Advertisement

Trans-interaction of risk loci 6p24.1 and 10q11.21 is associated with endothelial damage in coronary artery disease

Open AccessPublished:October 22, 2022DOI:https://doi.org/10.1016/j.atherosclerosis.2022.10.012

      Highlights

      • 6p24.1 locus containing rs6903956 is related to endothelial activation pathways.
      • Chromosomes 6p24.1 and 10q11.21 are spatially organized in endothelial cells.
      • 6p24.1 is in close genomic proximity to a weak promoter of CXCL12 that resides on 10q11.21
      • Risk allele ‘A’ of rs6903956 is associated with endothelial injury in CAD patients.

      Abstract

      Background and aims

      Single nucleotide polymorphism rs6903956 has been identified as one of the genetic risk factors for coronary artery disease (CAD). However, rs6903956 lies in a non-coding locus on chromosome 6p24.1. We aim to interrogate the molecular basis of 6p24.1 containing rs6903956 risk alleles in endothelial disease biology.

      Methods and Results

      We generated induced pluripotent stem cells (iPSCs) from CAD patients (AA risk genotype at rs6903956) and non-CAD subjects (GG non-risk genotype at rs6903956). CRISPR-Cas9-based deletions (Δ63-89bp) on 6p24.1, including both rs6903956 and a short tandem repeat variant rs140361069 in linkage disequilibrium, were performed to generate isogenic iPSC-derived endothelial cells. Edited CAD endothelial cells, with removal of ‘A’ risk alleles, exhibited a global transcriptional downregulation of pathways relating to abnormal vascular physiology and activated endothelial processes. A CXC chemokine ligand on chromosome 10q11.21, CXCL12, was uncovered as a potential effector gene in CAD endothelial cells. Underlying this effect was the preferential inter-chromosomal interaction of 6p24.1 risk locus to a weak promoter of CXCL12, confirmed by chromatin conformation capture assays on our iPSC-derived endothelial cells. Functionally, risk genotypes AA/AG at rs6903956 were associated significantly with elevated levels of circulating damaged endothelial cells in CAD patients. Circulating endothelial cells isolated from patients with risk genotypes AA/AG were also found to have 10 folds higher CXCL12 transcript copies/cell than those with non-risk genotype GG.

      Conclusions

      Our study reveals the trans-acting impact of 6p24.1 with another CAD locus on 10q11.21 and is associated with intensified endothelial injury.

      Graphical abstract

      Keywords

      1. Introduction

      Cardiovascular disease is the leading cause of mortality globally, with coronary artery disease (CAD) accounting for a majority of cardiovascular deaths [
      OECDOrganization WH
      Mortality from Cardiovascular Disease.
      ]. With the heritability of CAD estimated at up to 60% [
      • Zdravkovic S.
      • Wienke A.
      • Pedersen N.L.
      • Marenberg M.E.
      • Yashin A.I.
      • De Faire U.
      Heritability of death from coronary heart disease: a 36-year follow-up of 20 966 Swedish twins.
      ], genetic factors hold an important contribution to the risk of CAD along with other major etiologic determinants such as lifestyle and environmental factors.
      Genome-wide association studies (GWAS) have identified several common variants to be risk factors for CAD, but few have been functionally validated [
      • Schunkert H.
      • Konig I.R.
      • Kathiresan S.
      • Reilly M.P.
      • Assimes T.L.
      • Holm H.
      • et al.
      Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease.
      ,
      • Samani N.J.
      • Erdmann J.
      • Hall A.S.
      • Hengstenberg C.
      • Mangino M.
      • Mayer B.
      • et al.
      Genomewide association analysis of coronary artery disease.
      ,
      • Nikpay M.
      • Goel A.
      • Won H.H.
      • Hall L.M.
      • Willenborg C.
      • Kanoni S.
      • et al.
      A comprehensive 1,000 Genomes-based genome-wide association meta-analysis of coronary artery disease.
      ]. A single nucleotide polymorphism (SNP) rs6903956 on chromosome 6p24.1 was first reported as a susceptibility locus for CAD (adjusted odds ratio = 1.65, 95% CI 1.44–1.90, p = 2.55 × 10−13, Supplemental Table S1) in a GWAS on large cohort Chinese population [
      • Wang F.
      • Xu C.Q.
      • He Q.
      • Cai J.P.
      • Li X.C.
      • Wang D.
      • et al.
      Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
      ]. The minor allele at rs6903956 (A, 9.07% frequency) leads to increased risk of coronary atherosclerosis (Supplemental Table S1) [
      • Guo C.Y.
      • Gu Y.
      • Li L.
      • Jia E.Z.
      • Li C.J.
      • Wang L.S.
      • et al.
      Association of SNP rs6903956 on chromosome 6p24.1 with angiographical characteristics of coronary atherosclerosis in a Chinese population.
      ]. This finding has been subsequently corroborated by several independent studies on Asian populations, including Singaporean and Japanese individuals [
      • Guo C.Y.
      • Gu Y.
      • Li L.
      • Jia E.Z.
      • Li C.J.
      • Wang L.S.
      • et al.
      Association of SNP rs6903956 on chromosome 6p24.1 with angiographical characteristics of coronary atherosclerosis in a Chinese population.
      ,
      • Meng J.
      • Tan W.
      • Zhu Y.
      • Wang F.
      • Li X.
      • Zhang M.
      A coronary artery disease-associated SNP rs6903956 contributed to asymptomatic hyperuricemia susceptibility in Han Chinese.
      ,
      • Tayebi N.
      • Ke T.
      • Foo J.N.
      • Friedlander Y.
      • Liu J.
      • Heng C.K.
      Association of single nucleotide polymorphism rs6903956 on chromosome 6p24.1 with coronary artery disease and lipid levels in different ethnic groups of the Singaporean population.
      ,
      • Dechamethakun S.
      • Ikeda S.
      • Arai T.
      • Sato N.
      • Sawabe M.
      • Muramatsu M.
      Associations between the CDKN2A/B, ADTRP and PDGFD polymorphisms and the development of coronary atherosclerosis in Japanese patients.
      ]. In our Singaporean population, we leveraged on the ‘Singapore Coronary Artery Disease Genetics Study’ where CAD and normal control subjects from the National University Hospital Angiography Centre had been genotyped for a CAD GWAS [
      • Han Y.
      • Dorajoo R.
      • Chang X.
      • Wang L.
      • Khor C.C.
      • Sim X.
      • et al.
      Genome-wide association study identifies a missense variant at APOA5 for coronary artery disease in Multi-Ethnic Cohorts from Southeast Asia.
      ]. The rs6903956 variant was one of the constituent SNPs being interrogated in the genotyping array. We confirm that heterozygous AG at rs6903956 is significantly associated with CAD in Chinese (adjusted odds ratio = 2.26, 95% CI 1.09–4.6, p = 0.028), and not in the other ethnic groups when analyzed independently [
      • Tayebi N.
      • Ke T.
      • Foo J.N.
      • Friedlander Y.
      • Liu J.
      • Heng C.K.
      Association of single nucleotide polymorphism rs6903956 on chromosome 6p24.1 with coronary artery disease and lipid levels in different ethnic groups of the Singaporean population.
      ].
      SNP rs6903956 lies within the first intron of androgen-dependent tissue factor pathway inhibitor regulating protein (ADTRP). Its minor risk allele A is associated with decreased ADTRP mRNA expression in leukocytes [
      • Wang F.
      • Xu C.Q.
      • He Q.
      • Cai J.P.
      • Li X.C.
      • Wang D.
      • et al.
      Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
      ]. Plasma levels of ADTRP are significantly reduced in CAD patients compared to control subjects [
      • Ooi D.S.Q.
      • Ong S.M.
      • Eng M.H.
      • Chan Y.H.
      • Lee Y.S.
      • Low A.F.H.
      • et al.
      Detection of ADTRP in circulation and its role as a novel biomarker for coronary artery disease.
      ]. As risk variants identified by GWAS located in intronic regions of the genome are generally believed to confer risk of disease via disruption of gene regulatory elements [
      • Gallagher M.D.
      • Chen-Plotkin A.S.
      The post-GWAS era: from association to function.
      ], a recent study found that binding of GATA2 with a 519bp region containing rs6903956 non-risk allele G enhances ADTRP expression level in HeLa cells, suggesting a putative enhancer role of this region [
      • Luo C.
      • Tang B.
      • Qin S.
      • Yuan C.
      • Du Y.
      • Yang J.
      GATA2 regulates the CAD susceptibility gene ADTRP rs6903956 through preferential interaction with the G allele.
      ]. Genetic effects are often restricted to trait relevant cell types, making it important to have precise disease-specific tissue and cell types for expression quantitative trait loci (eQTL) mapping [
      • Torres J.M.
      • Gamazon E.R.
      • Parra E.J.
      • Below J.E.
      • Valladares-Salgado A.
      • Wacher N.
      • et al.
      Cross-tissue and tissue-specific eQTLs: partitioning the heritability of a complex trait.
      ,
      • Wang J Eric
      • Brandon Barbara
      • Hae Robert
      • et al.
      Imputing gene expression in uncollected tissues within and beyond GTEx.
      ]. However, there remains a lack of eQTL analysis of rs6903956 on relevant tissues from CAD patients.
      CAD is largely caused by atherosclerosis, which is a build-up of plaque inside the artery walls. Multiple cell types are involved in CAD pathogenesis, including endothelial cells, vascular smooth muscle cells, macrophages, and cells of the adaptive immune system [
      • Libby P.
      • Buring J.E.
      • Badimon L.
      • Hansson G.K.
      • Deanfield J.
      • Bittencourt M.S.
      • et al.
      ]. Functional causality of genetic risk variants could be better resolved with a focus on cell-type specific effects. For instance, the well-studied CAD risk locus 9p21.3 mediates its effect on ANRIL expression, and induced proinflammatory responses in endothelial cells, vascular smooth muscle cells and mononuclear cells [
      • Congrains A.
      • Kamide K.
      • Oguro R.
      • Yasuda O.
      • Miyata K.
      • Yamamoto E.
      • et al.
      Genetic variants at the 9p21 locus contribute to atherosclerosis through modulation of ANRIL and CDKN2A/B.
      ,
      • Zhou X.
      • Han X.
      • Wittfeldt A.
      • Sun J.
      • Liu C.
      • Wang X.
      • et al.
      Long non-coding RNA ANRIL regulates inflammatory responses as a novel component of NF-kappaB pathway.
      ,
      • Holdt L.M.
      • Hoffmann S.
      • Sass K.
      • Langenberger D.
      • Scholz M.
      • Krohn K.
      • et al.
      Alu elements in ANRIL non-coding RNA at chromosome 9p21 modulate atherogenic cell functions through trans-regulation of gene networks.
      ,
      • Lo Sardo V.
      • Chubukov P.
      • Ferguson W.
      • Kumar A.
      • Teng E.L.
      • Duran M.
      • et al.
      Unveiling the role of the most impactful cardiovascular risk locus through haplotype editing.
      ]. A common noncoding SNP rs17114036 at 1p32.2 is located in an endothelial-specific enhancer and regulates endothelial mechanotransduction mechanisms [
      • Krause M.D.
      • Huang R.T.
      • Wu D.
      • Shentu T.P.
      • Harrison D.L.
      • Whalen M.B.
      • et al.
      Genetic variant at coronary artery disease and ischemic stroke locus 1p32.2 regulates endothelial responses to hemodynamics.
      ]. Also, within our 6p24 locus of interest, another SNP rs9349379 in the third intron of gene encoding phosphatase and actin regulatory protein 1 (PHACTR1) demonstrates cis-regulation of endothelial expression of endothelin-1 (EDN1), a gene located 600 kb upstream of PHACTR1 [
      • Gupta R.M.
      • Hadaya J.
      • Trehan A.
      • Zekavat S.M.
      • Roselli C.
      • Klarin D.
      • et al.
      A genetic variant associated with five vascular diseases is a distal regulator of endothelin-1 gene expression.
      ]. It has been reported that ADTRP confers anti-coagulant protection in endothelial cells through regulation of tissue factor pathway inhibitor [
      • Lupu C.
      • Zhu H.
      • Popescu N.I.
      • Wren J.D.
      • Lupu F.
      Novel protein ADTRP regulates TFPI expression and function in human endothelial cells in normal conditions and in response to androgen.
      ]. Here, we prioritized endothelial disease biology in interrogating the functional basis of 6p24 locus that contains rs6903956.
      Given the challenges of obtaining patient coronary arteries and the tendency of primary endothelial cells to senesce in vitro, we leveraged induced pluripotent stem cells (iPSC) technology to generate human iPSC-derived endothelial cells. In analyzing nearby variants of rs6903956, a short tandem repeat variant rs140361069 was found to be the nearest variant in linkage disequilibrium (LD) with rs6903956 in East Asian populations. Hence, we employed CRISPR-Cas9 genome editing to delete small regions (63 - 89bp) covering both rs6903956 and rs140361069 at 6p24.1 in CAD patient iPSCs and control subject iPSCs. Using protocols that we had previously established [
      • Cheung C.
      • Bernardo A.S.
      • Trotter M.W.
      • Pedersen R.A.
      • Sinha S.
      Generation of human vascular smooth muscle subtypes provides insight into embryological origin-dependent disease susceptibility.
      ,
      • Narmada B.C.
      • Goh Y.T.
      • Li H.
      • Sinha S.
      • Yu H.
      • Cheung C.
      Human stem cell-derived endothelial-hepatic Platform for efficacy testing of vascular-protective metabolites from nutraceuticals.
      ], isogenic iPSC lines were differentiated using chemically defined factors to obtain consistent cultures of edited and unedited iPSC-derived endothelial cells. To decipher the cellular impact of 6p24.1 locus, we applied transcriptomic and epigenomic approaches on iPSC-derived endothelial cells to uncover potential genes regulated by the variants.

