Increased CD8⁺ tissue resident memory T cells, regulatory T cells and activated natural killer cells in systemic sclerosis lungs

Abstract

Objective

Multiple observations indicate a role for lymphocytes in driving autoimmunity in SSc. While T and NK cells have been studied in SSc whole blood and bronchoalveolar lavage fluid, their role remains unclear, partly because no studies have analysed these cell types in SSc-interstitial lung disease (ILD) lung tissue. This research aimed to identify and analyse the lymphoid subpopulations in SSc-ILD lung explants.

Methods

Lymphoid populations from 13 SSc-ILD and 6 healthy control (HC) lung explants were analysed using Seurat following single-cell RNA sequencing. Lymphoid clusters were identified by their differential gene expression. Absolute cell numbers and cell proportions in each cluster were compared between cohorts. Additional analyses were performed using pathway analysis, pseudotime and cell ligand-receptor interactions.

Results

Activated CD16⁺ NK cells, CD8⁺ tissue resident memory T cells and Treg cells were proportionately higher in SSc-ILD compared with HC lungs. Activated CD16⁺ NK cells in SSc-ILD showed upregulated granzyme B, IFN-γ and CD226. Amphiregulin, highly upregulated by NK cells, was predicted to interact with epidermal growth factor receptor on several bronchial epithelial cell populations. Shifts in CD8⁺ T cell populations indicated a transition from resting to effector to tissue resident phenotypes in SSc-ILD.

Conclusions

SSc-ILD lungs show activated lymphoid populations. Activated cytotoxic NK cells suggest they may kill alveolar epithelial cells, while their expression of amphiregulin suggests they may also induce bronchial epithelial cell hyperplasia. CD8⁺ T cells in SSc-ILD appear to transition from resting to the tissue resident memory phenotype.

Rheumatology key messages.

• Cytotoxic natural killer cells are observed in high frequency in SSc-ILD compared with healthy controls.

• Expression of amphiregulin is upregulated in natural killer cells and communicates with SSc-ILD basal cells.

• CD8⁺ and regulatory T cells are increased in number in SSc-ILD compared with healthy controls.

Introduction

Multiple observations point to an important role for lymphocytes in driving the autoimmune pathogenesis of SSc [1]. Activated T cells have been observed in affected tissues, both expanded and differentiated. Autoantibodies specific to SSc (e.g. anti-topoisomerase type 1/Scl-70) require T cell help for isotype class switching and maturation [2]. The skin in chronic graft vs host disease, a T cell–mediated disease in which the graft (donor) T cells attack the host, appears nearly histologically identical to SSc [3]. Cytotoxic T cells have been implicated in endothelial injury [4]. Additionally, the strongest genetic associations with SSc are in HLA class II genes, suggesting cognate T cell help in the maturation of autoantibody-producing B cells [5].

Lymphoid cells (LCs) in SSc have previously been studied primarily in peripheral blood and skin and to a lesser extent in bronchoalveolar fluid (BALF). Increased circulating CD4⁺ and decreased CD8⁺ T cells have been observed in SSc compared with control blood [6]. Another study showed that total T cell numbers were reduced [7]. Of the CD4⁺ T cell populations, Th2 and Th17 populations in blood and skin are associated with profibrotic phenotypes. In SSc blood and skin, Th2 cells are associated with profibrotic cytokines, IL-4, IL-5 and IL-13, while Th17 cells upregulate production of IL-17, leading to increased type I collagen [8, 9]. Studies of Tregs in SSc have yielded variable results; however, most studies have observed reduced numbers of Tregs in the blood and possible Treg dysfunction [10].

