| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Molecular Medicine |
From the Donald W. Reynolds Cardiovascular Clinical Research Center (R.T., R.A.W., T.Q.), Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, Calif; and GeneData (USA), Inc (A.L.), South San Francisco, Calif.
Correspondence to Thomas Quertermous, MD, Division of Cardiovascular Medicine Stanford University School of Medicine 300 Pasteur Dr, Falk CVRC Stanford, CA 94305. E-mail tomq1{at}stanford.edu
| Abstract |
|---|
|
|
|---|
Key Words: gene expression cardiac chambers microarrays regulatory networks development
| Introduction |
|---|
|
|
|---|
A number of methods have been developed and used over the past decade to characterize differences in gene expression pattern between various tissues or tissues treated differently. However, the recent utilization of microarray methodology has allowed the development of large amounts of informative data, via a high throughput analysis of gene expression termed transcriptional profiling. Transcriptional profiling with microarrays offers simultaneous expression analysis of thousands of genes, revealing unique biological insights through patterns of expression and suggesting functions of unknown genes. Microarrays have been applied to investigate basic issues in cardiovascular biology, including the transcriptional response to experimental myocardial ischemia12 and gene expression patterns in human heart failure.13 However, the value of such studies are limited due to a lack of understanding regarding basic cell signaling pathways that mediate such fundamental processes as myocardial cell differentiation, and the genetic basis of cellular and tissue-specific gene expression.
Utilizing a comprehensive mouse cDNA microarray containing 42 300 features representing over 25 000 unique genes and ESTs, we have performed microarray analysis of gene expression patterns in the four cardiac chambers and the interventricular septum. Using rigorous statistical tools for analysis of the microarray data, we have identified genes that are expressed in the heart overall, as well as genes with expression limited to a subset of chambers. We have been able to fit expression profiles of signaling molecules into existing models of molecular regulatory networks for the various regions of the heart, assigning new roles for known genes and suggesting roles for unknown genes. A searchable internet database, representing a comprehensive "atlas" of gene expression in the normal heart, has been constructed to allow other investigators to study the chamber-specific patterns of expression of genes of interest. Further study of genes in this database is likely to provide important new insights into cardiovascular genes involved in the development, function, and pathophysiology of the cardiovascular system.
| Materials and Methods |
|---|
|
|
|---|
25 000 unique genes and ESTs. It was constructed by combining the National Institutes of Aging (NIA) clone set (
15 000 clones) (http://lgsun.grc.nia.nih.gov/cDNA/ cDNA.html) and the RIKEN clone set (
21 000 clones) (http://fantom.gsc.riken.go.jp/doc/introduction.html#fantom), as well as
5000 genes that were donated by various investigators. A continuously updated and annotated list of the cDNAs included on this array is available at the Stanford Microarray Database (http://smd.stanford.edu).
RNA Preparation and Hybridization to Microarray
Fifteen 8-week-old C57Bl/6 female mice (http://labanimals. stanford.edu/index.html) were anesthetized with Avertin and perfused with normal saline after left ventricular puncture until liver blanching was noted. The four cardiac chambers and the interventricular septum of the mice were carefully dissected under a dissecting microscope, flash frozen in liquid nitrogen, and divided into three pools for further RNA isolation (5 mice per pool). The chambers were separated by first removing LA and RA from the base of the heart. The RV was then dissected off the LV/Septum by inserting iris scissors into the tricuspid valve opening and cutting around the interface of the RV and septum, leaving a
1.0- to 1.5-mm rim of RV tissue at the margins. The LV free wall was then dissected away from the septum by inserting iris scissors into the mitral valve opening and cutting around the interface of the LV and septum leaving a
1.0- to 1.5-mm rim of LV tissue at the margins. An oblong portion of septum was then dissected free of remaining ventricular tissues, with the interface regions discarded. Valvular apparatus and outflow tract were excluded from each sample. Total RNA was isolated using a modified two-step purification protocol employing homogenization (PRO250 Homogenizer, 10-mmx105-mm generator, PRO Scientific IN) in Trizol (Invitrogen) followed by purification over a Qiagen RNeasy column (Qiagen). First strand cDNA was synthesized from 15 µg of total RNA from each pool and from whole e17.5-day embryo for reference RNA in the presence of Cy3 or Cy5 dUTP, respectively, and hybridized to the microarray generating three biological replicates for each cardiac chamber (details provided at http://mousedevelopment.org/).
