- Email: [email protected]
Molecular Signature Analysis: Using the Myocardial Transcriptome as a Biomarker in Cardiovascular Disease
powerful technology are gene discovery and molecular signature analysis, and these two applications have different goals, statistical methods, and validation strategies.
Molecular Signature Analysis: Using the Myocardial Transcriptome as a Biomarker in Cardiovascular Disease
Michelle M. Kittleson and Joshua M. Hare*
Statistical Methods
With the emergence of microarray technology, it is now possible to simultaneously assess the expression of tens of thousands of gene transcripts, providing a resolution and precision of phenotypic characterization not previously possible. In the field of cardiomyopathy, microarray studies have largely focused on gene discovery, identifying differentially expressed genes characteristic of diverse disease states, through which novel genetic pathways and potential therapeutic targets may be elucidated. However, gene expression profiling may also be used to identify a pattern of genes (a molecular signature) that serves as a biomarker for clinically relevant parameters. One study thus far does demonstrate that a molecular signature can accurately identify etiology in cardiovascular disease, supporting ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy in heart failure. Microarray research in cardiomyopathy is still in its earliest stages. Nevertheless, the ultimate potential application of transcriptome-based molecular signature analysis is individualization of the management of patients with heart failure, whereby a patient with a newly diagnosed cardiomyopathy could, through molecular signature analysis, be offered an accurate assessment of prognosis and how individualized medical therapy could affect his or her outcome. (Trends Cardiovasc Med 2005;15:130–138) D 2005, Elsevier Inc.
Gene discovery focuses on identifying differentially expressed genes characteristic of different disease states, through which novel genetic pathways and potential therapeutic targets may be elucidated. Many statistical methods have been used to identify differentially expressed genes. All methods, however, rely on the same principles: comparing expression between two or more groups by taking into account the magnitude of the difference between groups and the variability of expression between groups while adjusting for multiple comparisons. The latter point is essential in microarray analyses, where the number of variables (i.e., thousands of genes) is orders of magnitude greater than the number of subjects. Significance analysis of microarrays (SAM) is one approach (Tusher et al. 2001) that has been used in many studies. Significance analysis of microarrays identifies genes with statistically significant changes in expression by identifying a set of gene-specific statistics (similar to the t-test, thus taking into account both magnitude of change and variability of expression) and a corresponding false discovery rate (similar to a P value adjusted for multiple comparisons).
Gene Discovery
Validation Michelle M. Kittleson and Joshua M. Hare are at the Department of Medicine, Cardiology Division, Johns Hopkins University School of Medicine, Baltimore, MD. This research was supported by National Institutes of Health grant 5RO1-HL-065455 (JMH). JMH is a recipient of a Paul Beeson Physician Faculty Scholars in Aging Research Award. MMK is a recipient of the Pearl M. Stetler Research Fund for Women Physicians Fellowship Award. * Address correspondence to: Joshua M. Hare, MD, Johns Hopkins Hospital, Ross 1059, 720 Rutland Avenue, Baltimore, MD, 21287. Tel.: (+1) 410-614-4161; fax: (+1) 443287-7945; e-mail: [email protected]. D 2005, Elsevier Inc. All rights reserved. 1050-1738/05/$-see front matter
130
With the emergence of microarray technology, it is now possible to simultaneously assess the expression of tens of thousands of gene transcripts, providing a resolution and precision of phenotypic characterization not previously possible. Because the state of the transcriptome in a given diseased tissue may contain a highly accurate representation of key biologic phenomena, patterns of gene expression have potential to provide insights not only into disease mechanisms but also to identify markers useful for diagnostic, prognostic, and therapeutic purposes. Thus, the two distinct major applications of this
Once differentially expressed genes are identified, the transcript abundance is routinely confirmed by a complementary method such as quantitative polymerase chain reaction (PCR), Northern blotting, or RNase protection assays (Cook and Rosenzweig 2002), considered technical validation. In cardiomyopathy, studies have relied mainly on quantitative PCR, with more than 80% agreement in all studies with the results of microarray hybridization. Less commonly, levels of the corresponding protein have also been measured, with less agreement between transcript and protein abundance, from TCM Vol. 15, No. 4, 2005
50% in one study (Chen et al. 2003b) to 67% in another (Margulies et al. 2005). This is not surprising, however, because differences in mRNA localization, processing, stability, translation efficiency, as well as posttranslational protein modification and interactions will all affect the measured protein abundance. This raises an important issue in microarray analyses focused on gene discovery. Because cellular processes are mainly mediated by proteins, mRNA changes unaccompanied by corresponding alterations in protein may not be meaningful. Furthermore, even if protein abundance confirms the gene expression levels noted by microarray analysis, a biologic role for these changes has still not been established. However, studies of gene discovery still have significant merit, in offering valuable hypothesisgenerating insight into possible mechanistic pathways that should be further elucidated with studies focused on establishing causality. On the other hand, as opposed to the technical validation described above, biologic validation can also be performed on microarray data with the use of pathways analysis. To date, this has not been used in the cardiomyopathy literature. Pathways analysis, through GenMapp, Ingenuity, and other software applications, identifies the biologically relevant networks that exist among differentially expressed genes. Although the insights obtained are only as accurate as the software’s database of known biologic pathways, pathways analysis is still a powerful means of organizing microarray data and will likely be used in future studies of gene discovery in cardiomyopathy.
Hall et al. 2004). Gene discovery has also provided new insights into rare diseases such as giant cell myocarditis (Kittleson et al. 2005b). Other studies have used more sophisticated techniques, focusing on three-way comparisons: comparing the differential gene expression of ischemic and nonischemic cardiomyopathy relative to nonfailing hearts (Kittleson et al. 2005a, Steenman et al. 2003) or failing and LVAD-supported hearts relative to nonfailing hearts (Margulies et al. 2005). These studies have provided insights into novel genetic pathways and therapeutic targets, and they also serve as the basis for studies involving molecular signature analysis. Failing Versus Nonfailing Hearts Our analysis comparing the gene expression of nonfailing hearts with that of both ischemic and nonischemic cardiomyopathy demonstrated that cardiomyopathies of different etiologies exhibit both shared and distinct changes in the genesis of heart failure. Remarkably, of more than 22,000 transcripts present on the Affymetrix microarray platform, only 288 (1% –2%) genes were differentially expressed in nonischemic and ischemic cardiomyopathy relative to nonfailing hearts, and only 41 of these genes were shared between ischemic and nonischemic cardiomyopathy (Figure 1). Despite the differences in sample size and statistical analyses, there is congruence between studies in the identity of differentially expressed genes and the magnitude of fold change in failing relative to nonfailing hearts (Table 1).
Significantly regulated genes across all studies are mainly those belonging to functional categories of cell growth and maintenance, cytoskeleton/sarcomere, metabolism, and signal transduction, and our previous analysis indicates that the differences in functional categories are not solely a function of their representation on the microarray (Kittleson et al. 2005a). This agreement across studies offers further validation for studies of differential gene expression. Therapeutic insights may be gleaned from the microarray analyses in failing and nonfailing hearts. For example, genes participating in fatty acid metabolism are upregulated in failing hearts, whereas those in glucose metabolism are downregulated (Kittleson et al. 2005a, Tan et al. 2002). Such a finding offers potential insight into drugs such as ranolazine, which shifts myocardial cells from fatty acid to glucose metabolism. This drug, currently being investigated as a treatment for myocardial ischemia (Chaitman et al. 2004), may also benefit patients with heart failure. Impact of Left Ventricular Assist Device Support A number of studies have also examined the changes in gene expression that occur after LVAD support (Blaxall et al. 2003, Chen et al. 2003a, Chen et al. 2003b, Hall et al. 2004). Through mechanical unloading of the failing ventricle, LVAD support results in beneficial hemodynamic, neurohormonal, structural, and biochemical changes, termed reverse remodeling (Margulies 2002). This phe-
Overview of Gene Discovery Studies in Cardiomyopathy In the field of cardiomyopathy, many microarray studies have focused on gene discovery. Some studies involved small sample sizes and binary comparisons such as failing and nonfailing hearts (Barrans et al. 2002, Boheler et al. 2003, Kaab et al. 2004, Kittleson et al. 2005b, Steenman et al. 2005, Tan et al. 2002, Yung et al. 2004), dilated and hypertrophic cardiomyopathy (Hwang et al. 2002), and before and after left ventricular assist device (LVAD) placement (Blaxall et al. 2003, Chen et al. 2003a, 2003b, TCM Vol. 15, No. 4, 2005
Figure 1. Identification of differentially expressed genes between nonfailing (NF) and nonischemic cardiomyopathy (NICM) samples and between NF and ischemic cardiomyopathy (ICM) samples. There were 257 genes differentially expressed between NF and NICM samples and 72 genes differentially expressed between NF and ICM samples. Of these, 41 were common to both NF and NICM and NF and ICM comparisons.
