Arrhythmia/Electrophysiology. Genetic Testing for Long-QT Syndrome Distinguishing Pathogenic Mutations From Benign Variants

Transcription

1 Arrhythmia/Electrophysiology Genetic Testing for Long-QT Syndrome Distinguishing Pathogenic Mutations From Benign Variants Suraj Kapa, MD*; David J. Tester, BS*; Benjamin A. Salisbury, PhD*; Carole Harris-Kerr, PhD; Manish S. Pungliya, MS; Marielle Alders, PhD; Arthur A.M. Wilde, MD, PhD; Michael J. Ackerman, MD, PhD Background Genetic testing for long-qt syndrome (LQTS) has diagnostic, prognostic, and therapeutic implications. Hundreds of causative mutations in 12 known LQTS-susceptibility genes have been identified. Genetic testing that includes the 3 most commonly mutated genes is available clinically. Distinguishing pathogenic mutations from innocuous rare variants is critical to the interpretation of test results. We sought to quantify the value of mutation type and gene/protein region in determining the probability of pathogenicity for mutations. Methods and Results Type, frequency, and location of mutations across KCNQ1 (LQT1), KCNH2 (LQT2), and SCN5A (LQT3) were compared between 388 unrelated definite (clinical diagnostic score 4 and/or QTc 480 ms) cases of LQTS and 1300 healthy controls for each gene. From these data, estimated predictive values (percent of mutations found in definite cases that would cause LQTS) were determined according to mutation type and location. Mutations were 10 times more common in cases than controls (0.58 per case versus 0.06 per control). Missense mutations were the most common, accounting for 78%, 67%, and 89% of mutations in KCNQ1, KCNH2, and SCN5A in cases and 95% in controls. Nonmissense mutations have an estimated predictive value 99% regardless of location. In contrast, location appears to be critical for characterizing missense mutations. Relative frequency of missense mutations between cases and controls ranged from 1:1 in the SCN5A interdomain linker to infinity in the pore, transmembrane, and linker in KCNH2. These correspond to estimated predictive values ranging from 0% in the interdomain linker of SCN5A to 100% in the transmembrane/linker/pore regions of KCNH2. The estimated predictive value is also high in the linker, pore, transmembrane, and C terminus of KCNQ1 and the transmembrane/linker of SCN5A. Conclusions Distinguishing pathogenic mutations from rare variants is of critical importance in the interpretation of genetic testing in LQTS. Mutation type, mutation location, and ethnic-specific background rates are critical factors in predicting the pathogenicity of novel mutations. Novel mutations in low estimated predictive value regions such as the interdomain linker of SCN5A should be viewed as variants of uncertain significance and prompt further investigation to clarify the likelihood of disease causation. However, mutations in regions such as the transmembrane, linker, and pore of KCNQ1 and KCNH2 may be defined confidently as high-probability LQTS-causing mutations. These findings will have implications for other genetic disorders involving mutational analysis. (Circulation. 2009;120: ) Key Words: genetic testing genetics ion channels long-qt syndrome mutations The modern era of genetic testing has afforded clinicians the opportunity to test for many genetically transmitted diseases. As genetic tests make their transition from investigational tests to bona fide, clinically indicated, and reimbursed tests, it is likely that increasing numbers of patients will undergo genetic analysis, for either specific diseases or eventually of their entire genome, potentially revealing numerous rare mutations of unknown clinical significance, otherwise known as variants of uncertain significance. Currently, over 1500 clinical genetic tests ( are available clinically and are offered through 600 diagnostic laboratories worldwide. 1 Editorial see p 1745 Clinical Perspective on p 1760 Interpreting the significance of a genetic test result in the context of insufficient clinical evidence for a particular Continuing medical education (CME) credit is available for this article. Go to to take the quiz. Received March 5, 2009; accepted August 17, From the Department of Medicine/Division of Cardiovascular Diseases (S.K., M.J.A.), Department of Molecular Pharmacology and Experimental Therapeutics/Windland Smith Rice Sudden Death Genomics Laboratory (D.J.T., M.J.A.), and Department of Pediatrics/Division of Pediatric Cardiology (M.J.A.), Mayo Clinic, Rochester, Minn; Clinical Genetics, PGxHealth, LLC, New Haven, Conn (B.A.S., C.H.-K.); Genaissance Pharmaceuticals, Inc, New Haven, Conn (M.S.P.); and Departments of Clinical Genetics (M.A.) and Cardiology (A.A.M.W.), Academic Medical Center, Amsterdam, the Netherlands. The online-only Data Supplement is available with this article at *The first 3 authors contributed equally to this work. Reprint requests to Michael J. Ackerman, MD, PhD, Director, Long QT Syndrome Clinic and the Mayo Clinic Windland Smith Rice Sudden Death Genomics Laboratory, Mayo Clinic, Guggenheim 501, 200 First St SW, Rochester, MN ackerman.michael@mayo.edu 2009 American Heart Association, Inc. Circulation is available at DOI: /CIRCULATIONAHA

2 Kapa et al LQTS Genetic Testing 1753 disease phenotype (or pretest probability of disease) is challenging for medical geneticists, disease subspecialists, and primary physicians. 2 A recent editorial by Hunter et al 3 detailed some of the challenges with genetic testing and the associated diagnostic implementation. For sudden death predisposing diseases like congenital long-qt syndrome (LQTS), in which highly effective medicine-, surgery-, and device-related therapies are available, proper interpretation of the genetic test is critical considering the diagnostic, prognostic, and therapeutic implications of a positive LQTS test. 4 Since the sentinel discovery of the first LQTSsusceptibility locus on chromosome 11 in and the discovery of mutations in key cardiac potassium and sodium channels in 1995, 6,7 LQTS has been understood as a cardiac channelopathy. To date, hundreds of LQTS-causing mutations in at least 12 LQTS-susceptibility genes have been described, and a litany of genotype-phenotype studies has revealed relatively gene-specific ECG patterns, arrhythmogenic triggers, risk for sudden death, and responses to pharmacotherapy Recognizing the essential role of LQTS genetic testing in the evaluation and management of patients, a handful of research laboratories conducted pseudoclinical testing from 1995 to In 2004, LQTS genetic testing for 5 LQTS-susceptibility genes, including the 3 major genes (KCNQ1 [LQT1], KCNH2 [LQT2], and SCN5A [LQT3]), became commercially available in North America. Mutations in the 3 major LQTS-susceptibility genes account for 70% to 75% of congenital LQTS cases; the 9 other minor genes contribute an additional 5%. 