Searching for the Elusive Regulatory Innate Lymphoid Cell

Maintenance of immune tolerance relies on molecular and cellular regulatory mechanisms that keep harmful immune responses in check. Various immune populations that limit innate and adaptive responses to maintain immune homeostasis have been identified. These include the well-studied CD4⁺ regulatory T cells (Tregs) in addition to noncanonical regulatory B cells, regulatory γδ cells, regulatory CD8⁺ T cells, regulatory dendritic cells (DCs), regulatory populations of macrophages (commonly referred to as M2 macrophages or alternatively activated macrophages), and, more recently, regulatory populations of innate lymphoid cells (ILCs) (1–3). These regulatory populations work directly or as part of cellular networks to limit inflammatory immune responses.

Although NK cells have been extensively studied, the discovery that they are part of a family of innate lymphocytes has changed our understanding of how the innate immune system contributes to immunity and tissue homeostasis. Knowledge of ILC diversity and functions is rapidly evolving, but ILCs are currently classified as NK cells, lymphoid tissue inducer cells, group 1 ILCs (ILC1), group 2 ILCs (ILC2), and group 3 ILCs (ILC3) (4, 5). NK cells express T-box transcription factor (TBET) and eomesodermin (EOMES), and produce IFN-γ and TNF-α (5). Similar to NK cells, ILC1s produce IFN-γ and TNF-α but are TBET⁺EOMES^low/− (5). ILC2s express GATA3 and the retinoic acid (RA) receptor (RAR)–related orphan receptor-α (RORα), and produce IL-4, IL-5, IL-9, and IL-13 (5). ILC3s and lymphoid tissue inducer cells express RORγt and IL-22 or IL-17 alone, or in combination, as well as GM-CSF (5). ILCs lack adaptive Ag-specific receptors but rapidly respond to signals within the microenvironment. Because of overlapping cytokine and transcription factor profiles, ILCs are often considered innate counterparts of effector T cells (5), with noncytotoxic helper ILC subsets (ILC1, ILC2, and ILC3) paralleling the CD4⁺ Th1, Th2, and Th17 cells, respectively, and NK cells having similar functions to CD8⁺ cytotoxic T cells (6).

In addition to these well described canonical ILCs, it has been long hypothesized that a regulatory ILC (ILCreg) population exists with analogous functions to CD4⁺ Tregs. Tregs have essential roles in regulating immunity and maintaining peripheral tolerance (5–9). Subsets of Tregs include natural CD4⁺FOXP3⁺ Tregs, induced CD4⁺FOXP3⁺ Tregs, and IL-10–producing type 1 regulatory (Tr1) cells (1, 10). Tregs employ a variety of mechanisms including production of immunosuppressive cytokines TGF-β and IL-10 (11, 12), metabolic disruption via the adenosine pathway (i.e., cAMP, CD39, CD73) (13–15), suppression by cytolysis of effector cells (16, 17), and modulation of APCs’ maturation and function (18–20). All ILC family members have been associated with negative regulation of immunity in certain contexts, often using similar mechanisms to those used by Tregs including contact-dependent inhibition of T cells, regulation of immune responses through IL-10 or TGF-β, and indirectly via modulation of APCs (21–23).

