CD4⁺ Group 1 Innate Lymphoid Cells (ILC) Form a Functionally Distinct ILC Subset That Is Increased in Systemic Sclerosis

Innate lymphoid cell (ILC) subsets that mirror Th cells in their effector cytokine profiles have recently emerged as central players in homeostatic and inflammatory conditions. Long-lived tissue-resident group 2 ILC (ILC2) constitutively produce IL-5 and play varied roles in maintaining immune and metabolic homeostasis (1–4); group 3 ILC (ILC3) maintain the integrity of the intestinal barrier, in part through production of IL-22 (5, 6). ILC2 and ILC3 can also express MHC class II and may act as APCs in the initiation of inflammation or in the maintenance of tolerance (7–10). In keeping with the functional parallels between ILC and Th cell subsets, group 1 ILC (ILC1) appear to be important in coordinating type 1 inflammatory responses (11, 12), ILC2 are required in numerous models of type 2 immunity (13–16), whereas ILC3, like Th17 and Th22 cells, are central players in IL-17– and/or IL-22–driven inflammation (17, 18).

In human studies, alterations in ILC frequencies in a number of diseases suggest that these innate populations may play key roles in the pathogenesis of human autoimmunity (19–27). How ILC subsets affect disease susceptibility, development, and progression remains undefined. However, in multiple sclerosis and psoriasis, treatments that decreased patients’ clinical severity scores also correlated with normalization of ILC frequencies (19, 21, 27). In psoriasis, increased frequencies of NKp44⁺ ILC3 during active disease, as well as normalization of these frequencies with treatment, occurred in the skin and peripheral blood. These studies indicate that systemic alterations in ILC subsets occur in autoimmunity and may present an important target for disease therapy. The regulation and function of these ILC populations, as well as the degree to which they parallel or differ from T cells, may have significant implications for the efficacy and side effect profile of novel therapeutic approaches.

We performed a comprehensive analysis of human peripheral blood ILC subsets and describe novel ILC1 populations that express CD4 and CD8α. Although these newly described ILC1 populations shared some characteristics with Th1 cells, CD4⁺ ILC1, in particular, were potent producers of TNF-α, GM-CSF, and IL-2 and showed considerable diversity in their chemokine and cytokine receptor expression. Our data illustrate that ILC1 cannot be thought of simply as innate equivalents of Th1 cells. We also show that peripheral blood CD4⁺, but not CD4⁻, ILC1 frequencies are altered in systemic sclerosis (SSc), a complex and poorly understood autoimmune disease characterized by fibrotic and vascular pathology. These data demonstrate a previously unappreciated heterogeneity in human peripheral blood ILC1 and suggest a role for ILC in the pathogenesis of SSc.

Abs used for flow cytometric analyses and cell sorting included the following: CCR10 PE (6588-5), CD117 (104D2) BV421/BV605/PE-Cy7, CD11c (3.9) FITC, CD123 (6H6) FITC, CD126 (UV4) biotin, CD127 (A019D5) BV650, CD130 (AM64) PE, CD14 (HCD14) FITC, CD16 (3G8) FITC/BV421, CD27 (O323) biotin, CD28 (CD28.2) BV421, CXCR3 (G025H7) BV421/PE-Cy7, CXCR5 (J252D4) BV421, CD19 (HIB19) FITC, CCR4 (L291H4) PerCP-Cy5.5/A647, CCR5 (HEK/1/85a) A647, CCR6 (G034E3) BV605, CCR7 (G043H7) BV421, CRTH2 (BM16) PerCP-Cy5.5/A647/BV421, CCR9 (L053E8) A647, CD3 (UCHT1) A700/BV421/FITC, NKp44 (P44-8) PE, CD34 (581) FITC, CD4 (OKT4) A700/BV605/BV785, CD45 (HI30) BV510, CD45RO (UCHL1) A700, CD45RA (HI100) BV421, CD56 (HCD56) A700/BV605, CD62L (DREG-56) PE, CD8α (RPA-T8) A700, CD94 (DX22) FITC, cutaneous lymphocyte Ag (CLA; HECA-452) biotin, FcεRIα (AER-37) FITC, GATA3 (TWAJ) PE, GM-CSF (BVD2-21C11) PE, granzyme A (CB9) A700, granzyme B (GB11) A647, IFN-γ (4S.B3) BV785, IL-13 (JES10-5A2) PE, IL-17A (BL168) A647, IL-2 (MQ1-17H12) A647, IL-22 (22URTI) PE-Cy7, TNF-α (MAb11) PE-Cy7, perforin (dG9) PE, T-bet (4B10) A647/PE-Cy7, and TCRαβ (IP26) PE (all from BioLegend); CD117 (104D2) PE-Cy7, CD2 (RPA-2.10) allophycocyanin, CD45 (HI30) e450, eomesodermin (Eomes; WD1928) PE, and Fixable Viability Dye eFluor 780 (all from eBioscience); and streptavidin PE and PE-Cy7, phospho-STAT1 (4a) PE, and phospho-STAT3 (4/P-STAT3) PerCp-Cy5.5 (all from BD Biosciences).

