Patient genotypes impact survival after surgery for isolated congenital heart disease

Page 1

CONGEN

ITALHEA

RT

J. MAXWELL CHAMBERLAIN MEMORIAL PAPER FOR CONGENITAL HEART SURGERY

Patient Genotypes Impact Survival After Surgeryfor Isolated Congenital Heart DiseaseDaniel Seung Kim, BS, Jerry H. Kim, MD, Amber A. Burt, MS, David R. Crosslin, PhD,Nancy Burnham, MSN, Donna M. McDonald-McGinn, MS, Elaine H. Zackai, MD,Susan C. Nicolson, MD, Thomas L. Spray, MD, Ian B. Stanaway, BS,Deborah A. Nickerson, PhD, Mark W. Russell, MD, Hakon Hakonarson, MD, PhD,J. William Gaynor, MD, and Gail P. Jarvik, MD, PhDDepartment of Medicine, Division of Medical Genetics, and Departments of Genome Sciences and Anesthesiology and Pain Medicine,University of Washington School of Medicine, Seattle, Washington; Divisions of Cardiothoracic Surgery, Genetics, and CardiothoracicAnesthesiology, and the Center for Applied Genomics, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania; andDepartment of Pediatrics and Communicable Diseases, Division of Pediatric Cardiology, University of Michigan Medical School, AnnArbor, Michigan

Background. Survival after cardiac surgery in infancyrequires adaptive responses from oxidative stress man-agement and vascular regulation pathways. We tested thehypothesis that genetic variation in these pathways in-fluences postoperative survival in nonsyndromic congen-ital heart disease children.

Methods. This is an analysis of a cohort of non-syndromic congenital heart disease patients who under-went cardiac surgery with cardiopulmonary bypass before6 months of age (n [ 422). Six single nucleotide poly-morphisms (SNPs) in six genes involved in oxidativestress and vascular response pathways, identified througha priori literature search, were tested for effects ontransplant-free survival. Survival curves, adjusting forconfounding covariates, were calculated using the Coxproportional hazard models.

Results. Long-term survival was strongly associatedwith vascular endothelial growth factor A gene SNPrs833069 (p [ 7.03310L4) and superoxide dismutase 2gene SNP rs2758331 (p [ 0.019). To test for joint effects

Accepted for publication March 5, 2014.

Presented at the Fiftieth Annual Meeting of The Society of ThoracicSurgeons, Orlando, FL, Jan 25–29, 2014. Winner of the J. MaxwellChamberlain Memorial Award for Congenital Heart Surgery.

Address correspondence to Dr Jarvik, Medical Genetics, Box 357720,University of Washington, Seattle, WA 98195-7720; e-mail: [email protected].

� 2014 by The Society of Thoracic SurgeonsPublished by Elsevier Inc

of the two SNPs on transplant-free survival, the geno-types were grouped to form a risk score reflecting thecumulative number of risk alleles (0 to 4 alleles perpatient). A higher risk score based on the VEGFA andSOD2 SNP genotypes was associated with worsetransplant-free survival (p [ 3.02310L4) after con-founder adjustment. The total burden of risk alleleswas additive; subjects with the highest risk score of 4(n [ 59 subjects, 14.2% of the cohort) had a totalcovariate-adjusted hazard ratio of 15.64 for worsetransplant-free survival.Conclusions. After cardiac surgery, infants who are

homozygous for the high-risk alleles for both the VEGFAand SOD2 SNPs have an approximately 16-fold increasedrisk of death or heart transplant, suggesting that geneticvariants are important modifiers of survival after surgeryfor congenital heart disease.

(Ann Thorac Surg 2014;98:104–11)� 2014 by The Society of Thoracic Surgeons

ongenital heart defects (CHDs) are the most com-

Cmon human birth defect. Approximately one third ofCHD cases require surgical intervention, with a majorityinvolving cardiopulmonary bypass (CPB). Long-termmortality in the postoperative stages remains consider-able, especially for the more severe heart defects [1].

Oxidative stress is considered to be a major factor aftercardiac surgery with CPB owing to postoperative organdysfunction [2]. The importance of oxidative stress inpostoperative outcomes has been demonstrated throughthe finding that allopurinol, which blocks free radical

formation and its resulting oxidative stress, is associatedwith decreased cardiac event rate after surgery with CPBin high-risk infants with hypoplastic left heart syndrome(HLHS) [3].Studies of postsurgical outcomes in pediatric patients

have successfully identified several genetic variants in-volved in vascular response pathways that affect long-term outcomes. First, endothelin-1 missense variant(EDN1 G5665T) has been associated with transplant-free

Ms McDonald-McGinn discloses a financial relation-ship with Natera.