      2. Materials and methods

      2.1 Study approvals and subject enrolment

      This study was approved by the Local Ethics Committee of Nanyang Technological University Singapore Institutional Review Board (IRB18/09/02 and IRB-2020-09-011), National Healthcare Group (DSRB: 2013/00937) and Agency for Science, Technology and Research (IRB Reference 2020-096), Singapore. This research complies with the Helsinki Declaration. Written informed consent was obtained from each participant after the nature and possible consequences of the studies have been explained.
      We leveraged on the ‘Singapore Coronary Artery Disease Genetics Study’ where CAD patients from the National University Hospital angiography center had been genotyped for rs6903956 alleles. CAD patients were diagnosed as non-ST-elevation myocardial infarction (NSTEMI) by angiography, while normal control subjects were based on self-declared medical history. Demographics of recruited subjects from whom samples were used for experimentation can be found in Supplemental Table S3.

      2.2 Peripheral blood mononuclear cell isolation and culture

      For sample collection, 10 ml of fresh blood was collected from each subject in heparin vacutainers and processed in the laboratory within 6 h. Upon centrifugation of blood using Ficoll-PaquePremium (GE Healthcare, catalog no.17-5442-03), a buffy coat layer containing peripheral blood mononuclear cells (PBMCs) were isolated. PBMC fractions were used for three purposes: (i) cultivated in cell culture to derive induced pluripotent stem cells; (ii) analysis of circulating endothelial cells; (iii) genotyping (Supplemental Methods). For cryopreservation, PBMCs were resuspended in heat-inactivated FBS (Thermo Fisher Scientific Life Sciences, catalog no. 10082139) with 10% DMSO, frozen first at 80 °C and then transferred to liquid nitrogen for storage.

      2.3 Sendai reprogramming of PBMCs to generate iPSCs

      iPSC were generated from donor PBMCs (Preprint Doi: 10.1101/2022.05.15.491987) by CytoTune-iPS Sendai Reprogramming (Thermo Fisher Scientific Life Sciences, catalog no. A16518). Please refer to Supplemental Methods for more details.

      2.4 Maintenance and characterization of induced pluripotent stem cells

      iPSCs were grown on Matrigel-coated plates (Corning, catalog no. 354230) in mTeSR1 medium. Cells were passaged every 4–5 days using ReLeSR (StemCell Technologies, catalog no. 05872). We performed characterization of our iPSCs by immunofluorescence, karyotyping and teratoma formation assay. Please refer to Supplemental Methods for more details.

      2.5 Endothelial differentiation from iPSC lines

      We followed our previously established differentiation protocols for lateral plate mesoderm derived endothelial cells by Cheung et al., 2012 and Narmada et al., 2016 [
      • Cheung C.
      • Bernardo A.S.
      • Trotter M.W.
      • Pedersen R.A.
      • Sinha S.
      Generation of human vascular smooth muscle subtypes provides insight into embryological origin-dependent disease susceptibility.
      ,
      • Narmada B.C.
      • Goh Y.T.
      • Li H.
      • Sinha S.
      • Yu H.
      • Cheung C.
      Human stem cell-derived endothelial-hepatic Platform for efficacy testing of vascular-protective metabolites from nutraceuticals.
      ]. Please refer to Supplemental Methods for more details.

      2.6 Endothelial characterizations

      Marker characterization of the endothelial cells from iPSCs were performed by flow cytometry and immunocytochemistry. Cells were washed and stained with primary antibodies, anti-human PECAM1 (CD31; Biolegend, catalog no. 102507) or anti-human ICAM1 (CD54; BioLegend, catalog no. 353111), in a staining buffer containing PBS with 2% heat-inactivated FBS, for 30 min in the dark. Following 30 min of antibody incubation at room temperature, stained cells were washed and resuspended in PBS containing 2% heat inactivated FBS. Flow cytometry data were collected on a BD LSRFortessa X-20 cell analyzer (Becton Dickinson) and analyzed using FlowJo v10.7.1 software (Becton Dickinson). Please refer to Supplemental Table S6 for details of primary antibodies.
      Endothelial cells from iPSCs were functionally characterized by tube formation. 50 μl of Matrigel was added into each 96-well plate well and allowed to solidify for 30 min at 37 °C. Endothelial cells were seeded onto Matrigel at a cell density of 25,000 cells per well in EGM-2 and monitored every hour by light microscopy. Positive and negative controls were human coronary artery endothelial cells (HCAEC, ATCC, catalog no. PCS-100-020) and our derived iPSCs respectively. Phase-contrast images were acquired at 10 × magnification using a Nikon Ti-E inverted microscope with MetaMorph version 7.8 (Molecular Devices, California, United States). We performed quantitative analysis of endothelial cord-like structures using Angiogenesis Analyzer on ImageJ [
      • Rueden C.T.
      • Schindelin J.
      • Hiner M.C.
      • DeZonia B.E.
      • Walter A.E.
      • Arena E.T.
      • et al.
      ImageJ2: ImageJ for the next generation of scientific image data.
      ].
      ELISA of interleukin-8 was rendered on conditioned media of our iPSC-derived endothelial cells. Cells were seeded at 100,000 cells per well of 6-well plate in 2 ml of EGM-2. Culture media were conditioned for 72 h and then harvested to undergo centrifugation at 13,000g for 10 min in order to remove cell debris. Supernatant fractions were collected for interleukin-8 ELISA (Abcam, catalog no. ab100575) according to manufacturer's instructions. Normalization of IL-8 measurements was done against total cell protein of each well. Cells were washed with PBS before lysing with 350 μl RLT for total protein quantification using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific Life Sciences, catalog no. 23227).

      2.7 RNA extraction and quantitative RT-PCR

      Total RNA was isolated using RNeasy Plus Mini kit (Qiagen, catalog no. 74134) as per the manufacturer's protocol, subsequently used to generate cDNA with LunaScript RT SuperMix Kit (New England Biolabs, catalog no. E3010S). Real-time PCR was performed using SYBR green gene expression assays (New England Biolabs, catalog no. M3003S) on a QuantStudio 6 instrument (Applied Biosystems). Gene expressions were normalized to endogenous GAPDH housekeeping gene. Please refer to Supplemental Table S6 for primer sequences.

      2.8 CRISPR-Cas9 genome editing at 6p24.1

      Cas9 guide RNAs (gRNA) with predicted breaks were identified using a computational algorithm scoring system (https://benchling.com/). A non-homology end joining (NHEJ) double sgRNA-guided Cas9 system was used. Two gRNA duplex oligos were subcloned and ligated with pMIA3 plasmid [
      • Ang L.T.
      • Tan A.K.Y.
      • Autio M.I.
      • Goh S.H.
      • Choo S.H.
      • Lee K.L.
      • et al.
      A Roadmap for human liver differentiation from pluripotent stem cells.
      ] (Addgene #109399) using T4 DNA ligase (New England Biolabs, catalog no. M0202S). 1.5 × 106 cells were pre-treated with 10 μM ROCKi (StemCell Technologies, catalog no. 72302) for an hour. Single cell suspension was prepared using accutase (StemCell Technologies, catalog no. 07922) and resuspended in 100 μl of nucleofection solution and 10 μg of pMIA3 plasmid containing the selected gRNA pairs. Cells were nucleofected using Amaxa4D nucleofector (Lonza, catalog no. AAF-1002B) and P3 primary kit (Lonza, catalog no. V4XP-3024) as per manufacturer's instructions and plated out on matrigel coated 6-well plate using mTeSR with CloneR supplement. After 48 h, cells were again pre-treated with ROCKi and a single cell suspension was prepared. Fluorescence activated cell sorting (FACS) was performed to enrich for targeted cells. RFP + cells were sorted and plated onto 6-well plates containing mTeSR with CloneR, P/S and Gentamicin. Colony formation was apparent from the individually sorted iPSCs 8 days after FACS. 24 single colonies then were manually picked into 12-well plates. Colonies were amplified and split with RelesR onto 6-well plates. Remaining cells were used for genomic DNA extraction and genotyping. Clones with successful CRISPR targeting were expanded. Refer Supplementary Fig. S2A for gRNA sequences.

      2.9 ENCODE and ChIP-Seq analysis

      H3k27Ac and H4K3me3 histone marks from ChIP-Seq datasets were visualized using Integrated Genomics Viewer (IGV). For 6p24 locus spanning 1 kb region on rs6903956 (hg38 chr6:11,773,800–11,774,800), bigWig fold change over control files of various tissues/cells were queried from Roadmap epigenomics database by the ENCODE Project Consortium [
      • Sloan C.A.
      • Chan E.T.
      • Davidson J.M.
      • Malladi V.S.
      • Strattan J.S.
      • Hitz B.C.
      • et al.
      ENCODE data at the ENCODE portal.
      ] (https://www.encodeproject.org/) with the following identifiers: ENCBS609ENC, ENCBS717AAA, ENCBS709TEL, ENCBS703ZDS, ENCBS899TTJ. Vascular endothelial datasets ChIP-Seq bigWig files were downloaded from GEO dataset (GSE131953). For 10q11 locus, analysis of H3K4me3 histone marks were conducted using coordinates hg38 chr10:44,297,249–44,312,248 for visualization of weak promoter lying ∼2 kb from fragment 4, and hg38 chr10:41,830,816–41,866,615 for visualization of super-enhancer region.
      Z-scores were computed following ENCODE Project Consortium published Z-score methods [
      • Consortium E.P.
      • Moore J.E.
      • Purcaro M.J.
      • Pratt H.E.
      • Epstein C.B.
      • Shoresh N.
      • et al.
      Expanded encyclopaedias of DNA elements in the human and mouse genomes.
      ]. For 6p24 locus, to investigate the H3K4me3 and H3k27ac signal on rs6903956 compared to other regions on ADTRP, UCSC tool bigWigAverageOverBed was used to compute the signal for every SNP lying on ADTRP gene (hg38 chr6:11,713,523–11,778,803). Log10 of these signals were obtained and a Z-score computed for rs6903956 compared to the other SNPs within ADTRP. SNPs with a raw signal of 0 were not annotated a Z-score value. For 10q11 locus, log10 signal for every SNP compared to the other SNPs lying on fragments 3–5 (hg38 chr10:44,297,249–44,312,248) was computed and the maximum Z-score for each fragment listed.

