Limits...
Non-additive genome-wide association scan reveals a new gene associated with habitual coffee consumption

View Article: PubMed Central - PubMed

ABSTRACT

Coffee is one of the most consumed beverages world-wide and one of the primary sources of caffeine intake. Given its important health and economic impact, the underlying genetics of its consumption has been widely studied. Despite these efforts, much has still to be uncovered. In particular, the use of non-additive genetic models may uncover new information about the genetic variants driving coffee consumption. We have conducted a genome-wide association study in two Italian populations using additive, recessive and dominant models for analysis. This has uncovered a significant association in the PDSS2 gene under the recessive model that has been replicated in an independent cohort from the Netherlands (ERF). The identified gene has been shown to negatively regulate the expression of the caffeine metabolism genes and can thus be linked to coffee consumption. Further bioinformatics analysis of eQTL and histone marks from Roadmap data has evidenced a possible role of the identified SNPs in regulating PDSS2 gene expression through enhancers present in its intron. Our results highlight a novel gene which regulates coffee consumption by regulating the expression of the genes linked to caffeine metabolism. Further studies will be needed to clarify the biological mechanism which links PDSS2 and coffee consumption.

No MeSH data available.


Manhattan plot of the genome-wide association analysis under the recessive model.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4997959&req=5

f1: Manhattan plot of the genome-wide association analysis under the recessive model.

Mentions: None of the discovery step genome-wide associations revealed evidence of residual stratification either in each separate cohort or after meta-analysis (genomic control λ between 0.99 and 1.02, see Table S2 for details). Meta-analysis between the results from the INGI cohorts revealed 21 SNPs under the recessive model to be genome-wide significant, with the top hit rs6568479 showing a p-value = 8.9 × 10−10 and an effect of 0.086 corresponding to a difference of 1.2 cups of coffee per day for the people homozygous for the G allele. No association were found using the dominant genetic model. Table 1 summarizes the results of the association analysis and replication for the 21 significant SNPs. Effect sizes in the two populations were extremely similar (CARL 0.0864, FVG 0.0861). Some of the SNPs showed significant results also under the additive model although p-values were higher. For this reason, we consider the recessive model to be the true genetic model. Figure 1 shows the Manhattan plot from the recessive model meta-analysis. As can be seen in Fig. 2 all SNPs fell in the same locus in chromosome 6 all inside the PDSS2 gene which codes for the coenzyme Q10 (Fig. 2 shows the regional plot for the same SNPs). Replication of the results from the 21 significant SNPs in the ERF cohort revealed that 5 out of 21 SNP had a p < 0.05 with a concordant although attenuated effect. All 5 SNPs were significant after pooling the results from the three cohorts with a p-value lower than the one observed in the discovery step. None of the identified SNPs are coding, however we sought to verify if any of these could be associated with an existing e-QTL of PDSS2. For this reason we downloaded the full results from the Gtex Database32 for all cis-eQTL SNPs tested against PDSS2 falling inside the two recombination hot-spots delimiting the association signal. Comparing the patterns of association for all the SNPs (see Materials and Methods for details) between the two recombination hot spots that show significant correlation with all the association patterns for the tested tissues shows correlation values varying between −0.27 for Adipose Subcutaneous tissue and −0.61 for Esophagus Mucosa tissue. In all cases correlation was negative indicating that people with a higher consumption of coffee have a lower expression of PDSS2. Supplementary Table S3 reports the Kendall’s values for each tissue. The annotation with Haploreg showed 7 further SNPs in strong LD with the 5 replicated ones which were not in included in the association analysis. Although none of the 12 SNPs resulted to be coding, 6 out of 12 are located inside enhancers regions active in numerous tissues (Table S4). Finally consistently with the analysis on the Gtex Database all SNPs showed significant association with tissue specific expression of PDSS2.


Non-additive genome-wide association scan reveals a new gene associated with habitual coffee consumption
Manhattan plot of the genome-wide association analysis under the recessive model.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4997959&req=5

f1: Manhattan plot of the genome-wide association analysis under the recessive model.
Mentions: None of the discovery step genome-wide associations revealed evidence of residual stratification either in each separate cohort or after meta-analysis (genomic control λ between 0.99 and 1.02, see Table S2 for details). Meta-analysis between the results from the INGI cohorts revealed 21 SNPs under the recessive model to be genome-wide significant, with the top hit rs6568479 showing a p-value = 8.9 × 10−10 and an effect of 0.086 corresponding to a difference of 1.2 cups of coffee per day for the people homozygous for the G allele. No association were found using the dominant genetic model. Table 1 summarizes the results of the association analysis and replication for the 21 significant SNPs. Effect sizes in the two populations were extremely similar (CARL 0.0864, FVG 0.0861). Some of the SNPs showed significant results also under the additive model although p-values were higher. For this reason, we consider the recessive model to be the true genetic model. Figure 1 shows the Manhattan plot from the recessive model meta-analysis. As can be seen in Fig. 2 all SNPs fell in the same locus in chromosome 6 all inside the PDSS2 gene which codes for the coenzyme Q10 (Fig. 2 shows the regional plot for the same SNPs). Replication of the results from the 21 significant SNPs in the ERF cohort revealed that 5 out of 21 SNP had a p < 0.05 with a concordant although attenuated effect. All 5 SNPs were significant after pooling the results from the three cohorts with a p-value lower than the one observed in the discovery step. None of the identified SNPs are coding, however we sought to verify if any of these could be associated with an existing e-QTL of PDSS2. For this reason we downloaded the full results from the Gtex Database32 for all cis-eQTL SNPs tested against PDSS2 falling inside the two recombination hot-spots delimiting the association signal. Comparing the patterns of association for all the SNPs (see Materials and Methods for details) between the two recombination hot spots that show significant correlation with all the association patterns for the tested tissues shows correlation values varying between −0.27 for Adipose Subcutaneous tissue and −0.61 for Esophagus Mucosa tissue. In all cases correlation was negative indicating that people with a higher consumption of coffee have a lower expression of PDSS2. Supplementary Table S3 reports the Kendall’s values for each tissue. The annotation with Haploreg showed 7 further SNPs in strong LD with the 5 replicated ones which were not in included in the association analysis. Although none of the 12 SNPs resulted to be coding, 6 out of 12 are located inside enhancers regions active in numerous tissues (Table S4). Finally consistently with the analysis on the Gtex Database all SNPs showed significant association with tissue specific expression of PDSS2.

View Article: PubMed Central - PubMed

ABSTRACT

Coffee is one of the most consumed beverages world-wide and one of the primary sources of caffeine intake. Given its important health and economic impact, the underlying genetics of its consumption has been widely studied. Despite these efforts, much has still to be uncovered. In particular, the use of non-additive genetic models may uncover new information about the genetic variants driving coffee consumption. We have conducted a genome-wide association study in two Italian populations using additive, recessive and dominant models for analysis. This has uncovered a significant association in the PDSS2 gene under the recessive model that has been replicated in an independent cohort from the Netherlands (ERF). The identified gene has been shown to negatively regulate the expression of the caffeine metabolism genes and can thus be linked to coffee consumption. Further bioinformatics analysis of eQTL and histone marks from Roadmap data has evidenced a possible role of the identified SNPs in regulating PDSS2 gene expression through enhancers present in its intron. Our results highlight a novel gene which regulates coffee consumption by regulating the expression of the genes linked to caffeine metabolism. Further studies will be needed to clarify the biological mechanism which links PDSS2 and coffee consumption.

No MeSH data available.