Distinct mechanisms to those employed by Tregs have also been identified. For example, ILC3s express MHC class II (MHC-II) and induce T cell anergy via MHC-II presentation of commensal bacteria Ag in the absence of costimulation (24). ILC2s can express OX40L, which promotes expansion of Th2 cells and Tregs during lung inflammation, displaying a paired role in regulating type-2 immune responses and Tregs (25). ILCs can also indirectly limit inflammation due to effects on other regulatory immune cells. For example, IL-9⁺ILC2s aided in the resolution of Th17-mediated inflammation in an arthritis model by enhancing suppressive capability of Tregs through ICOSL/ICOS and GITRL/GITR interactions (26) and Zhou et al. reported murine ILC3s secrete IL-2 to maintain survival of Tregs (27). Similarly, IL-33–activated ILC2s enhanced Treg expansion and function in adipose tissue at steady state, whereas IFN-γ from NK cells and ILC1s suppressed ILC2 activation during inflammation or obesity, thereby limiting ILC2-Treg crosstalk and type 2 immune responses (28, 29). Moreover, using a mouse model of allogeneic bone marrow transplant, IL-13⁺ILC2s increased myeloid derived suppressor cell recruitment to the gastrointestinal tract, which reduced pathogenic donor–derived Th1 and Th17 responses (30). Although best known for killing virally infected or cancerous cells, NK cells have established roles regulating antiviral T cell responses in both mice and humans (reviewed in Refs. 31–34). These and many other studies highlight central roles for ILCs in maintaining immune homeostasis.

Although the ability of ILCs to regulate immune responses has been widely observed, recent studies have attempted to identify unique features of ILCreg populations. Yet their existence has been highly controversial, due in part to overlapping properties with other ILC family members, lack of lineage defining transcription factors and limited studies identifying factors that promote ILCreg development or maintain immunoregulatory functions. In light of these observations, the present review focuses on advances of three types of regulatory ILCs: regulatory NK-like cells, ILCregs, and IL-10–producing ILC2s (ILC2₁₀). We contrast phenotypic properties and functions with canonical ILC family members, and highlight studies that examine factors governing development and function of ILCreg populations.

Mouse and human NK cells differ in many receptors that regulate their activity and define NK cell subsets. In humans, NK cells are broadly classified as CD56^dimCD16⁺ (CD56^dim) NK cells and CD56^brightCD16⁻ (CD56^bright) NK cells (35). Mice lack the expression of NCAM encoding for CD56, and murine NK cells are typically identified as CD3^– NK1.1⁺CD49b⁺NKp46⁺ (36). Mouse CD27⁺CD11b^– NK cells are similar in function to CD56^bright NK cells, whereas murine CD27^–CD11b⁺ NK cell functions overlap with highly cytotoxic CD56^dim NK cells (Fig. 1) (37). Cytotoxic CD56^dim NK cells, comprising 80–95% of peripheral blood NK cells, kill directly via various mechanisms (Fig. 1) (35, 38, 39). In contrast, circulating CD56^bright NK cells are generally associated as more immature NK cells (40, 41). Yet CD56^bright NK cells located in secondary lymphoid organs and tissues exhibit a more “regulatory” phenotype, and have sometimes been referred to as “NKregs” (35, 38, 42). Complicating NK cell biology is the tremendous phenotypic heterogeneity exhibited by this ILC family member, with estimates of ∼6,000–30,000 phenotypically distinct circulating NK cells within an individual (43), presenting a challenge in identifying populations with unique functions.

In addition to central functions in killing virally infected and cancerous cells, NK cells “tune” T cell responses in viral infections, autoimmunity, cancer, and transplantation (reviewed in Refs. 30, 40, and 41) and can regulate B cell responses directly by altering IgG or IgM secretion or indirectly via interactions with CD4⁺ T follicular helper (Tfh) cells (44–46). Regulation of T cells is controlled by germline-encoded activating and inhibitory receptors and coinhibitory or costimulatory molecules that dictate NK cell responsiveness (47). Mechanisms of NK cell regulation include direct killing of T cells, preventing development of central memory T cells, directing cytokine production of CD4⁺ and CD8⁺ T cells and preventing CD8⁺ T cell activation (48–56). Recently, NK cell suppression of T cells in LCMV infection was linked to CXCR3 expression that mediated murine NK cell relocalization from red pulp to white pulp or T cell zones of spleen (57). NK cell relocalization DCs by limiting their maturation and stimulatory capabilities or through direct DC killing (58–62). Despite the long-standing role for NK cells in regulating immune responses, identifying factors differentiating NK cells with regulatory potential from conventional NK cells engaged in antiviral and antitumor immunity has proven challenging.