The Appendix can be viewed in the online version ofthis article [http://dx.doi.org/10.1016/j.athoracsur.2014.03.017] on http://www.annalsthoracicsurgery.org.

0003-4975/$36.00http://dx.doi.org/10.1016/j.athoracsur.2014.03.017

Page 2

Abbreviations and Acronyms

CHD = congenital heart defectCPB = cardiopulmonary bypassEA = European ancestryEDN1 = endothelin-1HLHS = hypoplastic left heart syndromeMI = myocardial infarctionSNP = single nucleotide polymorphismSOD2 = superoxide dismutase 2TGA = transposition of the great arteriesTOF = tetralogy of FallotVEGFA = vascular endothelial growth factor A

105Ann Thorac Surg CHAMBERLAIN MEMORIAL PAPER KIM ET AL2014;98:104–11 VEGFA/SOD2 AND LONG-TERM SURVIVAL

CONGEN

ITALHEA

RT

survival in patients with functional single ventricle CHD,with the greatest effects in children with the most severephenotype, HLHS [4]. More recently, a randomized clin-ical trial reported that missense mutations that upregu-lated the renin-angiotensin-aldosterone system wereassociated with impaired ventricular remodeling, renalfunction, and somatic growth in infants with functionalsingle ventricle postcardiac surgery, highlighting the roleof vascular response genes on a wide spectrum of post-surgical outcomes [5].

Taken together, these studies suggest that oxidativestress and vascular response play important roles ininjury repair and long-term survival in the pediatric CHDpopulation. We sought to examine the effects of specificgenetic variants implicated in oxidative stress manage-ment and organ recovery on long-term survival in acohort of children with nonsyndromic CHD. Secondarily,we performed an analysis using a genetic risk score,reflecting the number of deleterious alleles each patient

Table 1. Description of Single Nucleotide Polymorphisms Studied

SNP GeneaVariantType Chr:Positionb

Major/MinorAllele

rs1051740 EPHX1 Tyr113His 1:224,086,256 A/G

rs5370 EDN1 Lys198Asn 6:12,404,241 T/G

rs833069 VEGFA Intronicc 6:43,850,557 A/G

rs2758331 SOD2 Intronicd 6:160,025,060 T/G

rs10776686 (CYP2E1) Intergenic 10:135,182,921 A/G

rs1001179 CAT 5’UTR 11:34,416,807 T/C

a Intergenic single nucleotide polymorphisms (SNPs) are represented in parinformation and annotation from reference assembly 36.3. c Intronic SNP,5’UTR SNP rs2010963, which has been reported to increase expression levelssuperoxide dismutase 2 (SOD2) Val16Ala missense SNP rs4880.

Chr ¼ chromosome; 3’UTR ¼ 3 prime untranslated region of a gene; 5’polymorphism.

has, to determine if the observed genotype effects wereindependent and additive.

Patients and Methods

Study DesignThis is an analysis of a previously described prospec-tive cohort [6–8] of 550 subjects collected to study neu-rodevelopmental dysfunction after CHD palliation. Thisspecific study sought to identify gene regions relatedto oxidative stress and vascular response potentiallyaffecting survival in infants after cardiac surgery withnonsyndromic CHD. We note that no genome-wide as-sociation analyses have been attempted on the phenotypeof long-term survival; this is solely a candidate genestudy.Of the 550 original subjects, 56 were removed owing

to likely genetic syndrome and an additional 72 wereremoved owing to lack of high-quality genotype data,leaving a total of 422 subjects available for analyses.Additional information on data collection (includinginclusion/exclusion criteria), operative management,genotyping, and analyses not presented in the mainmanuscript can be found in the online-only Appendix.

Single Nucleotide Polymorphism SelectionTo preserve statistical power, we selected six candidatesingle nucleotide polymorphism (SNPs) at six differentgenes involved broadly in oxidative and ischemic stressresponse (Table 1) a priori based on a systematic litera-ture review of published evidence from other in-vestigators, reporting that variants in these genes have afunctional impact potentially relevant to the outcomes

Minor AlleleFrequency Gene Name and Description

0.291 Epoxide hydrolase 1. Converts epoxides tonontoxic forms

0.187 Endothelin 1. Pro-peptide of endothelin 1,a potent vasoconstrictor

0.346 Vascular endothelial growth factor A. Growthfactor mediates vascular permeability andendothelium growth/apoptosis inhibition

0.436 Superoxide dismutase 2. Mitochondrial protein;converts superoxide byproducts to hydrogenperoxide

0.051 Cytochrome P450 2E1. Endoplasmicreticulum-associated enzyme, involvedin varied processes

0.159 Catalase. Key antioxidant heme enzyme presentin peroxisome

entheses naming the nearest gene, for example (CYP2E1). b Positionrs833069, is in strong linkage disequilibrium (LD [r2 ¼ 0.97]) with VEGFAof VEGFA. d Intronic SNP, rs2758331, is in strong LD (r2 ¼ 0.93) with

