Hypertrophic cardiomyopathy: from genetics to treatment


A. J. Marian, MD, Center for Cardiovascular Genetics, The Brown Foundation Institute of Molecular Medicine, The University of Texas Health Science Center and Texas Heart Institute at St. Luke’s Episcopal Hospital, 6770 Bertner Street, Suite C900A, Houston, TX 77030, USA. Tel.: 713 500 2350; fax: 713 500 2320; e-mail: ali.j.marian@uth.tmc.edu


Eur J Clin Invest 2010; 40 (4): 360–369


Background  Hypertrophic cardiomyopathy (HCM) is the prototypic form of pathological cardiac hypertrophy. HCM is an important cause of sudden cardiac death in the young and a major cause of morbidity in the elderly.

Design  We discuss the clinical implications of recent advances in the molecular genetics of HCM.

Results  The current diagnosis of HCM is neither adequately sensitive nor specific. Partial elucidation of the molecular genetic basis of HCM has raised interest in genetic-based diagnosis and management. Over a dozen causal genes have been identified. MYH7 and MYBPC3 mutations account for about 50% of cases. The remaining known causal genes are uncommon and some are rare. Advances in DNA sequencing techniques have made genetic screening practical. The difficulty, particularly in the sporadic cases and in small families, is to discern the causal from the non-causal variants. Overall, the causal mutations alone have limited implications in risk stratification and prognostication, as the clinical phenotype arises from complex and often non-linear interactions between various determinants.

Conclusions  The clinical phenotype of ‘HCM’ results from mutations in sarcomeric proteins and subsequent activation of multiple cellular constituents including signal transducers. We advocate that HCM, despite its current recognition and management as a single disease entity, involves multiple partially independent mechanisms, despite similarity in the ensuing phenotype. To treat HCM effectively, it is necessary to delineate the underlying fundamental mechanisms that govern the pathogenesis of the phenotype and apply these principles to the treatment of each subset of clinically recognized HCM.


Since its modern recognition approximately 50 years ago, hypertrophic cardiomyopathy (HCM) has continued to fascinate the clinicians and the researchers alike. The interest largely has stemmed from the unusual clinical and pathological manifestations of HCM. Our knowledge of the disease has paralleled the development of various medical disciplines and technologies. French pathologist Liouville, who has been credited with the first description of HCM, characterized it as cardiac contraction below the aortic valve in 1869 [1]. German pathologist Schmincke described HCM as diffuse muscular ‘hyperplasia’ at the left ventricular outflow tract (LVOT) in 1907 [2]. HCM remained a pathological entity until the mid 20th century. The development of advanced diagnostic tools including cardiac catheterization brought in a new phase in our understanding of the disease. Braunwald and colleagues described the dynamic obstruction of the LVOT in a subset of patients with and coined the term ‘idiopathic hypertrophic subaortic stenosis’ or ‘IHSS’ [3,4]. Morrow and colleagues described and established the surgical technique of trans-aortic septal myectomy for the relief of outflow tract obstruction in 1964 [5]. Likewise, systolic anterior motion of the anterior leaflet of mitral valve was recognized as a major contributor to LVOT obstruction and the accompanying mitral regurgitation. Hence, replacement of mitral valve with an artificial valve in patients with LVOT obstruction and mitral regurgitation was considered an effective treatment.

Advances in echocardiographic imaging further demonstrated the preponderance of cardiac hypertrophy to the interventricular septum and the asymmetric type of septal hypertrophy [6]. Today, echocardiography has become the most commonly used imaging modality in the evaluation of patients with HCM. Doppler echocardiography has provided a non-invasive method for evaluation of LVOT obstruction and diastolic function in HCM patients [7–10]. It has supplanted cardiac catheterization for the assessment of LVOT obstruction. The development of tissue Doppler imaging (TDI) further substantiated the presence of regional myocardial contraction and relaxation abnormalities in patients with HCM. The technique was applied to an early diagnosis of mutation carriers prior to expression of cardiac hypertrophy [11,12].