Complete RPMI medium (cRPMI) used for stimulations and resuspension included RPMI 1640 (Sigma), 50 U/ml penicillin/50 μg/ml streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, 1× MEM nonessential amino acids, 2 mM l-glutamine (all from Life Technologies), and 1 or 10% human serum (HS; Omega Scientific). FACS buffer included 1× PBS/1% BSA (w/v) with or without 0.5% sodium azide (w/v) (both from Sigma). MACS buffer was composed of 1× HBSS (HyClone)/5% FBS (Sigma)/20 mM HEPES/Pen-Strep/2 mM EDTA (Sigma). Sort buffer included 1× HBSS/2.5% FCS/Pen-Strep/20 mM HEPES.

Fresh whole blood and frozen PBMC from control subjects were obtained from healthy individuals who were participants in the Benaroya Research Institute (BRI) Healthy Control Registry and Biorepository without a self-reported personal history of autoimmunity. Matched controls that were compared with SSc patients also did not have a self-reported family history of autoimmunity. Frozen PBMC from SSc patients were obtained from subjects in the BRI Rheumatic Diseases Registry and Biorepository. Control and SSc cohorts were matched for age and gender in all studies. Frozen cord blood cell samples were from subjects in the BRI Control Biorepository. Protocols for the collection and use of samples in the BRI registries and Biorepositories were reviewed and approved by the BRI Institutional Review Board.

Human PBMC were isolated from peripheral blood by centrifugation over a Ficoll-Paque Plus (GE Healthcare) gradient. Previously frozen PBMC were quick thawed; prewarmed cRPMI with 10% HS and then 1% HS were added drop-wise to wash the cells.

Cells were stained with Fixable Viability Dye eFluor 780 (eBioscience) and then stained for surface markers for 25 min at room temperature (RT). Cells were then fixed with 2% paraformaldehyde (EMS)/PBS for 20 min at 4°C or fixed and permeabilized for intracellular/intranuclear staining. For intracellular cytokine staining, cells were permeabilized using BD Cytofix/Cytoperm, according to the manufacturer’s instructions, and then stained in 1× BD Fix/Perm buffer for 25 min at RT. For intranuclear staining, cells were fixed with 4% paraformaldehyde for 10 min at RT and permeabilized with ice-cold (−20°C) methanol (Sigma) for 15 min on ice. Intranuclear staining was then performed in Foxp3 buffer (eBioscience) for 45 min at RT. All flow cytometry data were acquired on a BD LSR II or BD Fortessa, and analyses were performed using FlowJo v10.

For cell sorting, freshly isolated PBMC were lineage depleted with FITC-conjugated Ab (anti-CD3, CD14, CD19, CD16, CD94, CD11c, CD123) and anti-FITC beads (Miltenyi Biotec), according to the manufacturer’s instructions, with minor modifications. Twenty microliters of beads/10⁷ cells were used in most experiments. After staining with viability dye and surface markers, cells were sorted on a BD FACSAria II using a 70-μm nozzle and collected into Eppendorf tubes containing cRPMI/10% HS.

CD4⁺ and CD4⁻ ILC1 were sorted from freshly isolated PBMC as described above, and ∼10,000 cells were added to a single well in a 96-well round-bottom plate coated with irradiated mouse embryonic fibroblasts (R&D Systems) and cultured for 7 d in cRPMI/10% HS with 20 ng/ml IL-2 (BioLegend) and 20 ng/ml IL-7 (BioLegend). Cells were stained with Fixable Viability Dye eFluor 780 (eBioscience) and then stained for surface markers and analyzed by flow cytometry.

