GNAS Gene Variants Affect β-blocker–related Survival after Coronary Artery Bypass Grafting

THE cardiac β-adrenergic receptor (βAR) signal transduction pathway regulates inotropy and chronotropy. βAR stimulation results in activation of the α subunit of the stimulatory G-protein (Gαs), which is coupled to adenylyl cyclase and increases cyclic adenosine monophosphate production. In coronary artery disease–associated ischemic heart failure, chronic stimulation of βARs by increased concentrations of circulating catecholamines evokes βAR desensitization through uncoupling of downstream signaling effectors.^1,2  βAR blocker use results in a restoration of βAR sensitivity and has been associated with slightly increased survival in patients after coronary artery bypass graft (CABG) surgery.³  However, substantial heterogeneity of individual responses to βAR blocker therapy exists and appear related in part to variation in the adrenergic pathway genes.⁴  Yet, results of studies addressing the association of genetic variation in the βAR with cardiac outcomes have been disappointing, and most presumptive associations have not been consistently replicated in later studies.^5–9 

Due to the direct signal transduction from βARs to the G-protein Gαs, Gαs might play a role in βAR blocker–related outcomes in patients undergoing CABG. Supporting this, in transgenic mice, overexpression of Gαs using a rat α-myosin heavy chain promoter¹⁰  increases heart rate and cardiac contractility in young transgenic animals but is associated with dilated cardiomyopathy at old age.¹¹  By contrast, chronic βAR blockade completely prevents the decrease in ejection fraction, cardiac dilation, and premature mortality, characteristic of older transgenic mice overexpressing Gαs.¹²  These data indicate significant physiological effects of increased Gαs expression. In humans, however, genetic overexpression experiments are difficult to perform and, therefore, it remains unknown whether altered Gαs expression affects the prognosis of humans with cardiac disease.

Gαs is imprinted in a tissue-specific manner. Although it is expressed primarily from the maternal allele in some select hormone-active tissues, it is biallelically expressed with an almost equal contribution of the maternal and paternal allele in most tissues.¹³  Although heterozygous GNAS mutations in specific tissues result in different kinds of endocrine disorders,¹⁴  germline variations affecting Gαs expression have not hitherto been identified. By sequencing the human GNAS gene, we recently identified two haplotype tagging single-nucleotide polymorphisms (SNPs), G(-1211)A and T2291C, which are located in regulatory regions and form three common haplotypes in the GNAS gene.^15,16  Haplotype *3 (A(-1211)/C2291) carriers showed altered transcription factor binding, increased Gαs expression, and enhanced Gαs-stimulated adenylyl cyclase activity. This was associated with a higher stroke volume and cardiac index as well as a lower N-terminal pro-brain natriuretic peptide concentration,^15,16  indicating that germline variants in regulatory regions of the Gαs gene serve as markers to investigate functional and clinical consequences of altered Gαs expression. These observations along with a pilot study of short-term mortality after CABG surgery¹⁷  led us to the hypothesis that GNAS variants associate with mortality during the 5 yr after CABG surgery and that this association is modified by the presence or absence of βAR blockade.¹¹  We further hypothesized that genetic variants would add additional predictive value over the EuroSCORE II surgical risk index, a validated and commonly used method for predicting mortality after CABG surgery.^18,19 

Patients aged 20 to 80 yr undergoing nonemergent primary CABG surgery using cardiopulmonary bypass without other concurrent surgery were prospectively enrolled (http://clinicaltrials.gov/show/NCT01258231) at two institutions (Brigham and Women’s Hospital, Boston, Massachusetts, and Texas Heart Institute, St. Luke’s Episcopal Hospital, Houston, Texas) between August 2001 and January 2009. Patients with a preoperative hematocrit of less than 25% or those having received a transfusion of leukocyte-rich blood products within 30 days before surgery were not enrolled. To avoid potential population stratification, analysis was restricted to subjects who self-reported four generations of grandparental European ancestry. Study protocols were approved by the respective Institutional Review Boards, and participants were enrolled after providing informed written consent. At each site, patient demographics, perioperative risk factors, medications, and postoperative outcomes were recorded using study-specific case report forms. Mortality was assessed by accessing hospital records and the Social Security Death Index.^† Death status was queried at 5 yr of follow-up or earlier in patients who had not completed 5 yr of follow-up at the time this analysis was performed.

