Congestive heart failure can be caused by a wide array of myocardial insults. Although this etiological heterogeneity has yet to be well characterized, the role of neurohormonal pathways in the progression of heart failure is well established 123. Prejunctional α₂-adrenergic receptors (α_2A and α_2C) regulate the release of norepinephrine from cardiac sympathetic nerves in a negative feedback manner. When these receptors are ablated in mouse models, uncontrolled norepinephrine release causes a lethal cardiomyopathy 45. Moreover, the released norepinephrine activates β₁-adrenergic receptors (β₁AR) expressed on cardiac myocytes which are coupled to stimulatory G-proteins that propagate signals to downstream effectors such as adenylyl cyclase, ion channels, and phospholipases 6. Prolonged activation of cardiomyocyte β₁AR signaling by any number of pharmacologic or genetic means typically results in hypertrophy, ventricular dysfunction, ventricular remodeling, or frank failure 7. Thus, sympathetic activation of the heart is coordinated in part by these two adrenergic receptors, in which genetic variation of expression or function could act to modify the progression of heart failure. And indeed, substantial variation in the progression, mortality, and treatment response of heart failure between otherwise similar individuals is well recognized 8910.

Previous studies of a common single nucleotide polymorphism C->G in the β₁AR gene (ADRB1) at codon 389, which results in an arginine for glycine substitution (Arg389Gly), have shown enhanced coupling of the Arg variant to G_s in recombinant cells 11. Arg389 is associated with differential exercise capacity in heart failure patients 12, response to beta-blockers 1314, hypertension 15, and risk of myocardial infarction 16. In transgenic mice, the Arg variant has been associated with early enhanced cardiac function but in older mice a predisposition to heart failure 17.

A common insertion/deletion polymorphism in African-Americans resulting in a consecutive four amino acid loss (322–325) in the α_2CAR gene (ADRA2C) has been found to reduce receptor function 18, leading to a loss of normal synaptic autoinhibitory feedback and concomitant enhanced release of norepinephrine 19. Given the potential synergistic effects of these polymorphisms in the ADRB1 and ADRA2C genes, two previous studies have investigated and identified epistatic interactions affecting heart failure risk 20 and response to β-blockers in heart failure patients 21. In both studies, only two putative functional polymorphisms (i.e., ADRB1 Arg389Gly and ADRA2C ins/del 322) were examined. Multiple additional polymorphisms in both of these two intronless genes have been recently identified and characterized in whole-gene transfection studies, some of which alter protein expression 2223. Here we examine whether there is evidence for more complex intragenic and intergenic epistatic effects on heart failure phenotypes and survival in these genes utilizing 16 DNA sequence variations in the ADRA2C and 17 DNA sequence variations in the ADRB1 genes.

Intergenic epistasis, or interaction between two genes, occurs when the phenotypic effects of a variation in one gene is affected by a variation in a second gene. This could result from conformational changes that prevent physical interaction between the two gene products, or from a change in the ability of one gene to regulate the expression of the other or, the pathologic pathway involves both genes in a manner that their downstream effects converge to alter critical events. Intragenic epistasis occurs when a variation in one location of a gene influences the phenotype differently depending on other variations within the same gene. This type of epistasis is most notable when two different amino acid substitutions within a gene result in differentially functioning the resulting protein. However, it is also seen to have an effect on gene expression and processing.

The 655 subject heart failure cohort was 69.2% male, with 47.9% having a history of hypertension, 84.0% receiving ACE inhibitor therapy and 69.6% receiving β-blocker therapy (Table 1). Ischemic cardiomyopathies made up 39.5% of the cohort, idiopathic dilated cariomyopaties comprised 58.3% of the cohort, and the remaining 2.1% of the cohort had other causes of heart failure, such as primary valvular defects. They had an average age of heart failure onset of 53.8 years and an average follow-up time of 3.16 years. Figure 1 shows a Kaplan-Meier curve of time from heart failure diagnosis to death or cardiac transplant for this cohort. The minor alleles and the frequencies of the 16 DNA sequence variations in the ADRA2C and 17 DNA sequence variations in the ADRB1 genes are displayed in Table 2. Eleven of the 16 variants genotyped in the ADRA2C gene and 11 of the 17 in the ADRB1 gene had frequencies greater than 1% and are considered polymorphic. Four polymorphisms, ADRA2C(-2579 T/C), ADRA2C(+964/965 ins/del), ADRB1(-517 T/C), and ADRB1(+145 A/G), were significantly out of Hardy Weinberg equilibrium (HWE) (P < 0.05). The linkage disequilibrium between polymorphisms is illustrated in Figure 2 and indicates that several polymorphisms within each gene have significant frequency correlations.

Given the large number of tests for main effects and epistatic effects using these 22 polymorphic loci we used both the false discovery rate and cross-validation methods to reduce the probability of false positive results. Table 3 provides a summary of the tests of association between the polymorphism and age of onset, LVEF, LV mass, and survival. Overall, only two polymorphisms had significant main effects on the survival distribution and 8 pairs of polymorphisms had epistatic effects on survival that were significant (after adjustment by FDR) and cross-validated.

