Chronic carvedilol treatment partially reverses the right ventricular failure transcriptional profile in experimental pulmonary hypertension

Abstract

Right ventricular failure (RVF) is the most frequent cause of death in patients with pulmonary arterial hypertension (PAH); however, specific therapies targeted to treat RVF have not been developed. Chronic treatment with carvedilol has been shown to reduce established maladaptive right ventricle (RV) hypertrophy and to improve RV function in experimental PAH. However, the mechanisms by which carvedilol improves RVF are unknown. We have previously demonstrated by microarray analysis that RVF is characterized by a distinct gene expression profile when compared with functional, compensatory hypertrophy. We next sought to identify the effects of carvedilol on gene expression on a genome-wide basis. PAH and RVF were induced in male Sprague-Dawley rats by the combination of VEGF-receptor blockade and chronic hypoxia. After RVF was established, rats were treated with carvedilol or vehicle for 4 wk. RNA was isolated from RV tissue and hybridized for microarray analysis. An initial prediction analysis of carvedilol-treated RVs showed that the gene expression profile resembled the RVF prediction set. However, a more extensive analysis revealed a small group of genes differentially expressed after carvedilol treatment. Further analysis categorized these genes in pathways involved in cardiac hypertrophy, mitochondrial dysfunction, and protein ubiquitination. Genes encoding proteins in the cardiac hypertrophy and protein ubiquitination pathways were downregulated in the RV by carvedilol, while genes encoding proteins in the mitochondrial dysfunction pathway were upregulated by carvedilol. These gene expression changes may explain some of the mechanisms that underlie the functional improvement of the RV after carvedilol treatment.

right ventricular failure (RVF) is a common cause of death of patients with severe pulmonary arterial hypertension (PAH) (53), and RV dysfunction has also been associated with increased mortality in patients with left ventricular failure (11, 29). However, despite its prognostic importance, therapies targeted to treat RVF are still lacking. We have reported that treatment with the nonselective adrenergic receptor antagonist carvedilol reverses established RVF in two different rat models of PAH (SU5416/hypoxia- and monocrotaline-induced PAH), and that this improvement in RV function was associated with reduced hypertrophy and improved capillary density of the myocardium (3). These changes were partially explained by an effect of carvedilol on the expression of genes which are likely potentially relevant for the adaptive response of the right ventricle (RV) to chronic pressure overload (3). However, the mechanisms by which carvedilol reverses maladaptive RV hypertrophy remain largely incomplete. Carvedilol has been shown to have pleiotrophic cardioprotective effects beyond heart-rate control. Indeed, a spectrum of anti-inflammatory, antioxidant, and antiapoptotic actions of carvedilol has been reported in vitro and in vivo (27, 51).

Based on microarray analysis, we have previously reported that RVF, in contrast to functional, compensated RV hypertrophy, is characterized by a distinct gene expression profile (6). The set of differentially expressed genes was designated the “RVF prediction set,” and we speculated that these genes might play a critical role in the transition between adaptive hypertrophy and failure (6). Based on a similar approach, we provide an in-depth assessment to show that carvedilol treatment affects the expression of a number of genes contained in the RVF prediction set (6). An initial analysis of the carvedilol-treated RVs showed that the gene expression profile resembled mostly that of the RVF tissue. However, an analysis beyond the boundaries of the prediction set revealed a small group of genes that were differentially expressed after carvedilol treatment.

METHODS

Justification of animal model.

The SuHx model of severe angio-obliterative PAH and RVF reproduces multiple salient features of human PAH and RVF. As previously described, the RV in the SuHx rat model responds to pulmonary hypertension with a robust degree of hypertrophy followed by dysfunction and failure (33). Furthermore, the RV of these rats is characterized by fibrosis, capillary rarefaction, and cardiomyocyte apoptosis (2), which results in decreased cardiac output and exercise capacity, as well as a marked RV dilation (3). Although the monocrotaline model of PAH also develops significant RV dysfunction, we do not use this model because it presents with multiple features that do not resemble the human disease (13); pertinently, treatment with monocrotaline generates liver damage and myocarditis (reviewed in Ref. 13).

Animal models.

Male Sprague-Dawley rats (Harlan Laboratories, Indianapolis, IN) weighing 200 g were injected subcutaneously with SU5416 suspended in 0.5% (wt/vol) carboxymethylcellulose sodium, 0.9% (wt/vol) sodium chloride, 0.4% (wt/vol) polysorbate 80, and 0.9% (vol/vol) benzyl alcohol in deionized water. Rats were given a single injection of SU5416 (20 mg/kg) at the beginning of the 6 wk experiment. The animals were then exposed to chronic hypoxia (simulated altitude of 5,000 m in a nitrogen dilution chamber) for 4 wk; thereafter the animals were kept at the altitude of Richmond, VA (45.7 m), for another 2 wk. Carvedilol (15 mg/kg; Sigma-Aldrich, St. Louis, MO) was dissolved in 20% dimethyl sulfoxide and water and administered once daily per oral gavage for 4 wk, beginning after return to room air breathing. All research was approved by the Virginia Commonwealth University Institutional Animal Care and Use Committee.