In SSc skin, an elevated number of CD8⁺ T cells have been observed in active disease, providing evidence for the presence of CD8⁺ topoisomerase type 1 reactive T cells in SSc [11]. In SSc-ILD, a shift towards the Th2 phenotype was observed in peripheral blood T cells in patients with SSc-associated interstitial lung disease (SSc-ILD) [12]. In addition, a correlation was observed between lower forced vital capacity and the degree of Th2 polarization [12]. On the other hand, a study that compared immune cells from BALF between SSc patients with and without lung fibrosis did not observe significant differences in the number of lymphocyte subpopulations, including CD4⁺ T, CD8⁺ T and NK cells [13].

Several studies have also shown alterations in NK cell populations in SSc. Reduced NK cell numbers in SSc patients have been observed in the blood of patients with limited scleroderma (lcSSc) [7]. Other investigators noted increased NK cells in the blood of diffuse scleroderma (dcSSc) patients [14]. What is more, SSc patients had reduced CD56^hi [15, 16] and dysfunctional NK cells [7, 14]. In contrast to these studies, SSc patients positive for Scl-70 or ACA showed an increased number of NK cells compared with ACA-negative SSc patients [15]. Despite the growing number of observations on T and NK cells, these subpopulations are not well described in the lungs of patients with SSc-ILD, a leading cause of death among SSc patients.

Methods

Lung explants from SSc-ILD patients and healthy controls (HCs) were processed for single-cell RNA sequencing using 10X Genomics. Detailed methods for immunohistochemistry, culturing of scleroderma skin fibroblasts with PDGF-DD and CITE-seq can be found in the Supplementary Data S1, available at Rheumatology online. Demographic data on the patient samples are described in Supplementary Table S1, available at Rheumatology online. This research was approved by the University of Pittsburgh Institutional Review Board (19100326-003).

Results

SSc-ILD lungs show altered lymphoid cell populations

Cells from each SSc-ILD and HC sample were transcriptionally profiled utilizing single-cell RNA sequencing (scRNA-seq) and clustered using Seurat (Supplementary Fig. S1, available at Rheumatology online). A cluster of LCs was initially isolated according to elevation of CD45 and further identified by plasma cell, T cell and NK cell markers. This cluster was subclustered into 16 subpopulations of LCs (Fig. 1A). Each subcluster included cells from multiple samples (Supplementary Fig. S2, available at Rheumatology online). Upon dividing cells based on their health status (SSc vs HC), Uniform Manifold Approximation and Projection (UMAP) subclustering showed several striking shifts in the lymphoid populations from SSc-ILD compared with HC lungs (Fig. 1B). Each subcluster of LCs was first identified according to their differentially expressed genes (Supplementary Table S2) through an extensive literature review, pathway analysis, and use of online compendia of human gene expression. Expression of key genes identified each cell population (Supplementary Fig. 2A, Supplementary Table S2, available at Rheumatology online).

Figure 1.

Dimensional reduction of scRNA-seq data of lymphoid populations from SSc-ILD. (A) UMAP of lymphoid subpopulations in SSc-ILD and HC cohorts. (B) UMAP of lymphoid subpopulations divided by SSc-ILD and HC cohorts. (C) Pseudotime of NK cell populations. The transition of dark colour in the cluster 0 CD16⁺ NK population into the bright yellow in the activated NK population observed in cluster 8 suggests the transition of NK cells from an inactive to an activated, cytotoxic phenotype in the SSc-ILD cohort. (D) Pseudotime of CD8⁺ T cell populations. The dark colour labelling the naïve CD8⁺ population suggests the start of the transition. Green dominates the CD8⁺ effector population while yellow is more prevalent in the CD8⁺ TRM