A number of studies reported unique cytokine profiles by NK cells that limit T cell responses, yet it is unclear how central these cytokines are to NK cell–mediated regulation. NK cells with nonclassical NK cell cytokine profiles (e.g., IL-10, TGF-β, GM-CSF, IL-22) have been observed and IL-10–producing NK cells have been studied dating back to the 1990s (63). Mehrotra et al. reported CD56⁺CD16⁺ NK cells cultured with IL-2 and IL-12 for 24 h expressed IL-10 (63). Separately, an induced IL-10⁺ NK cell population with reduced IFNG expression that maintained cytotoxic responses to K562 cells was observed (64). Functionally, IL-10⁺ NK cells suppressed proliferation of Ag-specific CD4⁺ T cells, and reduced IL-13 and IFN-γ production by T cells (64) (Fig. 2B). Subsequent studies identified mouse IL-10⁺NK1.1⁺NKp46⁺ NK cells with immunoregulatory roles in parasitic, viral, or bacterial infections. For example, treatment with IL-15 complex in experimental cerebral malaria induced IL-10⁺ NK cells that directly suppressed pathogenic IFN-γ⁺CD8⁺ T cells (65) (Fig. 2A). These IL-10⁺ NK cells were 80% positive for KLRG1 (killer cell lectin-like receptor G1) and CD11b, in line with a memory NK cell phenotype, and maintained expression of IFN-γ in contrast to IFN-γ^lowIL-10⁺ NK cells described in humans. In Prf1^−/− mice, CMV infection induced IL-10⁺ NK cells to directly suppress CD8⁺ T cell responses resulting in viral persistence (51) (Fig. 2A). Mouse IL-10⁺ NK cells that inhibited DC secretion of IL-12 expanded predominantly in the liver following Toxoplasma gondii infection (66). Expression of IL-10, however, is highly dependent on the type of infection, as local infection with influenza virus in mice did not induce IL-10⁺ NK cells (66).

In addition to IL-10, distinct cytokine profiles of “regulatory” NK-like cells have been observed in vitro, during infection, in cancer, and in the semiallogeneic environment of pregnancy (67–70). For example, regulatory NK-like cells that suppressed tumor-infiltrating lymphocytes also produced IL-22 and IL-9, with limited IFN-γ and TNF-α production (69). Specialized NK cells termed decidual NK (dNK) cells regulate fetal implantation and protect against maternal rejection of the fetus. These dNK cells comprise up to 70% of lymphocytes in the uterine mucosa and exhibit differential functions and limited cytotoxicity (71). Three distinct dNK cell populations were identified early in the maternal interface that all expressed CD49a (72) and in placental tissue two distinct CD56^bright NK cell populations were observed (73). EOMES^lowCD49a⁺ dNK cells exhibited high expression of inhibitor of DNA binding 3 (ID3) (73), a transcriptional regulator that suppresses T cell development and promotes CD56⁺ cell expansion from human fetal liver cells (74), and in mouse is necessary for TGF-β–induced generation of Tregs (75, 76). Separately, CD56^bright dNK cells were found to have low granzyme B, IFN-γ, and TNF-α expression compared with conventional NK cells, and expressed tissue repair factor AREG as well as VEGF (encodes vascular endothelial growth factor), which is important for the migration of trophoblast in late gestation (73). Additionally, GM-CSF expression by dNK cells aids in trophoblast implantation (77), a unique tissue-adapted function. Thus, production of noncanonical cytokines and factors can guide biological processes and NK-like regulatory populations can exhibit tissue-specific functions.

