Physiol. Genomics Ad Instruments
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Physiol. Genomics 34: 88-94, 2008. First published April 22, 2008; doi:10.1152/physiolgenomics.00299.2007
1094-8341/08 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Figures and Tables
Right arrow All Versions of this Article:
34/1/88    most recent
00299.2007v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kuner, R.
Right arrow Articles by Nabauer, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Kuner, R.
Right arrow Articles by Nabauer, M.
Received 20 December 2007; accepted in final form 19 April 2008.
Physiological Genomics 34:88-94 (2008)
1094-8341/08 $8.00 © 2008 American Physiological Society

Genomic analysis reveals poor separation of human cardiomyopathies of ischemic and nonischemic etiologies

Ruprecht Kuner 1,*, Andreas S. Barth 2,3,*, Markus Ruschhaupt 1,5, Andreas Buness 1, Ludwig Zwermann 3, Eckart Kreuzer 4, Gerhard Steinbeck 3, Annemarie Poustka 1, Holger Sültmann 1,* and Michael Nabauer 3,*

1 Division of Molecular Genome Analysis, German Cancer Research Center, Heidelberg, Germany
2 Division of Cardiology, Johns Hopkins University, Baltimore, Maryland
3 Department of Medicine I
4 Department of Cardiac Surgery, University Hospital Grosshadern, Ludwig Maximilians University
5 Department of Medical Informatics, Biometrics, and Epidemiology, Ludwig Maximilians University, Munich, Germany


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Clinically, the differentiation between ischemic (ICM) and nonischemic (NICM) human cardiomyopathies is highly relevant, because ICM and NICM differ with respect to prognosis and certain aspects of pharmacological therapy, despite a common final phenotype characterized by ventricular dilatation and reduced contractility. So far, it is unclear whether microarray-based signatures can be used to infer the etiology of heart failure. Using three different classification algorithms, we independently analyzed one cDNA and two publicly available high-density oligonucleotide microarray studies comprising a total of 279 end-stage human heart failure samples. When classifiers identified in a single study were applied to the remaining studies, misclassification rates >25% for ICM and NICM specimens were noted, indicating poor separation of both etiologies. However, data mining of 458 classifier genes that were concordantly identified in at least two of the three data sets points to different biological processes in ICM vs. NICM. Consistent with the underlying ischemia, cytokine signaling pathways and immediate-early response genes were overrepresented in ICM samples, whereas NICM samples displayed a deregulation of cytoskeletal transcripts, genes encoding for the major histocompatibility complex, and antigen processing and presentation pathways, potentially pointing to immunologic processes in NICM. Overall, our results suggest that ICM and NICM exhibit substantial heterogeneity at the transcriptomic level. Prospective studies are required to test whether etiology-specific gene expression patterns are present at earlier disease stages or in subsets of both etiologies.

dilated cardiomyopathy; ischemic cardiomyopathy; classification; functional genomics


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
HEART FAILURE due to dilated cardiomyopathy, characterized by impaired contractile function of one or both ventricles and dilatation of the cardiac chambers, is a leading cause of hospitalization and death in Western countries (14). Most frequently, it develops in the setting of coronary atherosclerosis, with ischemic cardiomyopathy (ICM) accounting for >50% of cases of heart failure (2). The second largest heart failure entity is dilated cardiomyopathies of nonischemic origin (NICM), comprising various subgroups such as valvular and hypertensive heart disease, viral myocarditis, hereditary cases, toxic forms (e.g., alcohol and anthracycline cardiomyopathy), as well as a subset of idiopathic cases in which no specific cause can be determined.

