Abstract

   Arterioles are small blood vessels located just upstream of capillaries
   in nearly all tissues. Despite the broad and essential role of
   arterioles in physiology and disease, current knowledge of the
   functional genomics of arterioles is largely absent. Here, we report
   extensive maps of chromatin interactions, single-cell expression, and
   other molecular features in human arterioles and uncover mechanisms
   linking human genetic variants to gene expression in vascular cells and
   the development of hypertension. Compared to large arteries, arterioles
   exhibited a higher proportion of pericytes which were enriched for
   blood pressure (BP)-associated genes. BP-associated single nucleotide
   polymorphisms (SNPs) were enriched in chromatin interaction regions in
   arterioles. We linked BP-associated noncoding SNP rs1882961 to gene
   expression through long-range chromatin contacts and revealed
   remarkable effects of a 4-bp noncoding genomic segment on hypertension
   in vivo. We anticipate that our data and findings will advance the
   study of the numerous diseases involving arterioles.

   Subject terms: Genetic association study, Hypertension, Vascular
   diseases, Epigenomics
     __________________________________________________________________

   Liu et al., report extensive maps of chromatin interactions,
   single-cell expression, and other molecular features in human
   arterioles and uncover mechanisms linking noncoding genetic variants to
   gene expression and the development of hypertension.

Introduction

   Arterioles are small blood vessels found just upstream of capillary
   beds in nearly all tissues^[59]1–[60]3. Arterioles have diameters of 10
   to 300 µm and consist of one layer of endothelial cells (EC) and one to
   a few layers of smooth muscle cells. The constriction and dilation of
   arterioles controls the blood flow to tissue regions. Therefore,
   arteriolar function is essential for maintaining proper distribution of
   blood across tissue regions at rest and changing blood flow in response
   to stimuli such as stress, exercise, and food ingestion^[61]1–[62]3.
   Additionally, arterioles are critical for determining systemic blood
   pressure (BP). BP is the product of total peripheral vascular
   resistance and cardiac output. Arterioles account for 60% to 70% of
   total peripheral vascular resistance in humans. Abnormalities in
   arterioles contribute to the development of major diseases such as
   hypertension, stroke, and many other microvascular complications of
   diabetes and hypertension^[63]4–[64]6.

   Despite the broad and essential role of arterioles in physiology and
   disease, current knowledge of the functional genomic features of
   arterioles is limited. This is in part because it is difficult to
   obtain arteriole samples for analysis, which requires skilled,
   microscopic tissue dissection^[65]5. Some functional genomic data is
   available for larger arteries with diameters of several mm or more. For
   example, the Encyclopedia of DNA Elements (ENCODE) project and the
   Genotype-Tissue Expression (GTEx) project include aorta, coronary
   artery, and tibial artery in their analysis^[66]7,[67]8. However,
   larger arteries and arterioles are structurally and physiologically
   highly distinct. Larger arteries do not contribute to the regulation of
   tissue perfusion or BP as arterioles do, and they contained more layers
   of smooth muscle cells and more elastic tissues than
   arterioles^[68]9,[69]10. Therefore, functional genomic insights
   obtained from the analysis of larger arteries may not be applicable to
   arterioles, leaving a critical knowledge gap.

   In this study, we established an extensive epigenomic landscape for
   human arterioles. We generated data for global and promoter-focused
   chromatin conformation, DNA methylation, and bulk and single-nucleus
   gene expression in human arterioles or tissue components of human
   arterioles. We carried out integrated analysis of these datasets with
   single nucleotide polymorphisms (SNPs) associated with vascular traits
   including BP identified by previous genome-wide association studies
   (GWAS). We further investigated the BP-associated, noncoding SNP
   rs1882961 that we found to form long-range chromatin interactions with
   NRIP1 promoter. Using genome editing in animal models and human induced
   pluripotent stem cells (hiPSCs), we uncovered the mechanism by which
   rs1882961 regulated NRIP1 expression and demonstrated robust effects
   for a rs1882961 orthologous site on BP in a rat model.

Results

The cellular composition and properties of human arterioles are distinct from
larger arteries

   We started by examining the cell type composition of human arterioles
   using single-nucleus RNA-seq (snRNA-seq). The arterioles were dissected
   from adipose tissues from surgical discards. Three arterioles, one from
   each subject, were pooled to provide sufficient material for snRNA-seq
   (Supplementary Data [70]S1). The arterioles were 100-150 μm in diameter
   and a few mm long.

   Despite the small amount of input tissue, we obtained snRNA-seq data
   meeting our quality criteria from 1205 nuclei. We detected several
   major cell types previously detected in the thoracic aorta using the
   same set of marker genes^[71]11 (Fig. [72]1A, B). The proportions of
   the cell types were substantially different between the arteriole and
   the aorta. To account for variable recovery rates for different cell
   types during nucleus isolation, we calculated proportions of cell types
   relative to EC in each tissue. The proportion of pericytes was higher
   in arterioles than in aorta (ratio = 1.31 in arterioles vs. 0.65 in
   aorta), whereas the proportion of vascular smooth muscle cells was
   lower in arterioles (ratio = 0.22 in arterioles vs. 11.19 in aorta)
   (Fig. [73]1C). For each cell type, the Spearman correlation
   coefficients of gene expression profiles between the arteriole and the
   aorta ranged from 0.52 to 0.76 (Fig. [74]1D). We examined the
   enrichment of genes reported by genome-wide association studies (GWAS)
   of arteriole-related traits in each cell type. GWAS for diastolic BP,
   systolic BP, pulse pressure, hypertension, stroke, diabetic
   nephropathy, diabetic retinopathy, and peripheral arterial disease were
   included in this analysis. We found that BP-associated GWAS genes were
   enriched in several arteriolar and aorta cell types but to
   substantially different extents (Fig. [75]1E). Genes associated with
   diastolic BP were enriched in arteriolar, but not aortic, endothelial
   cells, and they were substantially more enriched in arteriolar
   pericytes (β = 0.35) than aortic pericytes (β = 0.11). Genes associated
   with systolic BP and pulse pressure were enriched in aortic, but not
   arteriolar, smooth muscle cells and in arteriolar fibroblasts (β = 0.14
   and 0.32, respectively) much more than aortic fibroblasts (β = 0.09 and
   0.1) (Fig. [76]1E).

Fig. 1. The cellular composition and properties of human arterioles exhibit
distinct features compared with the aorta.

   [77]Fig. 1
   [78]Open in a new tab

   A UMAP plot showing major cell types identified in arteriole (this
   study) and thoracic aorta (Pirruccello, J.P. et al., 2022) snRNA-seq
   data. B Dot plot displaying expression of conventional marker gene
   across cell types. Dot size represents the percentage of cells
   expressing each gene, and color intensity indicates the average
   expression level. C Relative proportions of pericytes to ECs and VSMCs
   to ECs in arteriole and thoracic aorta, shown separately. D Heatmap
   showing pairwise gene expression correlations across cell types in
   arteriole and thoracic aorta, with color indicating Spearman
   correlation coefficients. E Dot plot showing significant associations
   between cell types and blood pressure-related traits from MAGMA cell
   typing analysis. Statistical significance was assessed using a gene set
   enrichment analysis based on the top 10% most specific genes for each
   cell type. P-values were adjusted using the Benjamini-Hochberg (BH)
   method to control the false discovery rate (FDR). Only associations
   with FDR < 0.05 are shown. Dot size represents the effect size (beta),
   and color intensity reflects the magnitude of the –log₁₀(p-value).
   Exact p values are in the Source Data file Cell type. EC endothelial
   cell, LEC lymphatic endothelial cell, VSMC vascular smooth muscle cell.

Chromatin interactions, transcriptome, and DNA methylation in human
arterioles

   We constructed reference profiles of chromatin interactions, gene
   expression, and DNA methylation in human arterioles. The source
   subjects included a mix of race and sex (Supplementary Data [79]S1).
   From each subject, one or more arterioles were separated into
   endothelial cells and endothelium-denuded arterioles (EDA), and one or
   more were analyzed as intact arterioles without the separation. The
   chromatin interaction analysis required more input material than what
   was available from a single subject. Therefore, arteriolar tissues from
   several subjects were combined to generate a reference chromatin
   interaction map for each tissue component.

