Human hypofunctional NCF1 variants promote pulmonary fibrosis in the bleomycin-induced mouse model and systemic sclerosis patients via expansion of SPP1⁺ monocytes-derived macrophages

Abstract

Objective

We assessed the role of a SLE causal hypofunctional variant, NCF1-p.Arg90His (p.R90H) substitution, in systemic sclerosis (SSc).

Methods

Association of NCF1-H90 with SSc was performed in case-control cohorts, bleomycin (BLM)-treated Ncf1-R90 C57BL/6 wildtype (WT) and Ncf1-H90 knock-in (KI) littermates. PBMC cell subsets were analyzed by CyTOF.

Results

The NCF1-H90 allele associated with risk for diffuse cutaneous SSc (dcSSc) in Chinese and European Americans, and lung fibrosis in Chinese SSc patients (OR = 2.09, P = 7.96E-10). Low copy number of NCF1 associated with lung fibrosis in European Americans (OR = 4.33, P = 2.60E-2). BLM-treated KI mice demonstrated increased pulmonary fibrosis, exhibiting activated type I interferon signature, elevated Spp1, Ccl2, Arg1, Timp1 and Il6 expression, enriched macrophage scores in lung tissues. In a longitudinal observation cohort, homozygous H90 SSc patients at baseline had increased ANA titers, ATA seropositivity and ACA seronegativity, increased incidence of lung fibrosis, and gender-age-lung physiology (GAP) index, elevated mRSS and elevated plasma osteopontin (OPN, SPP1), CCL2, ARG1, TIMP-1 and IL-6. These H90 SSc patients sustained elevated mRSS during follow-up years with decreased survival. The 0, 1 and 2 copies of H90 carriage in SSc PBMCs exhibited dose-dependent increases in profibrotic CD14⁺CD68⁺CD11b⁺Tim3⁺monocytes. Elevated OPN, CCL2 and ARG1 in CD68⁺CD11b⁺monocyte-derived macrophages (MoMs) from H90 patients were decreased after co-culturing with anti-CCL2 antibody.

Conclusion

Low NCF1 activity increases risk for the development of dcSSc and lung fibrosis via expanding profibrotic SPP1⁺MoMs in a CCL2-dependent manner, contributing to severity of lung fibrosis in both BLM-treated mice and SSc patients.

Introduction

Systemic sclerosis, characterized by excess deposition of extracellular matrix in the skin and internal organs, is often associated with interstitial lung disease (ILD)¹. While the disease mechanisms are completely unknown, genetic contribution of over 30 SSc-associated risk loci have been identified in large-scale association studies^2–4. We previously identified a hypofunctional variant (p.R90H) in NCF1, encoding a regulatory subunit of phagocyte NADPH oxidase 2 (NOX2), associated with robust risk of multiple autoimmune diseases including systemic lupus erythematosus (SLE), primary Sjögren’s syndrome (pSS) and rheumatoid arthritis (RA)⁵. Additionally, decreased copy numbers of NCF1 predisposed to SLE⁵. The hypofunctional H90-NCF1 is an established causal variant of SLE, affecting efferocytosis efficiency in macrophages⁶, and endosomal TLR-signaling in plasmacytoid dendritic cells (pDC)^7,8 and B cells resulting in breaks of immune tolerance and increased humoral immune responses to nucleic-acid containing antigens⁹. Considering pleiotropic effects of this variant and upregulated IFN-I signature commonly observed in patients affected with SSc, SLE and pSS^10,11, we hypothesized that NCF1-H90 variant contributes to susceptibility of SSc. SSc is classified into limited cutaneous (lcSSc) and diffuse cutaneous (dcSSc)¹² based on the extent of skin involvement. Patients affected with dcSSc tend to have rapid progression of multiple organ involvement and an early incidence of ILD — a leading cause of death in SSc ^13,14. A commonly used animal model of pulmonary fibrosis is a one-week BLM infusion by osmotic minipump in C57BL/6 mice¹⁵. Here we explored functional consequences NCF1-H90 variant in the development of disease manifestations in patients affected with SSc and in BLM-treated H90.B6 KI mice.

Results

Association of hypofunctional NCF1-H90 variant with SSc risks with stronger association in patients with dcSSc or lung fibrosis.

In 529 SSc patients and 3,020 controls, we observed association of H90 allele of NCF1 with SSc in Chinese patients (OR = 1.93, P = 5.97E-12), in which more robust association signals were observed with the dcSSc subset (OR = 2.25, P = 4.25E-11) and with lung fibrosis (OR = 2.09, P = 7.96E-10). Another variant, the GTGT sequence within exon 2 of NCF1, distinguishes a functional NCF1 gene from its neighboring NCF1B and NCF1C pseudogenes containing a GT deletion (ΔGT), resulting in copy number variations (CNVs)⁵. Because the CNVs were found in < 2% of East Asians and their impact on association analyses of NCF1-H90 in Asian data sets was expected to be negligible⁵, Chinese subjects were not assessed for the CNVs. However, >15% of European Americans (EA) exhibit CNVs of this locus⁵, hence we assessed CNVs in EA and genotyped the H90 variant only in subjects carrying normal copy numbers of NCF1. In EA, we confirmed the association between NCF1-H90 variant and dcSSc (OR = 2.48, P = 2.02E-3), between low CNV (one copy number of functional NCF1) and dcSSc (OR = 21.3, P = 1.71E-8), and between low CNV and lung fibrosis (OR = 4.33, P = 2.60E-2) (Table 1). While high CNVs (two copies of NCF1 and one or two copies of GTGT-containing NCF1B and NCF1C) protected against dcSSc in EA (OR = 0.35, P = 9.78E-5) (Table 1). These findings supported that abnormally low activity of NOX2, due to the hypofunctional NCF1-H90 variant⁶, predisposed to SSc, especially in patient subsets affected with either dcSSc or lung fibrosis.