      2.10 RNA-sequencing and analysis

      Total RNA was isolated from the iPSC-derived endothelial cells as per manufacturer's instructions (RNeasy Micro Kit, Qiagen, catalog no. 74004). PolyA library preparation and 150 base-pair paired end sequencing was performed by Novogene sequencing facility (Singapore) using an Illumina HiSeq sequencer. The average sequencing depth was 60 million reads. Reads were aligned using STAR aligner v2.4.1a to human genome (hg38) with default parameters. Read counts per gene were extracted from STAR output. Raw read counts were converted to log2-counts-per-million (logCPM) and the mean-variance relationship modelled with precision weights approach with voom transformation. For differential expression analysis, raw counts were normalized using edgeR R package v3.34.0 [
      • Robinson M.D.
      • McCarthy D.J.
      • Smyth G.K.
      edgeR: a Bioconductor package for differential expression analysis of digital gene expression data.
      ] using a trimmed mean of M values (TMM) and differentially expressed genes were identified using limma R package v3.46.0 [
      • Ritchie M.E.
      • Phipson B.
      • Wu D.
      • Hu Y.
      • Law C.W.
      • Shi W.
      • et al.
      Limma powers differential expression analyses for RNA-sequencing and microarray studies.
      ]. Heatmaps were generated using pheatmap R package v1.16.0. Principal component analysis plots were generated with scatterplot3d R package v0.3-41. Signed fold change “-log10(p-value)* sign(fold change)” was used as the input for GSEA. Gene ontology annotations [
      • Ashburner M.
      • Ball C.A.
      • Blake J.A.
      • Botstein D.
      • Butler H.
      • Cherry J.M.
      • et al.
      Gene ontology: tool for the unification of biology. The Gene Ontology Consortium.
      ] from Molecular Signatures Dataset [
      • Subramanian A.
      • Tamayo P.
      • Mootha V.K.
      • Mukherjee S.
      • Ebert B.L.
      • Gillette M.A.
      • et al.
      Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles.
      ] was used to perform Fast Gene Set Enrichment Analysis R package v1.18.0. The functional disease networks were generated through the use of IPA (QIAGEN Inc.) [
      • Kramer A.
      • Green J.
      • Pollard Jr., J.
      • Tugendreich S.
      Causal analysis approaches in ingenuity pathway analysis.
      ].

      2.11 Hi-C

      Hi-C was performed using the Arima-HiC kit (Arima Genomics, San Diego), according to the manufacturer's protocols up to purifying digested DNA. Purified DNA was then handed over to Integrated Genome Analytics Platform for steps up to library prep and sequencing. For creation and analysis of Hi-C contact maps, Raw fastq were first aligned to human reference genome hg38 in parallel using BWA-MEM (v0.7.15) with default parameters on each sample. Unmapped and abnormal chimeric reads were excluded [
      • Durand N.C.
      • Shamim M.S.
      • Machol I.
      • Rao S.S.
      • Huntley M.H.
      • Lander E.S.
      • et al.
      Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments.
      ]. Subsequently, duplicate reads and read pairs less than 2 kb apart were removed to avoid self-ligated fragments. The resulting.hic file contains filtered contact matrices, which were loaded into Juicebox for visualization.
      For identification of toplogically associating domains (TADs), the resulting.hic matrix files (MAPQ >30) were used as input for Arrowhead to identify TADs. Arrowhead ran automatically as part of Juicer pipeline post-processing. The following default parameters were set for Arrowhead: Resolution (r) = 5 kb; Normalization (k) = Knight-Ruiz balancing (KR); Size of sliding window (m) = 2000.
      For visualization of Hi-C output, Hi-C contact matrix are visualized using 3D genome browser and Juicebox software. Circular visualization of HUVEC (CC-2517) trans chromosomal interactions were visualized using Rondo [
      • Taberlay P.C.
      • Achinger-Kawecka J.
      • Lun A.T.
      • Buske F.A.
      • Sabir K.
      • Gould C.M.
      • et al.
      Three-dimensional disorganization of the cancer genome occurs coincident with long-range genetic and epigenetic alterations.
      ].

      2.12 3C-droplet digital PCR

      3C libraries were prepared as previously described [
      • Hagege H.
      • Klous P.
      • Braem C.
      • Splinter E.
      • Dekker J.
      • Cathala G.
      • et al.
      Quantitative analysis of chromosome conformation capture assays (3C-qPCR).
      ]. Briefly, 1 × 107 cells were cross-linked with 2% formaldehyde for 10 min and quenched with 0.125 mM glycine. Cells were lysed with 5 ml cold lysis buffer (10 mM Tris-HCl, pH 7.5; 10 mM NaCl; 5 mM MgCl2; 0.1 mM EGTA; 1x complete protease inhibitor; 11836145001 Roche) for 10 min. Cell nuclei were pelleted and resuspended in 500 μl 1.2x FastDigest Buffer (Thermo Fisher Scientific Life Sciences, catalog no. B64) with 0.3% SDS and incubated for 1 h at 37 °C. Triton X (final 2% v/v) was added and incubated for 1 h at 37 °C. Partially digested nuclei was incubated overnight at 37 °C with 400U of FastDigest HindIII (Thermo Fisher Scientific Life Sciences, catalog no. FD0504). To inhibit the restriction digestion, SDS was added (final 1.6% v/v) and incubated at 65 °C for 20–25 min, followed by Triton X (final 1% v/v) for an hour to sequester SDS. Digestion efficiencies were accessed using qPCR and only samples with the efficiency of restriction enzyme digestion above 60–70% were accepted for 3C analysis. For ligation of cross-linked and digested chromatin, samples were incubated at 16 °C for 4 h in the presence of 1.15x of T4 ligation buffer (New England Biolabs, catalog no. B0202S) and 100U T4 ligase. Proteinase K (Thermo Fisher Scientific Life Sciences, catalog no. EO0492) was added (300 μg final) and samples were de-crosslinked overnight in 65 °C, followed by RNase (Thermo Fisher Scientific Life Sciences, catalog no. EN0531) digestion (300 μg final). 3C samples were purified using phenol-chloroform extraction, and DNA concentration carefully determined via setting up qPCR reactions against reference genomic DNA of known concentration using “internal” primer sets amplifying GAPDH.
      Probe-based ddPCR was performed following manufacturer's protocol with reference to previously described 3C-digital PCR protocol [
      • Du M.
      • Wang L.
      3C-digital PCR for quantification of chromatin interactions.
      ]. Reactions were performed in total 24 μl volume using 12 μl of 2 x ddPCR Supermix for Probes (No dUTP) (Bio-Rad Laboratories, catalog no. 1863024), 900 nM of target primer pairs, 250 nM of probes and 400 ng of 3C template DNA. To check for efficiency of probes and primers, a control primer priming the region upstream of the constant fragment was used as positive control. To exclude false positive results caused by non-specific background, ddH2O was used as negative control. Reactions were performed under universal cycling conditions: 95 °C for 10 min, followed by 45 cycles at 94 °C for 30s and 60 °C for 2 min and with final enzyme denaturation at 98 °C for 10 min. Signal was quantified using QX200 Droplet Reader (Bio-Rad Laboratories) and data analysis performed using Quantasoft Analysis Pro (v1.7). Only droplets above the minimum amplitude threshold determined from negative control wells containing no template DNA were counted as positive. The interaction frequency (= target copy number per 100 ng sample) was calculated: 24 x copies/μL divided by 100 x amount of sample added to total reaction.

      2.13 Profiling of circulating endothelial cells in CAD patients

      100 μl of 1 million PBMCs was stained in the dark for 10 min at room temperature, followed by 20 min at 4 °C on an analog tube rotator with antibodies (Supplemental Table S6: Resource Table). After incubation, cells were washed and resuspended in 200 μl of PBS with 1% BSA for flow cytometry analysis. CECs were detected through the combined immunophenotypic profile of CD45-/CD31+/CD133-/DNA+. The number of CECs was expressed as cells per million of PBMCs. Flow Cytometry was performed using BD LSRFortessa X-20 and FACSDiva software (BD Biosciences) and data analyzed using FlowJo v10.7.1 software (Becton Dickinson). Each analysis included at least 30,000 events.

      2.14 RNA extraction and ddPCR in isolated CECs

      Total RNA was isolated using RNeasy Plus Micro kit (Qiagen, catalog no. 74034) as per the manufacturer's protocol, subsequently used to generate cDNA with LunaScript RT SuperMix Kit. Probe-based ddPCR was performed on QX200 Droplet Digital PCR System (ddPCR; Bio-Rad Laboratories) following manufacturer's protocol. Absolute quantity of DNA per sample (copies/μL) was processed using QuantaSoft (v.1.0.596) and converted to copies/CEC according to amount of input sample. Please refer to Supplemental Table S6 for primer sequences.

      2.15 Statistical analysis

      Statistical significance of differences between the cohorts were analyzed using GraphPad Prism version 9. Data normality was determined by Shapiro-Wilk test. Datasets with normal distributions were analyzed with unpaired Student's two-tailed t-tests to compare two conditions; one- or two-way analysis of variance (ANOVA) followed by post-hoc Tukey for datasets with more than two conditions. Non-parametric Mann–Whitney t-test or Kruskal-Wallis test were used for non-normally distributed data. Application of these statistical methods to specific experiments is noted in the figure legends. A p-value of less than 0.05 was considered significant. Results are depicted as either mean ± standard deviation (SD) or mean ± standard error of mean (SEM). Power analysis was performed to compute the sample size (n) required to detect an effect that is true. Significance of 5% (α = 0.05) and power of 80% (N = 0.80) were used.

      3. Results

      3.1 Generation of endothelial cell models from coronary artery disease and control subjects

      In our subject recruitment, CAD patients were diagnosed as non-ST-elevation myocardial infarction (NSTEMI). We selected gender-, age- and ethnicity-matched CAD patients (homozygous for risk allele A at rs6903956) and control subjects (homozygous for common allele G at rs6903956) for generation of their iPSCs (Supplemental Fig. S1A and Supplemental Table S3). CAD and control iPSC lines (n = 4 from 2 donors/group with 2 iPSC clones/donor) were created through Sendai-based reprogramming of their respective peripheral blood mononuclear cell (PBMC) samples. The iPSC lines were characterized and showed tightly packed iPSC colonies expressing classical markers of pluripotency (Fig. 1A). Gold standard in vivo teratoma formation assay confirmed the presence of differentiated tissues characteristic of each germ layer from the histology sections of the teratoma (Fig. 1B), validating pluripotency of our derived iPSC lines. Both CAD and control iPSCs were also karyotypically normal (Supplemental Fig. S1B).
      Fig. 1
      Fig. 1Derivation and characterization of induced pluripotent stem cells and endothelial cells from CAD patients and control subjects.
      (A) Immunostaining of pluripotency markers, SOX2 and NANOG, on iPSC colonies (scale bar, 100 μm). (B) In vivo teratoma formation assay of derived iPSCs. H&E staining of teratomas for differentiated tissues of three distinct germ layers (scale bar, 25 μm). (C) Top panel: stepwise protocol of iPSC differentiation toward endothelial lineage using chemically defined factors. Bottom panel: representative flow cytometry plots of iPSC-derived endothelial cells for PECAM1 expression before and after FACS. (D) Immunostaining for mature endothelial markers on iPSC-derived endothelial cells from CAD patients and control subjects (designated as CAD EC and control EC respectively). Negative controls were CAD and control iPSCs, while positive control was human coronary artery endothelial cells (HCAEC). (E) Flow cytometry characterization of proinflammatory marker ICAM1 on CAD EC and control EC. (F) ELISA-based measurement of secreted interleukin-8 concentrations in culture media of CAD EC and control EC, which were conditioned for 72 h. Bar graphs showing means with S.D. (n = 4 from 2 donors with 2 iPSC lines/donor), ***p ≤ 0.001, two-tailed t-test.
      Endothelial differentiation was then performed on CAD and control iPSCs using our established protocol [
      • Cheung C.
      • Bernardo A.S.
      • Trotter M.W.
      • Pedersen R.A.
      • Sinha S.
      Generation of human vascular smooth muscle subtypes provides insight into embryological origin-dependent disease susceptibility.
      ,
      • Narmada B.C.
      • Goh Y.T.
      • Li H.
      • Sinha S.
      • Yu H.
      • Cheung C.
      Human stem cell-derived endothelial-hepatic Platform for efficacy testing of vascular-protective metabolites from nutraceuticals.
      ]. To better mimic coronary vasculature, iPSCs were first differentiated to obtain a lateral plate mesoderm population that is the precursor tissue for cell lineages of the heart (Fig. 1C). Subsequently, a high purity of PECAM1-expressing endothelial cells (>98%) could be achieved after differentiation by chemically-defined conditions and fluorescence-activated cell sorting (Fig. 1C). Endothelial cells derived from CAD and control iPSCs (designated as CAD EC and control EC respectively) expressed mature endothelial markers such as PECAM1, VWF and NOS3 (Fig. 1D), to comparable levels as the positive control, human coronary artery endothelial cells (HCAEC). Functionally, all iPSC-derived endothelial cell lines were capable of forming cord-like structures in tube formation assays (Supplemental Fig. S1C). We then determined if CAD EC could recapitulate some disease-relevant hallmarks. We found that CAD EC were more proinflammatory than control EC by having greater ICAM1 expression (Fig. 1E), and secreting significantly higher amount of interleukin-8 (Fig. 1F), both of which are factors known to be implicated in atherosclerosis [
      • Tousoulis D.
      • Oikonomou E.
      • Economou E.K.
      • Crea F.
      • Kaski J.C.
      Inflammatory cytokines in atherosclerosis: current therapeutic approaches.
      ]. Next, we sought to establish the contribution of rs6903956 genotypes to endothelial phenotypes.

