A sex-specific evolutionary interaction between ADCY9 and CETP

  1. Isabel Gamache
  2. Marc-André Legault
  3. Jean-Christophe Grenier
  4. Rocio Sanchez
  5. Eric Rhéaume
  6. Samira Asgari
  7. Amina Barhdadi
  8. Yassamin Feroz Zada
  9. Holly Trochet
  10. Yang Luo
  11. Leonid Lecca
  12. Megan Murray
  13. Soumya Raychaudhuri
  14. Jean-Claude Tardif
  15. Marie-Pierre Dubé
  16. Julie Hussin

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This study presents evidence supporting population and sex-specific selection and an epistatic interaction between variants in the genes ADCY9 and CETP of pharmacogenetic importance. The confluence of evidence from population genetics, gene expression, functional experiments, and phenotypic association lends support to the paper's claims beyond what may be achieved by a single analysis. All three reviewers and I agreed that this work is of high interest in medical and population genetics and addresses a challenging topic in an impactful way.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Pharmacogenomic studies have revealed associations between rs1967309 in the adenylyl cyclase type 9 ( ADCY9 ) gene and clinical responses to the cholesteryl ester transfer protein (CETP) modulator dalcetrapib, however, the mechanism behind this interaction is still unknown. Here, we characterized selective signals at the locus associated with the pharmacogenomic response in human populations and we show that rs1967309 region exhibits signatures of positive selection in several human populations. Furthermore, we identified a variant in CETP , rs158477, which is in long-range linkage disequilibrium with rs1967309 in the Peruvian population. The signal is mainly seen in males, a sex-specific result that is replicated in the LIMAA cohort of over 3400 Peruvians. Analyses of RNA-seq data further suggest an epistatic interaction on CETP expression levels between the two SNPs in multiple tissues, which also differs between males and females. We also detected interaction effects of the two SNPs with sex on cardiovascular phenotypes in the UK Biobank, in line with the sex-specific genotype associations found in Peruvians at these loci. We propose that ADCY9 and CETP coevolved during recent human evolution due to sex-specific selection, which points toward a biological link between dalcetrapib’s pharmacogene ADCY9 and its therapeutic target CETP .

Article activity feed

  1. Version published to 10.7554/elife.69198 on eLife
  2. eLife

    Author Response:

    Reviewer #1:

    This is a very interesting manuscript that attempts to provide evidence of a case of evolutionary interaction (i..e. natural selection) between two human pharmacogenes: ADCY9 and CETP, suggesting also interaction with sex and a case for pleiotropy. This is likely one of the few examples (or maybe the only one) of evolutionary interaction between two genes in the field of human evolutionary genetics. The authors provide a set of genomics-based evidences to support their case of natural selection that include: (i) a replication of population genetics results in another Peruvian cohort, (ii) evidence of epistatic effect from RNAseq data on public datasets and on ad-hoc experiments, (iii) genotype/phenotypes associations in the UK Biobank. While the use of data from different sources (in opposition to a trans-omic approach in a Peruvian population) may be a weakness of the paper, it is important to recognize that performing a large set of experiments in the Peruvian cohort using the required sample sizes may be logistically prohibitive. Therefore, the author's approach is acceptable.

    Specifically, it is interesting that some of the phenotypes found in the UK Biobank is related with adaptation to high altitude (FVC)

    The only result that is difficult to explain of the difference in the level of long-range LD observed between males and females from Peru, both for the discovery and the replication datasets. The authors should elaborate more quantitatively on the plausibility of this finding.

    We thank the reviewer for highlighting our study’s novelty in the field of human evolutionary genetics. Indeed, the lack of a large, well-powered, accessible Peruvian biobank, and generally the overrepresentation of European ancestry resources in human genomics, is a weakness in most studies trying to address evolutionary and medical genetics questions in non-European populations. We hope our work will be cited as another example of why we need additional large and diverse cohorts in human genetics. This being said, the fact that we can report significant signals between our two genes of interest even in cohorts of unmatched ancestry is an argument for the widespread biological relevance of their interaction.

    A selective signal that varies according to sex is a plausible phenomenon, but they are very hard to detect, which means very little is known about these evolutionary events, particularly in humans, and therefore, quantitatively characterizing its likelihood is not straightforward. Such events can be explained by either (1) differences in survival between sexes or (2) by sexual selection, whereby individuals of one sex with specific genotypes would be more successful at reproducing. However, sexual selection alone is less likely here, as the favored parental combination would be equally likely to be transmitted to male or female offspring, so would not explain the preferential linkage between genotypes seen in males. Another explanation is differential survival of individuals receiving a specific combination, depending on their sex. In utero selection, whereby fetal survival chances depend on genotype combinations and sex of the foetus, or gamete selection, whereby reproductive cells with the advantageous genotype combination are more likely to give rise to a fertilized egg, are likely hypotheses. Finally, our results could be due to an ascertainment bias caused by increased mortality of individuals with a specific genotype combination in a sex-specific manner. We detail these hypotheses further in our revised manuscript, and highligth that these different scenarios will need to be investigated using simulations in future work to give a quantitative answer to this important question.