Stimulations were performed with 5 ng/ml rIL-6 (BioLegend) in cRPMI at 37°C/5% CO₂ for 20 min after a 30-min rest (10 × 10⁶ cells/ml in cRPMI/1% HS; 37°C/5% CO₂) and staining with viability dye and surface markers. PMA/ionomycin stimulations were performed in cRPMI/10% HS with 50 ng/ml PMA (Sigma), 750 ng/ml ionomycin (Sigma), and 10 μg/ml brefeldin A (Sigma) for 6 h at 37°C/5% CO₂.

ILC subsets were sorted as described above, and 2000 cells from each population were used to prepare cDNA using a SMARTer Ultra Low RNA Kit v2 (Clontech). A Nextera XT kit was used for library construction, and sequencing was performed on an Illumina HiSeq2500. Analysis was performed using the R package, DESeq2, and a false-discovery rate < 0.05 was considered significant. The sequences presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus under accession number GSE69596.

Statistical analyses were performed using Prism with the two-tailed unpaired Mann–Whitney test. The p values < 0.05 were considered significant.

Like their Th1, Th2, and Th17/Th22 counterparts, ILC subsets are categorized based on their expression of specific transcription factors and effector cytokines: ILC1 express T-bet and IFN-γ; ILC2 express GATA-3 and type 2 effector cytokines, such as IL-13 and IL-5; and ILC3 express RORγt and the cytokines IL-22 and/or IL-17 (28). Under this nomenclature, NK cells and lymphoid tissue inducers are considered ILC1 and ILC3, respectively. To better define the relative frequencies of ILC1, ILC2, and ILC3 in human peripheral blood, we first characterized the overall composition of ILC subsets by flow cytometry in a cohort of healthy subjects. Our gating strategy for the identification of peripheral blood ILC is shown in Fig. 1A. Although all ILC subsets express CD127 (IL-7Rα) but are negative for lineage-specific surface markers for T cells, B cells, monocytes, and dendritic cells, they can be further divided into groups by their differential expression of c-kit and CRTH2. Only ILC2 express CRTH2; within the CRTH2⁻ population, ILC1 are c-kit⁻, whereas ILC3 are c-kit⁺. In control subjects, very few peripheral blood ILC3 express NKp44. To exclude mature NK cells, we gated on CD56⁻ ILC1 in our analyses (29). Lack of expression of CD16, perforin, and granzyme B in CD56⁻ ILC1 further confirmed that these cells are distinct from mature NK cells (Supplemental Fig. 1), consistent with previous reports for ILC1 (29, 30). We were surprised to find that the peripheral blood ILC1 compartment contained CD4⁺ CD8⁻, CD4⁻ CD8⁺, and CD4⁻ CD8⁻ populations (referred to as CD4⁺, CD8⁺, and double-negative [DN] ILC1, respectively), whereas the ILC2 and ILC3 subsets did not express CD4 or CD8α (Fig. 1B). Although ILC subset frequencies varied considerably within a cohort of healthy controls (Table I), ILC frequencies were stable within an individual over time (Supplemental Fig. 2A). CD4⁺ ILC1 constituted more than half of peripheral blood ILC1 in the majority of healthy controls. In addition, ILC subset frequencies from fresh PBMC were comparable to frequencies following freeze-thawing (Supplemental Fig. 2B).

Further examination of ILC1 revealed that, like a recently described CD4⁺ CD3⁻ innate-like T cell population (31), CD4⁺ ILC1 expressed intracellular CD3ε, although we detected little or no expression of surface or intracellular TCRαβ on any peripheral blood ILC (Fig. 1C, 1D). Furthermore, intracellular CD3ε was present in CD4⁺ ILC1, as well as in a large percentage of all ILC1 subsets, but not in ILC2 or ILC3.

