Regulatory polymorphisms in key candidate genes for disease susceptibility and drug response: a mandate for valid genetic biomarkers

Biomarkers have moved into the center stage of personalized healthcare, promising optimal treatment and prevention strategies that are tailored to the individual. Genetic factors are thought to play significant roles in most complex diseases. Yet, genetic biomarkers have been gaining ground in clinical practice at a surprisingly slow pace. Among the many hurdles to overcome, incomplete understanding of the genetic basis of complex disorders looms high. Recent advances in genomics indicate a probable prevalence of regulatory polymorphisms over those encoding altered amino acid sequences, but a systematic exploration of the genetics of regulatory processes is still in its infancy.

Genetics of complex disorders: current limitations for genetic biomarkers

Among the many types of biomarkers, genetic sequence variants hold particular promise for predicting disease risk and guiding successful interventions. However, our understanding of the underlying molecular, genetic and biological mechanisms is still marginal. Despite tremendous technological advances, such as genome-wide association studies (GWAS), we have gained only a glimpse of candidate genes, while lacking understanding of how multiple genes interact to cause risk or afford protection. Current opinion holds that complex disorders result from numerous rare deleterious mutations that one can glean from GWAS with very large cohorts. With few frequently occurring variants found to contribute to disease risk, the development of clinically useful biomarkers proceeds at a slow pace, defying overly ambitious predictions of a boon in personalized healthcare.

Prevalence of regulatory polymorphisms

Earlier genetic studies have focused largely on coding SNPs (cSNPs; ‘SNP’ is used here to represent all types of sequence variants) that alter the amino acid sequence. However, hypothesis-neutral GWAS have overwhelmingly implicated SNPs located in introns, UTRs, promoter/enhancer regions and intergenic regions (possibly related to noncoding RNA genes) [1]. This finding should not have been a surprise, as gene regulation has been invoked as a main driver in evolution, outnumbering functional cSNP by a large margin [2]. Hence, we expect to find regulatory polymorphisms that have accumulated to a high frequency in human populations. Positive selection implies penetrance of the sequence variant conveying some advantage to the individual or the population at large. Possibly, the initial intent of GWAS, searching for frequent polymorphisms conveying disease risk, was doomed to fail from the onset, as one would expect negative selection and, hence, low frequencies for disease-causing variants. Low frequency of the identified candidate risk genes thus requires large cohorts for GWAS.

The physiological impact of regulatory polymorphisms is typically tissue specific [3], possibly one reason for the greatly lowered constraints for evolutionary development. We can, therefore, expect regulatory polymorphisms to have accumulated to high frequencies but their impact will be targeted to a defined selectable trait – an analysis of GWASs looking for the wrong traits will miss even highly penetrant and frequent regulatory polymorphisms. On the other hand, even if a polymorphism were under positive selection for some advantage, typically, it turns out to be a double-edged sword: what is good for one function (e.g., long endurance) may be detrimental for another (e.g., maximal strength). As a consequence, we can consider frequent regulatory polymorphisms as key factors in health but possibly also in long-term risk for disease. Clearly, rare mutations and frequent regulatory variants must be viewed together to assess individual risk.

Systematic search for regulatory polymorphisms

For a number of years, my laboratory has searched for regulatory polymorphisms in key genes involved in disease risk and drug response. We have developed a systematic approach for discovery of regulatory polymorphisms, using allelic mRNA expression in human target tissues, for example brain, liver, kidney and intestines, and scanning the gene locus for polymorphisms that account for differences in mRNA expression [4]. This has yielded a plethora of new regulatory variants in genes that had already been under intense study for some time (e.g., DRD2, TPH2, ACE, MAOA, CYP3A4, CETP, CHNRA5 and DAT [Wang D, Papp A, Pinsonneault J, Smith R, Sadee W, Unpublished Data] [5–8]. In a survey of 42 candidate genes, nearly half showed detectable allelic expression variation [4]. The newly discovered variants include those involved in transcription (rSNPs), but a surprisingly large portion affects mRNA processing, splicing, turnover and translation by SNPs residing in the transcribed portion of the gene. We have named these structural RNA SNPs (srSNPs) [3,4]. Allele frequencies were found to be high (5–50%) and, together with a location within large frequent haplotype blocks, suggest evolutionary selection. It is noteworthy that several of these regulatory polymorphisms represent a gain-of-function (increased expression; e.g., DRD2, TPH2 and CHNRA5), typically considered a rare event but consistent with evolutionary selection mechanisms. Some of these variants were subsequently found to be associated with clinical phenotypes, with high odds ratios and potential as biomarkers [5–8].