Quantitative Real-Time Reverse TranscriptasePolymerase Chain Reaction
Primers and probes for 10 representative differentially expressed genes were obtained from Applied Biosystems Assays-on-Demand. cDNA was synthesized from 5 µg of total RNA using MMLV reverse transcriptase (SuperScript II kit, Invitrogen). Amplification was performed in triplicate at 50°C for 2 minutes and 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Reactions without template and/or enzyme were used as negative controls. 18S ribosomal RNA was used as an internal control. A standard curve derived from e17.5 day mouse embryonic RNA was plotted for each target gene by linear regression using SPSS version 11.0 software (Applied Biosystems). RNA quantity was expressed relative to the corresponding 18S control. Fold differences were calculated by dividing the combined RA and LA by the combined LV and RV results, and plotted on a log10 scale. Primers and probes used are listed in the online data supplement available at http://www.circresaha.org.
Data Acquisition, Analysis, and Statistical Analysis
Image acquisition of the mouse cDNA microarrays was performed on an Agilent G2565AA Microarray Scanner System. Feature Extraction was performed with GenePix 4.0 software (Axon, Inc). Numerical raw data were migrated from GenePix, without processing, into an Oracle relational database (CoBi) that has been designed specifically for microarray data analysis (GeneData, Inc, USA). The data were then analyzed using Expressionist software (http://www. genedata.com/products/expressionist/). After background subtraction (by calculating median "local" background of neighboring 121 features) and dye bias normalization,14 poor quality features were excluded from further analysis. Features with low signal intensity in the reference channel were filtered if percentage coverage area was less than 40%, and if signal-to-noise ratio was less than 2.5. To obtain a manageable sized dataset for further analysis, a number of filtering algorithms were also utilized including filtering by valid-value-proportion. Features with valid values in at least 50% of the experiments and present in at least one of the replicates were retained for further analysis. For further statistical analysis of the data, a K-nearest-neighbor (KNN) algorithm was applied to impute for missing values.15 To identify genes with an expression level statistically different in at least one of the experimental groups, we used ANOVA (Figure 1). For two-group comparisons, we used t tests to generate a list of genes that differentiated between the experimental groups with high statistical significance (Figure 1). As an additional stringent statistical tool to validate the gene list generated by the ANOVA and two-groups t tests, we also used the significance analysis of microarrays (SAM) algorithm.16 This method uses replicate experiments to develop a measure of variance that is used to test whether observed differences in gene expression, in two cell-type partitions, are likely to be real (http://www-stat. stanford.edu/tibs/SAM/). For hierarchical clustering of the experiments, we used positive correlation for distance determination and required complete linkage, which uses the greatest distance between genes in two clusters to ascribe similarity. For 2-dimensional hierarchical clustering, we applied similar settings for clustering of the gene groups.
|
|
Heart Chamber Gene Expression Database
A database containing normalized relative gene expression values is available for the hybridizations presented in this article and can be searched interactively from a web browser (http://mousedevelopment. org). Expression patterns for any gene present on the array can be accessed by searching by gene name, accession number, or sequence BLAST.
| Results |
|---|
|
|
|---|
90% identity with the list generated by ANOVA, suggesting that for this dataset SAM was not greatly different from high stringency t tests. Hierarchical clustering of the experiments using the above list of genes revealed the expected pattern where each experiment clustered first with its biological replicate before clustering with the other cardiac chambers (Figure 2). Coclustering of particular chambers serves as a quality control for the entire study showing the high degree of correlation among the replicate experiments (Figure 2). Moreover, the pattern of clustering is consistent with underlying cardiac function, where the 2 atria, and the 2 ventricles, have similar patterns of gene expression (Figure 2). The left and right atria are shown to cluster together first, before clustering with the ventricles and the IVS. As would be expected, the degree of correlation of gene expression, demonstrated by the hierarchical clustering, is higher between the IVS and the left ventricle than between the two ventricles.
