Gain of function NOTCH4 variants disrupt angiogenesis in systemic sclerosis

Abstract

Objectives:

Vasculopathy and fibrosis are central to the pathogenesis of systemic sclerosis (SSc) and their genetic underpinnings are largely unknown. Here, we sought to examine the aetiology of severe vascular phenotypes and poorer outcomes in African American (AA) patients with SSc.

Methods:

The study focuses on AA patients with SSc who have more severe vascular phenotypes and poorer outcomes and combines genetics, single-cell RNA sequencing, functional assays, and a mouse model to explore the role of NOTCH4 in SSc vasculopathy and the potential for NOTCH4-directed therapies.

Results:

Gene-based testing identified NOTCH4 association at an exome-wide significance with SSc (P = 1.6 × 10⁻⁷) and patients with severe vascular disease (P = 3.5 × 10⁻⁷). The risk haplotype defined by the missense (c.2824C>T) and promoter (c.–117G>A) variants was enriched in AAs with SSc (11%) vs controls, and the population attributable risk due to this haplotype in AAs with SSc was 2.6%, which was 52-fold higher than in European Americans. The SSc-associated NOTCH4 variants increased NOTCH4 expression and signalling, leading to decreased angiogenesis and increased endothelial-to-mesenchymal transition (EndoMT). Nailfold capillary abnormalities, decreased angiogenesis, and fibrosis of the vascular lumen are commonly seen in SSc. Genetic, chemical, antibody, or Food and Drug Administration-approved drug inhibition of NOTCH4 signalling rescued angiogenesis and returned EndoMT to baseline.

Conclusions:

NOTCH4 variants are associated with SSc pathogenesis and vasculopathy, partly explaining the increased prevalence of SSc in AAs. The study highlights the need for further research and clinical trials in the inhibition of the NOTCH4 pathway as a strategy to treat the vascular and fibrotic manifestations of SSc.

INTRODUCTION

Systemic sclerosis (SSc, scleroderma) is a rare, multisystem disease that disproportionately affects women and individuals of African ancestry [1]. Significant disparities exist in SSc, with an increased disease burden (earlier age of onset, increased prevalence, and severity) and poorer outcomes (increased morbidity and mortality) among African American (AA) patients with SSc, as compared to European American (EA) patients [2–6]. The reasons for this increased prevalence, severity, and mortality of SSc in AA patients are largely unknown.

SSc is characterised by an autoimmune process, including a unique proliferative vascular disease, and varying degrees of tissue fibrosis [7–9]. In SSc, endothelial cell (EC) dysfunction may contribute to pathogenesis through endothelial-to-mesenchymal transition (EndoMT). This is associated with EC expression of mesenchymal markers such as α-smooth muscle actin (actin alpha 2 [ACTA2]), collagen deposition, and ultimately development of obliterative vascular lesions and tissue fibrosis [10]. Under homeostatic conditions, vascular hypoperfusion and hypoxia are potent triggers for angiogenesis [7,8]. In SSc, despite persistent hypoxia and the associated upregulation of proangiogenic cytokines such as vascular endothelial growth factor (VEGF) and platelet-derived growth factor, compensatory angiogenesis is impaired [10,11].

More than 90% of patients with SSc have a vasculopathy presenting clinically as Raynaud’s phenomenon (RP) associated with nailfold capillary abnormalities and rarefaction [8,12]. Other less common vascular phenomena include pulmonary vascular disease, microvascular disease of the heart, gastric antral vascular ectasias, or scleroderma renal crisis (SRC). Digital ischaemia complicated by fingertip ulcerations is observed in half of patients with SSc and is associated with a poorer quality of life as well as worse disease outcome [13,14]. Vascular involvement in SSc also leads to worsening cardiovascular disease and poor survival [8–10,15,16]. Pulmonary hypertension is the second leading cause of death in SSc [17]. SRC is a life-threatening complication of SSc, characterised by a small arterial vasculopathy leading to uncontrolled hypertension and rapidly progressive renal failure. SRC can be precipitated by glucocorticoids (GCs) [15,18]. Early institution of angiotensin converting enzyme inhibitor (ACEi) therapy can effectively block the hypertensive crisis as well as progression to renal failure associated with SRC [9,15].

NOTCH signalling, especially NOTCH1 and NOTCH4, along with NOTCH ligands, has been reported to play an important role in vascular development [19]. The NOTCH signalling pathway is evolutionarily conserved and is involved in embryogenesis, cell fate, differentiation, and proliferation [19]. NOTCH4 is a member of the NOTCH family of proteins [19] that can bind to 5 NOTCH ligands; its interaction with delta-like ligand 4 (DLL4) on an adjacent cell triggers a series of proteolytic cleavages by the γ-secretase complex and release of active NOTCH intracellular domain (NICD) from the cell membrane (Supplementary Fig S1) [11,19]. NICD translocates into the nucleus and induces transcriptional activation of Hairy/enhancer of split (HES), HES-related proteins (HEY), and DLL4 [11,19]. The NOTCH pathway has been implicated in SSc and in mouse models of SSc, but the functional role of NOTCH4 in SSc is not defined [20–26].

We examined the hypothesis that common genetic variants may identify genes that are enriched for rare, functional variants that could potentially be the cause of the increased disease burden seen in AAs with SSc. We used a combination of exome sequencing (ES) and a gene-based testing approach to discover rare genetic variants that are associated with increased risk for SSc. We then utilised a multipronged approach to study the mechanism by which these variants cause dysregulated angiogenesis, and tested genetic, antibody, and chemical inhibitors of the relevant pathway to rescue normal angiogenesis.

METHODS

Participants

This study included 379 AAs with SSc and 411 AA controls in the discovery cohort and 590 AAs with SSc and 360 AA controls in the replication cohort, enrolled from 24 academic centres participating in the Genome Research in African American Scleroderma Patient consortium [5,27,28]. All SSc subjects met the 1980 American College of Rheumatology (ACR) or the 2013 ACR/European League Against Rheumatism classification criteria for SSc or had at least 3 of the 5 features of CREST (calcinosis, Raynaud’s phenomenon, esophageal dysmotility, sclerodactyly, and telangiectasia) syndrome [29]. Definitions of each SSc disease subset evaluated in the analysis are further outlined in the Supplementary Appendix. Samples from unrelated, AA individuals who tested negative for antinuclear antibodies were included as controls [30]. The study was approved by the local ethics committee at every centre, and all participants provided written informed consent.

