Genetic regulation of protein variation in human IPSc - The contribution of protein-protein interactions to the attenuation of genetic effects

The genetic component of the molecular phenotypes has been studied extensively in the last decade. Multiple results show that the donor identity accounts for a significant part of transcriptome variation and identified thousands of cis expression Quantitative Trait Loci (eQTLs). While gene expression is informative for understanding mechanisms of gene regulation, most cellular phenotypes are occuring at the protein level and variation in transcript levels may not be reflected at the protein level and vice versa. Yet, the degree in which the proteome is affected is still poorly understood.

To examine the proteome variation and its coordination with gene expression, we generated matched proteomic (TMT Mass Spectrometry), transcriptomic (RNA-seq) and genomic datasets for 202 iPSC lines derived from 151 different donors as part of the Human Induced Pluripotent Stem Cells Initiative (HipSci).

Our data revealed significant effects of the donor identity on abundance variation across lines for more than 900 proteins and several protein complexes, many of these being independent of those observed at mRNA level. Next, we mapped cis quantitative trait loci at mRNA and protein level (pQTLs), which identified 533 pQTLs and 3,036 eQTLs (FDR 5%; out of 8,600 genes tested at both layers). More than 77% pQTLs are nominally significant (P<0.01 and effect sizes of the same direction) in eQTLs and 33% of eQTls are nominally significant in pQTLs. For hundreds of non-concordant QTLs, we identified transcript isoforms specific expression changes as the underlying cause.

We examined the propagation of eQTLs at protein level, and quantified the contribution of several factors. Among these, the membership in a protein complex appears to be the major factor mitigating the cis eQTLs, in agreement with the expected effect of constraints induced by the protein-protein interactions stoichiometry. Conversely, these interactions represent a mechanism of propagation in trans of the genetic effects. By linking cis-pQTL snps to proteins in trans we identified 70 trans-pQTLS with many of the cis-trans protein pairs being part of protein complexes.

Finally, the QTLs overlap with variants related to disease traits indicated a higher enrichment of GWAS tagging variants in pQTLs when comparing to eQTLs. We analyse whether the attenuation of genetic effects at protein level links to this difference. Relying on scores of CNV attenuation at protein level measured in cancer lines, and on a larger eQTL set (GTEx), we observe that the attenuated genes are significantly less associated to disease variants.