Inability of clinical association studies alone to identify functional genetic variants

Our result in such well-studied genes argue that clinical association studies have proven inadequate for the discovery of regulatory polymorphisms, even while such variants may be prominent, residing in obvious candidate genes identified by numerous studies. Yet, clinical association studies of traits that are far removed from any causative SNP continue to represent the main effort in finding candidate polymorphisms. For example, a promoter variant in SERT, encoding the serotonin transporter, has been implicated by hundreds of clinical trials as a risk factor in depression, possibly in conjunction with environmental factors. However, a recent meta-analysis failed to validate this claim [9]. The rush to implicate promoter variant SERT–LPR arose from molecular studies showing LPR to affect in vitro transcription [10]. Yet, the molecular genetic results have not been uniformly replicated, and we have found no evidence that the LPR affects mRNA expression in the raphe nuclei (in human autopsy tissues) where SERT is expressed and then distributed throughout the brain [11]. Whereas clinical association studies are under increasing scrutiny as to their validity, curiously, this is not the case for the molecular studies. A single in vitro result can trigger a storm of clinical trials, without any validation of the molecular basis of any effect in a physiological target tissue.

A second example is an InDel mutation in the angiotensin-converting enzyme gene (ACE), applied to countless clinical studies because ACE is strongly implicated in many disorders and the InDel is easily measurable and frequent. After more than 3000 clinical association studies, the current verdict is that the InDel is probably not functional and may, at best, partially represent a functional variant. Measuring allelic mRNA expression of ACE in heart tissues, we have identified regulatory SNPs nearly 2 kb upstream of the transcription start site that strongly affect mRNA expression [8]. These SNPs are highly prevalent only in the African–Americans tested, but not in Caucasians, and were significantly associated with coronary artery events in a hypertensive cohort. The impact of the SNPs on therapy with ACE inhibitors remains to be tested, but the results failed to show any evidence for a role of the InDel.

What must be done to exploit the full potential of genetic biomarkers?

We derive several guidelines, projections and recommendations from these studies discussed earlier:

• • The search for genetic biomarkers must be tailored to include regulatory polymorphisms that may not be directly related to disease risk and, in fact, could be ‘wellness genes’ under normal circumstance. Clinical association studies as currently conducted are ill suited to finding such variants;

• • Emphasis must be placed on multifactorial analysis, including environmental and genetic factors. Genetic variants should include rare mutations in combination with validated frequent regulatory polymorphisms, even if the latter show no disease association when analyzed alone;

• • Genetic biomarkers for optimizing drug therapy may have a greater chance to show high impact because the drug as an environmental stimulus taps into specific physiological mechanism, with fewer genes playing key roles;

• • Failure to clarify the causative relationship between a candidate SNP and its biological effect is detrimental, leading to unnecessary cost and endless clinical association studies. One must identify the causative polymorphism, particularly when a biomarker test involves a combination of several genes. Marker SNPs merely associated with the causative variant introduce additional noise that will be amplified with each gene added to the biomarker mix;

• • We must ask questions regarding to what extent each causative variant accounts for the interindividual variability in a given gene; if we capture only 30% of genetic factors in a given locus, the genetic impact is severely underestimated. Allelic mRNA analysis in target tissues is well suited to address this question.

We are entering a new era in medicine. Emphasis shifts to individualized disease management, early intervention and prevention. The development of genetic biomarkers must reflect the critical elements essential to attaining and maintaining personal wellness. An emphasis on disease-causing genetic variants alone is unlikely to yield success. For drug therapy, genetic biomarkers will assume increasing importance, being either specific to a given drug therapy or reflecting a disease component affected by the treatment. In either case, genetic biomarkers are only one aspect of all other factors, such as clinical traits/data, personal environment and culture. Effective biomarker panels incorporate all these variables. Therefore, we must assure that the genetic biomarkers reflect and fully represent valid biological processes so as to avoid adding more noise to a very noisy area of medicine, and fulfill the promise of personalized healthcare.

Financial & competing interests disclosure

Wolfgang Sadee is Professor of the Ohio State University. He is also Chief Scientific Officer of AIKO Biotechnology, Inc. He consults for pharmaceutical companies, but declares no external conflict of interest with respect to the content of this article. Several genetic variants mentioned in this article are included with current patent applications pursued by the Ohio State University. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.