In general, pharmacological treatment of overt heart failure is largely based on functional New York Heart Association (NYHA) status and differs little with respect to the underlying disease etiology. However, it has become increasingly clear that long-term prognosis and the response to certain pharmacological modalities may very well differ between ICM and NICM. For instance, response to digoxin and amiodarone was found to be better in nonischemic patients (1, 30). Likewise, large epidemiologic studies have shown that heart failure of ischemic origin is associated with a worse outcome compared with NICM (2, 18). Therefore, the differentiation between ischemic and nonischemic cases is clinically highly relevant but relies, so far, mainly on the results of a coronary angiography. While this invasive technique will correctly confirm (or rule out) significant coronary atherosclerosis in most cases, it might be misleading in up to 20% of patients who present after a myocardial infarction with spontaneous recanalization of the culprit coronary artery. Therefore, these patients will be misclassified as "nonischemic" in the absence of additional significant coronary stenoses (40). In this respect, genomic studies hold great promise to improve our current classification schemes. While we have recently reported a common and robust gene expression profile characteristic of failing human ventricular myocardium across >100 NICM and nonfailing (NF) samples of four independent microarray studies (4), it is currently less clear whether microarray-based signatures can also be used to infer the etiology of heart failure in end-stage NICM and ICM. Several pilot studies have suggested differences in genomic signatures between different etiologies of heart failure (7, 21, 22, 33), while others have failed to clearly differentiate between ischemic and idiopathic dilated cardiomyopathy on the basis of transcript profiling (12, 31). Therefore, we performed class prediction analysis independently in one cDNA and two publicly available high-density oligonucleotide microarray studies, comprising a total of 279 human myocardial samples. We then applied potential classifiers identified in a single study to the remaining data sets to test whether etiology-specific patterns of gene expression are discernible in end-stage human heart disease of ischemic and nonischemic origin.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
cDNA microarray study.
We performed a genomewide expression study that comprised septal myocardial samples from heart failure patients with end-stage NICM (n = 16) and ICM (n = 10). While the latter patients had a severe two- or three-vessel coronary disease (≥75% stenosis) and/or a history of myocardial infarction, coronary angiography showed no signs of significant (>50%) stenoses in NICM subjects. Detailed patient characteristics are given in Supplemental Table S1.1 All patients gave written informed consent. The investigation was approved by the Institutional Review Board and is in accordance with the Helsinki Declaration of 1975, as revised in 1983.

As a two-color microarray experiment, each NICM and ICM sample was labeled with Cy3 and hybridized against a common pool of RNA from 15 NF hearts labeled with Cy5. RNA amplification, labeling, and hybridization to RZPD Unigene 3.1 cDNA microarrays (37.5 K) were carried out as described previously (4). Preprocessing and most of the statistical analysis were done with Bioconductor (www.bioconductor.org). The cDNA microarray raw data were normalized with the "variance stabilization" method (VSN; Ref. 20) and the arrayMagic software package (10). Hierarchical clustering of genes and samples was based on the Euclidean distance measure. The platform iCHIP (Integration Centre of High Throughput Experiments; www.ichip.de) was used to export the microarray data via MAGE-ML to the repository ArrayExpress of the European Bioinformatics Institute (http://www.ebi.ac.uk/arrayexpress, accession number E-CVDE-1).

Etiologic classification of three microarray studies of human heart failure.
We classified NICM and ICM samples in one cDNA (E-CVDE-1) and two publicly available high-density oligonucleotide microarray studies (NCBI Gene Expression Omnibus database accession numbers GSE5406 and GSE1145). Data set GSE5406 represents the largest microarray study of human heart failure so far, comparing 108 ICM and 86 NICM samples hybridized to Affymetrix HG-U133A arrays (19). Data set GSE1145 consists of 32 ICM and 27 NICM samples hybridized to Affymetrix HG-U133 2.0 plus arrays (www.cardiogenomics.org). In summary, the present meta-analysis comprised a total of 150 ICM and 129 NICM samples from three independent microarray studies. More detailed information regarding these studies is provided in Table 1. Affymetrix probes and cDNA clones were merged between different microarray platforms with Bioconductor annotation packages (www.bioconductor.org).


View this table:
[in this window]
[in a new window]

  		Table 1. Characterization of human microarray studies focusing on different etiologies of heart failure

 
Three different classification methods [prediction analysis for microarrays (PAM; Ref. 34), random forest (RF; Ref. 9), and penalized logistic regression (PLR; Ref. 41)] were used to classify ICM and NICM samples in three independent microarray data sets. PAM uses denoised versions of the centroids as prototypes for each class. In the RF method, different decision trees are grown by introducing a random element into their construction. First, each tree makes a prediction of one class for each observation, and then these predictions are combined to get the final classification result. PLR combines a logistic regression model with a ridge penalty to classify microarray data. To estimate the fraction of misclassified samples, a complete cross-validation implemented in the Bioconductor package "MCRestimate" (28) was applied. This complete cross-validation includes an additional gene filtering step based on the genes with the highest variance. After estimating the misclassification rate in all three studies by the three aforementioned methods of classification, we then applied the most successful classifier (GSE1145, RF) as well as a published classifier of ICM vs. NICM (22) to the remaining microarray studies.