The relative risks (RR) associated with these polymorphisms are listed in Table 4. The two polymorphisms with single locus effects were ADRA2C(-1692 T/G) and ADRA2C(+964/965 ins/del). However, both of these polymorphisms were involved in significant epistatic interactions making the interpretation of the relative risks of their main effects impossible. Even though we equally tested for interactions within and among the two genes, somewhat surprisingly, we only found evidence of intergenic epistasis. The interaction between ADRA2C(+964/965 ins/del) and ADRB1(+145 A/G) arises from the 10:AG genotype class having elevated risk of death or transplant with RR = 5.12 (2.66,9.88). With respect to the interaction between ADRA2C(-230 T/C) and ADRB1(-517 T/C), the TC:CT genotype class has much higher risk of death or transplant with RR = 8.31 (3.32, 20.75) compared to the reference genotype class of TT:TT. The interaction between ADRA2C(-230 T/C) and ADRB1(-2297 T/G) appears to be due to the TC:TG genotype class (RR = 9.23 (3.67,23.16)), and the interaction between ADRA2C(-1692 T/G) and ADRB1(+145 A/G) is associated with an increased risk of death or transplant in the TG:AG genotype class with RR = 4.59 (2.31, 9.12). Likewise, the interaction between ADRA2C(-1692 T/G) and ADRB1(-3641 C/T) identifies the TG:CT genotype as the high risk group with RR = 3.56 (1.80, 7.05). Finally, the ADRA2C(-1692 T/G) and ADRB1(-3255 A/C) polymorphism appear to be interacting and again the TG genotype of the ADRA2C(-1692 T/G) gene has a significant increase risk in a particular ADRB1 genotype background – namely, AC genotype – such that the TG:AC genotype class has a relative risk of 3.02(1.53,5.97).

Given some degree of non-independence among these epistatic effects, we performed an a posteriori pooling of all high risk two locus genotype classes with RR's whose 95% confidence interval exceed 1.0 to attempt to identify the extent of overlap between the effects from the pairs of loci. These were compared to all other genotype classes combined, which provides a way to interpret these very specific sets of polymorphisms as if being used clinically. Here we found that the combined relative risk was 3.35 (1.82, 6.18).

In this study, we identified multiple polymorphisms within the ADRA2C and ADRB1 genes that interact to increase risk of death or transplant in heart failure patients. We investigated these genes because numerous studies in human patient populations 201214, animal models 17, and in vitro cell systems 18112223 have demonstrated their relevance to heart failure phenotypes or receptor expression/function.

As noted in the Results, four of the polymorphism significantly deviated from Hardy-Weinberg equilibrium. However, a previous study that genotyped the ADRA2C polymorphisms in a cohort of unaffected individuals did not show evidence of deviations from HWE 22. Similarly, genotyping of the ADRB1 polymorphisms in unaffected individuals for the Seattle SNPs database http://pga.gs.washington.edu/ do not show significant deviations from HWE. The HW deviations we observed could be due to an underlying association with heart failure 28 since our sample is exclusively heart failure patients and these genes have been previously associated with this disease, but random chance and genotyping error cannot be excluded.

The α₂ARs expressed on the presynaptic cardiac sympathetic nerves inhibit the release of norepinephrine when they are bound by the neurotransmitter 5, and thus provide a mechanism to regulate release of the neurotransmitter as sympathetic nervous system activity markedly increases in progressive heart failure. And indeed, in studies of gene-targeted mice where α₂ARs have been ablated, severe cardiomyopathy results due to norepinephrine cardiotoxicity exerted through myocyte β₁ARs 5. The human ADRA2C ins/del polymorphism previously denoted Del322-325, and here denoted ADRA2C(+964/965 ins/del) significantly reduces the function of α_2CAR receptors in transfected cells 18. While Caucasians have a very low prevalence of this allele 20 it appears that other common polymorphisms within this racial group, in the promoter and 3' UTR region of the gene, affect receptor expression 22. Of note, evidence suggests that there are few "spare receptors" in the complement of α_2CARs expressed in cardiac presynaptic nerves since heterozygous α_2CAR knock-out mice (~50% less receptor) also develop cardiomyopathy under pressure overload 29. Thus relatively small changes in expression due to polymorphisms may have physiologically relevant effects on the heart during heart failure progression. For the β₁AR, the most frequently studied polymorphic variation is at nucleotide 1165, where Arg or Gly can be commonly found at amino acid 389. This lies within a G-protein coupling domain, and in transfected cells the Arg variant exhibits enhanced coupling to adenylyl cyclase 11. In transgenic mice, with matched cardiac expression of the human β₁AR Arg or Gly389 receptors, Arg hearts have enhanced contractility, but progress to failure by 9-months of age, while Gly hearts show less contractile enhancement and no pathologic effects 17. The Arg389 phenotype also revealed a gene dose-response, and given that promoter SNPs of the ADRB1 gene alter expression in cell-based systems 23, the potential for cardiac relevance of non-coding β₁AR SNPs is apparent.

