Eomes⁻ phenotypic ILC1 (48, 50). These data suggest that Eomes and T-bet might achieve balance through mutual regulation. Developmentally, the exogenous factors that lead to differential regulation of Eomes and T-bet remain an active area of investigation. These factors likely include tissue factors such as TGF-β, as recently demonstrated in a unique ILC1-like population in the salivary gland that is T-bet- and Eomes- independent (51). Nonetheless, in physiologic conditions, both Eomes and T-bet together define the cytotoxic module, while Eomes is particularly critical for NK cells.

Module 1: Intracellular Pathogens and Viruses

ILC1 and Th1 are grouped together based on their shared production of IFN-γ and TNF-α, and their absent or reduced capacity for granular cytotoxicity compared to the Cytotoxicity module, likely due to their absent or reduced levels of perforin and granzyme B (13, 48) (Figure 2). In mouse, some ILC1 additionally appear to share expression of TRAIL (48, 52, 53). Underscoring the similarity in cytokine production with Cytotoxicity, Module 1 is similarly defined by the TF T-bet (48, 49, 54). Cells in this module are also similarly regulated by the cytokines IL-12, IL-18, and IL-15, though in human, CD127⁺ ILC1 preferentially respond to IL-12 and IL-18 while intraepithelial ILC1 respond to IL-15. While IL-12-mediated STAT4 activation causes Th1 polarization in vivo (55), this pathway is not required for ILC1 development (56). However, it does provide a functional proliferative signal to mouse adipose tissue ILC1 in an inflammatory context (56).

Engagement of Module 1 is physiologically associated with viral and intracellular pathogen immunity, but it is deregulated in Crohn’s disease and some forms of autoimmunity such as Type 1 Diabetes (4). Consistent with this, intraepithelial and CD127⁺ ILC1 have been reported to be increased in frequency in Crohn’s disease (17, 18, 57, 58). Increased frequencies of ILC1 have also been reported to correlate with chronic obstructive pulmonary disease (COPD) disease severity, presumably linked to viral infection (59).

T-bet has long been recognized as a key driver of Th1 differentiation, and is more recently recognized as a necessary factor for ILC1 differentiation (13, 48, 49, 54, 60). In the absence of T-bet, naïve T cells do not produce IFN-γ, and instead polarize to a Th2 effector program (61). Indeed, T-bet directly represses the development of Th2 (62, 63). The fate of ILC precursors unable to engage T-bet is currently unclear. Functionally, T-bet drives Module 1 effector cytokine production by binding to the promotor and regulatory elements of Il12rb2 and Ifng (64, 65). It also directly induces the expression of Runx3, an additional Ifng-activating TF. Supporting the role of Runx3 in Module 1, Runx3 deficient mice also fail to generate ILC1 (66).

While T-bet clearly has an important role in Module 1, the role of Eomes is more controversial. In mice, ILC1 are currently best defined by their lack of Eomes expression. However, in the absence of an Eomes fate-mapping mouse, it unclear if ILC1 always lack Eomes expression, or if they may sometimes express it. Recently, assay for transposase-accessible chromatin with high-throughput sequencing (ATACseq) data from mouse ILCs demonstrated graded expression of ATAC⁺ open chromatin surrounding Eomes, with highest expression in NK cells, lower expression in ILC1, and minimal expression in ILC2 and ILC3 (10), suggesting the potential to express Eomes may not be absent in ILC1. Furthermore, the “Eomes negative” definition of this module is not universally true, as some Th1 cells from both mouse and human express Eomes when activated (67, 68). In human, the two identified ILC1 subsets differ in their expression of EOMES and T-BET; CD127⁺ ILC1 exclusively expresses T-BET, while intraepithelial ILC1 expresses T-BET as well as higher levels of EOMES compared to CD127⁺ ILC1 and ILC3 cells (17, 18, 57). Notably, intraepithelial lymphocytes are well known to express highly activated phenotypes in steady-state as they are in direct contact with the epithelium and constitutively sense foreign metabolites and antigens (69). Collectively, these data suggest that cells in Module 1 critically rely on T-bet, likely for this TFs preferential induction of cytokine versus cytotoxic machinery, but may express Eomes in some circumstances, such as activation. Yet, unlike T-bet, Eomes is not required for development of Th1 or ILC1 and is unlikely to be a major functional regulator of this module.