Data mining was carried out with the classifiers derived from the three aforementioned studies and with 458 classifier genes that were concordantly identified in at least two of the three data sets. Statistical analysis of these genes was based on Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and the Gene Ontology (GO) domains "cellular component," "biological process," and "molecular function," implemented in "FatiGO" (3). Ingenuity Pathways Knowledge Base (Ingenuity, Mountain View, CA), the largest manual curated gene database, was used to identify literature-based gene-gene interaction networks. The utility of this database for microarray data of myocardial samples has been demonstrated previously (38).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 GRANTS
 REFERENCES
 
Classification of nonischemic and ischemic cardiomyopathies.
Applying three different classification algorithms to our own cDNA microarray data set (E-CVDE-1) and two larger, publicly available microarray data sets (GSE1145 and GSE5406), we found that on average one-fourth of the samples were misclassified (range 14–32%; Fig. 1 and Supplemental Figs. S1, S2, and S3). Because the definition of incorrectly classified samples was based on 50 resamplings, Supplemental Figs. S1, S2, and S3 show a graphical representation of the results of the cross-validation for each individual sample (i.e., in Supplemental Fig. S1, samples 93, 101, and 103–106 were consistently misclassified, while sample 102 shows borderline classification results, with an equal number of correct and incorrect classifications). Furthermore, these graphs demonstrate that the results of the classification were not very stable even in those cases where an apparent low misclassification rate was computed, e.g., with RF (14%) and PLR (17%) in data set GSE1145. This is best exemplified for the PLR method in GSE1145 (Supplemental Fig. S2), where one can appreciate a high variation even for those samples that were correctly classified. The variability suggests that the classifier itself is not very robust. This is supported by the fact that the results of the best classifier (correct classification of 86% of samples with RF in data set GSE1145) could not be reproduced in independent data sets, because misclassification rates were 34% and 35% in data sets GSE5406 and E-CVDE-1, respectively. As a limitation of the analysis of different microarray platforms, the 509 nonredundant classifier genes could only be matched to a subset of identifiers in the two other data sets (GSE5406: 414 genes and E-CVDE-1: 289 genes). To exclude that the poor classification result in data sets GSE5406 and E-CVDE-1 might be due to the lack of genes important for the classification process in these data sets, we applied the RF algorithm in GSE1145, using 414 genes available in GSE5406 and 289 genes available in data set E-CVDE-1. Because of overfitting, these results cannot be compared with the initial classification result for the data set. Therefore, we also performed the RF classification based on the 509 nonredundant classifier genes. There was no difference between the three resulting misclassification rates (data not shown). This suggests that the poor classification results obtained with the RF classifier from GSE1145 in data sets GSE5406 and E-CVDE-1 is not due to the lack of genes important for the classification algorithm. To further support this notion, we performed an additional analysis in the two largest data sets, GSE1145 and GSE5406, encompassing >90% of all 279 samples included in the present article. Based on ENTREZ IDs, we first determined the genes common to both data sets. Second, we applied the RF classification procedure in GSE1145, using only probe sets that were shared between both data sets. Supplemental Figure S5 shows that the classification results obtained in data set GSE1145 (misclassification rate 14%) are similar to our previous analysis using the complete data set (Fig. 1). However, application of this classifier to GSE5406 still yielded a poor classification, with a misclassification rate of 34%.


Figure 1
View larger version (27K):
[in this window]
[in a new window]

  		Fig. 1. Genomic classification of nonischemic (NICM) and ischemic cardiomyopathies (ICM) in 3 microarray studies. The misclassification rate and the number of genes used for classification were assessed individually for each study with 3 different methods (RF, random forest; PLR, penalized logistic regression; PAM, prediction analysis for microarrays). The results of the best classifier (86% correct classification with RF in data set GSE1145) could not be reproduced in the other 2 data sets. Likewise, misclassification rates up to 45% were observed when a previously published classifier for NICM and ICM (22) was applied to the remaining 3 data sets with PAM classification.

 
Likewise, we found no clear evidence that sex influences the misclassification rates for NICM or ICM samples (8). Even though men are in general overrepresented in both etiologies (across all studies ~60% and ~85% of subjects were male in the NICM and ICM groups, respectively), sex did not play a major role in the results of the current classification for NICM or ICM samples (Supplemental Table S1). This was also supported by a separate analysis of myocardial samples of male patients only that yielded similar results compared with analysis of the entire data set (Supplemental Fig. S6).

In an additional classification analysis, we examined whether a published 90-gene classifier for ICM and NICM (22) can correctly distinguish ICM from NICM in the three aforementioned microarray data sets. Here, the classifier genes were only reduced in data set E-CVDE-1 (35 genes). However, 40% of 279 myocardial samples were misclassified again, indicating that results from this study cannot be extrapolated to independent cardiomyopathy data sets (Fig. 1).

In contrast, by applying our previously described genomic classifier for heart failure (4) to the largest data set (GSE5406, Supplemental Fig. S4), we now achieved successful classification between NF and failing myocardium in a total of 347 of 360 samples, yielding an overall prediction accuracy of 96.4%, which is in sharp contrast to the high misclassification rate between NICM and ICM in the same microarray data set.

