Recent Highlights of ATVB: Functional Genomics and CRISPR Applied to Cardiovascular Research and Medicine

Introduction

The growing knowledge of the human genome sequence has spurred the interest in functional genomics.¹ Functional genomics aims to understand the complex relationship between genotype and phenotype. Using high-throughput methods, functional genomics has expanded biomedical science beyond reductionist approaches and strive to gain an unbiased view of the regulation of gene transcription, translation and protein-protein interactions on a genomic scale. A more focused functional genomics approach determines the functional impact of genomic variation, such as mutations and polymorphisms associated with complex traits and diseases. The endeavor of functional genomics is fueling an explosion of new insights into novel biology and therapeutic strategies.²

The purpose of this Recent Highlights article is to provide a concise but comprehensive overview of recent studies in the journal of Arteriosclerosis, Thrombosis, and Vascular Biology on functional genomics applied to cardiovascular research and medicine, while also highlighting cutting edge functional genomic tools and their applications.

Transcriptome and proteome profiling for characterization of molecular signature and discovery of novel biology

Discovering the molecular signature associated with a cellular or disease state by transcriptomic, proteomic, and metabolomic profiling is one of the main quests in functional genomic studies. Integrating these “omics” data not only allows for an unbiased interpretation of biological processes, but also facilitates hypothesis generation and target prioritization for deeper mechanistic studies.

Molecular profiling of specific cell types in tissue specimen provides insights into cell-type specific mechanisms in disease pathogenesis. Macrophages are intricately involved in atherogenesis, but how macrophages in vulnerable plaques differ from those in stable plaques remains undetermined.³ Microarray analysis of macrophages laser captured from symptomatic and asymptomatic human atherosclerotic plaques reveal differentially expressed genes.³ The transcriptome feature of symptomatic plaque macrophages corresponded to the upregulation of 7 functional pathways, including inflammation, lipid metabolism, hypoxic response, cell proliferation, apoptosis, antigen presentation, and cellular energetics, providing the transcriptomic signature of human plaque macrophages.³ Proteomic profiling of vascular cells in the entire vascular beds including macro- and micro- blood vessels and capillary vascular beds allows the characterization of the systemic nature of vascular diseases. Serra et al. developed whole-body differential perfusion with increasing concentrations of detergent buffer to selectively solubilize distinct layers of vascular bed tissue, including endothelial glycocalyx, endothelial cells, and vascular smooth muscle cells (VSMC). By incorporating quantitative proteomic analysis, the method allows for the detection of system-wide proteomic change in vascular beds in vivo.⁴ Multi-omics profiling, either at bulk or single cell level, that also captures spatial organization and function of specific cell types in tissue specimen will undoubtedly provide a more complete picture of diverse tissues and organisms.⁵

RNA-sequencing is widely used to discover novel transcripts and identify differentially expressed transcripts for functional follow-up. Dnm3os, one of the top differentially expressed long non-coding RNAs (lncRNAs) upregulated in bone marrow-derived macrophages from type 2 diabetic db/db mice, has been discovered as a positive regulator of macrophage inflammatory responses.⁶ LncRNA MYOSLID was highly induced in VSMC by overexpression of MYOCD – the master regulator for VSMC differentiation, and has been found to promote VSMC differentiation while inhibiting proliferation.⁷ LncRNA MALAT1 was repressed by increased matrix stiffness in cultured human VSMC. Knockdown of MALAT1 limited stiffness-induced proliferation and migration of human VSMC, and knockout of Malat1 was protective in in vivo vascular injury response.⁸ Despite success, how to prioritize potential functional lncRNAs from RNA-seq screen for further workup remains the key challenge.⁹ Detailed discussion on the decision scheme for the prioritization pipeline in published work and efforts to potentially establish standardized pipelines for scalable hypothesis creation will facilitate efficient prioritization and discovery of novel functional transcripts.

Interpreting GWAS findings with functional genomic data for prioritization of causal variants and genes

Genome-wide association studies (GWAS) have successfully identified a large number of genetic loci associated with risks of complex traits and cardiovascular diseases.¹⁰ Determining the molecular functions of these loci will provide new insights into novel biology and therapeutic targets.¹ The majority of genetic variants associated with complex diseases are found within noncoding regions of the genome, highlighting the necessity of understanding regulatory variation.