131
Table 1. Differentially expressed genes common to published reports of failing versus nonfailing hearts
Gene Name Apoptosis Pleiomorphic adenoma gene-like 1 TIA1 cytotoxic granule-associated RNA binding protein Catalytic activity ATPase, Na+/K+ transporting, beta 3 polypeptide Cell growth/maintenance Cyclin-dependent kinase inhibitor 1B Delta sleep-inducing peptide, immunoreactor rab6 GTPase activating protein Pleiotrophin Zinc finger protein 145 Cytoplasmic/ribosomal Translocated promoter region Cytoskeleton/sarcomere Collagen, type XXI, alpha 1 Myosin, light polypeptide 4, alkali; atrial, embryonic Myosin, heavy polypeptide 6, cardiac muscle, alpha Ficolin Pleckstrin homology-like domain, family A, member 1 Development Lumican Osteoblast-specific factor 2 Inflammation/immune response H factor 1 (complement) Metabolism Eukaryotic translation initiation factor 1A, Y chromosome F-box only protein 3 Heterogeneous nuclear ribonucleoprotein H3 Ornithine decarboxylase 1 Phospholipase A2, group IIA Signal transduction ATPase, H+ transporting, lysosomal interacting protein 2 Atrial natriuretic factor Natriuretic peptide precursor B Phosphoinositide-3-kinase, regulatory subunit, polypeptide 1 Phosphodiesterase 4B, cAMP-specific SH3 domain binding glutamic acid-rich protein-like Transcription Fragile X mental retardation 1 Nuclear receptor subfamily 3, group C, member 1
Gene symbol PLAGL1 TIA1
Kittleson et al. 2005a NICM-NF
Tan et al. 2002 NICM-NF
2.1 2.1
Barrans et al. 2002 NICM-NF
Yung et al. 2004 NICM-NF
Steenman et al. 2003 NICM-NF
2.2
2.7
CDKN1B
2.5
DSIPI
2.1
GAPCENA PTN ZNF145
2.1 2.2
TPR
2.2
COL21A1 MYL4
2.3
MYH6
3.7
5.3
FCN3 PHLDA1
2.6 5.1
7.7 5.43
LUM OSF-2
2.8 3.0
HF1
2.5
1.23
EIF1AY
2.2
1.78
FBXO3 HNRPH3
2.0 2.7
ODC1 PLA2G2A
2.0
ATP6IP2
2.1
2.3
2.03 1.29 1.74 3.29 2.33 2.02
3.8 12
2.79
2.3 2.4
1.36
2.5
SH3BGRL
3.1
FMR1 NR3C1
3.3 2.1
2.07
3.36
3.2 3.5
1.96
2.2 1.59 1.83 2.52
5.1
2.3
2.1 2.5
3.52 2.01
PDE4B
Steenman et al. 2003 ICM-NF
2.14
ATP1B3
HSCDDANF NPPB PIK3R1 2.3
Kittleson et al. 2005a ICM-NF
3.4 1.19
4.2 3.3
19.15
4.83 7.24 2.73
2.3 4.4 3.1
4.59 7.80 2.84
2.41 1.20
2.06 1.72
Used with permission from Kittleson et al. 2005a. ICM indicates ischemic cardiomyopathy; NF, nonfailing; NICM, nonischemic cardiomyopathy.
132
TCM Vol. 15, No. 4, 2005
notypic alteration, coupled with the availability of tissue samples obtained at the time of implantation, has offered a unique opportunity to study the transcriptomal shifts associated with reverse remodeling. Many of these analyses have demonstrated alterations in genes involved in vascular signaling, including downregulation of neuropilin-1, a vascular endothelial growth factor receptor (Hall et al. 2004), upregulation of endothelial nitric oxide synthase (Chen et al. 2003b), and upregulation of the APJ receptor for apelin, an endogenous cardiac inotrope present in the cardiac vasculature (Chen et al. 2003a). Thus, these unbiased approaches have identified significant alterations in genes that regulate vascular organization and endothelial function in response to mechanical unloading of the failing human heart. The most sophisticated of these analyses compared the gene expression of nonfailing hearts with that of failing and LVAD-supported hearts to identify adaptations that represent normalization of gene expression (Margulies et al. 2005). Of the 3088 transcripts that were differentially expressed in failing relative to nonfailing hearts, only 238 actually dem-
onstrated a consistent response to LVAD support, and of these, more than 75% demonstrated persistence or exacerbation of their heart failure expression pattern after LVAD support. This suggests that the alterations in gene expression after LVAD support are distinct from a return toward normalcy and may not represent a simple reversal of changes observed during disease progression (Margulies et al. 2005). These findings offer unique insights into the nature of LVAD-associated reverse remodeling and could theoretically be clinically useful in identifying individual patient responses to mechanical unloading. Implications for Molecular Signature Identification Although most microarray analyses in cardiomyopathy to date have focused on gene discovery, these studies nevertheless provide insight into the feasibility of molecular signature analysis. In two studies, heart failure of different etiologies demonstrated different patterns of gene expression in unsupervised analyses. An unsupervised analysis of gene expression does not take into
account a priori definitions such as clinical parameters of etiology or disease stage in the division of samples into groups. Rather, a global assessment of gene expression alone is used to determine the relatedness of samples, and the significance of the grouping is then assessed. In one study, the nonischemic cardiomyopathy samples demonstrated more extensive global changes in gene expression than ischemic cardiomyopathy samples after LVAD support (Blaxall et al. 2003). In another study, the overall gene expression of familial and alcoholic cardiomyopathy was distinct from that of idiopathic cardiomyopathy (Tan et al. 2002). Patients with heart failure of different clinical stages also exhibit different patterns of gene expression in unsupervised analyses. In one study of failing and nonfailing hearts, a distinct cluster of patients who were of the highest medical urgency status awaiting cardiac transplantation emerged in an unsupervised analysis (Steenman et al. 2005). In addition, in our study comparing the gene expression of ischemic and nonischemic cardiomyopathy relative to nonfailing hearts, an unsupervised clus-
Figure 2. Insight into the feasibility of molecular signature analysis from unsupervised analyses of gene discovery in failing and nonfailing hearts. Shown is a dendrogram derived from unsupervised hierarchical clustering of genes based on similarity in gene expression and relatedness of samples from nonfailing hearts and those with nonischemic cardiomyopathy (NICM). Each row represents a gene and each column represents a sample. The color in each cell reflects the level of expression of the corresponding gene in the corresponding sample, relative to its mean level of expression in the entire set of samples. Expression levels greater than the mean are shaded in blue, and those below the mean are shaded in red. Circled samples denote the predominant etiology clusters and samples labeled with an arrow fall outside of their appropriate cluster. Samples from patients who required an LVAD before cardiac transplantation (pre-LVAD) and samples from those who did not have an LVAD before transplantation (no LVAD) form distinct clusters, as indicated. Used with permission from Kittleson et al. 2005a.
TCM Vol. 15, No. 4, 2005
133
tering algorithm differentiated patients with cardiomyopathy who required an LVAD before cardiac transplantation from those who did not have an LVAD before transplantation, and these nonLVAD patients actually clustered with the nonfailing hearts (Figure 2). This implies that non-LVAD patients resemble nonfailing patients more than their failing counterparts who require an LVAD before cardiac transplantation. An examination of their baseline characteristics confirmed this: LVAD patients had higher pulmonary capillary wedge pressures and an increased need for intravenous inotropes, two known markers of poor prognosis. These findings suggest that gene expression can be correlated with clinically relevant parameters in patients with heart failure. However, because
these studies focused on gene discovery, the observations could not be applied prospectively to identify and validate a gene expression signature to distinguish subjects based on relevant clinical parameters, thus emphasizing the need for studies focused solely on molecular signature analysis.
Molecular Signature Analysis
Statistical Methods The goal of molecular signature analysis is to identify a pattern of gene expression that is associated with a clinical parameter such as etiology, prognosis, or response to therapy, thus providing diagnostic or prognostic precision not possible with standard clinical information. There are a number of methods that can be used for molecular signature analysis,
including partial least squares regression, neural networks, and shrunken centroids, and all rely on the same basic principles (Carey et al. 2005, Dudoit et al. 2003, Simon et al. 2003). First, samples are divided into groups based on a clinically relevant parameter such as disease etiology, prognosis, or response to therapy. Then a molecular signature is created by choosing genes whose expression is solidly associated with the parameter in question, by weighting genes based on their individual predictive strengths. Prediction analysis of microarrays (PAM) is one approach in molecular signature analysis (Tibshirani et al. 2002). Prediction analysis of microarrays uses the method of nearest shrunken centroids to identify and validate the smallest set of genes whose expression is associated with a predefined class
Figure 3. Depiction of PAM’s method of nearest shrunken centroids. Samples are first divided into classes based on a predefined parameter. In this case, there are four groups from a data set containing samples of four different types of small round blue cell tumors of childhood. The gray bars represent the standardized class centroid, the average expression of each gene in a given class divided by the within-class standard deviation (thus genes with stable expression have a greater contribution to the class centroid). The red bars represent the shrunken centroid, bde-noisedQ versions of centroids after subtraction of a given threshold determined by cross-validation. After shrinkage, a small number of genes remain and these act as prototypes for each class. Independent samples are classified based on their squared distance from each prototypic class shrunken centroid. Reproduced with permission from Tibshirani et al. 2002. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 99:6567–6572. Copyright 2002 National Academy of Sciences, U.S.A.