Table 1.

Association of NCF1 variant with SSc in different ancestry groups

SNP	Ancestry	Subtype	Subject		MAF %		OR	[95%Cl]	P
			Case	Control	Case	Control
rs201802880 (p.Arg90His, A/G)	Chinese	Total	332	2089	28.3	16.8	1.93	[1.60–2.32]	5.97E-12
		lcSSc	108	2089	26.4	16.8	1.79	[1.30–2.45]	3.14E-04
		dcSSc	170	2089	31.8	16.8	2.25	[1.77–2.87]	4.25E-11
		Lung	189	2089	29.9	16.8	2.09	[1.65–2.64]	7.96E-10
	European American	dcSSc	193	931	4.9	2.2	2.48	[1.39–4.42]	2.02E-03
CNV	Ancestry	Subtype	Subject		MAF %		OR	[95%Cl]	P
			Case	Control	Case	Control
1a	European American	dcSSc	14	3	3.1	0.1	21.3	[6.10–74.6]	1.71E-08
3b and 4c			13	184	2.9	8.2	0.35	[0.20–0.63]	9.78E-05
2d			197	931
			Presence	Absence	Presence	Absence
1a		Lung	11	3	5.2	1.3	4.33	[1.19–15.8]	2.60E-02
3b and 4c			6	7	2.9	2.9	1.06	[0.35–3.21]	1.00
2d			88	109

CNV = copy number variation; MAF = minor allele frequency; OR = odds ratio; Cl = confidence interval; dcSSc = diffuse cutaneous systemic sclerosis; lcSSc = limited cutaneous systemic sclerosis; Lung = case with lung fibrosis identified by chest X-ray and/or CT/HRCT. A total of 332 Chinese patients and 197 European American patients were enrolled, and 4 European American patients were excluded after quality control of the genotyping results. 54 Chinese SSc patients were excluded due to incomplete clinical information in subset analysis.

^aOne copy of functional NCF1.

^bTwo copies of functional NCF1 and one copy of GTGT-containing NCF1B or NCF1C.

^cTwo copies of functional NCF1 and two copies of GTGT-containing NCF1B or NCF1C.

^dTwo copies of functional NCF1 (wild type).

Increased pulmonary fibrosis and IFN-I scores in BLM-treated H90.B6 KI mice

To explore the link between NCF1-H90 variant and the development of SSc-associated lung disease, we used an established BLM mouse model^15,16 and delivered graded dozes of BLM (0, 50, 75, 100 U/kg) via osmotic minipumps to H90.B6 KI mice and R90.B6 WT littermates. Given that male B6 mice developed robust fibrosis in response to BLM treatment compared with B6 females^15,17, we used male mice first for a dose-finding experiment to study genetic effects then verified the optimum BLM dose in female mice. Compared to PBS-treated mice, BLM-treated littermates exhibited a dose-dependent increase of collagen deposition in lungs (Masson’s Trichrome-staining) (Figure 1A) and lung fibrosis (Ashcroft score) (Figure 1B left). Given significantly increased Ashcroft scores observed only in 100 U/kg BLM-treated male KI mice, we treated female KI and WT littermates with 100 U/kg BLM and confirmed significantly increased Ashcroft scores in H90.B6 KI female mice compared to their WT littermates (Figure 1B left).

Figure 1. Increased pulmonary fibrosis and type I IFN score in BLM-treated H90.B6 KI mice.

(A) Representative images of no or varying amount of BLM-treated left lung tissues of WT and KI littermates on Day 28 stained with H&E or with Masson’s trichrome. Intense blue Masson’s staining indicated the presence of collagen. Scale bar: 100 μm. (B) Left: quantification of fibrosis of lung tissues using Ashcroft score by a blinded pathologist. Middle and right: expression levels of fibrosis-related genes (Col1a1, Col1a2, Fn and Ctgf) and type I IFN-inducible genes (Irf7, Ifit44, Mx1, Isg15 and Oas1) in lung tissues were measured using real-time PCR and summarized as fibrosis scores and type I IFN scores, respectively. (C) Pearson correlations among Ashcroft score, fibrosis score and type I IFN score. (D) Unsupervised hierarchical clustering of 51 pathways derived from nCounter^® Fibrosis Panel (Nanostring Technologies) visualized as 8 distinct groups based on similar expression patterns. Ashcroft score of each mouse treated with graded dose of BLM were depicted below the heatmap. (E) Ashcroft score analysis among the 8 groups (upper part). Percentage of WT/KI and male/female mice number in groups treated with different BLM dosage or PBS (lower part). IFN, interferon; BLM, bleomycin; KI, knock-in; WT, wildtype; H&E, hematoxylin and eosin; PCR, polymerase chain reaction; PBS, phosphate buffer saline; M, male; F, female. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, P values between WT and KI were determined using two-tailed Student’s t-test, P values among more than two groups were determined using One-Way ANOVA. The correlation was calculated by linear regression test.

