Genetic and functional analysis of Raynaud’s syndrome implicates loci in vasculature and immunity

Summary

Raynaud’s syndrome is a dysautonomia where exposure to cold causes vasoconstriction and hypoxia, particularly in the extremities. We performed meta-analysis in four cohorts and discovered eight loci (ADRA2A, IRX1, NOS3, ACVR2A, TMEM51, PCDH10-DT, HLA, and RAB6C) where ADRA2A, ACVR2A, NOS3, TMEM51, and IRX1 co-localized with expression quantitative trait loci (eQTLs), particularly in distal arteries. CRISPR gene editing further showed that ADRA2A and NOS3 loci modified gene expression and in situ RNAscope clarified the specificity of ADRA2A in small vessels and IRX1 around small capillaries in the skin. A functional contraction assay in the cold showed lower contraction in ADRA2A-deficient and higher contraction in ADRA2A-overexpressing smooth muscle cells. Overall, our study highlights the power of genome-wide association testing with functional follow-up as a method to understand complex diseases. The results indicate temperature-dependent adrenergic signaling through ADRA2A, effects at the microvasculature by IRX1, endothelial signaling by NOS3, and immune mechanisms by the HLA locus in Raynaud’s syndrome.

Graphical abstract

Highlights

• •GWAS across four cohorts reveals eight loci for Raynaud’s syndrome
• •ADRA2A shows temperature-dependent smooth muscle contraction
• •Endothelial nitric oxide modulates the susceptibility to Raynaud’s syndrome
• •ADRA2A, NOS3, and ACVR2A can be targeted by existing pharmaceuticals

Raynaud’s syndrome (RS) manifests primarily as decreased blood flow in the fingers upon cold or stress exposure. Tervi, Ramste, et al. identified eight genes associated with RS. Follow-up studies validate causal genes important for temperature-dependent adrenergic signaling in microvasculature, endothelial signaling, and immunity, indicating a possible biological mechanism underlying RS.

Introduction

The autonomic nervous system controls physiological functions in the body that are not under direct voluntary control and are not typically consciously directed. Targets of the autonomic nervous system include body temperature, heart rate, respiration, bowel movements and digestion, sexual arousal, endocrine function, blood pressure regulation, and vascular tone. Diseases and problems in regulation by the autonomic nervous system are called dysautonomia, and they can affect many different functions of the autonomic nervous system, including vascular tone and blood pressure, as seen with Raynaud’s syndrome (RS).¹

RS has a clear and specific disease manifestation, where exposure to the cold increases the vascular tone of distal arteries, causing vasoconstriction and leading to cyanosis and hypoxia, particularly in the fingers and toes.² RS can be seen as an example of a disease with a clear component of dysautonomia. Furthermore, RS is a common phenomenon, with an estimated prevalence of 3%–5% in the global population.³ RS rarely causes clinically debilitating symptoms but is diagnosed with a single code in the international classification of diseases (ICD-10 I73.0⁴), making it possible to use electronic health records to find individuals with clinically significant RS and consequently understand RS manifestation, disease correlations, and the underlying biological mechanisms.

The comorbidities of RS include symptoms of pulmonary hypertension in a subset of patients, especially in patients with systemic sclerosis.^5,6,7,8 Consequently, RS can manifest as a comorbidity of diseases with substantial clinical significance, such as systemic sclerosis, lupus erythematosus, myalgic encephalomyelitis/chronic fatigue syndrome, and, most recently, long COVID.^9,10,11 Elucidating the disease mechanisms behind primary RS could provide insight into diseases with dysautonomia.¹²

RS is a common disease, and even though it is usually not life threatening, the symptoms can be very disabling. RS is classified as either primary RS, where there is no other underlying disease, or secondary RS, which manifests as a side effect of immune, cardiovascular, or connective tissue disorders.^2,3,12 Consequently, diagnosing secondary RS sometimes leads to finding the underlying systemic disease. RS has a relatively high hereditary component with estimated twin heritability between 54% and 65%,^13,14 suggesting that genetic studies of RS may reveal disease mechanisms, providing knowledge about biological risk factors, relationships with comorbid diseases, and information about possible drug targets for the development of future therapeutics. A recent single-cohort study in the UK Biobank (UKB) identified genetic association at ADRA2A, IRX1, and MICB loci,¹⁵ but ultimately, genetic studies across multiple different cohorts, healthcare systems, countries, and patient populations provide additional value to find robust and replicating signals. This study examines genetic associations of RS across multiple cohorts and with functional validation in cellular models.

Results

Eight loci associate with RS

Using genetic and electronic health record data from FinnGen data freeze 10 (R10), the UKB, the Estonian Biobank, and the Mass General Brigham (MGB) Biobank, we identified a total of 11,358 individuals with a diagnosis of RS and 1,106,871 controls. Demographic characteristics of the study cohorts showed that the majority of patients with RS were female (73.2%), in agreement with earlier reports,^3,16 and the mean disease diagnosis age was 49.6 years (Table S1).

Genome-wide association study (GWAS) of RS identified eight loci, which included a notable association at the α_2A-adrenergic receptor (ADRA2A) locus, which was seen independently at a genome-wide significant level (p ≤ 5 × 10⁻⁸) in all four cohorts and was further supported in the meta-analysis and ATAC data in pulmonary arteriolar smooth muscle cells (PaSMCs) (rs7090046, p meta-analysis = 3.93 × 10⁻⁴⁷, beta [SE] = 0.22 [0.02]; Figure 1; Table 1; Figures S1–S6 and S7A; Tables S2–S4). The same variant has also been associated with RS in the UKB.¹⁵

Figure 1.

Meta-analysis

(A) A Manhattan plot of RS meta-analysis (11,358 cases and 1,106,871 controls) combining UKB, FinnGen, MGB Biobank, and Estonian Biobank (EstBB) (b38).

(B and C) Locus zoom plots for (B) regional association at the ADRA2A locus (b38), lead variant rs7090046, and (C) regional association at the NOS3 locus (b38), lead variant rs3918226.

Table 1.

RS meta-analysis lead variants combining the UKB, FinnGen, MGB Biobank, and EstBB data (11,358 cases and 1,106,871 controls)

SNP	CHR	POS (b38)	Reference	Alt.	Combined Alt. AF	p value	Effect estimate (beta)	SE	Nearest gene
rs7090046	10	111,101,172	G	A	0.259	3.93 × 10−47	0.209	0.015	ADRA2A
rs7706161	5	4,047,779	A	G	0.261	4.02 × 10−26	−0.155	0.015	IRX1
rs3130968	6	31,097,294	C	T	0.131	7.48 × 10−16	0.157	0.019	C6orf15/HLA
rs3918226	7	150,993,088	C	T	0.073	1.37 × 10−9	0.148	0.024	NOS3
rs7559925	2	147,883,361	T	C	0.315	2.88 × 10−8	−0.079	0.014	ACVR2A
rs191137443	4	132,501,888	C	T	0.006	2.99 × 10−8	0.481	0.087	PCDH10-DT
rs2092504	1	15,190,856	C	T	0.423	3.08 × 10−8	−0.076	0.014	TMEM51
rs7601792	2	129,324,147	G	A	0.004	3.59 × 10−8	0.590	0.107	RAB6C

Base-pair positions reflect genome build 38. Alternative allele (Alt.) is the effect allele for the effect size. p values were calculated using METAL¹⁷ software.

To investigate the effect of the ADRA2A lead variant (rs7090046) on diagnosis age, we performed computational follow-up analyses using electronic health record data and examined if individuals carrying the risk allele for ADRA2A had an earlier disease onset. To do this, we conducted a survival analysis (age at the follow-up time and adjusted by sex, FinnGen cohort information, and first 10 genetic principal components) in FinnGen, which showed a higher incidence of RS for risk allele carriers over time (AA: hazard ratio [HR] [confidence interval (CI) 95%] = 1.44 [1.21–1.70], p = 2.54 × 10⁻⁵ and AG/GA: HR [CI 95%] = 1.25 [1.14–1.37], p = 1.38 × 10⁻⁶) (Figure S8A). Moreover, as RS is triggered by the cold, we examined if areas with longer winters and cold climates might have different allele frequencies (AFs) of the ADRA2A lead variant. Using data on the rs7090046 genotype and municipality of birth for 47,950 Finnish individuals from THL biobank, we compared the AF of the RS risk allele A between different municipalities of Finland. We identified that the AF was higher in the northern parts of Finland compared to the southern parts of Finland, showing a change in AF with latitude (north of Finland AF%: 27.3%–30.6%; south of Finland AF%: 18.2%–22.9%) (Figure S8B). The risk AF was also higher in the UKB in comparison to the other three other cohorts (UKB AF%: 31.0%, other cohorts AF%: 23.4%–24.7%) (Table S5).

In our meta-analysis, an additional seven loci were genome-wide significant (p ≤ 5 × 10⁻⁸, Table 1). The variant rs7706161 is an intergenic variant closest to and downstream of the IRX1 (Iroquois homeobox 1) gene. IRX1 encodes for a homeobox gene involved in finger development in model organisms.¹⁸ The lead variant, rs3130968 in chromosome 6, is located in the human leukocyte antigen (HLA) region downstream of HCG22 (HLA complex group 22) and upstream of the HLA-C and HLA-B genes. This variant has previously been associated with peripheral vascular disease.¹⁹ A lead variant in chromosome 7, rs3918226, is an intronic variant for a NOS3 (nitric oxide synthase 3/endothelial nitric oxide) 5′ regulatory region. Endothelial nitric oxide signaling (eNOS) is a canonical mechanism in vasoconstriction and dilation.^20,21,22 The NOS3 variant rs3918226 has been previously associated with coronary artery disease and hypertension, which can both manifest in individuals with RS,^23,24 and is proximate to a region of open chromatin in telomerase-immortalized human aortic endothelial cells (TeloHAECs) (Figure S7B).

