A recent project that illustrates the “Pharmaco-omics” research strategy that has proved to be so powerful in our studies of SSRI response in patients suffering from MDD is our 2016 “Molecular Psychiatry” (Gupta et al, 12:1717-1725, 2016) study which reported that, of the panel of LCECA platform plasma metabolites that we assayed at three time points for 300 MDD patients being treated with SSRIs, it was plasma serotonin that was most highly associated with virtually all of the clinical symptom-related outcomes and—furthermore—when we performed a GWAS for level of plasma serotonin as the phenotype, we identified two novel genes, TSPAN5 and ERICH3, that were genome-wide significant, both of which are highly expressed in the brain and which we could link functionally to serotonin biosynthesis in neuronally-derived cells. Finally, the ERICH3 SNPs from the GWAS were replicated in two other studies of SSRI response in MDD. These were striking and potentially significant examples of the power of joining multiple “omics” techniques—particularly against a background of the general lack of success of genomic techniques alone when applied to psychiatric illness.
Understanding each gene locus or gene cluster as an integrated regulatory: exampleCHRNA5-CHRNA3- CHRNB4 nicotinic receptor gene cluster
Introduction and Rationale. The vast majority of GWAS hits point to regulatory variants accounting for associations with complex traits, including drug response. Critical hub genes in biological networks tend to be under evolutionary selection, driving either purifying or positive selection. We have found that many drug target genes carry regulatory variants under positive selection, with high allele frequencies and influence on phenotypic traits. In most of these cases, more than one variant has arisen per gene locus in various populations. Finding the causative variants, and their interactions within the same gene locus, is critical for assessing the overall genetic influence, but this knowledge is still lacking for most genes. Even in heavy studied genes such as CYP3A4 and CYP2D6, we find substantial new regulatory variants that should be considered in biomarker panels (1,2). We posit that a full understanding of all functional variants and their interactions is key to evaluating the influence of each gene locus, and further, that dynamic interactions between gene loci (epistasis) cannot be fully revealed unless this overall influence is taken into account (statistically, a finding of epistasis is rapidly lost with decreasing LD of a surrogate marker with a causative SNP; ignoring the combined effect of multiple SNPs in a gene locus exacerbates this hurdle). As an example, we have selected the CHRNA5-CHRNA3-CHRNB4 nicotinic receptor gene cluster, containing protein coding and non-coding genes covered by long and frequent LD blocks, a signature of positive selection across the entire cluster (3).
Approach. We have develop an R package (K. Hartman) to canvass multiple databases including GTEx, dbGaP, ENCODE, 1,000 genomes project, etc, to extract LD, eQTLs, GWAS hits, and chromatin annotations, applicable to hundreds of genes at a time. This was applied to the nicotinic gene cluster to determine the main LD blocks and candidate variants associated with expression in body tissues and GWAS hits (3).
Results. In a first step, we have identified the main haplotype blocks and three SNPs, the regulatory variants rs880395 and rs1948, plus the known nonsynonymous CHRNA5 rs16969968, a known risk variant in nicotine dependence. Importantly, rs880395 affects the RNA expression of CHRNA3 and 5, and of an antisense RNA, while rs1948 affects specifically CHRNA3 only in the basal ganglia – a region important to nicotine dependence. The predominant LD structure in the gene cluster enables assignment of haplotypes and diplotypes with high confidence in a majority a study cohort. These results document the interactive nature of variants in this gene locus, modified in a tissue specific manner critical for determining influence on specific target traits.
Analysis of a GWAS cohort with nicotine dependence confirmed the known influence of the nsSNP alone, but haplotype and diplotype analyses reveal significant modulation of the influence of rs16969968 on nicotine dependence.
Conclusion. The CHRNA5-CHRNA3-CHRNB4 nicotinic receptor gene cluster represents a local regulome that should be considered for its overall influence on complex traits such as nicotine dependence. Operating via DNA looping over long distances, each regulatory variant can modify expression of multiple genes in an interactive fashion. We are now systematically expanding our approach to multiple gene loci and gene clusters, including cardiovascular and CNS disorder gene candidates, and genes in the innate immune system, which is under strong evolutionary constraints.