We next tested the effect of Ncf1-H90 variant on lung transcriptome in the BLM-induced SSc model. We measured transcript levels of fibrosis-related genes (Col1a1, Col1a2, Fn and Ctgf) and IFN-I-stimulated genes (Irf7, Ifit44, Mx1, Isg15, Oas1), summarized them as fibrosis score and IFN-I score for each mouse lung sample, respectively^18,19. The fibrosis score and its related-gene levels increased dose-dependently in BLM-treated male mice. Both male and female 100U/kg BLM-treated KI mice exhibited elevated fibrosis scores compared to their respective sex-matched WT littermates (Figure 1B middle and Supplementary Figure 1A). Additionally, H90.B6 KI mice had elevated IFN scores in lungs of BLM-treated male (75, 100 U/kg) and female (100 U/kg) mice compared to their WT littermates (Figure 1B right and Supplementary Figure 1B). Positive correlations between Ashcroft scores with either fibrosis scores or IFN-I scores were observed in both male and female mice (Figure 1C). Collectively, these data showed the R to H substitution of Ncf1 enhanced BLM-induced lung fibrosis in both sexes, and the extent of pulmonary fibrosis correlated with upregulated transcripts related to fibrosis and IFN-I pathways. We detected no measurable anti-nuclear antibody (ANA) activity in KI and WT mice treated with 100 U/kg BLM or PBS (Supplementary Figure 2).

To further understand additional pathways underlying lung fibrosis, we performed pathway analysis of 770 fibrosis-related genes (NanoString) in lung tissues from experimental mice. Unsupervised hierarchical clustering of the 51 pathway scores segregated the data into eight groups based on differential gene expression levels (Figure 1D). Group 1 only contained PBS-treated WT and KI mice with the lowest Ashcroft score, while Group 8 contained male mice treated with the highest dose 100 U/kg BLM (60% KI and 20% WT) and with 75 U/kg BLM (20% KI) exhibiting the highest Ashcroft score (Figure 1E). Lung samples from the female mice treated with 100 U/kg BLM were found in Group 3 and 4 (Figure 1E). An overall increase in Ashcroft scores was positively correlated with grouping from 1 to 8, and Ashcroft scores of Groups 3, 4 and 8 were significantly upregulated compared to Group 1 (Figure 1E). The most upregulated profibrotic pathways we observed in the highest Ashcroft score Group 8 had a higher representation of KI mice in BLM-treated lungs compared with a higher representation of WT mice in PBS-treated lungs.

Elevated expression of Spp1 and increased macrophage cell-type score in lung tissues of H90.B6 KI mice

Given that 100 U/kg BLM-treated KI mice of both sexes exhibited elevated Ashcroft scores compared to their similarly treated sex-matched WT littermates, we asked which genes and pathways were differentially expressed (DE) among the 770 fibrosis-related genes and 51 pathways. While transcript levels of 23 genes were significantly increased in the KI group (Figure 2A left and Supplementary Table 2), levels of 74 genes were significantly increased and 7 genes were significantly decreased in male mice (Figure 2A right and Supplementary Table 3). Among DE genes, 5 upregulated genes (Timp1, Ccl2, Il6, Arg1 and Spp1) exhibited both genotype and sex difference. As for DE pathways, a total of 14 pathways exhibited both genotype and sex difference (Figure 2B and Supplementary Figure 3). We further investigated the connection between DE pathways and genes based on NanoString annotation database. Epithelial–mesenchymal transition (EMT), a reported pathway of lung fibrosis, contains the pathway hallmark genes Spp1 and Timp1 in KI mice^20–22. The other three upregulated genes, Ccl2, Il6, and Arg1, exhibiting both sex and genotype effects, were in pathways involved in Cytokine, M2 activation and Neutrophil Degranulation (Figure 2C). Recognizing that ARG1 and SPP1 could be profibrotic markers of macrophages (termed as SPP1⁺ macrophage) in lung fibrosis^23,24, and signaling of CCL2/CCR2 in monocytes/macrophages could promote lung fibrosis²⁵, we assessed relative cell-type scores of lung tissues between KI mice and WT littermates, which showed increased macrophages and decreased B cells from KI group compared with those from WT group in both male and female mice treated with 100 U/kg BLM (Figure 2D and Supplementary Figure 4). Additionally, exhausted CD8⁺ T cells were significantly increased in male KI group (Supplementary Figure 4). The increased macrophages cell-type score in KI was consistent with the DE gene results (Ccl2, Arg1 and Spp1 increased in KI) and DE pathway results (EMT and M2 activation upregulated in KI), suggesting the SPP1⁺ monocytes/macrophages might be pivotal in the development of severe lung fibrosis observed in H90.B6 KI mice.

Figure 2. Increased expression of 5 genes and elevated macrophage cell type score in H90.B6 KI mice lung tissue by Nanostring fibrosis panel.

(A) Differentially expressed genes of 100 U/kg of BLM-treated H90.B6 KI mice compared to their WT littermates (left part) and male compared to female mice (right part). (B) Differential pathway scores between KI/WT mice (genotype difference) and male/female mice (sex difference) exhibiting 12 up-regulated and 2 down-regulated pathways. (C) Unsupervised hierarchical clustering of differentially expressed genes (either have genotype difference or sex difference) in pathways of EMT, Cytokine. Neutrophil degranulation and M2 activation. (D) Relative cell-type scores for macrophages between KI mice and WT littermates. FC, fold change; EMT, epithelial-to-mesenchymal transition; ECM, extracellular matrix; TLR, Toll-like receptor; TGF, transforming growth factor SASP, senescence-associated secretory phenotype; TILs, tissue infiltrating lymphocytes. * P < 0.05, P values between WT and KI littermates were determined using two-tailed Student’s t-test. The differentially expressed pathway was analyzed using linear regression model and Benjamini-Hochberg false discovery rate correction. Results were shown as mean ± SEM.