Lead variant rs7601684 in chromosome 2 is located upstream of RAB6C, a member of the RAS oncogene family. The other chromosome 2 lead variant, rs7559925, is an intronic variant for the ACVR2A gene (activin A receptor type 2A) that encodes for a receptor related to activins, which are part of the transforming growth factor beta family of proteins. The association with ACVR2A rs7559925 has been previously associated with immunological traits such as lymphocyte counts.²⁵ In addition, rs2092504 in chromosome 1, an intronic variant in transmembrane protein 51 (TMEM51), and rs191137443 in chromosome 4, upstream of the protocadherin 10 gene (PCDH10), were also genome-wide significant in our meta-analysis. To elucidate possible causal variants at the loci, we performed fine-mapping. The IRX1 locus had a credible set of 19 variants, of which rs7706161, rs72731435, and rs10512704 were the most likely causal variants (posterior inclusion probabilities (PIPs) of 0.13, 0.11, and 0.11, respectively) (Table S6A). The ADRA2A locus credible set included four variants, of which rs7090046 is shown to be the most likely causal variant, followed by rs1343449 (PIPs = 0.50 and 0.29, respectively) (Table S6B). We did not observe consistent fine-mapping results for the other loci.

Our data comprise RS from different countries, healthcare settings, and diagnoses from both primary and specialized care. To understand the robustness of the associations across these different settings, we compared the effect estimate of these genome-wide significant variants within the self-reported symptoms, primary-care-derived diagnoses, and diagnoses from specialized care in the UKB and FinnGen. The results show remarkably consistent effect estimates for each variant regardless of the diagnosis or self-reported source (Figure S9).

RS can manifest as the only symptom (primary RS) or can co-occur with another comorbid or even causal disease (secondary RS). To examine if our results were driven by primary RS or secondary RS, we conducted a sensitivity analysis GWAS of only primary RS cases in the UKB and FinnGen cohorts (3,568 and 1,418 cases, respectively). We used similar criteria for defining cases of primary RS to Hartmann et al.¹⁵ and as recommended by Wigley et al.,²⁶ removing individuals from cases who had a comorbidity diagnosis before the first diagnosis of RS (Table S7A). In the UKB, we replicated the previous findings for primary RS¹⁵ where ADRA2A and IRX1 loci remained significant and with similar effect sizes (rs7090046: p = 6.41 × 10⁻²¹, beta [SE] = 0.23 [0.02] and rs7706161: p = 3.33 × 10⁻¹⁶, beta [SE] = 0.20 [0.02], respectively; Figure S3). In FinnGen, the ADRA2A locus did not remain genome-wide significant but was still among the most strongly associated variants, with a similar effect size to the full sample (Figure S3, rs7090046: p = 2.66 × 10⁻⁷, beta [SE] = 0.22 [0.04]). Subsequently, we conducted a meta-analysis of primary RS from these two cohorts where ADRA2A, HLA, and IRX1 loci remained genome-wide significant (Figure S4; Tables S8 and S9; HLA locus: rs28752908, p = 1.59 × 10⁻⁸). We also performed a separate GWAS for only secondary RS cases in the UKB and FG (1,594 and 666 cases, respectively). Interestingly, no genome-wide significant hits were found for secondary RS in either the individual cohort GWAS or the meta-analysis (Figures S5 and S6; Tables S9 and S10).

The variants discovered in this GWAS are related to vascular tone and blood pressure (ADRA2A, NOS3, ACVR2A) or immune function (HLA) and even have approved pharmaceuticals that target these gene products or their ligands to affect vascular constriction. These include α-blockers that target the α_2A-adrenergic receptor protein encoded by ADRA2A, nitric oxide synthase inhibitors, fusion protein sotatercept²⁷ that traps activins, and growth differentiation factors. All these pharmaceuticals have a downstream effect on lowering vascular tone and, consequently, local or systemic blood pressure.^{28,29,30,31,32,33}

RS risk variants modulate gene expression in vasculature and connective tissue

To understand the functional consequences of the associating GWAS loci, we examined gene expression across human tissues and tissue specificity. Using data from GTEx (https://gtexportal.org/home/) and formal co-localization across all loci (Table S11), we observed that the lead variant affected the expression of ADRA2A in a tissue-specific manner in the tibial arteries (rs7090046, expression quantitative trait locus [eQTL] p = 1.3 × 10⁻¹³, Figure 2) in contrast to coronary arteries (p = 0.16) or aorta (p = 0.65). In addition, the lead variant for ADRA2A expression in GTEx in tibial arteries was rs1343449 (r² in Europeans with rs7090046 = 0.98), which was also the lead variant in the MGB Biobank and among the five top variants in each cohort and the credible set (Figure 2; Tables S2 and S6B). A formal co-localization analysis suggested a shared signal between RS and ADRA2A expression in tibial arteries specifically (posterior probability = 0.99, Figure 2; Table S11). The risk allele associated with higher RS risk associated with higher ADRA2A expression, in agreement with the reported role of increased adrenergic receptor signaling regulating vascular wall contraction.³⁴ Furthermore, the lead variant, rs7090046, affected the expression of another gene at the same ADRA2A locus (BBIP1) in a tissue-specific manner in the hippocampus, as shown previously¹⁵ (Figure S10).

Figure 2.

ADRA2A expression

(A and B) Genotype tissue expression for SNP rs7090046 in (A) tibial arteries and (B) across tissues from GTEx (NES, normalized expression values; m-value = posterior probability, p ≤ 0.05).

(C) The RS association co-localizes with eQTL signal in tibial arteries.

Created with Biorender.com.

In addition, co-localization indicated NOS3, IRX1, TMEM51, and ACVR2A as putative causal genes for the respective loci across multiple tissues (posterior probability > 0.85, Figure S10; Table S11), Notably, NOS3 (rs3918226) was expressed in the lung tissue and fibroblasts and IRX1 (rs7706161) was expressed in skeletal muscle tissue in GTEx (Figure S10; Table S11). As the tissues contain several different cell types, we performed stratified linkage disequilibrium score regression (LDSC) using ENCODE³⁵ to elucidate the overall relevance of different tissue types across the body in RS. We found the most significant enrichment in RS with SMC types, further suggesting a possible pathology mediated by SMCs (Table S12).

Disentangling genetic and phenotypic associations of primary and secondary RS

To understand RS as a phenotype and further untangle the genetic signal, we performed correlation analysis testing the phenotypic and genotypic associations between RS and known comorbidities. First, we performed multivariable logistic regressions (adjusted with age, sex, FinnGen cohort information, and the first 10 genetic principal components) in FinnGen to see if the same traits associate with RS as the ones in the UKB.¹⁵ RS was associated with multiple known autoimmune diseases, for example, systemic sclerosis and lupus, and with migraine, carpal tunnel syndrome, and purchase of beta blockers (Table S7B). Additionally, RS was inversely associated with type 1 and type 2 diabetes, as reported previously in the UKB.¹⁵ RS did not have an association with frostbite, cryoglobulinemia, or drug-induced autoimmune hemolytic anemia (Table S7B). As a sensitivity analysis, we repeated the multivariable logistic regression with primary RS in the FinnGen cohort. Systemic sclerosis, lupus, Sjögren’s syndrome, atherosclerosis, exposure to vibration, and autoimmune hemolytic anemia remained positively associated with primary RS (p < 0.05, Table S7C). Interestingly, both type 2 diabetes and purchase of beta blockers inversely associated with primary RS (Table S7C).

Second, we estimated the genetic correlation between known comorbidities and RS. First, we examined if our genome-wide significant variants replicated in known publicly available comorbidity GWASs. The HLA variant, rs3130968, reached genome-wide significance in lupus (FinnGen), Sjögren’s syndrome (UKB and FinnGen), atherosclerosis (FinnGen), carpal tunnel syndrome (FinnGen), hypothyroidism (FinnGen), type 1 diabetes (FinnGen), and polymyositis (FinnGen) (Table S7D). Our lead variant, rs7090046, at the ADRA2A locus reached statistical significance (p < 0.05) in lupus (UKB), carpal tunnel syndrome (FinnGen), dermatomyositis (FinnGen), and type 2 diabetes (FinnGen) (Table S7D) but did not reach genome-wide significance. Other genome-wide significant lead variants from our meta-analysis did not reach genome-wide significance in other comorbidity GWASs. Next, for genetic correlation analysis, we used the LDSC method,³⁶ which showed a positive genetic correlation between RS and hypothyroidism, migraine, Sjögren’s syndrome, and systemic sclerosis (p < 0.05, Table S7E). For primary RS, migraine and Sjögren’s syndrome were the only ones that remained positively associated (p < 0.05, Table S7F). Type 2 diabetes had an inverse genetic correlation with both RS’s wide definition and primary RS, replicating similar previous findings¹⁵ (p = 0.0003 and r² = −0.138 and p = 0.0002 and r² = −0.191, respectively).

ADRA2A and IRX1 are specifically expressed in microvasculature

Human arteries contain primarily fibroblasts, SMCs, pericytes (defined as the endothelial-like cells found around the microvasculature), and endothelial cells. While ADRA2A is expressed in various tissues, with likely pleiotropic roles, the disease-relevant and most significant eQTLs were observed in tibial arteries (Figure 2). As the ADRA2A eQTL analysis suggests that the lead variant affects expression chiefly in blood vessels in particular, we wanted to examine which cell types express the key targets, especially ADRA2A in the vascular wall. As RS is characterized by insufficient arterial blood flow in specific skin areas, such as the fingers and toes, we utilized available single-cell RNA sequencing (scRNA-seq) data from human skin samples and human arteries. We observed that ADRA2A was expressed in only a small cluster of SMCs, IRX in fibroblasts and NOS3 in endothelial cells (Figures 3B and S11).