Genetic and functional analysis

ES of the discovery cohort was performed, followed by custom capture and sequencing in the replication cohort. Lymphoblastoid cell lines (LCLs) carrying the NOTCH4 variants and EC lines were analysed using real-time polymerase chain reaction, RNA sequencing, DLL4 stimulation, flow cytometry, and ELISA (enzyme-linked immunosorbent assay). To assess angiogenesis, an in vitro tube formation assay was performed on SSc and control primary ECs (pECs) and EC line. Single-cell RNA-sequencing (scRNA–seq) was performed on skin biopsies of 27 SSc and 10 controls. SSc bulk RNA-seq and scRNA–seq, and assay for transposase-accessible chromatin (ATAC)-seq data were downloaded from the Gene Expression Omnibus, and chromatin immunoprecipitation (ChIP)-seq data were downloaded from the encyclopedia of DNA elements (ENCODE). A detailed description of the methods used for these procedures is available in Supplementary Appendix.

Mouse model

We generated a FVB/N background Tie2-tTA;TRE-Notch4 and genetic control: Tie2-tTA mice. Cryosections of the brains were used for immunofluorescence [31]. Further details are provided in Supplementary Appendix.

Statistical analysis

Gene-based testing for the 32 genes associated with SSc at a suggestive P value <10⁻⁶ in the genome-wide association study (GWAS) catalogue was performed using the sequence kernel association test (SKAT, Supplementary Table S1) [32]. Bonferroni’s multiple testing correction yielded a significance threshold of 0.0016 for the discovery and 0.05 for the replication cohort. For the SSc subset analysis, the significance threshold was lowered to 0.00014. For other analyses, a 2-tailed P value of .05 with Bonferroni correction for experiments with multiple comparisons was applied. Additional details are provided in Supplementary Appendix.

RESULTS

NOTCH4 association with SSc

A majority of the AAs with SSc carried a history of RP (95%) with 57.9% of them manifesting digital pitting scars, fingertip ulceration, or gangrene at some point during their disease course (Fig 1A,B, and Supplementary Table S1). A diagnosis of pulmonary arterial hypertension (PAH) was reported in 15.9%, SRC in 5.4%, and interstitial lung disease in 38.5% of subjects. We successfully sequenced 379 AAs with SSc and 411 AA controls and tested the rare functional variants in 32 genes previously reported in SSc (Supplementary Fig S2 and Supplementary Table S2) [33–37] using a gene-based association test. In the discovery cohort, we identified 7 genes with a P value <.05 (Table 1). After multiple testing correction, only the NOTCH4 gene association remained statistically significant (P = 9.9 × 10⁻⁴) (Table 1) and was replicated in an independent cohort (P = .02) as well as in the combined cohorts (P = 1.6 × 10⁻⁷) at an exome-wide significance level (Table 1). NOTCH4 is part of the major histocompatibility complex (MHC) [38,39] and HLA-II genes are the strongest risk factor in SSc.

Figure 1.

Functional role of the NOTCH4 variants. (A,B) Clinical images of a patient with SSc carrying the c.–117A allele (rs8192564) with severe Raynaud’s showing digital pitting, ulceration, and gangrene; (C) the 55 NOTCH4 variants observed in patients with SSc; (D-H) the 5′ UTR variant (c.–117G>A, rs8192564) and (I-M) nsSNV (c.2824C>T, rs17604492); (D) the ATAC-seq signal profile at the NOTCH4 locus in primary endothelial cells from patients with SSc and healthy controls (GSE163199) and the GR-a ChIP-seq in A549 cells; (E) the NOTCH4 gene expression by RT-PCR in the lymphoblastoid cell lines (LCLs) that are either WT or HET for the c.–117G>A variant in the absence of GC or FBS; (F) the NOTCH4 gene expression by RT-PCR in LCLs after addition of FBS and GC; (G) the NOTCH4 ELISA data in unstimulated LCLs; (H) the NOTCH4 expression in SSc skin, stratified by the c.–117G>A variant (GSE130955); (I) the change in glycine to arginine residue by the c.2824C>T variant; (J) the HEY2 gene expression by RT-PCR in unstimulated LCLs with serum that are either WT or HET for the c.2824C>T variant; (K,L) the Flow cytometry data for Hairy/enhancer of split (HES)1 and HEY2, unstimulated and after stimulation with DLL4 ligand, in LCLs that are either WT or HET for the c.2824C>T variant; (M) the HES4 and HES1 expression in SSc skin, stratified by the c.2824C>T variant (GSE130955); (N) the c.2824C>T and the c.–117G>A haplotype frequencies in SSc cases, controls, and reference European (EUR) and African (AFR) ancestral populations. The EUR comprised 136 individuals (51 IBS, 31 FIN, 12 CEU, 9 GBR, and 33 TSI) and the AFR comprised 471 individuals (34 ACB, 1 ASW, 93 ESN, 87 GWD, 77 LWK, 76 MSL, and 103 YRI) from the 1000 Genomes Project; (O) the odds ratios for the risk ‘TA’ haplotype in the different SSc vascular phenotypes as compared to the wild-type ‘CG’ haplotype. For panels (E-H) and (J-M), the upper, centre, and lower lines of the plot indicate 75%, 50%, and 25% quantiles, respectively. P values were calculated by 2-sided Mann-Whitney U test; ns > .05; * ≤ .05; ** ≤ .01; *** ≤ .001; **** ≤ .0001. RT-PCR, ELISA and Flow cytometry assays were run in triplicates, and the average of the 3 runs is shown. LCLs were serum starved for 2 hours before any of the above assays were performed. ACB, African Caribbean in Barbados; A549, lung carcinoma epithelial cell line; ANK, ankyrin; ASW, Americans of African Ancestry in Southwest USA; ATAC-seq, assay for transposase-accessible chromatin using sequencing; ChIP-seq, chromatin immunoprecipitation followed by sequencing; DLL4, delta-like canonical NOTCH ligand 4; EGF, epidermal growth factor; ELISA, enzyme-linked immunosorbent assay; FBS, fetal bovine serum; 5′UTR, 5′ untranslated region; GC, glucocorticoid; G942, glycine at position 942; GR, glucocorticoid receptor; HET, heterozygous; HEY, HES-related proteins; LNR, Lin12-NOTCH repeats; MFI, median fluorescence intensity; ns, not significant; nsSNV, nonsynonymous single-nucleotide variant; OR, odds ratio; PAH, pulmonary arterial hypertension; RT-PCR, reverse transcription polymerase chain reaction; R942, arginine at position 942; SRC, scleroderma renal crisis; SSc, systemic sclerosis; WT, wild-type.

Table 1.