   Two chromatin interaction assays were performed: a Micro-C assay for
   global mapping and a pan-promoter capture Micro-C assay for mapping
   chromatin interactions with promoters of more than 27,000 coding and
   noncoding genes.

   The Micro-C assay generated 315 million and 406 million unique read
   pairs in intact arterioles and EDA, respectively, resulting in the
   detection of an average of 1,037, 4,156, and 4,946 chromatin loops at
   4-kbp, 8-kbp, and 16-kbp resolution, respectively (Supplementary
   Datas [80]S2–[81]S9). All loops were intrachromosomal by the default
   settings of the analysis. The median loop size ranges from 96 kbp at
   4-kbp resolution to nearly 500 kbp at 16-kbp resolution (Supplementary
   Fig. [82]S1). It is likely that we only detected a portion of the loops
   in each tissue type because of the sequencing depth and the low amounts
   of input material limiting the library complexity. Despite this
   limitation, significant overlaps, defined as overlap of both contact
   regions that defined a loop, were observed between loops detected in
   the two tissue types (Fig. [83]2A). For example, at the 16 kbp
   resolution, 33% of all detected loops were shared between intact
   arterioles and EDAs (Fig. [84]2A). Only a few dozen loops were detected
   in arteriolar endothelial cells, suggesting that, even with pooling,
   the amount of material in the arteriolar endothelial cell sample was
   insufficient for detecting most chromatin loops.

Fig. 2. Chromatin contact regions in human arterioles contain DNA regulatory
elements and correlate with greater gene expression.

   [85]Fig. 2
   [86]Open in a new tab

   A Venn diagrams showing the overlap between loops identified in
   arterioles and EC-denuded arterioles (EDA) by Micro-C at 4 kb, 8 kb,
   and 16 kb resolutions. B Venn diagrams showing the overlap between
   interactions identified in arterioles and EDA by pan-promoter Capture
   Micro-C at 10 kb and 20 kb resolutions. C Bar plot showing the number
   of loops with regulatory elements present in one or both chromatin
   contact regions in arterioles and EDA across resolutions. “None,”
   “one,” and “both” indicate the presence of regulatory elements in
   neither, one, or both interacting regions, respectively. D Number of
   chromatin interactions classified by interaction type:
   promoter-promoter (PP), enhancer-promoter (EP), enhancer-enhancer (EE),
   enhancer-transcription factor binding site (ET), promoter-transcription
   factor binding site (PT), and transcription factor binding
   site-transcription factor binding site (TT). Categories are not
   mutually exclusive. E Expression levels of genes proximal to chromatin
   interactions (within 5 kb upstream or downstream of contact regions),
   grouped by interaction type (PP, EP, EE, ET, PT, TT). Controls include
   genes proximal to regulatory regions, including promoters, enhancers,
   or transcription factor binding sites, that were not in contact regions
   based on our data. Each box summarizes gene-level data. The number of
   genes (n) per category varies and is provided in the Source Data file.
   Boxes represent the interquartile range (IQR, 25th to 75th percentile),
   center line shows the median, and whiskers extend to 1.5 × IQR. F
   Boxplot showing gene expression levels grouped by the number of
   chromatin interactions involving gene promoter regions. Spearman
   correlation was applied. Each box summarizes gene-level data. The
   number of genes (n) per category is provided in the Source Data file.
   Boxes represent the IQR (25th to 75th percentile), center line shows
   the median, and whiskers extend to 1.5 × IQR.

   The pan-promoter capture Micro-C assay detected more than 100,000 and
   80,000 chromatin interactions with promoters at 10-kbp and 20-kbp
   resolution, respectively, in intact arterioles and EDAs (Supplementary
   Datas [87]S2, [88]S10–[89]S14). 60% to 73% of these contacts were
   intrachromosomal. On average, 48,220 intrachromosomal promoter
   interactions with distances between chromatin contacts from 10 kbp to 2
   million bp were included in the downstream analysis. Chromatin
   interactions were detected for the promoters of 4,595 to 10,443 of the
   20,089 protein-coding genes covered by the human pan-promoter probe
   panel in intact Arterioles and EDAs at 10-kbp or 20-kbp resolution. The
   median distance of the interacting regions was 102.1 kbp to 112.4kbp
   (Supplementary Fig. [90]S2). Substantial overlaps, defined by overlap
   of both contact regions, were observed between interactions detected in
   the two tissue types. At 10- and 20-kbp resolution, 26% to 34% of the
   interactions were shared between intact arterioles and EDAs
   (Fig. [91]2B). More interactions were detected at 1-, 2-, or 5-kbp
   resolution, but the percent of intrachromosomal interactions were below
   35% and the overlaps between the two tissues were less than 2%,
   suggesting greater noise in the result at these resolutions. No
   significant interactions were detected in the arteriolar endothelia
   sample likely because of, again, insufficient input material.

   Poly(A)-dependent bulk RNA-seq analysis (n = 3 each) (Supplementary
   Datas [92]S2, Fig. [93]S3) indicated that, as expected, the
   transcriptome profiles of arteriolar endothelial cells differed
   substantially from the intact arterioles and EDAs, with 2891 and 2939
   differentially expressed genes (FDR < 0.05), respectively
   (Supplementary Fig. [94]S3). The arteriolar endothelial cells were
   enriched for endothelial marker genes, while the arterioles and EDAs
   were enriched for marker genes for vascular smooth muscles
   (Supplementary Fig. [95]S4). 754 genes were differentially expressed
   between arterioles and EDAs (FDR < 0.05).

   Reduced representation bisulfite sequencing (RRBS) (Supplementary
   Data [96]S2) identified only a few differentially methylated regions
   (DMRs; FDR < 0.05) between arteriolar endothelial cells and either
   intact arterioles (1 DMR) or EDAs (25 DMRs). No DMRs were observed
   between EDAs and intact arterioles. These small differences may be
   attributed to the limited sample size. Similar to previous
   reports^[97]12,[98]13, the overall correlation between RNA abundance
   and gene promoter methylation was limited (Supplementary Fig. [99]S5).

Chromatin contact regions are enriched for regulatory elements and associated
with higher gene expression

   We obtained DNA regulatory elements including enhancers, promoters, and
   transcriptional factor binding sites from Ensembl^[100]14. These
   regulatory elements were defined as genomic regions with biochemical
   activity, identified through a four-step Regulatory Build
   process^[101]15, which integrated all available epigenomic data from
   the ENCODE and Roadmap Epigenomics projects. More than 90% of the
   chromatin loops in arterioles and EDAs, defined by global Micro-C,
   contained regulatory elements in both contact regions, 8.4% contained
   regulatory elements in one of the two contact regions, and only a small
   fraction did not contain any regulatory elements in their contact
   regions (Fig. [102]2C). The overlap of chromatin contact regions and
   regulatory elements was substantial as only 12% to 18% of the genome
   was covered by the contact regions and 16% covered by the regulatory
   elements. As we used regulatory elements from a broad range of cell
   types available in the Ensembl database, it is possible that the
   overlaps of arteriole chromatin contact regions with regulatory
   elements were overestimates, although the enrichment is likely true as
   the degree of enrichment is very high.

   We grouped the chromatin loops in arterioles and EDAs into those
   involving enhancer-enhancer (EE), promoter-promoter (PP),
   enhancer-promoter (EP), enhancer-transcription factor binding site
   (ET), promoter-transcription factor binding site (PT), and
   transcription factor binding site- transcription factor binding site
   (TT) interactions (Fig. [103]2D). All types of chromatin contacts were
   associated with higher expression of nearby genes (Fig. [104]2E).

   Based on the pan-promoter capture Micro-C analysis, the vast majority
   (89–96%) of chromatin regions interacting with gene promoters contained
   known regulatory elements (Supplementary Fig. [105]S6A). 72% of the
   chromatin regions interacting with gene promoters contained enhancers,
   and 15% contained transcriptional factor binding sites (Supplementary
   Fig. [106]S6B). RNA abundance was highly significantly correlated with
   the number of chromatin regions interacting with gene promoters
   (p < 2 × 10^−16) (Fig. [107]2F). The average number of chromatin
   interactions was 3.8 per gene.