40,000 women die from breast cancer annually in the US, and breast cancer is still the most common invasive cancer in women. 70-80% of these patients have estrogen receptor positive (ER+) disease which is currently treated with hormonal therapy, especially aromatase inhibitors (AIs) as first line therapy. Resistance to AIs is a major reason for disease recurrence and metastatic disease. AIs were designed specifically to inhibit aromatase (encoded by CYP19A1) activity, the enzyme that catalyzes the rate limiting step that converts androgens to estrogens. Estrogens are the major driving force for ER+ breast cancer. Among third generation AIs, letrozole and anastrozole are non-steroidal, while exemestane is steroidal. Prospective clinical trials have showed no differences in efficacy among these three AIs, perhaps due to the fact that results were always compared at the population level. However, our preliminary data indicate that knowledge of the host germline genetic background could help us to select AIs to treat specific individuals and to predict response to AI therapy. The purpose of this study is to understand how and why germline genetic variants might be associated with individual variation in in AI response. Understanding mechanisms for variation in AI response would help us design individualized therapies to treat ER+ breast cancer, especially for patients who are resistant to AI treatment
Since AIs block aromatase activity, suppression of estrogen formation is the pharmacodynamic effect of AIs. We set out to test whether genetic variation might contribute to AI efficacy through suppression of estrogen synthesis. To test that hypothesis, we took advantage of a series of large clinical trials for which we have extensive genomics and clinical response information. The first trial was conducted at Mayo Clinic, M.D. Anderson and MSK (Memorial Sloan Kettering) and the trial recruited over 800 post-menopausal women with primary early stage breast cancer treated with anastrozole. We have genome-wide SNP and hormone levels measured before and after patients had been treated with anastrozole for six months. We then performed a genome-wide association study (GWAS) for 624 of these patients for whom we had complete data to identify SNPs associated with estrogen level changes in women with ER+ breast cancer who were treated with anastrozole. Replication was then performed with disease recurrence as the phenotype for a second GWAS using samples from the MA.27 clinical trial with over 7500 patients randomized to anastrozole and exemestane treatment of post-menopausal women with ER+ breast cancer. That association study showed that a SNP in CSMD1 was associated with differential time to disease recurrence between the two AIs. That is, the variant genotype had a protective effect in the anastrozole arm only (Table 1). Functionally, the SNP influenced CSMD1 and, in turn, CYP19 gene expression, which contributed to SNP dependent response to anastrozole (Figure 1). Our data also suggested, surprisingly, that anastrozole might be an ERα ligand and function as ER degradator, especially in the presence of low dose E2. These findings have significant clinical implication: 1. The SNP can be used as a biomarker for selection of patients for anastrozole treatment. 2. The idea of combination of anastrozole and low dose estrogen to degrade ERα might be an alternative strategy to sensitize patients to anastrozole, especially in anastrozole resistant patients (Figure 2). 3. Our long term goal is to develop better therapeutic agents that can take advantage of our new findings.
ACCOuNT is a NIMHD funded Collaborative Consortium aimed at accelerating the pace of pharmacogenomic discovery and translation in African Americans. We are particularly interested in drug used in cardiovascular disease. We have brought together investigators at 2 major US cities (Chicago and DC), as well as clinical, pharmacogenomic and informatics experts to build a sustainable research consortium.
Figure 1. While both the discovery and translational projects are independent efforts, the overall interactions of these projects within ACCOuNT develop a pipeline from which newly discovered SNPs in the Discovery Project and SNPs curated via PharmGKB (Data Analysis and Harmonization Core) can be investigated within Translational Project. Data gather within the Translational Project can inform which drug are relevant candidate for investigation within the Discovery Project. Results of the Translational project will be evaluated by the Implementation Advisory Council (IAC), which will collaborate with regional community partners for implementation planning and pilot project development. The African American genomics database will house information generated via the Discovery Project and will become a resource for the greater pharmacogenomics community. Lastly, all efforts will have oversight for our CSAB within the Consortium core, which will also launch community-based pilot projects.
AA: African American, Dashed arrow show potential synergies, while solid arrows show planned interactions.
The goal of the NHLBI GO Exome Sequencing Project (ESP) is to discover novel genes and mechanisms contributing to heart, lung and blood disorders by pioneering the application of next-generation sequencing of the protein coding regions of the human genome across diverse, richly-phenotyped populations and to share these datasets and findings with the scientific community to extend and enrich the diagnosis, management and treatment of heart, lung and blood disorders.
Our October featured Investigator, Debbie Nickerson, along with Phil Green, Jay Shendure, are the three PIs heading the SeattleGO group in the NHLBI GO ESP. The other groups participating and collaborating are:
Clinical application of pharmacogenomics knowledge will result in less ‘trial and error’ prescribing and more efficacious, safer and cost-effective drug therapy. However, despite the major advances in PGx and several commercially available PGx tests, its implication in routine patient care remains limited. The Ubiquitous Pharmacogenomics Consortium (U-PGx; www.upgx.eu) is a European project bringing together a large group of pharmacogenomics experts from institutes in 10 different countries. U-PGx entails a 5 year, 15 million euro programme funded by the European Union’s Horizon 2020 research and innovation programme. The ultimate objective of the consortium is to make actionable pharmacogenomics data and effective treatment optimization accessible to every European citizen. Specifically, U-PGx will investigate if the emerging approach of pre-emptive genotyping of an entire panel of important PGx markers will result in an improved outcome for patients and is cost-effective.