UTR ¼ 5 prime untranslated region of a gene; SNP ¼ single nucleotide

Page 3

Table 2. Baseline and Operative Characteristics of the Cohort

Baseline Characteristics

CohortSubset

(n ¼ 422)

SexFemale 176 (41.7)Male 246 (58.3)

EthnicityAsian/Pacific Islander, Hispanic, or other

ancestry51 (12.1)

African ancestry, not Hispanic 98 (23.2)European ancestry, not Hispanic 273 (64.7)Gestational age, weeks 38.5 � 2.05Gestational weight, kg 3.15 � 0.62

Diagnostic classI: 2 ventricles, no arch obstruction 204 (48.3)II: 2 ventricles, arch obstruction 41 (9.7)III: 1 ventricle, no arch obstruction 46 (10.9)IV: 1 ventricle, arch obstruction 131 (31.1)

Specific CHD diagnosesHypoplastic left heart syndrome 130 (30.8)Tetralogy of Fallot 64 (15.2)Transposition of the great arteries 34 (8.1)Ventricle septal defect 40 (9.5)Ventricle septal defect, coarctation of aorta 19 (4.5)Single ventricle 30 (7.1)Other distinct CHD diagnoses 105 (24.9)

Preoperative intubation 119 (28.2)Preoperative length of stay, days 2.14 � 2.61Age at first operation, days 42.1 � 54.7Weight at first operation, kg 3.82 � 1.25Total CPB first operation, minutes 67.1 � 40.3Use of DHCA 259 (61.4)Total DHCA time in first operation, minutes 41.4 � 18.3Hematocrit level after hemodilution in first

operation, %27.8 � 4.1

Genetic risk score categorya

0 risk alleles 11 (2.6)1 risk allele 64 (15.2)2 risk alleles 138 (32.7)3 risk alleles 150 (35.5)4 risk alleles 59 (14.0)

Long-term mortality 47 (11.1)Time to mortality, years 9.32 � 3.48Heart transplants 4 (1.0)

a Number of vascular endothelial growth factor A (VEGFA) single nucle-otide polymorphisms (SNP) rs833069 or superoxide dismutase 2 (SOD2)SNP rs2758331 major alleles, which are both associated with worse long-

106 CHAMBERLAIN MEMORIAL PAPER KIM ET AL Ann Thorac SurgVEGFA/SOD2 AND LONG-TERM SURVIVAL 2014;98:104–11

CONGEN

ITALHEA

RT

(Appendix Table S1). Four of the six genes—epoxidehydrolase 1 (EPHX1), superoxide dismutase 2 (SOD2),cytochrome P450 2E1 (CYP2E1), and catalase (CAT)—areinvolved in oxidative stress. Vascular endothelial growthfactor A (VEGFA) and EDN1 are involved in the vascularresponse to low output and ischemic states associatedwith CPB; additionally, VEGFA may help promotevascular adaptation to hemodynamic alterations, and theEDN1 missense SNP has been associated with transplant-free survival in single-ventricle children [4]. All studiedSNPs were in Hardy-Weinberg equilibrium in controls.

AnalysisGenotypes were coded using an additive model, and allregression models included adjustment for confoundersbased on genetic ancestry and pertinent clinical cova-riates. Genetic ancestry was determined using previouslydescribed methods [9]. Owing to the mixed ancestry of thecohort (Table 2 for demographic information, includinggenetic ancestry), the first three principal component ei-genvectors from principal components analysis wereused as covariates to adjust for potential populationstratification [10]. We adjusted for multiple contrastsbased on the six SNPs by applying a Bonferroni correctionof a ¼ 0.05/6 ¼ 8.3�10�3.

Survival analyses and graphics were performed in R(http://www.r-project.org/). Time to long-term mortalitywas calculated from date of initial surgery to date ofdeath; these data include all deaths, including operativedeaths. A Cox proportional hazards model was used toevaluate the joint effect of the studied SNP and potentialcovariates. Output from the Cox proportional hazardsmodel was used for plotting of survival curves. An addi-tional analysis included cardiac transplant as an endpointin addition to any death using the described methods.Survival analyses were adjusted for the previously re-ported confounding variables: the first three principalcomponent eigenvectors for race, sex, gestational age,gestational weight, diagnostic class [11], preoperativeintubation, preoperative length of stay, age at first oper-ation, weight at first operation, total CPB time, use ofdeep hypothermic circulatory arrest, total deep hypo-thermic circulatory arrest time, and hematocrit at firstoperation [12]. A European ancestry (EA) only sensitivityanalysis for all outcomes was performed to ensure thedirection of SNP effects was consistent within the largestgenetic ancestry subgroup. A separate, diagnostic-classstratified analysis was performed for transplant-freesurvival to ensure consistency of SNP effects across thephysiologic range of heart defects.

term survival.