Module 2: Barrier Maintenance and Helminthes

ILC2, Th2, and Th9 share the same functional module based on their mutual production of signature cytokines IL-5 and IL-13 (70) (Figure 2). IL-25, IL-33, and TSLP are conserved cytokine regulators. Other factors associated with this module include IL-4, IL-9, GM-CSF, and the epidermal growth factor family member amphiregulin (71, 72). The γ_c family cytokines IL-4 and IL-9 in particular seem to be differently regulated in this module, as Th2 produce greater quantities of IL-4 (73–75), while ILC2 and Th9 produce IL-9 more readily (72). IL-4 signaling through STAT6 is also an important factor for Th2 and Th9 polarization in vitro, but this pathway is not universally required in vivo and is not required for the development of ILC2 (38, 76). However, STAT6 has a post-developmental role in ILC2 function, as STAT6-deficient ILC2 produce less IL-13 (38). Type 2 cytokines support tissue remodeling and the response to helminth infections, but are deregulated in the development of allergy and asthma (4, 77–79). Indeed, ILC2 are rare population in normal human tissues (33, 57), but are expanded in nasal polyps of allergic patients with chronic rhinosinusitis and the skin of patients with atopic dermatitis (20, 80, 81).

GATA3 is the master TF regulating Module 2 and is required for the development of ILC2 and Th2 (20, 65, 82–84). In both human and mouse, Il4, Il5, and Il13 are located in a gene cluster on the same chromosome and are mutually regulated by the locus control region within an intervening gene, Rad50 (70, 85, 86); Csf2 (encoding GM-CSF) is also located nearby on the same chromosome. GATA3 binds regulatory elements within this gene cluster and in the Rad50 locus control region as well as directly to the promoter of IL-5 and IL-13 (87). Consistent with the minimal expression of IL-4 by ILC2, GATA3 does not appear to bind directly to the IL-4 promoter and conditional deletion experiments demonstrate that it is not required for expression after Th2 development (87, 88). However, GATA3 conditional deletion demonstrates a common role for this TF in Module 2 production of IL-5, IL-13, and amphiregulin (88, 89).

Module 2 is a feed-forward module with a high degree of lineage stability, which in certain conditions, can be overcome to generate plasticity toward Module 1. Feed-forward loops depend on Module 2 cytokines and, at least for T cells, cell-intrinsic signaling. For T cells, IL-4 production by Th2 facilitates the polarization of more Th2 through STAT6 induction of GATA3 (76). Lineage stability is then enhanced by autoactivation of GATA3 once it overcomes a threshold mediated by the transcriptional regulator FOG-1 (90, 91). Although ILCs appear to develop in the absence of polarizing cytokines in the bone marrow, their activation in the periphery is similarly feed-forward. In mouse, ILC2-generated IL-13 acts on the epithelium to drive the differentiation of tuft cells from LGR5⁺ stem cells, which in turn generate IL-25 that activates ILC2 to produce more IL-5 and IL-13 (34–36).

Despite Module 2 lineage stability, strong Module 1 stimuli can induce plasticity. For example, viral infection with lymphocytic choriomeningitis virus (LCMV) causes Th2 to upregulate T-bet and subsequently produce IFN-γ in conjunction with IL-4 (92). Recently, mouse ILC2 were also shown to undergo similar T-bet-mediated plasticity in response to infection with respiratory syncytial virus (RSV), influenza virus, Haemophilus influenzae, and Staphylococcus aureus (59). Likewise, human ILC2 cultured in vitro with the danger cytokine IL-1 upregulate T-bet and IL-12Rβ2, prompting responsiveness to IL-12 and the development of IFN-γ and IL-13 dual-producing cells (93). Thus, Module 2 is typically self-perpetuating, but can occasionally be redirected toward Module 1.