Consortium projects, such as ENCODE (Encyclopedia of DNA Elements),¹¹ the Epigenomics Roadmap,¹² and FANTOM (Functional Annotation of the Mammalian Genome),¹³ have focused on producing large catalogs of functional elements in the genome that are used to annotate the putative regulatory function of genetic variants. The GTEx (Genotype-Tissue Expression) project allows eQTL (expression quantitative trait locus) analysis that associates genotypes to the levels and directionality of gene expression in population samples. Indeed, disease- and trait-associated genetic variants are enriched in tissue-specific epigenetic marks,¹² and eQTL variants in disease-relevant tissues and cell types are overrepresented among GWAS loci.^{14, 15}

Candidate causal genes at each of the GWAS loci are generally prioritized based on proximity to the lead SNP (single nucleotide polymorphism) by fine mapping, overlap with epigenetic regulatory marks, eQTL data, and long-range chromatin interactions of variants with gene promoters by chromosome conformation capture-based techniques. In addition, statistical colocalization analysis to estimate whether the variants for GWAS and eQTL signals are shared provides a stronger hypothesis that the gene expression change indicated by the eQTL is causally related to the GWAS.¹⁶ Additional tools, such as DEPICT (Data-Driven Expression-Prioritized Integration for Complex Traits)¹⁷ and FUMA (Functional Mapping and Annotation of GWAS)¹⁸ are also being developed and applied to prioritize variants and genes from known GWAS loci for functional follow-up.

There are continued efforts to detect novel gene-trait association. Importantly, most GWAS have mainly focused on populations of European ancestry. The inclusion of multiethnic and admixed populations will provide additional power to detect associations for genetic variants segregating at low frequency in European populations.¹⁹ A recent GWAS in participants of Han ethnicity has identified 3 novel intragenic SNPs associated with CAD (coronary artery diseases) and has shown that knockdown of the host genes, SCML4 and THSD7A, affected endothelial function and monocyte adhesion.²⁰ Furthermore, miRNAs²¹ and lncRNAs²² harboring GWAS variants have shown functional effects on trait-related cellular phenotypes, supporting their potential roles as the causal non-coding RNAs. In addition to human GWAS, QTL mapping in model organism using the inbred mice has successfully identified novel genetic variation and candidate genes regulating macrophage cholesterol metabolism,²³ efferocytosis,²⁴ apoptotisis,²⁵ and atherosclerosis susceptibility.^{24, 25} QTL studies in mice have a high degree of concordance with mapped phenotype regions in the human genome but require far fewer resources and time, complementing observations in human populations.²⁶

Although there is no guaranteed algorithm or community-curated standard to map GWAS variants to causal genes, fine mapping and other complementary functional genomic tools have provided first clue on the potential causal variants and genes, and the directionality of the effects. (Figure 1) Further validation in suitable experimental models is required to establish the causality and the functional mechanisms of the gene-to-disease association. The following section discusses examples of these efforts.

Figure 1. Interpreting GWAS findings with functional genomic tools for novel biology and translation.

Functional genomic data are essential for the functional annotation and prioritization of potential causal variants and genes. Perturbation of the candidate variants and genes in relevant tissues and cell types or model organisms establishes the causality and biological mechanisms of the variant-disease association. Integrating “omics” profiling facilitates hypothesis generation and mechanistic studies. ASE, allele-specific expression; CRISPR, clustered regularly interspaced short palindromic repeats; eQTL, expression quantitative trait loci; GWAS, genome-wide association studies; iPSC, induced pluripotent stem cells.

Interrogating the functional impact of candidate variants and genes

A noncoding variant can affect gene expression via epigenetic mechanisms, but may also serve as a proxy that links to other non-synonymous coding variants responsible for their causal effects on candidate gene expression. Perturbation of the genetic variants and gene-of-interest in relevant tissues and cell types or model organisms would therefore reveal the functional consequences of candidate variants and genes.

rs17514846, the lead SNP associated with CAD, is located within the intronic region of the FURIN gene.²⁷ Leukocytes from individuals carrying the risk allele (A) of rs17514846 show higher FURIN expression. An analysis of isogenic THP-1 monocytic cell lines created by CRISPR (clustered regularly interspaced short palindromic repeats)-mediated genome editing further supported that isogenic cells with the A/A genotype for rs17514846 had higher FURIN abundance than the isogenic cells with the C/C genotype. Lentivirus-mediated overexpression of Furin in RAW264.7 cells promoted monocyte/macrophage migration and proliferation while inhibiting apoptosis, providing a biological mechanism for the association between genetic variants in FURIN and CAD risk.²⁷ CAD-associated lead variant rs2487928 lies within the intronic regions of JCAD and is associated with JCAD expression in atherosclerotic arteries.²⁸ JCAD protein localizes to endothelial cell junctions and disruption of JCAD inhibited angiogenesis in vitro and in vivo,²⁹ At molecular level, JCAD interacts with large tumor suppressor kinase 2 and negatively regulates Hippo signaling leading to increased activity of Yes-associated protein, the transcriptional effector of the pathway.²⁸ It is, however, unknown whether rs2487928 is the causal variant without perturbation experiment.