Activated NK cells are expanded in SSc-ILD

In the 16 subpopulations of LCs, we identified three NK cell populations (Fig. 2B). CD16⁺ (FCGR3A), CD56^dim (NCAM1) NK cells in cluster 0 comprised the largest cluster in HC samples (Fig. 1A). Compared with other NK cell populations, CD16⁺CD56^dim NK cells in cluster 0 differentially expressed KLRF1 (NKp80), FGFBP2, FCGR3A, TYROBP, GNLY, SPON2 and PRF1 (Fig. 2B). Cluster 0 CD16⁺CD56^dim NK cells expressed fewer upregulated IFN genes than the other NK cell populations. CD69 and IL-2RB were also less robustly expressed compared with the other NK cell clusters (Fig. 2B). We identified two other NK cell clusters, a second population of CD16⁺CD56^dim NK cells in cluster 8 and a population of CD16⁻CD56^bright NK cells in cluster 12. CD16⁺CD56^dim NK cells in cluster 8 showed higher expression of GZMB, CCL3, CCL4L2, HAVCR2 and IRF8 (Fig. 2B). Consistent with elevated GZMB RNA experssion, increased granzyme B staining was observed in the NK cells from SSc-ILD lung explants compared with HC lungs on immunohistochemistry (Supplementary Fig. S3, available at Rheumatology online). Amphiregulin (AREG) was also upregulated in cluster 8 and in the top 20 differentially expressed genes (DEG) (Supplementary Table S2, available at Rheumatology online). The NK cells in this cluster were observed to be activated due to highly upregulated expression of IFNG (IFN-γ) as well as TAGAP, LAG3 and HLA-DRA [17]. Notably, expression of STAT3 and STAT4 were also elevated in this NK cell population.

Figure 2.

Gene expression associated with lymphoid populations. (A) Dot plot of DEG for each lymphoid population. (B) Dot plot of upregulated gene expression that differentiates between the three NK cell subpopulations. (C) Dot plot of upregulated genes, identifying CD8⁺ T cell subpopulations. (D) Dot plot of upregulated genes identifying CD4⁺ T cell subpopulations

Cluster 12 NK cells showed selectively upregulated expression of XCL1, XCL2, CCL3, TYROBP and FCER1G (Fig. 2B). NK cells in this cluster were especially different from the other NK clusters, showing highly upregulated expression of CD69, a tissue residency gene. Thus these cells are phenotypically tissue resident NK (TrNK) cells. IL-2RB, EOMES, KLRC1 (NKG2A) were also highly expressed, the latter two suggesting a unique population of NK cells with an adaptive-like phenotype [18].

Resting, tissue resident memory (TRM) and effector memory (EM) CD8⁺ T cells

Resting CD8⁺ T cells (cluster 1) comprised the second largest cluster, identified by upregulation of CD3D, CD8A, GZMH, NKG7 and KLRG1 [19] (Fig. 2C). Cluster 5 represented CD8⁺ TRM cells with increased gene expression of CD69, ITGAE and ZNF683 (HOBIT) and low-level expression of S1PR1 [20, 21]. KLRC1 was one of the most highly DEG, upregulated in both NK and CD8⁺ T cells. It is an inhibitory receptor expressed by cytotoxic lymphocytes with the capacity for self-tolerance [22]. Cluster 6 CD8⁺ T cells expressed GZMK, CCL5, CRTAM, DUSP2, CST7 and KLRG1, consistent with an effector memory population [21].

TRM, central memory (CM), Th17 and Treg CD4⁺ T cell subpopulations

Cluster 2 is a population of CD4⁺ T lymphocytes with modest gene expression of CD4 along with upregulation of CD69 and ITGAE (CD103) and downregulation of S1PR1, indicating a TRM population [23, 24] (Fig. 2D). CD4⁺ CM T cells are identified as cluster 4 by upregulation of CCR7 and SELL (CD62L) [25]. PRDM1 (BLIMP1) is also upregulated, suggesting a memory phenotype as well. Cluster 3 had very few DEG with very modestly upregulated gene expression, including IL7R (CD127), CCR6, and GZMK. Thus this cluster may represent mucosal-associated invariant T cells, an innate-like T cell population [26].