Representative genes with or without established roles in arteriolar function

   We examined promoter chromatin interactions, DNA methylation, and RNA
   abundance for several genes essential for arteriolar function. AGTR1
   (Angiotensin II Receptor Type 1) had greater expression in intact
   arterioles and EDA than in arteriolar endothelial cells (Fig. [108]3A).
   We detected interactions of AGTR1 promoter with intergenic genomic
   segments 100 kbp and 50 kbp upstream in intact arterioles and EDA,
   respectively (Fig. [109]3A). One of these interacting regions contained
   known enhancers based on Ensembl annotation. In EDA, AGTR1 promoter
   also interacted with a genomic region nearly 600 kbp downstream of
   AGTR1 (Fig. [110]3A). This genomic region contained known enhancers and
   binding sites for transcriptional factors and the CCCTC-binding factor
   (CTCF) important for chromatin conformation and transcriptional
   regulation.

Fig. 3. Chromatin interactions with gene promoters, DNA methylation, and mRNA
abundance of representative genes essential to arteriolar function.

   [111]Fig. 3
   [112]Open in a new tab

   A AGTR1 (Angiotensin II Receptor Type 1). B EDN1 (Endothelin 1). C NOS3
   (Nitric Oxide Synthase 3). Linux-based IGV was used to plot chromatin
   interactions based on pan-promoter capture Micro-C, RNA abundance based
   on poly(A)-dependent RNA-seq, and methylation levels based on RRBS.
   Only chromatin interactions with the promoter of the gene of interest
   are shown. The chromatin contact regions were compared with the
   regulatory elements as defined by Ensembl for overlaps. E enhancer, T
   transcriptional factor binding site, C CCCTC-binding factor (CTCF)
   binding site, O open chromatin region.

   In contrast, EDN1 (Endothelin-1) and NOS3 (Nitric Oxide Synthase 3)
   transcripts were abundant in intact arterioles and endothelial cells
   but much less so in EDA (Fig. [113]3B, C). For EDN1 promoter,
   interactions with a genomic segment 300 kbp downstream of the gene were
   detected in both arterioles and EDA (Fig. [114]3B). Additional
   interacting regions 600 kbp upstream and 100 kbp downstream of the gene
   were detected in arterioles and EDA, respectively. All these
   interacting regions contained known enhancers, open chromatin regions,
   or binding sites for CTCF. The genomic region surrounding EDN1 had more
   DNA methylation in EDA than in arterioles and endothelial cells
   (Fig. [115]3B). For NOS3 promoter, interacting regions were detected
   150 kbp downstream and 200 kbp upstream in arterioles and EDA,
   respectively, and these regions contained known enhancers, binding
   sites for transcriptional factors and CTCF, or open chromatin regions
   (Fig. [116]3C). The NOS3 gene body had more DNA methylation in EDA than
   in arterioles and endothelial cells.

   Next, we examined promoter chromatin interactions and DNA methylation
   for select genes without established roles in arteriolar function.
   Focusing on genes that were abundantly expressed in arteriolar tissues,
   we examined SNRNP27 (Small Nuclear Ribonucleoprotein U4/U6.U5 Subunit
   27) and LGI4 (Leucine Rich Repeat LGI Family Member 4) (Supplementary
   Fig. [117]S7). The promoter region of SNRNP27 interacted with genomic
   segments 126 kbp and 186 kbp downstream and 112 kbp upstream of the
   promoter region in both arterioles and EDA. The promoter region of LGI4
   interacted with a genomic segment 53 kbp downstream in both arterioles
   and EDA but interacted with several other genomic segments that were
   unique to arterioles or EDA.

   These findings reveal specific chromatin interactions and DNA
   methylation that may play a role in regulating the expression of
   several genes essential to arteriolar function or abundantly expressed
   but without established functional roles in arterioles.

Chromatin contact regions in human arterioles are enriched for BP-associated
SNPs

   To investigate the relation between arteriolar chromatin architecture
   and genetic determinants of arteriole-related phenotypes, we integrated
   the Micro-C and pan-promoter capture Micro-C data with an analysis of
   GWAS SNPs associated with arteriole-related traits, including systolic
   BP, diastolic BP, pulse pressure, essential hypertension, stroke,
   peripheral arterial disease, diabetic nephropathy, and diabetic
   retinopathy. By examining the overlap between these SNPs and chromatin
   interaction regions in both intact arterioles and EDA, we observed that
   SNPs associated with systolic BP, diastolic BP and pulse pressure were
   consistently enriched in chromatin contact regions compared to SNPs
   linked to non-arteriole-related traits (all p-values < 0.01)
   (Fig. [118]4A; Methods). Additionally, stroke-associated SNPs were
   significantly enriched in chromatin contacts in intact arteriole
   (p = 0.045), but not in EDA, consistent with a prominent role of
   endothelial cells in stroke pathogenesis. Our arteriole-based Micro-C
   data showed a higher capture of BP-associated SNPs per 1Mbp loop
   contact region compared to Hi-C data from tibial arteries and aorta in
   ENCODE^[119]7 (Fig. [120]4B), which might be due to differences between
   tissues, assay protocols, or both.

Fig. 4. Chromatin contact regions in human arterioles are enriched for SNPs
associated with blood pressure and stroke.

   [121]Fig. 4
   [122]Open in a new tab

   A Bar plot showing the percentage of SNPs covered by Micro-C or
   pan-promoter Capture Micro-C in arterioles and EDA, grouped by
   arteriole-related traits. Red dashed lines indicate the average SNP
   coverage across non-arteriole-related traits (Methods). One-sided
   binomial test; no multiple testing correction. B Bar plot of
   BP-associated SNPs density within 1 Mb contact regions in our Micro-C
   datasets compared to Hi-C from the tibial artery and aorta (ENCODE).
   Red dashed line represents average SNP density in Hi-C. C Venn diagrams
   showing overlap of SNP-gene pairs identified by chromatin interactions
   and GTEx artery eQTLs (aorta, tibial, coronary). Two-sided Fisher’s
   exact test against all arteriolar trait-associated SNPs linked to genes
   within ±2 Mbp; unadjusted p-values. D Bar plot of SNP-gene interaction
   distances, binned by distance categories. Only SNPs from
   arteriole-related traits (see A) are included. E Pathway enrichment of
   genes interacting with arteriole-related SNPs in pan-promoter Capture
   Micro-C, shown for arteriole and EDA. BH method applied for FDR
   correction. F Schematic of SNP-gene interaction patterns by SNP bin
   location (promoter vs. distal), with interaction counts in arterioles
   and EDA. G Percentage of covered SNPs by regulatory elements (promoter
   vs. distal). Red dashed lines represent overall SNP coverage. Two-sided
   Fisher’s exact test with BH correction; only significant enrichments
   with odds ratio > 1 are marked with asterisks. Rank plot of genes
   significantly upregulated in ECs (H) and fibroblasts (I) compared to
   all other cell types (one-sided Wilcoxon test; Bonferroni-corrected
   p-value < 0.05). Genes involved in SNP-promoter interactions are
   colored by occurrence. Enhancer SNP-promoter targets are highlighted
   with red circles. Top 10 SNP-gene pairs are labeled. J UMAP plots
   showing the expression of selected enhancer SNP-linked genes (ERG,
   FBLN5, SEMA4A and PDGFRA) across arterioles cell types (snRNA-seq
   data). Statistical significance is indicated as follows: *p < 0.05,
   **p < 0.01, and **p < 0.001. Exact p values are in the Source Data
   file.

   Further analysis of SNPs associated with arteriole-related traits and
   their interacting genes, captured through our chromatin contacts,
   identified a total of 10,217 SNP-gene pairs (Supplementary
   Data [123]S15). Notably, 8,011 of these pairs had not been previously
   reported in the GWAS Catalog or revealed through eQTL analysis by GTEx
   in large arteries, including aorta, tibial artery, and coronary artery,
   highlighting the additional information provided by chromatin
   interaction analysis. Nevertheless, the overlap between SNP-gene pairs
   identified by arteriolar chromatin interactions and large artery eQTL
   data was 938, a significant overlap compared to all arteriolar
   trait-associated SNPs linked to genes within 2 Mbp (p < 2 × 10^-16),
   supporting a regulatory role of chromatin interactions in eQTLs
   (Fig. [124]4C). Among these overlaps, 522 were not listed as nearby
   genes in the GWAS Catalog. Additionally, an average of 344
   long-distance SNP-gene interactions (>100 kb) were identified in the
   Micro-C data, while an average of 3,331 long-distance interactions were
   identified in the pan-promoter capture Micro-C data (Fig. [125]4D).
   Functional enrichment analysis of genes interacting with SNPs via
   chromatin contacts, based on the pan-promoter capture Micro-C data,
   showed significant enrichment in muscle system processes and tissue
   development in arterioles. In the EDA dataset, enriched pathways were
   related to cellular response to external stimuli, including hormone and
   peptide responses, apoptotic signaling and hypoxia (Fig. [126]4E).