Module 3: Extracellular Bacteria and Fungi

All ILC3 subsets, Th17, and Th22 belong to Module 3, based on production of cytokines IL-22, IL-17a, and/or IL-17f (Figure 2). IL-26 is additionally produced in human but is not conserved in mouse. In the T cell lineage, Th22 also produce IL-10, which is not made in substantial quantities but other cells (94). Yet, Module 3 shares additional factors, including inherent plasticity to Module 1 and regulation by the cytokines IL-23 and IL-1. IL-23 activation of STAT3 drives Th17 and Th22 polarization, but it is not required for ILC3 development. However, STAT3-deficient ILC3 are unable to respond appropriately to IL-23 and are functionally impaired (95, 96).

Module 3 is conventionally associated with immunity to extracellular pathogens and fungi, but this module is also recognized to have a role in tissue homeostasis. It is deregulated in the development of psoriasis, Crohn’s disease, and other forms of autoimmunity such as the multiple sclerosis model experimental autoimmune encephalomyelitis (EAE). Consistent with this, investigators have reported ILC3 expansion in: skin and blood of patients with psoriasis (97–99); blood, gut, synovial fluid, and bone marrow of patients with ankylosing spondylitis (100); and blood, lung, and gut of patients with common variable immunodeficiency (CVID) (101). Interestingly, several labs have reported that ILC3 are reduced rather than increased in Crohn’s disease relative to ILC1 (17, 18, 57, 58), and that remaining ILC3 are phenotypically altered and express less MHC-II (102).

The master transcription factor regulating Module 3 identity is RORγt because in its absence, Th17 and ILC3 do not develop (24, 103). After development, however, RORγt directly regulates few genes, though in Th17, these include the signature cytokines IL-17a and IL-17f as well as receptors for key activating cytokines IL-23 and IL-1 (104). In ILC3s, RORγt may have an even smaller effect, as conditional deletion of RORγt as well as pharmacologic blockade revealed no difference in IL-17a production, but did reduce IL-17a production in Th17 (105). Many other factors have been shown to regulate Th17, but the TF AHR in particular appears also to be shared by ILC3 (30, 106, 107). AHR is a ligand-dependent TF that binds pollutants, dietary metabolites, and bacterial metabolic byproducts (108). In conjunction with RORγt, AHR is required for the development of Th-22 and post-natal ILC3 and also directly controls IL-22 production (30, 106, 107, 109).

Both human and mouse Th17 as well as most ILC3 are further characterized by substantial plasticity to Module 1 through the acquisition of T-bet and reduction of RORγt expression. This was originally discovered in patients with Chron’s disease, who harbor a population of RORγt⁺ T-bet⁺ cells that produce both IL-17 and IFN-γ (110). Later, mouse lineage tracing experiments of RORγt expression demonstrated that Th17 lose RORγt and IL-17 in IL-12-rich inflammatory environments, but retain the Module 1 TF T-bet and IFN-γ effector profile (111). For the mouse CCR6⁻ lineage of ILC3, similar lineage tracing experiments revealed that plasticity occurs in steady state (13). Unlike transitioning Th17, however, RORγt⁺ T-bet⁺ NKp46⁺ ILC3 have lost the capacity to produce IL-17 and instead produce IL-22 and IFN-γ (24, 27–32). Mouse ILC3 plasticity is enhanced in certain microenvironments, such as the colon, in part through higher levels of IL-23 (27). Human NKp44⁺ ILC3 also exhibit substantial plasticity in vitro, induced by several cytokines including IL-12, IL-2, and IL-23 (18, 57, 112). Consequently, the reduction of ILC3 and increase in ILC1 in Chron’s disease is may reflect increased ILC3 plasticity that occurs in the inflammatory microenvironment.