Another CD4⁺ population, Th17 cells (cluster 10), expressed high levels of CCL20, CCR6 and KLRB1 with low-level, detectable expression of IL17A [27]. KLRB1 is associated with an active Th17 phenotype [27]. Tregs (cluster 9) were identified by upregulation of CD4, FOXP3 and IL2RA (CD25) (Fig. 2D). Cluster 7 DEG indicates a T cell population with many upregulated genes involved in protein folding and metabolism, labelled as endoplasmic reticulum (ER) stressed T cells; however, we were unable to further specify the phenotype of these cells (Fig. 2D).

Other populations identified in lung tissue

Four additional, lymphoid islands were identified by distinct DEG (Fig. 2A): plasma cells, proliferating lymphocytes, upregulated IFN T cells and innate lymphoid type 3 (ILC3) cells. ILC3 cells (cluster 15) were identified according to genes expressed by ILC3 in multiple tissues: KIT, TNFSF11 and TNFSF4 [28]. These genes comprise the top 5 DEG genes for cluster 15. LMNA is a lung-restricted ILC3 gene, indicative of tissue residency [28]. Average gene expressions in SSc-ILD and HCs for each cluster can be found in Supplementary Table S3, available at Rheumatology online.

Cell counts and population shifts

UMAP comparing SSc-ILD samples with HC samples demonstrated population differences in the subclusters (Fig. 1C). Cluster 0 CD16⁺ NK cells were significantly decreased in the SSc-ILD cohort compared with the HC group (Fig. 3A). This was accompanied by a strong trend approaching statistical significance towards increased numbers of activated CD16⁺ NK cells (cluster 8), clustering immediately adjacent to cluster 0. Trajectory analysis suggested that NK cells in HC lungs transition to an activated NK cell phenotype in SSc-ILD lungs (Fig. 1C).

Figure 3.

Median proportions of cells per cluster and altered gene expression in NK cells. (A) Graph depicting the proportion of cells in each cluster according to disease. Statistical significance is indicated by asterisks. (B) Violin plot of gene expression of IFN-γ-regulated genes between CD16⁺ NK cells (Fig. 1, cluster 0) and activated CD16⁺ NK cells (Fig. 1, cluster 8). Each dot represents a cell in that cluster

CD8⁺ TRM and Treg cells showed increased proportions in SSc-ILD compared with HC lungs (Fig. 3A). On the other hand, resting CD8⁺ T cells and ER stressed T cells showed statistically significantly decreased cell numbers in SSc-ILD compared with HCs (Fig. 3A). The Th17 cluster also trended towards decreased cell numbers in SSc-ILD compared with HC lungs. Trajectory analysis suggested that resting CD8⁺ T cells, seen mainly in HC lungs, progressively differentiate into CD8⁺ EM and then CD8⁺ TRM cells, seen mainly in SSc-ILD (Fig. 1D).

Although the proportion of CD4⁺ TRM cells did not show a significant shift in SSc-ILD compared with HC lungs, there was an apparent change in the state of these cells, indicated by a shift within the cluster comparing SSc-ILD with HCs (cluster 2, Fig. 1B). However, primarily one HC sample and two SSc-ILD samples contributed to most of the cells in this cluster comprising this apparent shift (Supplementary Fig. S2, available at Rheumatology online). Since this shift was seen in only a restricted number of the samples, we did not analyse this cluster further.

Tregs and CD8⁺ TRM T cells have upregulated genes associated with T cell exhaustion and fibrosis pathways. T cell receptor signalling, T cell exhaustion signalling, as well as Th1 and Th2 pathways were upregulated in CD8⁺ TRM cells (Supplementary Table S4A, available at Rheumatology online). Tregs showed upregulation of the same pathways, as well as Th17 and IL-6 signalling (Fig. 6C, Supplementary Table S4B, available at Rheumatology online). Full pathway analysis for both CD8⁺ TRM cells and Tregs are presented in Supplementary Table S4, available at Rheumatology online.

Figure 6.