Increased risk of dcSSc and pulmonary fibrosis in a longitudinal cohort of SSc patients carrying the NCF1-H90 variant

To determine if our findings in H90.B6 KI mice could be confirmed in SSc patients, we assessed the relationship between carriage of NCF1-H90 variant and clinical characteristics of patients with SSc in an independent, longitudinal observation cohort of 92 Chinese SSc patients (Table 2 and Supplementary Table 4). Among them, 47 were homozygous for NCF1-R90 (51.1%), 37 were NCF1-90R/H (40.2%) and 8 were homozygous for NCF1-H90 (8.7%). The minor allele frequency (MAF) of H90 variant was 28.8%, which was consistent with 28.3% depicted in Table 1. The baseline demographics and clinical manifestations at diagnosis are depicted in Table 2.

Table 2.

Patients baseline demographics, treatment and autoantibody

		R90 (n=47)	90R/H (n=37)	H90 (n=8)
Age (yrs)		52.0 (18.0–71.0)	48.0 (18.0–67.0)	50.5 (22.0–70.0)
Gender (F/M)		38/9	32/5	7/1
Disease duration (yrs)		7.4 (0.5–21.0)	7.0 (1.0–21.0)	6.9 (1.0–16.0)
SSc type (dcSSc/lcSSc)		25/22	25/12	7/1
Treatment
MMF (%)		12 (25.5%)	7 (18.9%)	3 (37.5%)
CTX (%)		19 (40.4%)	24 (64.9%)	7 (87.5%)
HCQ (%)		17 (36.2%)	20 (54.1%)	6 (75.0%)
MTX (%)		2 (4.3%)	4 (10.8%)	0 (0.0%)
Pred (%)
ANA titer		44 (93.6%)	36 (97.3%)	8 (100.0%)
	1:100(%)	25 (53.2%)	12 (32.4%)	1 (12.5%)
low	1:320(%)	6 (12.8%)	3 (8.1%)	1 (12.5%)
	1:640(%)	5 (10.6%)	4 (10.8%)	0 (0.0%)
high	1: 1000(%)	10 (21.3%)	11 (29.7%)	3 (37.5%)
	1:3200(%)	1 (2.1%)	7 (18.9%)	3 (37.5%)
ACA+ (%)		18 (38.3%)	6 (16.2%)	0 (0.0%)
ATA+ (%)		7 (14.9%)	13 (35.1%)	6 (75.0%)

yrs, years; F, female; M, male; SSc, systemic sclerosis; dcSSc, diffused cutaneous SSc; lcSSc, limited cutaneous SSc; MMF, mycophenolate mofetil; CTX, cyclophosphamide; HCQ, hydroxychloroquine; MTX, methotrexate; Pred, prednisone; ANA, anti-nuclear antibody; high ANA titer was defined as ≥ 1:1000, low ANA titer < 1:1000. ACA, anti-centromere antibody; ATA, anti-topoisomerase antibody.

No genotype-associated difference in age at diagnosis was observed (Figure 3A) with a female/male ratio of 4.2: 1 for R90, 6.4: 1 for 90R/H and 7.0: 1 for H90 genotype patients (Table 2). The MAF was 34.2% in dcSSc (n= 57) and 20.0% in lcSSc (n = 35). Given the increasing frequency of dcSSc observed in R90, 90R/H and H90 genotype groups, H90 variant carriers were more likely to develop diffuse cutaneous involvement (Figure 3B). HRCT imaging showed increasing frequencies of lung fibrosis in R90, 90R/H and H90 patients (Figure 3C). With respect to autoantibody profiles, the H90 patient group exhibited elevated ANA titers, increased anti-topoisomerase antibody (ATA) seropositivity and decreased anti-centromere antibody (ACA) seropositivity compared to the R90 group (Figure 3D). The extent of skin fibrosis assessed using the modified Rodnan Skin Score (mRSS) was significantly higher in the H90 group than either 90R/H group or R90 group at baseline (Figure 3E), which remained the highest during years of follow-up visits (Figure 3F). Importantly, H90 patients had a much larger GAP index (calculated by gender, age, forced vital capacity and diffusion capacity of lung for carbon monoxide) compared with either 90R/H patients or R90 patients at baseline (Figure 3G). Figure 3H depicts HRCT lung images from a typical patient of each genotype group representing the mean value of GAP index. The 0, 1 or 2 copies of H90 carriage dose-dependently associated with elevated mRSS and GAP index, indicating contribution of NCF1-H90 variant in disease severity and lung damage of SSc patients. While the overall survival rate of these 92 patients was 85.9% (79/92) during an average 5.4-year period, survival rate was 62.5% (5/8) in the H90 group, 83.8% (31/37) in the 90R/H group, and 91.5% (43/47) in the R90 group, respectively, indicating H90 carriage in SSc patients is associated with shorter survival (Figure 3I). Compared to the R90 group, survival rate analysis in different sub-groups showed a significantly decreased in H90 group in patients with high ANA titer and dcSSc, but no difference with H90 group in either ATA positive patients or patients with lung fibrosis (Supplementary Figure 5). Hence, the H90 carriage in SSc patients appeared to promote ATA seropositivity and lung fibrosis, resulting in a shortened survival.

Figure 3. Dose-dependent association of NCF1-H90 variant with increased skin thickness, lung fibrosis and mortality in Chinese SSc patients.