134
TCM Vol. 15, No. 4, 2005
(Figure 3 is a representative example of this method). In contrast to SAM and other methods of identifying differentially expressed genes, PAM focuses mainly on the stability of gene expression and on the smallest number of genes required to create a molecular signature. Validation Whereas the goal of gene discovery is to identify differentially expressed genes that offer insight into novel genetic pathways or cause-specific therapies, the goal of molecular signature analysis is to identify a pattern of genes that differentiates between clinical entities with a precision not possible based on standard clinical information. Thus, the identity of the mRNA transcripts in the signature or whether they are translated into protein may or may not have immediately discernable bearing on the utility of the pattern. Therefore, the validation strategy is also unique: testing the accuracy of the identified molecular signature in samples distinct from those used to create the signature. Molecular Signature Analysis in Neoplastic Disease In neoplastic disease, molecular signature analysis has proven useful in determining prognosis and response to therapy. In young women with breast cancer, a molecular signature predicted disease outcome better than standard criteria: a poor-prognosis signature was associated with a five-fold risk of distant metastases within 5 years, a difference that would justify early intensive adjuvant chemotherapy (Van de Vijver et al. 2002). Similar results have been obtained for acute leukemia (Bullinger et al. 2004, Holleman et al. 2004, Valk et al. 2004) and lymphoma (Dave et al. 2004, Rosenwald et al. 2002). These advances demonstrate that molecular signature analysis can augment current standard practices to better individualize management in neoplastic disease. It is essential to determine if this powerful technology can also be applied to diseases of the myocardium. Differentiating Heart Failure by Etiology We have identified a gene expression profile that differentiates the two major TCM Vol. 15, No. 4, 2005
forms of cardiomyopathy, ischemic and nonischemic, representing the first application of molecular signature analysis in cardiovascular disease (Kittleson et al. 2004). The analysis was performed on 48 samples: 25 obtained from patients at the time of LVAD placement or cardiac transplantation in patients without LVADs (end-stage samples), 16 from patients after LVAD placement (post-LVAD), and 7 from newly diagnosed patients with cardiomyopathy (from endomyocardial biopsy samples; Figure 4). With the use of PAM, a 16-sample training set was used to develop a prediction rule that was then tested in nine independent samples, including seven from a different institution (Chen et al. 2003b). A gene expression profile offered perfect prediction in these samples with 100% sensitivity and 100% specificity. We then performed this analysis over 210 random combinations of training and test sets, with an overall sensitivity and specificity both of 89%. We also tested the etiology signature in samples of different stages, from post-LVAD and newly diagnosed patients. In both cases, the etiology signature performed perfectly in nonischemic samples (specificity 100%) but only identified one of three ischemic samples correctly (sensi-
tivity 33%). This suggests that the etiology signature developed based on end-stage tissue was specific to disease stage in ischemic cardiomyopathy. Thus, future analyses will need to consider disease stage as a major factor. Age and sex may represent major confounding factors in differential gene expression of failing versus nonfailing hearts (Boheler et al. 2003). To address this issue for molecular signature determination, we stratified our analysis by clinical characteristics (Table 2). The sensitivity and specificity were not affected, indicating that the accuracy of the etiology signature was not an artifact of differences in baseline characteristics. This study represents the first use of molecular signature analysis in cardiovascular disease. Our findings support ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy.
Future Directions
Gene Discovery Versus Molecular Signature Analysis To determine if molecular signature analysis, in addition to gene discovery,
Figure 4. Study design to determine molecular signatures for ischemic and nonischemic cardiomyopathy. The samples were first divided into a training set of end-stage samples from Johns Hopkins Hospital (JHH) used to develop the etiology signature. The signature was then validated by determining its sensitivity and specificity in three test sets: end-stage samples from the University of Minnesota (UM), post-LVAD samples from both JHH and UM, and endomyocardial biopsy samples from newly diagnosed patients at JHH. Then the end-stage samples were randomly partitioned into training and test sets to identify the representative etiology signature and the overall sensitivity and specificity. Adapted with permission from Kittleson et al. 2004, Identification of a gene expression profile that differentiates ischemic and nonischemic cardiomyopathy. Circulation 110(22):3444-3451.
135
Table 2. Sensitivity and specificity of etiology prediction profile in strata defined by clinical covariates Sensitivity (%) Specificity (%) Overall 89 Age (y) z50 88 b50 100 Ejection fraction (%) z15 89 b15 100 Inotropic therapy Yes 100 No 89
89 80 90 89 83 100 60
Reproduced with permission from Kittleson et al. 2004, Identification of a gene expression profile that differentiates ischemic and nonischemic cardiomyopathy. Circulation 110(22):3444-3451.
is feasible and useful in myocardial diseases, it is essential to broaden the focus of gene expression research in cardiology. Currently, the focus is on the identification of differentially expressed genes and validation through confirmation of gene expression levels through quantitative PCR and protein products through Western blotting and immunofluorescence. To make the transition from gene discovery to molecular signature analysis, gene expression research in cardiology must develop a new focus: on determining the predictive power of gene expression. One important distinction in molecular signature analysis is the need first to identify a set of genes whose expression characterizes a predefined group of patients (with a clinically relevant distinction such as etiology, prognosis, or response to therapy) and then to test the predictive accuracy of this profile prospectively in an independent set of patients with varying phenotypes (Cook and Rosenzweig 2002). Another important distinction is that molecular signature analysis is based upon a pattern of gene expression rather than the identity of specific genes (Simon et al. 2003). The prediction algorithm is able to compare an unknown sample and determine how closely it resembles one pattern versus the other; the absolute expression of an individual gene carries relatively small
136
weight compared with the overall signature. Thus, validation cannot solely involve the confirmation of gene expression or gene product levels via a complementary technique such as quantitative PCR or immunofluorescence. Nevertheless, such validation can prove useful to address a different issue: whether the molecular signature offers utility independent of the microarray platform used to create it. This is important if diseasespecific platforms are developed, as in the oncology experience (Lossos et al. 2004, Paik et al. 2004). Sample Size in Microarray Experiments There is limited knowledge on the sample size required in microarray experiments. The largest microarray study in cardiomyopathy to date has involved 199 patient samples (Margulies et al. 2005), and in the oncology literature, each study has employed fewer than 300 patients. However, this should be considered a strength of these analyses: a succession of smaller studies, performed quickly and with the use of improving technology, will outperform larger studies locked into outdated approaches (Liu and Karuturi 2004). The Source of Tissue for Analysis To date, microarray analyses in cardiomyopathy have mainly used discarded myocardial tissue obtained at the time of cardiac transplantation or LVAD placement, and there are limitations to this approach. First, the tissue is obtained from patients late in the disease course and thus the conclusions may not be applicable to patients at an earlier stage of disease; as we demonstrated, a molecular signature based on etiology in end-stage cardiomyopathy is specific to disease stage (Kittleson et al. 2004). Second, explanted tissue is obtained from different areas of the left ventricle, and there is evidence, in mice, that regional differences in gene expression exist in the left ventricle (Mirotsou et al. 2003). Thus, in the future, microarray analyses in cardiomyopathy will ideally focus on endomyocardial biopsy tissue obtained from patients at earlier stages of disease. Although one commonly invoked limitation of gene expression research
in cardiovascular disease is the lack of ready access to human heart tissue samples, endomyocardial biopsies are frequently performed to evaluate newly diagnosed cardiomyopathy at our institution (Ardehali et al. 2004, Felker et al. 2000). In our experience, endomyocardial biopsy is a safe and well-tolerated procedure, with an overall mortality rate of 0.2%, a rate equivalent to that of other routinely performed catheterization procedures (Felker et al. 1999), and we have demonstrated that microarray hybridization from endomyocardial biopsies is feasible (Kittleson et al. 2004). Although endomyocardial biopsy is a safe procedure that could be more widely performed if a valuable prognostic or diagnostic test were developed, venipuncture is clearly more accessible. Therefore, it is essential to also test the utility of molecular signature analysis in peripheral blood leukocyte samples. In the cancer literature, molecular signatures derived from peripheral blood leukocytes offer comparable predictive accuracy with those from solid tumor samples in classifying subjects by cancer type and type of therapy (DePrimo et al. 2003, Twine et al. 2003). A recent study suggests that this may also be feasible in cardiovascular disease, as peripheral blood molecular signatures correlated with biopsy-proven allograft rejection in cardiac transplant recipients (Horwitz et al. 2004).
Summary
The current approach to the treatment of patients with cardiomyopathy lacks individualization. It is widely anticipated that the future management of patients with heart failure will be tailored based on individual assessments of prognosis and response to therapy. Currently, the ability to predict which newly diagnosed patients with cardiomyopathy will improve their ejection fraction and functional status, and which will go on to develop circulatory collapse and require cardiac transplantation, is still not possible. Studies of gene discovery could identify novel therapeutic targets for the treatment of heart failure, and molecular signature analysis could augment current clinical and imaging modalities used to determine prognosis and response to therTCM Vol. 15, No. 4, 2005
apy. Our identification of a molecular signature that differentiates cardiomyopathy by etiology supports ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy. The ultimate potential application of transcriptome-based molecular signature analysis is the individualization of the management of patients with heart failure. In the future, a patient with a newly diagnosed cardiomyopathy could, through molecular signature analysis, be offered an accurate assessment of prognosis and how individualized medical therapy could affect his or her outcome.
References Ardehali H, Qasim A, Cappola T, et al.: 2004. Endomyocardial biopsy plays a role in diagnosing patients with unexplained cardiomyopathy. Am Heart J 147:919–923. Barrans JD, Allen PD, Stamatiou D, et al.: 2002. Global gene expression profiling of end-stage dilated cardiomyopathy using a human cardiovascular-based cDNA microarray. Am J Pathol 160:2035–2043. Blaxall BC, Tschannen-Moran BM, Milano CA, Koch WJ: 2003. Differential gene expression and genomic patient stratification following left ventricular assist device support. J Am Coll Cardiol 41: 1096–1106.
Cook SA, Rosenzweig A: 2002. DNA microarrays: implications for cardiovascular medicine. Circ Res 91:559–564. Dave SS, Wright G, Tan B, et al.: 2004. Prediction of survival in follicular lymphoma based on molecular features of tumor-infiltrating immune cells. N Engl J Med 351:2159–2169. DePrimo SE, Wong LM, Khatry DB, et al.: 2003. Expression profiling of blood samples from an SU5416 phase III metastatic colorectal cancer clinical trial: a novel strategy for biomarker identification. BMC Cancer 3:3. Dudoit S, Gentleman RC, Quackenbush J: 2003. Open source software for the analysis of microarray data. Biotechniques (Suppl):45–51. Felker GM, Hu W, Hare JM, et al.: 1999. The spectrum of dilated cardiomyopathy. The Johns Hopkins experience with 1,278 patients. Medicine (Baltimore) 78:270–283. Felker GM, Thompson RE, Hare JM, et al.: 2000. Underlying causes and long-term survival in patients with initially unexplained cardiomyopathy. N Engl J Med 342:1077–1084. Hall JL, Grindle S, Han X, et al.: 2004. Genomic profiling of the human heart before and after mechanical support with a ventricular assist device reveals alterations in vascular signaling networks. Physiol Genomics 17:283–291. Holleman A, Cheok MH, Den Boer ML, et al.: 2004. Gene-expression patterns in drugresistant acute lymphoblastic leukemia cells and response to treatment. N Engl J Med 351:533–542.