Concluding Remarks

The identification of ILCs was a surprise in part because an entire lineage of cells was missed for many years, but also because it suggested that cytokines conventionally thought to be produced only by T cells could also be expressed by innate immune cells (113). Over recent years, the similarity between T cell and ILC development and effector programs has been increasingly appreciated. For ILCs, engagement of master TFs necessary for T helper cell polarization explains the convergent function of these cells in a single module, which notably each directly activate module-defining effector cytokines. In the future, emerging unique functions for ILCs are likely to be mediated by lineage-defining TFs, such as the master-ILC regulator Id2 and Ets family members (114), either directly or through reorganization of the accessible chromatin landscape.

Can ILCs perform all the functions of T cells? Thus far, there appear to be three functions that ILCs lack. First, ILCs by definition do not express the T cell receptor (TCR) and therefore cannot signal through it. Though obvious, this point is notable because gene expression and regulatory differences between ILCs and T cells may involve signaling factors differentially associated with TCR but not cytokine receptors. Consistent with this, we recently reported that the regulomes of human T helper cells were characterized by AP-1 and NFAT TF motifs compared to ILCs, core pathways activated by TCR signaling (115). Indeed, Il4 expression, which is increased in Th2 compared to ILC2, is well known to be regulated by AP-1 and NFAT (116).

Second, as immediately functional cells, ILCs lack the capacity for naivety, leading one group to propose that the naïve state is the defining feature of adaptive immune cells, rather than any particular effector function (117). As ILCs develop, they appear to acquire modular pathways de novo in the bone marrow, whereas naïve CD4⁺ T cells uniquely require TCR signaling and cytokine-mediated STAT-activation to generate chromatin landscapes supportive of Th- polarization (Figure 1). Interestingly, in-vitro generated Th cells also have the capacity to respond directly to cytokines via innate-like STAT- and NFκB-dependent, NFAT-independent mechanisms (118–122). In vivo, tissue-resident memory CD4⁺ T cell populations can also demonstrate similar innate-like cytokine activation (123, 124). Future comparisons between effector profiles of in-vivo generated T cells of different stages of maturity/tissue-residence and ILCs will likely be informative to further define the scope of shared immune modules.

Finally, to date, no NKT or ILC subset has been discovered to express the master regulatory TF FOXP3, which drives Treg development. In the future, it will be interesting to determine if a Regulatory Module driven by FOXP3 expression is incompatible with NKT and ILC lineage, and if so, the mechanism by which this occurs.

What do we know?

• ILCs and T cells share key transcription factors and effector functions.

• Transcription factors drive effector functions and define ILC and T cell subtypes.

• T helper cells go through a naive state while ILCs do not.

• ILCs and T cell functions are similar but not identical.

What is still unknown?

• To what extent do differences in cell surface receptor signaling produce functional differences between ILCs and T cells?

• How do effector programs of ILCs compare with those of naive, effector, and memory T cell subsets?

• Are there ILC counterparts to Tfh and Treg, and if not, why?

• What is the contribution of ILCs to human biology in homeostasis and disease?

List with all current abbreviations

APC

Antigen Presenting Cell

ATACseq

Assay for Transposase-Accessible Chromatin with high throughput sequencing

CLP

Common Lymphoid Progenitor

COPD

Chronic Obstructive Pulmonary Disease

CVID

Common Variable Immunodeficiency

EAE

Experimental Autoimmune Encephalomyelitis

Eomes

Eomesodermin

FasL

Fas Ligand

γ_c

Common gamma chain, IL-2rg

ILC

Innate Lymphoid Cell

ILL

Innate-Like Lymphocyte

LCMV

Lymphocytic Choriomeningitis Virus

NK

Natural Killer

NKT

Natural Killer T

RSV

Respiratory Syncytial Virus

TCR

T Cell Receptor

TF

Transcription Factor

Tfh

T follicular helper cell

Th

T helper

Treg

Regulatory T cell