      3.2 Effect of genome editing on cis gene expressions in iPSC-derived endothelial cells

      ADTRP mRNA and protein expression have been detected in human endothelial cells [
      • Lupu C.
      • Zhu H.
      • Popescu N.I.
      • Wren J.D.
      • Lupu F.
      Novel protein ADTRP regulates TFPI expression and function in human endothelial cells in normal conditions and in response to androgen.
      ,
      • Lupu C.
      • Patel M.M.
      • Lupu F.
      Insights into the functional role of ADTRP (Androgen-Dependent TFPI-regulating protein) in Health and disease.
      ,
      • Luo C.
      • Wang F.
      • Ren X.
      • Ke T.
      • Xu C.
      • Tang B.
      • et al.
      Identification of a molecular signaling gene-gene regulatory network between GWAS susceptibility genes ADTRP and MIA3/TANGO1 for coronary artery disease.
      ]. To determine ADTRP expression in our iPSC-endothelial cells, we demonstrated that ADTRP was detected in both CAD EC and control EC, as well as in the positive control HCAEC, but not in the iPSCs (Fig. 2A). We performed variant effect prediction [
      • McLaren W.
      • Gil L.
      • Hunt S.E.
      • Riat H.S.
      • Ritchie G.R.
      • Thormann A.
      • et al.
      The Ensembl variant effect predictor.
      ] and rs6903956 showed up largely as an intronic variant as expected, with a minor possibility of being a potential regulatory region variant and nonsense-mediated mRNA decay (NMD) transcript variant (Fig. 2B). Diving deeper into the possible regulatory functions of rs6903956, we examined chromatin immunoprecipitation sequencing dataset [
      • Sloan C.A.
      • Chan E.T.
      • Davidson J.M.
      • Malladi V.S.
      • Strattan J.S.
      • Hitz B.C.
      • et al.
      ENCODE data at the ENCODE portal.
      ,
      • Nakato R.
      • Wada Y.
      • Nakaki R.
      • Nagae G.
      • Katou Y.
      • Tsutsumi S.
      • et al.
      Comprehensive epigenome characterization reveals diverse transcriptional regulation across human vascular endothelial cells.
      ] for the presence of active enhancer (H3K27ac) and promoter (H3K4me3) histone markers in relevant tissues and vascular cell types. Spanning a 1000 bp region centered on rs6903956, there seemed to be negligible histone signals in coronary artery (Fig. 2C, Supplemental Fig. S2A) and human endothelial cells (Supplemental Fig. S2B). There were no obvious regulatory elements around rs6903956.
      Fig. 2
      Fig. 2Investigation of rs6903956 cis-gene regulation through variant effect prediction and genome editing on iPSCs.
      (A) Immunostaining of ADTRP in CAD and control iPSC-derived endothelial cells. Negative controls were iPSCs, while positive control was human coronary artery endothelial cells (HCAEC). Scale bar, 100 μm. (B) Annotation of rs6903956 by Ensembl Variant Effect Predictor. (C) Visualization of H3K4me3 and H3K27Ac histone marks on hg38 chr6:11,773,800–11,774,800 based on chromatin immunoprecipitation sequencing (ChIP-Seq) of various tissues/cells. (D) Hi-C contact map visualization of 6p24.1 locus in Control EC, CAD EC and HUVEC to reflect TADs involving rs6903956 in ADTRP and neighboring cis genes. (E) Regional association plot for ∼1 kb region centered on query variant rs6903956 in CAD GWAS of East Asian Population from LDproxy tool [
      • Machiela M.J.
      • Chanock S.J.
      LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants.
      ]. Proxy variants with R2 > 0.2 were highlighted and identified as in linkage disequilibrium with rs6903956. (F) Top panel: dual gRNA CRISPR-Cas9 deletion of 63–89bp regions flanking rs6903956 in WT CAD EC and WT Control EC. Bottom panel: sequencing chromatograms to validate CRISPR-Cas9 deletion in edited (Δ) versus unedited (UNΔ) endothelial cells. Two different sets of gRNAs were rendered on each iPSC-endothelial cell line to control for CRISPR-Cas9 off-targeting effects. (G) Quantitative RT-PCR of ADTRP and cis genes in WT, isogenic Δ and UNΔ CAD and Control EC. Bar graphs showing means with S.D. (n = 2–6, from 2 donor cell lines with 2 biological replicates per cell line generated by 2 different guide RNA pairs), ***p ≤ 0.001, one-way ANOVA.
      If rs6903956 does not reside in regulatory regions, one of its potential impacts could involve long-range chromatin interactions with other genetic loci. Changes in chromatin interaction dynamics may affect the accessibility of regulatory regions distal to the SNP site, resulting in multifactorial cell-wide effects [
      • Dixon J.R.
      • Selvaraj S.
      • Yue F.
      • Kim A.
      • Li Y.
      • Shen Y.
      • et al.
      Topological domains in mammalian genomes identified by analysis of chromatin interactions.
      ]. To understand tissue-specific chromatin architectures in our iPSC-derived endothelial model, we generated Hi-C genomic contact maps using CAD EC and control EC by mapping chromatin contacts genome-wide at a 10-kb resolution (Fig. 2D). Using in situ Hi-C data, we obtained an average of 101 million long-range (≥20 kb) intra-chromosomal contacts after aligning and filtering 301 million Hi-C read pairs for each sample. Enhancers find their target genes within topologically associated domains (TADs), which are demarcated by interactome boundaries that restrict chromatin interactions within a spatially confined compartment in the genome [
      • Nora E.P.
      • Lajoie B.R.
      • Schulz E.G.
      • Giorgetti L.
      • Okamoto I.
      • Servant N.
      • et al.
      Spatial partitioning of the regulatory landscape of the X-inactivation centre.
      ]. As TADs are known to be strongly conserved across species and cell types, we compared CAD EC and control EC Hi-C data with another well-characterized HUVEC line [
      • Rao S.S.
      • Huntley M.H.
      • Durand N.C.
      • Stamenova E.K.
      • Bochkov I.D.
      • Robinson J.T.
      • et al.
      A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping.
      ], which collectively reflected rather similar TAD characteristics (Fig. 2D).
      To gain a better resolution of the effects of rs6903956, we performed genome editing to compensate for our limitation of not having iPSC lines from the same clinical phenotype with both risk and non-risk alleles. The minor allele A at rs6903956 has been involved as a CAD susceptibility SNP primarily in Chinese populations [
      • Wang F.
      • Xu C.Q.
      • He Q.
      • Cai J.P.
      • Li X.C.
      • Wang D.
      • et al.
      Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
      ]. In analyzing nearby variants, a short tandem repeat variant, rs140361069, was found in weak linkage disequilibrium (LD) with rs6903956 in East Asian populations (R2 = 0.21), along with five other SNPs in strong LD (R2 > 0.8) (Fig. 2E). Hence, we generated isogenic iPSC lines with 63–89bp deletions on 6p24.1 containing rs6903956 and rs140361069 via Cas9-induced non-homologous end joining, with the use of a dual guide RNA (gRNA) targeting strategy [
      • Do P.T.
      • Nguyen C.X.
      • Bui H.T.
      • Tran L.T.N.
      • Stacey G.
      • Gillman J.D.
      • et al.
      Demonstration of highly efficient dual gRNA CRISPR/Cas9 editing of the homeologous GmFAD2-1A and GmFAD2-1B genes to yield a high oleic, low linoleic and alpha-linolenic acid phenotype in soybean.
      ] (Fig. 2F). To control for effects of CRISPR-Cas off-targeting, we designed two different pairs of gRNAs targeting the same region for both CAD and control iPSCs (Supplemental Fig. S2C). Guide RNAs were evaluated for DNA double strand break efficiencies in HEK293T cells using a disrupted GFP construct containing a 582bp ADTRP fragment (Supplemental Fig. S2D). In addition, we also generated unedited CAD and control iPSCs which were subjected to the same CRISPR-Cas9 nucleofection, but without a successful deletion at our target genomic region. Deletion of targeted genomic regions did not significantly impact iPSC growth and pluripotency marker expressions (TRA-1-60, OCT4, SOX2) (Supplemental Fig. S2E). These iPSC lines were further differentiated into endothelial cells. We henceforth designated WT CAD EC and WT Control EC as the parental wild-type lines. Those with successful deletions of rs6903956 and rs140361069 were Δ CAD EC and Δ Control EC, while those without successful deletions were UNΔ CAD EC and UNΔ Control EC (Fig. 2F).
      Our Hi-C contact maps earlier revealed rs6903956 to lie in TADs with neighboring genes HIVEP1, EDN1 and PHACTR1 (Fig. 2D). To identify if the deleted regions on 6p24.1 (Δ63-89bp) might affect expression of cis genes in the same TADs, we performed qRT-PCR with our edited and unedited endothelial cell lines. A previous study reported that presence of risk allele A at rs6903956 resulted in decreased mRNA expression of ADTRP in leukocytes [
      • Wang F.
      • Xu C.Q.
      • He Q.
      • Cai J.P.
      • Li X.C.
      • Wang D.
      • et al.
      Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
      ]. However, no significant difference was observed for ADTRP, HIVEP1 and EDN1 in our endothelial cell models (Fig. 2G). While there was an upregulation of PHACTR1 in Δ Control EC compared to all CAD endothelial cell lines, there were negligible differences of cis gene expressions between edited and unedited cell lines within their respective CAD and control groups.

      3.3 Genetically edited CAD endothelial cells reveal downregulation of molecular pathways relating to endothelial instability

