Cardiac hypertrophy is a complex and nonhomogenous response to various stimuli. In this study, we used high-density oligonucleotide microarray to examine gene expression profiles during physiological hypertrophy, pathological hypertrophy, and heart failure in Dahl salt-sensitive rats. There were changes in 404/3,160 and 874/3,160 genes between physiological and pathological hypertrophy and the transition from hypertrophy to heart failure, respectively. There were increases in stress response genes (e.g., heat shock proteins) and inflammation-related genes (e.g., pancreatitis-associated protein and arachidonate 12-lipoxygenase) in pathological processes but not in physiological hypertrophy. Furthermore, atrial natriuretic factor and brain natriuretic protein showed distinctive changes that are very specific to different conditions. In addition, we used a resampling-based gene score-calculating method to define significantly altered gene clusters, based on Gene Ontology classification. It revealed significant alterations in genes involved in the apoptosis pathway during pathological hypertrophy, suggesting that the apoptosis pathway may play a role during the transition to heart failure. In addition, there were significant changes in glucose/insulin signaling, protein biosynthesis, and epidermal growth factor signaling during physiological hypertrophy but not during pathological hypertrophy.

  • atrial natriuretic factor
  • brain natriuretic protein
  • insulin
  • apoptosis
  • microarray

cardiac hypertrophy is a complex response to various hypertrophic signals that promote the growth of cardiac myocytes. Multiple neurohumoral, hormonal, and mechanistic stimuli have been implicated in, and numerous interdependent pathways and molecules shown to be associated with, cardiac hypertrophy (10, 37). Accordingly, the responses of cardiac myocytes to different hypertrophic stimuli are not homogenous. Examples of stimulus-dependent hypertrophic responses can be found in both physiological and pathological hypertrophy. Exercise-induced cardiac hypertrophy is a good example of physiological hypertrophy, which is a favorable adaptive response of the heart to increases in bodily demand (8). In comparison, pathological hypertrophy is a maladaptive response to pathological stimuli, such as pressure or volume overload. These differences are particularly evident clinically, for pathological hypertrophy often progresses to heart failure, especially when pathological stimuli are persistent, whereas physiological hypertrophy usually does not. These examples support the notion that cardiac hypertrophies are not all the same, and suggest that stimulus-specific hypertrophic responses may be associated with distinct molecular changes (20, 27).

Previous studies of models of cardiac hypertrophy, using microarray technology, have yielded interesting but varied results (1, 4, 13, 16, 19, 23, 31, 32, 34, 50, 54). Studies of putative physiological stimuli showed increased expression of genes involved in the cell cycle, cell structure, intracellular signaling, protein synthesis, and metabolism (13, 16, 31, 34). Pathological stimuli, however, such as in the experimental animal myocardial infarction models, are associated with the increased expression of genes involved in inflammation, wound healing, structural proteins, metabolism, and survival (23, 32, 50). However, different transgenic mouse models of cardiac hypertrophy revealed no consistent change in any single gene, although there was a correlation between the number of differentially expressed genes and the degree of hypertrophy (1).

In this study, we used rat models of physiological hypertrophy and pathological hypertrophy that progress to heart failure in the Dahl salt-sensitive (DS) rat (25). Using microarray analysis, we examined differences in gene expression in physiological and pathological hypertrophy, and how the pattern of gene expression changes during the transition from hypertrophy to heart failure. Gene symbols are listed in Table 1.

View this table:
Table 1.



Exercise- and high-salt diet-induced cardiac hypertrophy.

All animal studies were conducted in accordance with the standards set forth in the Guide for the Care and Use of Laboratory Animals and were approved by our Institutional Animal Care and Use Committee. Physiological and pathological rat models of cardiac hypertrophy and heart failure were created as previously described (25). Briefly, physiological cardiac hypertrophy was generated by a vigorous daily exercise regimen for 6 wk. Pathological cardiac hypertrophy was generated by feeding a 6% NaCl diet for 6 wk. With a continued high-salt diet, these rats developed clinical heart failure after 15–18 wk. Control rats were sedentary, and age-matched DS rats were fed normal rat chow.