We next evaluated surface CD3ε and TCRαβ expression on sorted CD4⁺ and CD4⁻ ILC1 after a 7 d culture with irradiated feeder cells, IL-2, and IL-7 (Fig. 1E). The vast majority of sorted cells remained CD3ε⁻ and TCRαβ⁻, indicating that they are not activated T cells or NKT cells with low expression of TCR. A small percentage of CD3ε⁺ TCRαβ⁺ cells were present in both CD4⁺ and CD4⁻ ILC1 cultures, although it is unclear whether there was preferential expansion or survival of a small number of contaminating T lymphocytes or whether some ILC1 can indeed upregulate surface CD3 and TCR. It is interesting to note that a small percentage of cultured CD4⁻ ILC1 expressed surface CD3ε in the absence of TCRαβ. In addition, a fraction of CD4⁻ ILC1 upregulated CD56 postculture, suggesting that this marker may not distinguish between conventional NK cells and other ILC1.

These data demonstrate that the ILC1 subset contains distinct populations characterized by CD4 and CD8α surface expression, though the shared expression of intracellular CD3ε in ILC1 suggests a lineage relationship among the ILC1 populations that may be distinct from ILC2 and ILC3.

Effectors of type 1 inflammation, such as CD4⁺ Th1 cells, CD8⁺ T cells, and NK cells, are characterized by IFN-γ production, expression of the chemokine receptor CXCR3, and expression of the transcription factors T-bet and/or Eomes (32–35). To examine whether all ILC1 populations shared a type 1 phenotype, we first delineated the expression of T-bet and Eomes and contrasted this with the expression of the Th2-associated factor GATA-3 by flow cytometry. As expected, GATA-3 was expressed in ILC2 but was low or absent in ILC1 and ILC3 (Fig. 2A). Although CD4⁺ ILC1 expressed T-bet, but not Eomes, CD4⁻ ILC1 coexpressed Eomes and T-bet, most prominently in the CD8⁺ ILC1 population (Fig. 2B, 2C). All ILC1 subsets also expressed CXCR3, although CD8⁺ ILC1 contained the highest percentage of CXCR3⁺ cells (Fig. 2B, 2C).

To evaluate cytokine production from ILC1, we sorted ILC subsets from control subjects and analyzed intracellular cytokine production after PMA/ionomycin stimulation by flow cytometry. CD8⁺ and DN ILC1 were analyzed together as CD4⁻ ILC1 because of the lower frequency of these cells compared with CD4⁺ ILC1 in most individuals and the dominant Th1-type phenotype in both of these populations. CD4⁺ and CD4⁻ ILC1 produced IFN-γ but little or no IL-13, IL-17A, or IL-22 (Fig. 3A). Within CD4⁺ ILC1, IFN-γ production was largely restricted to CXCR3⁺ cells (Fig. 3B). CD4⁺ ILC1 were also potent producers of TNF-α, GM-CSF, and IL-2 (Fig. 3). To varying degrees, all other ILC subsets were also capable of producing TNF-α, GM-CSF, and IL-2 poststimulation (Fig. 3, Supplemental Fig. 3). Effector cytokine expression in peripheral blood ILC2 and ILC3 was consistent with previous reports (Supplemental Fig. 3) (30).

Although the CD4⁺, CD8⁺, and DN ILC1 populations that we describe shared numerous features with Th1 cells, CD4⁺ ILC1 expressed lower levels of Th1-associated factors than did CD4⁻ ILC1 and showed substantial functional heterogeneity. Coexpression of T-bet and Eomes in CD8⁺ ILC1 suggests that this subset shares more similarities with NK cells than do other ILC1 subsets. These data indicate that ILC1, particularly CD4⁺ ILC1, do not fit strictly into a Th1 paradigm.

Because CD4⁺ ILC1 demonstrated phenotypic and functional characteristics distinct from CD4⁻ ILC1 and had a lower frequency of cells expressing the Th1-associated factors T-bet, CXCR3, or IFN-γ, we were interested in whether chemokine receptors and activation markers might be able to further define functionally distinct CD4⁺ ILC1 populations. Differential patterns of chemokine receptor expression are important in determining lymphocyte migration to specific tissues and lymph nodes, and they can also be used to identify specific T cell subsets (36–42). Therefore, to provide some insight into potential functional differences among these ILC subsets, we performed a detailed examination of ILC chemokine receptor expression profiles (Fig. 4A, Table II).