15 The remaining 20% to 25% of LQTS continues to be genetically elusive. Genetic variants identified through genetic testing may represent known polymorphisms; known diseasecausing mutations; radical mutations, including splice, inframe, nonsense, and frame-shift mutations; silent variants; or rare, nonsynonymous single nucleotide polymorphisms/missense mutations that have not been identified or characterized previously. In addition to the published and Web-available compendia of putative LQTS-causing mutations, 2 compendia of rare potassium channel and sodium channel variants discovered among 800 ostensibly healthy volunteers have been published. 16,17 These control data not only established the ethnicity-dependent frequencies for common polymorphisms but also revealed that 5% of otherwise healthy individuals (3% to 4% of whites and 6% to 8% of blacks) may have what could be interpreted as a positive genetic test. 16,17 In other words, nearly 1 in 20 controls hosts a rare ( 0.5% allelic frequency, typically unique) genetic variant that alters the amino acid sequence of one of the channel proteins. Considering that LQTS has an estimated incidence of 1 in 2500 persons, the vast majority of these rare variants found in controls must be innocuous, having no clinical relevance and little or no functional difference from wild type. This so-called background genetic noise, now quantified for the LQTS genes, is almost certain to be present for virtually every disease-susceptibility gene and serves as a powerful reminder that genetic tests, like most diagnostic tests, are probabilistic rather than binary. Distinguishing pathogenic mutations from otherwise just rare, just there mutations is of paramount importance. Heterologous expression studies of every ion channel variant to assess in vitro whether the mutation confers a perturbed molecular/cellular phenotype are prohibitively time consuming, sometimes not translatable from heterologous system to human host, and thus not suitable for regulated, clinical testing. To date, 20% of all published LQTS-associated mutations have been characterized functionally. New variants are being discovered at a far greater rate than characterization experiments can be performed currently. Therefore, in an effort to enhance the diagnostic interpretability of the next novel genetic variant, we conducted a large multicenter case-control study examining the properties of mutations derived from high-probability clinical cases compared with those similarly rare variants derived from ostensibly healthy volunteers. Here, we have expanded on previous observations to quantify the signal-to-noise ratio in distinguishing a pathogenic mutation from an innocuous one based on mutation type and location. Methods Case-Control Study Design From 1997 to 2007, 1300 patients were referred to Mayo Clinic s Windland Smith Rice Sudden Death Genomics Laboratory (Rochester, Minn), the Cardiogenetics Clinic at the Academic Medical Center (Amsterdam, the Netherlands), or PGxHealth (New Haven, Conn) for comprehensive LQTS genetic testing after providing written, informed consent. In an effort to polarize the cohorts (cases versus controls) as cleanly as possible, only clinically definite unrelated index cases were examined, amounting to 388 of the total referral population, which increases the a priori or pretest probabilities that a case-derived mutation is indeed pathogenic. For the purpose of this case-control mutation analysis, clinically definite cases were defined as those with a clinical diagnostic score (Schwartz score) 4 or a corrected QT interval (QTc) 480 ms. These criteria have been shown to correlate with a high clinical probability for LQTS. 18 All cases were unrelated, and demographics, including ethnicity, age, and sex, were recorded at the time of sequencing. The spectrum and prevalence of case mutations, as well as mutation type and mutation location, were compared with the characteristics of the genetic variants found among controls. The controls consisted of 744 unrelated seemingly healthy volunteers analyzed previously for K-channel variants 16 and the 829 healthy volunteers analyzed previously for Na-channel variants, 17 along with 500 additional non-lqts volunteers sequenced for this study by PGxHealth (Table 1). DNA samples were obtained from the Human Genetic Cell Repository sponsored by the National Institute of General Medicine Sciences and the Coriell Institute for Medical Research (Camden, NJ), 16,17 as well as anonymized blood donors. An ECG showing a normal QT interval was not a prerequisite for subjects in this control cohort. Table 1 summarizes the ethnic distribution of all cases and controls. LQTS Genetic Testing The genomic DNA of all cases and controls was analyzed for mutations in the 60 translated exons and the splice-site regions of the 3 major LQTS-susceptibility genes: KCNQ1, KCNH2, and SCN5A, responsible for LQT1, LQT2, and LQT3, respectively. Mutation analyses were performed with polymerase chain reaction and either denaturing high-performance liquid chromatography, followed by automated DNA sequencing or direct high-throughput DNA sequencing. 19