Rats were anesthetized with an intraperitoneal injection of ketamine-HCl (50 mg/kg) and xylazine (10 mg/kg). Echocardiography was performed on rats in the prone decubitus position with a Hewlett-Packard Sonos 1500 sector scanner equipped with a 7.5-MHz phased-array transducer. Two-dimensionally guided M-mode tracings were recorded on strip-chart paper at a paper speed of 100 mm/s. Anterior and posterior wall thickness and left ventricle internal dimensions were measured according to the leading-edge method of the American Society of Echocardiography.

RNA isolation and Northern analysis.

Total RNA from ventricles was isolated as previously described (34). Briefly, the hearts were rapidly removed, and the atria and right ventricles were cut and snap frozen in liquid nitrogen. Total RNA was isolated using the TRIzol Reagent (Life Technologies) following the instructions of the manufacturer. Isolated RNA was further purified by a phenol-chloroform extraction, followed by ethanol precipitation. Northern analysis was performed according to the published protocol (48). The probes for mouse atrial natriuretic factor (ANF) and mouse brain natriuretic protein (BNP) and rat GAPDH probes were prepared as previously described (34).

Microarray analysis and statistics.

This study was conducted in collaboration with the National Heart, Lung, and Blood Institute-sponsored Program in Genomic Application (PGA) and was performed with strict quality assurance as set by the PGA. The details of the methods can be found at the CardioGenomics website (http://www.cardiogenomics.org). Generation of a biotinylated cRNA probe for the Affymetrix arrays was performed according to Affymetrix’s protocols. Affymetrix oligonucleotide Rat Genome U34A GeneChips (Affymetrix, Santa Clara, CA), which contain 8,799 probe sets for known genes and expressed sequence tags (ESTs), were used in our study. Normalization was done using invariant set normalization algorithms with dChip v.1.3, and perfect-match/mismatch difference model-based expression index was calculated (30). To identify the presence of transcripts, Affymetrix Microarray Suite (MAS v.5.01) was used. All genes identified as “absent” in all 24 arrays and control probe sets were excluded. We found 3,944 genes (44.82%) that were designated absent across all 24 arrays and 59 control probe sets. Probe sets from the same UniGene clusters (build no. 131 as of April 2004) were averaged. This process reduced the number of available transcripts further to 3,160 from 8,799. The remaining 3,160 genes were then statistically filtered for detecting significantly different expressed genes across the experimental conditions.

First, statistical filtering with two-way ANOVA was done as shown in Fig. 1. To assess the profiles of gene expression changes between physiological and pathological hypertrophy, we compared three groups at two time points, i.e., 3 and 6 wk (Fig. 1A). High-salt diet models of three time points were compared with control groups of the same ages (Fig. 1B) to evaluate the transition of pathological hypertrophy to heart failure. Then, q value was used to assess the false discovery rates with two-way ANOVA P values (44, 51). Significant genes from two-way ANOVA were exported to Spotfire DecisionSite 7.3 software (Spotfire, Sommerville, MA), where K-means cluster analysis was done.

Fig. 1.

Study design. Two-way factorial experimental designs for physiological vs. pathological hypertrophy (A) and hypertrophy vs. heart failure (B). Both treatment and time were used as main effects. Cont, control group; EX, exercise group; HS, high-salt diet group. C: survival curve of Dahl salt-sensitive (DS) rats fed high-salt diet. Solid arrow, time at which high-salt diet was initiated. Open arrows, times at which samples were obtained for microarray analysis.

Second, we compared each experimental condition at different time points to the age-matched control group with Welch’s t-test. A Venn diagram was created, with genes, that showed a minimum lower confidence bound of 1.5-fold in each comparison using dChip software. The P values of statistical comparison were exported to ErmineJ software (http://microarray.genomecenter.columbia.edu/ermineJ) to identify enriched Gene Ontology (GO) classes. We used GO classes of size k = 5–250. To test the significance of each class, we drew a random set of genes of the same size from the data and calculated the raw score r = −∑jlog(pj), where pj is the P value for each gene in the random sample. This procedure was done 200,000 times to calculate the distribution of raw score r under the null hypothesis. The overall significance for a class with a raw score r is calculated as the fraction of random trials resulting in a score higher than r (45).