|
Chamber-Specific Gene Expression
To identify genes primarily expressed in a particular chamber, we used the large dataset consisting of 23 607 features to perform t test comparisons between each chamber (3 replicates) and all other chambers combined (12 experiments). We repeated this procedure for each cardiac chamber and the interventricular septum (IVS) (Figure 1). From each of the 5 t tests performed, we chose those genes with statistically significant (P<0.01) higher expression levels in a particular chamber when compared with the other chambers (Figure 3, online Lists 1a through 1e). Interestingly, the IVS was found to share several genes with the left ventricle (LV) as well as the right ventricle (RV), but a number of genes were found to be uniquely expressed in the IVS alone (Figure 3).
Atrial and Ventricular Gene Expression
Our analyses demonstrated a very different pattern of gene expression between the atria and the ventricles. SAM analysis at a false detection rate of 2% identified 2460 genes with higher expression in the atria and 2970 genes with higher expression in the ventricles (Figure 4, online Lists 2a and 2b). Among the 2460 genes more highly expressed in the atria were those encoding transcription factors such as LIM proteins, cysteine rich proteins, and dickkopf homolog 3 (Dkk3); cytoskeletal genes such as myosin light chain, alkali, cardiac atria (Myla, MLC 1a), myosin light chain, regulatory A (MLC 2A) (see online List 8), actin related protein 2/3 complex; extracellular matrix molecules such as VCAM and EGF containing fibulin; growth related proteins such as cyclin I, growth arrest specific-1 and -6; metabolism-related molecules such as insulin-like growth factor binding protein-6; and several ESTs and uncharacterized genes. Genes with higher expression in the ventricles represented similar general categories of gene families. In comparison with atria, the ventricles had higher expression of the ventricular myosins Myosin light chain-1V (MLC 1v) and myosin light chain, phosphorylatable, cardiac ventricles (MLC 2V) (see online List 8) and several tropomyosins, as well as isoforms of actins and myosins, including cardiac
-actin, actinin
2, and myosin Ib and X. The higher relative expression of several of these myocyte-specific genes may be due to the higher myocyte contribution to tissue mass in the ventricular samples versus the atrial tissue. Several transcription factors were found to have higher expression in ventricles. These included LIM domain proteins such as Elfin, four and half LIM domain 2, and homeobox genes such as Iroquois-related homeobox-3 and -4. Expression patterns of other cardiac-specific genes such as those coding proteins important in calcium metabolism (sarcoplasmic reticulum calcium ATPase (SERCA) family), sodium and potassium handling (NaK-ATPase family), energy metabolism proteins, and several matrix proteins were also noted to be chamber specific.
|
Left and Right Heart Gene Expression
Direct comparison between the LV and the RV using SAM analysis revealed a differential pattern of expression for several important gene groups (Figure 4, online Lists 2c and 2d). As expected, the number of genes that differentiated between the two ventricles was much smaller than those differentiating between the atria and the ventricles. Direct comparison between the two atria also revealed a differential pattern of expression of several important gene groups (data not shown).
Chamber-Specific Distribution of Biologically Relevant Genes
The above analyses revealed differential expression of gene families that have been linked to critical aspects of myocardial development and function. Using gene ontology annotation, members of these gene families, as well as other families of biological interest such as transcription factors, homeobox genes, cytoskeletal proteins, and developmentally related genes, were compiled for a closer examination of their chamber-specific expression (see online data supplement). Similar analytical algorithms were applied to this smaller subset of genes as to the larger dataset. Many of these genes revealed statistically significant chamber-specific expression (online List 3). Two-dimensional hierarchical clustering with these genes demonstrated that they were able to accurately classify tissue samples regarding their chamber of origin (Figure 5). Direct comparisons between the atria and ventricles and LV versus RV using SAM revealed differential patterns of expression for members of these interesting protein families (online Lists 4 and 5). Some of the most interesting findings were the expression patterns of transcription factors. Genes with higher expression in the ventricle relative to atria included Irx3, Irx4, Nkx2.5, MAD homolog 3, and the paired-related homeobox-1. Important genes with higher relative expression in the atria include T-box 5 and Dkk3. Using the data obtained from this analysis, we have been able to fit specific expression profiles of known signaling molecules into existing models of molecular regulatory networks that maintain chamber-specific features in the adult heart (Figure 7).