3 1754 Circulation November 3, 2009 Table 1. Ethnic Distribution of Cases and Controls Controls, n Ethnicity Cases, n KCNQ1 KCNH2 SCN5A Black White Asian Hispanic Unknown/other Total Mutation Sequence Analysis Only genetic variants predicted to alter the open reading frame, thereby affecting the primary amino acid composition of the cardiac channel, were included. To be considered a possible pathogenic mutation in this study, the particular genetic variant found in a case must not have been observed in our control population. Variants seen in controls more than once were considered polymorphisms for purposes of this study and were excluded from analysis. Accordingly, common amino acid substitutions, also known as common polymorphisms, such as KCNH2-K897T and SCN5A-H558R did not influence the analyses of mutation type and topological locations in this study. The term mutation in this article is not meant to imply pathogenicity or even functional abnormality, merely rarity and predicted protein alteration as defined above. Mutations were then classified using standard nomenclature. 20 Splice-site, nonsense, frame-shift, and in-frame insertions/deletions were grouped together as radical mutations. Mutations were localized to exon and specific gene-encoded protein regions according to a combination of Swissprot ( ca.expasy.org/uniprot/) and recent studies of genomic and protein organization for the 3 genes The Swissprot database offers generally accepted residue ranges corresponding with each region and specialized subregions of the ion channel, used here with slight modifications. For KCNQ1 and KCNH2, mutations were characterized as being located in the N terminus (KCNQ1, amino acids 1 to 121; KCNH2, amino acids 1 to 403), the region between the N and C termini comprising the transmembrane, linker, and pore (KCNQ1, amino acids 122 to 348; KCNH2, amino acids 404 to 659), or the C terminus (KCNQ1, amino acids 349 to 676; KCNH2, amino acids 660 to 1159). 14 For SCN5A, mutations were characterized as being located in the N terminus (amino acids 1 to 126), interdomain linker (IDL I to II, amino acids 416 to 711; IDL II to III, amino acids 940 to 1200; and IDL III to IV, amino acids 1471 to 1523), transmembrane/linker (domain I, amino acids 127 to 415; domain II, amino acids 712 to 939; domain III, amino acids 1201 to 1470; and domain IV, amino acids 1524 to 1772), or C terminus (amino acids 1773 to 2016). Further subcategorizations were used for 4 specialized subregions: subunits assembly domain (SAD; amino acids 589 to 620) of the C terminus of KCNQ1, the Per-Arnt-Sim (PAS; amino acids 41 to 70) and PAS-associated C terminal (PAC; amino acids 92 to 144) subdomains of the N terminus of KCNH2, and the cyclic nucleotide-binding domain (cnbd; amino acids 742 to 842) of KCNH2. Statistical Analysis Variables were analyzed through the use of the Fisher exact test. Frequencies of each type of mutation were computed for all subjects and a whites-only subset analysis. Further analyses were performed within particular channel regions. Each mutation was counted every time it was seen because the statistics concern frequencies of people carrying particular classes of mutation. To estimate the likelihood of disease causation, a modification of the positive predictive value was used. We made the logical assumption that the incidence of benign, background mutations is the same for the case and control populations. We further made the simplifying assumption that all mutations found in controls are benign, background mutations, a reasonable assumption for the Figure 1. Topological depiction of all protein-altering mutations. Open circles represent case mutations; solid circles, rare (each observed only once) genetic variants observed among the 1300 healthy volunteers; and solid squares, the genetic variants/polymorphisms observed more than once among controls. All 259 distinct mutations, both radical and missense types, are included. Shaded regions: 1, SAD; 2, PAS; 3, PAC; and 4, cnbd. purpose of these analyses given the low prevalence of LQTS. On this basis, the case frequency of mutations in excess of the control frequency should represent the frequency of disease-causing mutations in cases. Using these principles, along with the genetic testing results of our study populations, we then calculated estimated predictive values (EPVs), the probability of pathogenicity of a mutation found in a case, where EPV (case frequency control frequency)/case frequency. The 95% confidence intervals (CIs) were calculated for all EPVs. Upper and lower bounds of the 95% CIs were calculated from the following formula: CI 1 1/(ê{ln (RR) z*[se(log RR)]}), where RR is the relative ratio (mutation frequency in cases divided by the mutation frequency in controls), z for 1 95%, and SE[log(RR)] is the SE around the log of RR. The EPVs calculated here are specific to clinically definite cases as defined above. Results Frequency of Mutations in Cases and Controls In total, 248 mutations (180 distinct: 129 missense mutations, 51 radical) were found among 224 of the 388 LQTS cases (58%). In contrast, only 79 unique mutations (77 missense) were found across the 3 genes among the 1300 controls (Figure 1 and Table I in the online-only Data Supplement). Although each control had no more than 1 mutation, 22 cases had 2 mutations across the 3 LQTS-causing genes, and 1 had 3. Nine cases had 2 mutations in KCNQ1, 2 had 2 mutations in KCNH2, and 1 had 2 mutations in SCN5A, whereas the remaining multiples included 1 gene. In addition, 31 single amino acid substitutions, 2 in-frame deletions, and 1 in-frame duplication were seen more than once in the controls (Figure 1 and Table I in the online-only Data Supplement). Mutations were significantly more common in cases than controls in each of the 3 genes but to different degrees (KCNQ1: 112 of 388 cases [0.29 per subject] versus 14 of 1344 controls [0.01/subject], P ; KCNH2: 108 of 388 cases [0.28 per subject] versus 28 of 1369 controls [0.02 per subject], P ; SCN5A: 28 of 388 cases [0.07 per subject] versus 37 of 1380 controls [0.03 per subject], P ). Table 2 shows how these mutation rates