Figure 3.

ADRA2A expression is restricted to SMCs located in distal arterioles

(A–D) Uniform manifold approximation and projection (UMAP) plot of (A) human skin and (B) human coronary artery scRNA-seq indicating ADRA2A expression, co-expression of NOTCH3 in the same cluster, and MYH11 expression as SMC cluster and all clusters in skin (C) and in the coronary artery dataset (D).

(E) A schematic of vascular wall and ADRA2A expression.

(F) RNAscope for ADRA2A in a dorsal hand biopsy.

(G) RT-PCR quantification of ADRA2A and NOTCH expression in human arterial SMCs and pulmonary arteriolar.

Mean ± SEM, ∗∗∗p < 0.0005, ∗∗p < 0.005, and ∗p < 0.05 (unpaired two-tailed Student’s t test).

Furthermore, the scRNA-seq data derived from skin tissue identified that ADRA2A is expressed by a subcluster of the MYH11-expressing SMCs that also express the microvascular marker NOTCH3, consistent with ADRA2A expression in the microvascular component of the SMC population (Figures 3A and 3C).^37,38 We further investigated the expression of ADRA2A in eight large arteries by examining an available composite dataset representing combined scRNA-seq data from these vessels. Again, ADRA2A was expressed in a small subcluster of MYH11-expressing SMCs that were NOTCH3 positive, consistent with ADRA2A expression in the adventitial microvascular component of the SMCs (Figures 3B and 3D). These data suggest that among different types of blood vessels, only microvascular cells express ADRA2A and are the likely causal cell population for RS. These findings correspond with earlier reports, showing that α_2A-adrenergic receptor can be detected by western blot only in arterioles.³¹

We then assessed the expression of ADRA2A in SMCs from multiple arterial tissues. ADRA2A was expressed in only one line. This line was derived from pulmonary microvasculature. The pulmonary arteriolar cells also express other markers of the microvascular SMC population identified by scRNA-seq, such as NOTCH3 (Figure 3G). Thus, this microvascular SMC line was chosen for subsequent functional studies.

In addition, we performed RNAscope in situ hybridization in both skin and coronaries, which provided evidence that ADRA2A is expressed in the SMCs of the arterioles, predominantly in deep layers of the epidermis in the skin (Figures 3 and S12).

We also wanted to understand how the other loci identified in our GWAS contribute to RS. Expression of IRX1, a transcription factor expressed in adipose, breast, kidney, and skin tissue, as identified in GTEx, was identified by RNAscope to be located in fibroblasts around the small arterioles and capillaries in the dermis of the skin (Figures S11 and S12). These data support the causality of IRX1 with the RS phenotype, as fibroblasts are also known to possess contractile capabilities.³⁹

CRISPR validation for ADRA2A and NOS3 causality

We then used the previously characterized primary PaSMCs to study ADRA2A gene causality in RS. To elucidate the mechanistic importance of the ADRA2A rs7090046 locus, we designed three CRISPR guides targeting the lead variant rs7090046 and used the CRISPRi-Cas9 machinery to interfere with signaling from this variant region. We observed a significant decrease in ADRA2A gene expression 5 days after lentiviral treatment on the PaSMCs suggesting that the region is needed for controlling ADRA2A expression (Figure 4).

Figure 4.

CRISPRi against rs7090046 supports causal role of ADRA2A

(A) Schematic of the CRISPRi experiment.

(B) RT-PCR quantification of ADRA2A expression in rs7090046 guide targeted cells vs. CTRL.

Mean ± SEM, ∗∗∗p < 0.0005, ∗∗p < 0.005, and ∗p < 0.05 (one-way ANOVA).

Also, one of the lead variants, rs3918226, was proximal to NOS3, an endothelially expressed gene encoding for nitric oxide, whose role is well characterized for endothelial cell-mediated vascular dilatation.⁴⁰ We performed CRISPRi, targeting guides to the region around the rs3918226 variant in TeloHAECs, and observed a ∼79% reduction in NOS3 expression, an effect similar to CRISPRi guides targeting the transcription start site of the NOS3 gene, which is ∼300 bp away (Figure S13). Previous studies have identified a binding site for an ETS family transcription factor, such as ELF1, immediately adjacent to the rs3918226 variant, with homozygosity of the minor T allele being associated with reduced NOS3 transcription and a higher risk of hypertension.^23,24

ADRA2A expression affects SMC contraction in temperature-dependent fashion

Finally, we examined how ADRA2A expression might alter SMC contractility in conditions mimicking RS and environmental cold stress. Earlier studies in temperature-dependent vascular contraction have supported the role of the adrenergic system but have nearly always focused on the role of ADRA2C and its temperature-dependent activation.^31,41,42 These earlier studies and our findings raise an interesting question: does ADRA2A directly affect vascular contraction? To test this, we used a collagen-based SMC contraction assay in ADRA2A-overexpressing or -silenced cells. We discovered that in cold conditions (+28°C), silenced small interfering RNA (siRNA)-ADRA2A-treated PaSMCs contracted significantly less (Figure 5A). Interestingly, ADRA2C silencing did not attenuate cold-induced contraction in a similar manner to ADRA2A (Figure 5A). In warm conditions, both ADRA2A and ADRA2C silenced SMCs contracted similarly (Figure 5B). Consequently, upon lentiviral overexpression of ADRA2A, we observed that the PaSMCs contracted significantly more (Figures 5C and 5D), suggesting that ADRA2A was affecting contraction in a dose-dependent fashion and responsible for cold-induced contraction.

Figure 5.

ADRA2A expression affects SMC contraction upon cold stimulus (+28°C)

(A) ADRA2A and ADRA2C silencing in cold exposure.

(B) ADRA2A and ADRA2C silencing in ambient conditions.

(C) ADRA2A and ADRA2C overexpression in cold exposure.

(D) ADRA2A and ADRA2C overexpression in ambient conditions.

Mean ± SEM, ∗∗∗p < 0.0005, ∗∗p < 0.005, and ∗p < 0.05 (unpaired two-tailed Student’s t test).

Discussion

We performed a meta-analysis of RS across four cohorts and identified genome-wide significant associations with RS at ADRA2A, HLA, NOS3, RAB6C, ACVR2A, PCDH10, TMEM51, and IRX1 loci. We examined the expression of genes at these loci in human tissues, including localization in the skin, and showed that, in particular, ADRA2A, IRX1, and NOS3 contribute to susceptibility to RS in a cell-type-specific fashion. Overall, the associations sketch a pathway where endothelial signaling through NOS3, perivascular expression of IRX1, and smooth muscle contraction by ADRA2A modulate RS susceptibility. Our findings indicate ADRA2A, HLA, and IRX1 in primary RS, overall clarifying the mechanisms that contribute to primary or secondary RS susceptibility.

The most prominent genetic association with RS was discovered in the ADRA2A locus and aligns with a previous study in the UKB.¹⁵ We showed this association to be independently genome-wide significant in all four cohorts, demonstrating the strength of evidence for the role of ADRA2A in RS. In silico follow-up analysis of ADRA2A RNA expression across tissues showed the highest expression in tibial arteries, and single-cell expression analysis from the skin further supported the role of SMCs as the key cell type for ADRA2A expression. Finally, a functional contraction assay in SMCs in cold conditions showed lower contraction in ADRA2A-deficient and higher contraction in ADRA2A-overexpressing SMCs. Overall, our findings indicate ADRA2A in RS and as a regulator of vascular contraction in SMCs in a temperature-dependent fashion.

The ADRA2A lead variants are located in a regulatory region that affected ADRA2A expression in distal arteries. In addition, we identified a subpopulation of SMCs that expresses the α_2A-adrenergic receptor. While the adrenergic system has been suggested as a potential pathological mechanism underlying RS,^{28,29,31,32,41,42,43,44,45,46,47,48,49} earlier functional studies have focused on the α_2C-adrenergic receptor. In this study, we assessed contraction SMCs after ADRA2A or ADRA2C siRNA knockdown in human SMCs. We saw a robust contraction upon cold exposure with ADRA2A knockdown, whereas ADRA2C silencing did not attenuate cold-induced contraction in a comparable manner to ADRA2A. Overall, these findings suggest an independent role of ADRA2A in vascular contraction and temperature-dependent control of vascular tone.

In cold or stress conditions, norepinephrine and epinephrine are released and bind to adrenergic receptors throughout the body, which exerts various effects: dilating pupils and bronchioles, increase in heart rate, and blood vessel constriction. Blood vessel constriction is mediated by the SMCs that express adrenergic receptors on their surface. In healthy patients without RS, there are mechanisms to prevent unwanted and excessive vessel contraction. First, the cell constriction (upon cold or stress) can be limited by the increased release of ligand by translocation of the α_2C-adrenergic receptor from the cytosol to the cell surface. Second, α_2A-adrenergic receptors are also expressed on the presynaptic membrane and function there as a negative feedback loop for catecholamine release.⁵⁰

Based on our results, we propose a model explaining the adrenergic pathomechanism of RS that underlines the power of human genetics-driven studies for understanding disease mechanisms. Our SMC contraction assays show that cold-induced SMC contractility was modified by ADRA2A expression. Furthermore, we combined functional studies with the genomics-driven discovery that genetic variation at the ADRA2A locus leads to increased expression of α_2A-adrenergic receptor in a population of SMCs. Such an increase in ADRA2A expression may sensitize these cells to adrenergic response and lead to increased signaling through the adrenergic pathway (Figure 6).

Figure 6.