Recent studies have attempted to identify the transcriptional state or surface markers that distinguish regulatory NK-like cells from conventional NK cells. Although numerous studies have demonstrated tumor-infiltrating NK cells commonly have reduced cytotoxic potential and poor cytokine production (reviewed in Ref. 78), regulatory NK-like populations that directly inhibit antitumor T cells have been reported by several groups (reviewed in Refs. 67 and 68). Morandi et al. reported circulating CD56^bright NK cells have increased CD73 in comparison with CD56^dim NK cells (79). CD73 hydrolyzes AMP to adenosine and is linked to Treg inhibition of DC maturation, polarizing DCs towards a suppressive phenotype, dampening T cell effector functions, and stabilizing Tregs (80). In breast cancer and sarcoma patients high NT5E (CD73) expression was noted and correlated with a worse prognosis (81). This was linked to tumor-infiltrating CD73⁺ NK cells that upregulated PD-1 and LAG3 after adenosine receptor binding (80, 81). Culture of NK cells from healthy donor blood with sarcoma and breast cancer resections induced CD73⁺ NK cells that produced IL-10 and TGF-β and inhibited T cell proliferation and IFN-γ expression (81) (Fig. 2B). Inhibiting STAT3 abrogated the ability of CD73⁺CD56⁺CD3⁻ NK-like cells to suppress T cells (81); however, blocking CD73 failed to restore T cell proliferation when CD73⁺ NK cells were present (Fig. 2B) (79). Characterization of CD73⁺ NK cells demonstrated an ∼8-fold increase in class I–restricted T cell–associated molecule (CRTAM) (81). CRTAM expression by NK cells is tissue specific (i.e., high expression by CD56^bright NK cells in healthy human lung) and can differ in disease contexts (82). Although CD56^bright NK cells in bone marrow of healthy individuals downregulated CRTAM (82, 83), expansion of a bone marrow CRTAM⁺CD56⁺ population that expressed IL-10 and TGF-β and exhibited decreased cytotoxicity with increased PD-1 expression was observed in acute lymphoblastic leukemia patients (84).

We identified a distinct regulatory NK-like population that suppressed CD4⁺ and CD8⁺ T cells and correlated with poorer clinical outcomes in ovarian cancer (Fig. 2B) (69). This NK-like ILCreg population displayed unique and overlapping properties with several ILC family members (69). This included expression of KIRs, NKp46, NKG2D/CD94, and CD7 in line with NK cells, but with a unique transcriptional profile, limited IFN-γ, and TNF-α and expression of CCL3, IL-9, and IL-22, associated with ILC1s, ILC2s, and ILC3s, respectively. No difference in FOXP3, IL-10, or TGFB1 expression was observed compared with conventional NK cells but regulatory NK-like cells upregulated ID2, which all ILCs express and is required for mouse NK cell maturation and cytotoxicity (85–87). Secreted factors did not appear involved in T cell regulation, but blocking the natural cytotoxicity receptor NKp46 abrogated their ability to suppress T cell proliferation (69).

NKp46, encoded by NCR1, is a conserved marker for NK cells and subsets of ILC3s in both mouse and human, and has also been associated with NK cell regulation of T cell responses (88, 89). Similar to our findings, NKp46 expression by regulatory NK-like cells was also observed in non–small-cell lung carcinoma, where a high proportion of circulating NKp46⁺CD56^dimCD16⁺ NK cells correlated with worse overall survival (90). These regulatory NKp46⁺ NK-like cells coexpressed PD-1 and TIM-3, and coculture with tumor-specific T cells suppressed IFN-γ secretion, an effect which was NKp46 dependent (90). Russick et al. characterized a similar NKp46⁺CTLA-4⁺ NK cell population in non–small-cell lung carcinoma (91). NKp46⁺CTLA-4⁺ NK cells reduced expression of MHC-II and CD86 on CD11c⁺ DCs, and also correlated with worse overall survival, further supporting regulatory NK-like cells may inhibit antitumor immunity (91).