4 Kapa et al LQTS Genetic Testing 1755 Table 2. Ethnic-Specific Differences in Mutation Rates for Cases and Controls Whites, n (%) Nonwhites, n (%) Gene Case Control Case Control KCNQ1 95/326 (0.29) 4/559 (0.007) 9/26 (0.35) 10/699 (0.015) KCNH2 89/326 (0.27) 11/559 (0.02) 9/26 (0.35) 16/724 (0.022) SCN5A 25/326 (0.077) 8/647 (0.012) 2/26 (0.077) 29/647 (0.045) Data are listed as the number of mutations seen/number of subjects tested, with rate (mutations per subject) in parentheses. Those of race defined as unknown or other (n 36 cases, n 86 controls) were not included in either group. break down by ethnicity, with typically lower mutation rates among white cases and controls than nonwhite cases and controls. The biggest ethnic difference was the greater mutation rate in nonwhite controls (0.08 per subject) than white controls (0.04 per subject; P ). Although 25% of all mutations in cases were considered radical mutations, such mutations were almost never observed among controls. In fact, the presence of a radical mutation (splice-site, nonsense, frame-shift, or in-frame insertions/deletions in KCNQ1 and KCNH2; in-frame insertions/deletions in SCN5A) confers an EPV 99% regardless of gene or gene region. The presence of a radical mutation in KCNQ1 confers an EPV of 100% (95% CI, 91 to 100); in KCNH2, an EPV of 99% (95% CI, 94 to 100); and in SCN5A, an EPV of 91% (95% CI, 10 to 99). Missense mutations were the most common type of mutation, representing 74% (184 of 248) of all case mutations, including 78% (87 of 112) of KCNQ1, 67% (72 of 108) of KCNH2, and 89% (25 of 28) of SCN5A mutations and 95% (77 of 79) of the control mutations. In contrast to radical mutations, the probability of pathogenicity, as captured in EPVs, for missense mutations is highly dependent on the gene and protein region in which the mutation is discovered, as detailed in the next section. Gene-Specific EPV Analysis for Missense Mutations KCNQ1 (LQT1-Susceptibility Gene) Figure 2 depicts the relative incidence of missense mutations in cases versus controls divided on the basis of location. For all regions within the KCNQ1-encoded Kv7.1 potassium channel, there is a higher relative frequency of missense mutations in cases than in controls for all ethnicities (3.5 times higher in cases than controls in the N terminus, 24 times high in the transmembrane/linker/pore-spanning region, and Figure 2. Mutation rates across gene-encoded regions. Depicted are the relative frequencies of mutations of all types (A) and missense only (B) in cases vs controls. Note the relatively higher rate of mutations of all types across most regions in cases than controls. Furthermore, in most regions, there is a relatively higher frequency of missense mutations in cases than controls. The exceptions are the N terminus and IDL regions of SCN5A. N indicates N terminus; T/L/P, transmembrane/linker/ pore; C, C terminus; and T/L, transmembrane/linker.

5 1756 Circulation November 3, 2009 Table 3. EPVs for Missense Mutations in Clinically Definite LQTS Cases Mutation Positive Protein and Location Case Control EPV (95% CI) KCNQ1 (Kv7.1) N terminus 1/388 1/ (0 98) Transmembrane/linker/pore 50/388 7/ (91 98) C terminus 36/388 6/ (89 98) SAD 7/388 0/ (64 100) C terminus outside SAD 29/388 6/ (86 98) KCNH2 (Kv11.1) N terminus 10/388 13/ (17 84) PAS/PAC domains 5/388 0/ (48 100) N terminus outside 5/388 13/ (0 74) PAS/PAC domains Transmembrane/linker/pore 47/388 0/ (95 100) C terminus 15/388 14/ (46 87) Cyclic nucleotide-binding 6/388 0/ (58 100) domain C terminus outside cnbd 9/388 14/ (0 81) SCN5A (NaV1.5) N terminus 0/388 2/ (0 95) Transmembrane/linker 12/388 5/ (67 96) IDL 5/388 22/ (0 53) C terminus 8/388 7/ (33 91) EPV (case rate control rate)/case rate. 22 times higher in the C terminus). The EPV for a mutation localizing anywhere in KCNQ1 was 96% (95% CI, 94 to 98) and for a missense mutation was 95% (95% CI, 92 to 97). Although the EPV, or probability for pathogenicity, for a missense mutation localizing to the N terminus of KCNQ1 was only 71% (95% CI, 0 to 98), an EPV 90% was suggested for missense mutations localizing to the transmembrane/linker/pore and C-terminus domains of KCNQ1 (Table 3). There were no control mutations localizing to the specialized SAD of the KCNQ1 C terminus, suggesting a high probability of pathogenicity of mutations in this region (nominal EPV 100%). The EPVs of missense mutations localizing to each region in KCNQ1 are shown in Figure 3. KCNH2 (LQT2-Susceptibility Gene) Mutations were discovered in cases throughout all regions of the KCNH2-encoded Kv11.1 potassium channel. In contrast, missense mutations discovered in controls localized only to the N and C termini of Kv11.1. The relative frequency of mutations discovered in the N and C termini in cases versus controls was 2.7 and 3.8 times, respectively. The EPV for any type of mutation localizing anywhere in KCNH2 was 93% (95% CI, 89 to 95) and for a missense mutation was 89% (95% CI, 84 to 93). There was an extremely high probability (EPV 100%; 95% CI, 95 to 100) of pathogenicity for missense mutations localizing to the linker, transmembrane, and pore regions of KCNH2 but lower EPVs for missense mutations in the N terminus (63%; 95% CI, 17 to 84) or C terminus of KCNH2 (74%; 95% CI, 46 to 87]). There were no Figure 3. EPVs for KCNQ1. (A), EPVs in clinically definite cases for a missense mutation occurring in each major structurefunction domain in the KCNQ1-encoded Kv7.1 potassium channel. The 95% CIs are given in parentheses. Open circles represent case mutations; solid circles, rare genetic variants observed only once among healthy volunteers; and solid squares, genetic variants/polymorphisms observed more than once among controls. (B), Shaded region 1 represents the SAD. control mutations localizing to the specialized PAS/PAC regions of the N terminus or the cnbd region of the C terminus, suggesting a very high EPV (100%; 95% CI, 78 to 100) for those subregions but reduced EPVs for the remainder of the N-terminal (26%; 95% CI, 0 to 74) and C-terminal (56%; 95% CI, 0 to 81) regions. These EPVs are summarized in Table 3 and Figure 4. SCN5A (LQT3-Susceptibility Gene) The greatest prevalence of missense mutations observed in the control population occurred in SCN5A, for which 3% of controls were positive for a mutation. The EPV for any type of mutation localizing anywhere in the SCN5A-encoded NaV1.5 sodium channel was 63% (95% CI, 40 to 77) and for a missense mutation was 60% (95% CI, 33 to 75). Although clustered in the 3 IDLs, control mutations were scattered throughout all gene regions, highlighting a significantly greater degree of genetic background noise in SCN5A than in either KCNQ1 or KCNH2. Moreover, given that the case frequency of missense mutations is lowest in SCN5A, the relative frequency of missense mutations found in cases compared with controls was quite low (Figure 2B). In fact, for both the N terminus and the IDLs, the control mutation rates exceeded those of the case mutation rates, making the calculated point estimates of EPVs for these regions less than the logical limit of 0%, presumably because of sampling error and perhaps the greater nonwhite population among the controls compared with the cases. Given that the third IDL region (DIII to DIV) may have different biophysical properties, making mutations there more likely to be pathogenic, we recalculated the EPV specific to the DIII to DIV region and the remainder of the IDLs separately. This did not significantly change the EPV conferred by mutations localizing to IDL1 (DI to DII) or IDL2 (DII to DIII), which was still below the logical limit of 0% (95% CI, 0 to 49). Similarly, the EPV of a mutation localizing to IDL3 (DIII to DIV), although 72%, had a large 95% CI with a lower bounds 0% and an