Schematic of the proposed RS-associated pathomechanism

We propose a possible mechanism based on our observations from genetic and contraction assays. In this hypothesized model, the pathomechanism of RS is dependent on ADRA2A expression. In healthy patients without RS, there are mechanisms to prevent unwanted and excessive vessel contraction. First, the cell constriction (upon cold or stress) can be limited by the increased release of ligand by translocation of the α_2C-adrenergic receptor from the cytosol to the cell surface. Second, α_2A-adrenergic receptors are also expressed on the presynaptic membrane and function there as a negative feedback loop for catecholamine release. In patients with RS, the expression of the ADRA2A gene, that encodes the α_2A-adrenergic receptor, is higher, and therefore the number of α_2A-adrenergic receptor available for ligand binding on microvascular SMCs is also higher. Increase in SMC ADRA2A expression sensitizes the postsynaptic system that aggravates the adrenergic effects of SMC contraction. In conditions such as cold and stress where more ligand is released, the contraction is further accentuated. Black arrows represent adrenergic downstream signaling strength.

In addition to the signal from the adrenergic system, our meta-analysis indicated signals that putatively participate in the regulation of vascular tone and, in particular, eNOS, encoded by NOS3. These findings clarify the pathological mechanisms with RS: nitric oxide itself is a well-known mediator of vasodilation in vessels.⁵¹ We show that the lead variant is a strong eQTL for NOS3 expression. In addition, targeting the lead variant rs3918226 using CRISPRi, we observe a significant decrease in NOS3 expression. Therefore, this genetic signal can be attributed as one of the underlying pathological mechanisms in RS.^52,53

These genetic associations may also elucidate the more specific disease mechanisms or even provide insight into the heterogeneity of the symptomatology in RS. For example, the signal from the HLA class I region on chromosome 6 points toward the possible immune, infectious, or autoimmune mechanisms in RS. It is still unclear, however, if this signal from autoimmune diseases that have secondary RS and not primary RS per se,^54,55 which was also indicated by Hartmann et al. in the UKB.¹⁵ In our meta-analysis for primary RS in the UKB and FinnGen, though not in the meta-analysis for secondary RS, the HLA region reached genome-wide significance, which could indicate a possible immune or infectious role for HLA in primary RS as well.

The gene IRX1, at one of the associated loci, has been implicated in embryonic development and the development of fingers and digits, which are affected in RS,^56,57 and as a tumor-suppressor gene.^58,59,60,61 As IRX1 based on our RNAscope is predominantly expressed in the upper layer of the epidermis, around smaller dermal capillaries, one possibility is that it may (as a transcription factor) act through regulating the fibroblast-to-myofibroblast phenotypic switch upon cold challenge, leading to a more contractile phenotype.⁶² On the other hand, as IRX1’s role during vertebrate limb embryogenesis is known,^63,64 it may as well play a role in the development of microvasculature, which is often observed to be tortuous in patients with RS.²⁶ Variants in our other associated loci, ACVR2A, RAB6C, PCDH10, and TMEM51, have been implicated in immunological traits, cell cycle, neurological disease, and contractile function in cardiomyocytes, respectively.^25,65,66,67

Three of the genes implicated in RS have approved pharmaceuticals: NOS3 is targeted by nitric oxide synthase inhibitors, the ACVR2A pathway is targeted by soluble receptor antagonists, and ADRA2A is targeted by alpha blockers. Our findings provide insight into better clinical recognition of primary and secondary RS and raise the possibility of repurposing approved drugs. Currently, alpha blockers, which can bind ADRA2A, or beta blockers that bind beta-adrenergic receptors are approved for both systemic and topical clinical use. A compelling avenue for future drug development may be harnessing a specific ligand for α_2A-receptor that could topically be applied, reaching the target superficial vasculature even without blood flow. Such drugs might reduce the duration or intensity of vascular contraction during an active RS episode. As alpha blockers alone only modestly affect RS symptoms, our data suggest that the approved drugs could be evaluated as a combination therapy as a prophylactic to prevent an RS attack. Furthermore, such topical applications might reduce potential off-target effects in tissues that are not part of the peripheral vasculature but might still express ADRA2A. The feasibility of the potential drug repurposing, including the possible off-target effects, needs to be evaluated by future clinical studies.

Limitations of the study

Our study should be interpreted within the following limitations. Only a subset of individuals with RS have symptoms that are recognized or need treatment, and we are potentially missing the more benign spectrum of symptoms in the current population. Our findings are limited to individuals with European ancestry. The case definition for genetic and epidemiological measures included primarily diagnosis codes. This may have resulted in biases or a lack of representation of the population-level prevalence of non-clinical RS. Our comparison between different sources of data within the UKB and FinnGen, however, showed that the effects were similar regardless of whether cases were defined using hospital records or primary care records (Figure S9). We conducted a sensitivity analysis including only primary RS or secondary RS cases to address the fact that RS can be triggered by other conditions as well. The used primary or secondary RS definition does not consider that primary RS can progress into secondary RS, meaning that a contributing other disease is diagnosed later on. Moreover, data from subcutaneous microvascular cells or microvascular cells from patients with RS were not available for this study, and any genetic or expression effects might be affected after disease onset in patients with RS, something that we cannot currently capture. Finally, our functional analyses focus mainly on the cold-induced biological mechanisms, and further studies are required to show whether the same mechanisms are in place when the symptoms of patients with RS are triggered by stress or emotions, and further studies are needed especially regarding IRX1’s role.

Conclusions

To summarize, we report here eight associations with RS with computational follow-up in all loci and functional follow-up in three loci. Our findings point toward the vascular pathophysiology underlying RS and specifically to the dysfunction of the autonomic nervous system and its dialogue with the vascular structure. Our results also suggest that the pathophysiological RS phenotype can be mediated through multiple, possibly independent, mechanisms involving adrenergic signaling, immune mechanisms, and NOS and agree with earlier studies.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
ADRA2A probe	ACDBio	602791
IRX1 probe	ACDBio	1043401
Bacterial and virus strains
Lentivirus 2ND generation pCMV-dR8.91	Gilbert et al.68 and Horlbeck et al.69	Addgene 8455
Lentivirus 2ND generation, pMD2.G	Trono lab, unpublished	Addgene 12259
NEB® Stable Competent E. coli, High Efficiency	NEB	C3040H
NEB®5-alpha Competent E. coli (High Efficiency)	NEB	C2987H
Biological samples
De-identified human skin biopsy samples	This paper	Stanford University, Department of Dermatology
Critical commercial assays
Cyto-Select 48-well Cell Contraction Assay Kit	Cell Biolabs	N/A
PrimeFlow™ RNA Assay Kit	ThermoFisher Scientific	88-18005-210
RNAscope™ 2.5 HD Assay-Red and ACD RNAscope 2.5 HD Duplex Assay	ACD Bio	N/A
RNAiMax	Invitrogen, Carlsbad, CA	13778150
Active Motif ATAC-seq kit	Active motif	53150
Deposited data
Meta-analysis summary statistics	This paper	Dryad open repository (https://doi.org/10.5061/dryad.1g1jwsv53)
FinnGen	Kurki et al.70	https://www.finngen.fi/en/access_results
UK Biobank	Bycroft et al.71	https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access
The Estonian Biobank	Leitsalu et al.72	releases@ut.ee
The Mass General Brigham Biobank	Boutin et al.73	https://www.partners.org/Medical-Research/Support-Offices/Human-Research-Committee-IRB/Default.aspx
GTEx eQTL data	GTEx Consortium74	https://gtexportal.org/home/
Skin scRNAseq data	He et al.75	Reposited in GEO under GSE147424.
ScRNA seq	CZI human vascular atlas	https://cellxgene.cziscience.com/
NOS3	This paper, IGVF database, https://data.igvf.org/Accession code: IGVFDS6983MMSZ	See Table S16
Experimental models: Cell lines
Pulmonary artery smooth muscle cells	Pulmonary Hypertension Breakthrough Initiative	https://ipahresearch.org/phbi-research/)
Human coronary artery smooth muscle cells	Cell applications Inc.	350-05a
TeloHAEC dox-inducible CRISPRi cell line	Schnitzler et al.76	N/A
HEK 293T cells	DuBridge et al.77 and Pear et al.78	ATCC CRL-3216
Oligonucleotides
CRISPRi Tiling Guides for CRISPRi-FlowFISH	This paper	See Table S15
ADRA2A siRNA	Dharmacon, ONTARGET plus SMART pool siRNA	L-005422-00-0005
ADRA2C siRNA	Dharmacon, ONTARGET plus SMART pool siRNA	L-005424-00-0005
ADRA2A	Taqman probe	Hs01099503
ADRA2C	Taqman probe	Hs03044628
NOTCH3	Taqman probe	Hs01128537
UBC	Taqman probe	Hs00824723
Recombinant DNA
SgOpti-blasticidin plasmid	Schnitzler et al.76	Modified from Addgene plasmid #85681 to contain blasticidin resistance gene (not puromycin resistance gene)
pBA904 with dCas9-KRAB	Replogle et al.79	Addgene, 122238
ADRA2A, Human Tagged ORF Clone in pLenti-C-Myc-DDK Lentiviral Gene Expression Vector	OriGene Technologies	NM_000681
ADRA2C, Human Tagged ORF Clone in pLenti-C-Myc-DDK Lentiviral Gene Expression Vector	OriGene Technologies	NM_000683
Software and algorithms
R	R version 4.0.0. | 4.0.1. | 4.2.2 | 4.3.2.	https://www.r-project.org/
METAL	Willer et al.17	https://genome.sph.umich.edu/wiki/METAL_Documentation
REGENIE	Mbatchou et al.80 v.2.2.4 |v.3.1.1 |v.3.0.3 |v.3.2.2	https://github.com/rgcgithub/regenie
LDSC	Bulik-Sullivan et al.36	https://github.com/bulik/ldsc
CRISPR Tiling Design	Fulco et al.81 Nasser et al.82	https://github.com/broadinstitute/CRISPRiTilingDesign/
CRISPRi-FlowFISH analysis pipeline	Fulco et al.81 Schnitzler et al.76	https://github.com/EngreitzLab/crispri-flowfish
Benchling guide Design (2024)	https://benchling.com.	gRNA sequences: Table S10