Mouse viral infection models that support NKp46 is important for NK cell inhibition of antiviral T cells (88, 92). In high dose LCMV infection, NKp46⁺NK1.1⁺ NK cells suppressed CD4⁺ and CD8⁺ proliferation and limited IFN-γ⁺ TNF-α⁺ expression by CD8⁺ T cells (88, 92). Separately, in vitro stimulation of NK cells with IL-15, IL-18, and IL-12 induced NK cells with an NKp46⁺CTLA-4⁺TIM-3⁺ phenotype that produced TGF-β and IL-10 (89) (Fig. 2A). These studies support regulatory NK-like cells may be identified by NKp46, PD-1, and CTLA-4 expression, but these molecules are also associated with conventional NK cells, limiting their application in differentiating conventional from regulatory NK-like cell populations.

Many questions remain, particularly what factors promote NK-like regulatory cells to be induced and produce noncanonical cytokines such as IL-10, TGF-β, IL-22, or GM-CSF. Are these NK cells that have acquired immunosuppressive functions based on microenvironment signals, or is their development induced from an ILC progenitor population? There are several similarities in phenotypes and mechanisms employed by regulatory-like NK cells and Tregs including IL-10, CTLA-4, PD-1, CD73, and TGF-β. Yet transcriptional networks that drive expression of these molecules have not been defined. Understanding similarities and differences of regulatory NK-like cells from Tregs and conventional NK cells, as well as defining transcription factors which drive suppressive functions or factors that promote regulatory NK-like cell development are needed to bring clarity to disparate observations of regulatory activity in NK cells.

Distinct IL-10–producing ILCregs have been observed by several groups. Kim et al. reported IL-10–producing ILCs marginally increased in a contact hypersensitivity model (93). Following this, an Id3⁺ ILCreg population was identified that produced IL-10 and TGF-β and limited ILC1s and ILC3s in acute intestinal inflammation (Fig. 3A) (94). ILCregs did not express lineage defining transcription factors or surface markers associated with other ILCs and were defined as CD127⁺IL-10⁺TGF-β⁺ using IL-10–GFP reporter mice (94). Although Id2, which is required for helper ILCs and terminal differentiation of NK cells was necessary for intestinal ILCregs, as the loss of Id3 impaired ILCreg development but not other ILCs (94, 95). TGF-β production sustained ILCregs whereas IL-10 production mediated suppression of ILC1s and ILC3s (94), mirroring Tr1 capacity to limit other CD4⁺ T helper subsets. Interestingly, in a separate mouse model of colorectal cancer, TGF-β–induced tumor-infiltrating IL-22⁺ ILC3s to acquire an IL-10⁺CD127⁺Id3⁺ ILCreg phenotype, suggesting that ILCregs may develop from canonical ILCs based on environmental factors (96).

Supporting potential for ILCregs to impact homeostasis in other tissues, Cao et al. identified a CD127⁺IL-10⁺Id3⁺ ILCreg population in murine kidney that limited ischemia reperfusion injury (Fig. 3A) (97). The frequency of ILCregs was 2.7% of renal ILCs in C57BL/6 mice (97). Adoptive transfer of expanded ILCregs decreased proinflammatory cytokine expression, as well as neutrophil and ILC1 infiltration (97). Human renal nephrectomy samples also contained an ILCreg population (4.3% of ILCs) but rather than being defined by ID3, expression of ILC2 activation markers CD25 and ICOS and the absence of CRTh2 and KLRG1 were used to define human renal ILCregs (97). This study noted IL-2C, a CD25 agonist, induced an ILCreg phenotype in several organs other than the kidney (97), yet the functions were not compared or contrasted across tissues.