Kittleson MM, Minhas KM, Irizarry RA, et al.: 2005b. Gene expression in giant cell myocarditis: altered expression of immune response genes. Int J Cardiol (in press). Liu ET, Karuturi KR: 2004. Microarrays and clinical investigations. N Engl J Med 350:1595–1597. Lossos IS, Czerwinski DK, Alizadeh AA, et al.: 2004. Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 350:1828–1837. Margulies KB: 2002. Reversal mechanisms of left ventricular remodeling: lessons from left ventricular assist device experiments. J Card Fail 8:S500–S505. Margulies KB, Matiwala S, Cornejo C, et al.: 2005. Mixed messages. Transcription patterns in failing and recovering human myocardium. Circ Res 1–8. Mirotsou M, Watanabe CMH, Schultz PG, et al.: 2003. Elucidating the molecular mechanism of cardiac remodeling using a comparative genomic approach. Physiol Genomics 15:115–126. Paik S, Shak S, Tang G, et al.: 2004. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817–2826. Rosenwald A, Wright G, Chan WC, et al.: 2002. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 346:1937–1947. Simon R, Radmacher MD, Dobbin K, McShane LM: 2003. Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 95:14–18.
Boheler KR, Volkova M, Morrell C., et al.: 2003. Sex- and age-dependent human transcriptome variability: implications for chronic heart failure. Proc Natl Acad Sci USA 100:2754–2759.
Horwitz PA, Tsai EJ, Putt ME, et al.: 2004. Detection of cardiac allograft rejection and response to immunosuppressive therapy with peripheral blood gene expression. Circulation 110:3815–3821.
Bullinger L, Dohner K, Bair E, et al.: 2004. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 350: 1605–1616.
Hwang JJ, Allen PD, Tseng GC, et al.: 2002. Microarray gene expression profiles in dilated and hypertrophic cardiomyopathic end-stage heart failure. Physiol Genomics. 10:31–44.
Steenman M, Lamirault G, Le Meur N, et al.: 2005. Distinct molecular portraits of human failing hearts identified by dedicated cDNA microarrays. Eur J Heart Fail 7:157–165.
Carey VJ, Gentry J, Whalen E, Gentleman R: 2005. Network structures and algorithms in Bioconductor. Bioinformatics 21:135–136.
Kaab S, Barth AS, Margerie D, et al.: 2004. Global gene expression in human myocardium–oligonucleotide microarray analysis of regional diversity and transcriptional regulation in heart failure. J Mol Med 82:308–316.
Tan FL, Moravec CS, Li J, et al.: 2002. The gene expression fingerprint of human heart failure. Proc Natl Acad Sci USA 99: 11387–11392.
Chaitman BR, Pepine CJ, Parker JO, et al.: 2004. Effects of ranolazine with atenolol, amlodipine, or diltiazem on exercise tolerance and angina frequency in patients with severe chronic angina: A randomized controlled trial. JAMA 291:309–316. Chen MM, Ashley EA, Deng DX, et al.: 2003a. Novel role for the potent endogenous inotrope apelin in human cardiac dysfunction. Circulation 108:1432–1439. Chen Y, Park S, Li Y, et al.: 2003b. Alterations of gene expression in failing myocardium following left ventricular assist device support. Physiol Genomics 14:251–260
TCM Vol. 15, No. 4, 2005
Kittleson MM, Ye SQ, Irizarry RA, et al.: 2004. Identification of a gene expression profile that differentiates between ischemic and nonischemic cardiomyopathy. Circulation 110:3444–3451. Kittleson MM, Minhas KM, Irizarry RA, et al.: 2005a. Gene expression analysis of ischemic and nonischemic cardiomyopathy: shared and distinct genes in the development of heart failure. Physiol Genomics 21: 299–307.
Steenman M, Chen YW, Le Cunff M, et al.: 2003. Transcriptomal analysis of failing and nonfailing human hearts. Physiol Genomics 12:97–112.
Tibshirani R, Hastie T, Narasimhan B, Chu G: 2002. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 99:6567–6572. Tusher VG, Tibshirani R, Chu G: 2001. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 98:5116–5121. Twine NC, Stover JA, Marshall B, et al.: 2003. Disease-associated expression profiles in peripheral blood mononuclear cells from patients with advanced renal cell carcinoma. Cancer Res 63:6069–6075.
137
Valk PJ, Verhaak RG, Beijen MA, et al.: 2004. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 350:1617–1628. Van De Vijver MJ, He YD, Van’t Veer LJ, et al.: 2002. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 347:1999–2009. Yung CK, Halperin VL, Tomaselli GF, Winslow RL: 2004. Gene expression profiles in end-stage human idiopathic dilated cardiomyopathy: altered expres-
Hsp20 and Its Cardioprotection Guo-Chang Fan, Guoxiang Chu, and Evangelia G. Kranias*
The small heat shock protein Hsp20, also referred to as P20/HspB6, is expressed in the brain, stomach, liver, lung, kidney, blood, smooth muscle, skeletal muscle, and cardiac tissue. Although Hsp20 is not heatinducible, several cellular signaling pathways appear to regulate its biologic functions. In recent years, tremendous advances have been made in elucidating the significance of Hsp20 in smooth muscle and its potential benefits on coronary vasculature. Of interest, recent findings have demonstrated that sustained b-adrenergic stimulation results in expression and phosphorylation of cardiac Hsp20. Moreover, Hsp20 overexpression enhances cardiac function and renders cardioprotection against b-agonist-mediated apoptosis and ischemia/reperfusion injury ex vivo and in vivo. This article reviews the new findings on translocation of Hsp20 in response to various stimuli and the multiple cellular targets of Hsp20, with special emphasis on its protective effects in the heart. (Trends Cardiovasc Med 2005;15:138–141) D 2005, Elsevier Inc. In mammalian species, the subfamily of small heat shock proteins (sHsps, 15-30 kDa) comprises of 10 known members (Kappe et al. 2003), namely, Hsp27 (Hsp25 of rodents, HspB1), myotonic dystrophy protein kinase binding protein (MKBP, HspB2), HspB3, aA-crystallin (HspB4), aB-crystallin (HspB5), Hsp20 (P20, HspB6), cvHsp (HspB7), Hsp22 (H11, HspB8), HspB9, and HspB10. According to their different patterns of
Guo-Chang Fan, Guoxiang Chu, and Evangelia G. Kramias are at the Department of Pharmacology and Cell Biophysics, University of Cincinnati College of Medicine, Cincinnati, Ohio. * Address correspondence to: Evangelia G. Kranias, PhD, Department of Pharmacology and Cell Biophysics, University of Cincinnati College of Medicine, 231 Albert Sabin Way, Cincinnati, OH 45267-0575, USA. Tel.: (+1) 513-558-2377; fax: (+1) 513-558-2269; e-mail: [email protected]. D 2005, Elsevier Inc. All rights reserved. 1050-1738/05/$-see front matter
138
gene expression and subcellular localization, Taylor and Benjamin (2005) classified these sHsps into two major categories: classes I and II. Class I sHsps (HspB1, HspB5, HspB6, and HspB8) are ubiquitously expressed, whereas class II members (HspB2, HspB3, HspB4, HspB7, HspB9, and HspB10) display tissue-restricted patterns of expression. Hsp20, also referred to as P20/HspB6, was originally copurified with the sHsps Hsp27/Hsp28 from rat and human skeletal muscles (Kato et al. 1994). Its genome contains 3 exons and 2 introns. Human, rat, and mouse Hsp20 are composed of 157, 162, and 162 amino acids, respectively. Although hyperthermia did not induce Hsp20 in rat tissue (Kato et al. 1994), heat pretreatment of swine carotid artery was associated with increased Hsp20 levels (O’Connor and Rembold 2002). This protein is detected in all tissues by a sandwich-type immunoassay system, reaching a maximal level of 1.3% of total protein in skeletal,
sion of apoptotic and cytoskeletal genes. Genomics 83:281–297.
PII S1050-1738(05)00061-7
TCM
heart, and smooth muscles (Kato et al. 1994), and its expression levels are altered during development (Inaguma et al. 1996, Verschuure et al. 2003). Recent studies have revealed the significance of Hsp20 in relaxation of vascular muscle (Beall et al. 1997, Rembold et al. 2000, 2001, Brophy et al. 2002, Flynn et al. 2003, Woodrum et al. 1999, Meeks et al. 2005) and inhibition of platelet aggregation (Kozawa et al. 2002). Of particular interest, our recent work (Chu et al. 2004) discovered a cardiac isoform of Hsp20 in mouse heart, which was inducible by isoproterenol stimulation, and comparison of its derived amino acid sequence with rat and human skeletal Hsp20 revealed 95% and 87% homology, respectively. Furthermore, cardiac Hsp20 overexpression renders protection against stressinduced injury in vitro and in vivo, through its antiapoptotic properties (Fan et al. 2004, 2005). Therefore, it is plausible that Hsp20 provides multifaceted beneficial effects in the heart. This brief review will highlight the translocation of Hsp20 in response to various stresses, interaction with its targets, cardioprotective effects, and other physiologic functions.