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Hanna M. Ollila (hanna.m.ollila@helsinki.fi).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The individual-level data used in this study cannot be deposited in a public repository because of the sensitive nature of the data. The FinnGen individual level data may be accessed through applications to the Finnish Biobanks’ FinnBB portal, Fingenious (www.finbb.fi). For the individual level data of the UKB, applications can be made through the UKB portal at https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access. For MGB Biobank, individual level data are available from the Mass General Brigham Human Research Office/Institutional Review Board at Mass General Brigham (contact located at https://www.partners.org/Medical-Research/Support-Offices/Human-Research-Committee-IRB/Default.aspx) for researchers who meet the criteria for access to confidential data. Lastly, for the EstBB, preliminary inquiries to access individual level data for scientific research can be sent to releases@ut.ee. In addition, summary level data have been deposited at Dryad open-access repository and are publicly available as of the date of publication (Dryad: https://doi.org/10.5061/dryad.1g1jwsv53). Accession numbers are listed in the key resources table.
• •The NOS3 dataset generated during this study is available at the IGVF database, https://data.igvf.org/ (IGVF: IGVFDS6983MMSZ).
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

De-identified human skin biopsy samples were received from Department of Dermatology at Stanford from patients undergoing skin surgery. Primary human PaSMCs obtained from Dr. Rabenovich were originally isolated from pulmonary arteries (<1mm) harvested from unused donor lungs all obtained de-identified from the Pulmonary Hypertension Breakthrough Initiative (https://ipahresearch.org/phbi-research/). Primary human coronary artery smooth muscle cells (HCASMCs) from normal human donor hearts were purchased from different manufacturers: Lonza, PromoCell, and Cell Applications. TeloHAECs were obtained from from the American Type Culture Collection (ATCC, CRL-4052).

Method details

Genetic analyses

Cohorts

FinnGen is a public-private partnership registry-based study of Finnish residents combining genetic and electronic health record data form different registers, for example, primary care and hospital in- and out-patient visits. The release 10 (R10) contains data on up to 412,181 participants, primarily of Finnish ancestry from newborns to the age of 104 at baseline recruitment. The aim of the study is to collect the data of 500,000 Finns representing 10% of the population of Finland (for more information see https://www.finngen.fi/en).

The UK Biobank (UKB) is a population-based study containing over 500,000 individuals of mainly European ancestry.⁷¹ The participants were recruited to the study between 2006 and 2010, were aged between 37 and 73 years of age and were residents of the United Kingdom. The study is a combination of, for example, different lifestyle measures, genotypes, electronic health record data, blood count data and questionnaire data, and the health record data is updated frequently to capture the health trajectories of participated individuals.

The Estonian Biobank is a population-based biobank with 212,955 participants in the current data freeze (2023v1). All biobank participants have signed a broad informed consent form and information on ICD-10 codes is obtained via regular linking with the national Health Insurance Fund and other relevant databases, with majority of the electronic health records having been collected since 2004.⁷²

The Mass General Brigham (MGB) Biobank (formerly Partners HealthCare Biobank) is a hospital-based cohort study from the MGB healthcare network in Boston (MA, USA) with electronic health record, genetic, and lifestyle data.^73,83,84 The MGB Biobank includes data obtained from patients in several community-based primary care facilities and specialty tertiary care centers in Boston, MA.^83,85

Ethics statement

Patients and control subjects in FinnGen provided informed consent for biobank research, based on the Finnish Biobank Act. Alternatively, separate research cohorts, collected prior the Finnish Biobank Act came into effect (in September 2013) and start of FinnGen (August 2017), were collected based on study-specific consents and later transferred to the Finnish biobanks after approval by Fimea (Finnish Medicines Agency), the National Supervisory Authority for Welfare and Health. Recruitment protocols followed the biobank protocols approved by Fimea. The Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa (HUS) statement number for the FinnGen study is Nr HUS/990/2017.

The FinnGen study is approved by Finnish Institute for Health and Welfare (permit numbers: THL/2031/6.02.00/2017, THL/1101/5.05.00/2017, THL/341/6.02.00/2018, THL/2222/6.02.00/2018,THL/283/6.02.00/2019, THL/1721/5.05.00/2019 and THL/1524/5.05.00/2020), Digital and population data service agency (permit numbers: VRK43431/2017–3, VRK/6909/2018–3, VRK/4415/2019–3), the Social Insurance Institution (permit numbers: KELA 58/522/2017, KELA 131/522/2018,KELA 70/522/2019, KELA 98/522/2019, KELA 134/522/2019, KELA 138/522/2019, KELA 2/522/2020,KELA 16/522/2020), Findata permit numbers THL/2364/14.02/2020, THL/4055/14.06.00/2020, THL/3433/14.06.00/2020, THL/4432/14.06/2020, THL/5189/14.06/2020, THL/5894/14.06.00/2020, THL/6619/14.06.00/2020, THL/209/14.06.00/2021, THL/688/14.06.00/2021, THL/1284/14.06.00/2021,THL/1965/14.06.00/2021, THL/5546/14.02.00/2020 and Statistics Finland (permit numbers: TK-53-1041-17 and TK/143/07.03.00/2020 (earlier TK-53-90-20)).

The Biobank Access Decisions for FinnGen samples and data utilized in FinnGen Data Freeze7 include: THL Biobank BB2017_55, BB2017_111, BB2018_19, BB_2018_34, BB_2018_67, BB2018_71, BB2019_7, BB2019_8, BB2019_26, BB2020_1, Finnish Red Cross Blood Service Biobank 7.12.2017, Helsinki Biobank HUS/359/2017, Auria Biobank AB17-5154 and amendment #1 (August 17 2020), Biobank Borealis of Northern Finland_2017_1013, Biobank of Eastern Finland 1186/2018 and amendment 22 §/2020, Finnish Clinical Biobank Tampere MH0004 and amendments (21.02.2020 06.10.2020), Central Finland Biobank 1–2017, and Terveystalo Biobank STB 2018001.2.

The activities of the EstBB are regulated by the Human Genes Research Act, which was adopted in 2000 specifically for the operations of the EstBB. Individual level data analysis in the EstBB was carried out under ethical approval 1.1–12/624 from the Estonian Committee on Bioethics and Human Research (Estonian Ministry of Social Affairs), using data according to release application 6–7/GI/16279 from the Estonian Biobank.

The North West Multi-centre Research Ethics Committee (MREC) has granted the Research Tissue Bank (RTB) approval for the UKB that covers the collection and distribution of data and samples (http://www.ukbiobank.ac.uk/ethics/). Our work was performed under the UKB application number 22627 (Principal Investigator Dr Matti Pirinen, FIMM). All participants included in the conducted analyses have given a written consent to participate.

The MGB Biobank has obtained a Certificate of Confidentiality. In addition, The MGB Biobank works in close collaboration with the Partners Human Research Committee (PHRC) (the Institutional Review Board). This collaboration has ensured that the Biobank’s actions and procedures meet the ethical standards for research with human subjects. Biobank patients are recruited from inpatient stays, emergency department settings, outpatient visits, and electronically through a secure online portal for patients. Recruitment and consent materials are fully translated in Spanish to promote patient inclusion. The systematic enrollment of patients across the MGB network and the active inclusion of patients from diverse backgrounds contribute to a Biobank reflective of the overall demographic of the population receiving care within the MGB network. Recruitment for the Biobank launched in 2009 and is ongoing through both in-person recruitment at participating clinics and electronically through the patient portal. The recruitment strategy has been described previously.⁸³ All recruited patients provided written consent upon enrollment, and are offered an option to refuse consent.

Genotyping and quality control

Genotyping in the FinnGen cohort was performed by using Illumina (Illumina Inc., San Diego, CA, USA) and Affymetrix arrays (Thermo Fisher Scientific, Santa Clara, CA, USA) and lifted over to Genome Reference Consortium Human Build version 38 (GRCh38/hg38).⁷⁰ Individuals with high genotype absence (>5%), inexplicit sex or excess heterozygosity (+-4 standard deviations) were excluded from the data.⁷⁰ Additionally, variants that had high absence (>2%), low minor allele count (<3) or low Hardy-Weinberg Equilibrium (HWE) (P < 1 × 10⁻⁶) were removed. More detailed explanations of the genotyping, quality control and the genotype imputation are provided elsewhere.⁷⁰ All individuals in the cohort were Finns and matched against the SiSu v4 reference panel (http://www.sisuproject.fi/).

All the EstBB participants have been genotyped at the Core Genotyping Lab of the Institute of Genomics, University of Tartu, using Illumina Global Screening Array v3.0_EST. Samples were genotyped and PLINK format files were created using Illumina GenomeStudio v2.0.4. Individuals were excluded from the analysis if their call-rate was <95%, if they were outliers of the absolute value of heterozygosity (>3SD from the mean) or if sex defined based on heterozygosity of X chromosome did not match sex in phenotype data. Before imputation, variants were filtered by call-rate <95%, HWE P-value < 1 × 10⁻⁴ (autosomal variants only), and minor allele frequency <1%. Genotyped variant positions were in build 37 and were lifted over to build 38 using Picard. Phasing was performed using the Beagle v5.4 software.⁸⁶ Imputation was performed with Beagle v5.4 software (beagle.22Jul22.46e.jar) and default settings. Dataset was split into batches of 5,000. A population-specific reference panel consisting of 2,695 WGS samples was utilized for imputation and standard Beagle hg38 recombination maps were used. Based on the principal component analysis, samples who were not of European ancestry were removed. Duplicate and monozygous twin detection was performed with KING 2.2.7,⁸⁷ and one sample was removed from the pair of duplicates. Analyses were restricted to individuals with European ancestry.