The discovery of the first causal mutation in familial HCM by Dr Seidman’s group in 1990 ushered in the era of molecular genetics [13]. It led to subsequent identification of over a dozen causal genes and several hundred mutations. The discoveries have raised the interest in genetic-based diagnosis, risk stratification and treatment. However, genetic studies have also elucidated the need for a comprehensive approach that utilizes not only the genetic information but also the clinical and other molecular data in the diagnosis and risk stratification. More recently, advances in deep re-sequencing technologies have afforded the opportunity to sequence all known causal genes, the exomes and even the entire genome in any given individual. The technological achievement, however, is expected to expose physicians to flurry of genetic data that they will need to interpret and apply to the care of the patients.

Sudden cardiac death (SCD) remains the most dreaded outcome of HCM and often the first presentation of the disease. Clinical studies have shown the effectiveness of automatic internal cardioverter/defibrillators (AICDs) in prevention of SCD, particularly in those at high risk [14,15]. The challenge is, however, to identify those who are the high risk of SCD. In 1995, Dr Sigwart introduced percutaneous trans-catheter septal ablation or alcohol septal ablation, which has been shown to be an effective method for reduction of the LVOT obstruction [16–19]. A recent noteworthy development is the effectiveness of experimental data in prevention and reversal of cardiac hypertrophy in HCM [20–24]. The findings in animal models of HCM have raised considerable interest in extending the potential beneficial effects of the novel pharmacological agents in treatment and prevention of HCM in humans.

Overview of clinical aspects of HCM

The HCM is a primary disease of cardiac myocytes that is morphologically characterized by concentric and often asymmetric cardiac hypertrophy and a non-dilated left ventricle with preserved or enhanced global systolic function. Pathologically, HCM is characterized by gross cardiac hypertrophy, myocyte hypertrophy and disarray, interstitial fibrosis, hypertrophy/hyperplasia of media of intramural coronary arteries and mitral valve leaflets anomalies. Cardiac myocyte disarray is the pathological hallmark of HCM and often comprises more than 20% of the ventricle.

Cardiac hypertrophy is the quintessential clinical diagnostic feature of HCM. Cardiac hypertrophy is concentric, but commonly asymmetric. The inter-ventricular septum is the predominant site of involvement. The estimated prevalence of HCM in the general population is approximately 1 : 500 [25]. However, the estimate is based on detection of a left ventricular wall thickness of ≥ 15 mm in a relatively young population. The true estimate may be higher because of penetrance. Many patients may show mild hypertrophy or develop hypertrophy later in life [26,27]. The underestimation may be offset by the fact that clinical diagnosis is not sufficiently robust to distinguish the true HCM from many HCM phenocopy conditions.

Most patients with HCM are asymptomatic or mildly symptomatic. Dyspnoea and chest pain are the most common symptoms. Dyspnoea may occur mainly because of high LV end diastolic pressure because of diastolic dysfunction, particularly during exercise. Chest pain may occur because of myocardial hypoperfusion and increased oxygen demand. Palpitations are common and often associated with light-headedness, dizziness and occasionally with syncope. Syncope is an infrequent, but serious symptom, as it often heralds SCD [28,29]. Atrial fibrillation and supraventricular arrhythmias are also common and are often associated with adverse clinical outcome [30].

Overall, HCM is a relatively benign disease with an annual mortality of slightly less than 1% in the unselected HCM population [31,32]. Unfortunately, SCD remains a major concern. It may be the first manifestation of the disease, particularly in the young and competitive athletes [33,34]. Therefore, the primary focus of physicians and patients alike is to determine the risk of SCD and intervene in those at high risk. None of the existing risk factors alone reliably predicts the risk of SCD in patients with HCM. It is important to gather comprehensive information and utilize the combined information in the evaluation of HCM patients. A prior history of sudden cardiac arrest (SCA), syncope caused by cardiac arrhythmias, repetitive non-sustained or sustained ventricular tachycardia, severe cardiac hypertrophy and a strong family history of SCD (causal and modifier genes) are important risk factors [29,35–39]. A pluralistic approach that includes the combination of these risk factors is necessary for proper identification of those at the risk of SCD and the need for an ICD implantation. The latter has been shown to be effective in reducing the risk of SCD as both secondary (prior history of SCA) and primary (no prior SCA) interventions [14,15]. In those with prior SCA, there is clear indication for implantation of ICD. Implantation of ICD for the primary prevention of SCD in those with one risk factor alone, such as severe cardiac hypertrophy, is subject to debate. There is less debate on the implantation of ICDs for the primary prevention of SCD in those who have two or more major risk factors. We advocate the pluralistic approach that is inclusive of all available data. Table 1 offers our simplified approach on weighing the specific risk factors for making a decision for an ICD implantation.