Gene-based association test of functional variants in African Americans with systemic sclerosis

Chr	Gene	Discovery			Replication		Combined
		SKAT P	SKAT PCorr	Variants	SKAT P	Variants	SKAT P	Variants
		SSc = 379/Ctrl = 411			SSc = 590/Ctrl = 360		SSc = 969/Ctrl = 771
6	NOTCH4	9.9 × 10−4	.03	49	.02	59	1.6 × 10−7 a	82
12	SOX5	.007	.22	9
6	HLA-DQB1	.01	.3	33
6	HLA-DPA1	.01	.3	36
6	HLA-DRB1	.01	.3	2
6	ANKS1A	.03	.81	31
6	HLA-DPB1	.03	.81	27
6	PSORS1C1	.05	1	29
6	HLA-DQA1	.05	1	29
6	PRDM1	.05	1	25
6	HLA-DRA	.05	1	8
3	DNASE1L3	.06	1	15
1	LAMC2	.06	1	40
12	ESYT1	.15	1	28
7	GRB10	.22	1	11
7	TNPO3	.27	1	10
7	IRF5	.33	1	4
2	STAT4	.39	1	7
17	IKZF3	.43	1	12
15	CSK	.45	1	10
1	CD247	.45	1	7
6	ATG5	.46	1	4
11	PHRF1	.46	1	15
16	IRF8	.48	1	6
6	PSORS1C2	.49	1	14
5	TNIP1	.59	1	17
2	RHOB	.65	1	10
17	GSDMA	.65	1	17
6	TNFAIP3	.65	1	18
16	ITGAM	.70	1	23
1	TNFSF4	.77	1	2
8	BLK	.79	1	20

Ctrl, Controls; P, P value; P_Corr, corrected P value for 32 comparisons; SKAT, sequence kernel association test; SSc, systemic sclerosis.

Bold statistically significant after multiple testing correction (P < .0016).

^aSignificant at exome-wide P value of 2.5 × 10⁻⁶.

To assess whether the association of NOTCH4 is independent of the HLA class II region, we conducted an exome-wide, single-variant association analysis. As expected, variants within the HLA region showed the strongest associations, with the top signal observed at rs36205178 (P = 1.1 × 10⁻¹³). We then performed a conditional gene-based association analysis adjusting for this top HLA variant (rs36205178), and the NOTCH4 association remained statistically significant (P = 2.2 × 10⁻⁵), confirming its independent association with SSc.

This gene-based NOTCH4 association was based on several variants within the NOTCH4 gene (Fig 1C). These NOTCH4 variants were either novel, rare, or low frequency except for 4 that were common (Fig 1C and Supplementary Fig S3A,B). Eighty-five per cent of these variants were missense substitutions (Supplementary Fig S3A), and almost all of them were predicted to be damaging by in silico analyses (Supplementary Fig S3C). Gene-based tests capture the combined effects of multiple rare and common genetic variants within a gene or genomic region. By evaluating the collective impact of these variants, gene-based testing increases the likelihood of detecting disease associations that may be too rare or have small effect sizes on single-variant analysis. To identify variants with strong effects, we performed, single-variant association analysis of the NOTCH4 gene and identified 2 common variants, c.2824C>T (p.Gly942Arg) and c.–117G>A, with exome-wide significance with SSc (P = 1.6 × 10⁻⁷; odds ratio = 1.95 and P = 1.2 × 10⁻⁶; odds ratio = 1.80, respectively; Supplementary Table S3). Even upon excluding these 2 variants, the gene-based NOTCH4 association with SSc remained statistically significant (P = 3.5 × 10⁻⁴, Supplementary Table S3), suggesting that the NOTCH4 association was not being driven solely by these 2 variants but rather was due to the burden of the variants across the NOTCH4 gene.

NOTCH4 haplotype increases risk in AAs with SSc

The c.2824C>T variant was more common in African ancestry populations than in European populations, and the c.–117G>A variant was slightly more common in European ancestry populations than in AAs (Supplementary Tables S4,S5 and Fig 1N). These 2 variants were independent in European populations (r² = 0.154229) and were in moderate linkage disequilibrium (LD) in African populations (r² = 0.864861). To address the LD structure in AA individuals specifically, we calculated the LD using data from 62 African Caribbean in Barbados and 60 Americans of African Ancestry in Southwest USA individuals, yielding an estimate of r² = 0.804969. Haplotype analysis of the c.2824C>T and c.–117G>A variants yielded a decreased frequency of the ancestral ‘CG’ haplotype (87.1% in SSc vs 92.4% in controls) and an increased frequency of the derived risk haplotype ‘TA’ (11% in SSc vs 6.1% in controls, Fig 1N). To assess their individual and combined effects, we ran a haplotype-based logistic regression adjusting for the top 10 principal components. The combined effect of the ‘TA’ risk haplotype was significant (P = 2.82 × 10⁻⁷), whereas the individual effects were not significant (Supplementary Table S6). The attributable risk of the ‘TA’ NOTCH4 haplotype (excess risk due to the haplotype) in AAs is more than double that of EAs (23.5% vs 10.9%, Supplementary Tables S7,S8) [34,40–42]. Overall, the population attributable risk (PAR) due to the risk ‘TA’ NOTCH4 haplotype was 2.6% in the AA SSc cohort and only 0.05% in the EA SSc cohort. The PAR in AAs with SSc by the NOTCH4 haplotype was 52-fold higher as compared to the EAs with SSc, providing insight into the genetic risk factors underlying the higher prevalence of SSc in AAs as compared to EAs.

NOTCH4 association with severe vasculopathy in SSc

Next, we investigated the association of the NOTCH4 gene with clinical subsets of SSc, given its phenotypic heterogeneity. Patients with SSc with a history of severe RP had the strongest association with NOTCH4 (P = 3.5 × 10⁻⁷, Table 2, Fig 1A,B). This association was absent in patients with mild RP. There was also a statistically significant association with the subset of patients defined by the presence of the African predominant antifibrillarin antibody and diffuse cutaneous involvement (Table 2). Interestingly, no association was seen with 2 other vascular phenotypes of SSc, PAH, and SRC, perhaps due to the small sample size and due to the cross-sectional nature of the clinical database. PAH is more common in patients with SSc who are older and have long-standing disease and needs longitudinal follow-up data. Regression analysis in the vascular subsets of SSc, using the NOTCH4 haplotypes, identified the ‘TA’ NOTCH4 risk haplotype to be enriched in patients with severe vasculopathy as evidenced by the simultaneous presence of severe RP, PAH, and SRC (odds ratio = 10.55; 95% CI 2–56, P = .007; Fig 1O and Supplementary Fig S4).

Table 2.