Translocation of Hsp20 in Response to Stress
Hsp20 is normally located in the cytoplasm of cardiac myocytes. After a heat stress, a subpopulation of Hsp20 migrates into the nucleus, whereas the other part remains in the cytoplasm. In very few cells, a faint sarcomeric association of Hsp20 is observed (Van de Klundert and de Jong 1999); however, Hsp20 is prominently translocated to the myofibrils in adult rat heart and skeletal muscle under ischemic conditions (Golenhofen et al. 2004). A similar phenomenon with Hsp20 redistribution to stress fibers has also been observed in the rat cardiac myoblast cell line H9C2 after proteasomal inhibition (Verschuure et al. 2002). Conversely, one study demonstrated that Hsp20 was predominantly in transverse bands, TCM Vol. 15, No. 4, 2005
Molecular Signature Analysis: Using the Myocardial Transcriptome as a Biomarker in Cardiovascular Disease
Michelle M. Kittleson and Joshua M. Hare*
Statistical Methods
With the emergence of microarray technology, it is now possible to simultaneously assess the expression of tens of thousands of gene transcripts, providing a resolution and precision of phenotypic characterization not previously possible. In the field of cardiomyopathy, microarray studies have largely focused on gene discovery, identifying differentially expressed genes characteristic of diverse disease states, through which novel genetic pathways and potential therapeutic targets may be elucidated. However, gene expression profiling may also be used to identify a pattern of genes (a molecular signature) that serves as a biomarker for clinically relevant parameters. One study thus far does demonstrate that a molecular signature can accurately identify etiology in cardiovascular disease, supporting ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy in heart failure. Microarray research in cardiomyopathy is still in its earliest stages. Nevertheless, the ultimate potential application of transcriptome-based molecular signature analysis is individualization of the management of patients with heart failure, whereby a patient with a newly diagnosed cardiomyopathy could, through molecular signature analysis, be offered an accurate assessment of prognosis and how individualized medical therapy could affect his or her outcome. (Trends Cardiovasc Med 2005;15:130–138) D 2005, Elsevier Inc.
Gene discovery focuses on identifying differentially expressed genes characteristic of different disease states, through which novel genetic pathways and potential therapeutic targets may be elucidated. Many statistical methods have been used to identify differentially expressed genes. All methods, however, rely on the same principles: comparing expression between two or more groups by taking into account the magnitude of the difference between groups and the variability of expression between groups while adjusting for multiple comparisons. The latter point is essential in microarray analyses, where the number of variables (i.e., thousands of genes) is orders of magnitude greater than the number of subjects. Significance analysis of microarrays (SAM) is one approach (Tusher et al. 2001) that has been used in many studies. Significance analysis of microarrays identifies genes with statistically significant changes in expression by identifying a set of gene-specific statistics (similar to the t-test, thus taking into account both magnitude of change and variability of expression) and a corresponding false discovery rate (similar to a P value adjusted for multiple comparisons).
Gene Discovery
Validation Michelle M. Kittleson and Joshua M. Hare are at the Department of Medicine, Cardiology Division, Johns Hopkins University School of Medicine, Baltimore, MD. This research was supported by National Institutes of Health grant 5RO1-HL-065455 (JMH). JMH is a recipient of a Paul Beeson Physician Faculty Scholars in Aging Research Award. MMK is a recipient of the Pearl M. Stetler Research Fund for Women Physicians Fellowship Award. * Address correspondence to: Joshua M. Hare, MD, Johns Hopkins Hospital, Ross 1059, 720 Rutland Avenue, Baltimore, MD, 21287. Tel.: (+1) 410-614-4161; fax: (+1) 443287-7945; e-mail: [email protected]. D 2005, Elsevier Inc. All rights reserved. 1050-1738/05/$-see front matter
130
With the emergence of microarray technology, it is now possible to simultaneously assess the expression of tens of thousands of gene transcripts, providing a resolution and precision of phenotypic characterization not previously possible. Because the state of the transcriptome in a given diseased tissue may contain a highly accurate representation of key biologic phenomena, patterns of gene expression have potential to provide insights not only into disease mechanisms but also to identify markers useful for diagnostic, prognostic, and therapeutic purposes. Thus, the two distinct major applications of this
Once differentially expressed genes are identified, the transcript abundance is routinely confirmed by a complementary method such as quantitative polymerase chain reaction (PCR), Northern blotting, or RNase protection assays (Cook and Rosenzweig 2002), considered technical validation. In cardiomyopathy, studies have relied mainly on quantitative PCR, with more than 80% agreement in all studies with the results of microarray hybridization. Less commonly, levels of the corresponding protein have also been measured, with less agreement between transcript and protein abundance, from TCM Vol. 15, No. 4, 2005
50% in one study (Chen et al. 2003b) to 67% in another (Margulies et al. 2005). This is not surprising, however, because differences in mRNA localization, processing, stability, translation efficiency, as well as posttranslational protein modification and interactions will all affect the measured protein abundance. This raises an important issue in microarray analyses focused on gene discovery. Because cellular processes are mainly mediated by proteins, mRNA changes unaccompanied by corresponding alterations in protein may not be meaningful. Furthermore, even if protein abundance confirms the gene expression levels noted by microarray analysis, a biologic role for these changes has still not been established. However, studies of gene discovery still have significant merit, in offering valuable hypothesisgenerating insight into possible mechanistic pathways that should be further elucidated with studies focused on establishing causality. On the other hand, as opposed to the technical validation described above, biologic validation can also be performed on microarray data with the use of pathways analysis. To date, this has not been used in the cardiomyopathy literature. Pathways analysis, through GenMapp, Ingenuity, and other software applications, identifies the biologically relevant networks that exist among differentially expressed genes. Although the insights obtained are only as accurate as the software’s database of known biologic pathways, pathways analysis is still a powerful means of organizing microarray data and will likely be used in future studies of gene discovery in cardiomyopathy.
Hall et al. 2004). Gene discovery has also provided new insights into rare diseases such as giant cell myocarditis (Kittleson et al. 2005b). Other studies have used more sophisticated techniques, focusing on three-way comparisons: comparing the differential gene expression of ischemic and nonischemic cardiomyopathy relative to nonfailing hearts (Kittleson et al. 2005a, Steenman et al. 2003) or failing and LVAD-supported hearts relative to nonfailing hearts (Margulies et al. 2005). These studies have provided insights into novel genetic pathways and therapeutic targets, and they also serve as the basis for studies involving molecular signature analysis. Failing Versus Nonfailing Hearts Our analysis comparing the gene expression of nonfailing hearts with that of both ischemic and nonischemic cardiomyopathy demonstrated that cardiomyopathies of different etiologies exhibit both shared and distinct changes in the genesis of heart failure. Remarkably, of more than 22,000 transcripts present on the Affymetrix microarray platform, only 288 (1% –2%) genes were differentially expressed in nonischemic and ischemic cardiomyopathy relative to nonfailing hearts, and only 41 of these genes were shared between ischemic and nonischemic cardiomyopathy (Figure 1). Despite the differences in sample size and statistical analyses, there is congruence between studies in the identity of differentially expressed genes and the magnitude of fold change in failing relative to nonfailing hearts (Table 1).
Significantly regulated genes across all studies are mainly those belonging to functional categories of cell growth and maintenance, cytoskeleton/sarcomere, metabolism, and signal transduction, and our previous analysis indicates that the differences in functional categories are not solely a function of their representation on the microarray (Kittleson et al. 2005a). This agreement across studies offers further validation for studies of differential gene expression. Therapeutic insights may be gleaned from the microarray analyses in failing and nonfailing hearts. For example, genes participating in fatty acid metabolism are upregulated in failing hearts, whereas those in glucose metabolism are downregulated (Kittleson et al. 2005a, Tan et al. 2002). Such a finding offers potential insight into drugs such as ranolazine, which shifts myocardial cells from fatty acid to glucose metabolism. This drug, currently being investigated as a treatment for myocardial ischemia (Chaitman et al. 2004), may also benefit patients with heart failure. Impact of Left Ventricular Assist Device Support A number of studies have also examined the changes in gene expression that occur after LVAD support (Blaxall et al. 2003, Chen et al. 2003a, Chen et al. 2003b, Hall et al. 2004). Through mechanical unloading of the failing ventricle, LVAD support results in beneficial hemodynamic, neurohormonal, structural, and biochemical changes, termed reverse remodeling (Margulies 2002). This phe-
Overview of Gene Discovery Studies in Cardiomyopathy In the field of cardiomyopathy, many microarray studies have focused on gene discovery. Some studies involved small sample sizes and binary comparisons such as failing and nonfailing hearts (Barrans et al. 2002, Boheler et al. 2003, Kaab et al. 2004, Kittleson et al. 2005b, Steenman et al. 2005, Tan et al. 2002, Yung et al. 2004), dilated and hypertrophic cardiomyopathy (Hwang et al. 2002), and before and after left ventricular assist device (LVAD) placement (Blaxall et al. 2003, Chen et al. 2003a, 2003b, TCM Vol. 15, No. 4, 2005
Figure 1. Identification of differentially expressed genes between nonfailing (NF) and nonischemic cardiomyopathy (NICM) samples and between NF and ischemic cardiomyopathy (ICM) samples. There were 257 genes differentially expressed between NF and NICM samples and 72 genes differentially expressed between NF and ICM samples. Of these, 41 were common to both NF and NICM and NF and ICM comparisons.