Animal models.

Both exercise (EX group) and high-salt diet (HS group) induced significant cardiac hypertrophy in DS rats compared with the control (Table 2). Hearts from both the EX group and the HS group showed significant increases in wall thickness by echocardiogram. The body and tissue weights of EX and HS rats showed significant increases in heart weight without changes in other tissue weights. These findings were consistent with cardiac hypertrophy with preserved cardiac function. Continuation of high-salt diet in the HS group resulted in clinical heart failure and death (Fig. 1C). These rat hearts showed significant increases in wall thickness and left ventricle chamber diameter and a decrease in fractional shortening, consistent with hypertensive dilated cardiomyopathy (Table 2). There were significant decreases in body weight due to cardiac cachexia, a nearly twofold increase in heart weight, and significant increases in the lung and liver weights, consistent with left- and right-sided heart failure, respectively.

View this table:
Table 2.

Echocardiographic findings and organ weights of different rat models

Genetic profiles of physiological and pathological cardiac hypertrophy.

Among 3,160 genes analyzed in HS and EX groups at 3 and 6 wk, 404 genes showed changes that were statistically significant at model q value <0.05. K-means clustering (k = 6) showed distinctive clustering for control, physiological hypertrophy, and pathological hypertrophy at 6 wk of stimuli (Fig. 2A). Most of the significant changes occurred after 6 wk of stimuli in HS and EX groups, suggesting that the majority of molecular changes occurred when there were significant phenotypic changes. Many of the genes were predominately altered in pathological hypertrophy but not in physiological hypertrophy (154/404; clusters 1 and 2). These included ANF, BNP, and GATA4, which are known to change during cardiac hypertrophy. Of note, these clusters include SOD2, ICAM1, HAND2, presenilin-1, and von Willebrand factor, which are known to be upregulated during oxidative stress. This finding suggests that pathological stimulus invokes the oxidative stress response, but physiological hypertrophy does not. A significant number of genes were changed in both physiological and pathological hypertrophy (159/404; clusters 5 and 6). This group of genes is comprised of known hypertrophic response proteins, such as MAP kinase 6 and 12, PECAM1, UCP2, phospholamban, Grb2, and GSK3β, suggesting that these genes are associated with both pathological and physiological hypertrophy. Only 91/404 genes were changed predominately in physiological hypertrophy (clusters 3 and 4). Genes in this group were predominately involved in metabolism and cell growth, further supporting the notion that physiological hypertrophy is associated with normal growth of the heart to keep up with the increased demand of the heart. ADRBK1 and MMP14 are in these clusters. Venn diagram analysis at 6 wk demonstrated that most of the gene changes occurred either only during physiological hypertrophy (74 genes) or only during pathological hypertrophy (120 genes), suggesting that conditions are characterized by distinct gene expressions (Fig. 2B). There were 50 genes that showed changes in both conditions.

Fig. 2.

Genetic expression profiles of cardiac hypertrophy and heart failure. A and B: K-mean clustering (A) and Venn diagram (B) of significantly altered genes in physiological vs. pathological hypertrophy. C and D: K-mean clustering (C) and Venn diagram (D) of significantly altered genes during transition from compensated cardiac hypertrophy to heart failure with high-salt diet. For K-mean clustering, the expression levels were standardized to mean 0 and standard deviation 1 before the cluster analysis in y-axis (n = 3). C6, control at 6 wk; E6, exercise for 6 wk; H6, high salt for 6 wk; H15, high salt for 15 wk.

Genetic profiles of pathological cardiac hypertrophy and heart failure.