Microarray analysis.

Total RNA was isolated from ∼30 mg of snap-frozen rat heart tissue using the Qiagen RNeasy Mini Kit (Qiagen, Valencia, CA). Hearts were homogenized with Buffer RLT and β-mercaptoethanol in an MP FastPrep-24 Lysing Matrix D tube (MP Biomedicals, Solon, OH), and then RNA was isolated and purified following the manufacturer's protocol. RNA concentration was determined with a NanoDrop ND-1000 (Thermo Fisher Scientific, Wilmington, DE). All samples had an A₂₆₀/A₂₈₀ ratio between 1.9 and 2.1.

The amplification and hybridization process is as follows: 500 ng of total RNA was amplified and labeled with Cyanine-5, and 500 ng of universal rat reference RNA (Stratagene, Santa Clara, CA) was amplified and labeled with Cyanine-3 using the Agilent QuickAmp Labeling kit (Agilent Technologies, Santa Clara, CA) to produce labeled cRNA following the manufacturer's protocol. After amplification and labeling, the dye incorporation was determined with a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE). All ratios were >8.0 pmol dye/μg cRNA per the manufacturer's recommendation. We combined 825 ng of sample and 825 ng of reference RNA and incubated that with an Agilent whole rat genome 4 × 44k microarray slide (Agilent Technologies, Wilmington, DE) for 17 h at 65°C. Following hybridization, slides were washed following the manufacturer's protocol and scanned with an Axon GenePix 4200A scanner (Axon Instruments, Union City, CA) at a resolution of 5 μM. The raw data were generated using GenePix Pro 5.0 software (Axon Instruments) and submitted to the Ramhorn Array Database, a VCU-specific implementation of the Longhorn Array Database (19).

Microarray data were retrieved from Ramhorn after filtering the data using a set of spot-quality metrics designed to ensure the reliability of the data. Features, individual spots on the microarray, were included in the dataset if they have a signal ≥1.5-fold above background in both the red and green channels. Those features with <50% good data, defined as passing aforementioned filtering conditions across the set of arrays, were eliminated. Technical replicates were averaged together by a best-effort averaging procedure.

Statistical analysis of biological replicates was performed using the Significance Analysis of Microarrays (SAM) algorithm (46) with a two-class paired unpaired design (SuHx + Carv RV vs. SuHx RV) with median centering to identify differentially expressed genes. Briefly, the SAM algorithm for a two-class comparison is as follows. First, the relative difference in expression between two sets of samples is calculated for each gene; this is the observed relative difference. To calculate the expected relative difference, data labels are permuted and the relative difference is calculated for each gene. The expected relative difference for a gene will be the average relative difference across all permutations. The observed relative difference is plotted against the expected relative difference. Genes that fall outside of a threshold value, Δ, will be called significant. Δ-Values were chosen to give an acceptable false discovery rate (FDR) of < 5%. For each gene, we report the FDR-equivalent P value, the q value as a percentage. The q value is defined as the minimum FDR of a given individual test at which a gene may be called significant.

The gene list was clustered by hierarchical clustering methods in Cluster 3.0 (9) and visualized in Java Tree View (39). Based on literature search results and pathway analysis results obtained using Ingenuity Pathway Analysis [(IPA); Ingenuity Systems, Redwood City, CA] and the PANTHER classification system (30, 44), we selected genes for confirmation by quantitative RT-PCR.

For prediction analyses, raw expression data files were uploaded into R (36) and normalized with the marray package (50) by the Lowess normalization algorithm and then exported to BRB Array Tools. Lowess intensity-dependent normalization was used to adjust for differences in labeling intensities of the Cy3 and Cy5 dyes. The adjusting factor varied over intensity levels (40). We identified genes that were differentially expressed between the three classes by an F test, using an α-level of 0.0001 for univariate testing, and performing 1,000 permutations for multivariate analysis.

The class prediction model was developed using diagonal linear discriminant analysis (37), nearest neighbor classification (7), and nearest centroid classification. The models incorporated genes that were previously deemed significant. We estimated the prediction error of each model using leave-one-out cross-validation (LOOCV) as described by Simon et al. (42). For each LOOCV training set, the entire model building process was repeated, including the gene selection process. We also evaluated whether the cross-validated error rate estimate for a model was significantly less than one would expect from random prediction. The class labels were randomly permuted, and the entire LOOCV process was repeated. The significance level is the proportion of the random permutations that gave a cross-validated error rate no greater than the cross-validated error rate obtained with the real data. We used 2,000 random permutations.