Connectome analysis and Ingenuity Pathway Analysis (IPA) of Treg cell DEGs. (A) Circos plot of cell–cell edges (interactions) between Treg (Fig. 1, cluster 9) genes (ligand) and remaining lymphoid populations (receptor) in SSc-ILD cohort only. (B) Circos plot of cell–cell edges between Treg genes (receptor) and remaining lymphoid populations (ligand) in SSc-ILD cohort only. (C) Abbreviated IPA of top pathways from DEG (full pathways are provided in Supplementary Table S4C)

CITE-seq confirmed T and NK cells subpopulations

Our findings using scRNA-seq were further confirmed using CITE-seq on an additional two HC and two SSc-ILD lung samples, allowing us to compare both transcriptomic data and expression of protein surface markers to identify the subpopulations (Supplementary Table S5, available at Rheumatology online). Subpopulations of T cells expressing CD8A mRNA expressed CD8 protein, and NK cells expressing NCAM1 and low-level FCGR3A mRNAs expressed CD56 and dimly CD16 proteins, respectively (Supplementary Fig. S4, available at Rheumatology online). In addition, protein expression of the CD45 isoform CD45RA, typically expressed on naïve T cells, marked resting CD8 T cells (cluster 2), while CD45RO, typically marking memory T cells, was expressed by TRM and EM T cells [29]. CD45RA, also a marker of mature NK cells (CD56^dim CD16⁺), was expressed by NK cell clusters, whereas CD45RO, rarely expressed on NK cells, was detected at a much lower level [30] (Supplementary Fig. S4, available at Rheumatology online).

Activated NK cells suggest pro-inflammatory, profibrotic interactions with T cells via pathway analysis

Several pathways associated with inflammation and T cell differentiation were highly upregulated in the transcriptome of the activated NK cell population (cluster 8). Upregulated pathways included crosstalk between dendritic cells and natural killer cells and IL-23, IL-6, and TREM2 signalling pathways. In addition, pathways implicated in fibrosis including the hepatic fibrosis and pulmonary healing signalling pathways were upregulated. HMGB1, another driver of inflammation, and the senescence pathways were also upregulated in SSc-ILD-activated NK cells (Supplementary Table S4C, available at Rheumatology online).

Interactions between NK cells and with other cell populations

Receptor–ligand interactions were visualized among the lymphoid subpopulations as well as among all cell types represented in the lung tissue to assess how lymphoid populations may promote profibrotic pathways. We initially focused on analysing receptor–ligand interactions increased in SSc-ILD lymphoid populations: activated CD16⁺ NK cells, CD8⁺ TRM and Tregs. Upregulated IFNG (encoding IFN-γ, Fig. 3B) expressed by activated SSc-ILD CD16⁺ NK (Fig. 1A, cluster 8) was predicted to interact with IFNGR1 (IFN-γ receptor 1) expressed by Th17, TrNK and CD4⁺ TRM cells and IFNGR2 expressed by plasma cells (Fig. 4A). Activated NK cells also showed upregulated chemokines, CCL3 (Fig. 4C), CCL4 (Fig. 4C), CCL5 and CCL3L1, all predicted to interact with CCR5 receptors expressed by CD8⁺ EMT cells. CCL5 is also predicted to interact with SDC1 expressed by plasma cells (Fig. 4A). Upregulated expression of HLA-A, -B and -C was predicted to interact with CD3D expressed by multiple LCs.

Figure 4.