Values are n (%) or mean � SD.

CHD ¼ congenital heart defect; CPB ¼ cardiopulmonarybypass; DHCA ¼ deep hypothermic circulatory arrest.

Results

Demographic and clinical variables are presented inTable 2. Of the 422 nonsyndromic children studied, 16died operatively and 31 additional children died duringthe follow-up period, with an average time to mortality of9.32 years; an additional 4 patients underwent cardiactransplant. Number of transplants or death by diagnosticclass was 8 (3.9%) for class 1; 2 (4.9%) for class 2; 7 (15.2%)

for class 3; and 34 (25.9%) for class 4. Median follow-uptime for all subjects was 10.18 years.Survival analyses demonstrated that SNPs at two

different loci were associated with long-term survival:

Page 4

Table 3. Single Nucleotide Polymorphism Results (p �0.05) for Long-Term Mortality

SNP GeneEA Only

HR (95% CI)aFull Cohort

HR (95% CI)aEA Onlyp Valueb

Full Cohortp Value

rs833069 VEGFA 0.25 (0.105–0.609) 0.37 (0.211–0.659) 0.0021 7.03�10�4

rs2758331 SOD2 0.29 (0.133–0.636) 0.52 (0.304–0.900) 0.0019 0.019

a Hazard ratios (HR) and 95% confidence intervals (CI) were calculated using Cox proportional hazards methods for the outcome of long-term survival,adjusting for the covariates listed in Methods. b Analyses performed on the majority genetic ancestry group (European ancestry [EA], non-Hispanic) ofthe cohort (n ¼ 273 subjects), using Cox proportional hazards methods for the outcome of long-term survival, adjusting for the covariates listed in themethods.

SNP ¼ single nucleotide polymorphism; SOD2 ¼ superoxide dismutase 2; VEGFA ¼ vascular endothelial growth factor A.

107Ann Thorac Surg CHAMBERLAIN MEMORIAL PAPER KIM ET AL2014;98:104–11 VEGFA/SOD2 AND LONG-TERM SURVIVAL

CONGEN

ITALHEA

RT

VEGFA intronic SNP rs833069 (hazard ratio [HR] 0.37, p ¼7.03�10�4) and SOD2 intronic SNP rs2758331 (HR 0.52, p¼0.019; Table 3). For both SNPs, each copy of theminor allelewas associated with an increase in long-term survival. TheVEGFA effect was significant after Bonferroni correction,whereas the SOD2 effect was not. In the case of VEGFA SNPrs833069, each “G” allele led to a dose-dependent increasein survival probability (Fig 1A). A similar effect was notedfor SOD2 SNP rs2758331 (Fig 1B). We note that whentesting for transplant-free survival, both VEGFA and SOD2remained significant (Appendix Table S2).

To test for joint effects of the two SNPs in VEGFA andSOD2 on long-term transplant-free survival, the geno-types were grouped to form a risk score reflecting thecumulative number of risk alleles (0 to 4 alleles per pa-tient). For both VEGFA SNP rs833069 and SOD2 SNPrs2758331, the minor allele was low risk, whereas the morecommon variant was the high-risk allele for survival(Table 3 and Fig 1). A higher risk score based on theVEGFA and SOD2 SNP genotypes was associated with

Fig 1. (A) Vascular response–related vascular endothelial growth factor A (VAA [n ¼ 193]; dotted line ¼ AG [n ¼ 171; broken line ¼ GG [n ¼ 58]), ars2758331 (solid line ¼ TT [n ¼ 132]; dotted line ¼ GT [n ¼ 208]; broken licurves show a dose-dependent effect of each SNP. Note the range of the y-

worse transplant-free survival (p ¼ 3.02�10�4), adjusting forconfounders. The total burden of risk alleles was additive;patients with the highest risk score of 4 (n ¼ 59 subjects,14.2% of the cohort) had a total covariate-adjusted HR of15.64 for worse transplant-free survival (Fig 2).To rule out effects of residual population stratification,

we performed a separate sensitivity analysis on the ma-jority EA subset of the cohort (n ¼ 273 subjects) for thesignificant SNPs (Table 3). These EA only analyses showconsistency of the effect size and directions with thecomplete cohort. Thus, our results are unlikely to besecondary to population stratification.To examine the potential for genotype confounding,

whereby our identified VEGFA and SOD2 SNPs areassociated with less severe CHD, therefore leading toincreased survival, we performed separate Cochran-Armitage tests for both genotypes and diagnostic class.From these analyses, we found that the VEGFASNP rs833069 minor allele was associated with higherrisk diagnostic class subgroups (p ¼ 9.12�10�8; see

EGFA) single nucleotide polymorphism (SNP) rs833069 (solid line ¼nd (B) oxidative stress–related superoxide dismutase 2 (SOD2) SNPne ¼ GG [n ¼ 82]) genotypes predict long-term survival. The survivalaxis, Survival Probability.