Table 1.   Risk factors for SCD in patients with HCM
Major impact (Recommend AICD implantation)
 Priori episode of SCA
 History of recurrent syncope (suspected due to cardiac arrhythmias)
 Sustained and repetitive non-sustained ventricular tachycardia
Moderate impact (alone not a sufficient indication for AICD implantation, but recommend AICD implantation when present in conjunction with another risk factors)
 Family history of SCD (more than 1 victim of SCD)
  Causal mutations, including double mutations
  Modifier genes
 Severe cardiac hypertrophy
Modest impact (unsettled effect on the risk of SCD)
 Left ventricular outflow tract obstruction
 Abnormal blood pressure response to exercise
 Early onset clinical manifestations
 Myocardial ischemia
 Interstitial fibrosis/disarray

Molecular genetics of HCM

The HCM is familial in approximately half the cases and sporadic in the other half. The mode of inheritance is autosomal dominant. Hence, every offspring has a 50% chance of inheriting the mutation and therefore of developing the disease. As penetrance is age-dependent, many family members may not express the phenotype at the time of examination and may be falsely considered ‘unaffected’. Thus, periodic evaluation of the ‘unaffected’ family members is necessary, as some may develop HCM later in life [40].

The discovery of a missense mutation in the MYH7 gene, which encodes the β-myosin heavy chain (MyHC), about 2 decades ago, provided the first clue to the molecular genetic basis of HCM [13]. The discovery had a watershed effect, as it led to subsequent discoveries of more than a dozen causal genes and several hundred mutations (Table 2). Collectively, the findings have established HCM as a disease of sarcomeric protein: primarily, the thick filaments, to a lesser extent, the thin filaments and uncommonly, the Z disk proteins. MYH7 and MYBPC3 encoding β-MyHC and myosin binding protein-C (MyBP-C) respectively are the two most common genes for HCM [27,41,42]. Mutations in MYH7 and MYBPC3 are responsible for HCM in approximately half of patients [27,41,42]. Several hundred mutations in MYH7 and MYBPC3 already have been published or are listed in the public databases. Majority of the causal mutations in MYH7 are point mutations leading to substitution of one amino acid by another (missense mutations). However, a significant number of the causal mutations in MYBPC3 are insertion/deletion mutations, which often lead to a frame shift and premature truncation of the protein. TNNT2, TNNI3, TPM1 and ACTC1, which code for cardiac troponin T, cardiac troponin I, α-tropomyosin and cardiac α-actin respectively collectively account for about 10–15% of HCM cases [43–46]. Mutations in TNNT2, TNNI3, TPM1 and ACTC1 are mostly missense mutations. The rest of the known causal genes are uncommon causes of HCM, each being responsible for < 1% of HCM cases. The spectrum of the causal mutations in HCM was recently expanded to include mutations in the Z disk proteins MYOZ2 and TCAP, encoding myozenin 2 and telethonin respectively [47,48]. Mutations in MYOZ2 and TCAP are uncommon causes of HCM [47,48]. However, they highlight the increasing recognition of the Z disc proteins as signalling molecules involved in regulating cardiac hypertrophic response. Collectively, the known causal genes account for about 2/3rd of all HCM cases. Thus, 1/3rd of the causal genes for HCM are yet to be identified. One may surmise that each of the remainder causal genes is also responsible for a small fraction of HCM cases.

Table 2.   Causal genes for HCM
 β-Myosin heavy chainMYH7∼25%
 Myosin binding protein CMYBPC3∼25%
 Cardiac troponin TTNNT2∼3–5%
 Cardiac troponin ITNNC1∼3–5%
 Myozenin 2 (calsarcin 1)MYOZ21 : 250
 Myosin light chain 1MYL3Rare
 Myosin light chain 2MYL2Rare
 Myosin light chain kinase 2MYLK2Rare
 α-Myosin heavy chainMYH6Rare
 Cardiac troponin CTNNC1Rare
 Caveolin 3CAV3Rare