rs9390459 is a synonymous variant in STXBP5 and the lead SNP associated with altered plasma von Willebrand factor concentration.³⁰ However, functional genomic data do not support rs9390459 as a non-coding variant that affect DNA epigenetic modification, transcriptional factor binding, distal interaction, or eQTL. Instead, a non-synonymous SNP rs1039084, in high LD (linkage disequilibrium) with the lead SNP rs9390459, is the causal variant for a decreased thrombotic phenotype.³⁰ Zhu et al. knocked the rs1039084 minor allele of human STXBP5 into the orthologous mouse Stxbp5 locus by CRISPR and found that mice carrying the minor allele showed lower plasma von Willebrand factor concentration, prolonged bleeding and decreased thrombosis, phenotypes consistent with human minor allele carriers.³⁰ Morris et al. also suggested that non-synonymous variant rs1051338, in high LD with the CAD lead SNP in LIPA gene, is associated with lower LIPA expression and enzymatic activity in lysosomes of human monocyte-derived macrophages,³¹ but causal modeling has not been performed.³²

rs12740374 has previously been identified as the causal variant responsible for the association of the SORT1 locus with low-density lipoprotein cholesterol and CAD.³³ rs12740374 is a noncoding SNP that lies ~120 kb away from the promoter of the SORT1 gene.³³ The minor allele of rs12740374 creates a binding site for CCAAT-enhancer-binding protein transcription factors, resulting in liver-specific transcriptional activation of SORT1.³³ Wang et al. further demonstrated a direct causal link between rs12740374 and SORT1 by perturbing the site of the SNP and evoking an alteration of SORT1 expression.³⁴ CRISPR-targeting of the rs12740374 minor allele sequence in the liver of a locus-humanized transgenic mouse model results in a profound reduction of hepatic SORT1 protein expression, supporting the causality of the variant.³⁴

Much progress has been made but it remains challenging to conclusively confirm the causal variants, causal genes, and the biological mechanisms of gene-to-disease association. Integrating state-of-the-art tools, such as human induced pluripotent stem cell (iPSC) technology and CRISPR genome editing techniques, will maximize the value of functional genomic data and accelerate its translation.

Human induced pluripotent stem cell differentiation for functional interrogation of GWAS variants and candidate genes

The differentiation of human iPSC to cell types of relevance to cardiovascular biology and diseases offers a powerful platform with enormous potential for disease modeling, drug screening and cell therapeutics. The combination of human iPSC technology with genome editing technique allows the generation of isogenic cell lines that differ in single genetic changes for causal modeling of candidate variants and genes, offering a new tool linking genotypes to phenotypes in the study of human cell biology and disease. (Figure 1)

We have previously generated iPSC-derived macrophages with knockout of the LIPA gene in human macrophages.³⁵ Macrophages with loss-of-function of LIPA showed expected phenotype of impaired hydrolysis of cholesteryl ester and lysosomal cholesteryl ester accumulation, but the expression of ATP-binding cassette transporter ABCA1 and cholesterol efflux capacity were not affected.³⁵ Impaired ABCA1 expression and function were, however, observed in skin fibroblasts derived from Cholesteryl Ester Storage Disease patients with loss-of-function mutations of LIPA,³⁵ suggesting potential cell type-specific differences in LIPA metabolic phenotypes and the value of iPSC-derived cells in the study of human cell biology.

Since iPSC differentiation provides an unlimited source of subject genotype-specific cells, iPSC-derived hepatocytes-like cells^{36, 37} and adipocytes³⁷ from a cohort of healthy subjects have been used to create a hepatocyte- and adipocyte-specific eQTL database that is complementary to traditional eQTL studies using bulk tissues. The iPSC-hepatocyte cohort successfully identify a large number of eQTL genes shared with the GTEx liver cohort, as well as many new eQTL genes. The iPSC-hepatocytes, however, poorly model the strong eQTL for rs12740374 genotype and SORT1 expression observed in both whole-liver samples and primary human hepatocytes.³⁴ This suggests that iPSC-derived cells can be useful for establishing causal SNP-gene relationships for many, but not all loci. The eQTL database of iPSC-derived cells provides a useful resource for planning experiments to model these relationships.³⁴

ATVB has recently published an “ATVB in Focus” series on iPSC-derived cell types relevant to atherothrombosis. The series includes 6 review articles on the differentiation, characterization and application of iPSC-derived endothelial cells,^{38, 39} VSMC,⁴⁰ macrophages,⁴¹ megakaryocytes and platelets,⁴² and hepatocytes.⁴³ We therefore refer to these contributions to guide further reading.