Page 5

Fig 2. Genetic risk score reflecting number ofvascular endothelial growth factor A(VEGFA) and superoxide dismutase 2 (SOD2)risk alleles carried by each patient is predic-tive of long-term survival. Note the range ofthe y-axis, Survival Probability. (Solid blackline ¼ double homozygote protective [n ¼ 11];dashed purple line ¼ one risk allele [n ¼ 64];dotted green line ¼ two risk alleles [n ¼ 138];broken blue line ¼ three risk alleles [n ¼150]; dashed red line ¼ double homozygousdeleterious [n ¼ 59].)

108 CHAMBERLAIN MEMORIAL PAPER KIM ET AL Ann Thorac SurgVEGFA/SOD2 AND LONG-TERM SURVIVAL 2014;98:104–11

CONGEN

ITALHEA

RT

Appendix Table S3). Further analysis of specific di-agnoses found that the VEGFA minor allele was in excessin HLHS (all diagnostic class 4), whereas the VEGFAmajor allele was positively associated with both tetralogyof Fallot (TOF) and transposition of the great arteries(TGA [see Appendix Tables S4 to S6]). The vast majorityof both the TOF (67 of 69) and TGA (49 of 52) subgroupswere categorized into diagnostic class 1. The SOD2genotype was not associated with diagnostic class.

We conducted a sensitivity analysis, stratifying the pri-mary analysis by diagnostic class, to evaluate potentialconfounding (Appendix Table S7) by severity of CHD.From these generally underpowered analyses, we notedthat VEGFA and SOD2 genotypes remained marginallysignificant (VEGFA p ¼ 0.037 and SOD2 p ¼ 0.066) whereasthe genetic risk score remained highly significant (p ¼0.0067) for class 4 subjects. Genetic risk score was alsosignificant for class 3 subjects (p ¼ 0.017). The other effectswere not significant, although coefficients trended in thecorrect direction for the remainder of diagnostic classes.Given this, and that class was a covariate in our originalregression model, the association of VEGFA genotype withCHD class did not account for the association of risk scorewith survival.

Comment

We have performed a candidate SNP analysis for genesinvolved in oxidative stress and injury repair pathwaysin the prediction of long-term survival of children withnonsyndromic CHD. The SNPs at two genes, VEGFA andSOD2, were associated with long-term survival. More-over, we have shown through a genetic risk score thatthe effects of the VEGFA and SOD2 SNPs are cumula-tive—the more copies of the deleterious alleles of eitherSNP a patient has, the lower their probability of survival.

Through exclusion of subjects with chromosomal orgenetic anomalies, we have obtained results that aremore accurate representations of the pediatric popula-tion with nonsyndromic CHD, as genetic anomalies arefrequently associated with poorer outcomes in childrenwith CHD [8].In our analyses, we have identified minor alleles at

VEGFA and SOD2 as both being significantly protectiveagainst long-term mortality (VEGFA p ¼ 7.03�10�4 andSOD2 p ¼ 0.019). We also have demonstrated that theeffects of both SNPs are independent and additive whenconsidered together as a genetic risk score (p ¼ 3.02�10�4,HR 15.64 in the highest risk group [n ¼ 59] comparedwith the group with no risk alleles [n ¼ 11]). The VEGFASNP studied, rs833069, is in strong linkage disequilibrium(r2 ¼ 0.97) with one 5’UTR SNP, rs2010963, whose minorallele has been associated with higher VEGFA expression[13]; thus, the SNP studied is associated with higherVEGFA expression. Prior experiments injecting exogenousvascular endothelial growth factor (VEGF) into rats foundan increase in cardiac vascular permeability and cellulardamage, which is hypothesized to be a mechanismthrough which VEGF causes injury after myocardialinfarction (MI) [14]. However, other studies have foundthat lower endogenous VEGF levels are associated withan increased risk of adverse cardiovascular events afterMI [15] and that exogenous VEGF can rescue cardiacfunction after MI [16].Recent animal evidence suggests that whereas long-

term VEGFA expression (eg, from a plasmid) is detri-mental to heart remodeling after MI owing to the sideeffects of hypotension and edema, short-term, pulselikeVEGFA expression that more closely matches in vivoVEGF dynamics improves survival at 1-year follow-upwhile avoiding the previous VEGF-associated sideeffects [17]. Taken together, we hypothesize that our