|
|
Quantitative Real-Time RT-PCR Confirms the Accuracy of Microarray Hybridization Results
Differential expression of 10 representative genes from a number of pathways was confirmed by qRT-PCR. The genes assayed included Myla (myosin light chain, alkali, cardiac atria), sarcolipin, Dkk3, Sdccag28 (serologically defined-colon-CA-antigen-28), Pdlim3 (PDZ and LIM-domains-3), Ednrb (endothelin receptor b), Lmcd1 (LIM and cysteine-rich domains 1), robo4 (roundabout homolog 4), Nd-1 (Ivns1abp, Unigene Mm.33764), and CD36 antigen. Overall, there was good correlation between the two methods, with the qRT-PCR data showing greater measured differences in most cases (Figure 6).
|
Chamber Classification by ESTs
To identify genes with an expression level statistically different in at least one of the experimental groups, we used ANOVA (P<0.01) and multiclass SAM analysis. The resultant list of
4200 genes included approximately 900 ESTs and uncharacterized features. Hierarchical clustering of the experiments using these ESTs revealed an accurate clustering of the cardiac chambers (data not shown). This finding suggests that uncharacterized genes included on the array have chamber-specific gene expression, which may warrant further detailed analysis.
| Discussion |
|---|
|
|
|---|
Embryonic heart development is characterized by morphogenetic events that are orchestrated by complex patterns of gene expression.10,1720 Patterned differences in gene expression, regulated by known developmental transcription factor families, are important for cell movement and response to local inductive signals. In general, such developmental pathways are not commonly considered in the context of adult tissue function, although reactivation of embryonic signaling is often linked to pathophysiology. However, a striking finding of these studies has been the documentation of sustained expression of critical developmental factors in the adult cardiac tissues. Certainly, different cardiac chambers are exposed to different hemodynamic forces, and required to adopt their specialized functions early in development. Thus, transcription factors that regulate chamber-specific developmental events may also determine transcription of molecules that define unique aspects of the contractile and cytoskeletal apparatus, energy management, ion-channel function, and electrophysiological properties. By combining information regarding known signaling pathways that are used in cardiac development with the information provided through these studies documenting chamber-specific differential gene expression, it is possible to propose a paradigm of transcriptional regulatory networks that maintain chamber-specific features in the adult heart (Figure 7).
Data presented in this study and in Figure 7 suggests important roles for wnt signaling pathways in cardiac chamber maintenance or function. Wnt pathways are among the most important determinants of embryonic patterning, and the Drosophila prototype gene Wingless (Wg) is critical for heart formation.2124 Coordinated expression of the mammalian wnt antagonists, Crescent and dickkopf-1 (Dkk1), in the cardiac crescent overcomes repressive wnt signaling and allows commitment of cells to the cardiac fate.25,26 Mice that lack the disheveled-2 gene, a downstream signaling molecule in the wnt pathway, develop with outflow tract abnormalities.27,28 Any ongoing role of wnt signaling in the maintenance of cell identity in adult heart tissues has not been clearly described.