The previous studies of interactions between one ADRA2C and one ADRB1 polymorphism illustrate that the physiological effects of variations in these two genes act synergistically to increase risk of heart failure 20 and influences patient responsiveness to beta-blocker therapy as assessed by short-term improvements in LVEF 21. Our study indicates that there may be several other polymorphisms within these genes that have important physiological and clinical consequences for heart failure patients. Specifically, the ADRA2C(-1692 T/G) was involved in three epistatic interactions with ADRB1 polymorphisms (+145 A/G, -3641 C/T, and -3255 A/C) and the ADRA2C(-230 T/C) polymorphism was involved in two epistatic interactions with ADRB1 polymorphisms (-2297 T/G and -517 T/C). Only one ADRB1 polymorphism (+145 A/G) was involved in more than one epistatic interaction with the ADRA2C polymorphisms (-1692 T/G and +964/965 ins/del). In total, three polymorphisms in the ADRA2C gene and five polymorphisms in the ADRB1 gene appear to be implicated in epistatic effects on heart failure survival.

Because of the relatively large number of polymorphisms investigated in each gene one natural analytical approach might have been to investigate haplotypes within each gene and to test for potential interaction between haplotypes in their influence on risk. However, one of the main drawbacks of such an approach is reducing the haplotype space to the relevant set with phenotypic effects that could potentially interact within or across genes. Since we did not find any evidence of significant intragenic epistasis, it becomes even more difficult to analyze the diverse set of haplotypes – many with low frequency – to identify potentially relevant interactions across genes.

One of the short-comings of genetic association studies is that they have often failed to replicate and Manly 30 suggests that internal validation, common to good experimental practices, is one way to avoid the publication of spurious findings. In our study, we used cross-validation methods to significantly reduce the chance of false positives. Cross-validation methods were developed as a way to incorporate a measure of predictive accuracy (and correspondingly, a measure of prediction error) for an estimated model based on its performance predicting the outcome for independent test cases 31. During the last decade, cross-validation methods have been used widely for everything from robust variable selection in gene expression array studies 32 to reducing false positives in gene-gene interaction studies 3334 to evaluating the predictive accuracy of molecular or genetic classifiers of disease before clinical implementation 35. It has become a standard in the field of metabolomic 36, proteomic 3738, and transcriptomic 39 studies because of its ease of execution and its emphasis on prediction in independent test cases as a method of discriminating between true associations and false associations.

The epistatic interactions identified here occur between -1692, -230, and the +964/965 polymorphisms of ADRA2C and -3641, -3255, -2297, -517 and +145 polymorphisms in ADRAB1. We have previously generated constructs that mimic these polymorphisms and by transient transfection of cells ascertained expression phenotypes (5' and 3' -flanking regions) or signaling phenotypes (nonsynonymous coding polymorphisms) 11222318. All but one of the ADRA2C 5' promoter and 5' UTR polymorphisms found here that interact with ADRAB1 have been found to influence α_2CAR expression in these model systems. (The ADRA2C(-1692) polymorphism has not been studied in this manner.) And, the +964/965 deletion polymorphism results in depressed agonist-promoted function 18. For the ADRAB1, the aforementioned polymorphisms of the promoter region (except for -1294 which has not been studied) have also been shown to alter β₁AR expression 23. The nonsynonymous polymorphism at nucleotide position +145 (representing Gly49) results in a β₁AR that undergoes enhanced agonist-promoted downregulation compared to the Ser49 receptor 40, an important phenotype since downregulation is a protective mechanism in heart failure. Taken together, then, there is biologic plausibility in the epistatic interactions that were observed. The fact that there is a relatively large fraction of the patients that do not have these specific combinations, and yet they do display variability in heart failure progression, indicates additional genetic (and likely non-genetic) causes for heterogeneity. Within the axis which we are currently exploring, there are a number of other genes to be considered. As additional information from fine mapping and cell-based studies is obtained, there is the possibility for enhanced predictive power and an assignment of risk alleles for a greater percentage of heart failure patients. Nevertheless, the current work shows that epistatic interactions between α_2CAR and β₁AR polymorphisms affect heart failure survival, and further confirm the notion that this complex syndrome is modified by multiple polymorphisms in multiple genes.

Although we did not observe intragenic epistatic interactions between the ADRA2C and ADRB1 genes, we did observe multiple polymorphisms acting synergistically between the ADRA2C and ADRB1 genes to increase risk of death or cardiac transplant in heart failure patients. This underscores the complexity of the genetic factors that affect the progression of this syndrome.

The authors declare that they have no competing interests.

SLRK performed statistical analysis and drafted the manuscript. RK performed statistical analysis and assisted in drafting the manuscript. MK, BA, and GG developed genotyping methods for the ADRA2C and ADRB1 polymorphisms and performed the genotyping of the subjects. HH and LW recruited subjects and interpreted echocardiogram results. KC managed the subject clinical and endpoint data. GWD designed the study and assisted in drafting the manuscript. SL identified the ADRA2C and ADRB1 polymorphisms and their haplotypes, developed genotyping methods, participated in the design of the study and interpretation of results, and assisted in drafting the manuscript. All authors read and approved the final manuscript.