Gene-based NOTCH4 association with clinical and autoantibody subsets of African Americans with systemic sclerosis

	Combined
	N (%)	SKAT P
Ctrl	771 (100)
Overall SSc	969 (100)	1.6 × 10−7 a
Skin extentb
lcSSc	369 (41.6)	7.8 × 10−4
dcSSc	518 (58.4)	1.5 × 10−5
Organ involvement
PAHc	147 (15.9)	.23
ILDd	324 (38.5)	4.4 × 10−3
SRCc	50 (5.4)	.52
Raynaudse
Mild (with/without vasodilator)	386 (42.1)	.05
Severe (with digital pitting scars, tip ulcerations, or gangrene)	531 (57.9)	3.5 × 10−7 a
Antibodies
ACAf	73 (7.7)	.68
ATAg	252 (26.5)	1.8 × 10−4
AFAh	139 (15.5)	4.0 × 10−6
ARAi	125 (13.6)	.19

ACA, anticentromere antibody; AFA, antifibrillarin antibody; ARA, anti-RNA polymerase III antibody; ATA, antitopoisomerase I antibody; Ctrl, Controls; dcSSc, diffuse cutaneous systemic sclerosis; ILD, interstitial lung disease; lcSSc, limited cutaneous systemic sclerosis; N (%), number of samples and percentage; P, P value; PAH, pulmonary arterial hypertension; SKAT, sequence kernel association test; SRC, scleroderma renal crisis; SSc, systemic sclerosis. Bold statistically significant after multiple testing correction (P < 1.4 × 10⁻⁴).

^aSignificant at exome-wide P value of 2.5 × 10⁻⁶.

^bData available on 887 patients with SSc.

^cData available on 925 patients with SSc.

^dData available on 842 patients with SSc.

^eData available on 917 patients with SSc.

^fData available on 948 patients with SSc.

^gData available on 950 patients with SSc.

^hData available on 894 patients with SSc.

ⁱData available on 918 patients with SSc.

Exploring the functional role of NOTCH4 variants in SSc

To determine the functional role of NOTCH4 variants, we focused on the 2 independently associated NOTCH4 variants identified in the single-variant association analysis, c.–117G>A and c.2824C>T (Fig 1C and Supplementary Table S3). These 2 NOTCH4 variants showed exome-wide significant associations, with effect sizes approaching 2 and relatively high allele frequencies, increasing the likelihood of identifying carriers among patients. Based on these factors, we prioritised these variants for subsequent functional studies. Chromatin accessibility at the site of the c.–117G>A promoter variant was confirmed by ATAC-seq in pECs from both patients with SSc and controls (Fig 1D). Using a transcription factor binding database, we bioinformatically predicted GC receptor (GR)-α to have a higher binding affinity for the risk allele, c.–117A, than the wild-type (WT) allele (Supplementary Fig S5). ChIP-seq data from ENCODE confirmed GR-α binding at the c.–117G>A variant site (Fig 1D). Use of GCs has been shown to independently increase the risk for SRC in patients with SSc [43]. GCs act by binding to the GR, which translocates into the nucleus and binds to the GC response elements in the promoter regions of target genes [44]. Because serum used in tissue culture media contains GCs, we examined NOTCH4 expression in LCLs in serum-free condition and did not observe an increase in expression between the c.–117G>A genotype LCLs (Fig 1E). However, on culturing the LCLs with serum, the c.–117A LCLs had a 2.3-fold higher NOTCH4 expression as compared to the WT LCLs. NOTCH4 expression further increased 7-fold in the c.–117A LCLs on treatment with GCs as compared to c.–117A LCLs without GCs (Fig 1F). The NOTCH4 protein level was also increased by the c.–117A allele (Fig 1G). Analysis of SSc skin expression data confirmed that patients with the c.–117A allele had higher NOTCH4 expression than the samples with the reference allele G (Fig 1H). To determine whether the effect of GCs on increasing NOTCH4 expression was specific to LCLs or extended to other cell types, we examined pECs—specifically, primary human umbilical vein ECs (HUVECs)—following GC treatment. Upon GC stimulation of the HUVECs, we observed a statistically significant increase in NOTCH4 mRNA levels (P < .0001; fold change = 1.8), along with a corresponding increase in NOTCH4 protein levels (P = .04; fold change = 1.4) (Supplementary Fig S6A,B). Notably, this GC-induced upregulation of NOTCH4 in pECs mirrored the effect observed in LCLs.

To assess functional effects of the c.2824T allele (p.942Arg; Fig 1I), we examined HEY2 (a NOTCH4 effector gene) expression in the unstimulated LCLs with serum and observed that the c.2824T LCLs had higher HEY2 transcripts, suggesting increased NOTCH4 signalling (Fig 1J). Next, we stimulated the LCLs with a NOTCH4 ligand DLL4 for 24 hours and assessed changes in NOTCH4 downstream effector genes. Following DLL4 stimulation, LCLs carrying the c.2824T variant had increased HES1 and HEY2 protein expression compared to WT samples, consistent with increased NOTCH4 signalling (Fig 1K,L). Furthermore, HES4 and HES1 transcripts were increased in SSc skin of c.2824T samples as compared to the c.2824C samples, consistent with increased NOTCH4 signalling (Fig 1M).

NOTCH4 expression in SSc skin

On analysing in silico SSc gene expression databases [45–49], we identified an increase in the expression of NOTCH4 and its downstream effector genes in the skin in not only AA individuals but in individuals of non-African ancestries as well (Fig 2A–E). In contrast, NOTCH1, NOTCH2, and NOTCH3 expression was not increased and rather was either the same or decreased in the skin of SSc subjects as compared to controls (Fig 2F). To further explore NOTCH4 expression in SSc skin, we performed scRNA–seq (Supplementary Table S9) [50]. We identified ECs as the major cell type expressing NOTCH4 (Fig 3A,B, and Supplementary Fig S7). Moreover, NOTCH4 was the predominant NOTCH receptor expressed in ECs in the skin (Supplementary Fig S8).

Figure 2.

Expression analysis of NOTCH4 and downstream pathway genes in SSc skin. (A-C) The expression of NOTCH4 gene in control and SSc in skin in 3 published datasets of all ancestries and African ancestry. (D) and (E) HES4 and HEY2 gene expression in control and SSc in skin. (F) The expression of NOTCH1, NOTCH2, and NOTCH3 genes in control and SSc in skin in 3 published datasets of all ancestries. Number of samples for GSE181549- All ancestries SSc=295, control=44; African Americans SSc=44, control=11. Number of samples for GSE130955- All ancestries SSc=58, control=33; African Americans SSc=11, control=6. Number of samples for GSE58095- All ancestries SSc=58, control=39; African Americans SSc=9, Control=11. The upper, centre, and lower lines of the plot indicate 75%, 50%, and 25% quantiles, respectively. P values were calculated by 2-sided Mann-Whitney U test; ns > .05; * ≤ .05; ** ≤ .01; *** ≤ .001; **** ≤ .0001. HES, Hairy/enhancer of split; HEY, HES-related proteins; ns, not significant; SSc, systemic sclerosis.