U-PGx uses a multifaceted approach consisting of four components to achieve this goal (fig 1). The first component focuses on developing enabling tools required to integrate PGx test results into the electronic healthcare record and computerized decision support systems, taking into account the differences in health care models, languages and laws across the EU. These enabling tools consist of IT solutions, PGx testing infrastructure, PGx educational programmes for healthcare professionals, and the translation of the Dutch PGx guidelines into local languages. This component will pave the way for the execution of component two. Component 2 entails implementation of PGx testing of a panel of pharmacogenes into clinical practice in a large cluster-randomised controlled trial (n=8,000) in 7 European countries(Netherlands, Spain, UK, Italy, Austria, Greece and Slovenia) , assessing the aggregate impact of multiple pharmacogenomics interventions on patient outcomes. Additional outcomes include, cost-effectiveness and process indicators for implementation and provider adoption of PGx. A third component applies innovative methodologies such as NGS and systems pharmacology to discover additional variants associated with drug response and to elucidate drug-drug-gene interactions. The final, fourth, component assures ethical proceeding of the project and spearheads outreaching and educational activities to influential stakeholders.
Dr. Monte's research gathers clinical, genomic, cytochrome (CYP) metabolism phenotype, and metabolomic data in a prospective trial of a patients beginning metoprolol therapy for uncontrolled hypertension. These factors will be compared to identify which variables predict the drug response to metoprolol. These data will be integrated into a model to predict systolic blood pressure reduction due to metoprolol. This innovative approach integrates clinical, genomic, CYP metabolism phenotype, and metabolomic data to predict the drug response of metoprolol. The integrative approach can be applied to other diseases and therapeutics to improve drug efficacy and safety. This may eliminate time intensive up-titration, eliminate therapies that are destined to be ineffective, and minimize adverse drug effects.
A clinical example of the proposed integrated method for predicting drug response. (1) Patients are given 10 mg of hydrocodone. (2) Clinical phenotypes are captured fully and completely. These may include (among others) development of ADRs, chronicity of treatment, ethnic differences, and demographic factors. (3) Association studies may contribute to characterization of the clinical phenotypes (e.g. RNA-sequencing may help distinguish chronicity of treatment). (4) The drug response is categorized into phenotypically pertinent groups. (5) Relevant biological pathways are identified and linked by individual metabolomic markers. (6) Stratification of drug response is refined by accounting for biological pathway polymorphisms and controlled for phenotypic variables captured in #2 above. (7) The final stepwise model is built, allowing for a high, although not perfect, receiver operating characteristic (ROC).
Experimental MIRD entails using hypercapnia (rebreathing 5% carbon dioxide) to measure baseline minute ventilation response and depression in the response after morphine dose. The first aim is to test the hypothesis that Caucasian race (genetically defined using ancestry information markers using a Genome Wide Association Study array) and female sex contribute to MIRD risk. The second aim addresses the contribution of specific genetic variants and their interactions ( ATP binding cassette ABCB1, Fatty Acid Amide Hydrolase/FAAH and Mu-opioid receptor /OPRM1) to inter-patient variability in MIRD. Analysis is done using logistic regression after population stratification using known clinical predictors like morphine doses, hyperoxemia, pain scores, co-administration of sedatives and significant variables from Aim 1 as covariates. Furthermore, this project explores associations with MIRD for variants in select genes involved in the opioid-MIRD and morphine pharmacokinetic (PK) pathway using a discordant phenotype approach to maximize identification of associations. Morphine concentration data is analyzed using non-linear modeling to evaluate genetic effects on PK/PD.
We reported association for μ1 opioid receptor OPRM1 A118G SNP and clinical respiratory depression in our population. This variant results in decreased μ-receptor binding potential in the brain and increases morphine requirement. Multivariable logistic regression showed that the risk of MIRD in patients with AA genotype was significantly higher (odds ratio 5.6, 95% CI: 1.4–37.2, P=0.030). Presence of G allele was associated with higher pain scores (effect size 0.73, P=0.045). (attached figure). We also published our findings of associations of ABCC3 variants with postoperative MIRD and morphine pharmacokinetics in children. The ATP binding Cassette gene ABCC3 which is a hepatic efflux transporter of morphine metabolites, was found to affect morphine 3 and 6 glucuronide formation clearances. This supports findings in a tonsillectomy population which is under study at our institution, where ABCC3 variant is associated with MIRD. This is the first study to report association of ABCC3 variants with opioid-related RD, and morphine metabolite formation (in two independent surgical cohorts). Moreover, we identified a region of FAAH variants in the region +/- 5kb of the FAAH gene with regulatory function, associated with depression of the hypercarbic response after morphine administration, and post-operative vomiting. Patients with clinical respiratory depression also had depressed hypercarbic responses, which indicated our genetic associations can identify subclinical respiratory depression. This manuscript is under preparation. These findings bring us closer to my goal of understanding predictors of MIRD in children, and personalization of opioid analgesia to improve opioid safety.