Genetic studies, in addition to establishing heterogeneity of the causal genes, have also illustrated considerable allelic heterogeneity. Several hundred causal mutations already have been identified and very few are recurring mutations in a small number of families. Thus, the frequency of each causal mutation is relatively low and most mutations are ‘private’ mutations. The presence of a large number of causal mutations necessitates direct sequencing of a large number of putative causal genes to identify the causal mutation in a given individual and hence makes the task of genetic screening tedious. In conjunction with the genetic heterogeneity of HCM, phenotypic expression of HCM also exhibits a high level of variability. Phenotypic variability also limits the utility of genetic testing in predicting the clinical outcome. Members of a single family, who share the same causal mutation and a significant fraction of their genomes, exhibit considerable differences in the severity of cardiac hypertrophy or the risk of SCD. Variability in the phenotypic expression of HCM is because of multiple factors including heterogeneity of the causal genes and mutations, presence of multiple functional variants in the sarcomeric proteins, double modifier genes and alleles, epigenetic factors and post-translation modifications of the proteins and possibly the environmental factors. Modifier genes/variants, which are neither necessary nor sufficient to cause HCM, influence severity of cardiac hypertrophy as well as risk of SCD [49–53]. In a recent genome-wide analysis, we mapped five modifier loci for human HCM [49,50]. Each locus imparted considerable effects on phenotypic expression of cardiac hypertrophy. The effect sizes of the modifier alleles in the homozygous form were even greater than the effect-size of the causal mutation (as the causal mutation in an autosomal dominant disease is present in heterozygous form only) [49,50]. The effects of epigenetic factors, microRNAs, post-translational modification of proteins as well as the environmental factors on phenotypic expression of HCM remain to be delineated.

The genetic studies also have brought to light the shortcomings of the clinical diagnosis of HCM, as the findings show that an ‘HCM-like’ phenotype, i.e. phenocopy, could arise from storage diseases and metabolic abnormalities, which are discussed later. Give the prevalence of the phenocopy conditions, approximately 5–10% of the clinically diagnosed cases of HCM are probably not HCM, but rather are phenocopy, such as storage diseases. Probably the best-known phenocopy condition is Fabry disease, which is caused by the deficient activity of α-galactosidase A [54,55]. Cardiac hypertrophy in Fabry disease is often indistinguishable from that in HCM. The significance of the distinction is obvious as the two conditions have completely different treatments. Mutations in PRKAG2, which encodes the γ2 subunit of AMP Kinase, cause cardiac hypertrophy, which is also clinically misdiagnosed as HCM [56–58]. The additional phenotypes associated with PRKAG2 are conduction defects and a pre-excitation pattern on the electrocardiogram.


The initial genetic discoveries placed emphasis primarily on HCM caused by mutations in the protein constituents of the thick and thin filaments of sarcomeres. The ensuing mechanistic studies elucidated a diverse array of functional defects including altered Ca2+ sensitivity of myofibrillar ATPase activity and force generation [59–72]. For example, we showed that the Ca2+ sensitivity of myofibrillar ATPase activity in the β-MyHC-Q403 transgenic rabbits, which express the β-MyHC as in the humans [73], was reduced early and prior to expression of hypertrophy (Fig. 1) [62,74]. In contrast, the Ca2+ sensitivity of myofibrillar protein ATPase activity and force generation were enhanced in the cTnT-Q92 transgenic mice [75,76]. The functional phenotypes were also found for HCM-causing mutations in TPM1 and TNNI3 [67,77–84]. Together, the findings illustrate the diversity of the molecular pathways involved in the pathogenesis of human HCM. In accordance with these findings, we promote the notion that the pathogenesis of human HCM involves – at least partially – distinct mechanisms. Accordingly, clinically diagnosed HCM probably entails multiple conditions that all share cardiac hypertrophy as their common gross phenotype.

Figure 1.

 Contrasting effects of mutations on calcium sensitivity of myofibrillar ATPase activity. Calcium sensitivity of myofibrillar ATPase activity is reduced in the β-myosin heavy chain-Q403 (MyHC-Q403) transgenic rabbits as compared to non-transgenic rabbits. In contrast, it was increased in the cardiac troponin T-Q92 (cTnT-Q92) transgenic mice.

Sarcomeres are conventionally recognized as the structural and mechanical units of cardiac myocytes and less so as the modulators of muscle trophic signalling. The M line and Z disk of the sarcomere are emerging as important cell signalling hubs [85]. For example, the Z-disk proteins are implicated in regulating transcriptional regulators of cardiac hypertrophic response including regulation of the calcineurin-nuclear factor of activated T cells pathways and stretch responsive signalling molecules [86]. Identification of mutations in Z disk proteins, such as MYOZ2 and TCAP in patients with HCM has further emphasized the role of sarcomere in modulating hypertrophic signalling pathways [48,87].