    Reviewer #2:

    This work attempts to identify possible functional links between two pharmacogenetic relevant loci, ADCY9 and CETP, by using signals of positive selection as a starting point. Starting with the 1000 Genomes (1kG) dataset, the authors identify complementary signals of positive selection in the 1kG Peruvian cohort (PEL) using iHS and PBS analyses, specifically in a LD block in ADCY9. They use these results then to support investigating possible coevolution between ADCY9 and CETP in the form of long-range linkage disequilibrium (LRLD), a clever way to identify possible cosegregation between variants that should otherwise not be present. This analysis is particularly apt since two regions have already been identified, one of which is now suggested to have experienced a rapid increase in allele frequency. The authors not only find evidence of LRLD between SNPs in ADCY9 and CETP, but they also identify these results occur in a sex-specific manner. The authors then begin investigating possible functional connections between these two loci, specifically in the form of gene expression analyses. Using both human cell ADCY9 knockdown lines and GEUVADIS/GTEx data, the authors identify that ADCY9 impacts CETP expression, both broadly and through a specific interaction between rs1967309 and rs158477. And lastly, the authors further investigate potential interactions between ADCY9 and CETP via epistatic association analyses using UK BioBank and GTEx data. Encouragingly, among a handful of relevant phenotypes, they find marginally significant, sex-specific interaction effects with CAD and Lp(a), and in both cases the direction of effects match those seen in their LRLD analyses -- eg rs1967309-AA + rs158477-GG containing a protective effect in males.

    Overall, the authors make a strong case that there is coevolution occurring between ADCY9 and CETP. That they are able to also continuously replicate some of their findings using an independent dataset, LIMAA, also strengthens their results. The authors do acknowledge that their sample sizes may be limiting their power at times, but they point out that finding multiple, concordant marginally significant results may represent an unlikely outcome. However, currently these are presented as just observations and not as a formal, integrated test. Therefore we cannot adjudicate whether these concordant, marginally significant results are occurring more than we would expect by chance.

    It is worth noting that by beginning with two loci, the authors are able to make use of 'pairwise' approaches such as LRLD and epistatic analyses. Normally, these types of tests come with large multiple testing burdens due to the rapid increase in test combinations. In fact, a priori one would potentially not expect such analyses to perform well with the limited sample sizes. However, by beginning with a hypothesis that focused on two loci, the authors are able to overcome this normal statistical challenge.

    It is also worth noting that these results are only possible due to the inclusion of diverse datasets. The initial selection signals would not have been identified if the datasets only contained individuals of European ancestry. Additionally, even with the limited sample sizes, the authors are still able to identify statistically significant results. Therefore this work is another example of what can be gained when incorporating more diverse human genomics datasets.

    In terms of identifying a functional or molecular link between ADCY9 and CETP, the authors have begun the work of finding some possible connections between these two loci. However, this goal was not completely met with the current work. There is a clear effect of ADCY9 on CETP gene expression, though whether this is ultimately a direct effect or indirect is unclear. And while the epistatic analyses using phenotypes such as clinical outcomes and biomarkers are encouraging, and concordant in terms of direction of effects, they still do not elucidate a mechanism by which the variants of interest in ADCY9 and CETP are functionally interacting. The finding that sex plays a role in this interaction, and may be important to the mechanistic link as well, is an important result though.

    We thank the reviewer for their careful assessment of our work. Indeed, our main finding here is an example of how population genetics (and specifically, the study of selective signals from genetic data in multiple populations) can help with understanding functional significance of association studies findings (here in the context of pharmacogenomics), which could be made possible thanks to important genomics resources such as the 1000 Genomes Project. Similarly, the functional characterization was made possible thanks to GTEx, and allowed us to explore multiple tissues to understand the association found, which would be otherwise impossible for a single group to perform for a specific research question like this one. We are therefore grateful to all researchers, institutions and funders that contributed in putting together these resources.

    We are happy that the reviewer sees the value in beginning with a well-defined two-loci hypothesis. Hypothesis-driven science using large, publicly available data resources has a very important place in biomedical research. It is true, however, that we do not provide an integrated test here to quantitatively evaluate the probability of our multi-level signals in the different datasets used. This is mainly because there’s no well-established framework to account for all confounders and ascertainment biases from different cohorts in an integrated way, and combining evidence from cell lines experiments with high-throughput data signals is not trivial. Therefore, we took the approach of reporting each line of evidence based on an appropriate experiment and statistical test, following a logical multi-step procedure, where each step is based on the hypotheses generated in the previous experiment/analysis. In the revised manuscript, we added a flowchart figure, showing for each step (natural selection, co-evolution, transcriptomics and pan-phenotypic analyses), the datasets used, and the key results obtained (needed to move forward to next steps), identifying the steps where we considered sex in the analysis. We hope this will help the reader to recognize that the likelihood of finding, by chance alone, interaction effects between our two genes in multiple independent experiments and datasets is very low, providing strong evidence for a functional genetic interaction.