Consistent with a Th1 phenotype, ILC1 had the highest frequency of CXCR3-expressing cells; however, CD4⁺ ILC1, which had a lower frequency of CXCR3⁺ cells than CD4⁻ ILC1, demonstrated the highest level of heterogeneity in chemokine receptor expression. A higher frequency of CD4⁺ ILC1 than CD4⁻ ILC1 expressed CLA, a skin-homing receptor, and a significant percentage of CD4⁺, but not CD4⁻, ILC1 expressed CCR4, CCR10, and CXCR5. Of particular note, the coexpression of CCR7 and CD62L on a high frequency of all ILC1 populations distinguished ILC1 from ILC2 and ILC3, which expressed CD62L but not CCR7.

ILC2, like Th2 cells, were nearly all CCR4⁺. ILC2 also had the highest frequency of cells positive for CCR4, CCR6, and CCR9, chemokine receptors important in homing to barrier surfaces of the skin, lung, and gut (43, 44). ILC3 were very heterogeneous in their chemokine receptor expression; however, other than CD62L, CLA was the most frequently expressed surface receptor on ILC3 of the chemokine receptors and adhesion molecules examined.

As with the chemokine receptors, we found substantial heterogeneity in the expression of costimulatory markers among the different peripheral blood ILC subsets (Fig. 4B, Table II). The majority of all ILC1 subsets expressed CD28, CD27, and CD2, costimulatory markers that are constitutively expressed on T cells (45–47). In contrast, ILC2 expressed CD28, but not CD27 or CD2, whereas a fraction of ILC3 expressed CD2 and CD28 but not CD27. Interestingly, we found that ILC1, but not ILC2 or ILC3, expressed CD45RO, a marker used to identify memory T lymphocyte subsets (Fig. 4B, Table II) (48). When we examined ILC subsets in cord blood, all ILC subsets were present and expressed similar patterns of CD28 compared with adult PBMC, yet CD45RO was low to absent in cord blood ILC (Supplemental Fig. 4).

These data suggest unique functions or migratory patterns for CD4⁺ ILC1 compared with other ILC1 populations, but the similarities in CCR7, costimulatory marker, and CD45RO expression in ILC1 suggest common modes of regulation among the ILC1 subsets.

To further examine factors that may differentially regulate the ILC subsets, we examined whether these subsets were responsive to IL-6, a cytokine that is thought to be important in autoimmune disease pathogenesis (49). Only CD4⁺ ILC1 had a high frequency of cells expressing both IL-6Rα and gp130 (Fig. 5A, 5C). In some individuals, ILC2 also had substantial expression of IL-6Rα, although the majority of IL-6Rα⁺ ILC2 did not coexpress gp130. Importantly, IL-6 responsiveness, as measured by STAT-3 and STAT-1 phosphorylation after IL-6 stimulation, paralleled the frequency of IL-6Rα and gp130 coexpression on the various ILC subsets (Fig. 5B, 5C). Thus, compared with other ILC subsets, CD4⁺ ILC1 are uniquely IL-6 responsive.

Although CD4⁺ ILC1 exhibited substantial heterogeneity in surface marker expression, we were interested in determining whether there was a core transcriptional profile that distinguished CD4⁺ ILC1 from CD4⁻ ILC1. To address this question, we performed RNA sequencing (RNAseq) on peripheral blood CD4⁺ and CD4⁻ ILC1 isolated from three healthy control subjects. We found that a total of 66 genes was differentially expressed between CD4⁺ and CD4⁻ ILC1, with 50 genes that were upregulated and 16 that were downregulated in CD4⁺ ILC1 compared with CD4⁻ ILC1 (Fig. 6). A variety of cellular processes were represented by these differentially expressed genes; only a small number were immune-regulatory genes (50, 51). Of these immune-related genes, CD4, CD8α, CD8β, and CRTAM are also differentially expressed between CD4⁺ and CD8⁺ T lymphocytes (52). In addition, the IL-18R subunit IL18R1 and AICD genes were expressed more highly in CD4⁺ ILC1. Although the best characterized role for AICD is in B cell class switching, several reports also suggest that it is involved in innate antiviral immunity (53). These data indicate that a small set of genes does differentiate CD4⁺ from CD4⁻ ILC1.