Figure 3.

Single-cell transcriptomic and pathway analysis of SSc and control skin. (A) The UMAP plot of subclustered skin cells of SSc and control subjects, with colours identifying each subpopulation. (B) The density plot for NOTCH4 gene expression, and the lighter colour indicates increased expression. (C) The UMAP plot of reclustered endothelial cells in SSc and control subjects and (D) the density plot for NOTCH4 gene expression with the lighter colour indicating increased expression. (E,F) The violin plots for NOTCH4 expression in SSc and control subjects. Cell types are colour coded. Tip cells were only seen in patients with SSc. (G) Volcano plot of genes upregulated and downregulated in the NOTCH4^hi expressing cluster as compared to NOTCH4^low nonexpressing endothelial cells. Genes upregulated and part of the angiogenesis pathway are highlighted. Vertical lines indicate a logFC threshold of 0.3, and P values are corrected for multiple comparisons. (H) The top Gene Ontology (GO) pathways for biological processes for the NOTCH4 expressing endothelial cells. (I) Gene expression of NOTCH4 and extracellular matrix-related genes in the NOTCH4^hi endothelial cells cluster. Lighter colour indicates increased expression. (J) The UMAP plot of the arterial, mature capillary, and tip endothelial cell cluster that were the main endothelial cell subpopulations expressing NOTCH4 (NOTCH4^hi). (K) The pseudotime analysis of the arterial, mature capillary, and tip endothelial cell cluster with the trajectory indicated by the black line. Cells lighter in colour are earlier in the progression and are transitioning along the trajectory line to later cell states depicted by darker colour. (L) The trajectory as indicated by the black line, with cells coloured as SSc or controls. ACTA2, actin alpha 2; COL1A1, collagen type I alpha 1 chain; EC_APOE, endothelial cells expressing apolipoprotein E; FC, fold change; NRP1, neuropilin-1; OXPHOS, oxidative phosphorylation ; PcV, postcapillary venule; PDGFB, platelet-derived growth factor subunit B; SSc, systemic sclerosis; TGF, transforming growth factor; UMAP, uniform manifold approximation and projection; VEGFA, vascular endothelial growth factor A; VEGFR2, vascular endothelial growth factor receptor 2.

NOTCH4 expression in ECs and angiogenesis

Subclustering of ECs identified tip, arterial, and capillary cells expressing high levels of NOTCH4 (NOTCH4^hi) (Fig 3C–F and Supplementary Figs S8,S9). The tip ECs spearhead vascular sprouts and are involved in angiogenesis, were predominantly seen in patients with SSc (Fig 3F) and expressed increased levels of DLL4 and KDR (the receptor for VEGF; Supplementary Fig S10). Next, we compared gene expression from the NOTCH4^hi ECs with ECs expressing lower levels of NOTCH4 (NOTCH4^low) and performed differential expression and pathway analyses. NOTCH4^hi ECs expressed higher levels of DLL4, KDR, HEY1, and PDGFB genes (Fig 3G). Angiogenesis and neovascularisation pathways had the highest enrichment ratio; the signaling pathway of vascular endothelial growth factor A (VEGFA) and its primary receptor, VEGFR2 (the VEGFA-VEGFR2 signaling pathway) had the lowest P value (Fig 3H). Thus, angiogenesis emerged as a major pathway associated with NOTCH4^hi ECs. We replicated these findings using a large, published study of scRNA–seq data of SSc skin (Supplementary Fig S11) [51]. The angiotensin II receptor type I pathway was the third most enriched pathway, which was striking given the role of angiotensin II in fibrosis and the use of ACEi drugs in the treatment of patients with SRC (Fig 3H).

NOTCH4 inhibition restores angiogenesis

NOTCH4^hi ECs were enriched for angiogenesis pathway genes; we therefore hypothesised that increased NOTCH4 signalling was driving the dysregulated angiogenesis observed in SSc. To test this hypothesis, we used a Matrigel tube formation assay, a widely used in vitro assay for angiogenesis, in which EC lines form tubules (Fig 4A and Supplementary Movie S1). Treatment of EC lines with recombinant DLL4 stimulated the NOTCH4 pathway, and led to inhibition of angiogenesis, represented by decreased tube length, loss of nodes, junctions, and meshes (Fig 4A,B, Supplementary Fig S12A–C, and Supplementary Movie S2). However, pretreatment of the EC lines with a NOTCH4-specific blocking antibody or DAPT, a γ-secretase inhibitor (GSI), or nirogacestat [52], a small-molecule, selective GSI approved by the Food and Drug Administration for the treatment of Desmoid tumours, followed by recombinant DLL4 stimulation, prevented the loss of nodes, junctions and meshes and decreased tube length induced by DLL4 (Fig 4A,B, Supplementary Fig S12A–C, Supplementary Movies S3–S5).

Figure 4.

Effect of NOTCH4 on angiogenesis and EndoMT assays. (A,B) The decrease in tube formation in the endothelial cell line after NOTCH4 stimulation by its ligand DLL4. Using anti-NOTCH4 blocking antibody, DAPT (a γ-secretase inhibitor) or Nirogacestat before NOTCH4 stimulation by its ligand DLL4, restored the normal tube formation. (C,D) Decrease in tube formation parameters after NOTCH4 stimulation by its ligand DLL4. NOTCH4 knockdown before NOTCH4 stimulation by DLL4 did not affect tube formation. (E,F) Decrease in angiogenic sprouting after NOTCH4 stimulation by its ligand DLL4. Using Nirogacestat before NOTCH4 stimulation by its ligand DLL4 restored the angiogenic sprouting. (G) SSc primary endothelial cells had higher NOTCH4 expression than healthy controls. (H) Tube formation in primary endothelial cells from patients with SSc that were WT for the c.–117G > A variant. (I) Tube formation in primary endothelial cells from patients with SSc that were HET for the c.–117G > A variant. (J) The tube formation using primary endothelial cells from an unaffected control. (K) SSc primary endothelial cells (WT and HET for the c.–117G > A variant) had decreased tube formation than unaffected controls. SSc primary endothelial cells HET for the c.–117G > A variant had decreased tube formation at baseline compared to SSc primary endothelial cells that were WT for the c.–117G > A variant (H,I). (L,M) Blocking NOTCH4 signalling using either an anti-NOTCH4 blocking antibody or Nirogacestat rescued tube formation in the SSc primary endothelial cells. (N) Increase in EndoMT markers, NOTCH3 and SNAI2, in endothelial cell line after NOTCH4 stimulation by its ligand DLL4. Using Nirogacestat before NOTCH4 stimulation by its ligand DLL4 restored the NOTCH3 and SNAI2 levels to the untreated levels. (O,P) Decrease in expression of ACTA2, COL1A1, and SNAI1 after Nirogacestat treatment in primary endothelial cells from patients with SSc that were HET for the c.–117G > A variant. For (B,D,F,G,N), the upper, centre, and lower lines of the plot indicate 75%, 50%, and 25% quantiles, respectively. P values (B,D,F,G,K-M,N) were calculated by a 2-sided Mann-Whitney U test, ns > .05; * ≤ .05; ** ≤ .01; *** ≤ .001; **** ≤ .0001. P values for panel (P) were calculated using 1-sample t-test (2 tailed) ns > .05; * ≤ .05; ** ≤ .01; *** ≤ .001; **** ≤ .0001. ACTA2, actin alpha 2; COL1A1, collagen type I alpha 1 chain; DAPT, γ-secretase inhibitor N-[N-(3, 5-difluorophenacetyl)-l-alanyl]-s-phenylglycinet-butyl ester; DLL4, delta-like canonical NOTCH ligand 4; EndoMT, endothelial-to-mesenchymal transition; HET, heterozygous; HULEC, human lung microvascular endothelial cells; MFI, median fluorescence intensity; ns, not significant; pEC, primary endothelial cells; SNAI1, Snail Family Transcriptional Repressor 1; SNAI2, Snail Family Transcriptional Repressor 2; SSc, systemic sclerosis; WT, wild-type.