Distinct biological processes in ischemic and nonischemic dilated cardiomyopathies.
To obtain etiology-specific transcriptional patterns in NICM vs. ICM, we performed GO and KEGG pathway analyses with classifiers derived from each data set separately. Furthermore, to define a common, minimal denominator across independent data sets, we also performed a GO analysis with 458 classifier genes that were concordantly deregulated in at least two of the three microarray studies (Fig. 2A and Supplemental Tables S2 and S3). As a result, we found an overrepresentation of cytoskeletal transcripts, genes encoding for the major histocompatibility complex, and antigen processing and presentation pathways in NICM samples. In contrast, the GO groups of "inflammatory response," "cytokine activity," and "G protein-coupled receptor binding" were enriched in ICM samples (Fig. 2B). Further analysis of the classifier genes characteristic of ICM samples revealed an overrepresentation of these transcripts in a prominent gene-gene interaction network associated with the immediate-early response genes FOS and JUNB (Fig. 3A). Of note, many of the genes comprised in this gene network have previously been implicated in human cardiomyopathies, including CCL2 (4), ATF3 (4), SFRP4 (4), LAMP2 (39), IGFBP (6), and TIMP1 (5). Upregulation of these ICM classifier genes could be linked to a subset of ICM specimens, as suggested by the hierarchical clustering in data set GSE1145 (Fig. 3B) with separation of two distinct clusters comprising predominantly ICM samples on the left (11 ICM vs. 3 NICM) and a cluster with a nearly equal number of ICM and NICM samples on the right.


Figure 2
View larger version (21K):
[in this window]
[in a new window]

  		Fig. 2. A: functional analysis consisted of 458 classifier genes shared by at least 2 of the 3 microarray studies. The Venn diagram depicts the overlap of classifier genes obtained with RF classification. B: ICM classifier genes showed an overrepresentation of cytokine signaling pathways and inflammatory and chemotactic genes, whereas classifier genes characteristic of NICM displayed a deregulation of cytoskeletal transcripts, genes encoding for the major histocompatibility complex, and antigen processing and presentation pathways (P < 0.05 by Fisher exact test for all groups shown). KEGG, Kyoto Encyclopedia of Genes and Genomes.

 

Figure 3
View larger version (45K):
[in this window]
[in a new window]

  		Fig. 3. A: Ingenuity pathway analysis suggested an overrepresentation of ICM classifier genes in a gene-gene interaction network with the 2 immediate-early response genes FOS and JUNB. The genes shown in gray are derived from the ICM classifier list (Supplemental Table S3). The other genes were suggested from the literature as potential interaction partners. B: unsupervised clustering using Euclidean distance for the genes of the Ingenuity network in data set GSE1145. Red indicates upregulation and blue indicates downregulation of these genes. NICM samples are represented by yellow and ICM samples by green bars. Upregulation of these ICM classifier genes could be linked to a subset of ICM specimens, as the clustering suggested separation of samples into 2 distinct clusters, 1 on the left comprising predominantly ICM samples (11 ICM and 3 NICM) and 1 on the right with a nearly equal number of ICM and NICM samples.

 
Notably, t-test-based methods did not allow a comprehensive analysis of the different etiologies since statistical analysis of microarrays (SAM) analysis (35) resulted only in a list of 13 genes in the largest data set (GSE5406) with negligible overlap to the other microarray data sets (data not shown).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
GRANTS
 REFERENCES
 
Classifying end-stage ischemic and nonischemic cardiomyopathies.
Clinically, the differentiation into NICM and ICM is highly relevant because, in nonischemic cases, secondary causes (e.g., myocarditis and excessive alcohol consumption) need to be identified, while ischemic patients, on the contrary, may benefit from revascularization and secondary preventive pharmacotherapy with lipid-lowering drugs and antiplatelet agents. However, correct classification between ischemic and nonischemic cases is difficult, as highlighted by a study using gadolinium-enhanced cardiovascular magnetic resonance imaging that suggested an incorrect assignment of ischemic and nonischemic heart failure etiology in 13% of heart failure patients (24). Likewise, the Assessment of Treatment with Lisinopril and Survival study (ATLAS) showed an incorrect classification of NICM and ICM cases in up to 28% based on autopsy data (36). In this respect, incorrect assignment of disease etiology based on coronary angiography would not only be misleading in patients in whom spontaneous revascularization of the coronary artery occurs after a myocardial infarction (40), but also in patients who are classified as NICM and who show major coronary findings at autopsy (36).