      To elucidate wider effects of the deleted region of 6p24.1 locus, RNA-sequencing was conducted on WT CAD EC with risk genotype (AA) and WT Control EC with non-risk genotype (GG) (n = 3 biological replicates from 3 independent differentiation batches), as well as their edited (Δ) and unedited (UNΔ) counterparts (n = 6 from 2 cell lines generated by 2 different guide RNA pairs, 3 independent differentiation batches/cell line). Principle component analyses demonstrated minimal variance between biological replicates (Supplemental Fig. S3A). There was also a high degree of similarity between different pairs of gRNAs targeting the same region, suggesting minimal off-targeting effect (Supplemental Fig. S3A). Differential expression analysis revealed 183 upregulated genes and 270 downregulated genes in Δ CAD EC compared to UNΔ CAD EC (Fig. 3A, top). Most of these differentially expressed genes were similarly upregulated or downregulated when WT CAD EC dataset was used in comparison to Δ CAD EC (Supplemental Fig. S3B). Considering that double-strand breaks induced by Cas9 might give rise to DNA damage [
      • Haapaniemi E.
      • Botla S.
      • Persson J.
      • Schmierer B.
      • Taipale J.
      CRISPR-Cas9 genome editing induces a p53-mediated DNA damage response.
      ], we proceeded with gene set enrichment analyses [
      • Subramanian A.
      • Tamayo P.
      • Mootha V.K.
      • Mukherjee S.
      • Ebert B.L.
      • Gillette M.A.
      • et al.
      Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles.
      ,
      • Mootha V.K.
      • Lindgren C.M.
      • Eriksson K.F.
      • Subramanian A.
      • Sihag S.
      • Lehar J.
      • et al.
      PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes.
      ] using the differentially expressed genes between isogenic UNΔ EC and Δ EC. Enrichment analyses comparing Δ CAD EC against UNΔ CAD EC revealed downregulation of processes/pathways relating to vasculature development, abnormal vascular physiology, endothelial cell migration and sprouting angiogenesis, etc (Fig. 3B). This might suggest that the deleted region on 6p24.1 containing AA risk genotype in Δ CAD EC had led to a less proliferative/activated endothelial phenotype. On the other hand, there were substantially more differentially regulated genes between Δ Control EC compared to UNΔ Control EC (Fig. 3A, bottom). However, their gene set enrichment analyses did not reveal processes/pathways that were directly indicative of perturbed vascular biology (Supplemental Fig. S3C). The deleted region on 6p24.1 containing GG non-risk genotype in Δ Control EC might instead impact broadly on other fundamental biological processes. Therefore, we postulated that the rs6903956 G-to-A allele substitution could underlie a gain-of-function pathogenic role, as removal of AA risk genotype in Δ CAD EC, but not removal of GG non-risk genotype in Δ Control EC, resulted in downregulation of endothelial activation mechanisms. Nonetheless, we also recognized that the difference between Δ CAD EC and UNΔ CAD EC versus the difference between Δ Control EC and UNΔ Control EC likely indicated the SNP effects with potential gene-environment interactions [
      • Joseph P.G.
      • Pare G.
      • Anand S.S.
      Exploring gene-environment relationships in cardiovascular disease.
      ,
      • Talmud P.J.
      Gene-environment interaction and its impact on coronary heart disease risk.
      ] inherent with the CAD background, as well as the cumulative effect together with other CAD risk alleles [
      • Nikpay M.
      • Goel A.
      • Won H.H.
      • Hall L.M.
      • Willenborg C.
      • Kanoni S.
      • et al.
      A comprehensive 1,000 Genomes-based genome-wide association meta-analysis of coronary artery disease.
      ,
      • McPherson R.
      • Tybjaerg-Hansen A.
      Genetics of coronary artery disease.
      ].
      Fig. 3
      Fig. 3Transcriptomic analysis of isogenic edited and unedited iPSC-derived endothelial cells.
      (A) Volcano plots showing the adjusted p-values and log2 fold change values of genes in Δ CAD EC versus UNΔ CAD EC (top), as well as in Δ Control EC versus UNΔ Control EC (bottom). Differentially expressed genes are indicated by red (upregulated) and blue (downregulated) dots. (B) Top 20 ontology gene sets normalized enrichment scores from Gene Set Enrichment Analysis, ranked according to p-value. Signed fold change of genes differentially expressed in Δ CAD EC versus UNΔ CAD EC was used as input. (C) An interactome of differentially expressed genes in Δ CAD EC versus UNΔ CAD EC, arising from the ‘cardiovascular system development and function’ network based on Ingenuity Pathway Analysis. Genes indicated in red and blue were upregulated and downregulated in Δ CAD EC respectively. (D) Parallel coordinate plot for curated genes that showed gene expressions in Δ CAD EC ‘normalized’ to the levels of that in WT Control EC. (E) Heatmap visualization of curated genes from (D). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
      Next, we used a double-pronged approach to identify candidate effector genes due to the deleted region on 6p24.1. First, we performed ‘diseases and functions’ network analysis by Ingenuity Pathway Analysis on the differentially regulated genes between Δ CAD EC and UNΔ CAD EC. We found that ‘cardiovascular system development and function’ was among the most dysregulated network in Δ CAD EC (Supplemental Fig. S3D). Using the genes from this network, we derived a network interactome that showed predominately downregulated genes, with the CXC chemokine ligand 12 (CXCL12) being one of the central elements of this interactome (Fig. 3C). Secondly, based on parallel coordination plots of gene expression dynamics, we curated candidate genes whereby deletion of 6p24.1 containing AA risk genotype in Δ CAD EC could restore expressions to the levels found in WT Control EC harboring GG non-risk genotype (Fig. 3D). A heatmap visualization helped us appreciate the candidate genes that were normalized to the levels of WT Control EC upon removal of risk genotype in Δ CAD EC (Fig. 3E). Many of these genes have known roles in atherosclerosis (CXCL12, CXCR4, FOXP2, COL15A1, IGFBP7), angiogenesis/VEGF pathway (DMKN, SFRP1, MAP2, SULF1) and metabolic diseases (SUNCR1, NPY, HNMT, SIX1) (Supplemental Table S2). Others are involved in fundamental cellular processes such as transcription regulation and inflammation. Consistently, CXCL12 surfaced as a potential candidate gene. CXCL12 resides within chromosome 10q11.21 that has been reported as a risk locus for CAD susceptibility [
      • Samani N.J.
      • Erdmann J.
      • Hall A.S.
      • Hengstenberg C.
      • Mangino M.
      • Mayer B.
      • et al.
      Genomewide association analysis of coronary artery disease.
      ,
      • Samani N.J.
      • Deloukas P.
      • Erdmann J.
      • Hengstenberg C.
      • Kuulasmaa K.
      • McGinnis R.
      • et al.
      Large scale association analysis of novel genetic loci for coronary artery disease.
      ]. Three CXCL12 variants (rs266089, rs1065297, rs10793538) have been reported to show an association with CAD in the Chinese Han population [
      • Zhang J.
      • Ma H.
      • Gao J.
      • Kong S.
      • You J.
      • Sheng Y.
      Variants in the CXCL12 gene was associated with coronary artery disease susceptibility in Chinese Han population.
      ]. We then genotyped our iPSC lines to ascertain if they had underlying CXCL12 variants. All our iPSC lines harbored common alleles at those CXCL12 SNPs with no known association with CAD, except for one of the control iPSC lines with heterozygous ‘GA’ genotype at rs266089 that has been postulated to confer CAD risk (Supplemental Table S3). However, there has been no eQTL analysis to establish whether these CXCL12 SNPs may result in change of CXCL12 expression. We were motivated by the possibility of inter-chromosomal (trans) interaction of CAD susceptibility loci linking 6p24.1 and 10q11.21.

      3.4 Trans interaction of 6p24.1 and 10q11.21 in endothelial cells

      We performed data mining of Hi-C data from Rao et al. [
      • Rao S.S.
      • Huntley M.H.
      • Durand N.C.
      • Stamenova E.K.
      • Bochkov I.D.
      • Robinson J.T.
      • et al.
      A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping.
      ] to probe the chromatin landscape between 6p24.1 and 10q11.21 in HUVEC dataset. Strikingly, we observed that both chromosomal regions were closely intertwined in 3D space, probably due to the presence of a super enhancer region on 10q11.21, spanning around 30 kb (Fig. 4A). Super enhancers are known to play key roles in organizing gene expression patterns that regulate cell identity [
      • Wang X.
      • Cairns M.J.
      • Yan J.
      Super-enhancers in transcriptional regulation and genome organization.
      ]. To examine if such super enhancer activity was detected in other endothelial models, we analyzed a database of ChIP-Seq on various human vascular endothelial cell lines [
      • Nakato R.
      • Wada Y.
      • Nakaki R.
      • Nagae G.
      • Katou Y.
      • Tsutsumi S.
      • et al.
      Comprehensive epigenome characterization reveals diverse transcriptional regulation across human vascular endothelial cells.
      ]. Indeed, there was general increase in H3K27Ac enhancer marks at our observed super enhancer region on 10q11.21 in all vascular endothelial cell types (i.e. HCCaEC, HCoAEC, HaoEC, HPAEC, HENDC) except IMR90, a lung fibroblast cell line (Supplemental Fig. S4A).
      Fig. 4
      Fig. 4Frequent chromatin contacts between 6p24.1 and 10q11.21.
      (A) Chord diagram visualizing HUVEC dataset from Rao et al. [
      • Rao S.S.
      • Huntley M.H.
      • Durand N.C.
      • Stamenova E.K.
      • Bochkov I.D.
      • Robinson J.T.
      • et al.
      A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping.
      ]. Hi-C inter-chromosomal loops between 6p24.1 and 10q11.21 on hg19. Only significant interactions (q-value <0.01) are shown. Presence of a super enhancer region on hg38: chr10:41,830,816- 41,866,615 (boxed in red) where most of the chromatin contacts on 6p24.1 and 10q11.2 are anchored on, including CXCL12 and ADTRP. (B) Chromatin conformation capture (3C)-droplet digital PCR of WT Control EC and WT CAD EC. Anchoring on HindIII fragment harboring rs6903956 (constant fragment; 2279bp resolution), we probed for HindIII fragments 2–6 lying on 5′ untranslated region of CXCL12. Line graph showing means with S.E.M. (n = 4 from 2 donor cell lines with 2 technical replicates per cell line), ****p ≤ 0.0001, two-way ANOVA comparing WT CAD EC with WT Control EC for each individual fragment. (C) ChIP-Seq of H3K4me3 histone marks in vascular endothelial cells [
      • Nakato R.
      • Wada Y.
      • Nakaki R.
      • Nagae G.
      • Katou Y.
      • Tsutsumi S.
      • et al.
      Comprehensive epigenome characterization reveals diverse transcriptional regulation across human vascular endothelial cells.
      ] on hg38 chr10:44,297,249–44,312,248: Common Carotid Artery ECs (HCoAEC, n = 4), Coronary Artery ECs (HCCaEC, n = 6), Human Aortic ECs (HAoEC, n = 3), Human Pulmonary Artery ECs (HPAEC, n = 2), Human Endocardiac Cells (HENDC, n = 3). Presence of a weak promoter ∼2 kb downstream of fragment 4 was boxed in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
      While this super enhancer on 10q11.21 could have facilitated long-range physical interaction with 6p24.1, we asked if 6p24.1 covering rs6903956 could be in close genomic proximity with regulatory elements activating CXCL12 transcription in CAD EC. Chromatin conformation capture (3C) was conducted on WT CAD EC with risk AA and WT Control EC with non-risk GG (n = 4 from 2 donor cell lines with 2 technical replicates per cell line) in order to probe interactions between a constant HindIII fragment covering rs6903956 (fixed interaction hotspot) and the entire CXCL12 target genomic region covering 27 HindIII cutting sites (fragments 1–28) (Supplemental Fig. S4B). The complexity of the human genome makes it difficult to quantify true ligation frequencies due to other random competing interactions, especially since ligation frequency is generally low between two separate chromosomes [
      • Simonis M.
      • Kooren J.
      • de Laat W.
      An evaluation of 3C-based methods to capture DNA interactions.
      ,
      • McCord R.P.
      • Kaplan N.
      • Giorgetti L.
      Chromosome conformation capture and beyond: toward an integrative view of chromosome structure and function.
      ]. Furthermore, the inherent bias posed by 3C technique toward cis interactions due to the power-law decay model makes it only possible for the most frequent long-range interactions to be accurately quantified [
      • McCord R.P.
      • Kaplan N.
      • Giorgetti L.
      Chromosome conformation capture and beyond: toward an integrative view of chromosome structure and function.
      ]. Therefore, instead of conventional TaqMan qPCR, we switched to droplet digital PCR, which delivered high precision of absolute ligation product copy numbers, without the numerous normalization controls that 3C-qPCR requires [
      • Du M.
      • Wang L.
      3C-digital PCR for quantification of chromatin interactions.
      ], allowing us to accurately quantify ligation frequency using 3C. To make up the large number of cells (10 million) required for 3C, we pooled five independently differentiated batches of iPSC-derived endothelial cells, at the same time controlling for batch variations. Our findings demonstrated frequent interactions between our control fragment containing rs6903956 with the entire CXCL12 genomic region (Supplemental Fig. S4B). In particular, fragment 4 which lies within the 5’ untranslated region of CXCL12 appeared to have the strongest interaction signals with the control fragment containing rs6903956 in WT CAD EC (Supplemental Fig. S4B). We repeated the experiment with prioritization on fragments 2–6 on WT CAD EC with risk alleles and WT Control EC with non-risk alleles. Similar results of interaction frequency peaking around fragment 4 in WT CAD EC were reproduced (Fig. 4B).
      To understand if regions around fragment 4 could be functionally regulating CXCL12 expression, we reanalyzed ChIP-Seq dataset of vascular endothelial cells [
      • Nakato R.
      • Wada Y.
      • Nakaki R.
      • Nagae G.
      • Katou Y.
      • Tsutsumi S.
      • et al.
      Comprehensive epigenome characterization reveals diverse transcriptional regulation across human vascular endothelial cells.
      ]. The data predicted weak and consistent promoter (H3K4me3) histone peaks lying on fragment 5, located 2 kb from fragment 4 (Fig. 4C, Supplemental Fig. S4C). Taken together, CAD background coupled with AA risk genotype at rs6903956 seemed to confer greater interaction frequency near a weak promoter on CXCL12.