DNA was extracted from leukocytes using standard protocols. Two GNAS gene SNPs (Chr. 20): G(-1211)A (rs6123837) and T2291C (rs6026584) and two ADRB2 gene SNPs (Chr. 5), Arg16Gly (nucleotide 46 A/G, rs1042713), and Gln27Glu (nucleotide 79 C/G, rs1042714) were genotyped. The method of “slowdown polymerase chain reaction”²⁰  was used to amplify promoter and intron 1 fragments comprising GNAS SNPs as previously described.¹⁶ GNAS and ADRB2 SNPs were genotyped using genotyping assays from Applied Biosystems (Carlsbad, CA). GNAS haplotypes were inferred using the Bayesian statistical-based program PHASE, version 2.1,²¹  resulting in three common haplotypes (*1, *2, and *3). Individual diplotypes (i.e., haplotype pairs) were constructed from the respective genotypes (see table, Supplemental Digital Content 1, http://links.lww.com/ALN/B35). All SNPs were in Hardy–Weinberg equilibrium.

Data are presented as mean ± SD unless stated otherwise. Univariate comparisons of 5-yr genotype- or diplotype-related mortality were performed using Cox proportional hazard statistics. Kaplan–Meier plots were used to show the relation between genotypes or diplotypes and cumulative mortality. Log-rank tests for trend were performed for genotype- or diplotype-dependent survival using an additive model due to a gene-dose effect. After demonstrating a significant βAR blocker × diplotype interaction on the univariate (P = 0.018) as well as the multivariate level (P = 0.015), we generated two multivariable (Cox proportional hazards) predictor models of mortality in patients receiving βAR blockers. The first model was generated using clinical and demographic variables that were associated (P < 0.1) in univariate analysis as well as covariates being part of the EuroSCORE II with time to death during the first 5 yr after surgery. The second multivariable model incorporated the EuroSCORE II score. Patients who did not die during the 5 yr after CABG were censored at 5 yr or at the last date of follow-up if that was less than 5 yr after surgery.

Proportional hazards assumptions were obtained by computing correlations between time and partial residuals, which are calculated separately for each genetic predictor. Partial residuals were plotted against survival time to test the proportional hazards assumption by adding a smoother to the residual plots. In addition, general linear regression analysis was performed for partial residuals against time. A nonzero correlation was regarded as evidence against the proportionality assumption.

This study takes the advantage of an ongoing registry and longitudinally updated data set, CABG Genomics, which has been used for previous analyses of genetic associations. We are aware of potential problems of multiplicity in a data set over time, but adjusting the α error for previous analyses is difficult and has traditionally not been done. Hence, neither attempts were made to adjust the error rate for past or future analyses of these data, nor did we consider multiple genotype predictors.

Data were analyzed using the SPSS software package, version 19.0 (IBM, New York, NY). A two-tailed P value of less than 0.05 was considered statistically significant.

Patient characteristics stratified by GNAS genotypes are shown in table 1. Whereas only a significant association of T2291C and sex could be detected, no association of GNAS genotypes and variables of cardiovascular disease prevalence were identified.

During the up to 5-yr postoperative follow-up period, 10.9% of subjects died. Median follow-up time of surviving patients was 5.0 yr (range, 2.4 to 5.0 yr). Univariate analysis demonstrated that GNAS variants, but not ADRB2 variants, had a significant association with all-cause mortality (table 2).