By convention, new diagnostic tests should be validated against an established gold standard. Because several clinical modalities including coronary angiography and magnetic resonance imaging (MRI) have been shown to lack the sensitivity to correctly classify the etiology of heart failure in all cases, the most appropriate reference standard is postmortem autopsy. As a limitation, autopsy data were only available for a subset of cases in our own cDNA microarray study (Supplemental Table S1) and have not been provided in any of the published studies dealing with different etiologies of heart failure (19, 22). Thus, in the present study, phenotypic classification was largely based on information routinely used by clinicians to assign patients to ischemic or nonischemic disease etiologies. However, it is of interest to note that even in those cases that could be diagnosed unequivocally as ICM, misclassification based on microarray data was common. As an example, 5 of 10 ICM samples were misclassified in the cDNA microarray study—despite the fact that these patients had a history of prior myocardial infarctions, coronary artery bypass graft surgeries, and three-vessel coronary disease on coronary angiograms. Several explanations are possible. First, a severe ICM is expected to exhibit significant heterogeneity at the molecular level, reflecting areas of nonischemic, ischemic, stunned, or hibernating myocardium, all of which show distinct genomic profiles in animal models and human heart failure (15–17). In agreement with this hypothesis, we noted significant heterogeneity of ischemia marker genes in samples from ICM patients (Fig. 3). The fact that those ischemia marker genes were also upregulated in some NICM samples goes along with the fact that coronary microvascular dysfunction has been shown to be prevalent in NICM as well (26), leading to defined areas of ischemia that might show a gene expression profile similar to ICM. Recent MRI studies, showing patchy or longitudinal striae of midwall enhancement with a distribution pattern different from the fibrotic areas observed in coronary artery disease, support this hypothesis (24).

Second, it might be possible that the underlying etiology is blurred in end-stage human heart failure in a substantial number of cases. After an initial cardiac insult, producing a decline in pump function, myocardial remodeling and dilatation in ischemic as well as nonischemic heart failure is aggravated by systemic activation of neurohormonal systems. Hence, the concept of the final common pathway of cardiomyopathic diseases has been established, in which dilatation is a nonspecific phenotype of severely stressed myocardium, irrespective of the initiating cardiac insult (23). In line with this notion, all major pharmacological interventions known to increase survival in heart failure patients (NICM and ICM) aim at disrupting the vicious circle of neurohormonal dysregulation, i.e., they inhibit either the adrenergic or the renin-angiotensin-aldosterone system. Thus expression patterns in end-stage human heart disease might be dominated by common heart failure pathways, as demonstrated by our previously described genomic classifier for heart failure, which was derived from NICM vs. NF tissue samples but discriminated ICM from NF myocardium as well (Ref. 4; Supplemental Fig. S4).

As a limitation, we did not attempt to stratify the NICM cases for postmyocarditis, end-stage hypertensive cardiomyopathy, or alcohol-induced or familial forms of NICM. We felt that the clinical information provided was insufficient even for our own patients, with the large majority being classified as "idiopathic" NICM. Future studies are necessary to separate those entities, preferably at earlier disease stages by examining endomyocardial biopsies.

From a methodological point of view, integrating independent data sets into a common analysis remains a challenge, even though some strategies have been proposed (32, 37). However, it remains questionable whether the gain of information obtained from examining a larger number of samples can compensate for the loss of information resulting from data reduction. In the present study, we avoided the loss of information and used our own as well as publicly available data sets to validate our results. Of note, this approach was successfully applied in previous studies (4, 29). However, for ICM and NICM, we were neither able to identify a novel robust classifier in three independent data sets nor to apply a published classifier from a fourth study (22). Our additional analysis, where only common genes of independent data sets were used for classification, argues against a technical limitation for the poor misclassification rate. In summary, we presented multiple lines of evidence for a poor separation of ICM and NICM etiologies based on independent microarray studies and different classification methods.

Distinct biological processes in ischemic and nonischemic dilated cardiomyopathies.
Atherosclerosis is currently considered to be an inflammatory disease in which the initiation and progression of an atherosclerotic toward an unstable plaque is driven by leukocyte recruitment mediated by various inflammatory mediators, including chemotactic cytokines or chemokines (11). Consistent with this hypothesis, ICM samples were characterized by the functional GO classes of "inflammatory response," "chemotaxis," "cytokine activity," and "myeloid cell differentiation" (Fig. 2). Ingenuity pathway analysis suggested a gene-gene interaction network centered around the immediate-early response genes FOS and JUNB. It has been shown that brief episodes of myocardial ischemia activate a distinct genetic program promoting cardioprotection and cell survival, with the immediate-early response genes being among the most consistent findings of ischemic preconditioning (15, 27).