      3.5 rs6903956 risk alleles are associated with vascular injury in patients with CAD

      To test the effect of risk allele A at rs6903956, we measured circulating endothelial cells (CECs) as a biomarker for vascular injury [
      • Hebbel R.P.
      Blood endothelial cells: utility from ambiguity.
      ], in the blood of CAD patients stratified by genotypes. CECs are damaged endothelial cells found in the peripheral blood, which have been detached from the blood vessel lining because of vascular injury [
      • Hebbel R.P.
      Blood endothelial cells: utility from ambiguity.
      ,
      • Blann A.D.
      • Woywodt A.
      • Bertolini F.
      • Bull T.M.
      • Buyon J.P.
      • Clancy R.M.
      • et al.
      Circulating endothelial cells. Biomarker of vascular disease.
      ,
      • Chioh F.W.
      • Fong S.W.
      • Young B.E.
      • Wu K.X.
      • Siau A.
      • Krishnan S.
      • et al.
      Convalescent COVID-19 patients are susceptible to endothelial dysfunction due to persistent immune activation.
      ]. We were mindful of our small sample size for iPSC-EC studies. Therefore, sample size derivation analysis was performed to ensure that validation experiments in CECs were adequately powered. We estimated fold effect size between risk genotypes AA/AG and non-risk genotype GG, and their respective standard deviations based on flow cytometry detection of circulating endothelial cells in 8 samples per genotype group (Supplemental Table S4). The minimum number of samples needed to achieve statistical significance of p < 0.05 with a power of 0.8 was 20 per genotype. With all CAD patients age- and gender-matched, we were able to achieve appropriate sample sizes in non-risk GG (n = 24) and risk AA/AG (n = 31) genotypes (Supplemental Table S5), assuming a dominant model. Remarkably, patients with risk genotypes AA/AG had significantly higher number of CECs compared with patients with non-risk GG genotype (Fig. 5A). When we compared the number of CECs of AA and AG groups separately, it reflected a trend where ‘A’ risk allele might act through an additive manner, although not significant (Supplemental Fig. S5). Our current results proved heightened levels of vascular injury associated with rs6903956 ‘A’ risk allele.
      Fig. 5
      Fig. 5Profiling of circulating endothelial cells reveals association of rs6903956 risk allele with vascular injury.
      (A) Association between number of circulating endothelial cells (CECs) per million PBMCs in patient samples and their genotypes at rs6903956. Bar graphs showing means with S.D. (n = 31 ‘GA/AA’ and n = 24 ‘GG’), **p ≤ 0.01, Mann–Whitney t-test. (B) Droplet digital PCR of CXCL12 transcript copies from isolated CECs pooled from ‘GG’ (n = 20) and ‘GA/AA’ (n = 27) patient samples. Bar graph showing means with S.D., *p ≤ 0.05, Mann–Whitney t-test.
      To validate variation of endothelial expression of CXCL12 due to rs6903956, CECs were isolated by fluorescence-activated cell sorting and grouped according to the presence and absence of risk alleles (n = 27 ‘GA/AA’ and n = 20 ‘GG’). We performed droplet digital PCR to quantify CXCL12 transcript copies due to the very low numbers of isolated CECs from the blood. We observed about 10 folds more CXCL12 copy number per CEC in risk genotypes ‘AA/AG’ compared with non-risk ‘GG’ genotype (Fig. 5B). These results strengthen the association between rs6903956 ‘A’ allele and higher CXCL12 expression.

      4. Discussion

      Our genetic experimentation using patient iPSC-derived endothelial cells and data mining offers new insights into the molecular basis of susceptibility SNP rs6903956 in endothelial biology. We propose a mechanistic model in which trans-chromosomal impact on the atherosclerosis-implicated gene CXCL12 could be one of the pathways through which 6p24.1 covering rs6903956 exerts its deleterious effects, as evidenced by elevated levels of vascular injury in patients associated with the A risk allele. Facilitated by CRISPR-Cas9 deletions (Δ63-89bp) on 6p24.1 including rs6903956 in our iPSC-derived endothelial cells, we demonstrated that presence of risk genotype AA at rs6903956 dysregulates vascular physiology transcriptional networks. On the other hand, removal of the non-risk GG genotype at rs6903956 had minimal effect on endothelial-specific pathways. Upregulation of CXCL12 gene expression was related to increased inter-chromosomal interaction frequency of 6p24.1 with AA risk genotype near a putative CXCL12 promoter region. Notably, a link between rs6903956 and vascular injury was validated by an association of risk allele A with higher numbers of damaged CECs detected.
      Causal non-coding SNPs can interfere with normal gene regulation by being regulatory loss- or gain-of-function mutations. On certain occasions, non-coding variants also have the potential to interfere with chromatin topological domain architecture by repositioning regulatory elements between domains or disrupting TAD boundaries, inducing ectopic gene activation and causing misexpression [
      • Spielmann M.
      • Mundlos S.
      Looking beyond the genes: the role of non-coding variants in human disease.
      ]. We found previously that many cardiovascular disease-associated SNPs contribute to variation of gene expressions within the same TAD as the variants concerned [
      • Tan W.L.W.
      • Anene-Nzelu C.G.
      • Wong E.
      • Lee C.J.M.
      • Tan H.S.
      • Tang S.J.
      • et al.
      Epigenomes of human hearts reveal new genetic variants relevant for cardiac disease and phenotype.
      ]. However, our data across disease and control isogenic (Δ63-89bp) endothelial cell lines suggest that rs6903956 may not affect the expressions of cis genes (i.e., ADTRP, HIVEP1, EDN1 and PHACTR1) within the same TAD. On the other hand, we demonstrated the feasibility of obtaining high-resolution trans interaction by coupling chromatin conformation capture (3C) with droplet digital PCR. Higher inter-chromosomal contact frequency was found between 6p24.1 containing rs6903956 and a 1,792bp region lying within 5’ untranslated region of CXCL12, which appeared to be a weak promoter. We note that rs6903956 is in linkage disequilibrium with rs140361069, a short tandem repeat, which was missing in our CAD EC harboring homozygous A alleles at rs6903956. Sun et al. had shown that some disease-associated tandem repeats may be located with chromatin domain boundaries, affecting insulation of genomic neighborhoods [
      • Sun J.H.
      • Zhou L.
      • Emerson D.J.
      • Phyo S.A.
      • Titus K.R.
      • Gong W.
      • et al.
      Disease-associated short tandem repeats Co-localize with chromatin domain boundaries.
      ]. It remains to be elucidated how G-to-A substitution at rs6903956 and/or missing tandem repeats at rs140361069 change chromatin dynamics, leading to elevated interaction frequency with a distal promoter site on CXCL12. No H3K27Ac enhancer histone marker has been detected at rs6903956 in vascular cells or tissues. Recent studies support the notion that enhancer activity might be the result of a synergistic action between multiple acetylation events at different histone sites, rather than the sole action of H3K27Ac [
      • Zhang T.
      • Zhang Z.
      • Dong Q.
      • Xiong J.
      • Zhu B.
      Histone H3K27 acetylation is dispensable for enhancer activity in mouse embryonic stem cells.
      ]. Further investigation on other histone variants in endothelial cell system is necessary to substantiate the regulatory role of rs6903956 as an enhancer.
      Our Hi-C data mining suggested there was an endothelial-specific super-enhancer on 10q11.21 that might be responsible for the regular contacts between 6p24.1 and 10q11.21. Super-enhancers have been thought to facilitate cell-identity specific regulatory response via the formation of phase separation bodies [
      • Hnisz D.
      • Shrinivas K.
      • Young R.A.
      • Chakraborty A.K.
      • Sharp P.A.
      A phase separation model for transcriptional control.
      ]. We find it fascinating that 6p24.1 and 10q11.21, both known loci harboring genetic polymorphisms associated with CAD, are closely interacting in 3D space. It will be worth investigating the potential global regulation of super-enhancers and how CAD risk loci on different chromosomes can mutually regulate gene expressions in chromatin space, opening the door for discovery of coregulated gene clusters via super-enhancer-mediated inter-chromosomal interaction.
      In this study, rs6903956 was associated with increased CXCL12 expression in endothelial cells. Serum CXCL12 levels are markedly increased in patients with cardiovascular diseases [
      • Sjaarda J.
      • Gerstein H.
      • Chong M.
      • Yusuf S.
      • Meyre D.
      • Anand S.S.
      • et al.
      Blood CSF1 and CXCL12 as causal mediators of coronary artery disease.
      ,
      • Tavakolian Ferdousie V.
      • Mohammadi M.
      • Hassanshahi G.
      • Khorramdelazad H.
      • Khanamani Falahati-Pour S.
      • Mirzaei M.
      • et al.
      Serum CXCL10 and CXCL12 chemokine levels are associated with the severity of coronary artery disease and coronary artery occlusion.
      ], in which endothelial-derived CXCL12 is a key driver of atherosclerosis and one of the contributors to serum CXCL12 levels [
      • Döring Y.
      • Vorst EPCvd
      • Duchene J.
      • Jansen Y.
      • Gencer S.
      • Bidzhekov K.
      • et al.
      CXCL12 derived from endothelial cells promotes atherosclerosis to drive coronary artery disease.
      ]. CXCL12 has previously been associated with a salutary effect on atherosclerotic plaque stability, possibly due to cell-specific atheroprotective effects and CXCL12's role in recruitment of bone marrow-derived cells to sites of injury [
      • Zernecke A.
      • Weber C.
      Chemokines in atherosclerosis.
      ,
      • Wang Y.
      • Huang J.
      • Li Y.
      • Yang G.Y.
      Roles of chemokine CXCL12 and its receptors in ischemic stroke.
      ]. In contrast, other studies have shown that inflammation mediated by CXCL12 and its receptors have long been linked to vascular injury, involving processes of monocyte differentiation and macrophage infiltration, aggravating pro-inflammatory responses on vascular endothelium [
      • Gencer S.
      • Evans B.R.
      • van der Vorst E.P.C.
      • Doring Y.
      • Weber C.
      Inflammatory chemokines in atherosclerosis.
      ]. CXCL12 is also involved in other pro-atherogenic processes, including hyperlipidemia and insulin resistance that may in turn affect endothelial integrity [
      • Gao J.H.
      • Yu X.H.
      • Tang C.K.
      CXC chemokine ligand 12 (CXCL12) in atherosclerosis: an underlying therapeutic target.
      ]. To validate our findings, we measured CECs as a biomarker to examine the functional consequence of rs6903956 on vascular injury in CAD patients, as CECs have been shown to be an indicator of vascular injury as they are shed into the circulation following vascular damage [
      • Farinacci M.
      • Krahn T.
      • Dinh W.
      • Volk H.D.
      • Dungen H.D.
      • Wagner J.
      • et al.
      Circulating endothelial cells as biomarker for cardiovascular diseases.
      ,
      • Hill J.M.
      • Zalos G.
      • Halcox J.P.
      • Schenke W.H.
      • Waclawiw M.A.
      • Quyyumi A.A.
      • et al.
      Circulating endothelial progenitor cells, vascular function, and cardiovascular risk.
      ]. CECs are valuable markers of vascular dysfunction in a variety of vascular disorders including myocardial infarction, acute ischemic stroke [
      • Nadar S.K.
      • Lip G.Y.
      • Lee K.W.
      • Blann A.D.
      Circulating endothelial cells in acute ischaemic stroke.
      ], atherosclerosis, vasculitis, coronary artery disease [
      • Schmidt D.E.
      • Manca M.
      • Hoefer I.E.
      Circulating endothelial cells in coronary artery disease and acute coronary syndrome.
      ] etc. Interestingly, we detected a significantly higher number of damaged CECs in patients harboring risk alleles A at rs6903956 than those with homozygous non-risk alleles G. Furthermore, CECs isolated from patients with risk genotypes AA/AG expressed higher levels of CXCL12 gene expression compared with non-risk GG group.
      There are limitations with our study. Firstly, given that we did not create iPSC lines from the same clinical phenotype harboring both risk and non-risk alleles, we could not attribute genetic effect solely due to rs6903956 genotypes. Instead, the results of deleting 6p24.1 region with ‘AA’ risk alleles in isogenic cell lines (Δ CAD EC vs. UNΔ CAD EC) must be interpreted with potential gene-environment interactions [
      • Joseph P.G.
      • Pare G.
      • Anand S.S.
      Exploring gene-environment relationships in cardiovascular disease.
      ,
      • Talmud P.J.
      Gene-environment interaction and its impact on coronary heart disease risk.
      ] inherent with the CAD background. Secondly, we have not validated if the 6p24.1 locus where rs6903956 resides has potential enhancer activity or converges on binding sites of chromosomal architectural proteins which may account for the differences of long-range chromatin interaction dynamics. Finally, sexual dimorphisms are well-known in the pathways relevant to endothelial dysfunctions [
      • Davis C.M.
      • Fairbanks S.L.
      • Alkayed N.J.
      Mechanism of the sex difference in endothelial dysfunction after stroke.
      ,
      • Vinas J.L.
      • Porter C.J.
      • Douvris A.
      • Spence M.
      • Gutsol A.
      • Zimpelmann J.A.
      • et al.
      Sex diversity in proximal tubule and endothelial gene expression in mice with ischemic acute kidney injury.
      ], however we have only studied gender-matched samples from male participants. In the GWAS that discovered rs6903956 as a susceptibility locus for CAD [
      • Wang F.
      • Xu C.Q.
      • He Q.
      • Cai J.P.
      • Li X.C.
      • Wang D.
      • et al.
      Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
      ], it was reported that the odds ratio for allele A was higher in the female population (OR = 1.63) compared to the male population (OR = 1.44) but this difference was not significant (p = 0.32, Breslow-Day tests). Further studies are required to derive sex-dependent effects on rs6903956-associated mechanisms.
      Overall, our study demonstrates the advantage of gene editing in human iPSCs to uncover functional effects of a non-coding disease-associated SNP relevant to endothelial biology in patients with CAD. Implementation of genetic testing is cost-effective for triaging patients based on individual genetic susceptibilities. Our findings of rs6903956-associated pathways may improve risk stratification and guide the design of therapeutic trials aiming at mitigating CAD. This will serve as a proof-of-principle that patient-derived iPSCs can offer valuable insights into precision medicine approaches of genotype-guided therapies.