Page 6

109Ann Thorac Surg CHAMBERLAIN MEMORIAL PAPER KIM ET AL2014;98:104–11 VEGFA/SOD2 AND LONG-TERM SURVIVAL

CONGEN

ITALHEA

RT

findings reflect a spatiotemporally restricted increasedlevel of VEGFA gene expression in patients with theprotective allele, which, when endogenously expressedand regulated, improves vascular response to ischemicconditions and increases myocardial function after sur-gery, leading to improved long-term outcomes. Withinthis context, our findings could represent a potentialtherapeutic target using the short-term, pulselike VEGFadministration for clinical follow-up after surgical palli-ation of nonsyndromic CHD. The SOD2 intronic variantanalyzed, rs2758331 was studied because of strong link-age disequilibrium (r2 ¼ 0.93) with a well-studied SOD2Val16Ala missense SNP, rs4880, which has been reportedto increase enzyme activity approximately 33% [18],consistent with a protective improvement in the oxida-tive stress response.

Post-hoc literature review found a prior report of theVEGFA major allele (associated with decreased VEGFAexpression [13], and decreased survival probability inour data) being associated with TOF [19]. Our datareplicate this finding, as the VEGFA major allele wasassociated with both TOF and TGA, which representedrelatively minor (low morbidity/low mortality) heartdefects in our cohort (vast majority diagnostic class 1).Additionally, the VEGFA minor allele, which is associ-ated with increased VEGFA expression [13] andincreased survival in our data, was associated with themost severe heart defect, HLHS (all 130 subjects withHLHS were in the highest risk diagnostic class 4). Hadthe VEGFA minor allele been associated with both lowerclass, less severe CHD and improved survival, the classassociation might have led to a false positive survivalassociation. However, we find that the VEGFA minorallele is associated with the most severe CHD, andindependently (as demonstrated by both the adjustmentfor diagnostic class in the original analyses and thesignificant and protective effects of VEGFA in the strat-ified sensitivity analysis of class 4 subjects) is protectiveagainst death or heart transplant in these highest riskchildren. Thus, we conclude that there are not con-founding effects of VEGFA (or SOD2, which was notassociated with diagnostic class) on survival, as theprotective effects of VEGFA were strongest in childrenwith the most severe CHD. Further research into themechanisms through which VEGFA expression in-fluences the type and severity of CHD is nowwarranted.

Some limitations of this study must be considered. First,power was limited owing to the size of the cohort andthe lack of comparable cohorts with which to considerpooling data. We addressed this by limiting the numberof hypotheses, including not attempting genome-wideanalyses. As noted, all SNPs tested had or taggedknown functional effects, improving the probability oftrue associations. However, true positives with smallereffect sizes may have been missed. Second, although thecohort is largely of EA, we used data from subsets of allgenetic ancestries and adjusted for this through usage of astandard statistical method for adjusting for populationstratification [10] and performed sensitivity analyses in

the majority EA subset of the cohort. These analysessuggest that the observed associations are not due topopulation stratification. Finally, our literature-identifiedand tested SNPs do not represent all biologically plau-sible candidates. We could only test variants of interestthat were represented or in high linkage disequilibriumwith SNPs on our genotype chip, limiting hypothesestested.More work must be done to establish these findings

before they can be implemented in the clinical setting.First, these results should be replicated in an independentcohort; unfortunately, no comparable genetic study ofnonsyndromic CHD patients is available at this time.Moreover, it must be noted that any replication of ourwork would require an inception cohort with DNA ob-tained before the first surgery. Failure to obtain DNAbefore the first surgery will likely lead to a survivor bias,as those with the deleterious alleles identified in our workwill die at a rate disproportionate to the rest of the cohort.In conclusion, the results presented offer evidence that

long-term survival of children with nonsyndromic CHDis likely affected by variation in genes involved inoxidative stress and vascular response mechanisms.Given the high incidence of CHD and the frequent needfor surgical palliation, further molecular follow-up ofthese genes is imperative. Identification of these candi-date genes and their differential susceptibility to oxidativeand ischemic stress provides a potential window intonovel pathways that can aid in the development of ther-apies and preventive strategies to aid in decreasing themorbidity and mortality of necessary cardiac surgery ininfants with nonsyndromic CHD.

We would like to thank all subjects and families for theirparticipation. Genotyping was performed by the Center forApplied Genomics at the Children’s Hospital of Philadelphia.This work was supported by a grant from the Fannie E. RippelFoundation, an American Heart Association National Grant-in-Aid (9950480N), HL071834 from the National Institutes ofHealth, and a Washington State Life Sciences Discovery Awardto the Northwest Institute for Genetic Medicine. Daniel SeungKim was supported by 1F31MH101905-01.

References

1. Feinstein JA, Benson DW, Dubin AM, et al. Hypoplastic leftheart syndrome: current considerations and expectations.J Am Coll Cardiol 2012;59(Suppl):1–42.