The UKB genotyped 488,477 participants: 49,950 on the Affymetrix (Thermo Fisher Scientific) UK BiLEVE Axiom Array and 438,427 on the highly similar Affymetrix UK Biobank Axiom Array. These arrays captured up to 825,927 SNPs and short indels, with variants prioritized for known coding variants, those previously associated with disease and ancestry-specific markers that provide a good imputation backbone. DNA was extracted from blood samples taken at baseline interviews, between 2006 and 2010, and genotyping was carried out in 106 sequential batches, giving genotype calls for 812,428 unique variants in 489,212 participants. After removing high missingness and very rare variants, as well as poor-quality samples, these genotypes were phased using SHAPEIT3 and imputed to the Haplotype Reference Consortium (HRC) and to a merged UK10K and 1000 Genomes phase 3 reference panel,⁸⁸ both in genome assembly GRCh37 using IMPUTE2. This resulted in 93,095,623 autosomal SNPs, short indels and large structural variants in 487,442 individuals and 3,963,705 markers on the X chromosome. For more details, see Bycroft et al. (2018).⁷¹

The MGB Biobank genotyped 53,297 participants on the Illumina Global Screening Array ('GSA') and 11,864 on Illumina Multi-Ethnic Global Array ("MEG"). The GSA arrays captured approximately 652K SNPs and short indels, while the MEG arrays captured approximately 1.38M SNPs and short indels. These genotypes were filtered for high missingness (>2%) and variants out of HWE (P < 1 × 10⁻¹²), as well as variants with an AF discordant (P < 1 × 10⁻¹⁵⁰) from a synthesized AF calculated from GnomAD subpopulation frequencies and a genome wide GnomAD model fit of the entire cohort. This resulted in approximately 620K variants for GSA and 1.15M for MEG. The two sets of genotypes were then separately phased and imputed on the TOPMed imputation server (Minimac4 algorithm) using the TOPMed r2 reference panel. The resultant imputation sets were both filtered at an R2 > 0.4 and a MAF >0.001 (minor allele frequency), and then the two sets were merged/intersected resulting in approximately 19.5M GRCh38 autosomal variants. The sample set for analysis here was then restricted to just those classified as EUR (N = 54,452) according to a metric of being +/− 2 SDs of the average EUR sample’s principal components 1 to 4 in the HGDP reference panel.

Phenotype definition

We built the phenotype for RS using ICD10 code I73.0 in all our cohorts (see Table S1 for the total number of cases and controls in each cohort used). Additionally, we used the self-reported measures and primary care codes of RS in the UKB, and the ICD-9 code 4430 in the FinnGen and the UKB.

For FinnGen phenotype definition, we used both the hospital record data (inpatient N = 88 (4.2%) cases and outpatient N = 1,204 (57.8%) cases) and primary care data (N = 792 (38.0%)). Most of the cases (N = 2,025 (87.75%)) in the EstBB were from primary care data setting, and the rest were from hospital record data (N = 180 (12.25%)). In the case of MGB, all the cases were obtained from hospital record data (inpatient N = 251 (13.2%) cases and outpatient N = 1,656 (86.8%) cases).

From the UKB data, we obtained both self-reported and electronic health record data for disease definitions. To define the phenotypes, we used data from the self-report non-cancer illness codes (data field 20002), which were assessed during the baseline interview, hospital inpatient records (HES; data field 41234) and primary care diagnosis records (data field 42040). For RS, code 1561 was used from the self-reported data. From the hospital inpatient data, we included individuals as a case for the phenotype if they had I73.0 ICD-10 or 4430 ICD-9 diagnosis code. In the primary care data, diagnoses are coded using the NHS-specific Read v2 or CTV3 codes instead of the ICD-coding. We used the following Read codes to define the respective phenotype.

• (1)Read v2: “G730.”, “G7301”, “G7300” or “G730z”
• (2)Read CTV3 (v3): “G730.”, “XE0VQ”, “G7300”, “G730z”, “G7301”, “X7051” or “XE0XA”

With this definition for RS, we identified 5,162 cases and 440,833 controls of European ancestry. Most of the cases for RS came from the primary care data (N = 2,953 (57.2%)) and hospital inpatient data (N = 1,664 (32.2%)).

Overall, of our RS cases 5,770 (50.8%) come from primary care data, 5,043 (44.4%) from hospital record data and 545 (4.8%) from self-reported data. Individuals who did not have the codes mentioned above were used as controls.

In FinnGen and UKB, we defined primary RS cases by only including RS cases who did not have a diagnosis of a comorbidity prior or at the same time as RS (list of comorbidity codes Table S7A). Comorbidities and risk conditions were defined similarly as Hartmann et al. (2023)¹⁵ and recommended by Wigley et al. (2016)²⁶ With these definitions, we ended up with 1,418 cases in FinnGen and 3,568 cases in the UKB. Secondary RS cases were defined as individuals who had RS ICD-10 I73.0 diagnosis code and a predefined comorbidity condition prior to or at the same time as the RS ICD-10 code (Table S7A). With this definition we identified 1,594 cases and 666 cases of secondary RS in the UKB and FinnGen, respectively. Secondary RS cases were removed from primary RS analyses and primary RS cases were removed from secondary RS analyses.

GWAS

For the FinnGen cohort, GWAS was conducted using the REGENIE (v.2.2.4) pipeline for R10 data⁸⁰ (https://github.com/FINNGEN/regenie-pipelines). Analysis was adjusted for age at death or end of follow up (12/31/2021), sex, genotyping batches and the first 10 genetic principal components. Firth approximation was applied for variants with association P-value <0.01.

The UKB GWA analyses were performed using REGENIE v.3.1.1.⁸⁰ The whole-genome regression model (step 1) was created using 524,307 high-quality genotyped SNPs (bi-allelic; MAF ≥1%; HWE P > 1 × 10⁻⁶; present in all genotype batches, total missingness <1.5% and not in a region of long-range LD⁸⁹ with the leave-one-chromosome-out (--loocv) option enabled. We corrected for the following covariates.

• (1)age at follow-up end (2019/08/18) or death (if earlier than follow-up end), calculated as the difference in years between the 15th day of month and year of birth (data fields 52 and 34, respectively) and the follow-up end or death date.
• (2)sex (data field 31)
• (3)genotyping array (categorical), derived from genotyping batch (data field 20000), as “UKB BiLEVE” (batches −11 to −1), “UKB Axiom release 1” (1–22) and “UKB Axiom release 2” (23–95).
• (4)genetic principal components 1 to 10 (data field 22009)
• (5)center of baseline visit (categorical; data field 54)

The GWA (step 2) was performed using v3 imputed genotypes⁷¹ for chromosomes 1–22 and X with the approximate Firth correction applied for variants with association P-value <0.05 (default setting), using the flags --firth, --approx and --firth-se. After analysis with REGENIE, we excluded results for imputed variants with MAF <0.1% and/or imputation INFO <0.3.

Association analysis in the EstBB was carried out for all variants with an INFO score >0.4 using the additive model as implemented in REGENIE v3.0.3 with standard binary trait settings.⁸⁰ Logistic regression was carried out with adjustment for current age, age², sex and 10 first genetic principal components as covariates, analyzing only variants with a minimum minor allele count of 2.

In the MGB Biobank, the association analysis was carried out using REGENIE v3.2.2,⁸⁰ with covariates of Age, Sex, Genotype-Chip and five first principal components of ancestry calculated on just the analysis set (EUR only) of samples, analyzing only variants with a minimum minor allele count of 10.

Manhattan-plots for all the cohorts’ GWASs and meta-analyses were plotted using R version 4.0.1 (packages: qqman and RColorBrewer).

Meta-analysis

Meta-analyses were conducted using METAL (https://genome.sph.umich.edu/wiki/METAL_Documentation)¹⁷ with standard settings, and tracking allele frequency with the AVERAGEFREQ option and analyzing heterogeneity between used summary statistics with the ANALYZE HETEROGENEITY option. Summary statistics from different cohorts were matched against rsIDs and the final meta-analysis summary statistics are in GRCh38. Both sample size based meta-analysis and effect estimate based analyses were run (Figure 1; Tables 1 and S3). Meta-analysis was run to all RS (11,358 cases and 1,106,871 controls) summary statistics (UKB, FG, EstBB and MGB Biobank), primary RS summary statistics (UKB and FG) and secondary RS summary statistics (UKB and FG). Locus zoom plots from the meta-analysis results were created using the LocusZoom web browser (https://github.com/statgen/locuszoom).⁹⁰

eQTL and co-localization analyses

We conducted the eQTL analysis for our lead variant (rs7090046) at the ADRA2A locus by using the web browser of the GTEx project (https://gtexportal.org/home/).⁷⁴ Co-localization analyses were performed using the coloc R package (v5.1.0.1)^91,92 in R v4.2.2. We extracted all variants in a 200kb region centered on the lead variant and imported the same region from GTEx v8⁷⁴ eQTL association statistics. We then tested the co-localization between RS and each of the 49 tissues and generated co-localization plots using the LocusCompareR R package (v1.0.0)⁹³ using LD r² from 1000 Genomes⁸⁸ European-ancestry samples. Co-localization was repeated to all lead variants in all genome-wide loci in the meta-analysis. Those genes that reached over 0.85 posterior probability were used for visualizing the results as co-localization plots (Figure S10). Results from the HLA-area were not included in the correlation plots due to the high LD structure of the area and unspecific results in the co-localization analysis (Table S11).