6 Kapa et al LQTS Genetic Testing 1757 Figure 5. EPVs for SCN5A. EPVs in clinically definite cases for a missense mutation occurring in each major region in the SCN5A-encoded NaV1.5 sodium channel. The 95% CIs are given in parentheses. Open circles represent case mutations; solid circles, rare genetic variants observed only once among healthy volunteers; and solid squares, the genetic variants/polymorphisms observed more than once among controls. Figure 4. EPVs for KCNH2. (A), EPVs in clinically definite cases for a missense mutation occurring in each major region in the KCNH2-encoded Kv11.1 potassium channel. The 95% CIs are given in parentheses. Open circles represent case mutations; solid circles, rare genetic variants observed only once among healthy volunteers; and solid squares, the genetic variants/polymorphisms observed more than once among controls. (B), Shaded regions 1 and 2 are the specialized PAS and PAC regions, respectively; shaded region 3 represents the specialized cnbd. upper bounds of 98%, likely secondary to small sample size. The EPV was moderate for the C-terminus region (75%; 95% CI, 33 to 91; Figure 5). The region in SCN5A with the greatest association between mutation discovery and disease causation was the transmembrane/linker region, in which the EPV was 88% (95% CI, 67 to 96). Discussion Part of the clinical utility of a diagnostic test is the ability to recognize whether results of a particular clinical test fit within a normal range defined by population studies of presumably nondiseased patients. However, further clinical studies may redefine the normal range when taking into account other measures of clinical likelihood. One hallmark example is the cholesterol panel, the acceptable results of which have come to be dependent on the clinical likelihood of myocardial disease. 24 Thus, the test has limited value as a standalone or screening measure of risk but is invaluable in the management of at-risk patients. It is becoming increasingly clear that these concepts extend to genetic testing because the presence of a mutation alone needs to be considered in clinical and epidemiological contexts. Although once hoped to be binary in interpretation and still often incorrectly viewed as such, genetic tests must be viewed fundamentally as probabilistic tests. Indeed, genetic screening raises several new questions relative to patient management because genetic testing is meant variably to verify clinical suspicion or to assess susceptibility to disease. Certain mutations, particularly radical mutations, may suggest a high likelihood of disease, regardless of other clinical evidence. However, some uncharacterized missense mutations may be difficult to interpret. Recent efforts in various diseases have concentrated on methods for determining the probability of mutation pathogenicity. 2,6 Very few studies address the interpretability of novel, rare genetic variants in the clinical expression of genetically associated diseases. One of the largest population studies of this, done on the BRCA1 and BRCA2 genes for breast and ovarian cancer, suggested that disease causation may be difficult to assess and depends in part on the availability of a large control population to assess the relative frequency of novel missense mutations. 25 However, for diseases such as LQTS in which disease incidence is smaller and phenotypic expression can be surreptitious, outcomes of overinterpretation may lead to unnecessary invasive therapies and underinterpretation may, although rarely, lead to death. Therefore, the ability to accurately distinguish pathogenic mutations from innocuous mutations is vital. One of the main limitations in achieving this level of differentiation for individual mutations is the relative paucity of independent validations of pathogenicity with heterologous expression studies. In fact, 20% of all published LQTS-causative mutations have been studied functionally in this manner by all of the channel function research laboratories to date throughout the world combined. Furthermore, these studies may be hard to control because of different cellular expression systems and the many other variables that may affect interpretation of findings. Thus, by design, all such functional testing was explicitly ignored in our study to allow us to focus instead on developing evalu-

7 1758 Circulation November 3, 2009 ative criteria for the probability of pathogenicity based solely on mutation type and location. Many mutations encountered during ongoing LQTS genetic testing continue to represent novel missense mutations (data not shown). As we have shown, the occurrence of this type of mutation does not necessarily imply disease causation, given the background noise rate of 4% among whites and up to 8% among nonwhites gleaned here from an examination of the 3 main genes in 1300 controls. Thus, even when the pretest probability of LQTS is high, other factors still need to be taken into account during interpretation of the genetic test and counseling of the patient. The results of our study raise several key issues relative to genetic testing in general and LQTS genetic testing in particular. First, there exists background genetic noise in the general population, and the largest component of this noise consists of very rare missense mutations. The only way to estimate the amount of noise is by complete testing of large numbers of ostensibly healthy controls. To the best of our knowledge, the only other disease-susceptibility gene that has undergone this level of scrutiny among ostensibly healthy subjects is the BRCA1 breast cancer susceptibility gene. It would be expected that the same frequency of noise in the general population should also exist in patients with disease; thus, the presence of a mutation alone, despite a high index of clinical suspicion, may not be sufficient to establish pathogenicity. In the case of LQTS, the established yield of genetic testing among clinically irrefutable cases of LQTS is 70% to 75%. 