To apply these phenotypic analyses to a systemic disease, we chose to examine ILC subset frequencies in SSc, an autoimmune disease in which vascular dysfunction and fibrosis of the skin and other organs lead to substantial morbidity and mortality (54). Although there is clearly immune dysregulation in SSc, the mechanisms involved in disease pathogenesis are poorly understood. Type 2 inflammatory cytokines, particularly IL-13, are thought to be central drivers of fibrosis in SSc, although Th17 and Th22 cells have also been implicated in SSc disease pathogenesis (55–58). When we compared ILC subset frequencies in PBMC from patients with SSc with those from age- and gender-matched control subjects, we were surprised to find that, although ILC2 frequencies were not significantly different, ILC1 and NKp44⁺ ILC3 frequencies were increased and NKp44⁻ ILC3 frequencies were decreased in SSc relative to the total peripheral blood ILC (Lin⁻ CD127⁺) population and total live CD45⁺ lymphocytes (Fig. 7A). Analyses of some of the subjects shown in Fig. 7A included staining for CD4. In this cohort, we found that the enrichment of ILC1 in SSc was primarily attributable to changes in CD4⁺ ILC (Fig. 7B). These data examining CD4⁺ and CD4⁻ ILC1 were confirmed in a second age- and gender-matched cohort, which demonstrated increased frequencies in CD4⁺ and CD4⁻ ILC1 in SSc within the total ILC population but significant increases only in CD4⁺ ILC1 within total live CD45⁺ lymphocytes (Fig. 7B). Expression patterns of the costimulatory markers CD2, CD27, and CD28 in ILC subsets from SSc were similar to those seen in controls (Fig. 7C).

Because IL-6 signaling has been implicated in inflammation and fibrosis in SSc (56, 59–61), we also investigated IL-6Rα expression on ILC subsets in the peripheral blood of SSc patients compared with age- and gender-matched controls. Interestingly, we found a decrease in the percentage of CD4⁺ ILC1 coexpressing IL-6Rα and gp130 in SSc and a decrease in IL-6Rα mean fluorescence intensity (MFI) on CD4⁺ ILC1 in SSc compared with controls (Fig. 7D).

Through a comprehensive evaluation of chemokine receptors, adhesion molecules, costimulatory markers, and cytokines expressed by human peripheral blood ILC, we characterized distinct patterns within the ILC subsets that may further delineate their migration and function. Of particular note, we discovered a previously unappreciated heterogeneity in the ILC1 subset, which included CD4⁺, CD8⁺, and DN populations. We were surprised to find that CD4⁺ and CD4⁻ ILC1, but not ILC2 or ILC3, expressed significant levels of intracellular CD3ε, suggesting a developmental relationship among the ILC1 populations that may be distinct from ILC2 and ILC3. Previous work demonstrated that NKp44⁻ ILC3 have the capacity to differentiate into NKp44⁺ ILC3 or ILC1 in vitro, depending on the cytokine milieu (30). Whether the ILC1 subsets that we describe are distinct from those that differentiate from ILC3 remains unclear.

Some NK cells also express intracellular CD3 components (62–64); however, the cell subsets in our analyses are distinct from classic NK cells based on a lack of CD16, CD56, perforin, and granzyme B expression (29). Unlike CD4⁺ ILC1, CD8⁺ ILC1 had high frequencies of Eomes and T-bet coexpression, suggesting that they may share significant phenotypic and functional characteristics with NK cells. In contrast to CD8⁺ and DN ILC1, which displayed prominent Th1-like phenotypes, a substantial percentage of CD4⁺ cells lacked CXCR3, T-bet, and IFN-γ expression. Interestingly, CD4⁺ ILC1 were more potent producers of TNF-α, GM-CSF, and IL-2 and also included populations expressing CCR4, CCR10, and CXCR5. Given these phenotypic and functional differences between CD4⁺ and CD4⁻ ILC1, CD4⁺ ILC1 may represent a distinct subset, whereas DN and CD8⁺ ILC1 are likely more related to one another. A re-examination of these different ILC1 populations that incorporates functional analyses beyond the Th1 paradigm may better define conditions in which ILC1 are simply a source of IFN-γ, as well as when they may modulate inflammatory responses through other pathways.