2. Laffey JG, Boylan JF, Cheng DCH. The systemic inflamma-tory response to cardiac surgery: implications for the anes-thesiologist. Anesthesiology 2002;97:215–52.

3. Clancy RR, McGaurn SA, Goin JE, et al. Allopurinol neuro-cardiac protection trial in infants undergoing heart surgeryusing deep hypothermic circulatory arrest. Pediatrics2001;108:61–70.

4. Kirshbom PM, Mahle WT, Joyner RW, et al. Theendothelin-1 G5665T polymorphism impacts transplant-freesurvival for single ventricle patients. J Thorac CardiovascSurg 2008;136:117–22.

5. Mital S, Chung WK, Colan SD, et al. Renin-angiotensin-aldosterone genotype influences ventricular remodeling ininfants with single ventricle. Circulation 2011;123:2353–62.

6. Gaynor JW, Gerdes M, Zackai EH, et al. Apolipoprotein Egenotype and neurodevelopmental sequelae of infant

Page 7

110 CHAMBERLAIN MEMORIAL PAPER KIM ET AL Ann Thorac SurgVEGFA/SOD2 AND LONG-TERM SURVIVAL 2014;98:104–11

CONGEN

ITALHEA

RT

cardiac surgery. J Thorac Cardiovasc Surg 2003;126:1736–45.

7. Gaynor JW, Nord AS, Wernovsky G, et al. Apolipoprotein Egenotype modifies the risk of behavior problems after infantcardiac surgery. Pediatrics 2009;124:241–50.

8. Kim DS, Stanaway IB, Rajagopalan R, et al. Results ofgenome-wide analyses on neurodevelopmental phenotypesat four-year follow-up following cardiac surgery in infancy.PLoS ONE 2012;7:e45936.

9. Kosoy R, Nassir R, Tian C, et al. Ancestry informative markersets for determining continental origin and admixture pro-portions in common populations in America. Hum Mutat2009;30:69–78.

10. Price AL, Patterson NJ, Plenge RM, Weinblatt ME,Shadick NA, Reich D. Principal components analysis correctsfor stratification in genome-wide association studies. NatGenet 2006;38:904–9.

11. Clancy RR, McGaurn SA, Wernovsky G, et al. Preoperativerisk-of-death prediction model in heart surgery with deephypothermic circulatory arrest in the neonate. J ThoracCardiovasc Surg 2000;119:347–57.

12. Gaynor JW, Mahle WT, Cohen MI, et al. Risk factors formortality after the Norwood procedure. Eur J CardiothoracSurg 2002;22:82–9.

13. Vailati FB, Crispim D, Sortica DA, Souza BM, Brondani LA,Canani LH. The C allele of -634G/C polymorphism in the

VEGFA gene is associated with increased VEGFA geneexpression in human retinal tissue. Invest Ophthalmol VisSci 2012;53:6411–5.

14. Weis S, Shintani S, Weber A, et al. Src blockade stabilizes aFlk/cadherin complex, reducing edema and tissue injuryfollowing myocardial infarction. J Clin Invest 2004;113:885–94.

15. Matsudaira K, Maeda K, Okumura N, et al. Impact of lowlevels of vascular endothelial growth factor after myocardialinfarction on 6-month clinical outcome. Circ J 2012;76:1509–16.

16. Guo HD, Cui GH, Yang JJ, et al. Sustained delivery of VEGFfrom designer self-assembling peptides improves cardiacfunction after myocardial infarction. Biochem Biophys ResComm 2012;424:105–11.

17. Zangi L, Lui KO, Von Gise A, et al. Modified mRNA directsthe fate of heart progenitor cells and induces vascularregeneration after myocardial infarction. Nat Biotech 2013;31:899–907.

18. Bastaki M, Huen K, Manzanillo P, et al. Genotype-activityrelationship for Mn-superoxide dismutase, glutathioneperoxidase 1 and catalase in humans. PharmacogenetGenom 2006;16:279–86.

19. Lambrechts D. Low expression VEGF haplotype increasesthe risk for tetralogy of Fallot: a family based associationstudy. J Med Genet 2005;42:519–22.

DISCUSSION

DR JOHN G. COLES (Toronto, Ontario, Canada): Thank you forallowing me to discuss this excellent presentation. I have nodisclosures.

Let me first point out that this work by Mr Kim and colleaguesis highly innovative because it is the first clinical study of its kind toidentify mechanistic pathways that are specifically relevant tocongenital heart disease. Redox stress, for example, can be gener-ated by chronic hypoxia that occurs in congenital heart disease,and is further aggravated upon repair with reoxygenation. Thestudy therefore has a potential to stimulate further investigationsinto the molecular pathways and useful target discovery effortsthat are unique to our specialty. The manuscript provided a veryarticulate and balanced discussion of the translational significanceof their results.