Fine-mapping, survival analysis and allele frequency map

Fine-mapping was performed on all predefined genome-wide LD Blocks (https://bitbucket.org/nygcresearch/ldetect-data/src/master/EUR/fourier_ls-all.bed) with 5 or more variants that had a genome-wide significant P-value (P < 5 × 10⁻⁸) and passed a MAF filter of 0.01 using Sum of Single Effects (SuSiE) Regression on Summary Statistics.^94,95 The R package "susieR" was used to run the analysis, with the function susie_rss(), utilizing default parameters with R v.4.0.0. Participants of the UK Biobank of European ancestry were used as the LD reference panel for the analysis.

To assess risk allele effect over time in the lead variant at the ADRA2A locus, rs7090046, we used multivariable Cox regression analysis⁹⁶ in FinnGen. Individuals were divided into homozygote for the risk allele (AA), heterozygote (AG/GA) and homozygote for the reference allele (GG). As the follow-up time we used age at the end of follow-up period or death (12/31/2021), and adjusted the model with sex, cohort (biobanks and studies incorporated in FinnGen) and first 10 genetic principal components. Multivariable Cox regression was run with R v.4.3.2. (R-packages: data.table, lubridate, ggplot2, dplyr, survival, survminer, survMisc, stringr, stringi and RColorBrewer).

The frequency of allele A of rs7090046 was estimated for each municipality of Finland. We used data on rs7090046 genotype and municipality of birth for 47,950 Finnish individuals from the THL (Finnish Institute for Health and Welfare, Finland - THL) biobank (project no. 2019_44). For each municipality, the allele frequency estimate was based on 652 nearest individuals to the center point of the municipality, and the location of an individual was the center point of their municipality of birth. In case there were more than 652 individuals born in the municipality, a random subsample was chosen. In case there were less than 652 individuals born in the municipality, randomly chosen individuals from nearest other municipalities were included until the target sample size of 652 was achieved. The map of Finland with the boundaries of the municipalities was generated by the geoBoundaries R-package (https://www.geoboundaries.org/).

Genetic correlation

Genetic correlation between RS and comorbidities was performed by using the LD score regression method provided by the Broad Institute of Massachusetts Institute of Technology (MIT) and Harvard and MRC Integrative Epidemiology Unit, University of Bristol.³⁶ For RS, we used the meta-analysis summary statistics both for all RS and primary RS. Summary statistics for comorbidities were obtained from the FinnGen defined phenotypes from hospital data (https://www.finngen.fi/en/access_results), and from UKB freely available summary statistics provided the Neale lab in the Broad Institute, Cambridge, MA, USA (https://pheweb.org/UKB-Neale/). HapMap 3 SNP list and European LD score files, which are provide with the software, were used in our LD Score regression analyses.

Phenotypic correlations

To assess phenotypic correlation between RS and comorbidities in FinnGen we performed a multivariable logistic regression using R v.4.3.2 (R packages: data.table, lubridate and dplyr). The model was adjusted with age at the end of follow-up (12/31/2021) or death, sex, cohort and the first 10 genetic principal components. Association analysis was repeated for both all RS and primary RS. For comorbidities, the following FinnGen defined endpoints were used: M13_SJOGREN, I9_ATHSCLE, L12_LUPUS, G6_CARPTU, G6_MIGRAINE, VWXY20_EXPOS_VIBRATI, M13_SYSTSLCE, E4_HYTHY_AI_STRICT, E4_DM2_STRICT, M13_RHEUMA, M13_DERMATOPOLY, D3_HAEMOLYTICANAEMIA, M13_POLYMYO, T1D_STRICT, D3_CRYOGLOBU, ST19_FROSTBITE and D3_AIHA_DRUG. Further information for these phenotypes can be found from https://r10.risteys.finregistry.fi/. For beta-blocker purchase, ATC-codes starting with C07 (beta-blocking agents) were used from the KELA (the Social Insurance Institution of Finland) drug purchase registry, which is incorporated in the FinnGen data.

Functional assays

RNA extraction and RT-PCR

RNA was isolated according to manufacturer’s instructions using RNeasy plus micro kit (Qiagen, #74034). The quality of the RNA was determined with Nanodrop ND-1000 (Thermo Fisher Scientific), and 500 ng of total RNA was used for cDNA synthesis using High-capacity RNA-to-cDNA kit (Life Technologies, #4388950) on a BIO-RAD C1000 thermal cycler. RT-qPCR was performed using Taqman probes for ADRA2A (Hs01099503), ADRA2C (Hs03044628) and NOTCH3 (Hs01128537) according to the manufacturer’s instructions on a ViiA7 Real-Time PCR system (Applied Biosystems, Foster City, CA). GAPDH and UBC were used to normalize relative expression levels. The 2–ΔΔCt method was used to quantify relative gene expression levels. Technical and experimental triplicates of Ct values were averaged for each sample and normalized to the housekeeping gene. Expression levels of mRNA are presented as fold change (control group = 1).

SiRNA and lentiviral overexpression

For siRNA transfection, cells were 60% confluent when treated with siRNA or scramble control to a final concentration of 20 nM with RNAiMax (Invitrogen, Carlsbad, CA). The siRNAs for ADRA2A (Cat # L-005422-00-0005) and ADRA2C (Cat # L-005424-00-0005) were purchased from Dharmacon (ONTARGET plus SMART pool siRNA). Cells were treated with an equimolar combination of Silencer and Scramble and collected 72 h after transfection.

For overexpression studies ADRA2A (Human Tagged ORF Clone in pLenti-C-Myc-DDK Lentiviral Gene Expression Vector, NM_000681) and ADRA2C (Human Tagged ORF Clone in pLenti-C-Myc-DDK Lentiviral Gene Expression Vector, NM_000683) plasmids were purchased from OriGene Technologies.^68,69 To package viruses 8.5 × 10⁵ HEK293T cells^77,78 plated in each well of a six-well plate. The following day, lentiviral gene expression vectors were co-transfected with second generation lentivirus packaging plasmids, pMD2.G and pCMV-dR8.91, into cells using Lipofectamine 3000 (Thermo Fisher, L3000015) according to the manufacturer’s instructions. ViralBoost Reagent (AllStem Cell Advancements, VB100) was added (1:500) with fresh media after 5 h. Supernatant containing viral particles was collected 48 h after transfection and filtered. PaSMCs were transduced with high MOI and treated for 12 h and collected 72 h after transfection.

BulkATAC

ATAC-seq was performed with slight modifications to the published protocol^97,98 using the Active Motif ATAC-seq kit (Cat #53150). Briefly, PaSMCs (passages 5) were cultured in SMC media until ∼75% confluence. Approximately 50 000 fresh cells in three separate samples were collected by centrifugation at 500g and washed twice with cold 1 × PBS. Nuclei-enriched fractions were extracted with cold lysis buffer containing 10 mM Tris–HCl, pH7.4, 10 mM NaCl, 3 mM MgCl2 and 0.1% IGEPAL (octylphenoxypolyethoxyethanol), and then suspended in transposition reaction buffer containing Tn5 transposases (Illumina Nextera). Transposition reactions were incubated at 37 °C for 30 min, followed by DNA purification. Libraries were initially PCR amplified using Nextera barcodes and High-Fidelity polymerase. The number of cycles was empirically determined from an aliquot of the PCR mix, by calculating the Ct value at 25–30% maximum Rn for each library preparation. The final amplified library was again purified using the Zymo DNA Clean-up and Concentration kit and purified using SPRI beads, and the DNA was evaluated and quantified using Bioanalyzer and nanodrop. Libraries were multiplexed and then sequenced on Novaseq for 150-bp paired-end sequencing. Raw fastq files were evaluated with fastqc, and then low-quality bases and adaptor contamination were trimmed by cutadapt. Reads were mapped to hg19 using bowtie2. Duplicate reads were marked by Picard Markduplicate module and removed with unmapped or mitochondrial reads by samtools. bedtools was used to generate BED file from filtered reads followed by Tn5 shifting with awk. macs2 callpeak with --broad and SICER-rb.sh with “W200 G600 E1000” parameters were used for peak calling. Assay was performed in three separate sets. Bigwig files were generated for IGV visualization.

RNAscope in situ hybridization

De-identified human skin biopsy samples from the dorsal hand were received from Department of Dermatology at Stanford from patients undergoing skin surgery. They were fixed in 4% PFA overnight and then was held in PBS twice for 10 min and once in 100% EtOH. Next the samples were dehydrated in rising alcohol series and embedded in paraffin. 8 μm sections were used for the RNA Scope in situ hybridization assay.

Additionally, frozen sections of human coronary arteries were processed according to the manufacturer’s instructions, and all reagents were obtained from ACD Bio (Newark, CA). Sections were incubated with commercially available probes against human ADRA2A (#602791) and IRX1 (#1043401). Colorimetric assays were performed per the manufacturer’s instructions using RNAscope 2.5 HD Assay-Red and ACD RNAscope 2.5 HD Duplex Assay.

Single-cell RNA sequencing (scRNAseq) from human tissue

scRNAseq data was obtained from the human vascular atlas available on https://cellxgene.cziscience.com/. Skin scRNAseq data was obtained from publication by He et al.⁷⁵ reposited in GEO (GEO: GSE147424). Both datasets were then analyzed using the R package Seurat, for skin, the 17 separate individuals were created into a merged Seurat object.⁹⁹ The number of genes, number of unique molecular identifiers and percentage of mitochondrial genes were plotted to identify outliers. Analysis and visualization of gene expression and generation were performed using Seurat’s function such as “FeaturePlot”, and “FindMarker”.