26 This implies that even among absolute cases, there is an estimated 6% chance among white subjects that a positive genetic test is a false positive. In addition, this background noise rate is ethnicity dependent, and there is a greater chance (up to 10%) for false positives to occur among blacks receiving the same LQTS genetic test. As the a priori or pretest clinical probability for a diagnosis of LQTS decreases, the probability that the identified mutation is simply just there, just rare rather than being an LQTS-causative mutation increases. In other words, EPVs (or probability of pathogenicity) for mutations found in lower-probability patients would be correspondingly lower, to the extreme of effectively 0%, except perhaps for the most critical protein regions, if testing were done as part of a universal screening program for this 1:2500 disease. However, the higher the certainty of disease is, the more likely it is that a mutation is causative and that the EPVs provide an upper limit on those probabilities. In calculating the EPVs for mutations occurring in each gene region, we sought to provide a means of estimating whether a variant discovered in a proband may be causative, although this does not equate to diagnosing the proband. Ideally, strong linkage analysis and good functional studies would be available for every mutation, but this is usually not possible, certainly not at the moment of first discovery. Thus, the EPV is helpful in guiding initial decision making in the absence of more conclusive data on each particular mutation. Second, the nature of the detected mutation matters greatly in predictions of the probability of pathogenicity. In particular, radical mutations, specifically nonsense, frameshift, and splice-site mutations, have a nearly 100% EPV for disease pathogenicity involving the 2 major potassium channels implicated in LQTS. In this study, one fourth of the case mutations were of this variety, enabling high-confidence positive test interpretations for this subset. For LQTS genetic testing, the challenge lies in the discernment of rare, single amino acid substitutions (missense mutations), which make up 75% of the positive test results. Third, the probability of pathogenicity for a particular missense mutation may be affected profoundly by its location in the protein. 27 In the case of LQTS genetic testing, the specific location of a mutation in the complex structure of the voltage-gated ion channel confers markedly different probabilities of pathogenicity. Thus, knowledge about the specific gene-encoded regions (ie, structure-function domains) is a critical determinant underlying the probability of pathogenicity. In this study, missense mutations represent highprobability disease mutations when localized to a channel s pore, transmembrane-spanning domains, and other critical subdomains, whereas extraordinary caution is necessary when interpreting the potential significance of similarly rare single amino acid substitutions in other locations such as the IDL I to II of the SCN5A-encoded NaV1.5 sodium channel. Thus far, secondary analyses involving amino acid conservation, inspection of physicochemical properties, and in silico predictions of impact on secondary structure have not helped to further refine these estimates of pathogenicity in low-epv domains such as the C terminus of Kv11.1 or the DI to II and DII to III linkers of NaV1.5 (data not shown). Limitations The cases in this study represented a referral population not necessarily evaluated clinically at either LQTS specialty center. Previous studies have demonstrated a 70% to 75% yield from LQT1 through LQT3 genetic testing among single center derived, definite cases. 26 The case yield in this study was 58%, indicating that some of the cases may not have been of the highest a priori clinical probability. It can also be assumed that not all mutations found in our cases are pathogenic. In fact, this is a central premise of and motivation for our study. Specifically, the overall yield of 6% among the entire control cohort implies that 10% of the case mutations elucidated in this study may be false positives. This estimate may be high for 2 reasons. First, our case population has a much higher representation of whites, in whom the background rates are lower but the overall case mutation yield is similar. This same skew in ancestry also suggests that the EPVs in this article are relatively underestimated for whites and overestimated for nonwhites. To estimate these ethnicspecific EPVs would require significantly larger sample sizes than are currently available. Second, the 10% false-positive estimate above may be overestimated and our EPVs underestimated because this article is based on a historical control population. The next 1300 controls would presumably have considerably fewer novel variants because many of the rare variants would have been observed before in controls, or even in cases. Second, the 79 rare mutations found in the control cohort are not guaranteed to all be 100% benign. Because a normal