This work advances, I think for the first time, the concept ofusing personalized genomics in surgery for congenital heartdisease, and that leads me to my first question, Dr Kim. In termsof clinical utility or implementation of these results, are youprepared at this stage to use a patient’s genomic profile withrespect to these two deleterious alleles, or survival alleles, interms of stratifying your surgical pathways for given patients? Sothe first question then relates to clinical implementation of yourfindings.

Secondly, could you please comment on strategies that couldbe used to validate the genes implicated in your analysis otherthan using a new inception cohort, which is problematic forvarious reasons? I assume the two genes implicated in the anal-ysis represent gain of function mutations since they are known tobe cardioprotective in the literature. Specifically, could youcomment on the approach of creating patient-specific car-diomyocytes from these patients harboring deleterious homo-zygous double mutations to test for a confirmatory phenotypein vitro, thereby adding further validity to your clinical results?Again, thank you very much for the excellent and informativepresentation.

MR KIM: With regard to the first question, the implementationof a patient’s genetic information into the clinic, I would like to

say that it’s promising, but not yet. As mentioned previously,validation and replication in an independent cohort is absolutelynecessary before we move forward into any sort of imple-mentation phase. Although these results are very promising, wemust exercise caution and not overinterpret.

With regard to validation, there are several projects under wayto validate these with data that are existing; however, I can’tcomment further on that, and if you have any questions we can talkoffline about that.

With regard to patient-specific cardiomyocytes, induced pro-genitor stem cells and these sorts of things are extremely promisingand are one avenue through which we could pursue validation, butI cannot comment too much further on that because I am not anexpert in these technologies.

DR PETER MCKEOWN (Nicholasville, KY): I want to compli-ment the authors for a really excellent paper, and ask a simplequestion, that if a lot of the injury is related to reperfusion, wouldthere be some simple strategy such as xanthine oxidase in-hibitors or allopurinol that might prevent that before we get to apoint where we can change the genomic strategies?

MR KIM: That is an excellent question. With regard to theclinical management, I would defer that to Dr Gaynor. With re-gard to the allopurinol, Dr Spray has published a paper on that inthe past showing that it had neurocardiac protective effects.

DR DAVID H. HARPOLE (Durham, NC): This is an excellentpaper and I just had a couple of questions for the authors. DrKim, it was unclear to me, were these DNA samples from theheart biopsies or were they from white blood cells in the pe-ripheral blood?

MR KIM: It is from a peripheral blood sample.

DR HARPOLE: Which is what I assumed, which is the way mostof these are done when you are looking at germline type issues.And so one thing I would ask, since these are all patients who

Page 8

111Ann Thorac Surg CHAMBERLAIN MEMORIAL PAPER KIM ET AL2014;98:104–11 VEGFA/SOD2 AND LONG-TERM SURVIVAL

CONGEN

ITALHEA

RT

have had neonatal heart surgery, it would be very interesting todo small biopsies and to compare the single nucleotide poly-morphisms (SNPs) in the primary heart versus the peripheralblood, because you might actually see an upregulation or a highernumber of abnormalities in the muscle of these children withcongenital heart disease and increase the power of your data,because when you are using peripheral blood, in developmentaldefects one would see a differential expression of these SNPs.

And number two is that, as you increase your population ofpatients, it would be most interesting to take blood samples fromnormal children under age 6 months and compare them to yourdata and use germline-wide association studies where you just usethe outcome or how they have done at 5 to 6 years to separate thetwo groups, and you would probably identify much larger cohortsof genes, which may be important in the differentiation of childrenwho do well and don’t do well, take those and then re-put themback into our model as a way.

I think it’s very excellent that you picked two specifichypothesis-driven pathways, but you could probably identify a lotmore information, which may strengthen your model in the future.Thank you.

MR KIM: Thank you very much for the comments.

DR CARL L. BACKER (Chicago, IL): Congratulations on a verywell presented, elegant, and intriguing study. My question is, didyou look at the structural heart disease in the patients youevaluated? Was there any correlation between the number ofhigh-risk alleles and the complexity of the congenital heart dis-ease? You could postulate that if all the patients who had aventricular septal defect closed at 6 months of age had no allelesand the hypoplastic left heart syndrome patients all had fouralleles that the structural heart disease would also influence youroutcomes. Thank you for a very intriguing study.

MR KIM: Thank you very much for the question. And if it’spossible to look at my supplementary slides, I have a table there.We did think about this, and what we wanted to see and makesure is that that group with zero risk alleles, in which there were11 patients, they were a good representation of high-risk ordeleterious structural defects. In fact, 8 of those 11 patients whowere in that low-risk allele group had hypoplastic left heartsyndrome or a variant of it.