In NICM, GO analysis suggested involvement of immune processes with upregulation of major histocompatibility complex genes. In line with this finding, KEGG pathway analysis revealed "antigen processing and presenting pathways" to be more prominent in NICM than ICM. Of note, idiopathic nonischemic dilated cardiomyopathy has been proposed as a disease with autoimmune features (25). The finding that immune processes were involved in both ICM and NICM is consistent with our previous study (4), in which we found a deregulation of the immune system to be characteristic of end-stage human failing myocardium. In addition to deregulation of immune processes in NICM, it is of particular interest to find deregulation of cytoskeletal and sarcomere transcripts as well (Supplemental Tables S2 and S3), especially as a wide variety of studies point to a link between mutations in cytoskeletal genes and the initiation of dilated cardiomyopathy (13). However, one has to keep in mind that our classification results and the low number of differentially expressed genes suggest substantial similarities in the transcriptomic response of ICM and NICM. Transcriptional differences between the two etiologies do exist, but these differences seem to be rather small when the analysis is based on the preassigned clinical categories. Thus prospective studies are required to identify specific subgroups of patients and test whether distinct transcriptomic patterns and related biological processes carry a different prognosis and require individualized therapies (i.e., platelet inhibition for NICM patients with an ischemic gene expression profile).


   GRANTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
REFERENCES
 
This study was supported by a grant of the German Federal Ministry for Education and Research supporting the German National Genome Research Network [BMBF Grant 01GS0109 (G. Steinbeck, M. Nabauer); NGFN Grant 01GR0101 (R. Kuner, A. Buneß, H. Sültmann); NGFN Grant 01GR0459 (M. Ruschhaupt)] and a grant from the Deutsche Forschungsgemeinschaft [DFG; Grant BA 3341/1-1 (A. S. Barth)].


   ACKNOWLEDGMENTS
 
The authors are thankful to Jürgen Eils and Chris Lawerenz from the Division of Theoretical Bioinformatics in the German Cancer Research Centre (DKFZ) regarding their support for iCHIP (Integration Centre of High Throughput Experiments; www.ichip.de).


   FOOTNOTES
 
Address for reprint requests and other correspondence: A. S. Barth, Dept. of Cardiology, Johns Hopkins Univ., Ross 844, 720 Rutland Ave., Baltimore, MD, 21205 (e-mail: abarth3{at}jhmi.edu); R. Kuner, Division of Molecular Genome Analysis, German Cancer Research Center, Im Neuenheimer Feld 580, 69120 Heidelberg, Germany (e-mail: r.kuner{at}dkfz.de).

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

* R. Kuner, A. S. Barth, H. Sültmann, and M. Nabauer contributed equally to this work. Back