Enhancer SNP-promoter interactions linked to cell type specific gene
expression

   In our pan-promoter capture Micro-C data, each interaction includes a
   captured promoter bin and a distal interacting bin (Supplementary
   Datas [127]S11–[128]S14). An average of 86% of the SNPs in the captured
   SNP-gene pairs were in the distal interacting bin (6192 out of 7216 and
   5567 out of 6427) (Fig. [129]4F). These SNPs were significantly
   enriched in enhancer regions (all p-values < 0.001), suggesting that
   SNPs frequently exert their influence on gene promoter via enhancer
   regions when mediated by chromatin interactions (Fig. [130]4G).

   To assess potential impact of SNPs on gene expression through promoter
   interaction in a cell type specific manner, we investigated two
   scenarios: 1) arteriole-specific SNP-gene interactions involving EC
   enriched genes (Fig. [131]4H), and 2) SNP-gene pairs shared between
   arteriole and EDA that involve genes enriched in non-EC cell types
   (Fig. [132]4I). Among the 1367 EC-enriched genes identified from
   arteriole snRNA-seq data (Bonferroni-adjusted p-value < 0.05 and |log2
   fold change | > 0.25), 102 genes were involved in SNP-gene
   interactions, with 26 of them having promoter interactions with
   enhancer-located SNPs. Notable examples include rs2836411-ERG and
   rs2244643-FBLN5 (Fig. [133]4H). ERG regulates angiogenesis and vascular
   stability by promoting VE-cadherin expression and Wnt/β-catenin
   signaling, ensuring endothelial integrity and survival^[134]16,[135]17.
   FBLN5 has been reported to be involved in the adaptive response of
   endothelial cells to hypoxia^[136]18. Both ERG and FBLN5 showed strong
   endothelial expression (Fig. [137]4J), and ERG was significantly more
   highly expressed in arterioles compared to DEA in bulk RNA-seq
   (FDR = 0.002). Among non-EC cell types, fibroblasts had the largest
   number of SNP-interacting genes enriched in this cell type, including
   top enhancer SNP-gene interactions such as rs2251636-SEMA4A and
   rs11725926-PDGFRA (Fig. [138]4I). SEMA4A has been reported to promote
   fibrosis across multiple disease contexts^[139]19,[140]20, while PDGFRA
   is involved in fibroblast-to-myofibroblast transition^[141]21.
   Together, these findings underscore the potential of chromatin
   interactions to mediate the cell type-specific regulatory effects of
   genomic variants on gene expression.

BP-associated SNP rs1882961 interacts with NRIP1 promoter in human arterioles
and regulates NRIP1 expression

   The above findings suggested that SNPs might influence
   arteriole-related phenotypes by regulating genes whose promoters
   interacted with the SNPs across long genomic distances. In a proof of
   principle study, we experimentally investigated a selected
   BP-associated SNP. There were 8790 pairs of arteriolar trait-associated
   SNPs and genes whose promoters interacted with the genomic regions
   containing these SNPs based on our arteriolar and EDA pan-promoter
   capture Micro-C dataset. We filtered these pairs based on both
   biological and technical criteria (Fig. [142]5A). The biological
   criteria included BP-associated SNPs, promoters of protein-coding
   genes, low promoter methylation levels (beta value < 0.2), and RNA
   abundance above a minimum threshold (average log[2](FPKM + 1) > 1).
   Technical criteria included not being in linkage disequilibrium with
   other SNPs for ease of genomic editing, cross-species conservation of
   the SNP genomic region to facilitate studies in animal models, and
   genes paired with only one SNP to avoid confounding effects from
   multiple variants. Furthermore, we focused on SNP-gene pairs for which
   the SNP is more than 100 kbp away from the gene as it is particularly
   challenging to identify the effector genes for these SNPs and
   understand the genomic mechanisms involved. These filters narrowed the
   list of SNP-gene pairs down to 89, from which we performed experimental
   studies on rs1882961 and NRIP1 based on additional considerations of
   GWAS significance, mapping specificity (single SNP-gene association),
   gene expression levels in snRNA-seq data, and known functions of the
   genes involved.

Fig. 5. BP-associated noncoding SNP rs1882961 interacts with NRIP1 promoter
119 kbp away and influences NRIP1 expression.

   [143]Fig. 5
   [144]Open in a new tab

   A Prioritization of SNP-gene pairs for experimental analysis. B The
   genomic region containing rs1882961 interacted with NRIP1 promoter 119
   kbp away from it in human endothelium-denuded arterioles, based on
   pan-promoter capture Micro-C analysis. C Local gene organization around
   rs1882961 in the human genome. D Deletion of DNA segment containing
   rs1882961. 39b, original iPSC cell line; X33 and X85, rs1882961-deleted
   cell lines. E Reconstitution of the rs1882961 locus containing either
   homozygous low BP allele or high BP allele. 39b, original iPSC cell
   line; X85, rs1882961-deleted cell line; Y series, cell lines with
   reconstituted low-BP rs1882961 allele; Z series, cell lines with
   reconstituted high-BP rs1882961 allele. See Supplementary Fig. [145]S8
   for sequence confirmation. F Homozygous high-BP allele of rs1882961,
   reconstituted in hiPSCs, increased NRIP1 and SAMSN1 expression in
   isogenic hiPSC-derived vascular smooth muscle cells (iVSMCs). Each line
   of hiPSCs was differentiated to iVSMCs in five independent rounds with
   three independent wells in each round, with each well considered a
   replicate. For SAMSN1 and NRIP1 expression, N = 13 for low BP and
   N = 15 for high BP; for USP25, N = 15 for low BP and N = 15 for high
   BP; Data are presented as Mean ± SEM. *, p = 0.0095 for SAMSN1 and
   p = 0.0371 for NRIP1 -, two-tailed unpaired t-test. G Region-capture
   Micro-C analysis showed greater chromatin interactions between the
   rs1882961 region and the promoters of SAMSN1 and NRIP1 in iVSMCs with
   the high-BP allele of rs1882961, compared to iVSMCs with the low-BP
   allele. Each square color box in the zoom-in images is 5 kbp x 5 kbp.

   rs1882961 is associated with systolic BP^[146]22. The genomic region
   containing rs1882961 interacted with NRIP1 promoter region in EDA
   (Fig. [147]5B). The transcription start site of NRIP1 is 119 kbp from
   rs1882961 (Fig. [148]5C). NRIP1 encodes nuclear receptor interacting
   protein 1, also known as receptor-interacting protein 140 (RIP14).
   RIP140 is a coregulatory for most nuclear receptors including retinoic
   acid receptor, estrogen receptor, and thyroid hormone receptor, many of
   which could influence vascular function^[149]23. However, the
   involvement of NRIP1 in blood pressure regulation and the role of
   rs1882961 in the regulation of NRIP1 are unknown.

   We experimentally examined the allelic effect of rs1882961 on the
   expression of NRIP1 and other local genes SAMSN1 and USP25 in human
   vascular smooth muscle cells (Fig. [150]5C). To that end, we generated
   isogenic hiPSCs precisely edited to contain homozygous rs1882961-C
   (low-BP allele) or rs1882961-T (high-BP allele) (Fig. [151]5D, E;
   Supplementary Figs. [152]S8, [153]S9). The edited hiPSCs were
   differentiated to vascular smooth muscle cells (iVSMCs) (Supplementary
   Fig. [154]S10A, B). The successful generation of iVSMCs was confirmed
   through the expression of VSMC marker genes, including Actin Alpha 2,
   Smooth Muscle (ACTA2), Calponin 1 (CNN1), and Smoothelin (SMTN)
   (Supplementary Fig. [155]S10C). In iVSMCs, NRIP1 and SAMSN1 were
   significantly upregulated in cells with the high-BP allele of rs1882961
   compared with the low-BP allele (Fig. [156]5F). USP25 was not
   differentially expressed. We performed region-capture Micro-C analysis
   to capture any genomic segment interacting with rs1882961 or promoters
   of its neighboring genes. The analysis revealed greater chromatin
   interactions between the rs1882961 genomic region and the promoter
   regions of NRIP1 and SAMSN1, but not USP25, in iVSMCs with the high-BP
   allele than the low-BP allele (Fig. [157]5G).