(A-D) The age at diagnosis, SSc subtype (dcSSc or lcSSc), proportions of lung fibrosis identified by HRCT, and autoantibody profiles of consecutive SSc patients enrolled in a longitudinal observational cohort stratified by genotypes. High ANA titer was defined as ≥ 1:1000. (E) A dose-dependent association of NCF1-H90 variant with elevated mRSS of patients at enrollment. (F) Elevated mRSS of NCF1-H90 SSc patients during follow-up visits. The decreased numbers of patients in yearly follow-up visits stratified by genotypes, were due to either mortality or limited duration of follow-up. (G) The presence and the extent of lung fibrosis were evaluated using GAP index (rang from 0 to 8, including sex, age, FVC and DL_CO). A dose-dependent association of NCF1-H90 variant with elevated GAP index of patients at enrollment that the GAP index was recorded as 0 for SSc patients without lung fibrosis. (H) Representative HRCT images including upper lobes, middle lobes, and lower lobes from one patient of each genotype group representing the mean value of GAP index. (I) Kaplan–Meier survival curve of these 92 SSc patients. SSc, systemic sclerosis; NCF1, neutrophil cytosolic factor 1; R90, homozygous for NCF1-R90; 90R/H, heterozygous for NCF1-90R/H; H90, homozygous for NCF1-H90; HRCT, high-resolution computed tomography; ILD, interstitial lung disease; ANA, anti-nuclear antibody; ATA, anti-topoisomerase antibody; ACA, anti-centromere antibody; mRSS, modified Rodnan skin score; GAP, gender-age-lung physiology; FVC, forced vital capacity; DL_CO, diffusion capacity of lung for carbon monoxide; ns, no significance. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, P values among the three groups were determined using Fisher’s exact test for categorical variables or using One-Way ANOVA with Bonferroni’s multiple-comparison test for continuous variables. The association of H90 copy numbers with mRSS or GAP index was calculated by linear regression test. Results were shown as mean ± SEM.

Elevated CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes in PBMCs from NCF1-H90 SSc patients

To further explore the NCF1-H90 variant involvement in SSc disease progression, we comprehensively profiled PBMCs from sex- and age-matched 7 healthy controls (HCs) and 14 SSc patients from the three groups of genotypes collected at their follow-up visits using cytometry by time-of-flight (CyTOF). The demographics and clinical manifestations of these subjects at the time of blood draw are depicted in Supplementary Table 5. The incidence of lung fibrosis was 33.3%, 80.0% and 100.0% for R90, 90R/H and H90 group, respectively. Similarly, H90 group had significantly higher mRSS (38.33 ± 6.66) and GAP index (4.00 ± 1.73) than R90 or 90R/H group.

Based on expression levels of lineage-associated markers, PBMCs were partitioned into 38 cell clusters (Supplementary Table 6) visualized on a t-distributed stochastic neighbor embedding (t-SNE) map (Figure 4A), which were annotated into 10 major immune cell lineages: CD4⁺ T cells, CD8⁺ T cells, DNT cells, DPT cells, NKT cells, monocytes, B cells, NK cells, dendritic cells and stem cells. While the cell-type composition in PBMCs varied substantially among the five groups, the frequency of monocytes was significantly increased in H90 patients compared to either R90 patients or HC-R90 subjects, and exhibited an H90 dose-dependent increase in SSc patients (Figure 4B, Supplementary Figure 6–7). Moreover, cluster 37 (defined as CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes) were significantly increased in the SSc-H90 group compared with the other four groups (Figure 4C), exhibiting an H90 dose-dependent increase in both percentage and absolute number of CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes. The percentage of CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes was positively correlated with mRSS and GAP index, indicating their profibrotic property (Figure 4C). The lack of difference between HC-R90 and HC-90R/H group provided the rationale to combine them in subsequent analyses. Additionally, the frequency of CD8⁺CD3⁺CCR5⁺CD28⁺CXCR3⁺CD31⁺ T cells (cluster 15) and CD69⁺CD34⁺CCR4⁺CXCR3⁺CD31⁺ stem cells (cluster 25) were remarkedly decreased in the PBMCs of SSc patients compared to those from HCs (Supplementary Figure 7).

Figure 4. Dose-dependent effects of NCF1-H90 variant in increased CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes in PBMCs of SSc patients.

(A) Left: tSNE plot of the CyTOF data of CD45⁺ cells from PBMCs of 14 SSc patients and 7 age- and sex-matched HCs. A total of 38 clusters were identified and annotated into a specific immune cell type according to the known marker expression. Right: tSNE plot of all CD45⁺ cells colored by the expression levels of CD45, CD4, CD19, CD68, CD14 and CD56. (B) Frequency of 10 main types of immune cells in PBMCs of SSc-R90 (n=6), SSc-90R/H (n=5), SSc-H90 (n=3), HC-R90 (n=5) and HC-90R/H (n=2). (C) Left: tSNE plot of all CD45⁺ cells from PBMCs of R90, 90R/H, H90 patients and HCs. Middle: Proportions of 38 cell clusters in the five groups. Right: Proportions and numbers of cluster 37 (CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes) associated with H90 copy numbers, mRSS and GAP index. (D) Elevated expression levels of CD68, Tim3, CD86 and CD204 in SSC patients were associated with copy numbers of H90 variant. tSNE, t-distributed stochastic neighbor embedding; CyTOF, cytometry by time-of-flight; HC, healthy control. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, P values among the four groups were determined using One-Way ANOVA with Bonferroni’s multiple-comparison test. The association of H90 copy numbers with the percentage of monocytes, percentage and number of C37 or expression of cell markers was calculated by linear regression test. The correlation was calculated by linear regression test. Results were shown as mean ± SEM.