Besides confirming ILCs with an ILCreg phenotype are present in human nephrectomy and gut samples, few studies have identified human ID3⁺ILCreg populations or assessed their functions. O’Connor et al. recently reported a CD127^lowCD74⁺CXCR5⁺CD7⁺CCR6⁺ID3⁺ follicular ILCreg (ILC_FR) in human tonsils with a transcriptional profile similar to Id3⁺ILCregs (98). TGF-β expression by ILC_FR cultured with Tfh cells and B cells from germinal centers reduced IgG secretion, decreased IL-21, and increased soluble CD40L (Fig. 3B) (98). ILC_FR transcription factor profile and function were contrasted with ILC3s, and observed ILC3s increased IgG production when cultured with Tfh and B cells contrary to mouse studies that demonstrated MHC-II⁺ ILC3s interactions with Tfh cells result in reduced colonic IgA (98, 99). These findings and others demonstrate ILCs influence immune responses within lymphoid tissues and control developing immune responses by acting on multiple adaptive immune populations.

Despite intriguing findings on ILCregs, limited studies have identified a distinct ILCreg population, and the role of Id3⁺ ILCregs in immune responses remains highly controversial. Bando et al. could not replicate findings of Id3⁺ ILCregs in the intestine using IL-10–eGFP reporter mice bred to C57BL/6J mice (100). IL-10 expressing ILCs in this study did not express the same surface markers and cytokines defined by Wang et al. with less than 1% of Lin⁻CD127⁺GATA3⁻RORγt⁻ cells having a phenotype consistent with ILCregs (100). Instead ILC2s were the predominant source of IL-10–producing ILCs in the intestine (100). Lineage tracing studies may yield insights into whether the microenvironment redirects an ILC2 or ILC3 to acquire an ILCreg phenotype and downregulate canonical ILC markers. Yet similar to regulatory NK-like cells, lack of consistent markers and transcription factor expression that differentiates ILCregs from other ILCs presents a challenge.

Landmark studies have demonstrated that natural helper cells, nuocytes, and innate helper 2 cells collectively referred to as ILC2s have the potential to produce IL-10 (101, 102). ILC2₁₀s have been referred to as ILC2₁₀ or ILC2regs interchangeably. Early reports of helminth expulsion observed ILC2s from spleen produced IL-10 (102), a finding that was also observed in allergic rhinitis (103–105). More recently, an ILC2₁₀ population was observed in lungs of mice treated with either IL-33 and IL-2 or with chronic exposure to the protease-allergen papain (103). These ILC2₁₀s limited lung inflammation and eosinophil recruitment (Fig. 3C) (103). Removal of stimuli resulted in contraction of ILC2₁₀, yet these ILCs remained as tissue sentinels, poised to rapidly respond and limit pathogenic inflammatory immune responses (103).

The ILC2_10s population described by Seehus et al. paralleled a regulatory ILC2 population observed in the lung tissue of house dust mite (HDM)–treated mice and patients with chronic rhinosinusitis with nasal polyps (CRSwNP) that suppressed the proliferation of CD4⁺ T cells and ILC2s (104). ILC2₁₀s were rarely observed in healthy subjects or control-treated mouse lungs, lacked expression of FOXP3, and maintained expression of ILC2-associated transcription factors such as GATA3 (103, 104). In contrast to ILC2₁₀ characterized in lung, ILC2₁₀ in CRSwNP expressed CTLA-4 and CD25, and downregulated genes encoding ILC2-associated molecules CRTH2, CD127 and CD117 (104).