Connectome analysis of activated NK cells. Cell–cell edges (interactions) represent high likelihood and specificity of interaction between ligand and receptor. (A) Circos plot of cell–cell edges between cluster 8 activated NK cell genes (ligand) and remaining lymphoid populations (receptor) in SSc-ILD cohort only. (B) Circos plot of cell–cell edges between cluster 8 activated NK cell genes (receptor) and remaining lymphoid populations (ligand) in SSc-ILD cohort only. (C) Violin plot showing CCL3, CCL4 and CD226 gene expressions in SSc-ILD and HC groups

Activated CD16⁺ NK cells were predicted to receive HLA class 1 signals from resting CD8⁺ T cells, proliferating NK and T cells, and Tregs to KLRD1, KIR2DL3 and KIR3DL2. These three genes encode inhibitory receptors for NK cell activation. Activated CD16⁺ NK cells were also predicted to receive signals from several integrin receptors: ITGAX, ITGA4 and ITGB2 from multiple upregulated integrin ligands, including CD40L (from CD4⁺ TRM), NCAM1 and ICAM1 (from TrNK), ICAM2 (from plasma cells) and ICAM3 (from Th17 cells) (Fig. 4B).

We then looked at NK cell interactions with other cell populations in the lung, examining the dataset before extracting the lymphoid compartment (Supplementary Fig. S1, available at Rheumatology online). Focusing first on epithelial cell populations, AREG, encoding amphiregulin, a ligand for the epidermal growth factor receptor (EGFR), was highly upregulated in SSc-ILD in NK cells (Fig. 5A, C and D). Connectome analysis showed AREG ligand interacting with EGFR on goblet, basal, and alveolar epithelium type 1 (AT1) and EGBB3 on goblet, AT1, ciliated, and AT2 cells (Fig. 5A). GZMB was seen to interact with the PGRMC1 receptor on AT1, AT2, and basal cells and IL7, IL18, and MDK from ciliated, basal, and goblet cells, respectively (Fig. 5B).

Figure 5.

Connectome analysis of NK cells with epithelial cell types in SSc-ILD lung. (A) Circos plot of cell–cell edges (interaction) between cluster 10 NK cell genes (ligand) and epithelial cell populations (receptor) in SSc-ILD cohort only. (B) Circos plot of cell–cell edges between cluster 10 NK cells (Fig. S1) genes (receptor) and epithelial cell populations (ligand) in SSc-ILD cohort only. (C) Violin plot depicting AREG expression from SSc-ILD NK cells subpopulations. (D) Violin plots depicting AREG expression in the total NK cell population and EGFR upregulated in SSc-ILD basal and goblet cell populations

Focusing the connectome analyses on vascular (Supplementary Fig. S5, available at Rheumatology online) and fibroblast cell (Supplementary Fig. S6, available at Rheumatology online) populations revealed upregulated NK PDGF-D interaction with PDGRFB on pericytes and fibroblasts; TGFB1 interaction with ACVRL1 (ALK-1) on capillary endothelial cells and ENG on arterial endothelial cells; and IFNG on capillary and arterial endothelial cells, as well as other interactions. To explore the potential role of PDGF-DD in mediating fibroblast activation, we treated pulmonary fibroblasts from SSc-ILD lungs with PDGF-DD compared with PDGF-AB. PDGF-DD induced expression of collagen type X alpha 1 chain (Col10A1), while alpha-1 type I collagen (Col1A1) and smooth muscle actin alpha 2 (ACTA2) were unchanged (Supplementary Fig. S4C, available at Rheumatology online). PDGF-AB showed no effect on either of the collagen genes or smooth muscle actin, suggesting an alternate activation pathway. Notably, Col10A1 is strongly upregulated in myofibroblasts in both SSc-ILD and in SSc skin [31, 32].

Interactions between Tregs and other lymphoid cells

We next examined Treg ligand–receptor interactions. CD70 ligand on Tregs was shown to interact with CD27 found on Tregs themselves, as well as on CD4⁺ CM, CD4⁺ TRM and plasma cells (Fig. 6A). This same interaction was observed when analysing receptors on SSc-ILD Tregs (Fig. 6B). Recent studies suggest that upregulated CD70 and interaction with CD27 inhibits Treg suppressive activity [33, 34]. Pathway analysis indicated T cell activation and exhaustion (Fig. 6C, Supplementary Table S4C, available at Rheumatology online).