Treatment with NOTCH4-specific blocking antibody or DAPT or nirogacestat, in the absence of DLL4, did not alter tube formation, number of nodes, number of junctions, or meshes (Supplementary Fig S13). We were also able to rescue tube formation by silencing NOTCH4 expression using siRNA, thereby preventing the loss of nodes, junctions, and meshes and decreasing tube length after DLL4 stimulation (Fig 4C,D, Supplementary Fig S12D–F, Supplementary Movies S6–S9). We utilised a spheroid-based sprouting assay [53], an alternate assay to study sprouting angiogenesis, and observed the same results as the tube formation assay. Recombinant DLL4 treatment of EC lines inhibited angiogenesis, and pretreatment with nirogacestat prevented the decrease in branch length, number of branches, junctions, and average signal intensity (Fig 4E,F, Supplementary Fig S12G–I). These results suggested that blocking NOTCH4 signalling chemically, genetically, or by an antibody rescued and restored angiogenesis.

Establishing the effect of increased NOTCH4 expression in EC lines, next we investigated the effect of NOTCH4 expression in the pECs derived from patients with SSc (Supplementary Table S10). SSc pECs had increased NOTCH4 expression as compared to unaffected control pECs (Fig 4G). To evaluate whether increased NOTCH4 signalling inhibited angiogenesis and to examine the effect of SSc-associated NOTCH4 variants, we performed a Matrigel tube formation assay on pECs derived from patients with SSc. Patients with SSc were stratified based on the presence of the NOTCH4 promoter variant c.–117A (heterozygous, HET) allele or the c.–117G (wild-type, WT) allele. It is important to note that the patients with SSc who were WT for the NOTCH4 promoter variant still carried one of the several other SSc-associated NOTCH4 coding variants identified in the gene-based analysis (Fig 1C). We observed a decreased number of tubes, nodes, junctions, and meshes in the Matrigel tube formation assay in untreated patient with SSc samples as compared to controls, and these differences were greater in promoter HET SSc pECs (n = 6) as compared to promoter WT SSc pECs (n = 5, Fig 4H–K and Supplementary Fig S14A, Supplementary Movies S10,S11). Treatment of the SSc pECs with either a NOTCH4-specific blocking antibody or nirogacestat increased the number of tubes, nodes, junctions, as well as meshes in both the promoter HET SSc pECs and promoter WT SSc pECs (Fig 4L,M, and Supplementary Fig S14B,C, Supplementary Movies S12–S15). NOTCH4 gene silencing using siRNA led to an increase in the number of tubes, nodes, junctions, and meshes in promoter HET SSc pECs, although these changes did not reach statistical significance (Supplementary Fig S14D). Thus, blocking NOTCH4 signalling rescued tube formation and restored angiogenesis in SSc pECs.

NOTCH4 leads to EndoMT transition

Pathway analysis of NOTCH4^hi ECs further identified profibrotic pathways (TGF-beta signalling and primary focal segmental glomerulosclerosis) and the epithelial to mesenchymal transition pathway, although with a low enrichment ratio (Fig 3H). Next, we examined the NOTCH4^hi ECs for mesenchymal and extracellular matrix (ECM) genes. A subcluster of ECs, while expressing endothelial markers (Supplementary Fig S15) simultaneously, expressed the mesenchymal cell marker (ACTA2), collagen genes (COL1A1, COL3A1, COL4A1, COL6A1), EndoMT transcription factors (Snail Family Transcriptional Repressor 2 [SNAI2], TWIST1), and EndoMT surrogate genes (platelet-derived growth factor receptor beta, NOTCH3; Fig 3I) [54,55]. Applying pseudotime analysis, a computational method utilising single-cell transcriptomics to arrange cells based on their progression through dynamic cellular processes, we observed a linear transition from NOTCH4^hi ECs to NOTCH3^hi ECs (Fig 3J–L) [56]. We conducted a ligand-receptor analysis to further explore the link between these NOTCH4^hi ECs and the NOTCH3^hi EC cluster and revealed DLL4 and JAG1 as ligands on NOTCH4^hi ECs (sender cells) that were interacting with NOTCH3 as a receptor on NOTCH3^hi ECs (receiver cells) (Supplementary Fig S16) [57]. This interaction was further validated by stimulation of EC lines with recombinant DLL4, leading to a time-dependent increase in NOTCH3 expression (Supplementary Fig S17).

Defective angiogenesis has been reported in SSc, and the antiangiogenic SSc pECs have been claimed to undergo EndoMT leading to a myofibroblast-like morphology and phenotype [58].

We had identified SNAI2 and NOTCH3 in a subcluster of NOTCH4^hi ECs from the scRNA–seq analysis of SSc skin (Fig 3I). We stimulated the EC lines with DLL4 to activate the NOTCH4 pathway, and observed a statistically significant increase in NOTCH3 and SNAI2 by flow cytometry. Treatment of these EC lines with nirogacestat reduced the protein expression of NOTCH3 and SNAI2 back to baseline levels (Fig 4N). HET patient with SSc pECs treated with nirogacestat reduced the expression of ACTA2, COL1A1, and SNAI1 genes (Fig 4O,P). These findings support the role of NOTCH4 in EndoMT in SSc and in vitro inhibition of NOTCH4 signalling using nirogacestat reverses EndoMT.

