A genome-wide association study of mass spectrometry proteomics using a nanoparticle enrichment platform

Abstract

Most studies to date of protein quantitative trait loci (pQTLs) have relied on affinity proteomics platforms, which provide only limited information about the targeted protein isoforms and may be affected by genetic variation in their epitope binding. Here we show that mass spectrometry (MS)-based proteomics can complement these studies and provide insights into the role of specific protein isoform and epitope-altering variants. Using the Seer Proteograph nanoparticle enrichment MS platform, we identified and replicated new pQTLs in a genome-wide association study of proteins in blood plasma samples from two cohorts and evaluated previously reported pQTLs from affinity proteomics platforms. We found that >30% of the evaluated pQTLs were confirmed by MS proteomics to be consistent with the hypothesis that genetic variants induce changes in protein abundance, whereas another 30% could not be replicated and are possibly due to epitope effects, although alternative explanations for nonreplication need to be considered on a case-by-case basis.

Fig. 1. Two-dimensional Manhattan plot.

Grid plot of the genomic position of the variant (SNP position) versus the position of the gene coding for the pQTL protein (protein position). The cis-pQTLs are in red and the trans-pQTLs in blue (plot data in Supplementary Table 2).

Fig. 2. Properties of 364 pQTLs discovered in a GWAS with MS proteomics.

a–c, Scatterplots of effect size (β) versus minor allele frequency (MAF) (a), pQTL effect size (β) from the discovery (Tarkin) versus from the replication (QMDiab) study (b) and EAFs for Tarkin versus QMDiab (c). All the protein associations that reached a significance level P < 5 × 10⁻⁸ in the discovery study are represented, except for b, where only replicated pQTLs are shown (plot data in Supplementary Table 2).

Fig. 3. MSPA scores and statistical power to replicate as a function of pQTL rank.

a–g, Scatterplots of the power to replicate a pQTL from the deCODE SOMAscan (a) and the UKB-PPP Olink (e) studies against the ranks of the affinity proteomics pQTLs. Scatterplots of individual MSPA scores against the ranks of the affinity proteomics pQTLs of the deCODE SOMAscan (b) and the UKB-PPP Olink (f) studies, starting with the lowest P value: 120 out of 319 pQTLs (37.6%) for deCODE and 167 out of 392 pQTLs (42.6%) for UKB-PPP had >99% power at a significance level of P < 0.05/319 and P < 0.05/319 for SOMAscan and Olink, respectively, colored to indicate likely protein abundance QTLs (MSPA score >0.8; green) and likely epitope effect-driven pQTLs (MSPA score <0.2; red). MSPA scores were limited to 46 pQTLs that were reported on the same variant in deCODE (c) and UKB-PPP (g). Scatterplots of the effect size (β) of the 46 pQTLs reported deCODE and UKB-PPP (d) (plot data in Supplementary Tables 9, 10 and 12).

Fig. 4. Peptide level analysis of the rs2052534 SPINK5 pQTL.

a–c, Violin plots of a splice QTL from GTEx (a), SPINK5 at the protein level (b) and SPINK5 at the peptide level for representative peptides of the four protein groups (c). d, Forest plot of the peptide level meta-analysis. e, UniProt isoforms and protein groups. f, Gene structure indicating the alternative splice forms (black, green, red and blue) from Ensembl with the position of rs2052534 indicated. The C-allele of rs2052534 leads to a higher exon excision ratio and results in lower levels of the peptide QVQNEAEDAK, which is specific for the intron-retaining Q9NQ38-3 isoform (note the reverse order of the alleles in a versus b and c). This SPINK5 pQTL illustrates how peptide level information can be used to identify different isoform QTLs. However, it also reveals current limitations in the automated quantification of such isoforms, because this information did not translate to the protein level QTL, which in this case reflects the average of the isoform signals. A multiple alignment of all detected peptides is shown in Supplementary Table 11, violin plots for all peptide associations are provided as Supplementary Data 5 and regional association plots are in Supplementary Figure 16.