Interactions between CD8⁺ TRM cells and other lymphoid cells

Connectome analysis of SSc-ILD CD8⁺ TRM showed some similar interactions as those seen in NK and Treg populations. Upregulated CCL4 and CCL5 ligands interacted with CCR5 and SCD1. CD8⁺ TRM cells were predicted to receive CD70 signals from activated NK cells (Supplementary Fig. S7A, available at Rheumatology online) and to receive CCL3, CCL4 and CCL5 signals from multiple lymphoid populations to CCR5 (Supplementary Fig. S7B, available at Rheumatology online).

Altered SSc-ILD phenotypes do not correlate with patient age

The number of peripheral NK cells increases with age in healthy adults while their cytotoxicity diminishes [35]. Naïve T cells decrease with age as the thymus shrinks over time. T cell subsets like CD8⁺ memory T cells expand over the course of a lifetime and express more senescent markers and carry impaired cell proliferation [36]. Because our healthy cohort was younger on average compared with the SSc-ILD group, we tested if increased T cell and NK cell subsets correlated with patient age. Neither the proportion of cells in cluster 8 CD16⁺ NK cells nor cluster 5 CD8⁺ TRM cells showed a significant correlation with age [r(17) = −0.32, P = 0.18 and r(17) = −0.22, P = 0.36, respectively].

Discussion

In addition to discovering several populations of NK cells, CD4⁺ T cells and CD8⁺ T cells in both SSc-ILD and HC lungs, we noted shifts in several lymphoid populations, notably increased proportions of activated NK, CD8⁺ TRM and Treg cells in SSc-ILD lungs. These were accompanied by remarkable changes in cell pathways and predicted ligand–receptor interactions, implicating LCs in multiple aspects of pathogenesis.

Activated NK cells expressing high IFN-γ were seen almost exclusively in SSc-ILD. The CD16⁺CD56^dim NK cells associated with SSc-ILD clustered separately due to upregulation of several pro-inflammatory genes, including IFN-γ, IFN regulatory factor 8 (IRF8) and GZMB, the most cytotoxic enzyme made by human NK cells [37]. IRF8 deficiency in humans has been shown to impair NK function and reduce the cytotoxicity of NK cells [38]. NK cells can directly influence adaptive immune responses, involving CD8⁺ T and CD4⁺ T cells through cell-mediated cytotoxicity and the killing of activated T cells and indirectly through killing of endothelial cells via activation by CX3CL1, leading to vascular injury [39].

There have been conflicting data on the frequency and functionality of NK cells in SSc, studied primarily in whole blood. One study observed increased NK cells and markers of early activation in NK cells from patients with dcSSc [14]. We observed very few activated NK cells in HC lungs. Trajectory analysis suggested that CD16⁺CD56^dim NK cells in SSc-ILD are derived from HC CD16⁺CD56⁺ NK cells that have undergone an activation step associated with upregulation of IFNG, pro-inflammatory chemokines, and GZMB. Pathway analysis supported IL-23 signalling, a previously described regulator of IFN-γ by NK cells [40], via upregulated STAT3 in SSc-ILD, CD16⁺CD56^dim NK cell activation. However, we did not see upregulated IL-23A expression in any cell scRNA-seq population (not shown).

Although T cells can also produce IFN-γ and granzyme B, SSc-ILD CD16⁺CD56^dim NK cells expressed far higher levels of both these genes. In NK cells, IFN-γ is induced by type I IFN via STAT 4 signalling and by IL-12 and IL-18 [41]. Overproduction of IFN-γ reinforces T follicular cell differentiation, leading to germinal cell formation, and B cell proliferation and differentiation [41]. Furthermore, IFN-γ enhances macrophage function, increases major histocompatibility complex expression and mobilization of leukocytes to sites of infection.