When we analyzed the expression of 42 markers in each cluster, expression levels of CD68, Tim3, CD86 and CD204 were significantly higher in the H90 patient group than the HC group. Moreover, there was an H90 dose-dependent increase in the expression of CD68, CD86 and CD204 in SSc patients, which were all cell markers for macrophages (Figure 4D). The lack of H90-dose-dependency of Tim3 could be contributed by its expression in clusters 34 and 36 in addition to cluster 37. These CyTOF findings indicated the expanded monocyte/macrophages in PBMCs might contribute to lung fibrosis in H90 patients.

CCL2-dependent elevated levels of osteopontin, CCL2 and ARG1 in monocyte-derived macrophages from H90 patients

To explore functional consequences of the elevated monocyte subset and its underlying mechanism of regulation, we further collected peripheral blood from three H90 patients, six R90 patients and seven HCs to obtain plasma, PBMCs and MoMs. The plasma levels of CCL2, osteopontin (OPN encoded by SPP1), ARG1, TIMP-1 and IL-6 in H90 patients were significantly elevated than in either R90 patients or HCs, which was consistent with previous results that H90.B6 KI mice showed higher expression of Ccl2, Spp1, Arg1, Timp1 and Il6 at the transcript level (Figure 5A). No significant difference of OPN, CCL2, ARG1, TIMP-1 and IL-6 plasma level was found among the three genotype groups in newly recruited 20 HCs (Figure 5A and Supplementary Figure 8A), indicating the increased expression of these 5 genes in H90 group during the disease state.

Figure 5. Monocyte-derived macrophages from PBMCs of NCF1-H90 SSc patients exhibited elevated expression of OPN in a CCL2-dependent manner.

(A) Elevated plasma levels of OPN, CCL2, ARG1, TIMP-1 and IL-6 from H90 patients (n=3) compared to those from either R90 patients (n=6) or HC (n=7) from the subjects enrolled in the CyTOF study. An additional 20 HCs stratified by genotypes exhibited no significant difference in OPN plasma levels. (B) The CD68⁺CD11b⁺ MoMs (lower) differentiated from CD14⁺ monocytes (upper) cultured in DMEM supplemented with M-CSF. (C) Levels of OPN in MoMs from R90 and H90 SSc patients correlated with mRSS and GAP index. (D) Expression levels of CD204, CD86, Tim3, CD163, CD206 and MERTK in OPN⁺ MoMs of R90 and H90 patients. (E-F) The expression level of OPN in MoMs was remarkedly higher in H90 patients than R90 by both flow cytometry and immunofluorescence, which was decreased in MoMs co-cultured with anti-human CCL2 antibody but not with anti-human IgG control. Bar: 100 μm; green, OPN; blue, DAPI. (G) Elevated culture supernatant levels of CCL2, OPN and ARG1 in H90 MoMs were significantly decreased with the addition of anti-human CCL2 antibody. SSc, systemic sclerosis; R90, subjects with NCF1-R90; 90R/H, subjects with NCF1-90R/H; H90, subjects with NCF1-H90; HC, healthy control; ns, no significance; MoMs, monocyte-derived macrophages; OPN, osteopontin. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, P values among three groups were determined using One-Way ANOVA with Bonferroni’s multiple-comparison test. P values between R90 and H90 were determined using two-tailed Student’s t-test, P values among anti-CCL2 group, IgG control group and non-anti-CCL2 group were determined using One-Way Repeated Measures ANOVA. The correlation was calculated by linear regression test. Results were shown as mean ± SEM.

Next, the monocytes in PBMCs were compared between R90 and H90 patients. Similar to the CyTOF results, both the percentage and absolute number of CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes were significantly larger in the H90 group than the R90 group (Supplementary Figure 9A). Moreover, expression levels of CD11b, CD204, CD86 and Tim3 in CD14⁺ monocytes were significantly higher in H90 patients (Supplementary Figure 9B). Further, CD68⁺CD11b⁺ MoMs were cultured from CD14⁺ monocytes isolated from PBMCs (Figure 5B), and the expression of OPN was significantly increased in MoMs from H90 patients than R90 patients, which significantly correlated with mRSS and GAP index (Figure 5C). For the OPN⁺ MoMs, surface expressions of CD204, Tim3, CD163, CD206, but not CD86, were significantly elevated in the H90 group (Figure 5D), and the supernatant levels of active TGFβ1, TIMP-1 and IL-6 were significantly elevated in H90 group (Supplementary Figure 8B). The addition of monoclonal anti-human MCP-1/CCL2 antibody, but not the control antibody, suppressed not only elevated levels of CCL2 but also OPN and ARG1 in the culture system of H90 MoMs compared to those of R90 MoMs (Figure 5E-G), indicating a critical role of the chemokine CCL2 in regulating activation of SPP1⁺ macrophages. However, TGFβ1, TIMP-1 and IL-6 were not suppressed after anti-CCL2 treatment (Supplementary Figure 8B).

Discussion

Here, we showed the hypofunctional H90 NCF1 allele is one of the strongest common causal risk variants for SSc, by using the BLM-treated H90.B6 mouse model for both in vivo and in vitro studies, and by confirming mechanistic findings from mice to patients with SSc (Figure 6). We confirmed the previously reported association between the NCF1-H90 variant and SSc in Japanese²⁶ using Chinese and EA population, and observed more robust association signals in patients affected with either lung fibrosis or dcSSc.

Figure 6. Overview of the experimental workflow.