Comparisons of ILC1 with ILC2 and ILC3 also highlighted other phenotypic features that were distinct among these populations. Only the ILC1 subsets express the T cell memory marker CD45RO. In addition, a large percentage of CD4⁺ and CD4⁻ ILC1 coexpressed CD62L and CCR7, which was absent on ILC2 and ILC3. In this manner, the ILC1 subsets closely resemble central memory T lymphocytes in phenotype (65). CCR7 is involved in the trafficking of certain T cell subsets to secondary lymphoid organs; the expression of the CCR7 ligands CCL19 and CCL21 on nonhematopoietic cells in the lymph node allows colocalization of CCR7⁺ T cells and activated dendritic cells to more efficiently initiate an immune response (66, 67). Although other ILC subsets are also present in lymphoid organs, such as the tonsils (68), CCR7 expression on ILC1 may suggest that this population is more actively recruited to lymphoid organs, or preferentially localized to lymph node T cell areas, where they may be more intimately associated with activated dendritic cells and T cells. Recently, a distinct subset of memory CD4⁺ T lymphocytes was described that migrates from the skin to draining lymph nodes in a CCR7-dependent manner and then re-enters the peripheral circulation, which allows this subset to provide help to distal lymphoid and cutaneous tissues (69). It is interesting to speculate that a similar scenario may exist for ILC1. Distinct patterns of expression of the costimulatory markers CD2, CD27, and CD28 also suggest that ILC1 may differ from ILC2 and ILC3 in how they are activated, as well as in the cell types with which they interact.

An examination of peripheral blood ILC in SSc, a disease with prominent fibrotic features, revealed that ILC1 and NKp44⁺ ILC3 were increased in frequency, whereas NKp44⁻ ILC3 were decreased in subjects with SSc. Additionally, the increase in ILC1 frequencies was primarily attributable to changes in CD4⁺ ILC1, which express lower levels of T-bet and IFN-γ than CD4⁻ ILC1. This is of particular interest because T-bet–deficient mice display increased sensitivity to the bleomycin-induced dermal sclerosis model of SSc in a T cell–independent manner (70). In light of numerous studies in mouse models and in humans that implicate TNF-α and GM-CSF in the pathogenesis of several autoimmune diseases (71, 72), it is also intriguing that multiple ILC subsets can produce these cytokines and that CD4⁺ ILC1 appear to be more potent producers of TNF-α and GM-CSF than are CD4⁻ ILC1.

CD4⁺ ILC1 were also the most responsive to IL-6, another cytokine that is thought to play an important role in autoimmunity (49). IL-6 and soluble IL-6Rα were reported to be elevated in SSc (56, 59, 61, 73, 74). However, we were surprised to find that the frequencies of CD4⁺ ILC1 expressing IL-6Rα were decreased in SSc. Although these data suggest that increased IL-6 signaling in CD4⁺ ILC1 may not be critical to SSc pathogenesis, CD4⁺ ILC1, as well as ILC2, which can also express significant levels of IL-6Rα, may serve as sources of soluble IL-6Rα, which can amplify IL-6 responses in inflammatory conditions (75). IL-6Rα was shown to be downregulated upon T cell stimulation; thus, decreased IL-6Rα expression on ILC in SSc could also result from increased ILC activation during disease (76).

Because SSc is a very heterogeneous disease that may involve distinct pathogenic mechanisms in early versus late disease, it will be important to understand how ILC frequencies change in different stages and subsets of disease. We defined important phenotypic and functional features of ILC subsets in the peripheral blood. More in-depth analyses of how the ILC populations that we describe may differ from those in tissue will better define their roles in autoimmune disease pathogenesis and help to establish whether ILC2 are functionally different in SSc.

Currently, tocilizumab, an anti–IL-6R mAb, and abatacept (CTLA4-Ig), which blocks interactions between CD28 and CD80 or CD86, have shown promise in small studies (77–79), and both are in clinical trials for SSc. The differential expression of IL-6Rα and CD28 on ILC subsets suggests that not all ILC will be affected by these biologics to the same degree. Dysregulation of the innate immune compartment may be an important process that underlies the development of autoimmunity; thus, the ability of these therapies to appropriately target innate cells, such as ILC, is likely to have an important impact on therapeutic efficacy. Evaluation of how ILC subsets change during therapy, as well as whether changes in these subsets correlate with clinical response, will give insight into the mechanisms of disease and therapeutic efficacy and may aid the design of the next generation of biologics.

We thank Thien-Son Nguyen and all members of the BRI clinical core who aided in subject recruitment and samples management; Vivian Gersuk and the BRI genomics and bioinformatics core staff who performed the library construction, sequencing, and data processing for the RNAseq; and Alyssa Sheih for critical review of the manuscript.

The authors have no financial conflicts of interest.