    However, the reviewer is right that, despite indicating that the SNP rs1967309 in ADCY9 is involved in a functional mechanism related to CETP expression and that this mechanism implicates sex as a modulator, our study does not establish the exact molecular mechanism at play here. However, the identification of several relevant tissues in GTEx helps us focus our efforts for the next steps, and very importantly, our results establish that future experiments will now need to take sex into account. In our revised manuscript, we have added discussion points about what our results bring to this ongoing research for precision medicine.

    Reviewer #3:

    The authors have analysed genomic data from populations of South Americans to assess the genetic and functional link between ADCY9 and CETP. These two genes, which are both on chromosome 16 separated by over 50Mbp, show weak but significant long range linkage disequilibrium in some subpopulations (ie from Peru). The genetic link between these genes (and SNPs), despite being weak, is suggestive of positive selection for haplotypes that appear with higher frequency in the population compared to populations from Africa, Asia and Europe. What is surprising is the sex-specific linkage, with most if not all of the association signal being driven by male samples. The explanation behind this remains open, however, with multiple explanations from population dynamics and drift, to potential functional benefits selecting this association.

    Strengths:

    The work carefully assesses this linkage through robust statistical frameworks. Despite weak effects and low sample sizes, there is a replicable signal in two other populations. The data is also easily accessible and I appreciate the author's documentation of their work.

    Weaknesses:

    The effects are still weak, and may still be explained by multiple factors that aren't directly addressed in the work. Most of the caveats of the study have been pointed out by the authors in their discussion, yet some of these detract from their claims and findings. If this is a sexually dimorphic trait, or a sex-specific effect, most of the functional analyses are shown without this distinction.

    We thank the reviewer for their positive feedback and for bringing to our attention the critical point of functional analyses not being done in line with our sexually dimorphic findings. This comment led us to perform additional analyses that highly strengthened the manuscript.

    The preliminary sex-stratified analysis we had performed in our RNAseq discovery cohort (GEUVADIS) were not conclusive, which means we did not pursue these in our replication cohorts (GTEx and CARTaGENE). We have now performed further analyses in GTEx and CARTaGENE that produced very interesting new results, in line with the sex-specific nature of the evolutionary genetics results obtained, that we are happy to report in a revised manuscript. In summary, most sex-combined results were driven by males, but stratifying by sex in GTEx samples revealed additional tissues in which females only show a significant interaction (from tibial artery and heart tissues), and very intriguingly, the sign of interaction effect is reversed in this case. We tested for a three-way interaction (SNP1 x SNP2 x sex) in tissues harboring a convincing sex-specific pattern in stratified analyses. We also note that the donor for the HepG2 cell line, used in our knockdown experiments, was a male and so we highlight in our revised manuscript that future experiments should consider cells from both male and female donors in multiple tissues to better understand the molecular interaction.

  3. eLife

    Evaluation Summary:

    This study presents evidence supporting population and sex-specific selection and an epistatic interaction between variants in the genes ADCY9 and CETP of pharmacogenetic importance. The confluence of evidence from population genetics, gene expression, functional experiments, and phenotypic association lends support to the paper's claims beyond what may be achieved by a single analysis. All three reviewers and I agreed that this work is of high interest in medical and population genetics and addresses a challenging topic in an impactful way.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    This is a very interesting manuscript that attempts to provide evidence of a case of evolutionary interaction (i..e. natural selection) between two human pharmacogenes: ADCY9 and CETP, suggesting also interaction with sex and a case for pleiotropy. This is likely one of the few examples (or maybe the only one) of evolutionary interaction between two genes in the field of human evolutionary genetics. The authors provide a set of genomics-based evidences to support their case of natural selection that include: (i) a replication of population genetics results in another Peruvian cohort, (ii) evidence of epistatic effect from RNAseq data on public datasets and on ad-hoc experiments, (iii) genotype/phenotypes associations in the UK Biobank. While the use of data from different sources (in opposition to a trans-omic approach in a Peruvian population) may be a weakness of the paper, it is important to recognize that performing a large set of experiments in the Peruvian cohort using the required sample sizes may be logistically prohibitive. Therefore, the author's approach is acceptable.

    Specifically, it is interesting that some of the phenotypes found in the UK Biobank is related with adaptation to high altitude (FVC)

    The only result that is difficult to explain of the difference in the level of long-range LD observed between males and females from Peru, both for the discovery and the replication datasets. The authors should elaborate more quantitatively on the plausibility of this finding.

