Reconstructing the blood metabolome and genotype using long-range chromatin interactions

Tayaza Fadason, William Schierding, Nikolai Kolbenev, Jiamou Liu, John R. Ingram, Justin M. O’SullivanbioRxiv 2019 · DOI

TL;DR

This study integrates multiple data types, including 3D genome structure (Hi-C) and gene expression data (eQTLs), to link genetic variations (SNPs) associated with blood metabolites to the genes they regulate. This approach significantly expands our understanding of how genetic variations influence metabolic pathways, revealing novel gene-metabolite associations and potential therapeutic targets.

What Problem Does This Paper Address?

Metabolic pathways are crucial for human health, but the genetic influences on these pathways are often poorly understood. Previous studies, like genome-wide association studies (GWAS), have identified genetic variants (SNPs) associated with blood metabolite levels. However, a significant portion of these SNPs are located in non-coding regions of the genome, making it challenging to pinpoint the specific genes they regulate. Traditional methods often rely on proximity, which is insufficient given the growing evidence for long-range gene regulatory interactions mediated by 3D genome structure, such as chromatin looping. This paper aims to bridge this gap by systematically identifying the genes regulated by metabolite-associated SNPs, even when they are physically distant, by leveraging insights from the 3D organization of the genome.

Methods

The researchers employed a multi-pronged approach:

  • Chromatin Interaction Mapping (Hi-C): They used Hi-C data from eight human cell lines to identify which genomic regions (fragments) are in physical proximity to the metabolite-associated SNPs. This is a key technique for understanding the 3D organization of the genome and how distant regulatory elements can interact with genes.
  • Expression Quantitative Trait Loci (eQTL) Analysis: They integrated eQTL data from the Genotype-Tissue Expression (GTEx) project across 48 human tissues. eQTLs link genetic variants to gene expression levels, helping to identify genes that are regulated by the identified SNPs.
  • Functional Annotation: To understand the biological relevance of the identified genes, they used several databases and methods:
  • Gene Ontology (KEGG): To classify genes into metabolic and other biological pathways.
  • Protein Classification (Protein Atlas): To categorize genes based on protein function.
  • Drug Target Analysis (DGIdb): To identify genes whose products are targeted by existing drugs.
  • Literature Text Mining (PubMed): To find existing scientific literature supporting the identified SNP-gene-metabolite associations.

Datasets

  • Metabolite-associated SNPs: Data from Shin et al. (2014) identifying 145 genome-wide significant SNPs associated with blood metabolites.
  • Chromatin Interaction Data: Hi-C data from eight human cell lines (GM12878, HeLa, HMEC, HUVEC, IMR90, K562, KBM7, and NHEK).
  • eQTL Data: Genotype-Tissue Expression (GTEx) v7 data across 48 human tissues.
  • Gene Ontology Database: Kyoto Encyclopedia of Genes and Genomes (KEGG).
  • Protein Classification Database: The Human Protein Atlas.
  • Drug Target Database: Drug Gene Interaction database (DGIdb).
  • Literature Database: MEDLINE via PubMed APIs.

Key Insights and Ideas

  • 3D Genome Structure is Crucial: The study strongly emphasizes that understanding long-range chromatin interactions is essential for accurately linking genetic variants to the genes they regulate, especially for non-coding SNPs.
  • Expanding the Regulatory Landscape: By incorporating Hi-C data, they significantly increased the number of metabolite-associated SNPs linked to regulated genes (eQTLs), more than doubling the findings from previous proximity-based approaches.
  • Metabolism is Highly Networked: The identified genes are not only enriched in metabolic pathways but also in broader categories like disease development and plasma proteins, highlighting the interconnectedness of metabolic regulation.
  • Drugability of Metabolic Genes: A substantial portion of the identified genes are targets for existing drugs, suggesting potential for developing new therapies for metabolic disorders.
  • Bridging Knowledge Gaps: The integration of diverse data sources, particularly the literature mining, helps to connect previously disparate pieces of information and identify novel gene-metabolite relationships.

Key Results

  • They identified 130 (~90%) of the 145 metabolite-associated SNPs as marking eQTLs, a significant increase from the ~39% reported previously.
  • These 130 SNPs (eSNPs) were associated with the expression of 577 genes (eGenes) through 612 unique eSNP-eGene pairs across 48 human tissues.
  • Altogether, they identified 2,757 eSNP-eGene interactions, with 2,136 being novel discoveries.
  • The eGenes were significantly enriched in metabolism-related pathways (37.8%), as well as in categories related to enzymes, plasma proteins, disease development, and potential drug targets.
  • 147 (25.5%) of the identified eGenes encode products targeted by at least one drug. Notably, 101 of these druggable eGenes had not been previously linked to genetic variants associated with metabolism.
  • Literature text mining supported 41 eGene-metabolite pairs, with 14 of these being novel findings not reported by Shin et al.

Limitations and Future Work

  • Literature Support Stringency: The text mining approach had stringent criteria (requiring terms in title/abstract), which may have led to missing some valid associations.
  • Unknown Metabolites: The identity of some associated metabolites was not always known, which could limit the interpretation of some findings.
  • Tissue Specificity: While the study analyzed 48 tissues, the eSNP-eGene associations were found to be tissue-specific, suggesting that a comprehensive understanding would require even broader tissue coverage.
  • Future Directions: The authors suggest that these novel SNP-gene-metabolite associations provide a valuable resource for further empirical testing to delineate the roles of these genes in metabolism and for investigating unknown metabolites. The identified druggable genes also offer promising avenues for drug development for metabolic diseases.

Summary updated

Abstract

Abstract The mechanisms of metabolism comprise a large number of biochemical pathways with a myriad of poorly characterised genetic influences. In this study, we perform a systematic integration of chromatin interaction (Hi-C), expression quantitative trait loci (eQTL), gene ontology, drug interaction, and literature-supported connections to deconvolute the genetic regulatory influences of 145 blood metabolite-associated single nucleotide polymorphisms (SNPs). We identify 577 genes that are regulated via chromatin looping to 130 distal and proximal SNPs across 48 different human tissues. The affected genes are enriched in categories that include metabolism, enzymes, plasma proteins, disease development, and potential drug targets. These novel SNP-gene-metabolite associations are a valuable resource for understanding the molecular mechanisms guiding pathologic metabolite levels in human tissues, and for further investigation into disease diagnosis and therapy.
BibTeX
@misc{fadason2019reconstructing,
  author = {Tayaza Fadason and William Schierding and Nikolai Kolbenev and Jiamou Liu and John R. Ingram and Justin M. O’Sullivan},
  title = {{Reconstructing the blood metabolome and genotype using long-range chromatin interactions}},
  year = {2019},
  doi = {10.1101/656132},
  howpublished = {bioRxiv},
}
Figure