GNAS variants showed a significant association with mortality for the G(-1211)A variant (GG: 13.2%, AG: 9.8%, and AA: 9.6%; P = 0.012; fig. 1A) and the T2291C variant (TT: 18.9%, CT: 12.1%, and CC: 9.2%; P = 0.0003; fig. 1B). Estimation of diplotypes confirmed these results with 5-yr mortality for diplotype *1/*1: 18.9%, *2/*1: 13.7%, *2/*2: 9.3%, *3/*1: 10.6%, *3/*2: 9.1%, and *3/*3: 9.6% (overall P = 0.0006; fig. 1C). Consideration of clinical covariates revealed a hazard ratio of 2.2 for homozygous *1 versus *3 (95% CI, 1.2 to 3.8; P = 0.006).

We next examined GNAS genotype association with 5-yr mortality stratified by βAR blocker use. When we analyzed patients without βAR blocker therapy, we observed no genotype effect on 5-yr mortality (fig. 1, D–F). In contrast, there was a strong association with mortality in patients receiving βAR blockers. Five-year mortality was associated with genotypes of the GNAS variants G(-1211)A (GG: 14.1%, AG: 9.2%, and AA: 6.9%; P = 0.0014; fig. 1G), T2291C (TT: 19.4%, CT: 11.5%, and CC: 8.2%; P = 0.0002; fig. 1H), and estimated diplotypes (*1/*1: 19.4%, *2/*1: 13.5%, *2/*2: 9.2%, *3/*1: 9.7%, *3/*2: 8.6%, and *3/*3: 6.9%; P < 0.0001; fig. 1I). In multivariable analysis, after verifying proportional hazard assumptions, these findings remained present after accounting for clinical factors previously found to be associated with mortality and revealed increased mortality for *2/*1 and *1/*1 patients compared with that for homozygous *3 patients (table 3). We also estimated the additional value of GNAS haplotype data on the widely used EuroSCORE II to predict survival after CABG surgery. As expected, we not only observed strong predictive value of the EuroSCORE II but also observed that GNAS haplotypes were prognostic factors for 5-yr survival, independent of the EuroSCORE II. Although the EuroSCORE II resulted in a hazard ratio of 1.07 (95% CI, 1.06 to 1.09; P < 0.001), patients homozygous for the *1 GNAS haplotype had a more than three-fold increased 5-yr mortality compared with that for patients homozygous for *3 (hazard ratio, 3.14; 95% CI, 1.57 to 6.31; P = 0.001).

In this cohort, βAR blocker use per se was not associated with overall survival (fig. 2A). However, stratification by GNAS haplotypes demonstrated marked differences in the effect of βAR blocker use on survival. Significantly better survival was seen for patients using βAR blockers and being *3 positive (91.3%). In contrast, reduced survival was observed in patients not possessing the *3 haplotype irrespective of βAR blocker therapy (85.6 to 85.9%) or in haplotype *3 positive patients without βAR blocker therapy (85.8%; fig. 2B; P = 0.0085).

Our data show that GNAS haplotypes associate with significantly different rates of death during the 5 yr after CABG surgery. GNAS haplotype *1 or *2 carriers show an increased risk of death compared with the risk by *3 haplotypes. Importantly, this effect is more pronounced in patients receiving βAR blockers and is independent of other traditional risk factors.

These findings also suggest that routine use of βAR blockers in GNAS haplotype *3 negative patients undergoing cardiac surgery may not provide significant long-term mortality benefit. Furthermore, we observed that the addition of GNAS diplotypes to the well-validated EuroSCORE II provides additional predictive value for long-term postoperative mortality. The frequency of homozygosity of the *1 haplotype, similar to that observed in a different smaller CABG cohort,¹⁷  establishes the GNAS haplotype as a common disease-linked haplotype with a novel association to a long-term outcome, whereas an association to cardiovascular disease prevalence has not been detected.