The expanding CRISPR toolbox for genome engineering

The advent of CRISPR genome editing techniques has greatly revolutionized functional genomic research. (Figure 1) CRIPSR/Cas9 (CRISPR Associated Protein 9) system permits rapid generation of novel animal models with a specific genetic background in one-step by co-injecting Cas9 mRNA, guide RNA, and DNA repair templates to zygotes.⁴⁴ The microinjection can be done in C57BL/6 zygote, a strain commonly used in the studies of atherosclerosis and metabolic diseases,⁴⁵ eliminating the need for extensive backcrossing.

Wang et al. generated intermedin knockout mice using CRIPSR and found that intermedin induced quiescent endothelial cells to proliferate, resulting in continuous vascular lumen expanding and a more effective blood perfusion.⁴⁶ Because of founder mosaicism in CRISPR-derived mouse model, it is recommended that each F1 mouse is treated as the start of an individual mutant line. This study has also demonstrated that multiple mutant lines have the same phenotypes.⁴⁶ Mouse models with fluorescent tagging of endogenous proteins can be useful to spatiotemporally track protein expression, localization, and binding activity in vivo. Lyu et al. generated a mouse model with CRISPR-mediated knock-in of epitope tags (carboxyl-terminal 3×FLAG or 3×HA) at the C-terminal end of the Myocd locus,⁴⁷ which allows for detection of nuclear localization of Myocd and associated DNA binding.⁴⁷

Somatic editing with CRISPR has also been established. Jarrett et al. developed a new method for generating an atherosclerosis mouse model through somatic deletion of Ldlr in liver of adult mice by single injection of adeno-associated viral vectors expressing Cas9 and a guide RNA targeting the Ldlr gene.⁴⁸ The edited mice were compared with germline Ldlr^−/− mice, and mice with overexpression of PCSK9 (proprotein convertase subtilisin/kexin type 9) that promotes degradation of low-density lipoprotein receptor in the liver,^{49, 50} models currently used to study atherosclerosis. Notable sexual dimorphism was observed, wherein somatic editing of Ldlr was superior for Ldlr removal in male mice, PCSK9 overexpression was more effective in female mice.⁴⁸ Both models provide useful alternatives to the use of Ldlr^−/− mice in atherosclerosis research.

The CRISPR toolbox is expanding. By fusing catalytically inactive Cas9 (dCas9) protein to cytosine deaminase domain, this “base editor” can convert cytosine bases to thymine bases in the genomic DNA near the guide RNA target site, allowing precise knock-in of specific nucleotide changes without inducing double-strand breaks.⁵¹ Chadwick et al. introduced site-specific nonsense mutations into the PCSK9 gene in the liver of adult mice, resulting in substantially decreased plasma PCSK9 protein concentrations by >50% and reduced plasma cholesterol concentrations by ~30%, demonstrating the ability to precisely introduce therapeutically relevant nucleotide variants into the genome in somatic tissues of adult mammals.⁵² By fusing different transcription regulators to the dCas9 protein, CRISPRi and CRISPRa can induce transcriptional silencing and activation, respectively.⁵³ CRISPRi allows for precise perturbation of lncRNA transcription without eliminating the lncRNA genes, therefore well suited for the study of lncRNA function. Chowdhury et al. used CRISPRi to knockdown a natural antisense transcripts from the tie1 locus – tie1 antisense (tie1AS), and found that knockdown of tie1AS led to increased tie1 mRNA abundance and tie1-mediated vascular phenotypes in zebrafish.⁵⁴

In addition to targeted genome editing, CRISPR has been harnessed to achieve large-scale functional screens. The ultimate aim for such screens is to identify novel genes and genetic variants that influence a specific phenotype in an unbiased fashion with genome-wide coverage. The application of CRISPR screen has been extensively covered in other review articles,⁵⁵ and we expect CRISPR screens that annotate gene function at scale to accelerate many research projects in the field of cardiovascular science.

Summary

Novel genetic loci associated with cardiovascular diseases will continue to emerge through GWAS in diverse populations. The post-GWAS functional studies will dedicate to delineating the causal genetic variants and genes, and the underlying biological mechanisms. Novel functional genomic technologies at single-cell and single-molecule level will further reveal cell type-specific regulatory mechanisms.⁵⁶ The promise of functional genomics is to understand the function of all components of the genome, and ultimately, translate genetic findings toward clinical applications in the diagnosis and treatment of both rare and common cardiovascular diseases.