A 4-bp segment encompassing the rs1882961 orthologous site alters BP in SS
rats

   We tested the effect of the rs1882961 genomic site on BP and Nrip1
   expression in vivo. We mapped rs1882961 to position chr11:15,091,368 in
   the rat genome, which is 112 kbp from the TSS of Nrip1 (Fig. [158]6A;
   Supplementary Fig. [159]S11). A 500-bp region around the SNP site is
   conserved across vertebrates. For the 60 bp around the SNP site, only 5
   mismatches are present between the human and rat reference genomes
   (Supplementary Fig. [160]S11). We successfully deleted a 4-bp segment
   that encompassed the rs1882961 orthologous site from the genome of the
   Dahl salt-sensitive (SS) rat using CRISPR/Cas9 (Fig. [161]6A;
   Supplementary Fig. [162]S11). The mutant strain was designated
   SS-Δrs1882961. We phenotyped female rats as the hiPSCs used in the
   study above were derived from a female subject. Mean arterial pressure,
   systolic BP, diastolic BP, and pulse pressure, but not heart rate, in
   SS-Δrs1882961^−/− rats was significantly lower than wild-type
   littermate SS rats after the rats had been fed a high-salt (4% NaCl)
   diet (Fig. [163]6B–[164]F). The difference of mean arterial pressure
   occurred primarily in the dark phase (i.e., active phase) in a
   light-dark cycle, reaching approximately 15 mmHg during part of the
   dark phase (Fig. [165]6G).

Fig. 6. Deletion of a 4-bp noncoding genomic segment containing the rs1882961
orthologous site attenuates salt-induced hypertension in the Dahl
salt-sensitive (SS) rat.

   [166]Fig. 6
   [167]Open in a new tab

   A Comparative mapping and deletion of a 4-bp rat genomic segment
   containing the rs1882961 orthologous site. See Supplementary
   Fig. [168]S11 for additional detail and confirmation of deletion. The
   image at the top was generated using the UCSC Genome Browser,
   [169]http://genome.ucsc.edu. The deletion of the 4-bp noncoding genomic
   segment in the SS rat attenuated salt-induced hypertension (B–G). B
   Mean arterial pressure (MAP). C Systolic blood pressure (SBP). D
   Diastolic blood pressure (DBP). E Pulse pressure. F Heart rate. G
   Hourly averages of MAP on days 12 to 14 on a 4% NaCl high salt diet. WT
   wild-type littermate SS rats, Δrs961 SS-Δrs1882961^−/− rats. For B–G,
   n = 7 for WT and 8 for Δrs961. Data are presented as mean ± SEM. #,
   p < 0.05, ##, p < 0.01 for WT vs. Δrs961 by two-way RM ANOVA; *,
   p < 0.05 vs. WT by Holm-Sidak test. H Nrip1 expression in the mesentery
   was lower in SS-Δrs1882961^−/− rats compared to WT. In normal salt,
   n = 7 for WT and n = 9 for Δrs961; in high salt, n = 4 for WT and n = 5
   for Δrs961. Data are presented as Mean ± SEM. *, p < 0.05, Two-way
   ANOVA followed by Holm-Sidak test. I Vascular smooth muscle sensitivity
   to nitric oxide (NO) in the 3^rd order mesenteric artery was improved
   in SS-Δrs1882961^−/− rats compared to WT. Spermine NONOate, an NO
   donor. N = 6 for WT and 7 for Δrs961 rats. Data are presented as
   mean ± SEM. ##, p < 0.001 for WT vs. Δrs961; *, p < 0.001 for 10^-7 and
   10^-6 M and p = 0.019 for 10^-5 M; two-way RM ANOVA, followed by
   Holm-Sidak test.

   Nrip1 expression in mesenteric tissue was up-regulated in response to
   the high-salt diet but to a lesser extent in SS-Δrs1882961^−/− rats,
   resulting in significantly lower expression in SS-Δrs1882961^−/− rats
   than in wild-type SS rats on the high-salt diet (Fig. [170]6H).
   However, there was no change in Nrip1 expression in the aortic tissue
   of SS-Δrs1882961-/- rats compared to WT (Supplementary Fig. [171]S12).
   Vasodilation of the third-order mesenteric arteries in response to
   spermine NONOate, an nitric oxide (NO) donor, was significantly
   improved in SS- Δrs1882961^−/− rats compared with SS littermates
   (Fig. [172]6I), suggesting that vascular smooth muscle sensitivity to
   NO is improved by rs1882961 orthologous segment deletion.

   These studies demonstrate that a 4-bp segment of noncoding DNA
   encompassing the rs1882961 orthologous site regulates the expression of
   Nrip1 112 kbp away in vivo and influences vascular function and the
   development of salt-induced hypertension.

Discussion

   Long-range chromatin interactions play an important role in the
   regulation of gene expression^[173]24,[174]25. We have found that, in
   human arterioles, regulatory elements are prevalent in chromatin
   contact regions and the frequency of long-range chromatin interaction
   with gene promoters is highly significantly correlated with RNA
   abundance. These findings indicate that long-range chromatin
   interaction may be a here-to-fore unrecognized mechanism underlying the
   essential physiological function and broad pathophysiological relevance
   of arterioles. In addition, we found pericytes to be a prominent cell
   type in arterioles, and genes enriched in arteriolar pericytes
   significantly overlapped with genes linked to diastolic BP-associated
   genomic variants. These findings suggest that the functional
   significance of arteriolar pericytes warrants greater attention in
   future studies. We envision that our extensive dataset will drive
   forward mechanistic studies of hypertension, stroke, microvascular
   complications of diabetes and hypertension, and other arteriole-related
   diseases in new directions.

   Long-range chromatin interactions are especially relevant to
   understanding the mechanisms by which noncoding sequence variants
   identified by GWAS influence their associated traits as many GWAS
   variants are located 10 s of kbp from their nearest
   genes^[175]26,[176]27. The approach that we used in this study
   integrates in-depth chromatin conformation analysis in trait-relevant
   human tissues with analysis of GWAS-nominated SNPs, in vivo and in
   vitro genome editing, and functional investigation. The analysis
   revealed significant enrichment of SNPs associated with systolic BP,
   diastolic BP, pulse pressure, and stroke in chromatin contact regions
   in arteriolar tissues, supporting the relevance of arteriolar chromatin
   interactions in the genetic regulation of these specific traits.
   Further integration with snRNA-seq data highlighted the potential of
   arteriolar chromatin interactions to mediate the cell type-specific
   regulatory effects of vascular trait-associated genomic variants on
   gene expression. As demonstrated by our proof of principle study of
   rs1882961, this integrated approach is powerful and can be applied
   broadly to bridge the critical gap between genetic discoveries and
   physiological understanding.

   It is remarkable that deletion of a mere 4 bp at a genomic site that is
   more than 100 kbp from the transcription start site of the nearest
   protein-coding gene changes hourly BP by up to 15 mmHg. The high-salt
   dietary challenge and possibly the genomic background of the SS rat may
   have contributed to the uncovering of this remarkable BP effect. By
   extrapolation, the SNP rs1882961 and other GWAS-nominated sequence
   variants may have greater effects on BP and other associated traits in
   groups of humans with permissive genomic backgrounds and exposed to
   certain stressors than the typically minute effect reported by GWAS. In
   addition, such robust phenotypic changes, along with the changes in
   Nrip1 expression in the mutant rat and edited cells, not only reenforce
   the significance of long-range chromatin interaction in genetic and
   molecular regulation but also support its relevance to the regulation
   of physiological function. It is important to recognize that the robust
   BP effects were observed in the context of salt-induced hypertension in
   this study, and it remains to be investigated whether these effects are
   relevant to other types of hypertension.