131
Table 1. Differentially expressed genes common to published reports of failing versus nonfailing hearts
Gene Name Apoptosis Pleiomorphic adenoma gene-like 1 TIA1 cytotoxic granule-associated RNA binding protein Catalytic activity ATPase, Na+/K+ transporting, beta 3 polypeptide Cell growth/maintenance Cyclin-dependent kinase inhibitor 1B Delta sleep-inducing peptide, immunoreactor rab6 GTPase activating protein Pleiotrophin Zinc finger protein 145 Cytoplasmic/ribosomal Translocated promoter region Cytoskeleton/sarcomere Collagen, type XXI, alpha 1 Myosin, light polypeptide 4, alkali; atrial, embryonic Myosin, heavy polypeptide 6, cardiac muscle, alpha Ficolin Pleckstrin homology-like domain, family A, member 1 Development Lumican Osteoblast-specific factor 2 Inflammation/immune response H factor 1 (complement) Metabolism Eukaryotic translation initiation factor 1A, Y chromosome F-box only protein 3 Heterogeneous nuclear ribonucleoprotein H3 Ornithine decarboxylase 1 Phospholipase A2, group IIA Signal transduction ATPase, H+ transporting, lysosomal interacting protein 2 Atrial natriuretic factor Natriuretic peptide precursor B Phosphoinositide-3-kinase, regulatory subunit, polypeptide 1 Phosphodiesterase 4B, cAMP-specific SH3 domain binding glutamic acid-rich protein-like Transcription Fragile X mental retardation 1 Nuclear receptor subfamily 3, group C, member 1
Gene symbol PLAGL1 TIA1
Kittleson et al. 2005a NICM-NF
Tan et al. 2002 NICM-NF
2.1 2.1
Barrans et al. 2002 NICM-NF
Yung et al. 2004 NICM-NF
Steenman et al. 2003 NICM-NF
2.2
2.7
CDKN1B
2.5
DSIPI
2.1
GAPCENA PTN ZNF145
2.1 2.2
TPR
2.2
COL21A1 MYL4
2.3
MYH6
3.7
5.3
FCN3 PHLDA1
2.6 5.1
7.7 5.43
LUM OSF-2
2.8 3.0
HF1
2.5
1.23
EIF1AY
2.2
1.78
FBXO3 HNRPH3
2.0 2.7
ODC1 PLA2G2A
2.0
ATP6IP2
2.1
2.3
2.03 1.29 1.74 3.29 2.33 2.02
3.8 12
2.79
2.3 2.4
1.36
2.5
SH3BGRL
3.1
FMR1 NR3C1
3.3 2.1
2.07
3.36
3.2 3.5
1.96
2.2 1.59 1.83 2.52
5.1
2.3
2.1 2.5
3.52 2.01
PDE4B
Steenman et al. 2003 ICM-NF
2.14
ATP1B3
HSCDDANF NPPB PIK3R1 2.3
Kittleson et al. 2005a ICM-NF
3.4 1.19
4.2 3.3
19.15
4.83 7.24 2.73
2.3 4.4 3.1
4.59 7.80 2.84
2.41 1.20
2.06 1.72
Used with permission from Kittleson et al. 2005a. ICM indicates ischemic cardiomyopathy; NF, nonfailing; NICM, nonischemic cardiomyopathy.
132
TCM Vol. 15, No. 4, 2005
notypic alteration, coupled with the availability of tissue samples obtained at the time of implantation, has offered a unique opportunity to study the transcriptomal shifts associated with reverse remodeling. Many of these analyses have demonstrated alterations in genes involved in vascular signaling, including downregulation of neuropilin-1, a vascular endothelial growth factor receptor (Hall et al. 2004), upregulation of endothelial nitric oxide synthase (Chen et al. 2003b), and upregulation of the APJ receptor for apelin, an endogenous cardiac inotrope present in the cardiac vasculature (Chen et al. 2003a). Thus, these unbiased approaches have identified significant alterations in genes that regulate vascular organization and endothelial function in response to mechanical unloading of the failing human heart. The most sophisticated of these analyses compared the gene expression of nonfailing hearts with that of failing and LVAD-supported hearts to identify adaptations that represent normalization of gene expression (Margulies et al. 2005). Of the 3088 transcripts that were differentially expressed in failing relative to nonfailing hearts, only 238 actually dem-
onstrated a consistent response to LVAD support, and of these, more than 75% demonstrated persistence or exacerbation of their heart failure expression pattern after LVAD support. This suggests that the alterations in gene expression after LVAD support are distinct from a return toward normalcy and may not represent a simple reversal of changes observed during disease progression (Margulies et al. 2005). These findings offer unique insights into the nature of LVAD-associated reverse remodeling and could theoretically be clinically useful in identifying individual patient responses to mechanical unloading. Implications for Molecular Signature Identification Although most microarray analyses in cardiomyopathy to date have focused on gene discovery, these studies nevertheless provide insight into the feasibility of molecular signature analysis. In two studies, heart failure of different etiologies demonstrated different patterns of gene expression in unsupervised analyses. An unsupervised analysis of gene expression does not take into
account a priori definitions such as clinical parameters of etiology or disease stage in the division of samples into groups. Rather, a global assessment of gene expression alone is used to determine the relatedness of samples, and the significance of the grouping is then assessed. In one study, the nonischemic cardiomyopathy samples demonstrated more extensive global changes in gene expression than ischemic cardiomyopathy samples after LVAD support (Blaxall et al. 2003). In another study, the overall gene expression of familial and alcoholic cardiomyopathy was distinct from that of idiopathic cardiomyopathy (Tan et al. 2002). Patients with heart failure of different clinical stages also exhibit different patterns of gene expression in unsupervised analyses. In one study of failing and nonfailing hearts, a distinct cluster of patients who were of the highest medical urgency status awaiting cardiac transplantation emerged in an unsupervised analysis (Steenman et al. 2005). In addition, in our study comparing the gene expression of ischemic and nonischemic cardiomyopathy relative to nonfailing hearts, an unsupervised clus-
Figure 2. Insight into the feasibility of molecular signature analysis from unsupervised analyses of gene discovery in failing and nonfailing hearts. Shown is a dendrogram derived from unsupervised hierarchical clustering of genes based on similarity in gene expression and relatedness of samples from nonfailing hearts and those with nonischemic cardiomyopathy (NICM). Each row represents a gene and each column represents a sample. The color in each cell reflects the level of expression of the corresponding gene in the corresponding sample, relative to its mean level of expression in the entire set of samples. Expression levels greater than the mean are shaded in blue, and those below the mean are shaded in red. Circled samples denote the predominant etiology clusters and samples labeled with an arrow fall outside of their appropriate cluster. Samples from patients who required an LVAD before cardiac transplantation (pre-LVAD) and samples from those who did not have an LVAD before transplantation (no LVAD) form distinct clusters, as indicated. Used with permission from Kittleson et al. 2005a.
TCM Vol. 15, No. 4, 2005
133
tering algorithm differentiated patients with cardiomyopathy who required an LVAD before cardiac transplantation from those who did not have an LVAD before transplantation, and these nonLVAD patients actually clustered with the nonfailing hearts (Figure 2). This implies that non-LVAD patients resemble nonfailing patients more than their failing counterparts who require an LVAD before cardiac transplantation. An examination of their baseline characteristics confirmed this: LVAD patients had higher pulmonary capillary wedge pressures and an increased need for intravenous inotropes, two known markers of poor prognosis. These findings suggest that gene expression can be correlated with clinically relevant parameters in patients with heart failure. However, because
these studies focused on gene discovery, the observations could not be applied prospectively to identify and validate a gene expression signature to distinguish subjects based on relevant clinical parameters, thus emphasizing the need for studies focused solely on molecular signature analysis.
Molecular Signature Analysis
Statistical Methods The goal of molecular signature analysis is to identify a pattern of gene expression that is associated with a clinical parameter such as etiology, prognosis, or response to therapy, thus providing diagnostic or prognostic precision not possible with standard clinical information. There are a number of methods that can be used for molecular signature analysis,
including partial least squares regression, neural networks, and shrunken centroids, and all rely on the same basic principles (Carey et al. 2005, Dudoit et al. 2003, Simon et al. 2003). First, samples are divided into groups based on a clinically relevant parameter such as disease etiology, prognosis, or response to therapy. Then a molecular signature is created by choosing genes whose expression is solidly associated with the parameter in question, by weighting genes based on their individual predictive strengths. Prediction analysis of microarrays (PAM) is one approach in molecular signature analysis (Tibshirani et al. 2002). Prediction analysis of microarrays uses the method of nearest shrunken centroids to identify and validate the smallest set of genes whose expression is associated with a predefined class
Figure 3. Depiction of PAM’s method of nearest shrunken centroids. Samples are first divided into classes based on a predefined parameter. In this case, there are four groups from a data set containing samples of four different types of small round blue cell tumors of childhood. The gray bars represent the standardized class centroid, the average expression of each gene in a given class divided by the within-class standard deviation (thus genes with stable expression have a greater contribution to the class centroid). The red bars represent the shrunken centroid, bde-noisedQ versions of centroids after subtraction of a given threshold determined by cross-validation. After shrinkage, a small number of genes remain and these act as prototypes for each class. Independent samples are classified based on their squared distance from each prototypic class shrunken centroid. Reproduced with permission from Tibshirani et al. 2002. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 99:6567–6572. Copyright 2002 National Academy of Sciences, U.S.A.
134
TCM Vol. 15, No. 4, 2005
(Figure 3 is a representative example of this method). In contrast to SAM and other methods of identifying differentially expressed genes, PAM focuses mainly on the stability of gene expression and on the smallest number of genes required to create a molecular signature. Validation Whereas the goal of gene discovery is to identify differentially expressed genes that offer insight into novel genetic pathways or cause-specific therapies, the goal of molecular signature analysis is to identify a pattern of genes that differentiates between clinical entities with a precision not possible based on standard clinical information. Thus, the identity of the mRNA transcripts in the signature or whether they are translated into protein may or may not have immediately discernable bearing on the utility of the pattern. Therefore, the validation strategy is also unique: testing the accuracy of the identified molecular signature in samples distinct from those used to create the signature. Molecular Signature Analysis in Neoplastic Disease In neoplastic disease, molecular signature analysis has proven useful in determining prognosis and response to therapy. In young women with breast cancer, a molecular signature predicted disease outcome better than standard criteria: a poor-prognosis signature was associated with a five-fold risk of distant metastases within 5 years, a difference that would justify early intensive adjuvant chemotherapy (Van de Vijver et al. 2002). Similar results have been obtained for acute leukemia (Bullinger et al. 2004, Holleman et al. 2004, Valk et al. 2004) and lymphoma (Dave et al. 2004, Rosenwald et al. 2002). These advances demonstrate that molecular signature analysis can augment current standard practices to better individualize management in neoplastic disease. It is essential to determine if this powerful technology can also be applied to diseases of the myocardium. Differentiating Heart Failure by Etiology We have identified a gene expression profile that differentiates the two major TCM Vol. 15, No. 4, 2005
forms of cardiomyopathy, ischemic and nonischemic, representing the first application of molecular signature analysis in cardiovascular disease (Kittleson et al. 2004). The analysis was performed on 48 samples: 25 obtained from patients at the time of LVAD placement or cardiac transplantation in patients without LVADs (end-stage samples), 16 from patients after LVAD placement (post-LVAD), and 7 from newly diagnosed patients with cardiomyopathy (from endomyocardial biopsy samples; Figure 4). With the use of PAM, a 16-sample training set was used to develop a prediction rule that was then tested in nine independent samples, including seven from a different institution (Chen et al. 2003b). A gene expression profile offered perfect prediction in these samples with 100% sensitivity and 100% specificity. We then performed this analysis over 210 random combinations of training and test sets, with an overall sensitivity and specificity both of 89%. We also tested the etiology signature in samples of different stages, from post-LVAD and newly diagnosed patients. In both cases, the etiology signature performed perfectly in nonischemic samples (specificity 100%) but only identified one of three ischemic samples correctly (sensi-
tivity 33%). This suggests that the etiology signature developed based on end-stage tissue was specific to disease stage in ischemic cardiomyopathy. Thus, future analyses will need to consider disease stage as a major factor. Age and sex may represent major confounding factors in differential gene expression of failing versus nonfailing hearts (Boheler et al. 2003). To address this issue for molecular signature determination, we stratified our analysis by clinical characteristics (Table 2). The sensitivity and specificity were not affected, indicating that the accuracy of the etiology signature was not an artifact of differences in baseline characteristics. This study represents the first use of molecular signature analysis in cardiovascular disease. Our findings support ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy.