Dkk3, which also acts as a potent inhibitor of wnt signaling, is expressed in cells of the bulbis cordis and sinus venosus in the 9-day-old embryo. However, its expression becomes restricted to the atria and endocardial cushions by embryonic day 12.5 (as is Dkk1), suggesting that it plays a role in the establishment of atrial identity.29 Our findings show that Dkk3 maintains this highly localized expression in the adult atrium, with little expression in the ventricle, suggesting that it may also be important for the maintenance of atrial identity by antagonizing wnt signaling.9 In keeping with this possibility is the observation that dishevelled-1 (Dvl1) is significantly more highly expressed in the ventricles than in the atria. An appealing hypothesis is suggested, such that wnt signaling is opposed in the atria by Dkk3 and continues unopposed in the ventricles as evidenced by high levels of Dvl1 mRNA, resulting in wnt pathway input into the maintenance of phenotype in the adult cardiac chambers. Wnt mediators often operate in combination with other pathways such as LIM proteins. Recent studies have shown the four and a half LIM-only protein 2 (FHL2) is a novel ß-catenininteracting protein and coactivator.30 Our results demonstrate that FHL2 has higher expression in the ventricles.
A number of LIM domain proteins were identified in these studies and likely serve critical overlapping functions in different chambers (Figure 7). Many known LIM proteins have proven roles in differentiation, function, and the maintenance of phenotype of portions of the cardiovascular system. These include Cipher31 muscle LIM protein (MLP)7,32 and CRP1 and 2, among others.33 In our study, LIM proteins were found to have unique expression patterns among the cardiac chambers (Figure 5, online List 3). Consistent with prior studies, our results reveal that FHL2 and Elfin are highly expressed in the ventricles.9,34,35 Perhaps more interestingly, these studies show that thymus LIM protein is expressed in the heart with a ventricular-restricted pattern. Cardiac expression of this LIM gene family member has not been previously reported. We have identified four other poorly characterized LIM domain proteins with restricted expression patterns. Further study of these members of this important gene family is likely to provide significant insights into developmental and functional features of the ventricle.
We have identified hundreds of other named genes and uncharacterized ESTs, which have restricted expression patterns within the heart. Many of these ESTs are assigned to Unigene clusters with limited sequences, so it is not possible to assign them to structural protein families. Further studies of these genes will likely lead to identification of additional new gene families and pathways involved in the development, function, and pathophysiology of the cardiovascular system.
Previous studies using microarrays and transcriptional profiling have investigated gene expression differences between the different cardiac chambers.9,11,36 We have compared our data with that obtained by these other groups, and found that genes identified in their studies as having large expression differences between the heart chambers were also identified in the analysis presented here (online Lists 8 and 9). However, in looking at the overall data, there are significant differences in gene expression patterns among these different studies, due to differences in experimental design, methodology, and data analysis.
By using microarrays to analyze gene expression patterns in the heart on a genome wide scale, we have identified large numbers of genes, which are differentially regulated and likely play important roles in the development, function, and pathophysiology of the cardiovascular system. Although definitive experiments are required to fully characterize the function of such candidate genes, grouping of genes on the basis of their coordinated expression allows the development of hypotheses regarding signaling functions. This study provides a comprehensive gene expression atlas of the mammalian heart, establishing a strong foundation for further study of cardiac development and disease states.
| Acknowledgments |
|---|
| Footnotes |
|---|
Original received July 1, 2003; revision received September 17, 2003; accepted October 15, 2003.