  5. eLife

    Reviewer #2 (Public Review):

    This work attempts to identify possible functional links between two pharmacogenetic relevant loci, ADCY9 and CETP, by using signals of positive selection as a starting point. Starting with the 1000 Genomes (1kG) dataset, the authors identify complementary signals of positive selection in the 1kG Peruvian cohort (PEL) using iHS and PBS analyses, specifically in a LD block in ADCY9. They use these results then to support investigating possible coevolution between ADCY9 and CETP in the form of long-range linkage disequilibrium (LRLD), a clever way to identify possible cosegregation between variants that should otherwise not be present. This analysis is particularly apt since two regions have already been identified, one of which is now suggested to have experienced a rapid increase in allele frequency. The authors not only find evidence of LRLD between SNPs in ADCY9 and CETP, but they also identify these results occur in a sex-specific manner. The authors then begin investigating possible functional connections between these two loci, specifically in the form of gene expression analyses. Using both human cell ADCY9 knockdown lines and GEUVADIS/GTEx data, the authors identify that ADCY9 impacts CETP expression, both broadly and through a specific interaction between rs1967309 and rs158477. And lastly, the authors further investigate potential interactions between ADCY9 and CETP via epistatic association analyses using UK BioBank and GTEx data. Encouragingly, among a handful of relevant phenotypes, they find marginally significant, sex-specific interaction effects with CAD and Lp(a), and in both cases the direction of effects match those seen in their LRLD analyses -- eg rs1967309-AA + rs158477-GG containing a protective effect in males.

    Overall, the authors make a strong case that there is coevolution occurring between ADCY9 and CETP. That they are able to also continuously replicate some of their findings using an independent dataset, LIMAA, also strengthens their results. The authors do acknowledge that their sample sizes may be limiting their power at times, but they point out that finding multiple, concordant marginally significant results may represent an unlikely outcome. However, currently these are presented as just observations and not as a formal, integrated test. Therefore we cannot adjudicate whether these concordant, marginally significant results are occurring more than we would expect by chance.

    It is worth noting that by beginning with two loci, the authors are able to make use of 'pairwise' approaches such as LRLD and epistatic analyses. Normally, these types of tests come with large multiple testing burdens due to the rapid increase in test combinations. In fact, a priori one would potentially not expect such analyses to perform well with the limited sample sizes. However, by beginning with a hypothesis that focused on two loci, the authors are able to overcome this normal statistical challenge.

    It is also worth noting that these results are only possible due to the inclusion of diverse datasets. The initial selection signals would not have been identified if the datasets only contained individuals of European ancestry. Additionally, even with the limited sample sizes, the authors are still able to identify statistically significant results. Therefore this work is another example of what can be gained when incorporating more diverse human genomics datasets.

    In terms of identifying a functional or molecular link between ADCY9 and CETP, the authors have begun the work of finding some possible connections between these two loci. However, this goal was not completely met with the current work. There is a clear effect of ADCY9 on CETP gene expression, though whether this is ultimately a direct effect or indirect is unclear. And while the epistatic analyses using phenotypes such as clinical outcomes and biomarkers are encouraging, and concordant in terms of direction of effects, they still do not elucidate a mechanism by which the variants of interest in ADCY9 and CETP are functionally interacting. The finding that sex plays a role in this interaction, and may be important to the mechanistic link as well, is an important result though.

  6. eLife

    Reviewer #3 (Public Review):

    The authors have analysed genomic data from populations of South Americans to assess the genetic and functional link between ADCY9 and CETP. These two genes, which are both on chromosome 16 separated by over 50Mbp, show weak but significant long range linkage disequilibrium in some subpopulations (ie from Peru). The genetic link between these genes (and SNPs), despite being weak, is suggestive of positive selection for haplotypes that appear with higher frequency in the population compared to populations from Africa, Asia and Europe. What is surprising is the sex-specific linkage, with most if not all of the association signal being driven by male samples. The explanation behind this remains open, however, with multiple explanations from population dynamics and drift, to potential functional benefits selecting this association.

    Strengths:

    The work carefully assesses this linkage through robust statistical frameworks. Despite weak effects and low sample sizes, there is a replicable signal in two other populations. The data is also easily accessible and I appreciate the author's documentation of their work.

    Weaknesses:

    The effects are still weak, and may still be explained by multiple factors that aren't directly addressed in the work. Most of the caveats of the study have been pointed out by the authors in their discussion, yet some of these detract from their claims and findings. If this is a sexually dimorphic trait, or a sex-specific effect, most of the functional analyses are shown without this distinction.

  7. Version published to 10.1101/2021.05.12.443794v2 on bioRxiv
  8. Version published to 10.1101/2021.05.12.443794v1 on bioRxiv