Future Directions
Gene Discovery Versus Molecular Signature Analysis To determine if molecular signature analysis, in addition to gene discovery,
Figure 4. Study design to determine molecular signatures for ischemic and nonischemic cardiomyopathy. The samples were first divided into a training set of end-stage samples from Johns Hopkins Hospital (JHH) used to develop the etiology signature. The signature was then validated by determining its sensitivity and specificity in three test sets: end-stage samples from the University of Minnesota (UM), post-LVAD samples from both JHH and UM, and endomyocardial biopsy samples from newly diagnosed patients at JHH. Then the end-stage samples were randomly partitioned into training and test sets to identify the representative etiology signature and the overall sensitivity and specificity. Adapted with permission from Kittleson et al. 2004, Identification of a gene expression profile that differentiates ischemic and nonischemic cardiomyopathy. Circulation 110(22):3444-3451.
135
Table 2. Sensitivity and specificity of etiology prediction profile in strata defined by clinical covariates Sensitivity (%) Specificity (%) Overall 89 Age (y) z50 88 b50 100 Ejection fraction (%) z15 89 b15 100 Inotropic therapy Yes 100 No 89
89 80 90 89 83 100 60
Reproduced with permission from Kittleson et al. 2004, Identification of a gene expression profile that differentiates ischemic and nonischemic cardiomyopathy. Circulation 110(22):3444-3451.
is feasible and useful in myocardial diseases, it is essential to broaden the focus of gene expression research in cardiology. Currently, the focus is on the identification of differentially expressed genes and validation through confirmation of gene expression levels through quantitative PCR and protein products through Western blotting and immunofluorescence. To make the transition from gene discovery to molecular signature analysis, gene expression research in cardiology must develop a new focus: on determining the predictive power of gene expression. One important distinction in molecular signature analysis is the need first to identify a set of genes whose expression characterizes a predefined group of patients (with a clinically relevant distinction such as etiology, prognosis, or response to therapy) and then to test the predictive accuracy of this profile prospectively in an independent set of patients with varying phenotypes (Cook and Rosenzweig 2002). Another important distinction is that molecular signature analysis is based upon a pattern of gene expression rather than the identity of specific genes (Simon et al. 2003). The prediction algorithm is able to compare an unknown sample and determine how closely it resembles one pattern versus the other; the absolute expression of an individual gene carries relatively small
136
weight compared with the overall signature. Thus, validation cannot solely involve the confirmation of gene expression or gene product levels via a complementary technique such as quantitative PCR or immunofluorescence. Nevertheless, such validation can prove useful to address a different issue: whether the molecular signature offers utility independent of the microarray platform used to create it. This is important if diseasespecific platforms are developed, as in the oncology experience (Lossos et al. 2004, Paik et al. 2004). Sample Size in Microarray Experiments There is limited knowledge on the sample size required in microarray experiments. The largest microarray study in cardiomyopathy to date has involved 199 patient samples (Margulies et al. 2005), and in the oncology literature, each study has employed fewer than 300 patients. However, this should be considered a strength of these analyses: a succession of smaller studies, performed quickly and with the use of improving technology, will outperform larger studies locked into outdated approaches (Liu and Karuturi 2004). The Source of Tissue for Analysis To date, microarray analyses in cardiomyopathy have mainly used discarded myocardial tissue obtained at the time of cardiac transplantation or LVAD placement, and there are limitations to this approach. First, the tissue is obtained from patients late in the disease course and thus the conclusions may not be applicable to patients at an earlier stage of disease; as we demonstrated, a molecular signature based on etiology in end-stage cardiomyopathy is specific to disease stage (Kittleson et al. 2004). Second, explanted tissue is obtained from different areas of the left ventricle, and there is evidence, in mice, that regional differences in gene expression exist in the left ventricle (Mirotsou et al. 2003). Thus, in the future, microarray analyses in cardiomyopathy will ideally focus on endomyocardial biopsy tissue obtained from patients at earlier stages of disease. Although one commonly invoked limitation of gene expression research
in cardiovascular disease is the lack of ready access to human heart tissue samples, endomyocardial biopsies are frequently performed to evaluate newly diagnosed cardiomyopathy at our institution (Ardehali et al. 2004, Felker et al. 2000). In our experience, endomyocardial biopsy is a safe and well-tolerated procedure, with an overall mortality rate of 0.2%, a rate equivalent to that of other routinely performed catheterization procedures (Felker et al. 1999), and we have demonstrated that microarray hybridization from endomyocardial biopsies is feasible (Kittleson et al. 2004). Although endomyocardial biopsy is a safe procedure that could be more widely performed if a valuable prognostic or diagnostic test were developed, venipuncture is clearly more accessible. Therefore, it is essential to also test the utility of molecular signature analysis in peripheral blood leukocyte samples. In the cancer literature, molecular signatures derived from peripheral blood leukocytes offer comparable predictive accuracy with those from solid tumor samples in classifying subjects by cancer type and type of therapy (DePrimo et al. 2003, Twine et al. 2003). A recent study suggests that this may also be feasible in cardiovascular disease, as peripheral blood molecular signatures correlated with biopsy-proven allograft rejection in cardiac transplant recipients (Horwitz et al. 2004).
Summary
The current approach to the treatment of patients with cardiomyopathy lacks individualization. It is widely anticipated that the future management of patients with heart failure will be tailored based on individual assessments of prognosis and response to therapy. Currently, the ability to predict which newly diagnosed patients with cardiomyopathy will improve their ejection fraction and functional status, and which will go on to develop circulatory collapse and require cardiac transplantation, is still not possible. Studies of gene discovery could identify novel therapeutic targets for the treatment of heart failure, and molecular signature analysis could augment current clinical and imaging modalities used to determine prognosis and response to therTCM Vol. 15, No. 4, 2005
apy. Our identification of a molecular signature that differentiates cardiomyopathy by etiology supports ongoing efforts to incorporate expression-profiling-based biomarkers in determining prognosis and response to therapy. The ultimate potential application of transcriptome-based molecular signature analysis is the individualization of the management of patients with heart failure. In the future, a patient with a newly diagnosed cardiomyopathy could, through molecular signature analysis, be offered an accurate assessment of prognosis and how individualized medical therapy could affect his or her outcome.
References Ardehali H, Qasim A, Cappola T, et al.: 2004. Endomyocardial biopsy plays a role in diagnosing patients with unexplained cardiomyopathy. Am Heart J 147:919–923. Barrans JD, Allen PD, Stamatiou D, et al.: 2002. Global gene expression profiling of end-stage dilated cardiomyopathy using a human cardiovascular-based cDNA microarray. Am J Pathol 160:2035–2043. Blaxall BC, Tschannen-Moran BM, Milano CA, Koch WJ: 2003. Differential gene expression and genomic patient stratification following left ventricular assist device support. J Am Coll Cardiol 41: 1096–1106.
Cook SA, Rosenzweig A: 2002. DNA microarrays: implications for cardiovascular medicine. Circ Res 91:559–564. Dave SS, Wright G, Tan B, et al.: 2004. Prediction of survival in follicular lymphoma based on molecular features of tumor-infiltrating immune cells. N Engl J Med 351:2159–2169. DePrimo SE, Wong LM, Khatry DB, et al.: 2003. Expression profiling of blood samples from an SU5416 phase III metastatic colorectal cancer clinical trial: a novel strategy for biomarker identification. BMC Cancer 3:3. Dudoit S, Gentleman RC, Quackenbush J: 2003. Open source software for the analysis of microarray data. Biotechniques (Suppl):45–51. Felker GM, Hu W, Hare JM, et al.: 1999. The spectrum of dilated cardiomyopathy. The Johns Hopkins experience with 1,278 patients. Medicine (Baltimore) 78:270–283. Felker GM, Thompson RE, Hare JM, et al.: 2000. Underlying causes and long-term survival in patients with initially unexplained cardiomyopathy. N Engl J Med 342:1077–1084. Hall JL, Grindle S, Han X, et al.: 2004. Genomic profiling of the human heart before and after mechanical support with a ventricular assist device reveals alterations in vascular signaling networks. Physiol Genomics 17:283–291. Holleman A, Cheok MH, Den Boer ML, et al.: 2004. Gene-expression patterns in drugresistant acute lymphoblastic leukemia cells and response to treatment. N Engl J Med 351:533–542.