1 The online version of this article contains supplemental material. Back


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 

  1. Adams KF Jr, Gheorghiade M, Uretsky BF, Young JB, Ahmed S, Tomasko L, Packer M. Patients with mild heart failure worsen during withdrawal from digoxin therapy. J Am Coll Cardiol 30: 42–48, 1997.[Abstract]
  2. Alla F, Briancon S, Juilliere Y, Mertes PM, Villemot JP, Zannad F. Differential clinical prognostic classifications in dilated and ischemic advanced heart failure: the EPICAL study. Am Heart J 139: 895–904, 2000.[Web of Science][Medline]
  3. Al-Shahrour F, Minguez P, Tarraga J, Medina I, Alloza E, Montaner D, Dopazo J. FatiGO +: a functional profiling tool for genomic data. Integration of functional annotation, regulatory motifs and interaction data with microarray experiments. Nucleic Acids Res 35: W91–W96, 2000.[CrossRef]
  4. Barth AS, Kuner R, Buness A, Ruschhaupt M, Merk S, Zwermann L, Kaab S, Kreuzer E, Steinbeck G, Mansmann U, Poustka A, Nabauer M, Sultmann H. Identification of a common gene expression signature in dilated cardiomyopathy across independent microarray studies. J Am Coll Cardiol 48: 1610–1617, 2006.[Abstract/Free Full Text]
  5. Barton PJ, Birks EJ, Felkin LE, Cullen ME, Koban MU, Yacoub MH. Increased expression of extracellular matrix regulators TIMP1 and MMP1 in deteriorating heart failure. J Heart Lung Transplant 22: 738–744, 2003.[CrossRef][Web of Science][Medline]
  6. Barton PJ, Felkin LE, Birks EJ, Cullen ME, Banner NR, Grindle S, Hall JL, Miller LW, Yacoub MH. Myocardial insulin-like growth factor-I gene expression during recovery from heart failure after combined left ventricular assist device and clenbuterol therapy. Circulation 112: I46–I50, 2005.[Web of Science][Medline]
  7. Blaxall BC, Tschannen-Moran BM, Milano CA, Koch WJ. Differential gene expression and genomic patient stratification following left ventricular assist device support. J Am Coll Cardiol 41: 1096–1106, 2003.[Abstract/Free Full Text]
  8. Boheler KR, Volkova M, Morrell C, Garg R, Zhu Y, Margulies K, Seymour AM, Lakatta EG. Sex- and age-dependent human transcriptome variability: implications for chronic heart failure. Proc Natl Acad Sci USA 100: 2754–2759, 2003.[Abstract/Free Full Text]
  9. Breiman L. Random Forest. Machine Learning 45: 5–32, 2001.[CrossRef][Web of Science]
  10. Buness A, Huber W, Steiner K, Sultmann H, Poustka A. arrayMagic: two-colour cDNA microarray quality control and preprocessing. Bioinformatics 21: 554–556, 2005.[Abstract/Free Full Text]
  11. Charo IF, Taubman MB. Chemokines in the pathogenesis of vascular disease. Circ Res 95: 858–866, 2004.[Abstract/Free Full Text]
  12. Chen MM, Ashley EA, Deng DX, Tsalenko A, Deng A, Tabibiazar R, Ben-Dor A, Fenster B, Yang E, King JY, Fowler M, Robbins R, Johnson FL, Bruhn L, McDonagh T, Dargie H, Yakhini Z, Tsao PS, Quertermous T. Novel role for the potent endogenous inotrope apelin in human cardiac dysfunction. Circulation 108: 1432–1439, 2003.[Abstract/Free Full Text]
  13. Chien KR. Genomic circuits and the integrative biology of cardiac diseases. Nature 407: 227–232, 2000.[CrossRef][Medline]
  14. Dec GW, Fuster V. Idiopathic dilated cardiomyopathy. N Engl J Med 331: 1564–1575, 1994.[Free Full Text]
  15. Deindl E, Schaper W. Gene expression after short periods of coronary occlusion. Mol Cell Biochem 186: 43–51, 1998.[CrossRef][Web of Science][Medline]
  16. Depre C, Kim SJ, John AS, Huang Y, Rimoldi OE, Pepper JR, Dreyfus GD, Gaussin V, Pennell DJ, Vatner DE, Camici PG, Vatner SF. Program of cell survival underlying human and experimental hibernating myocardium. Circ Res 95: 433–440, 2004.[Abstract/Free Full Text]
  17. Depre C, Tomlinson JE, Kudej RK, Gaussin V, Thompson E, Kim SJ, Vatner DE, Topper JN, Vatner SF. Gene program for cardiac cell survival induced by transient ischemia in conscious pigs. Proc Natl Acad Sci USA 98: 9336–9341, 2001.[Abstract/Free Full Text]
  18. Follath F, Cleland JG, Klein W, Murphy R. Etiology and response to drug treatment in heart failure. J Am Coll Cardiol 32: 1167–1172, 1998.[Abstract/Free Full Text]
  19. Hannenhalli S, Putt ME, Gilmore JM, Wang J, Parmacek MS, Epstein JA, Morrisey EE, Margulies KB, Cappola TP. Transcriptional genomics associates FOX transcription factors with human heart failure. Circulation 114: 1269–1276, 2006.[Abstract/Free Full Text]
  20. Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M. Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 18, Suppl 1: S96–S104, 2002.[Abstract]
  21. Hwang JJ, Allen PD, Tseng GC, Lam CW, Fananapazir L, Dzau VJ, Liew CC. Microarray gene expression profiles in dilated and hypertrophic cardiomyopathic end-stage heart failure. Physiol Genomics 10: 31–44, 2002.[Abstract/Free Full Text]
  22. Kittleson MM, Ye SQ, Irizarry RA, Minhas KM, Edness G, Conte JV, Parmigiani G, Miller LW, Chen Y, Hall JL, Garcia JG, Hare JM. Identification of a gene expression profile that differentiates between ischemic and nonischemic cardiomyopathy. Circulation 110: 3444–3451, 2004.[Abstract/Free Full Text]