Our experiments could be summarized into four parts: 1. The NCF1-H90 allele is one of the strongest common risk variants for dcSSc and lung fibrosis in a case-control association study; 2. The increased lung fibrosis in the BLM-treated H90.B6 mouse model confirms Ncf1-H90 is a causal risk variant. Differential expression of BLM-treated lung tissues from R90.B6 vs. H90.B6 and males vs. female mice revealed elevated genes, pathways and infiltrating leukocytes associated with lung fibrosis; 3. The genetic effect of NCF1-H90 in promoting high ANA titers, seropositivity of ATA, seronegativity of ACA, dcSSc, lung fibrosis and high disease activity is confirmed in a longitudinal observation cohort of Chinese SSc patients; 4. CyTOF cell subset analysis of SSc PBMCs identifies a profibrotic monocyte subset that MoM cultures verify the profibrotic SPP1⁺ MoMs contributes to lung fibrosis.

SSc, systemic sclerosis; HC, healthy control; NCF1, neutrophil cytosolic factor 1; dcSSc, diffused cutaneous SSc; lcSSc, limited cutaneous SSc; ATA, anti-topoisomerase antibody; mRSS, modified Rodnan skin score; GAP, gender-age-lung physiology index; KI, knock-in; WT, wildtype; HET, heterozygote; IFN, interferon; CyTOF, cytometry by time-of-flight; MoM, monocyte-derived macrophage.

As lung biopsies are seldom performed in SSc patients, to investigate the relationship between NCF1-H90 and lung fibrosis, we explored murine lung tissues by using a widely accepted BLM-induced fibrosis model by comparing R90.B6 WT to their H90.B6 KI littermates. Several studies have shown that BLM delivery by osmotic minipump in B6 mice had similarity to human SSc-ILD^16,27. The male B6 mice receiving BLM via the pump model, showed a dose-dependent increase in lung fibrosis¹⁵, which was consistent with our findings that BLM resulted in lung fibrosis in a dose-dependent manner, and 100 U/kg was the strongest dosage (Figure 1B). Using the Ncf1-H90 genetically engineered B6 mice, we observed significant enhanced lung fibrosis and IFN-I signature in KI mice compared to WT mice of both sexes. The association of NCF1-H90 variant and IFN-I signature has been reported in patients with autoimmune diseases such as SLE and RA^28,29, as well as in H90.B6 KI mouse model showing lupus-like features induced by pristane⁶. Also, Taroni et al. identified a common pathogenic signature related to an “innate immune-fibrotic axis” that included IFN-I and alternatively activated profibrotic macrophages in multiple SSc tissues including lung, skin and PBMCs³⁰. We showed the genetic effect of Ncf1-H90 variant amplified the impact of the IFN-I signature and the immune-fibrotic axis in lung fibrosis induced by BLM-treatment.

We detected elevated levels of Spp1, Arg1, Ccl2, Timp1 and Il6 in lung tissues from KI mice compared to those from WT mice using NanoString. SPP1/OPN is considered to be a marker of profibrotic macrophages in IPF³¹ and is up-regulated in patients with SSc-ILD^32,33. An important role of SPP1 in SSc-ILD is supported by the discovery of a profibrotic macrophage population marked by expression of SPP1, FN1 and ARG1 (termed SPP1⁺ macrophages) that drives a CXCL4-dependent myofibroblast activation in fibrosis²⁴. Lung fibroblast migration could be promoted by monocytes via OPN secretion, and the process could be amplified by IL-6³⁴. Moreover, the elevated serum OPN and the IL-6 responsive plasma cell signature, could predict lung function deterioration in SSc-ILD patients^34,35. In our study, the plasma levels of OPN and IL-6 were higher in SSc patients than HC, particularly in patients carrying H90 variant, whose lung fibrosis was more severe with higher GAP index. Additionally, Spp1 was one of the top 3 differentially expressed genes in the EMT pathway, which was the most upregulated pathway in KI mice (Figure 2B-C). The transcriptional activation of EMT pathway is shown in SSc-ILD using scRNA-seq³⁶, and the SPP1⁺ macrophages could induce EMT of alveolar epithelial cells, resulting in pulmonary fibrosis³⁷.

The top differentially expressed gene in cytokine pathway, Ccl2 (Figure 2C), was another important marker associated with SSc, especially for patients with lung fibrosis, dcSSc at early stages or ATA⁺ SSc^38,39. High circulating CCL2 levels predict ILD progression and poorer survival in SSc⁴⁰. Conversely, after mycophenolate mofetil treatment, the expression of CCL2 and the numbers of macrophages in skin biopsies decrease⁴¹. In our study, CCL2 plasma level was significantly increased in H90 patients, who had higher mRSS and ATA seropositivity with poor survival. Another differentially expressed gene Arg1 is a profibrotic marker of macrophages, activated by Ly6C⁺ monocytes, which plays roles in lung fibrosis²³.

After anti-CCL2 treatment, the increased secretion of CCL2, OPN and ARG1, but not TIMP-1 and IL-6, in MoMs from H90 patients were suppressed, hence there was a correlation among CCL2, OPN and ARG1, which was partially investigated in previous studies. In synovial fluids of psoriatic arthritis, the transcripts for SPP1 and CCL2 were most highly upregulated by monocytes/macrophages, and both proteins were elevated⁴². In a human monocyte cell line, SPP1 knockdown in macrophages decreased secretion of ARG1⁴³. In our study, we further extended the effect of NCF1-H90 variant in amplifying the CCL2-SPP1-ARG1 axis in monocytes/macrophages in the pathogenesis of lung fibrosis in SSc patients, suggesting that targeting this axis might be efficacious in preventing the progression of lung fibrosis.