Notch4 upregulation in mice leads to EndoMT transition

To ascertain if upregulation of Notch4 in ECs leads to EndoMT in vivo, we examined the Tie2-tTA;TRE-Notch4* mouse model, in which a constitutively active Notch4 (Notch4*) transgene is expressed specifically in ECs in a temporarily regulatable manner [59–63]. When Notch4* is turned on from birth, by withdrawing tetracycline water from the Tie2-tTA;TRE-Notch4* newborn mice, these mice became moribund and on harvesting Tie2-tTA;TRE-Notch4* brains had developed advanced arteriovenous malformations (AVMs) [60–63]. We performed immunofluorescent staining on frozen brain sections with antibodies to ETS (erythroblast transformation-specific)-related gene (Erg), a nuclear EC marker [64], and cluster of differentiation 31 (CD31), a membrane EC marker to identify ECs, along with Acta2. We focused on the thin-walled blood vessels with a single layer of endothelium without a detectable smooth muscle cell layer (Fig 5A). Representative images from the CD31+Erg+ thin-walled blood vessel regions also expressing Acta2 were observed in 80.7% of the sections from the Tie2-tTA;TRE-Notch4* mutant mice as compared to Tie2-tTA control littermate (Fig 5B–F). These data suggest that Notch4 gain-of-function Tie2-tTA;TRE-Notch4* mice had increased Acta2 expression in CD31+Erg+ ECs and promoted EndoMT.

Figure 5.

Notch4 overexpression mouse model develops arteriovenous malformations and EndoMT. Costaining of Acta2 and CD31 in thin-walled vessels of constitutively Notch4* expressing brain endothelial cells in Tie2-tTA;TRE-Notch4* mutant mice and Tie2-tTA littermate control mice. (A) Representative low-resolution images of mutant Tie2-tTA;TRE-Notch4* mouse brains with DAPI in blue, Acta2 in red, and CD31 in green. (B) Quantification of the percentage of thin-walled vessels containing at least 1 area that costains for Acta2 and CD31. n = 4 Tie2-tTA control mice and n = 6 Tie2-tTA;TRE-Notch4* mutant mice, P = .0009 by Student’s t-test. (C) Thin-walled blood vessels in 1 of 4 Tie2-tTA control mice brain. (D-F) Thin-walled blood vessels in Tie2-tTA;TRE-Notch4* from 3 of 6 mutant mice. Acta2 is shown as red; CD31 is shown as green; Erg is shown as white. Acta2, actin alpha 2; CD31, cluster of differentiation 31 (PECAM1); DAPI, 4′,6-diamidino-2-phenylindole; EndoMT, endothelial-to-mesenchymal transition; Erg, ETS (erythroblast transformation-specific)-related gene; TRE, tetracycline response element.

DISCUSSION

Here, we report the presence of rare, functional variants in the NOTCH4 gene in a large cohort of AAs with SSc. We demonstrate not only that the collective burden of rare, functional variants in NOTCH4 contributes to disparities in SSc frequency and severity in the AA population, but that this enrichment is observed in clinical subsets of SSc with a severe vascular phenotype. This is the first reported association of the c.2824C>T (p.Gly942Arg) and c.–117G>A variants with SSc. Functional interrogation of these specific regulatory and coding NOTCH4 variants highlights their role in activating the NOTCH4 signalling pathway that transcends ancestry. We show that this NOTCH4 activation leads to disruption of angiogenesis and EndoMT in both SSc pECs and a mouse model. Inhibition of this activated NOTCH4 pathway using a blocking antibody, a chemical inhibitor, genetic silencing, or an Food and Drug Administration-approved drug, all led to rescuing of angiogenesis and restoration of EndoMT back to baseline. Cellular and molecular dissection of NOTCH4 signalling has important implications for vascular biology and informs our understanding of SSc vascular disease while suggesting novel therapeutic targets for a disorder with substantial unmet needs.

Although exonic variants (both synonymous and nonsynonymous) of NOTCH4 have previously been associated with SSc susceptibility in early-onset SSc [65] and in adult-onset disease in EA and Han Chinese populations [66], we demonstrate here that rare functional NOTCH4 variants contribute to SSc disease burden in the AA population, and present data implicating NOTCH4 in the development of SSc vascular phenotypes. Herein, we provide what we believe is the first molecular determination of the role of NOTCH4 in disrupting angiogenesis and promoting EndoMT in SSc, and the first evidence supporting NOTCH4-directed treatment strategies. We previously demonstrated that class II MHC genes account for some of the increased SSc risk in the AA population [27], and here we demonstrate that a second gene in the MHC locus contributes to this important health disparity. The NOTCH4 ‘TA’ risk haplotype elucidated in this manuscript increases SSc risk in both EA and AA patients, but the PAR is 52 times higher in AAs compared to EAs, thus providing a clue regarding the increased prevalence of SSc in the AA population. The association of the NOTCH4 gene in both EA and AA individuals, along with its increased expression in SSc skin across both groups, suggests a role for NOTCH4 in SSc pathogenesis that is independent of ancestry.

At the time this study was conceived, only 5 SSc GWASs were listed in the GWAS Catalogue [32], and the 32 genes identified in those studies were included in our analysis. Since then, 10 additional SSc GWASs have been reported [67–76] (Supplementary Table S11), including the largest SSc GWAS to date in individuals of European ancestry [71] and a major study in Japanese cohorts [74]. From these newer studies, we identified 113 additional genes and performed gene-based testing using SKAT to assess their association with SSc. Five genes—STAT1 (P = .02), IL12RB1 (P = .02), HOPX (P = .02), ARL14 (P = .03), and TRPC6 (P = .03)—showed nominal significance in the discovery cohort, but none remained significant after correction for multiple testing (Supplementary Table S12). Further investigations in patients with SSc of diverse ancestral backgrounds are needed to clarify the contribution of functional variants in these SSc-associated genes.

Of the 4 NOTCH genes, we found that only NOTCH4 expression was increased in SSc skin, specifically in ECs. Our data support a model in which the NOTCH4 protein suppresses angiogenesis, thus contributing to vascular compromise in the face of immune-mediated damage. Using several distinct approaches, we were able to inhibit the NOTCH4 signalling and demonstrate the restoration of angiogenesis in EC lines and SSc pECs. In the setting of increased NOTCH4 expression, there is decreased angiogenesis along with increased EndoMT, and the proangiogenic compensatory mechanisms, such as elevated VEGF levels [58,77] that are reported in SSc, may be ineffective in increasing angiogenesis (Fig 6) [78]. The noncoding c.117G>A variant marks a GR binding site that confers exaggerated NOTCH4 expression, while the c.2842C>T variant increases NOTCH4 signalling in cell lines, thus providing an explanation for the lack of a compensatory angiogenic response in patients with SSc. In patients with SSc, long-term, high-dose GCs are generally avoided due to a concern for their potential role in triggering SRC [9,15]. We demonstrate that GCs increase NOTCH4 expression in both LCLs and pECs (Fig 1F and Supplementary Fig S6). Although this effect has been reported in other systems [79], our study highlights that this increase in expression is further amplified in the presence of the SSc risk genotypes (Fig 1F). Our results suggest that the presence of NOTCH4 risk variants should be evaluated as a potential biomarker for identifying patients at high risk for developing SRC.