   This study is one of the first to apply Micro-C and capture Micro-C
   techniques to analyze human tissues^[177]28,[178]29. The data was
   robust, considering the limited input material compared with studies
   using cultured cells. We performed our chromatin interaction analyses,
   as well as DNA methylation analysis, in bulk tissues. While techniques
   for analyzing chromatin conformation and DNA methylation in single
   nuclei have been reported^[179]30–[180]32, they require much more input
   materials than available from human arterioles, and data from these
   assays is sparser than from bulk tissue assays. Moreover,
   single-nucleus methods are not available for promoter capture Micro-C.
   Meanwhile, the robustness of our bulk tissue data is enhanced by the
   relative simplicity of cell type composition in arteriole compared to
   more complex organs. Additionally, we reduced the tissue complexity by
   analyzing separate tissue components in the current study.

   We analyzed arterioles isolated from subcutaneous fat. Pathological
   changes in subcutaneous vessel structure and physiology are
   well-established to reflect the influence of cardiovascular risk
   factors and prevalent cardiovascular disease^[181]33–[182]37.
   Physiological similarities and differences exist between arterioles in
   subcutaneous fat and those in visceral fat or internal
   organs^[183]34,[184]38,[185]39. It would be valuable to analyze
   arterioles in other tissue depots in future studies. Nevertheless, the
   reference molecular profiles generated in the current study will help
   to drive studies of arteriole-related traits and diseases in general as
   our study of rs1882961 has demonstrated. In addition, we chose to use
   subcutaneous fat because of the relative ease of obtaining such
   specimens from human participants, including via a minimally invasive
   biopsy in healthy individuals^[186]40. This means our data will be
   directly relevant to large-scale studies of human arterioles in which
   the ease of tissue procurement is particularly important.

Methods

   The human research protocols were approved by the Medical College of
   Wisconsin Institutional Review Board (IRB). The animal protocols were
   approved by the Institutional Animal Care and Use Committee at the
   Medical College of Wisconsin (AUA00000206, AUA00002214).

Human arterioles

   Human arterioles were obtained from discarded subcutaneous adipose
   tissue from surgery as described with modifications^[187]40. All
   protocols were approved by the Medical College of Wisconsin
   Institutional Review Board (IRB). Subject inclusion criteria were age
   18-89 and without major cardiovascular diseases except for
   hypertension. Exclusion criteria were coronary heart disease, heart
   failure, COVID19 infection within one-year, current tobacco use, and
   cancer chemotherapy drug exposure. The subjects involved in this study
   included a mix of age, race, sex, and hypertension status for the goal
   of generating reference omic profiles for subjects meeting the
   inclusion/exclusion criteria. Detailed information about the subjects
   is available in Supplementary Data [188]S1. Immediately upon receiving
   the discarded tissue, arterioles were isolated in cold (4 °C) HEPES
   buffer by removing the surrounding fat and connective tissue. These
   intact isolated arterioles were snap-frozen in liquid nitrogen and
   stored at −80 °C until use. For arterioles undergoing endothelial cell
   scraping, the isolated arterioles were cut open along their
   longitudinal direction with fine scissors and fixed lumen side-up with
   fine pins on a peri dish coated with silicone gel. After washing two
   times with PBS, the endothelial layer at the top of the lumen side was
   scrapped with a blade. After scraping, the blade was rinsed in an
   Eppendorf tube containing 700 µl of PBS; 800 µl of PBS was used to
   rinse the scrapped lumen side and collected into the same Eppendorf
   tube. The tube was spun at 300 g at 4 °C for 10 min. The pellet with
   100 µl of supernatant was snap-frozen in liquid nitrogen and stored at
   −80 °C until use. The vascular tissue that remained after the scraping
   was also collected as endothelium-denuded arteriole (EDA).

Micro-C assay and data analysis

   Micro-C assay^[189]29 was performed using Dovetail® Micro-C Kit
   following manufacturer’s User Guide for Tissue and Blood to create
   Micro-C libraries. Briefly, tissues were ground into fine powder with
   CryoGrinder^TM. The ground tissue was crosslinked with disuccinimidyl
   glutarate and formaldehyde sequentially. The genomic DNA was digested
   in situ with MNase Enzyme mix to 20%–40% mononucleosomes. Genomic DNA
   was released from cells and put through a sequential process of binding
   to chromatin capture beads, end polishing, bridge ligation,
   intra-aggregate ligation, crosslink reversal, and DNA purification.
   Library was made from the purified DNA and subject to sequencing with
   Novaseq 6000 platform for approximately 800 M 150 PE reads.

   Pair-end sequencing reads were cleaned with Trim Galore (v0.6.2) to
   remove adaptor and low-quality reads. We used JuicerBox, which was a
   one-click pipeline for processing HiC or Micro-C data, to generate hic
   file for direct visualization^[190]41. We followed Dovetail Micro-C
   Data Processing Guide to generate.mcool file and used it for loop
   calling with Mustache^[191]42.

Pan-promoter capture Micro-C assay and data analysis

   Pan-promoter capture Micro-C library was created using the Micro-C
   libraries generated as described above, Dovetail® human Pan Promoter
   Enrichment panel, and Dovetail Target Enrichment kit. The panel
   included 161,144 probes designed to cover 84,643 promoters for 19725
   protein-coding genes and 7630 long noncoding genes in human based on
   Ensembl GRCh38 version 86. The libraries were sequenced using the
   Illumina NovaSeq sequencers.

   The data were analyzed following Dovetail Promoter Panel Data
   Processing Guide to generate mapped.PT.bam file. Direct chromatin
   interactions were then identified by CHiCAGO at different resolutions
   with the mapped.PT.bam file^[192]43. To ensure our dataset aligned with
   the most current promoter definitions, we updated the transcript
   promoter annotations using the latest Ensembl GRCh38 version 112.
   Following this update, we retained 99% of the original chromatin
   interactions that captured promoter regions (Supplementary
   Data [193]S10). These retained interactions were used in the subsequent
   analysis. Genomic overlap analysis was performed using the
   GenomicRanges R package (v1.50.2). Regulatory element annotations were
   downloaded from Ensembl. To maintain consistency with the promoter
   definition used in Dovetail, promoter regions in the pan-promoter
   Capture Micro-C data were defined as 1000 bp upstream and 500 bp
   downstream of transcript TSSs.

Hi-C data source and access

   Pre-processed Hi-C loop data for the aorta and tibial artery were
   obtained from the ENCODE consortium. The data were accessed in the form
   of bedpe files from the following ENCODE experiment accessions:
   ENCSR797MWY, ENCSR554SLA (aorta), and ENCSR601RFS (tibial artery).

Poly(A)-dependent RNA-seq and data analysis

   Poly(A)-dependent RNA-seq experiment and data analysis were performed
   as described^[194]40. An in-house pipeline was used to analyze RNA-seq
   data. Briefly, raw RNA-Seq sequence reads were pre-processed using Trim
   Galore (v0.6.2)
   ([195]http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/).
   Adapters and reads with low quality (base quality < 20) were removed
   prior to further analysis. The trimmed sequence reads longer than 20 bp
   were mapped to the human genome (hg38) using hisat2 (v2.2.1)^[196]44.
   Transcript assembly and quantification were performed using StringTie2
   (v2.2.1)^[197]45, based on the GENCODE reference annotation (v38). R
   package DESeq2 was used to identify differentially expressed genes
   (DEGs; FDR < 0.05), accounting for inter-subject variability (modeled
   as ~ Subject + Tissue)^[198]46.

Reduced representation bisulfite sequencing (RRBS) and data analysis

   DNA methylation profiles at single-base resolution were analyzing using
   multiplexed RRBS as we described^[199]12,[200]13,[201]47. Trimmed
   sequences were mapped to the human reference genome (hg38) using
   Bismark (v0.16.1)^[202]48. To account for inter-subject variability,
   modified metilene (Version 0.23) with paired Wilcoxon signed-rank test
   was applied to identify differentially methylated regions (DMRs) de
   novo^[203]49. Methylation rate of each region was calculated as the
   average methylation rate of each site across the region. Differentially
   methylated regions (DMR) were identified based on the following
   criteria: a minimum mean methylation difference between groups of 0.05,
   a minimum of three CpGs within the identified regions, and an
   FDR < 0.05 from the Wilcoxon signed-rank test.