IL-2, IL-4, IL-27 and soluble neuromedin U (NMU) have been associated with promoting ILC2₁₀; however, IL-33 and RA have consistently promoted IL-10–producing ILCs in mouse and human (Fig. 3C, 3D) (100). Seehus et al. demonstrated RA-induced production of IL-10 but not IL-13 in mouse ILC2s (103). Similarly, Morita et al. reported in vitro stimulation with RA-induced IL-10 production by human ILC2s (104). Increased protein expression of retinaldehyde dehydrogenase 1 (RALDH1), an enzyme that converts retinal to RA, was observed in nasal epithelial cells of CRSwNP patients, suggesting epithelial cells synthesize RA during inflammation to promote ILC2₁₀ as a means of re-establishing tissue homeostasis (Fig. 3D) (104). RA signals through RAR, which is expressed by all human ILCs (106). IL-10 production was blocked in a dose-dependent manner by addition of a pan-RAR inhibitor to human ILC2s stimulated with IL-2, IL-33, and RA (104). TGF-β in combination with RA inhibited IL-10 production by ILC2s, whereas IL-2 and RA synergized to promote IL-10 production (103). Bando et al. reported IL-2, IL-4, IL-27, and soluble NMU promote IL-10 production by mouse ILC2s, but not by ILC1s, ILC3s, or NK cells, and further showed IL-10 acts in a positive feedback loop to promote additional ILC2₁₀ (Fig. 3C) (100). Separately, ILC2₁₀ were induced by IL-4 in a murine model of AHR where ILC2₁₀ dampened ILC2 effector functions leading to decreased lung inflammation (107). Cytokines and metabolites that promote ILC2₁₀ may therefore differ across tissues, highlighting the need to understand how microenvironment signals are integrated to promote ILC2₁₀ as part of circuits that maintain tissue homeostasis.

Recent reports of ILC2₁₀s have emphasized potential therapeutic applications. In agreement with earlier studies, Golebski et al. confirmed IL-33 and RA induced ILC2₁₀ in vitro and reported KLRG1 is required for ILC2₁₀s (105). Single-cell analysis of ILC2₁₀ demonstrated increased expression of genes implicated in the retinol metabolism pathway including RARA and RARG (105). IL-10⁺KLRG1⁺ ILC2s decreased activation of naive CD4⁺ T cells, limited Th1 and Th17 cell responses, and maintained epithelial cell integrity (105). In addition, the authors observed patients with grass pollen allergy had lower levels of IL-10⁺KLRG1⁺ ILC2s compared with healthy controls (105). In a double-blind, placebo-controlled clinical trial, ILC2₁₀ were restored following grass pollen allergen-specific immunotherapy (AIT). Separately, IL-10⁺CTLA-4⁺ ILCs increased by 3.2% after 2 y in HDM allergic rhinitis receiving AIT (108). Together, the frequency of ILC2₁₀ increases in response to AIT for allergic responses, and approaches which promote their activity may have therapeutic applications in allergic airway inflammation.

The immunoregulatory potential of ILC2₁₀ was also highlighted in mouse islet allograft and traumatic brain injury models (109, 110). In vivo IL-33 administration to mice receiving allogenic islets resulted in expansion of ILC2s and Tregs, and enhanced islet allograft survival (109). Two subtypes of ILC2s were identified within the islet allograft: ILC2₁₀s and non–ILC2₁₀s (109). Adoptive transfer of ILC2₁₀ prolonged islet allograft survival, supporting the therapeutic potential of harnessing ILC2₁₀ in islet transplantation (Fig. 3C) (109). Separately, AMP-activated protein kinase (AMPK) was linked to ILC2₁₀ in a mouse model of traumatic brain injury (110). Inhibition of AMPK increased all ILC subsets within the meninges, whereas AMPK activation selectively increased numbers of ILC2₁₀ at the expense of ILC1s and ILC3s (110). Use of AMPK activator metformin led to expansion of ILC2₁₀ and a corresponding decrease in ILC1s and ILC3s, demonstrating potential of AMPK-activating therapeutics to enhance ILC2₁₀ expansion (110).

