The Effect of Genetic Variation on the Placental Transcriptome in Humans

Abstract

The knowledge of genetic variants shaping human placental transcriptome is limited and they are not cataloged in the Genotype-Tissue Expression project. So far, only one whole genome analysis of placental expression quantitative trait loci (eQTLs) has been published by Peng et al. (2017) with no external independent validation. We report the second study on the landscape of placental eQTLs. The study aimed to generate a high-confidence list of placental cis-eQTLs and to investigate their potential functional implications. Analysis of cis-eQTLs (±100 kbp from the gene) utilized 40 placental RNA sequencing and respective whole genome genotyping datasets. The identified 199 placental cis-eSNPs represented 88 independent eQTL signals (FDR < 5%). The most significant placental eQTLs (FDR < 10^-5) modulated the expression of ribosomal protein RPL9, transcription factor ZSCAN9 and aminopeptidase ERAP2. The analysis confirmed 50 eSNP-eGenes pairs reported by Peng et al. (2017) and thus, can be claimed as robust placental eQTL signals. The study identified also 13 novel placental eGenes. Among these, ZSCAN9 is modulated by several eSNPs (experimentally validated: rs1150707) that have been also shown to affect the methylation level of genes variably escaping X-chromosomal inactivation. The identified 63 placental eGenes exhibited mostly mixed or ubiquitous expression. Functional enrichment analysis highlighted 35 Gene Ontology categories with the top ranking pathways "ruffle membrane" (FDR = 1.81 × 10^-2) contributing to the formation of motile cell surface and "ATPase activity, coupled" (FDR = 2.88 × 10^-2), critical for the membrane transport. Placental eGenes were also significantly enriched in pathways implicated in development, signaling and immune function. However, this study was not able to confirm a significant overrepresentation of genome-wide association studies top hits among the placental eSNP and eGenes, reported by Peng et al. (2017). The identified eSNPs were further analyzed in association with newborn and pregnancy traits. In the discovery step, a suggestive association was detected between the eQTL of ALPG (rs11678251) and reduced placental, newborn's and infant's weight. Meta-analysis across REPROMETA, HAPPY PREGNANCY, ALSPAC cohorts (n = 6830) did not replicate these findings. In summary, the study emphasizes the role of genetic variation in driving the transcriptome profile of the human placenta and the importance to explore further its functional implications.

Keywords: ALSPAC; HAPPY PREGNANCY; REPROMETA; cis-eQTL; complex traits; human placenta.

FIGURE 1

Overview of the placental cis-eQTL analysis outcome. (A) Manhattan plot representing the landscape of P-values from the discovery analysis. The association P-values for the three eSNPs-eGene pairs selected for validation experiments are shown for the discovery (green) and validation (blue) analysis. The red line indicates the statistical significance threshold (FDR 5%). (B) The proportion of eGene expression (R²) explained by the eSNPs detected in this study. (C) The expression levels of identified placental eGenes, shown in read counts from the RNA-Seq dataset (Sõber et al., 2015). (D) The overlap of identified placental eGenes between the current study and Peng et al. (2017). (E) Profile of mRNA and protein expression of the placental eGenes according to Human Protein Atlas (www.proteinatlas.org).

FIGURE 2

Genomic regions surrounding the eSNP-eGene pairs selected for validation (A) ZSCAN9, (B) ERAP2, and (C) ALPG. Distances between the genes are drawn in approximate scale and the eGene of the region is highlighted with an increased font. eSNPs identified in the discovery study are shown above the genes. The eSNP chosen for experimental validation is boxed and its proxy SNPs in LD (r² > 0.8) are shown in gray below the genes. The landscape of reported genetic associations with common traits and diseases demonstrated at the bottom part of each subfigure was derived from the GWAS catalog and literature reports. The respective references are provided in Supplementary Table S9. ALPG, alkaline phosphatase, germ cell; ALPI, alkaline phosphatase, intestinal; ALPP, alkaline phosphatase, placental; AS, ankylosing spondylitis; chr, chromosome; CID, chronic inflammatory disease; DIS3L2, DIS3 like 3′-5′ exoribonuclease 2; ECEL1, endothelin converting enzyme like 1; eQTL, expression quantitative trait loci; ERAP1, Endoplasmic Reticulum Aminopeptidase 1; ERAP2, endoplasmic reticulum aminopeptidase 2; GWAS, genome-wide association study; IBD, inflammatory bowel disease; kb, kilobasepairs; LD, linkage disequilibrium; LNPEP, leucyl and cystinyl aminopeptidase; NKAPL, NFKB activating protein like; PE, preeclampsia; PGBD1, piggyBac transposable element derived 1; PsA, psoriatic arthritis; pulm. fn, pulmonary function; SCLC, Squamous cell lung carcinoma; SDS, standard deviation score; UTI, urinary tract infection; ZKSCAN4, zinc finger with KRAB and SCAN domains 4; ZKSCAN8, zinc finger with KRAB and SCAN domains 8; ZSCAN9, zinc finger and SCAN domain containing 9; SCAN12, zinc finger and SCAN domain containing 12; ZSCAN16, zinc finger and SCAN domain containing 16; ZSCAN26, zinc finger and SCAN domain containing 26; ZSCAN31, zinc finger and SCAN domain containing 31.

FIGURE 3

Experimental validation and genetic association testing with newborn parameters. (A) Comparative effect sizes from the RNA-Seq discovery and Taqman RT qPCR experiments for the eSNP-eGene associations selected for the experimental validation. For the estimation of fold changes, median expression level in the placentas with the major homozygote genotype was considered as the reference. Statistical analysis for association testing was performed under linear regression using additive model adjusted by the newborn sex, pregnancy complication group and labor activity. As the validation dataset was not matched for gestational age at delivery, this parameter was additionally incorporated as a covariate in the validation step. The shown P-values have been corrected multiple testing using FDR method. n, number of samples. (B) Effect of the ALPG c.-318 G > A (rs11678251) eSNP on the offspring growth parameters at birth and during infancy in the REPROMETA dataset. The data was available for 336 newborns at delivery, and their follow-up data at the of 6 (n = 233) and 12 (n = 216) months of age. Genetic association testing was performed using linear regression under recessive model adjusted by fetal sex. In testing the newborn parameters, gestational age at delivery was used as an additional covariate. The obtained nominal P-values < 0.05 were considered as supportive for the trend of a tested association. Beta values reflect the estimated effect of the AA-homozygosity on the tested parameter.