      Financial support

      The National Research Foundation, Singapore (Project Number 370062002 ) funded the Singapore Coronary Artery Disease Genetics Study (SCADGENS) and genotyping of the participants. The team from Nanyang Technological University Singapore was funded by an Academic Research Fund Tier 1 grant ( 2018-T1-001-030 ) from the Ministry of Education, Singapore, Human Frontier Science Program Research Grant ( RGY0069/2019 ), and the Nanyang Assistant Professorship. K.Y.T. is supported by NTU Research Scholarship. H.H.L. is supported by the Institute of Molecular and Cell Biology (IMCB) Scientific Staff Development Award (SSDA) for her part-time Ph.D. A.K.K.T. is supported by IMCB, A*STAR , Precision Medicine and Personalised Therapeutics Joint Research Grant 2019, the 2nd A*STAR-AMED Joint Grant Call 192B9002 and NUHSRO/2021/035/NUSMed/04/NUS-IMCB Joint Lab/LOA .

      CRediT authorship contribution statement

      Kai Yi Tay: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. Kan Xing Wu: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – review & editing. Florence Wen Jing Chioh: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – review & editing. Matias Ilmari Autio: Methodology, Writing – review & editing. Nicole Min Qian Pek: Data curation, Writing – review & editing. Balakrishnan Chakrapani Narmada: Methodology, Writing – review & editing, Methodology, Writing – review & editing. Sock-Hwee Tan: Writing – review & editing. Adrian Fatt-Hoe Low: Methodology, Writing – review & editing. Michelle Mulan Lian: Methodology, Writing – review & editing. Elaine Guo Yan Chew: Methodology, Writing – review & editing. Hwee Hui Lau: Data curation, Methodology, Writing – review & editing. Shih Ling Kao: Resources, Writing – review & editing. Adrian Kee Keong Teo: Methodology, Funding acquisition, Supervision, Writing – review & editing. Jia Nee Foo: Methodology, Supervision, Writing – review & editing. Roger Sik Yin Foo: Methodology, Funding acquisition, Supervision, Writing – review & editing. Chew Kiat Heng: Formal analysis, Resources, Funding acquisition, Project administration, Writing – review & editing. Mark Yan Yee Chan: Formal analysis, Resources, Funding acquisition, Project administration, Supervision, Writing – review & editing. Christine Cheung: Conceptualization, Formal analysis, Investigation, Methodology, Resources, Visualization, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review & editing, All authors approved the submitted manuscript version.

      Declaration of competing interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgements

      We thank all patients and healthy donors who have participated in this study. Special thanks to the team at National University Hospital for coordinating clinical sample collection, Ms Kee Bee Leng for conducting interview with each participant, Mr Tan Wei Heng, Ms Konstanze Tan and Mr Marcus Teo for experimental assistance, Asst Prof Melissa Fullwood and Dr Wilson Lek Wen Tan for discussion on ChIA-PET and Hi-C visualizations, Dr Stefano Perna for discussion on ChIP-Seq Z-score.