Single-nucleus RNA-seq and data analysis

   Human arterioles were broken up using a beads method, and nuclei were
   extracted using the ‘Frankenstein’ protocol. Briefly, the arterioles
   were broken up in 0.5 ml Nuclei EZ Lysis buffer with Precellys
   Evolution Touch and tissue homogenizing ceramic beads at a setting of
   6500 rpm for 20 sec. Another 3.5 ml Nuclei EZ Lysis buffer was added to
   tissue homogenized buffer and incubated for 5 min on ice before being
   filtered through a 70 µm strainer mesh. The extracted nuclei were
   washed once with Nuclei Wash and Suspension Buffer. Single-nucleus
   RNA-seq libraries were prepared using Chromium Single Cell 3’ Reagent
   Kits V3 (10x Genomics, Pleasanton, CA), following the manufacturer’s
   user guide. snRNA-seq libraries were sequenced using the Illumina
   NovaSeq sequencers to generate approximately 30 GB or 100 million
   paired reads of data per library.

   Raw fastq files were processed using CellRanger (v7.0.0) to generate
   feature matrices using 10x Genomics reference genome assembly hg38
   (refdata-gex-GRCh38-2020-A). DoubletFinder R package (v2.0.3) was used
   to remove potential doublets, with the doublet rate set to max(0.3,
   #cells*8*1e-6)^[204]50. Cells with less than 200 genes or a
   mitochondrial percentage larger than 20% were excluded from the
   analysis.

   For quality-controlled snRNA-seq data, the gene count matrix was
   log-transformed and scaled. The top 2000 most variable genes were
   identified using the FindVariableFeatures function from Seurat R
   package (v4.3.0.1)^[205]51, and then were used for principal component
   analysis (PCA). Clustering was performed using the top 20 principal
   components and the Louvain algorithm, with resolution adjusted from 0.5
   to 2 in increments of 0.5. The optimal resolution was selected based on
   Clustree (v0.5.1) analysis^[206]52. To annotate identified clusters,
   differentially expressed genes (DEGs) between clusters were identified
   by Wilcoxon rank sum test using FindAllMarkers function
   (logfc.threshold = 0.25, min.pct = 0.25, only.pos = TRUE). Statistical
   significance was determined as a Bonferroni-adjusted p-value < 0.05 and
   an absolute log2 fold change >0.25. Canonical markers from PanglaoDB
   and CellMarker were referenced to annotate major cell
   types^[207]53,[208]54.

   Preprocessed snRNA-seq data for the thoracic aorta was downloaded from
   [209]https://singlecell.broadinstitute.org/single_cell/study/SCP1265,
   and cell clusters and annotations from the original study were
   applied^[210]11.

   To compare cell type expression similarity between arterioles and the
   thoracic aorta, pseudo-bulk RNA-seq profiles were generated for each
   cell type within each dataset using AverageExpression function. DEGs
   across cell types were calculated using the FindAllMarkers function,
   with significance determined by a Bonferroni-adjusted p-value < 0.05.
   Common DEGs across datasets were then used for pairwise Spearman
   correlation analysis to compare cell types between the two datasets
   based on their pseudo-bulk expression profiles.

MAGMA cell typing

   Preprocessed study-level GWAS summary statistics for blood pressure
   regulation and hypertension end-organ damage traits were downloaded
   from the GWAS Catalog ([211]https://www.ebi.ac.uk/gwas/). This included
   systolic and diastolic blood pressure from studies by Keaton JM et
   al.^[212]55, Surendran P et al.^[213]56, and Wojcik GL et al.^[214]57;
   pulse pressure from Keaton JM et al.^[215]55 and Surendran P et
   al.^[216]56; essential hypertension from Wojcik GL et al.^[217]57;
   stroke from Malik R et al.^[218]58 and Mishra A et al.^[219]59;
   peripheral arterial disease from Sakaue S et al.^[220]60; diabetic
   nephropathy from Sakaue S et al.^[221]60 and Zhou W et al.^[222]61; and
   diabetic retinopathy from Backman JD et al.^[223]62 and Zhou W et
   al.^[224]61. Cell type-trait association analysis was performed using
   the celltype_associations_pipeline from the MAGMA.Celltyping R package
   (v2.0.11)^[225]63. The top 10% mode was used to focus on the most
   enriched genes, as it also has a stronger ability to identify
   associations and aligns more closely with other gene-trait association
   methods^[226]64. Significance was determined using a false discovery
   rate (FDR) of <0.05.

Integrated data analysis

   GWAS summary statistics for arteriole-related traits were downloaded
   from the GWAS Catalog ([227]https://www.ebi.ac.uk/gwas/), including
   systolic BP, diastolic BP, pulse pressure, essential hypertension,
   stroke, peripheral arterial disease, diabetic nephropathy, and diabetic
   retinopathy. Only SNPs with a p-value < 5e-8 were retained for further
   analysis. To assess whether the percentage of SNPs covered by Micro-C
   or pan-promoter Capture Micro-C in arterioles and EDAs was
   significantly higher in arteriolar traits, we selected traits not
   directly or primarily related to the arteriole and calculated their SNP
   coverage percentages. These traits included birth weight, bone density,
   breast cancer, colorectal cancer, Alzheimer’s disease, Parkinson’s
   disease, multiple sclerosis, rheumatoid arthritis, asthma, psoriasis,
   and chronic obstructive pulmonary disease. A one-sided binomial test
   was applied to compare SNP coverage for each arteriolar trait against
   the average percentage across non-arteriole-relevant traits.

   The ClusterProfiler R package (v4.6.2) was used for pathway enrichment
   analysis on genes associated with arteriole-related SNPs through
   chromatin interactions captured by pan-promoter Capture
   Micro-C^[228]65. Low-abundance genes, defined as those with an average
   FPKM value less than 0.1 in the bulk RNA-seq data, were excluded from
   the analysis.

   Following the same pipeline as before, SNP regulatory annotation was
   performed using Ensembl annotation^[229]14. The coverage of SNPs in
   each regulatory category by promoter Capture Micro-C was calculated
   separately for promoter bins and distal bins. A Fisher’s exact test was
   applied to assess whether SNPs were enriched in specific regulatory
   categories based on their bin location.

   Visualization was performed using ggplot2^[230]66.

SNP editing in hiPSCs

   The generation of isogenic hiPSCs with homozygous low-BP and high-BP
   alleles for the SNP rs1882961 was achieved using an efficient two-step
   approach as we described^[231]26. Briefly, 3 µM each for two synthetic
   sgRNAs (Synthego) targeting genomic regions flanking rs1882961
   (Supplementary Data [232]S16) and 3 µM spCas9-2NLS nuclease protein
   (Synthego) with Lipofectamine 2000 transfection reagent (Invitrogen)
   were used to delete a segment of DNA around rs1882961. After 48 hrs of
   transfection, cells underwent clonal expansion in 96-well plates using
   Hana Cell Sorter and Single Cell Dispenser (Bio-Techne). Following
   single cell seeding, single cell clones were further grown for 10-14
   days and characterized for the deletion of SNP region using quick
   extraction of DNA and PCR (Supplementary Data [233]S16). hiPSC clones
   showing the intended deletion based on quick extraction of DNA and PCR
   were further characterized for the deletion by PCR and sanger
   sequencing. hiPSC line with the deletion of the region around rs1882961
   was tested for their genetic stability by molecular karyotyping using
   hPSC Genetic Analysis Kit (STEMCELL). For the second step of 2-step SNP
   editing, homology dependent knock-in approach was used with a gRNA
   targeting the junction of rs1882961 region deletion site and ssDNA
   donor fragment for low-BP rs1882961-C or high-BP rs1882961-T alleles
   with 100-200 bp long homology arms at the 3’ and 5’ sides of rs1882961
   (Supplementary Data [234]S16). The 500 ng ssDNA donor for either the
   low-BP rs1882961-C or high-BP rs1882961-T alleles, along with 3 µM each
   for two sgRNAs targeting junction region and 3 µM spCas9-2NLS Nuclease
   protein, were transfected into the hiPSC line with the selected clone
   for rs1882961 SNP region deletion from the first step using
   Lipofectamine 2000 transfection reagent. Single-cell clonal expansion
   and initial screening to identify isogeneic hiPSC lines were then
   performed as described in the first step. Isogenic hiPSC lines with the
   reconstitution of homozygous low-BP or high-BP alleles were selected,
   confirmed, and assessed for their pluripotency, differentiation
   potential, and genetic stability by molecular karyotyping using hPSC
   Genetic Analysis Kit and Giemsa-based karyotyping.