According to our current understanding, cardiac hypertrophy in HCM is the consequence of activation of various signalling pathways activated by the causal mutations in sarcomeric proteins (excluding the storage diseases). Genetic mutations affect sarcomeric protein structure or expression level (initial phenotype) that affects molecular and cellular functions, such as actin and myosin interactions and myofibrillar ATPase activity (Functional phenotype). Functional defects initiate a cascade of events that lead to expression of intermediary phenotypes, such as activation of signalling pathways involved in cardiac hypertrophy and expression of molecular markers of cardiac hypertrophy. The ensuing phenotype is the consequence of interactions between various intermediary molecular pathways with the background in which they operate as well as with the environmental factors. Thus, cardiac hypertrophy and other clinical phenotype are secondary or distant phenotype (Fig. 2). Given the secondary nature of cardiac hypertrophy and the diversity and partial independence of the molecular pathways involved in the pathogenesis of HCM, clinically recognized HCM may comprise multiple different subtypes that all grossly manifest with cardiac hypertrophy. The main implication of this point is in the development of an effective therapy for HCM. One has to either target a pathway that is common to various sub-forms of HCM or intervene at specific pathways that are involved in the pathogenesis of each subset of HCM.

Figure 2.

 Pathogenesis of HCM. Mutations in sarcomeric proteins impair the protein structure and function and provide the initial impetus for expression and activation of the intermediary molecular phenotype. The intermediary molecular phenotype including activation of the hypertrophic signalling molecules induce cardiac hypertrophy and other histological and morphological phenotypes in HCM.

Management and treatment

For an extensive review, the readers are referred to two recent reviews on pharmacological and non-pharmacological therapies including experimental therapies in HCM [88,89].

As HCM is a genetic disease and commonly familial with an autosomal dominant mode of inheritance, it is important to evaluate the genetically related family members for HCM [90,91]. As expected in an autosomal dominant disease, half of the offspring carry the causal mutation and therefore, are at risk of HCM. However, because of age-dependence of penetrance, many mutation carries may not exhibit a phenotype early in life. In addition, many may express the phenotype, but remain asymptomatic and hence, typically are undiagnosed unless screened. Considering the above and because SCD is often the first manifestation of HCM, it is important to evaluate family members of the probands periodically. The typical evaluation of HCM probands and their related family members includes history taking and physical examination, construction of a pedigree, obtaining 12-lead electrocardiogram, Holter monitoring and echocardiography. As noted earlier, TDI of the left ventricle in the familial setting could help identify the mutations carriers prior to development of cardiac hypertrophy [11,12,92]. TDI initially in a transgenic rabbit model of human HCM and subsequently in humans showed that myocardial systolic and diastolic tissue Doppler velocities are reduced in the mutation carriers prior to development of discernible cardiac hypertrophy [11,12,92].

With the advent of deep sequencing or massively parallel DNA sequencing technology, genetic-based diagnosis is increasingly becoming practical. The deep sequencing technology affords the opportunity to sequence the entire genome of an individual or only the target regions of interest. The targeted approach requires enrichment of the genomic regions of interest, such as all exons in the genome (exome) or exons of specific candidate genes prior to deep sequencing. The cost of sequencing per nucleotide is several orders of magnitude cheaper than the Sanger sequencing method, as the output is much larger. Whole genome sequencing is more desirable and yet at the present time, it is costly and time consuming, although the cost is dropping drastically [93]. Likewise, bioinformatics analysis of the huge amount of data generated is complex. Deep sequencing leads to identification of a large number of sequencer variants including a small fraction of sequencing errors. Therefore, for a clinical application, at the present time, it is necessary to validate the sequence variants identified by the deep sequencing technology by the Sanger sequencing technique. The clinical application of deep sequencing technology to genetic screening of HCM patients and families is currently being evaluated.