Upregulated GZMB strongly suggests that the activated NK population has enhanced killer activity. PRF1, encoding perforin, was also most highly expressed by the NK cell populations and modestly upregulated in both cluster 0 HC and cluster 8 SSc-ILD NK cells. Ligand–receptor plots indicated multiple potential interactions regulate NK activity in SSc-ILD lungs. CD226, an activating receptor for NK cells, was expressed more highly on activated SSc-ILD NK cells (Fig. 4C). HLA molecules protect cells from NK cell killing. The loss of AT1 cells in SSc-ILD suggests these might be targets of activated NK cells [32]. HLA-G was most highly expressed by AT1 cells and strongly downregulated in SSc-ILD [42]. HLA-G inhibits NK cell activation through KIR2DL4 [43], suggesting that its downregulation might enable NK cell killing of AT1 cells. Similarly, NK cells in primary biliary cirrhosis (PBC) can kill biliary epithelial cells [44]. Furthermore, in PBC, NK cell killing releases autoantigen and activates autoreactive T cells.

Notably, NK cells in SSc-ILD also showed dramatically increased expression of the gene encoding amphiregulin (AREG), a ligand for the EGFR shown on connectome analysis to interact with basal cells, goblet cells, AT1, AT2, and ciliated cells. This observation suggests that AREG secreted by SSc-ILD NK cells might be stimulating increased basal cell proliferation, a feature of SSc-ILD lungs [42]. Indeed, AREG can induce basal cell as well as goblet cell hyperplasia [45].

Trajectory analysis suggests a global T cell stimulus is driving the loss of resting CD8⁺ T cells and expansion of CD8⁺ TRM cells. The expanded population of CD8⁺ TRM cells in SSc-ILD suggest these cells may also play a role in pathogenesis. CD8⁺ TRM cells can persist in the tissues in the absence of ongoing antigen stimulation [46]. TRM cells can respond to self-antigen and induce autoimmunity as well as contribute to pulmonary inflammation and fibrosis [47]. IL-10 and TGF-β are considered upstream regulators of TRM differentiation.

Most reports suggest there are fewer Tregs in SSc compared with HCs [48]. Tregs were increased in the SSc-ILD group, and this observation may account for fewer circulating Tregs. Tregs upregulated genes in the IL-6 signalling pathway, known to promote T cell proliferation. Treg/Th17 cell skewing was opposite from what might be expected from the effects of IL-6 and IL-23. IL-6 suppresses FOXP3, which in turn decreases peripheral induced Tregs and increases Th17 cells. IL-6 is essential in differentiation into as well as maintenance of Th17 cells [49]. This increased ratio between Th17 and Treg subsets has been observed in SSc patients [50]. IL-6 from fibroblasts in inflamed joints has been identified to promote Treg conversion to Th17 cells [49].

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The data underlying this article are available in the article and in its online supplementary material.

Authors’ contributions

Cristina Padilla: Seurat and connectome analyses, statistics, and writing bulk of the paper and editing; Eleanor Valenzi: provided data and initial analysis with Seurat; Tracy Tabib: analysis with pseudotime, processing samples for single cell RNA-seq; Banafsheh Nazari: granzyme B stain of lung samples and writing portion of methods and results; John Sembrat: collection of lung samples, providing demographic table, and review of paper; Mauricio Rojas: reviewing paper, edits, and expertise in analyzing different genes; Robert Lafyatis: provided guidance, assistance creating figures, reviewing/editing paper, providing references.

Funding

This work was supported by the National Heart, Lung, and Blood Institute (5T32 HL007563, to C.P.) and the National Institute of Arthritis and Musculoskeletal and Skin Diseases (2P50 AR060780, to R.L.).

Disclosure statement: R.L. has received grants or contracts from Corbus, Formation, Moderna, Regeneron, AstraZeneca, Pfizer and Mitsubishi and consulting fees from Pfizer, Bristol Myers Squibb, Boehringer Ingelheim, Formation, Sanofi, Boehringer Mannheim, Merck, Genentech/Roche and Biogen. The remaining authors have declared no conflicts of interest.