Differentiation of hiPSCs to vascular smooth muscle cells (iVSMCs)

   The differentiation was performed as we
   described^[235]26,[236]67,[237]68. Briefly, hiPSCs were treated with
   mTeSR™ plus medium with 10 µM Rock inhibitor Y-27632 (STEMCELL
   Technologies) for 24 h and then N2B27 medium (Life Technologies) plus
   8 µM CHIR99021 (Selleck Chemicals) and 25 ng/ml BMP4 (PeproTech) for 3
   days to generate mesoderm cells. For VSMC induction, mesoderm cells
   were grown with N2B27 medium supplemented with 10 ng/ml PDGF-BB
   (PeproTech) and 2 ng/ml Activin A (PeproTech) for 2 days, and N2B27
   supplemented with 2 ng/ml Activin A and 2 µg/ml Heparin (STEMCELL
   Technologies) for 5 days. VSMCs were enriched by removing CD144 + cells
   using CD144 magnetic beads. Enriched VSMC cells were further cultured
   in N2B27 medium supplemented with 10 ng/ml PDGF-BB, 2 ng/ml Activin A
   and 1 µM [238]PD032590 (Selleck Chemicals) for 6 days for contractile
   VSMC formation.

Real-time PCR

   RNA extraction and quantitative real-time PCR analysis were performed
   as described^[239]69. Briefly cells were harvested in Trizol reagent
   (Life Technologies) for total RNA isolation. Approximately 1 µg of
   total RNA was converted to cDNA using random hexamer primers and
   RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific).
   Quantitative PCR (qPCR) was performed on QuantStudio6 Real-Time PCR
   System (Applied Biosystems) using SYBR Green Master Mix (Applied
   Biosystems) and primers for neighboring genes. Primer sequences are
   shown in Supplementary Data [240]S16. Expression levels of mRNAs were
   normalized to the endogenous control 18S using ΔΔCt method. Statistical
   analyses were performed utilizing GraphPad Prism 9 (Domatics).

Region-capture Micro-C and data analysis

   The assay and data analysis were performed as described with some
   modifications^[241]26,[242]28,[243]70. Briefly, Dovetail® Micro-C kit
   was used to create Micro-C libraries following manufacturer’s User
   Guide. Region capture Micro-C libraries were then generated with
   reagents from Dovetail® targeted enrichment panels. The panel of 80 bp
   probes covers the rs1882961 genomic region and promoters of adjacent
   protein coding genes. The libraries were sequenced using the Illumina
   NovaSeq sequencers. Pair-end sequencing reads were cleaned with Trim
   Galore! to remove adaptor and low-quality reads. We used Juicer to
   generate.hic file and Juicebox for direct visualization^[244]41.

Animals

   The Institutional Animal Care and Use Committee (IACUC) at Medical
   College of Wisconsin approved all animal related protocols
   (AUA00000206, AUA00002214). The Dahl salt-sensitive rat (SS/JrHsdMcwi)
   was used to generate SS-Δrs1882961^−/− strain. Because the hiPSCs used
   in the current study were derived from a female subject, female
   SS-Δrs1882961^−/− rats and wild type littermates were phenotyped
   accordingly. Animals were housed in a temperature-controlled
   environment with a 12-h light/dark cycle. Rats had free access to water
   and a custom AIN-76A purified rodent chow containing 0.4% NaCl (Dyets,
   Inc) unless otherwise specified.

Generation of SS-Δrs1882961^−/− rat

   CRISPR/SpCas9 methods were used to delete a noncoding genomic segment
   in the Dahl salt-sensitive rat (SS/JrHsdMcwi). The LiftOver tool at
   UCSC Genome Browser placed the non-coding syntenic region to the human
   rs1882961 at mRatBN7.2 chr11:15,091,368. SpCas9 (QB3 MacroLab,
   University of California, Berkely) was mixed with a sgRNA targeting the
   sequence GATGGCCTGCCAGGACAGAGTGG (protospacer adjacent motif underlined
   and target base in bold) and injected into fertilized SS/JrHsdMcwi
   strain rat embryos. A founder was identified harboring a 4-bp deletion
   (mRatBN7.2 chr11:15,091,367-15,091,370 and was backcrossed to the
   parental strain to establish a breeding colony. This SS-Δrs1882961^−/−
   strain is registered as SS-Del(11p)3Mcwi (RGDID: 407572511).
   Heterozygous SS-Δrs1882961^+/- breeders were maintained on a purified
   0.4% NaCl diet AIN-76A (Dyets, Inc).

Telemetry measurement of BP

   Continuous measurement of BP in conscious, freely moving female rats
   (n = 7 for WT and 8 for Δrs961) was performed using radiotelemetry as
   described^[245]71,[246]72. At 7 weeks of age, rats were anesthetized
   with a mixture of ketamine (75 mg/kg), xylazine (10 mg/kg), and
   acepromazine (2.5 mg/kg) intramuscular injection, and a telemetry
   transmitter (model HD-S10, Data Systems International) was implanted in
   the right carotid artery to measure arterial blood pressure. After one
   week of recovery, continuous BP monitoring was started. Following three
   days of stable baseline BP recording, the diet was switched to an
   AIN-76A diet containing 4% NaCl (high salt diet) for 14 days. The
   animal experiments were performed in accordance with the
   recommendations in the Guide for the Care and Use of Laboratory Animals
   of the National Institutes of Health. The animal protocols were
   approved by the Institutional Animal Care and Use Committee at the
   Medical College of Wisconsin (AUA00000206, AUA00002214).

Vasoreactivity analysis

   Vascular function analysis was performed as described^[247]40,[248]73.
   Immediately after euthanasia, third order mesenteric arteries from
   female rats (n = 6 for WT and 7 for Δrs961) on an AIN-76A diet
   containing 4% NaCl for 14 days were dissected and used in the current
   study. Isolated artery segments were transferred to an organ bath
   filled with physiological salt solution (PSS) consisting of 119 mM
   NaCl, 4.7 mM KCl, 1.6 mM CaCl[2], 0.03 mM EDTA, 1.17 mM MgSO[4], 24 mM
   NaHCO[3], 1.18 mM KH[2]PO[4], and 5 mM glucose and mounted between two
   glass cannulas in a culture myograph chamber (DMT, 204CM). The myograph
   chamber was transferred to a stage of an inverted microscope attached
   to a video camera, video monitor, and video-measuring device (model
   VIA-100; Boeckeler). Arterioles were in PSS buffer bubbled with a gas
   mixture of 21% O[2]-5% CO[2]-74% N[2] at 37 °C. The arteriole was
   allowed to equilibrate at an intraluminal pressure of 30 mmHg supplied
   by a reservoir system for 30 min and followed by 60 mmHg for another
   30 min. After equilibration, U46619 (Catalog No. 16450, Cayman
   Chemical) was used to pre-constrict the arteriole to 50% to 70% of
   maximum diameter, followed by administration of 0.1 nM -10 μM spermine
   NONOate (Catalog No. 82150, Cayman Chemical), a nitric oxide (NO)
   donor, to construct cumulative concentration-response curves.
   Vasodilatory response to NONOate is reported as % of maximum dilation,
   which was defined as (vessel diameter – pre-constricted
   diameter)/(maximum diameter – preconstricted diameter), where maximum
   diameter was the resting diameter or the diameter following the
   addition of papaverine, whichever was greater.

Supplementary information

   [249]Supplementary information^ (2.4MB, pdf)
   [250]Supplementary Tables^ (20.9MB, xlsx)
   [251]Peer Review file^ (503.3KB, pdf)
   [252]Supplementary Data^ (47.4KB, pdf)

Source data

   [253]Source Data^ (24.5MB, xlsx)

Acknowledgements