Genetic changes that are associated with the transition from compensated cardiac hypertrophy to decompensated heart failure showed 874/3,160 genes that were changed with a statistical significance of model q value <0.05. K-mean clustering into six distinct patterns revealed that most of the genes (369/874 genes) were changed in both hypertrophy and heart failure (clusters 5 and 6; Fig. 2C). These genes included hypertrophic signaling genes, such as GATA4, RAB7, NRAS, GNA12, STAT3, STAT5B, FYN, CRKO, MYCN, PTEN, AKT1, and IL6ST/gp130. These findings suggest that these pathways are significantly activated during the transition from compensated pathological hypertrophy to decompensated heart failure. There were also several genes that encode for channels, such as VDAC1 and CAVNβ2, that suggest the possible role of these channel proteins in the process. There were 205/874 genes altered predominantly in hypertrophy (clusters 3 and 4) and 300/874 genes altered predominately in heart failure (clusters 1 and 2). These included genes associated with oxygen regulation and deprivation, such as heme oxygenase-1, HAGH, thioreductase-2, SOD2, and cytochrome b5, suggesting that oxygen deprivation is an important aspect of pathological hypertrophy and heart failure. Furthermore, several stress-reactive genes, such as heat shock proteins (hsp40, hsp70), and inflammation-related genes, such as Alox12 and PAP, were upregulated in these groups. The fold changes of all the significant genes were described in Supplement Table S1 (available at the Physiological Genomics web site).1 Venn diagram analysis revealed that most of the gene changes occurred in 6-wk high-salt diet (120 genes) and 15-wk high-salt diet groups (89 genes) when there were obvious phenotypic changes (Fig. 2D). ANF was the only gene that changed in all three pathological groups but not in physiological groups (Fig. 2D and Fig. 3).

Fig. 3.

Specific gene changes during various conditions. A: Northern analysis of atrial natriuretic factor (ANF) and brain natriuretic protein (BNP) from different groups. B: quantitative analysis of ANF and BNP expression in different groups (n = 3). C: quantitative analysis of pancreatitis-associated protein (PAP) and arachidonate 12-lipoxygenase (Alox12) in different groups (n = 3). Bars indicate SE.

Specific genes of interest.

ANF and BNP have been shown to be upregulated during hypertrophy and heart failure (3, 26, 28, 29, 33, 49). In our study, there was significant upregulation of ANF mRNA expression in pathological hypertrophy (6.1-fold) and an even greater increase in heart failure (14.3-fold) (Fig. 3, A and B). ANF was also increased in physiological hypertrophy (2.3-fold), although significantly less compared with pathological conditions. There was a significant increase in the tissue mRNA BNP level during pathological hypertrophy (3.0-fold) but not physiological hypertrophy. Interestingly, there was a significant downregulation of the tissue mRNA BNP level during the transition to decompensated heart failure to the baseline level (Fig. 3, A and B). Because these two genes show very distinctive change patterns during hypertrophy/heart failure, the absolute levels as well the ANF-to-BNP ratio may be useful in understanding the current therapeutic strategy for the treatment of decompensated heart failure.

A significant difference between pathological and physiological hypertrophy was the change in acute stress reactants and inflammatory-related genes in pathological hypertrophy. Pathological hypertrophy resulted in upregulation of 84 genes, compared with 19 genes in physiological hypertrophy (see Supplemental Table S1). In addition, the state of decompensated heart failure resulted in an even more impressive activation of stress reactants. Moreover, there was extremely high activation of inflammatory response genes, such as PAP (123-fold) and Alox12 (25-fold), during heart failure (Fig. 3C). These findings suggest a potentially important role of inflammation during the transition from compensated hypertrophy to decompensated heart failure.

Hypertrophic signaling pathways.