Kittleson MM, Minhas KM, Irizarry RA, et al.: 2005b. Gene expression in giant cell myocarditis: altered expression of immune response genes. Int J Cardiol (in press). Liu ET, Karuturi KR: 2004. Microarrays and clinical investigations. N Engl J Med 350:1595–1597. Lossos IS, Czerwinski DK, Alizadeh AA, et al.: 2004. Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 350:1828–1837. Margulies KB: 2002. Reversal mechanisms of left ventricular remodeling: lessons from left ventricular assist device experiments. J Card Fail 8:S500–S505. Margulies KB, Matiwala S, Cornejo C, et al.: 2005. Mixed messages. Transcription patterns in failing and recovering human myocardium. Circ Res 1–8. Mirotsou M, Watanabe CMH, Schultz PG, et al.: 2003. Elucidating the molecular mechanism of cardiac remodeling using a comparative genomic approach. Physiol Genomics 15:115–126. Paik S, Shak S, Tang G, et al.: 2004. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817–2826. Rosenwald A, Wright G, Chan WC, et al.: 2002. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 346:1937–1947. Simon R, Radmacher MD, Dobbin K, McShane LM: 2003. Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 95:14–18.
Boheler KR, Volkova M, Morrell C., et al.: 2003. Sex- and age-dependent human transcriptome variability: implications for chronic heart failure. Proc Natl Acad Sci USA 100:2754–2759.
Horwitz PA, Tsai EJ, Putt ME, et al.: 2004. Detection of cardiac allograft rejection and response to immunosuppressive therapy with peripheral blood gene expression. Circulation 110:3815–3821.
Bullinger L, Dohner K, Bair E, et al.: 2004. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 350: 1605–1616.
Hwang JJ, Allen PD, Tseng GC, et al.: 2002. Microarray gene expression profiles in dilated and hypertrophic cardiomyopathic end-stage heart failure. Physiol Genomics. 10:31–44.
Steenman M, Lamirault G, Le Meur N, et al.: 2005. Distinct molecular portraits of human failing hearts identified by dedicated cDNA microarrays. Eur J Heart Fail 7:157–165.
Carey VJ, Gentry J, Whalen E, Gentleman R: 2005. Network structures and algorithms in Bioconductor. Bioinformatics 21:135–136.
Kaab S, Barth AS, Margerie D, et al.: 2004. Global gene expression in human myocardium–oligonucleotide microarray analysis of regional diversity and transcriptional regulation in heart failure. J Mol Med 82:308–316.
Tan FL, Moravec CS, Li J, et al.: 2002. The gene expression fingerprint of human heart failure. Proc Natl Acad Sci USA 99: 11387–11392.
Chaitman BR, Pepine CJ, Parker JO, et al.: 2004. Effects of ranolazine with atenolol, amlodipine, or diltiazem on exercise tolerance and angina frequency in patients with severe chronic angina: A randomized controlled trial. JAMA 291:309–316. Chen MM, Ashley EA, Deng DX, et al.: 2003a. Novel role for the potent endogenous inotrope apelin in human cardiac dysfunction. Circulation 108:1432–1439. Chen Y, Park S, Li Y, et al.: 2003b. Alterations of gene expression in failing myocardium following left ventricular assist device support. Physiol Genomics 14:251–260
TCM Vol. 15, No. 4, 2005
Kittleson MM, Ye SQ, Irizarry RA, et al.: 2004. Identification of a gene expression profile that differentiates between ischemic and nonischemic cardiomyopathy. Circulation 110:3444–3451. Kittleson MM, Minhas KM, Irizarry RA, et al.: 2005a. Gene expression analysis of ischemic and nonischemic cardiomyopathy: shared and distinct genes in the development of heart failure. Physiol Genomics 21: 299–307.
Steenman M, Chen YW, Le Cunff M, et al.: 2003. Transcriptomal analysis of failing and nonfailing human hearts. Physiol Genomics 12:97–112.
Tibshirani R, Hastie T, Narasimhan B, Chu G: 2002. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 99:6567–6572. Tusher VG, Tibshirani R, Chu G: 2001. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 98:5116–5121. Twine NC, Stover JA, Marshall B, et al.: 2003. Disease-associated expression profiles in peripheral blood mononuclear cells from patients with advanced renal cell carcinoma. Cancer Res 63:6069–6075.
137
Valk PJ, Verhaak RG, Beijen MA, et al.: 2004. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 350:1617–1628. Van De Vijver MJ, He YD, Van’t Veer LJ, et al.: 2002. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 347:1999–2009. Yung CK, Halperin VL, Tomaselli GF, Winslow RL: 2004. Gene expression profiles in end-stage human idiopathic dilated cardiomyopathy: altered expres-
Hsp20 and Its Cardioprotection Guo-Chang Fan, Guoxiang Chu, and Evangelia G. Kranias*
The small heat shock protein Hsp20, also referred to as P20/HspB6, is expressed in the brain, stomach, liver, lung, kidney, blood, smooth muscle, skeletal muscle, and cardiac tissue. Although Hsp20 is not heatinducible, several cellular signaling pathways appear to regulate its biologic functions. In recent years, tremendous advances have been made in elucidating the significance of Hsp20 in smooth muscle and its potential benefits on coronary vasculature. Of interest, recent findings have demonstrated that sustained b-adrenergic stimulation results in expression and phosphorylation of cardiac Hsp20. Moreover, Hsp20 overexpression enhances cardiac function and renders cardioprotection against b-agonist-mediated apoptosis and ischemia/reperfusion injury ex vivo and in vivo. This article reviews the new findings on translocation of Hsp20 in response to various stimuli and the multiple cellular targets of Hsp20, with special emphasis on its protective effects in the heart. (Trends Cardiovasc Med 2005;15:138–141) D 2005, Elsevier Inc. In mammalian species, the subfamily of small heat shock proteins (sHsps, 15-30 kDa) comprises of 10 known members (Kappe et al. 2003), namely, Hsp27 (Hsp25 of rodents, HspB1), myotonic dystrophy protein kinase binding protein (MKBP, HspB2), HspB3, aA-crystallin (HspB4), aB-crystallin (HspB5), Hsp20 (P20, HspB6), cvHsp (HspB7), Hsp22 (H11, HspB8), HspB9, and HspB10. According to their different patterns of
Guo-Chang Fan, Guoxiang Chu, and Evangelia G. Kramias are at the Department of Pharmacology and Cell Biophysics, University of Cincinnati College of Medicine, Cincinnati, Ohio. * Address correspondence to: Evangelia G. Kranias, PhD, Department of Pharmacology and Cell Biophysics, University of Cincinnati College of Medicine, 231 Albert Sabin Way, Cincinnati, OH 45267-0575, USA. Tel.: (+1) 513-558-2377; fax: (+1) 513-558-2269; e-mail: [email protected]. D 2005, Elsevier Inc. All rights reserved. 1050-1738/05/$-see front matter
138
gene expression and subcellular localization, Taylor and Benjamin (2005) classified these sHsps into two major categories: classes I and II. Class I sHsps (HspB1, HspB5, HspB6, and HspB8) are ubiquitously expressed, whereas class II members (HspB2, HspB3, HspB4, HspB7, HspB9, and HspB10) display tissue-restricted patterns of expression. Hsp20, also referred to as P20/HspB6, was originally copurified with the sHsps Hsp27/Hsp28 from rat and human skeletal muscles (Kato et al. 1994). Its genome contains 3 exons and 2 introns. Human, rat, and mouse Hsp20 are composed of 157, 162, and 162 amino acids, respectively. Although hyperthermia did not induce Hsp20 in rat tissue (Kato et al. 1994), heat pretreatment of swine carotid artery was associated with increased Hsp20 levels (O’Connor and Rembold 2002). This protein is detected in all tissues by a sandwich-type immunoassay system, reaching a maximal level of 1.3% of total protein in skeletal,
sion of apoptotic and cytoskeletal genes. Genomics 83:281–297.
PII S1050-1738(05)00061-7
TCM
heart, and smooth muscles (Kato et al. 1994), and its expression levels are altered during development (Inaguma et al. 1996, Verschuure et al. 2003). Recent studies have revealed the significance of Hsp20 in relaxation of vascular muscle (Beall et al. 1997, Rembold et al. 2000, 2001, Brophy et al. 2002, Flynn et al. 2003, Woodrum et al. 1999, Meeks et al. 2005) and inhibition of platelet aggregation (Kozawa et al. 2002). Of particular interest, our recent work (Chu et al. 2004) discovered a cardiac isoform of Hsp20 in mouse heart, which was inducible by isoproterenol stimulation, and comparison of its derived amino acid sequence with rat and human skeletal Hsp20 revealed 95% and 87% homology, respectively. Furthermore, cardiac Hsp20 overexpression renders protection against stressinduced injury in vitro and in vivo, through its antiapoptotic properties (Fan et al. 2004, 2005). Therefore, it is plausible that Hsp20 provides multifaceted beneficial effects in the heart. This brief review will highlight the translocation of Hsp20 in response to various stresses, interaction with its targets, cardioprotective effects, and other physiologic functions.
Translocation of Hsp20 in Response to Stress
Hsp20 is normally located in the cytoplasm of cardiac myocytes. After a heat stress, a subpopulation of Hsp20 migrates into the nucleus, whereas the other part remains in the cytoplasm. In very few cells, a faint sarcomeric association of Hsp20 is observed (Van de Klundert and de Jong 1999); however, Hsp20 is prominently translocated to the myofibrils in adult rat heart and skeletal muscle under ischemic conditions (Golenhofen et al. 2004). A similar phenomenon with Hsp20 redistribution to stress fibers has also been observed in the rat cardiac myoblast cell line H9C2 after proteasomal inhibition (Verschuure et al. 2002). Conversely, one study demonstrated that Hsp20 was predominantly in transverse bands, TCM Vol. 15, No. 4, 2005