It is unknown whether specific transcription factors are required for ILC2₁₀ activity. ILC2₁₀ do not express FOXP3 and few studies have identified gene expression patterns that direct the divergence of ILC2s to ILC2₁₀. Howard et al. demonstrated IL-4–mediated induction of mouse ILC2₁₀ via an IL-4–STAT6 axis corresponded with upregulation of c-musculoaponeurotic fibrosarcoma (cMaf) and B lymphocyte-induced maturation protein-1 (Blimp-1) (107). Knockdown of cMaf or Blimp-1 prevented IL-10 production by ILC2₁₀ (107), in line with established roles for cMaf and Blimp-1 in regulating IL-10 (111–114) and paralleling Tr1 cells that are Blimp-1 dependent (115). Yet, BLIMP1 was not differentially expressed in KLRG1⁺ ILC2₁₀ in individuals with grass pollen allergic rhinitis, nor was HELIOS and IRF4 (105). Morita et al. detected upregulation of HELIOS in IL-10⁺ ILC2s, but confirmed no difference at the protein level (104). In lung ILC2₁₀ of IL-33–treated mice, higher expression of forkhead box F1 (Foxf1), activating transcription factor 3 (Atf3), Klf2 (Krüppel-like factor 2) and Id3 was observed compared with IL-10⁻ ILC2s, indicating potential roles in ILC2₁₀ development or function (103). Although the function of Foxf1 and Atf3 in ILCs have not been reported, KLF2 regulates NK cell proliferation and survival and Id3 was also observed in ILCregs and NK-like regulatory cells (73, 95, 103). However, Id3 was not differentially expressed in ILC2₁₀ in HDM-treated mice or in vitro–expanded human ILC2₁₀, indicating Id3 is not required for ILC2₁₀ function. Elucidating the transcription factor network that directs the development or function of regulatory ILC2₁₀ will be important moving forward to delineate the role of ILC2₁₀ in maintaining immune homeostasis.

ILCs are often considered the innate counterpart of T cells due to overlapping transcription factor and cytokine expression profiles. Yet a developmentally distinct ILCreg subset central to immune tolerance similar to natural CD4⁺FOXP3⁺ Tregs has not been established. Questions remain as to whether any of these ILCreg subsets can be considered distinct from canonical ILCs, or if instead ILCreg populations acquire immunosuppressive functions from microenvironment signals. ILCs display a high degree of plasticity, acquiring new phenotypes and functions based on factors within the tissue niche. Evidence to date mostly supports conversion of canonical ILC subsets to ILCreg populations. Indeed, in many ways ILCs are similar to macrophages, existing on a spectrum of phenotypes and functions. Of note, Id2 is important for both ILC and myeloid lineage development, including monocyte to macrophage differentiation (116, 117) and blocks T cell development (118, 119). ILC2s could potentially mirror alternatively activated/M2-like macrophages, which can express IL-10 and develop downstream of cytokines including IL-4 and IL-13 (120, 121). The paradigm of ILCs mirroring T cell subsets may not accurately reflect tissue-adapted functions of ILCs in all cases, and perhaps looking to myeloid studies may provide new insights into ILC development and function.

Overlapping phenotypic properties and IL-10 production by ILCregs and ILC2₁₀ make distinguishing these two ILCreg populations challenging. Of note, ILC2₁₀ downregulated CRTH2 and other canonical ILC2 markers (104), and Id3 expression has been reported in both populations. More research is needed to confirm whether the ILCregs and ILC2₁₀ are distinct or rather the same population with slightly altered phenotypic profiles based on the tissue or model employed. Similar to studies of regulatory NK-like populations, lack of specific markers or transcription factors that distinguish inflammatory from regulatory populations creates a challenge for interpreting findings.

Beyond gaps in understanding development of ILCreg populations, the stability of regulatory NK-like cells, ILCregs, and ILC2₁₀ is unclear. Whether removal of stimuli results in regulatory populations reverting back to canonical ILCs has not been determined. Further investigation is needed to assess epigenetic regulation of transcription factor and cytokine loci. With the advancement of single-cell RNA sequencing and proteomic approaches, addressing ILCreg stability within the context of heterogenous ILC populations in health and disease is now possible. Determining the stability or plasticity of ILCreg populations will enhance our understanding of immune tolerance mechanisms and inform if harnessing or inhibiting ILCreg populations may have applications in new therapeutic approaches.

We thank Sarah J. Colpitts for careful reading of the manuscript. Figures were created with BioRender.com.

The authors have no financial conflicts of interest.