Timp1 was the top differentially expressed gene in our KI mice, playing a pivotal role in maintaining ECM homeostasis and has been implicated as a key non-invasive diagnostic and prognostic marker for lung fibrosis⁴⁴. The profibrotic TIMP-1 production by monocytes was upregulated in SSc compared with RA or HC, and was mediated by TLR8⁴⁵. Moreover, elevated serum TIMP-1 in SSc associated with the disease severity⁴⁶. In our study, we also observed plasma and culture supernatant levels of TIMP-1 were significantly elevated in H90 group than R90 group of SSc patients (Figure 5A and Supplementary Figure 8B). While the role of TIMP-1 in SSc has been clearly demonstrated, whether elevated circulating levels of TIMP-1 associated with SLE, lupus nephritis and/or disease activity have yielded inconsistent findings^47–49, which could be attributed to disease heterogeneity and/or small sample sizes of each study. While the H90 risk variant predisposed to both SSc and SLE, TIMP-1 might impact the disease pathogenesis differently.

We observed an increased macrophage cell-type score in lung tissues of H90.B6 KI mice (Figure 2D). Monocytes/macrophages have been implicated in the pathogenesis of SSc, especially in lung involvement^50,51. Multiple immune cell types are affected by H90 variant, including pDCs⁷, macrophages⁶, B cells⁹ and T cells⁵² in SLE, SSc and RA. Given that SSc has a shared genetic basis and IFN-I signature with SLE, and the induced IFN-I signature by H90 variant is found in both the two diseases^6,28, the risk allele might be involved in SSc and SLE both through macrophages by elevated IFN-I signature produced mainly by pDCs and neutrophils. In renal biopsies from patients with lupus nephritis, CD163⁺ M2-like macrophages are the dominant subpopulation, and M2a subpopulations were associated with disease progression⁵³. The current study showed elevated transcripts of 5 genes (Spp1, Ccl2, Arg1, Timp1 and Il6) and macrophage infiltration in the lung tissues of BLM-treated H90 KI mice. The protein products of these 5 genes were elevated in plasma samples of H90 SSc patients. The MoMs from H90 patients also had increased secretion of these 5 gene products, and had increased expression of CD163, CD204 and CD206 (M2 markers) (Figure 5D,G and Supplementary Figure 8B). Taken together, M2-like macrophage infiltration into tissue damage sites appeared to be a common theme in both SLE and SSc, in which the H90 risk variant enhanced the disease manifestations.

Limitations of this study include the following: (1) The principal components analysis could not be performed to correct for population admixture due to our candidate gene approach. However, we established a mice model to confirm the strong association between H90 variant and SSc development. (2) The lack of availability of lung biopsy samples from SSc-ILD precluded studies of genetic effects on lung transcriptome of patients with or without H90 variant carriage. To this end, we explored DE genes in lung tissues using a murine BLM-induced fibrosis model based on H90 KI mice, and the results of mice lung transcriptome was consistent with the CyTOF results of PBMCs from SSc patients. (3) The skin lesions in BLM-treated mice was unable to assess, because the skin fibrosis peaked in the early phase of the model (10 days) while the lung fibrosis was increased in the late phase (28 days)¹⁵. Our primary focus is to assess genotype effects on lung fibrosis, we sacrificed the mice at day 28 post-BLM treatment and observed no significant change in dermal thickness between BLM- and PBS-treated mice. We would assess genetic effects of H90 variant on skin fibrosis in future work. (4) We were unable to isolate sufficient numbers of the newly CyTOF-defined CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes from SSc PBMCs to differentiate into MoMs to establish their direct relationship. However, the dose-dependent increase in the total number of monocytes from the H90 group, who had higher GAP index and mRSS (Figure 4B), and the positive correlation of proportions of newly defined monocyte subset with GAP index and mRSS (Figure 4C) inferred a relationship between the CD14⁺CD68⁺CD11b⁺Tim3⁺ monocytes and the profibrotic SPP1⁺ MoMs cultured from total monocytes isolated from SSc PBMCs.

In summary, the hypofunctional NCF1-H90 variant was associated with susceptibility to SSc, especially for dcSSc with lung fibrosis. H90 patients had an abnormal increased monocyte subset in peripheral blood, which could differentiate into profibrotic SPP1⁺ macrophages in a CCL2-dependent manner. The CCL2-SPP1-ARG1 axis is a potential therapeutic target in SSc, particularly in the dcSSc patients carrying H90 variant.

Key messages.

What is already known about this subject?

The hypofunctional p.Arg90His (p.R90H, rs201802880) substitution encoded in neutrophil cytosolic factor 1 (NCF1), resulting in reduced NOX2 activity and induced type I interferon (IFN-I) signature, strongly associates with risk of multiple autoimmune diseases.

What does this study add?

• The hypofunctional NCF1-H90 variant strongly associated with dcSSc and lung fibrosis in Chinese and European Americans. The H90 variant was dose-dependently associated with increased skin thickness, lung fibrosis and mortality in a longitudinal Chinese cohort.

• The C57BL/6 (B6) mouse with a KI H90 variant in the Ncf1 locus demonstrated severe pulmonary fibrosis and IFN-I signature upon bleomycin treatment. The H90.B6 KI mice exhibited elevated expression of Spp1, Ccl2, Arg1, Timp1 and Il6, along with increased macrophage cell-type score in lung tissues.

• SSc patients carrying two copies of NCF1-H90 variant demonstrated increased CD14⁺CD68⁺CD11b⁺Tim3⁺monocytes/macrophages in PBMCs. The monocyte-derived macrophages from H90 patients exhibited more profibrotic, with higher expression of osteopontin in a CCL2-dependent manner.

How might this impact on clinical practice or future developments?

SSc patients carrying the hypofunctional NCF1 variants have increased risk for active disease course and lung fibrosis, and CCL2-SPP1-ARG1 axis is a potential therapeutic target for these patients.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.