  23. Mann DL, Bristow MR. Mechanisms and models in heart failure: the biomechanical model and beyond. Circulation 111: 2837–2849, 2005.[Free Full Text]
  24. McCrohon JA, Moon JC, Prasad SK, McKenna WJ, Lorenz CH, Coats AJ, Pennell DJ. Differentiation of heart failure related to dilated cardiomyopathy and coronary artery disease using gadolinium-enhanced cardiovascular magnetic resonance. Circulation 108: 54–59, 2003.[Abstract/Free Full Text]
  25. Mobini R, Maschke H, Waagstein F. New insights into the pathogenesis of dilated cardiomyopathy: possible underlying autoimmune mechanisms and therapy. Autoimmun Rev 3: 277–284, 2004.[CrossRef][Web of Science][Medline]
  26. Neglia D, L'Abbate A. Coronary microvascular dysfunction and idiopathic dilated cardiomyopathy. Pharmacol Rep 57, Suppl: 151–155, 2005.[Web of Science][Medline]
  27. Plumier JC, Robertson HA, Currie RW. Differential accumulation of mRNA for immediate early genes and heat shock genes in heart after ischaemic injury. J Mol Cell Cardiol 28: 1251–1260, 1996.[CrossRef][Web of Science][Medline]
  28. Ruschhaupt M, Huber W, Poustka A, Mansmann U. A compendium to ensure computational reproducibility in high-dimensional classification tasks. Stat Appl Genet Mol Biol 3: Article 37, 2004.
  29. Schneider J, Ruschhaupt M, Buness A, Asslaber M, Regitnig P, Zatloukal K, Schippinger W, Ploner F, Poustka A, Sültmann H. Identification and meta-analysis of a small gene expression signature for the diagnosis of estrogen receptor status in invasive ductal breast cancer. Int J Cancer 119: 2974–2979, 2006.[CrossRef][Web of Science][Medline]
  30. Singh SN, Fletcher RD, Fisher SG, Singh BN, Lewis HD, Deedwania PC, Massie BM, Colling C, Lazzeri D. Amiodarone in patients with congestive heart failure and asymptomatic ventricular arrhythmia. Survival Trial of Antiarrhythmic Therapy in Congestive Heart Failure. N Engl J Med 333: 77–82, 1995.[Abstract/Free Full Text]
  31. Steenman M, Lamirault G, Le Meur N, Le Cunff M, Escande D, Leger JJ. Distinct molecular portraits of human failing hearts identified by dedicated cDNA microarrays. Eur J Heart Fail 7: 157–165, 2005.[Abstract/Free Full Text]
  32. Tan AC, Naiman DQ, Xu L, Winslow RL, Geman D. Simple decision rules for classifying human cancers from gene expression profiles. Bioinformatics 21: 3896–3904, 2005.[Abstract/Free Full Text]
  33. Tan FL, Moravec CS, Li J, Apperson-Hansen C, McCarthy PM, Young JB, Bond M. The gene expression fingerprint of human heart failure. Proc Natl Acad Sci USA 99: 11387–11392, 2002.[Abstract/Free Full Text]
  34. Tibshirani R, Hastie T, Narasimhan B, Chu G. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 99: 6567–6572, 2002.[Abstract/Free Full Text]
  35. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 98: 5116–5121, 2001.[Abstract/Free Full Text]
  36. Uretsky BF, Thygesen K, Armstrong PW, Cleland JG, Horowitz JD, Massie BM, Packer M, Poole-Wilson PA, Ryden L. Acute coronary findings at autopsy in heart failure patients with sudden death: results from the assessment of treatment with lisinopril and survival (ATLAS) trial. Circulation 102: 611–616, 2000.[Abstract/Free Full Text]
  37. Warnat P, Eils R, Brors B. Cross-platform analysis of cancer microarray data improves gene expression based classification of phenotypes. BMC Bioinformatics 6: 265, 2005.[CrossRef][Medline]
  38. Wittchen F, Suckau L, Witt H, Skurk C, Lassner D, Fechner H, Sipo I, Ungethum U, Ruiz P, Pauschinger M, Tschope C, Rauch U, Kuhl U, Schultheiss HP, Poller W. Genomic expression profiling of human inflammatory cardiomyopathy (DCMi) suggests novel therapeutic targets. J Mol Med 85: 253–267, 2007.[Web of Science]
  39. Yang Z, McMahon CJ, Smith LR, Bersola J, Adesina AM, Breinholt JP, Kearney DL, Dreyer WJ, Denfield SW, Price JF, Grenier M, Kertesz NJ, Clunie SK, Fernbach SD, Southern JF, Berger S, Towbin JA, Bowles KR, Bowles NE. Danon disease as an underrecognized cause of hypertrophic cardiomyopathy in children. Circulation 112: 1612–1617, 2005.[Abstract/Free Full Text]
  40. Yousef ZR, Redwood SR, Bucknall CA, Sulke AN, Marber MS. Late intervention after anterior myocardial infarction: effects on left ventricular size, function, quality of life, and exercise tolerance: results of the Open Artery Trial (TOAT Study). J Am Coll Cardiol 40: 869–876, 2002.[Abstract/Free Full Text]
  41. Zhu J, Hastie T. Classification of gene microarrays by penalized logistic regression. Biostatistics 5: 427–443, 2004.[Abstract]



This article has been cited by other articles:


Home page
 CirculationHome page
R. Sarwar and S. A. Cook
Genomic Analysis of Left Ventricular Remodeling
Circulation, August 4, 2009; 120(5): 437 - 444.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Figures and Tables
Right arrow All Versions of this Article:
34/1/88    most recent
00299.2007v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kuner, R.
Right arrow Articles by Nabauer, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Kuner, R.
Right arrow Articles by Nabauer, M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Visit Other APS Journals Online
Copyright © 2008 by the American Physiological Society.