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
K. Gambetta, M. K. Al-Ahdab, M. N. Ilbawi, N. Hassaniya, and M. Gupta Transcription repression and blocks in cell cycle progression in hypoplastic left heart syndrome Am J Physiol Heart Circ Physiol, May 1, 2008; 294(5): H2268 - H2275. [Abstract] [Full Text] [PDF] |
||||
![]() |
B. Burstein, E. Libby, A. Calderone, and S. Nattel Differential Behaviors of Atrial Versus Ventricular Fibroblasts: A Potential Role for Platelet-Derived Growth Factor in Atrial-Ventricular Remodeling Differences Circulation, April 1, 2008; 117(13): 1630 - 1641. [Abstract] [Full Text] [PDF] |
||||
![]() |
P. Philip-Couderc, N. I. Tavares, A. Roatti, R. Lerch, C. Montessuit, and A. J. Baertschi Forkhead Transcription Factors Coordinate Expression of Myocardial KATP Channel Subunits and Energy Metabolism Circ. Res., February 1, 2008; 102(2): e20 - e35. [Abstract] [Full Text] [PDF] |
||||
![]() |
A. S. Barth, R. Kuner, A. Buness, M. Ruschhaupt, S. Merk, L. Zwermann, S. Kaab, E. Kreuzer, G. Steinbeck, U. Mansmann, et al. Identification of a Common Gene Expression Signature in Dilated Cardiomyopathy Across Independent Microarray Studies J. Am. Coll. Cardiol., October 17, 2006; 48(8): 1610 - 1617. [Abstract] [Full Text] [PDF] |
||||
![]() |
I. KEHAT, T. HASIN, and A. ARONHEIM The Role of Basic Leucine Zipper Protein-Mediated Transcription in Physiological and Pathological Myocardial Hypertrophy Ann. N.Y. Acad. Sci., October 1, 2006; 1080(1): 97 - 109. [Abstract] [Full Text] [PDF] |
||||
![]() |
R. Tabibiazar, R. A. Wagner, A. Deng, P. S. Tsao, and T. Quertermous Proteomic profiles of serum inflammatory markers accurately predict atherosclerosis in mice Physiol Genomics, April 13, 2006; 25(2): 194 - 202. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. C. Wu, J. M. Spin, F. Cao, S. Lin, X. Xie, O. Gheysens, I. Y. Chen, A. Y. Sheikh, R. C. Robbins, A. Tsalenko, et al. Transcriptional profiling of reporter genes used for molecular imaging of embryonic stem cell transplantation Physiol Genomics, March 13, 2006; 25(1): 29 - 38. [Abstract] [Full Text] [PDF] |
||||
![]() |
S. L. Merklinger, R. A. Wagner, E. Spiekerkoetter, A. Hinek, R. H. Knutsen, M. G. Kabir, K. Desai, S. Hacker, L. Wang, G. M. Cann, et al. Increased Fibulin-5 and Elastin in S100A4/Mts1 Mice With Pulmonary Hypertension Circ. Res., September 16, 2005; 97(6): 596 - 604. [Abstract] [Full Text] [PDF] |
||||
![]() |
R. Tabibiazar, R. A. Wagner, E. A. Ashley, J. Y. King, R. Ferrara, J. M. Spin, D. A. Sanan, B. Narasimhan, R. Tibshirani, P. S. Tsao, et al. Signature patterns of gene expression in mouse atherosclerosis and their correlation to human coronary disease Physiol Genomics, July 14, 2005; 22(2): 213 - 226. [Abstract] [Full Text] [PDF] |
||||
![]() |
P. Ellinghaus, R. J. Scheubel, D. Dobrev, U. Ravens, J. Holtz, J. Huetter, U. Nielsch, and H. Morawietz Comparing the global mRNA expression profile of human atrial and ventricular myocardium with high-density oligonucleotide arrays J. Thorac. Cardiovasc. Surg., June 1, 2005; 129(6): 1383 - 1390. [Abstract] [Full Text] [PDF] |
||||
![]() |
S.-K. Park and T. A. Prolla Gene expression profiling studies of aging in cardiac and skeletal muscles Cardiovasc Res, May 1, 2005; 66(2): 205 - 212. [Abstract] [Full Text] [PDF] |
||||
![]() |
R. Tabibiazar, R. A. Wagner, J. M. Spin, E. A. Ashley, B. Narasimhan, E. M. Rubin, B. Efron, P. S. Tsao, R. Tibshirani, and T. Quertermous Mouse Strain-Specific Differences in Vascular Wall Gene Expression and Their Relationship to Vascular Disease Arterioscler. Thromb. Vasc. Biol., February 1, 2005; 25(2): 302 - 308. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. M. Spin, S. Nallamshetty, R. Tabibiazar, E. A. Ashley, J. Y. King, M. Chen, P. S. Tsao, and T. Quertermous Transcriptional profiling of in vitro smooth muscle cell differentiation identifies specific patterns of gene and pathway activation Physiol Genomics, November 17, 2004; 19(3): 292 - 302. [Abstract] [Full Text] [PDF] |
||||
| ||||||||