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      References

        • OECD
        • Organization WH
        Mortality from Cardiovascular Disease.
        2020
        • Zdravkovic S.
        • Wienke A.
        • Pedersen N.L.
        • Marenberg M.E.
        • Yashin A.I.
        • De Faire U.
        Heritability of death from coronary heart disease: a 36-year follow-up of 20 966 Swedish twins.
        J. Intern. Med. 2002; 252: 247-254
        • Schunkert H.
        • Konig I.R.
        • Kathiresan S.
        • Reilly M.P.
        • Assimes T.L.
        • Holm H.
        • et al.
        Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease.
        Nat. Genet. 2011; 43: 333-338
        • Samani N.J.
        • Erdmann J.
        • Hall A.S.
        • Hengstenberg C.
        • Mangino M.
        • Mayer B.
        • et al.
        Genomewide association analysis of coronary artery disease.
        N. Engl. J. Med. 2007; 357: 443-453
        • Nikpay M.
        • Goel A.
        • Won H.H.
        • Hall L.M.
        • Willenborg C.
        • Kanoni S.
        • et al.
        A comprehensive 1,000 Genomes-based genome-wide association meta-analysis of coronary artery disease.
        Nat. Genet. 2015; 47: 1121-1130
        • Wang F.
        • Xu C.Q.
        • He Q.
        • Cai J.P.
        • Li X.C.
        • Wang D.
        • et al.
        Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population.
        Nat. Genet. 2011; 43: 345-349
        • Guo C.Y.
        • Gu Y.
        • Li L.
        • Jia E.Z.
        • Li C.J.
        • Wang L.S.
        • et al.
        Association of SNP rs6903956 on chromosome 6p24.1 with angiographical characteristics of coronary atherosclerosis in a Chinese population.
        PLoS One. 2012; 7e43732
        • Meng J.
        • Tan W.
        • Zhu Y.
        • Wang F.
        • Li X.
        • Zhang M.
        A coronary artery disease-associated SNP rs6903956 contributed to asymptomatic hyperuricemia susceptibility in Han Chinese.
        Lipids Health Dis. 2015; 14: 33
        • Tayebi N.
        • Ke T.
        • Foo J.N.
        • Friedlander Y.
        • Liu J.
        • Heng C.K.
        Association of single nucleotide polymorphism rs6903956 on chromosome 6p24.1 with coronary artery disease and lipid levels in different ethnic groups of the Singaporean population.
        Clin. Biochem. 2013; 46: 755-759
        • Dechamethakun S.
        • Ikeda S.
        • Arai T.
        • Sato N.
        • Sawabe M.
        • Muramatsu M.
        Associations between the CDKN2A/B, ADTRP and PDGFD polymorphisms and the development of coronary atherosclerosis in Japanese patients.
        J. Atherosclerosis Thromb. 2014; 21: 680-690
        • Han Y.
        • Dorajoo R.
        • Chang X.
        • Wang L.
        • Khor C.C.
        • Sim X.
        • et al.
        Genome-wide association study identifies a missense variant at APOA5 for coronary artery disease in Multi-Ethnic Cohorts from Southeast Asia.
        Sci. Rep. 2017; 717921
        • Ooi D.S.Q.
        • Ong S.M.
        • Eng M.H.
        • Chan Y.H.
        • Lee Y.S.
        • Low A.F.H.
        • et al.
        Detection of ADTRP in circulation and its role as a novel biomarker for coronary artery disease.
        PLoS One. 2020; 15e0237074
        • Gallagher M.D.
        • Chen-Plotkin A.S.
        The post-GWAS era: from association to function.
        Am. J. Hum. Genet. 2018; 102: 717-730
        • Luo C.
        • Tang B.
        • Qin S.
        • Yuan C.
        • Du Y.
        • Yang J.
        GATA2 regulates the CAD susceptibility gene ADTRP rs6903956 through preferential interaction with the G allele.
        Mol. Genet. Genom. 2021; 296: 799-808
        • Torres J.M.
        • Gamazon E.R.
        • Parra E.J.
        • Below J.E.
        • Valladares-Salgado A.
        • Wacher N.
        • et al.
        Cross-tissue and tissue-specific eQTLs: partitioning the heritability of a complex trait.
        Am. J. Hum. Genet. 2014; 95: 521-534
        • Wang J Eric
        • Brandon Barbara
        • Hae Robert
        • et al.
        Imputing gene expression in uncollected tissues within and beyond GTEx.
        Am. J. Hum. Genet. 2016; 98: 697-708
        • Libby P.
        • Buring J.E.
        • Badimon L.
        • Hansson G.K.
        • Deanfield J.
        • Bittencourt M.S.
        • et al.
        Atherosclerosis. Nat Rev Dis Primers. 2019; 5: 56
        • Congrains A.
        • Kamide K.
        • Oguro R.
        • Yasuda O.
        • Miyata K.
        • Yamamoto E.
        • et al.
        Genetic variants at the 9p21 locus contribute to atherosclerosis through modulation of ANRIL and CDKN2A/B.
        Atherosclerosis. 2012; 220: 449-455
        • Zhou X.
        • Han X.
        • Wittfeldt A.
        • Sun J.
        • Liu C.
        • Wang X.
        • et al.
        Long non-coding RNA ANRIL regulates inflammatory responses as a novel component of NF-kappaB pathway.
        RNA Biol. 2016; 13: 98-108
        • Holdt L.M.
        • Hoffmann S.
        • Sass K.
        • Langenberger D.
        • Scholz M.
        • Krohn K.
        • et al.
        Alu elements in ANRIL non-coding RNA at chromosome 9p21 modulate atherogenic cell functions through trans-regulation of gene networks.
        PLoS Genet. 2013; 9e1003588
        • Lo Sardo V.
        • Chubukov P.
        • Ferguson W.
        • Kumar A.
        • Teng E.L.
        • Duran M.
        • et al.
        Unveiling the role of the most impactful cardiovascular risk locus through haplotype editing.
        Cell. 2018; 175 (e20): 1796-1810
        • Krause M.D.
        • Huang R.T.
        • Wu D.
        • Shentu T.P.
        • Harrison D.L.
        • Whalen M.B.
        • et al.
        Genetic variant at coronary artery disease and ischemic stroke locus 1p32.2 regulates endothelial responses to hemodynamics.
        Proc. Natl. Acad. Sci. U. S. A. 2018; 115: E11349-E11358
        • Gupta R.M.
        • Hadaya J.
        • Trehan A.
        • Zekavat S.M.
        • Roselli C.
        • Klarin D.
        • et al.
        A genetic variant associated with five vascular diseases is a distal regulator of endothelin-1 gene expression.
        Cell. 2017; 170: 522-533 e15
        • Lupu C.
        • Zhu H.
        • Popescu N.I.
        • Wren J.D.
        • Lupu F.
        Novel protein ADTRP regulates TFPI expression and function in human endothelial cells in normal conditions and in response to androgen.
        Blood. 2011; 118: 4463-4471
        • Cheung C.
        • Bernardo A.S.
        • Trotter M.W.
        • Pedersen R.A.
        • Sinha S.
        Generation of human vascular smooth muscle subtypes provides insight into embryological origin-dependent disease susceptibility.
        Nat. Biotechnol. 2012; 30: 165-173
        • Narmada B.C.
        • Goh Y.T.
        • Li H.
        • Sinha S.
        • Yu H.
        • Cheung C.
        Human stem cell-derived endothelial-hepatic Platform for efficacy testing of vascular-protective metabolites from nutraceuticals.
        Stem Cells Transl Med. 2017; 6: 851-863
        • Rueden C.T.
        • Schindelin J.
        • Hiner M.C.
        • DeZonia B.E.
        • Walter A.E.
        • Arena E.T.
        • et al.
        ImageJ2: ImageJ for the next generation of scientific image data.
        BMC Bioinf. 2017; 18: 529
        • Ang L.T.
        • Tan A.K.Y.
        • Autio M.I.
        • Goh S.H.
        • Choo S.H.
        • Lee K.L.
        • et al.
        A Roadmap for human liver differentiation from pluripotent stem cells.
        Cell Rep. 2018; 22: 2190-2205
        • Sloan C.A.
        • Chan E.T.
        • Davidson J.M.
        • Malladi V.S.
        • Strattan J.S.
        • Hitz B.C.
        • et al.
        ENCODE data at the ENCODE portal.
        Nucleic Acids Res. 2016; 44: D726-D732
        • Consortium E.P.
        • Moore J.E.
        • Purcaro M.J.
        • Pratt H.E.
        • Epstein C.B.
        • Shoresh N.
        • et al.
        Expanded encyclopaedias of DNA elements in the human and mouse genomes.
        Nature. 2020; 583: 699-710
        • Robinson M.D.
        • McCarthy D.J.
        • Smyth G.K.
        edgeR: a Bioconductor package for differential expression analysis of digital gene expression data.
        Bioinformatics. 2010; 26: 139-140
        • Ritchie M.E.
        • Phipson B.
        • Wu D.
        • Hu Y.
        • Law C.W.
        • Shi W.
        • et al.
        Limma powers differential expression analyses for RNA-sequencing and microarray studies.
        Nucleic Acids Res. 2015; 43: e47
        • Ashburner M.
        • Ball C.A.
        • Blake J.A.
        • Botstein D.
        • Butler H.
        • Cherry J.M.
        • et al.
        Gene ontology: tool for the unification of biology. The Gene Ontology Consortium.
        Nat. Genet. 2000; 25: 25-29
        • Subramanian A.
        • Tamayo P.
        • Mootha V.K.
        • Mukherjee S.
        • Ebert B.L.
        • Gillette M.A.
        • et al.
        Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles.
        Proc. Natl. Acad. Sci. U. S. A. 2005; 102: 15545-15550
        • Kramer A.
        • Green J.
        • Pollard Jr., J.
        • Tugendreich S.
        Causal analysis approaches in ingenuity pathway analysis.
        Bioinformatics. 2014; 30: 523-530
        • Durand N.C.
        • Shamim M.S.
        • Machol I.
        • Rao S.S.
        • Huntley M.H.
        • Lander E.S.
        • et al.
        Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments.
        Cell Syst. 2016; 3: 95-98
        • Taberlay P.C.
        • Achinger-Kawecka J.
        • Lun A.T.
        • Buske F.A.
        • Sabir K.
        • Gould C.M.
        • et al.
        Three-dimensional disorganization of the cancer genome occurs coincident with long-range genetic and epigenetic alterations.
        Genome Res. 2016; 26: 719-731
        • Hagege H.
        • Klous P.
        • Braem C.
        • Splinter E.
        • Dekker J.
        • Cathala G.
        • et al.
        Quantitative analysis of chromosome conformation capture assays (3C-qPCR).
        Nat. Protoc. 2007; 2: 1722-1733
        • Du M.
        • Wang L.
        3C-digital PCR for quantification of chromatin interactions.
        BMC Mol. Biol. 2016; 17: 23
        • Tousoulis D.
        • Oikonomou E.
        • Economou E.K.
        • Crea F.
        • Kaski J.C.
        Inflammatory cytokines in atherosclerosis: current therapeutic approaches.
        Eur. Heart J. 2016; 37: 1723-1732
        • Lupu C.
        • Patel M.M.
        • Lupu F.
        Insights into the functional role of ADTRP (Androgen-Dependent TFPI-regulating protein) in Health and disease.
        Int. J. Mol. Sci. 2021; 22
        • Luo C.
        • Wang F.
        • Ren X.
        • Ke T.
        • Xu C.
        • Tang B.
        • et al.
        Identification of a molecular signaling gene-gene regulatory network between GWAS susceptibility genes ADTRP and MIA3/TANGO1 for coronary artery disease.
        Biochim. Biophys. Acta, Mol. Basis Dis. 2017; 1863: 1640-1653
        • McLaren W.
        • Gil L.
        • Hunt S.E.
        • Riat H.S.
        • Ritchie G.R.
        • Thormann A.
        • et al.
        The Ensembl variant effect predictor.
        Genome Biol. 2016; 17: 122
        • Nakato R.
        • Wada Y.
        • Nakaki R.
        • Nagae G.
        • Katou Y.
        • Tsutsumi S.
        • et al.
        Comprehensive epigenome characterization reveals diverse transcriptional regulation across human vascular endothelial cells.
        Epigenet. Chromatin. 2019; 12: 77
        • Dixon J.R.
        • Selvaraj S.
        • Yue F.
        • Kim A.
        • Li Y.
        • Shen Y.
        • et al.
        Topological domains in mammalian genomes identified by analysis of chromatin interactions.
        Nature. 2012; 485: 376-380
        • Nora E.P.
        • Lajoie B.R.
        • Schulz E.G.
        • Giorgetti L.
        • Okamoto I.
        • Servant N.
        • et al.
        Spatial partitioning of the regulatory landscape of the X-inactivation centre.
        Nature. 2012; 485: 381-385
        • Rao S.S.
        • Huntley M.H.
        • Durand N.C.
        • Stamenova E.K.
        • Bochkov I.D.
        • Robinson J.T.
        • et al.
        A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping.
        Cell. 2014; 159: 1665-1680
        • Do P.T.
        • Nguyen C.X.
        • Bui H.T.
        • Tran L.T.N.
        • Stacey G.
        • Gillman J.D.
        • et al.
        Demonstration of highly efficient dual gRNA CRISPR/Cas9 editing of the homeologous GmFAD2-1A and GmFAD2-1B genes to yield a high oleic, low linoleic and alpha-linolenic acid phenotype in soybean.
        BMC Plant Biol. 2019; 19: 311
        • Haapaniemi E.
        • Botla S.
        • Persson J.
        • Schmierer B.
        • Taipale J.
        CRISPR-Cas9 genome editing induces a p53-mediated DNA damage response.
        Nat. Med. 2018; 24: 927-930
        • Mootha V.K.
        • Lindgren C.M.
        • Eriksson K.F.
        • Subramanian A.
        • Sihag S.
        • Lehar J.
        • et al.
        PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes.
        Nat. Genet. 2003; 34: 267-273
        • Joseph P.G.
        • Pare G.
        • Anand S.S.
        Exploring gene-environment relationships in cardiovascular disease.
        Can. J. Cardiol. 2013; 29: 37-45
        • Talmud P.J.
        Gene-environment interaction and its impact on coronary heart disease risk.
        Nutr. Metabol. Cardiovasc. Dis. 2007; 17: 148-152
        • McPherson R.
        • Tybjaerg-Hansen A.
        Genetics of coronary artery disease.
        Circ. Res. 2016; 118: 564-578
        • Samani N.J.
        • Deloukas P.
        • Erdmann J.
        • Hengstenberg C.
        • Kuulasmaa K.
        • McGinnis R.
        • et al.
        Large scale association analysis of novel genetic loci for coronary artery disease.
        Arterioscler. Thromb. Vasc. Biol. 2009; 29: 774-780
        • Zhang J.
        • Ma H.
        • Gao J.
        • Kong S.
        • You J.
        • Sheng Y.
        Variants in the CXCL12 gene was associated with coronary artery disease susceptibility in Chinese Han population.
        Oncotarget. 2017; 8: 54518-54527
        • Wang X.
        • Cairns M.J.
        • Yan J.
        Super-enhancers in transcriptional regulation and genome organization.
        Nucleic Acids Res. 2019; 47: 11481-11496
        • Simonis M.
        • Kooren J.
        • de Laat W.
        An evaluation of 3C-based methods to capture DNA interactions.
        Nat. Methods. 2007; 4: 895-901
        • McCord R.P.
        • Kaplan N.
        • Giorgetti L.
        Chromosome conformation capture and beyond: toward an integrative view of chromosome structure and function.
        Mol. Cell. 2020; 77: 688-708
        • Hebbel R.P.
        Blood endothelial cells: utility from ambiguity.
        J. Clin. Invest. 2017; 127: 1613-1615
        • Blann A.D.
        • Woywodt A.
        • Bertolini F.
        • Bull T.M.
        • Buyon J.P.
        • Clancy R.M.
        • et al.
        Circulating endothelial cells. Biomarker of vascular disease.
        Thromb. Haemostasis. 2005; 93: 228-235
        • Chioh F.W.
        • Fong S.W.
        • Young B.E.
        • Wu K.X.
        • Siau A.
        • Krishnan S.
        • et al.
        Convalescent COVID-19 patients are susceptible to endothelial dysfunction due to persistent immune activation.
        Elife. 2021; : 10
        • Spielmann M.
        • Mundlos S.
        Looking beyond the genes: the role of non-coding variants in human disease.
        Hum. Mol. Genet. 2016; 25: R157-R165
        • Tan W.L.W.
        • Anene-Nzelu C.G.
        • Wong E.
        • Lee C.J.M.
        • Tan H.S.
        • Tang S.J.
        • et al.
        Epigenomes of human hearts reveal new genetic variants relevant for cardiac disease and phenotype.
        Circ. Res. 2020; 127: 761-777
        • Sun J.H.
        • Zhou L.
        • Emerson D.J.
        • Phyo S.A.
        • Titus K.R.
        • Gong W.
        • et al.
        Disease-associated short tandem repeats Co-localize with chromatin domain boundaries.
        Cell. 2018; 175: 224-238 e15
        • Zhang T.
        • Zhang Z.
        • Dong Q.
        • Xiong J.
        • Zhu B.
        Histone H3K27 acetylation is dispensable for enhancer activity in mouse embryonic stem cells.
        Genome Biol. 2020; 21: 45
        • Hnisz D.
        • Shrinivas K.
        • Young R.A.
        • Chakraborty A.K.
        • Sharp P.A.
        A phase separation model for transcriptional control.
        Cell. 2017; 169: 13-23
        • Sjaarda J.
        • Gerstein H.
        • Chong M.
        • Yusuf S.
        • Meyre D.
        • Anand S.S.
        • et al.
        Blood CSF1 and CXCL12 as causal mediators of coronary artery disease.
        J. Am. Coll. Cardiol. 2018; 72: 300-310
        • Tavakolian Ferdousie V.
        • Mohammadi M.
        • Hassanshahi G.
        • Khorramdelazad H.
        • Khanamani Falahati-Pour S.
        • Mirzaei M.
        • et al.
        Serum CXCL10 and CXCL12 chemokine levels are associated with the severity of coronary artery disease and coronary artery occlusion.
        Int. J. Cardiol. 2017; 233: 23-28
        • Döring Y.
        • Vorst EPCvd
        • Duchene J.
        • Jansen Y.
        • Gencer S.
        • Bidzhekov K.
        • et al.
        CXCL12 derived from endothelial cells promotes atherosclerosis to drive coronary artery disease.
        Circulation. 2019; 139: 1338-1340
        • Zernecke A.
        • Weber C.
        Chemokines in atherosclerosis.
        Arterioscler. Thromb. Vasc. Biol. 2014; 34: 742-750
        • Wang Y.
        • Huang J.
        • Li Y.
        • Yang G.Y.
        Roles of chemokine CXCL12 and its receptors in ischemic stroke.
        Curr. Drug Targets. 2012; 13: 166-172
        • Gencer S.
        • Evans B.R.
        • van der Vorst E.P.C.
        • Doring Y.
        • Weber C.
        Inflammatory chemokines in atherosclerosis.
        Cells. 2021; 10
        • Gao J.H.
        • Yu X.H.
        • Tang C.K.
        CXC chemokine ligand 12 (CXCL12) in atherosclerosis: an underlying therapeutic target.
        Clin. Chim. Acta. 2019; 495: 538-544
        • Farinacci M.
        • Krahn T.
        • Dinh W.
        • Volk H.D.
        • Dungen H.D.
        • Wagner J.
        • et al.
        Circulating endothelial cells as biomarker for cardiovascular diseases.
        Res Pract Thromb Haemost. 2019; 3: 49-58
        • Hill J.M.
        • Zalos G.
        • Halcox J.P.
        • Schenke W.H.
        • Waclawiw M.A.
        • Quyyumi A.A.
        • et al.
        Circulating endothelial progenitor cells, vascular function, and cardiovascular risk.
        N. Engl. J. Med. 2003; 348: 593-600
        • Nadar S.K.
        • Lip G.Y.
        • Lee K.W.
        • Blann A.D.
        Circulating endothelial cells in acute ischaemic stroke.
        Thromb. Haemostasis. 2005; 94: 707-712
        • Schmidt D.E.
        • Manca M.
        • Hoefer I.E.
        Circulating endothelial cells in coronary artery disease and acute coronary syndrome.
        Trends Cardiovasc. Med. 2015; 25: 578-587
        • Davis C.M.
        • Fairbanks S.L.
        • Alkayed N.J.
        Mechanism of the sex difference in endothelial dysfunction after stroke.
        Transl Stroke Res. 2013; 4: 381-389
        • Vinas J.L.
        • Porter C.J.
        • Douvris A.
        • Spence M.
        • Gutsol A.
        • Zimpelmann J.A.
        • et al.
        Sex diversity in proximal tubule and endothelial gene expression in mice with ischemic acute kidney injury.
        Clin. Sci. (Lond.). 2020; 134: 1887-1909
        • Machiela M.J.
        • Chanock S.J.
        LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants.
        Bioinformatics. 2015; 31: 3555-3557