To identify groups of genes that are altered during various stimuli, we used resampling-based gene score calculating method to define significantly altered gene clusters, based on GO classification. During pathological hypertrophy, genes associated with the apoptosis pathway showed statistically significant changes. Analysis of genes classified as apoptosis genes revealed that during pathological hypertrophy, there was a highly statistically significant change in these genes (Table 3). Genes associated with regulation of apoptosis, apoptosis inhibitor activity, anti-apoptosis, regulation of apoptosis, and regulation of programmed cell death showed statistical changes during pathological hypertrophy. Interestingly, only apoptosis inhibitor activity and anti-apoptosis showed statistically significant changes during physiological hypertrophy (Table 3). This finding suggests that there is increased apoptotic modulation during pathological hypertrophy that may play a role during the transition from pathological hypertrophy to heart failure.

View this table:
Table 3.

Representative gene groups during physiological and pathological hypertrophy

In comparison, we identified several signaling pathways that were statistically changed only during physiological hypertrophy (Table 3). For example, the gene clusters associated with glucose/insulin signaling, protein biosynthesis, and epidermal growth factor (EGF) signaling showed statistically significance changes during physiological hypertrophy, suggesting that these pathways may be involved predominately during physiological hypertrophy. Interestingly, EGF signaling activation occurred only early during physiological hypertrophy (i.e., the 3-wk exercise group). A complete list of enriched GO classes are described in Supplement Table S2.


In this study, we analyzed gene expression in two distinct models representative of physiological and pathological cardiac hypertrophy. Despite development of a comparable degree of hypertrophy, there were significant differences in molecular changes associated with specific stimuli, suggesting that different genetic programs may be altered in the two models. Specific changes occurred in a wide range of genes involved in hypertrophic signaling, metabolism, cell structural organization, cell death, stress responses, and ion channels that may be important in understanding the pathophysiology of these conditions. Because pathological hypertrophy usually progresses to heart failure, a better characterization of the changes seen in pathological compared with physiological hypertrophy may provide clues to the specific pathways involved in the progression to heart failure.

Our findings suggest that pathological processes, such as pathological hypertrophy and heart failure, are characterized by an increase in the stress genes, such as genes involved in inflammation, oxidative stress, and acute stress response. Physiological hypertrophy did not manifest these changes in stress genes. Also, our data are consistent with previous findings (54) that PAP and Alox12 increase significantly during the transition to heart failure in Dahl rats. PAP is a 16-kDa secretory protein belonging to the C-type lectin family (14) that is strongly upregulated by proinflammatory mediators, such as the cytokine IL-6 (6). PAP is speculated to be involved in cell proliferation (39), the inhibition of apoptosis (12, 42, 43), and anti-inflammatory function (18, 55). Alox12 is a member of a family of lipoxygenases that metabolize arachidonic acid (5). Alox12 activation may play a role in ANG II-induced vascular smooth muscle cell hypertrophy (41). Furthermore, mice deficient in the leukocyte-type 12-lipoxygenase have impaired protection after ischemic preconditioning, suggesting that leukocyte 12-lipoxygenase is an important mediator of preconditioning (17).

The transition from pathological hypertrophy to heart failure has been shown to involve cardiac apoptosis in various human and animal models of heart failure (24). In this study, we found that various components of apoptosis pathways were significantly changed during pathological processes, particularly during pathological hypertrophy. Previously, we demonstrated that a proapoptotic milieu is induced during pathological hypertrophy that may cause cardiac myocytes to be more sensitive to apoptotic stimuli (25). In addition, in microarray analysis of various transgenic mice that are associated with hypertrophy and heart failure, the Gαq-transgenic mouse heart revealed an increased expression of apoptosis genes (1). This finding suggests that the apoptosis pathway may be associated with significant changes during pathological hypertrophy and may contribute to the development of heart failure in this model.