8 Kapa et al LQTS Genetic Testing lead ECG with a normal QTc was not a prerequisite for the ostensibly healthy volunteers who made up the control cohort, it is possible that some of the controls may have LQTS; therefore, their particular rare variant could be falsely considered part of the background genetic noise. For example, we recently found in an expanded cohort that the background rate may be 3% among whites (data not shown). Although acknowledging that some mutations among controls may represent incompletely penetrant mutations that may be pathogenic in another patient, the average penetrance of these variants must be extremely low. Otherwise, the prevalence of LQTS must be far greater than believed or is plausible. Based on an incidence of 1 in 2500 of LQTS among the general population, we can be 98% confident that no more than 2 mutations listed in the control compendia are strongly pathogenic mutations rather than background noise. Among the 79 mutations found in the 1300 controls, one that is extremely suspicious for possible pathogenic status is the KCNQ1-A300T missense mutation that localizes to the pore-forming region. In addition, the radical T A splice mutation in SCN5A might lead to loss of channel function, which might be predicted to confer a Brugada syndrome phenotype. Despite representing arguably the second largest and most expansive examination of background genetic noise among any of the clinically available genetic tests, perhaps the biggest limitation of this study is still the relatively small sample size ( 3000 reference alleles) of the controls. Given the overall low control mutation level, on which genetic testing generally draws its strength, the EPVs of some regions may be strongly influenced by small numbers, as is reflected in the sometimes broad CIs. Given that a timely functional assay does not exist to assess whether a particular variant of uncertain significance can be upgraded or downgraded on the basis of its degree of channel perturbation, further comparative genomic investigations of cases and controls, among both white and nonwhite populations, are needed to improve further the diagnostic accuracy of the LQTS genetic test. Despite these limitations, this study demonstrates clearly that ethnicity, mutation type, and mutation location, particularly channel structure-function domains, are critical determinants of the probability of pathogenicity of a particular mutation in the 3 major LQTS-causing genes. Conclusions When evaluating the results of any genetic test for any disease, we should weigh such factors as location, the possible role of ethnicity, and the possible existence of mutations in other, untested genes in the balance. Genetic testing has the potential to offer valuable clinical information that can facilitate early clinical intervention. In the case of LQTS, the diagnostic, prognostic, and therapeutic implications of genetic testing are now viewed as fundamental to the clinical evaluation. 28,29 However, the presence of a genetic mutation alone cannot supplant clinical evidence, and even in robust cases of LQTS, a positive genetic test result must be carefully reviewed. Indeed, genetic tests must be viewed in most cases by clinicians as probabilistic tests, not binary (yes/no, positive/negative) tests, which places genetic testing in good company with virtually every diagnostic test that we order. Sources of Funding The comparative genomic analyses were performed in Dr Ackerman s research program with support from the Mayo Clinic Windland Smith Rice Comprehensive Sudden Cardiac Death Program. Dr Wilde s research program is supported by Interuniversity Cardiology Institute the Netherlands project 27 and the Leducq program grant Alliance Against Sudden Cardiac Death. Disclosures Drs Salisbury and Harris-Kerr are employees of PGxHealth, which offers the FAMILION LQTS test, and stockholders of the parent company, Clinical Data. Dr Ackerman is a consultant for PGxHealth and chairs its FAMILION Medical/Scientific Advisory Board (approved by Mayo Clinic s Medical Industry Relations Office and Conflict of Interests Review Board). In addition, cardiac channel gene screen and know-how relating to long QT genetic testing license agreements, resulting in consideration and royalty payments, were established between Genaissance Pharmaceuticals (now PGx- Health) and Mayo Medical Ventures (now Mayo Clinic Health Solutions) in The other authors report no conflicts. References 1. GeneTests. Medical Genetics Informational Resource (database online). Seattle, Wash: University of Washington; Available at: GeneTests. Accessed October 1, Evans JP, Skrzynia C, Burke W. The complexities of predictive genetic testing. BMJ. 2001;322: Hunter DJ, Khoury MJ, Drazen JM. Letting the genome out of the bottle: will we get our wish? N Engl J Med. 2008;358: Shimizu W. The long QT syndrome: therapeutic implications of a genetic diagnosis. Cardiovasc Res. 2005;67: Keating M, Atkinson D, Dunn C, Timothy K, Vincent GM, Leppert M. Linkage of a cardiac arrhythmia, the long QT syndrome, and the Harvey ras-1 gene. Science. 1991;252: Wang Q, Shen J, Splawski I, Atkinson D, Li Z, Robinson JL, Moss AJ, Towbin JA, Keating MT. SCN5A mutations associated with an inherited cardiac arrhythmia, long QT syndrome. Cell. 1995;80: Curran ME, Splawski I, Timothy KW, Vincent GM, Green ED, Keating MT. A molecular basis for cardiac arrhythmias: HERG mutations cause long QT syndrome. Cell. 1995;80: Zareba W, Moss AJ, Schwartz PJ, Vincent GM, Robinson JL, Prior SG, Benhorin J, Locati EH, Towbin JA, Keating MT, Lehmann MH, Hall WJ, Andrews ML, Napolitano C, Timothy K, Zhang L, Medina A, MacCluer JW. Influence of the genotype on the clinical course of the long QT syndrome. N Engl J Med. 1998;339: Schwartz PJ, Prior SG, Spazzolini C, Moss AJ, Vincent GM, Napolitano C, Denjoy I, Guicheney P, Breithardt G, Keating MT, Towbin JA, Beggs AH, Brink P, Wilde AAM, Toivonen L, Zareba W, Robinson JL, Timothy KW, Corfield V, Wattanasirichaigoon D, Corbett C, Haverkamp W, Schulze-Bahr E, Lehmann MH, Schwartz K, Coumel P, Bloise R. Genotype-phenotype correlation in the long QT syndrome: gene specific triggers for life threatening arrhythmias. Circulation. 2001;103: Priori SG, Schwartz PJ, Napolitano C, Bloise R, Ronchetti E, Grillo M, Vincentini A, Spazzolini C, Nastoli J, Bottelli G, Folli R, Cappelletti D. Risk stratification in the long-qt syndrome. N Engl J Med. 2003;348: Zareba W. Genotype-specific ECG patterns in long QT syndrome. J Electrocardiol. 2006;39:S101 S Napolitano C, Bloise R, Priori SG. Gene-specific therapy for inherited arrhythmogenic disorders. Pharmacol Ther. 2006;110: Tan HL, Bardai A, Shimizu W, Moss AJ, Schulze-Bahr E, Noda T, Wilde AAM. Genotype-specific onset of arrhythmias in congenital long QT syndrome: possible therapy implications. Circulation. 2006;114: Moss AJ, Shimizu W, Wilde AAM, Towbin JA, Zareba W, Robinson JL, Qi M, Vincent M, Ackerman MJ, Kaufman ES, QHofman N, Seth R, Kamakura S, Miyamoto Y, Goldenberg I, Andrews ML, McNitt S. Clinical aspects of type-1 long QT syndrome by location, coding type, and bio-