Experiments from transgenic mice indicate the importance of the Gαs protein in the maintenance and augmentation of cardiac function.^11,12,22  We recently identified three common haplotypes in the GNAS gene encoding Gαs¹⁶  with haplotype-specific differences in GNAS promoter activity.^16,17  In human myocardium, we could demonstrate the greatest Gαs protein expression and signal transduction in haplotype *3 carriers, followed by haplotypes *2 and *1.¹⁷  Interestingly, diminished Gαs expression in haplotype *1 carriers was associated with a higher βAR density,¹⁶  a feature pathognomonic of βAR up-regulation.²³  However, it has to keep in mind that those experiments were carried out in human atrial tissue and it has been shown that there are differences in the Gαs expression levels between cardiac atria and ventricular myocardium.²⁴ 

Experiments with transgenic mice overexpressing Gαs showed increased sensitivity to catecholamines as a result of enhanced βAR coupling to various effector pathways, specifically adenylyl cyclase and L-type calcium channels.^10,11,25  However, the effects of chronically enhanced β-adrenergic signalling over the lifespan of the animal may be deleterious, particularly in the face of ineffective βAR-desensitization mechanisms, as these mice developed cardiomyopathy at older age.²⁶  These experiments in transgenic animals are in contrast to our findings, where we identified a 30% increase of Gαs expression in humans in homozygous haplotype *3 carriers,¹⁶  which was also associated with a survival benefit in patients with enhanced Gαs expression (haplotype *3). In transgenic mice, however, Gαs expression was up-regulated more than three-fold compared with that in wildtype mice.²²  Therefore, we hypothesize that higher Gαs expression in homozygous haplotype *3 carriers might result in a sensitization of βAR, resulting in a greater number of βAR in the “high-affinity state” resulting in a survival benefit.¹⁰ 

Our findings showing a survival benefit from βAR blockade in patients with haplotype *3 (increased Gαs expression), compared with other patients (fig. 2B), are in agreement with chronic βAR blockade in transgenic mice overexpressing Gαs resulting in protection from evoked heart failure.^{10–12,22,25} We, therefore, speculate that resensitization of βAR by βAR blockade in patients with increased Gαs expression results in a long-term patient benefit.

Another issue that has to be considered is the potential effect of altered Gαs expression in ischemia–reperfusion damage during CABG surgery. Many pathological processes contribute to ischemia- and reperfusion-associated heart injury. For example, ischemia during extracorporal circulation is associated with decreased adenylate cyclase activity and intracellular cyclic adenosine monophosphate concentrations and may result in increased vascular permeability, endothelial cell inflammation, an imbalance between vasodilating and vasoconstricting factors, and the activation of coagulation and the complement system.²⁷  Therefore, increased cardiac Gαs concentration with increased intracellular cyclic adenosine monophosphate levels during that period could evoke beneficial effects regarding attenuated reperfusion injury and inflammatory response.

We hypothesized that enhanced Gαs-mediated signal transduction might yield a more beneficial outcome after CABG, particularly under βAR blockade, and this hypothesis was tested, first in a pilot study,¹⁷  and now in a large multicenter cohort of patients. In the pilot study including 185 patients under chronic βAR blockade, Gαs expression was diminished in homozygous haplotype *1 patients and this was associated with an increased risk of death compared with the risk in homozygous haplotype *3 carriers, with an independent hazard ratio of 3.1.¹⁷  The current study in 1,627 patients with a 5-yr follow-up now extends these pilot data in a new cohort of well-characterized patients and confirms the primary hypothesis. In a multivariable analysis, homozygous *1 patients had a doubled risk of death compared with that of homozygous *3 patients and stratification by the presence or absence of βAR blockade tripled this risk, independently of conventional risk factors.

