Identification of Common and Rare Genetic Variation Associated With Plasma Protein Levels Using Whole-Exome Sequencing and Mass Spectrometry

Abstract

Background: Identifying genetic variation associated with plasma protein levels, and the mechanisms by which they act, could provide insight into alterable processes involved in regulation of protein levels. Although protein levels can be affected by genetic variants, their estimation can also be biased by missense variants in coding exons causing technical artifacts. Integrating genome sequence genotype data with mass spectrometry-based protein level estimation could reduce bias, thereby improving detection of variation that affects RNA or protein metabolism.

Methods: Here, we integrate the blood plasma protein levels of 664 proteins from 165 participants of the Tromsø Study, measured via tandem mass tag mass spectrometry, with whole-exome sequencing data to identify common and rare genetic variation associated with peptide and protein levels (protein quantitative trait loci [pQTLs]). We additionally use literature and database searches to prioritize putative functional variants for each pQTL.

Results: We identify 109 independent associations (36 protein and 73 peptide) and use genotype data to exclude 49 (4 protein and 45 peptide) as technical artifacts. We describe 2 particular cases of rare variation: 1 associated with the complement pathway and 1 with platelet degranulation. We identify putative functional variants and show that pQTLs act through diverse molecular mechanisms that affect both RNA and protein metabolism.

Conclusions: We show that although the majority of pQTLs exert their effects by modulating RNA metabolism, many affect protein levels directly. Our work demonstrates the extent by which pQTL studies are affected by technical artifacts and highlights how prioritizing the functional variant in pQTL studies can lead to insights into the molecular steps by which a protein may be regulated.

Keywords: exons; genotype; plasma; proteomics; quantitative trait loci.

Figure 1.

Study overview: 165 individuals from The Tromsø Study were followed from 1994–2013. Between 1994 and 1995, blood plasma and whole blood were collected; blood plasma and whole blood were processed and subsequently used for protein quantification by mass spectrometry and whole exome sequencing, respectively. These analyses identified 5,608 peptides and 664 proteins from plasma, and 3,148,863 variants from whole blood, across all individuals.

Figure 2.

Description of protein and genotype data: (A) Cumulative distribution plot showing the number of peptides identified in at least N samples. 5,052 peptides were identified in at least 8 samples (blue), 3,394 peptides were identified in at least 82 samples (red), and 1,430 peptides were identified in all 165 samples (green). (B) Histogram showing the number of peptides identified for each of the 664 parent proteins. A mean of 8.45 peptides per parent protein were identified (dotted line). (C) Bar plot showing q-values from Reactome pathway analysis of the significantly enriched top level groups in the Reactome event hierarchy. The significance threshold of −log₁₀(0.05) is shown by the red dotted line. (D) Histogram of the minor allele frequencies in this study for all 3,148,863 genetic variants identified across individuals. (E) Bar plot of the number of identified genetic variants within each SnpEff annotation. The number of variants with each annotation is also listed next to each bar.

Figure 3.

Pathways identified from rare variation analyses: (A) An overview of the lectin complement pathway showing the relationship between FCN3 (Ficolin 3; teal) and the complement pathway. Nominal p-values are shown for the association between rare variation at the FCN3 locus and levels of the complement pathway proteins. C4, C3, C5, C8, and C6 were associated at a nominal P < 0.05 (purple), C2, C9, C7, or C5b were not associated (gray). (B) STRING database diagram of the five proteins associated with rare SERPINA1 variation (each labeled with their nominal association p-value). Connections between proteins are colored based on their evidence (see legend and STRING documentation).

Figure 4.

Putative functional variant analyses: (A) Cartoon illustrating the genomic locations of variants with particular annotations and mechanisms, relative to the gene body of the pQTL. For example, indel annotated variants were only located within gene exons, but variants that have an underlying mechanism of “isoform” could be found in introns, exons, or the 3’ UTR. The three pQTLs where the PFV was a large genic deletion are not illustrated. (B) Stacked bar plot of the number of PFVs associated with each mechanism, subset by whether the mechanism affects the RNA molecule or the protein directly. (C) Stacked bar plot of the number of PFVs with each SnpEff annotation, subset by whether the PFVs’ mechanism affects the RNA molecule, the protein directly, or is unknown. (D) Stacked bar plot of the number of PFVs that were eQTLs in GETx, subset by whether the PFVs’ mechanism affects the RNA molecule, the protein directly, or is unknown.