9 1760 Circulation November 3, 2009 physical function of mutations involving the KCNQ1 gene. Circulation. 2007;115: Tester DJ, Ackerman MJ. Novel gene and mutation discovery in congenital long QT syndrome: let s keep looking where the street lamp standeth. Heart Rhythm. 2008;5: Ackerman MJ, Tester DJ, Jones GS, Will ML, Burrow CR, Curran ME. Ethnic differences in cardiac potassium channel variants: implications for genetic susceptibility to sudden cardiac death and genetic testing for congenital long QT syndrome. Mayo Clin Proc. 2003;78: Ackerman MJ, Splawski I, Makielski J, Tester D, Will M, Timothy K, Keating M, Jones G, Chadha M, Burrow C. Spectrum and prevalence of cardiac sodium channel variants among black, white, Asian, and Hispanic individuals: implications for arrhythmogenic susceptibility and Brugada/long QT syndrome genetic testing. Heart Rhythm. 2004;1: Schwartz PJ, Moss AJ, Vincent GM, Crampton RS. Diagnostic criteria for the long QT syndrome: an update. Circulation. 1993;88: Tester DJ, Will ML, Ackerman MJ. Mutation detection in congenital long QT syndrome: cardiac channel gene screen using PCR, dhplc, and direct DNA sequencing. Methods Mol Med. 2007;128: Antonarakis SE. Recommendations for a nomenclature system for human gene mutations: Nomenclature Working Group. Hum Mutat. 1998; 11: Splawski I, Shen J, Timothy KW, Vincent M, Lehmann MH, Keating MT. Genomic structure of three long QT syndrome genes: KVLQT1, HERG, and KCNE1. Genomics. 1998;51: Wang Q, Li Z, Shen J, Keating MT. Genomic organization of the human SCN5A gene encoding the cardiac sodium channel. Genomics. 1996; 34: Neyroud N, Richard P, Vignier N, Donger C, Denjoy I, Demay L, Shkolnikova M, Pesce R, Chevalier P, Hanique B, Coumel P, Schwartz K, Guicheney P. Genomic organization of the KCNQ1 K channel gene and identification of C-terminal mutations in the long-qt syndrome. Circ Res. 1999;84: Grundy S, Cleeman J, Bairey Merz C, Brewer Jr H, Clark L, Hunninghake D, Pasternak R, Smith S Jr, Stone N. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. J Am Coll Cardiol. 2004;44: Easton DF, Deffenbaugh AM, Pruss D, Frye C, Wenstrup RJ, Allen-Brady K, Tavtigian SV, Monteiro AN, Iversen ES, Couch FJ, Goldgar DE. A systematic genetic assessment of 1,433 sequence variants of unknown clinical significance in the BRCA1 and BRCA2 breast cancer-predisposition genes. Am J Hum Genet. 2007;81: Tester DJ, Will ML, Haglund C, Ackerman MJ. Effect of clinical phenotype on yield of long QT-syndrome genetic testing. J Am Coll Cardiol. 2006;47: Worth CL, Bickerton GR, Schreyer A, Forman JR, Cheng TM, Lee S, Gong S, Burke DF, Blundell TL. A structural bioinformatics approach to the analysis of nonsynonymous single nucleotide polymorphisms (nssnps) and their relation to disease. J Bioinform Comput Biol. 2007; 5: Lehnart SE, Ackerman MJ, Benson DW Jr, Brugada R, Clancy CE, Donahue JK, George AL Jr, Grant AO, Groft SC, January CT, Lathrop DA, Lederer WJ, Makielski JC, Mohler PJ, Moss A, Nerbonne JM, Olson TM, Przywara DA, Towbin JA, Wang LH, Marks AR. Inherited arrhythmias: a National Heart, Lung, and Blood Institute and Office of Rare Diseases workshop consensus report about the diagnosis, phenotyping, molecular mechanisms, and therapeutic approaches for primary cardiomyopathies of gene mutations affecting ion channel function. Circulation. 2007;116: Zipes DP, Camm AJ, Borggrefe M, Buxton AE, Chaitman B, Fromer M, Gregoratos G, Klein G, Moss AJ, Myerburg RJ, Priori SG, Quinones MA, Roden DM, Silka MJ, Tracy C. ACC/AHA/ESC 2006 guidelines for management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: a report of the American College of Cardiology/ American Heart Association Task Force and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Develop Guidelines for Management of Patients with Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death): developed in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society. Circulation. 2006;114:e385-e484. CLINICAL PERSPECTIVE Genetic testing for congenital long-qt syndrome has become an established part of the clinical armamentarium for physicians encountering patients with a personal or family history of ventricular tachyarrhythmias and sudden cardiac death. However, the presence of a positive genetic test result must be reviewed carefully. Most mutations identified in cases of long-qt syndrome are novel missense mutations and thus must be distinguished as either background noise or function-altering/disease-causing mutations. Specifically, even in the general healthy population, genetic variation, including known polymorphisms and novel, rare variants, may be demonstrated. Thus, the presence of a genetic mutation should be interpreted as a probabilistic measure of likelihood of disease rather than as a binary indicator of its presence or absence. This work indicates that mutation type, mutation location, and ethnic-specific background rates are critical factors in predicting the pathogenicity of novel mutations. However, in certain cases, distinguishing pathogenic mutations from rare variants may not be possible without additional functional characterization or linkage analyses. Although mutations localizing to certain specialized regions may be expected with high probability to be disease causing, mutations in other regions should be seen as variants of uncertain significance and should be interpreted in the clinical context and lead to further investigation rather than prompting assumption of disease presence and resulting therapeutic interventions. This classification of genetic test results as a probabilistic rather than absolute measure of disease causation has potential implications when counseling patients, especially as genetic testing becomes less expensive, more easily available, and more commonly used. These findings may also have implications for other genetic disorders involving mutational analysis where benign genetic variation is sure to exist and must be distinguished from disease-causing mutations. Go to to take the CME quiz for this article.