The clinical utility of genetic screening is best illustrated in identification of the family members or individuals who do not carry the causal mutations and hence, are at an extremely low risk for HCM (the risk is not completely absent because of possible presence of a second mutation and/or technical errors). In familial HCM, genetic screening of the clinically normal family members could lead to early identification of those with the causal mutation. The early genetic identification could lead to frequent evaluation of such individuals for development of the clinical or echocardiographic phenotype. The latter could provide for the opportunity to intervene early to reduce the risk of SCD. This application of genetic testing is relevant whenever the causal gene/mutation is known or one of the known causal genes is responsible for HCM in the family. The causal genes, however, is unknown in about 40% for HCM cases. Genetic screening of the known causal genes will not lead to identification of the causal mutation in approximately 40% of the families or individuals. It is also important to note that not all genetic variants including non-synonymous variants in sarcomeric genes cause HCM and many could be benign variants. Therefore, identification of a non-synonymous variant in a known gene for HCM in an individual alone is not sufficient to establish the causal role. Co-segregation of the variant with the inheritance of HCM in the familial setting, its absence in a large cohort of ethnically matched control individuals and functional studies are often necessary to establish the causality. Thus, clinicians need to train themselves with adequate knowledge for the proper interpretation of the genetic screening.

Effective treatment of HCM requires targeting the specific pathways that are involved in its pathogenesis. Unfortunately, the phenotype-based diagnosis does not delineate the specific pathways that are involved in the pathogenesis of the phenotype. Accordingly, the current pharmacological treatment of human HCM is largely empirical and none has been shown to prevent, attenuate or reverse HCM in humans or impact the prognosis. We advocate the doctrine that HCM, as it is clinically recognized, could result from multiple independent mechanisms and that elucidation of the specific pathways involved in the pathogenesis of each genetic subset is a pre-requisite for an effective therapy. Currently, β-blockers are the mainstay of treatment and the first choice, unless they are contra-indicated. Treatment with β-blockers improves ventricular relaxation, increases diastolic filling time and reduces susceptibility to ventricular and supra-ventricular arrhythmias. Calcium channel blockers that do not have significant vasodilatory effects are also beneficial because of their negative inotropic and chronotropic effects as well as improvement of myocardial diastolic properties. They are used whenever β-blockers are not tolerated or in conjunction with the β-blockers, but not in those with LVOT obstruction. Disopyramide, when used in conjunction with β-blockers, is partially effective in reducing LVOT obstruction and improving symptoms [94]. However, it has some side effects including anti-cholinergic side effects, such as tachycardia, blurred vision and dry mucosa. Overall, the currently used pharmacological agents, while partially effective for symptomatic relief, do not directly target the specific pathways involved in the pathogenesis of HCM. In addition, they are not effective in inducing regression of cardiac hypertrophy or fibrosis in HCM.

Experimental data in animal models of human HCM have shown beneficial effects of 3-hydroxy-3-methyglutaryl-coenzyme A (HMG-CoA) reductase inhibitors (statins), angiotensin II receptor blockade, mineralocorticoid receptor antagonists and anti-oxidant N-acetylcystein (NAC) in prevention of the evolving phenotype and reversal of the established phenotype in HCM [20–24,95]. Clinical utility of these agents in treatment of HCM in humans is subject to investigation and remains to be established [88].


The current clinical diagnosis of HCM is based on detection of cardiac hypertrophy, which is neither specific nor sensitive [26,27,54,55]. The advent of deep sequencing technology affords the opportunity for DNA-based diagnosis, particularly in the early diagnosis and prior to and independent of the clinical phenotype. Physicians and researcher need to garner the knowledge to differentiate the disease-causing variants from those that are either benign or modify phenotypic expression of the phenotype.

Our premise is that HCM is not a single disease entity, as it is currently recognized and treated. Effective treatment of human patients with HCM requires elucidation of molecular pathogenesis of each subset of HCM and targeting the specific pathways involved in the pathogenesis. The alternative approach is to target mechanisms that are common to pathological cardiac hypertrophic growth, such as increased oxidative stress [20–22,96,97]. However, extension of the findings in animal models to humans is expected to be a challenging task and would require large multi-centre efficacy studies.


This study was supported by grants from the NIH/NHLBI, Clinical Scientist Award in Translational Research from the Burroughs Wellcome Fund and The TexGen Fund from the Greater Houston Community Foundation.


Center for Cardiovascular Genetics, The Brown Foundation Institute of Molecular Medicine, The University of Texas Health Science Center, Houston, TX 77030, USA (A. J. Marian).