Figure 6.

NOTCH4 signalling, angiogenesis and EndoMT. VEGF gradient promotes angiogenic sprouting by tip cell formation and stalk migration. An increase in NOTCH4 signalling results in an increase in SNAI2, NOTCH3, PDGFRB, and ACTA2 resulting in EndoMT and inhibition of angiogenesis, whereas inhibition of NOTCH4 signalling results in sprouting angiogenesis and a decrease in EndoMT gene expression. Created with BioRender.com. ACTA2, actin alpha 2; EndoMT, endothelial-to-mesenchymal transition; KDR, Kinase Insert Domain Receptor (VEGFR2); PDGFRB, platelet-derived growth factor receptor beta; sFLT1, soluble Fms-related Receptor Tyrosine Kinase 1 (VEGFR1); SNAI2, Snail Family Transcriptional Repressor 2; VEGF, vascular endothelial growth factor.

Our data also suggest an explanation for the phenomenon of EndoMT and its potential relevance in SSc. Bioinformatic analysis of single-cell transcriptomics indicates that NOTCH4^hi ECs likely transition to NOTCH3^hi ECs and express mesenchymal markers. We further validated the dose-dependent NOTCH4 EC to NOTCH3 EC transition and demonstrated in both EC line and pECs from patients with SSc the increase in mesenchymal markers upon activation of the NOTCH4 signalling. Furthermore, the heterozygous SSc pECs had increased mesenchymal markers and treatment with nirogacestat was able to normalise them. Previous studies have shown the colocalisation of EC marker (CD31) and alpha-smooth muscle actin (α-SMA) in skin sections from patients with SSc [80–82]. Our work suggests that increased NOTCH4 expression leading to EndoMT is the source of increased ECM deposition in the SSc vessel wall, leading to lumen obliteration and tissue hypoxia (Fig 6). This vascular defect in the setting of a dysregulated repair mechanism could lead to tissue fibrosis as evidenced in SSc.

Using a murine model to overexpress Notch4 transgene in the neonatal period, we demonstrated the development of brain AVMs, accompanied by the expression of mesenchymal markers in the ECs of the mutant mice, confirming Notch4-mediated EndoMT. Previously, overexpression of the Notch4 transgene in older mice resulted in AVMs in the skin, liver, and uterus [60–62]. These mice became sick, exhibiting weight loss, respiratory distress, lethargy, and a dishevelled coat [59]. In patients with hypoxic pulmonary hypertension, increased expression of NOTCH4 in the media of pulmonary vasculature and upregulation in lung tissues have been reported [83]. NOTCH4 signalling pathway has been associated with the development of hereditary haemorrhagic telangiectasias that is characterised by the presence of multiple AVMs along with PAH [62,84]. AVMs lack intervening capillaries and result in direct connections between arteries and veins. Telangiectasias are small AVMs that are characteristically found on the hands, face, lips, tongue, buccal and gastrointestinal (GI) mucosa [84,85]. Patients with SSc characteristically have weight loss with vascular involvement manifesting as telangiectasias, PAH, SRC, and vascular involvement in the GI tract [9,77,86,87]. The presence of telangiectasias in SSc, their morphological resemblance to AVMs, and the development of AVMs in the Notch4 transgenic mouse collectively implicates NOTCH4 as a key contributor to the pathogenesis of vascular abnormalities in SSc.

It is noteworthy that angiogenesis plays an important role in the pathophysiology of multiple diseases and that NOTCH4 variants have been associated with susceptibility to several of these conditions. To investigate the functional relevance of these variants, we examined expression changes using the publicly available genotype-tissue expression (GTEx) dataset. In age-related macular degeneration and type I diabetes, progression to blindness is driven by increased angiogenesis in the retina, and the reported risk alleles are associated with decreased NOTCH4 expression (Supplementary Fig S18A,B) [88,89]. Conversely, in lupus and eczema, the risk alleles correlate with increased NOTCH4 expression, similar to observations in SSc, and would be predicted to decrease angiogenesis (Supplementary Fig S18C) [90,91]. Taken together, our results not only implicate NOTCH4 in SSc vasculopathy but also potentially establish a new paradigm in which NOTCH4 signalling plays a central role in regulating angiogenesis in several diseases.

Our in vitro studies suggest that, at least in the case of SSc, NOTCH4 may present a therapeutic target, because nirogacestat, an oral, selective GSI, blocked the NOTCH4-induced defect in angiogenesis. A recent study using nirogacestat showed promising results in the treatment of desmoid tumours, which express high levels of NOTCH1, without displaying any major side effects [52]. Our study illustrates the value of studying diseases in underrepresented minorities not only as a way of understanding human biology more broadly, but as a first step towards personalised medicine for us all.

Supplementary Material

Supplementary materials

Supplementary material associated with this article can be found in the online version at doi:10.1016/j.ard.2025.12.015.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Systemic sclerosis (SSc) prevalence varies based on ancestry. African Americans have an earlier onset and a more severe form of SSc.

• Vasculopathy manifesting clinically in telangiectasias, nailfold capillary abnormalities, and Raynaud’s phenomenon, followed by intimal fibrosis of the vessel lumen, is a hallmark feature of SSc.

• Previous studies have failed to identify the genetic cause of vasculopathy and fibrosis in SSc.

WHAT THIS STUDY ADDS

• NOTCH4 variants are associated with an increased risk of SSc and an increased risk of SSc with severe vasculopathy.

• An SSc-associated NOTCH4 haplotype has a higher population attributable risk in African Americans as compared to European Americans.

• Increased NOTCH4 signalling decreases angiogenesis and leads to endothelial-to-mesenchymal (EndoMT) transition, regardless of ancestry.

• Blocking NOTCH4 signalling rescues angiogenesis and reverses EndoMT.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• It identifies NOTCH4 signalling as a major contributor towards vasculopathy in SSc and other diseases.

• Blocking NOTCH4 may present a therapeutic target in SSc to correct the defect in angiogenesis and luminal fibrosis.

• This study may pave the way for personalised medicine in SSc.

Data availability statement

The sequence data and the corresponding phenotypic information utilised in this study are being submitted to the database of Genotypes and Phenotypes (dbGaP), and the study accession number will be updated as it becomes available.