Smooth muscle cell contraction assay

PaSMCs were transfected with either scrambled siRNA or siADRA2A, siADRA2C or combination of both. Following 48 h of transfection, cells were trypsinized and collected to be used for a collagen-based cell contraction assay Cyto-Select 48-well Cell Contraction Assay Kit (Cell Biolabs, San Diego, CA). In the assay, a mixture of the pulmonary SMCs (3 × 10⁶/mL) and cold Collagen Gel Working Solution was incubated in a 48-well dish at 37°C for 1 h to induce optimal polymerization following the manufacturer’s instructions. Next, cell culture medium with added SMC contraction agent or without was added on top of each well already containing the polymerized cell and collagen mixture. The cells were then incubated at either 28°C or at 37°C, 5% CO₂. All conditions were performed in triplicate wells and experiments were repeated. After 48 h, the Keyence slide scanning microscope BZ-X810 was used to image the wells and cell contraction was measured using ImageJ by drawing the outlines of the gel, calculating the gel area, and comparing it to the well area.

Generation and analysis of CRIPSRi in PaSMC lines

Genome editing of the region around rs7090046 was performed by CRISPRi/dCas9-KRAB system as previously reported.^79,100 The guide RNAs targeting this SNP were designed using Benchling online tools. Synthesized oligos were then cloned into pBA904 vector backbone containing dCas9-KRAB and lentivirus was packaged as described above. For the CRISPR interference experiment, PaSMCs cells were seeded into 6 well plate (8×10⁵ cells/well) and triplicate wells were used for all conditions. The next day, cells were transduced with the virus for 12 h with 8 μg/mL polybrene. The cells were cultured for an additional 5 days with medium change until RNA was extracted. GuideRNA sequences are listed in Table S14.

CRISPRi-FlowFISH in TeloHAEC cells

The CRISPRi-FlowFISH process promotes chromatin repression in specified regions using a KRAB-dCas9 complex, which combines the KRAB effector domain with a deactivated Cas9 protein and a pool of guide RNAs that are virally infected into a population of cells. In this experiment, we used a TeloHAE cell line with the KRAB-dCas9-IRES-BFP expression cassette downstream of a doxycycline-inducible promoter, as previously described.⁷⁶ Using CRISPRi-FlowFISH, we evaluated the effect size of CRISPRi silencing of a 300 bp region around the rs3918226 variant, compared to the NOS3 transcription start site (TSS) (500 bp region) and negative control CRISPRi guides. Guides were designed using our standard pipeline (https://github.com/EngreitzLab/CRISPRDesigner), which was previously described,^81,82 with 15 guides targeting the rs3918226 variant, 15 guides targeting the NOS3 TSS, 40 non-targeting control guides (guides which do not have a close match to any region in the human genome⁶⁸) and 40 safe-targeting guides (targeting non-functional genomic sites) (Tables S15 and S16). We excluded gRNAs with low specificity scores or low-complexity sequences as previously described.⁸²

Briefly, TeloHAEC CRISPRi cells were infected via lentivirus containing the pool of CRIPSRi guides, maintaining a coverage of at least 500 cells per guide. The experiment was set up with two biological replicates. TeloHAEC cells were infected at a density of 1.5 x 106 cells/mL per mL with polybrene (10 μg/mL) and a volume of lentivirus containing our pool of guides, to give an infection rate of ∼15%. Cells were infected in 12-well plates which were centrifuged for 2 h at 2000 r.p.m, at 30°C, before being incubated at 37°C. Blasticidin (final concentration 15 μg/mL) was applied 24 h later and selection for infected cells was carried out for 6 days. After day 6, the dose of blasticidin was reduced to 3 μg/mL, to allow the cells to proliferate and expand while not losing the integration of the sgRNA. Doxycycline (2 μg/mL) was added for 72 h to induce the CRISPRi machinery (with a continued maintenance dose of blasticidin (3 μg/mL)).

To quantify the results on gene expression, we performed fluorescence in situ hybridization (FISH) which utilizes fluorescent probes to label the specified RNA. We used the PrimeFlow RNA Assay Kit from (Cat # 88-18005-210, ThermoFisher Scientific), according to the manufacturer’s instructions, with some minor modifications, as previously described.¹⁰¹ We used 6 million cells per reaction, with two technical RNA-FlowFISH replicates for each biological replicate. We performed the fixation and permeabilization steps in one tube for each biological replicate, before splitting the reactions into individual 1.5 mL tubes prior to the target probe hybridization step in the protocol. During the target probe hybridization, the control reaction (termed ‘unstained’) is incubated with 100 μl of Target Probe Diluent without target probes, while the FlowFISH reaction tubes are stained with the following probes: NOS3 (gene of interest) was labeled with ‘Type-1’ probeset (Cat # VA1-11705-PF, ThermoFisher Scientific) and RPL13A (‘housekeeping’ control gene) was labeled with ‘Type-4’ probeset (Cat # VA4-13187-PF, ThermoFisher Scientific). The purpose of the ‘unstained’ control reaction is to assess the background fluorescence signal from non-specific binding of label probes. Each sample (including the control ‘unstained’ sample) was stained with the fluorescent Label Probes (Alexa Fluor 647 for NOS3 and Alexa Fluor 488 for RPL13A). Finally, we added a total of five washes with 40°C wash buffer following the final wash described in the manufacturer’s protocol. This is to remove an excess unbound label probe.

We observed an approximately 5.9-fold-change signal for NOS3 in TeloHAEC CRISPRi cells with probes applied (stained) versus cells without target gene probes applied (unstained) (Figure S10). We used a SONY MA900 Cell Sorter to perform fluorescence-activated cell sorting (FACS) which split the cell population into four bins based on RNA expression levels. From there, high-throughput sequencing was used to identify the frequency of gRNAs in each expression bin, which was then used to determine the effect each gRNA has on gene expression, as previously described.^81,82 We scaled the effect size of each gRNA in the screen linearly so that the strongest guide at the TSS of the target gene has an 85% effect on expression of the target gene.⁸² This accounts for non-specific probe binding that occurs in the RNA FISH assay and it is based on our observation that promoter CRISPRi typically shows 80–90% knockdown by qPCR. Significance was determined via an unpaired two-tailed t-test, comparing targeting guides to the negative control guides. The data is reposited in IGVF database, https://data.igvf.org/(IGVF: IGVFDS6983MMSZ).

Cell culture and sample processing

Primary human PaSMCs obtained from Dr. Rabenovich were originally isolated from pulmonary arteries (<1mm) harvested from unused donor lungs all obtained de-identified from the Pulmonary Hypertension Breakthrough Initiative (https://ipahresearch.org/phbi-research/). For the experiments, total RNA was isolated from confluent cells at passage 5 and 6.

TeloHAEC CRISPRi cells were grown in VascuLifeⓇ VEGF Endothelial Medium Kit (Cat # 102970–922, Lifeline Cell Technology). Cells were split with 0.05% Trypsin-EDTA (Cat # 25300120, ThermoFisher Scientific), approximately every 2–3 days as needed, when approaching confluence.

Quantification and statistical analysis

Statistics of computational analyses

Computational analyses were conducted using REGENIE,⁸⁰ METAL,¹⁷ LDSC³⁶ and R (versions specified above in the methods details section). REGENIE is a C++ program designed for whole genome regression modeling of large-scale GWAS.⁸⁰ It uses generalized mixed linear (GML) model for binary or quantitative phenotypes accounting for population structure and relatedness.⁸⁰ METAL is software used for meta-analysis of GWAS summary statistics,¹⁷ combining summary statistics that are weighted proportional to the square root of each GWAS sample size. It allows meta-analysis using either test-statistic or P-values and direction of effect, calculating a Z score for each variant and performing a Z-test to calculate the P-value of a combined effect, with heterogeneity of effect calculated using a Cochran’s Q-test.¹⁷ LDSC can be used to study partitioned heritability and genetic correlation between traits using GWAS summary statistics quantifying the individual SNP effects and confounding factors.³⁶ Co-localization analysis estimates whether a specific genetic variant is responsible for both the GWAS association and the regional eQTL.⁹² Co-localization analyses were performed using R “coloc” package^91,92 using the population enumeration approach. For fine-mapping we used SuSiE^94,95 based on a sparse multiple regression with the R package "susieR". Multivariable logistic regression in FinnGen was used to study the association between traits adjusting for age at the end of follow-up (12/31/2021) or death, sex, cohort and the first 10 genetic principal components, and using the glm() function in R. Cox regression analysis⁹⁶ was used to study the effect of the risk allele of rs7090046 variant over time in FinnGen adjusting for age at the end of follow-up (12/31/2021) or death, sex, cohort and the first 10 genetic principal components. The R package “survival” and the function coxzph() were used to conduct the Cox regression. For both multivariable logistic regression and Cox regression, the P-values were calculated using a Wald test.

Statistics and reproducibility for the functional studies

All statistical analyses for the functional studies were conducted using GraphPad Prism software version 9 (Dotmatics Inc). Difference between two groups were determined using an unpaired two-tailed Student’s t test. Differences between multiple groups were evaluated by one-way analysis of variance (ANOVA) followed by Dunnett’s post-hoc test after the sample distribution was tested for normality. P-values <0.05 were considered statistically significant. All error bars represent standard error of the mean. Number of stars for the P-values in the graphs: ∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05. No statistical method was used to predetermine sample size, which was based on extensive prior experience with this model.

Published: August 13, 2024

Supplemental information

Table S6. Fine-mapping results, related to Figure 1

Table S6A. Fine-mapping results from Raynaud’s syndrome meta-analysis - IRX1 locus. Table S6B. Fine-mapping results from Raynaud’s syndrome meta-analysis - ADRA2A locus

Table S7. Genetic and phenotypic associations with Raynaud’s syndrome, related to Figure 1

Table S7A. Diagnosis codes used for primary and secondary Raynaud’s syndrome definition. Table S7B. Multivariable logistic regression results for Raynaud’s syndrome. Table S7C. Multivariable logistic regression results for primary Raynaud’s syndrome. Table S7D. Phenome-wide associations of eight Raynaud’s syndrome lead variants with secondary comorbidities. Table S7E. LD score regression analysis results for Raynaud’s syndrome. Table S7F. LD score regression analysis results for primary Raynaud’s syndrome