From a pharmacogenetic point of view, our observations are important as βAR blockers are the cornerstone therapy for patients with coronary artery disease, and the use of βAR blockers in the perioperative period remains a topic of intense research.²⁸  Although the consistent efficacy of βAR blockers has been questioned, variability of effects was suggested to relate in part to genetic heterogeneity.^29,30  On the basis of our observations in the current study, future studies of βAR blockers should include accounting for genotype structure of genes in the βAR pathway. Importantly, this study questions the value of routine use of βAR blockers in all patients undergoing cardiac surgery, an issue which was already raised during the early termination of the PeriOperative ISchemic Evaluation trial,³¹  in which patients who received βAR blockers had an increased risk of strokes. However, the investigational βAR blocker bucindolol, which in the Beta-Blocker Evaluation of Survival Trial trial³²  did not show an overall survival benefit compared with placebo, was associated with fewer adverse events in patients with a βAR genetic variant.³³  Thus, without pharmacogenetic data, there exists a tradeoff between the benefit of a medication in a minority of patients and the possibly associated risks and costs in the majority of patients.

Finally, despite improved survival in a pilot study of patients with various Gly16Gly genotypes undergoing CABG surgery,¹⁷  we did not find an association of ADRB2 polymorphisms with long-term survival in the current study. Therefore, we cannot draw a definitive conclusion of whether ADRB2 genotypes influence outcome after CABG surgery. However, a recent study³⁴  of patients with acute coronary syndrome does not support a significant impact of ADRB2 genotypes on outcome in American patients of Caucasian ethnicity.

Our study should be interpreted in the context of some potential limitations. The study of this longitudinal observational cohort was not designed to investigate βAR blocker–specific effects on long-term survival. Due to a high level of adherence to βAR blocker guideline recommendations, only a small fraction of subjects were left untreated with βAR antagonists. This precluded a formal analysis to investigate the statistical interaction of genotype with βAR antagonist treatment. Nevertheless, we identified genetic subgroups with worse outcomes despite βAR antagonist treatment, in whom more aggressive interventions to improve survival might be warranted. In addition, we had no information on the various βAR blockers used in this cohort and thus cannot ascertain agent-specific effects, if present. Moreover, we were unable to account for changes in medications during follow-up time. Furthermore, classifying patients by discharge medication status is a well-recognized and often-used approach because most patients remain on their discharge regimen after hospital stay.³⁵ 

This is an analysis of an ongoing study in a longitudinal cohort for which we have previously reported associations between genetic variants and atrial fibrillation,³⁶  myocardial injury,^37–39  ventricular dysfunction,^40,41  inflammation,⁴²  and up to 5-yr postoperative all-cause mortality.^41,43–45  The current study’s assessment of GNAS variants associated with 5-yr mortality did not adjust for previous analyses of these other gene-association studies in the CABG Genomics Program cohort. Finally, it may be speculated that one or several yet unidentified functional SNPs in adrenergic or other pathway genes such as the A2B adenosine receptor or the β1-receptor could influence our results.^6,46  However, due to our hypothesis-driven approach following a pilot study,¹⁷  analysis of those genes was not the aim of our current study but should be considered with future studies. Moreover, besides the β-adrenergic signaling pathway, Gαs is also activated by several other G-protein–coupled receptors,for example, including the serotonin and adenosine receptors.⁴⁷  It is, therefore, conceivable that it is not solely the β-receptor–mediated signal transduction pathway that is responsible for the observed effects but also cumulative or single effects mediated via other Gαs-mediated signaling pathways.

In conclusion, GNAS haplotypes independently associate with an increased risk of death after primary CABG surgery. These results are most pronounced in patients receiving βAR blockers and strengthen the rationale for personalized treatment to decrease medication side effects and improve outcomes.

Supported by Harvard Clinical and Translational Science Center, National Center for Research Resources, Boston, Massachusetts; National Institute of Health grants, Bethesda, Maryland; American Heart Association Scientist Development Grant; the Department of Anesthesiology, Perioperative, and Pain Medicine, Brigham and Women’s Hospital, Boston, Massachusetts; the Klinik für Anästhesiologie und Intensivmedizin, Universität Duisburg-Essen and University Hospital Essen, Essen, Germany; and Baylor College of Medicine Division of Cardiovascular Anesthesia at the Texas Heart Institute, Saint Luke’s Episcopal Hospital, Houston, Texas.

The authors declare no competing interests.