In our study, there was a significant change in several signaling pathways during physiological hypertrophy. The insulin pathway plays an important role in normal growth and development in various tissues (7). It is also a major regulator of insulin-like growth factor (IGF)-1 and phosphatidylinositol 3-kinase (PI3-kinase), which in turn regulate many important biological responses, such as cell growth and survival (9, 53). In the heart, the IGF and PI3-kinase pathway has been implicated in normal cardiac growth and the development of physiological hypertrophy (34, 35, 48). Overexpression of IGF receptor-1 induces compensated cardiac hypertrophy resembling physiological hypertrophy in a PI3-kinase pathway-dependent manner (34). Furthermore, cardiac-specific overexpression of dominant negative PI3-kinase is associated with hypertrophy in response to pressure overload via aortic banding, but not to exercise-induce cardiac hypertrophy (35). In fact, the mTOR-S6K-E1BP pathway downstream from the PI3-kinase pathway has been shown to be important for protein biosynthesis during physiological hypertrophy (15, 38, 47). Our current data are consistent with our previous findings in mouse models suggesting that the insulin-IGF-PI3-kinase pathway axis plays a dominant role in physiological hypertrophy. Furthermore, our data suggest that other pathways, such as the EGF pathway, might be involved in physiological hypertrophy. The EGF pathway has been shown to cause hypertrophy via shedding of EGF by a disintegrin and metalloproteinase domain 12 (ADAM12) or via cyclic AMP stimulation (2, 40).

ANF and BNP are a family of structurally related peptides that participate in the integrated control of renal and cardiovascular function. ANF is mainly secreted by the atria in response to the stretch that occurs with the increased left atrial pressure associated with congestive heart failure (CHF). Its level is known to rise early in the course of heart failure and has been used as a marker for the diagnosis of asymptomatic left ventricular dysfunction with a sensitivity and specificity of >90% (29). BNP is synthesized predominantly in ventricular myocardium in response to increased wall stress, hypertrophy, and volume overload (26). The plasma concentrations of both ANF and BNP are increased in patients with either asymptomatic or symptomatic left ventricular dysfunction, permitting their use in diagnosis as well as providing prognostic information in patients with chronic CHF, including those receiving therapy with β-blockers and angiotensin-converting enzyme inhibitors (3, 28, 33, 49). In addition, Nesiritide (recombinant human BNP) produces significant improvement in various hemodynamic parameters and was shown to be an effective therapy in the treatment of decompensated heart failure in randomized clinical trials (11, 22, 36).

In our model of decompensated heart failure, the increase in ANF was significantly greater than that observed during pathological hypertrophy. Interestingly, BNP levels during this state were significantly lower than in pathological hypertrophy and closer to the normal level. This finding contrasts with the plasma BNP levels in patients that present with heart failure. However, our finding regarding BNP is supported by the fact that tissue mRNA BNP levels may not correlate well with plasma BNP levels, and that posttranslational control may play a major role in the regulation of BNP gene expression in heart (52). This might explain the effectiveness of Nesiritide during decompensated heart failure, despite high circulating plasma BNP levels. Of note, there are other studies showing that the mRNA level of BNP was not elevated at a compensatory hypertrophic stage, but only at a decompensated stage in Dahl salt-sensitive rats (46, 56). However, it would be difficult to make a comparison with these papers, since the mRNA levels for the “decompensated group of rats” were measured after only 12 wk of high-salt diet (compared with 15–18 wk in our study). At this stage, there was no deterioration in systolic cardiac function, which was comparable to our rats at 6 wk of high-salt diet (56). In our study, ejection fraction was significantly depressed, and the rats were clinically in heart failure (Table 2), which is consistent with previously published data on these rat models during the transition to heart failure (21, 25).

In conclusion, our study shows that there are distinct genetic regulations that appear to be specific as to whether hypertrophic stimuli are physiological or pathological. Although further studies are needed to confirm these findings, we believe this study provides comprehensive gene alterations as well as significantly altered pathways during various forms of cardiac hypertrophy. These differences may give insight into why the propensity for cardiac hypertrophy to progress to heart failure differs for different stimuli. In addition, the molecular switching that may occur during the transition could be a key component of understanding the molecular mechanism of heart failure and to finding therapeutic targets for the treatment of heart failure.


This study was supported in part by grants from the American Heart Association (0030278N; to P. M. Kang) and the National Heart, Lung, and Blood Institute (U01-HL-66582; to S. Izumo).


We thank Ellen Gower for editorial assistance.



View Abstract