We have previously described a set of 450 genes that were most discriminatory between normal RVs, adaptive nondysfunctional hypertrophic RVs (induced by chronic hypoxia PH), and failing RVs (induced by SU5416/hypoxia), and we designated this group of genes the RVF prediction set (6). By comparing the expression of carvedilol-treated failing RVs to the RVF prediction dataset, we determined to which expression pattern carvedilol-treated RV tissues samples most resembled and identified the effects that carvedilol treatment had among this relatively small subset of genes. The RVF prediction set, derived from normal, chronic hypoxia (nondysfunctional RV hypertrophy) and RVF (SuHx RV) microarrays, acted as a training data set against which the SuHx rats treated with carvedilol (hereafter SuHx + Carv) data set (predictor data set) was compared.

We employed a class comparison algorithm that uses a t-test to compare the expression of each gene between the SuHx and SuHx + Carv microarrays data sets (41). Those genes with a P value below a given threshold value were accepted as significantly changed. In addition, multivariate permutation testing was performed to control for the FDR (21, 40). For each permutation in the multivariate permutation testing algorithm, the data labels were changed and the P value was recalculated for each gene. The P value threshold was set at 0.001 to allow no more than 10 false-positive genes and a proportion of false-positive genes <5%.

All microarray data discussed have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO)(8) database and are accessible through GEO Series accession number GSE42579 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE42579).

Gene expression studies.

Quantitative real-time PCR (qRT-PCR) was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA) and Power SYBR Green PCR Master Mix (Applied Biosystems) along with rat-specific primers (Invitrogen, Carlsbad, CA). A dissociation profile was generated after each run to verify specificity of amplification. α-Actinin was used as a housekeeping gene.

Isolation of mitochondria and mitochondrial oxidative phosphorylation.

Isolation of a single population of cardiac mitochondria was conducted as described previously (12). Oxygen consumption by intact mitochondria was measured using a Clark-type oxygen electrode (Strathkelvin Instruments, North Lanarkshire, UK) as previously reported (12). Briefly, substrates for complex I (20 mM glutamate), complex II (20 mM succinate with 7.5 μM rotenone), and complex IV (1 mM N,N,N′,N′-tetramethyl-p-phenylenediamine/20 mM l-ascorbate with 7.5 μM rotenone) were used, and state 3 (0.2 mM ADP-stimulated), state 4 (ADP-limited) respiration, respiratory control ratio, maximal rate of state 3 respiration (2 mM ADP), and ADP/O ratio were determined.

Protein isolation.

Whole cell lysates were isolated and used for enzyme-linked immune sorbent assays (ELISAs). We homogenized 50 mg of tissue in radioimmunoprecipitation assay (Sigma-Aldrich) buffer containing 1 mM Na₃VO₄ (New England Biolabs, Ipswich, MA), 10 μM PMSF (Sigma-Aldrich), 10 μg/ml aprotinin (Sigma-Aldrich), 50 μM PR-619 (LifeSensors, Malvern, PA), 1% phosphatase inhibitor cocktail (Sigma-Aldrich), and 1% protease inhibitor cocktail (Sigma-Aldrich). Protein concentration was calculated with the DC protein assay kit (Bio-Rad, Hercules, CA).

Western blot analysis.

Western blot analysis was performed using whole cell lysate as described previously (12) using the NuPAGE system (Invitrogen, Carlsbad, CA). Immunodetection was performed with commercially available antibodies (PGC-1α #2178, α-actinin #3134; Cell Signaling Technology, Danvers, MA).

20S proteasome assay.

20S proteasome activity was measured from 50 μg whole cell lysate using the 20S proteasome assay kit (Chemicon International, Temecula, CA) following the manufacturer's protocol. The assay is based on detection of the fluorophore 7-amino-4-methylcoumarin (AMC) after cleavage from the labeled substrate LLVY-AMC. Free AMC fluorescence was measured using a 380/460-nm filter set on a SpectraMax GeminiXS fluorescence microplate reader (Molecular Devices, Sunnyvale, CA) and quantified with SOFTMax PRO software (Molecular Devices).

Ubiquitin ELISA.

Concentration of total ubiquitinated (poly- and monoubiquitinated) was measured with the UbiQuant ELISA kit (LifeSensors) from whole cell lysates following the manufacturer's protocol. Chemiluminescence was measured on a Synergy HT multimode microplate reader (BioTek Instruments, Winooski, VT) and quantified in Gen5 data analysis software (BioTek Instruments).

Statistical analysis.

Results are reported as means ± SE or as fold-changes ± SE unless otherwise specified. Differences between groups were assessed with one-way ANOVA and Newman-Keuls post hoc tests using GraphPad Prism (La Jolla, CA). A P value < 0.05 was accepted as significant.

RESULTS

Prediction analysis.

Using the one-nearest neighbor, three-nearest neighbors, nearest centroid, and linear discriminant analysis algorithms (7), we assigned each of the arrays in the SuHx + Carv data set a label of either mostly resembling “normal,” “hypertrophy,” and “failure” based on the similarity between the 450 genes from the training data set and the 450 genes in the predictor (SuHx + Carv) data set (Table 1). As can be seen in Fig. 1 and Table 1, each of the RV SuHx + Carv arrays was classified as most closely resembling failure by all of the algorithms. In other words, it was apparent that the expression of the genes in the SuHx + Carv dataset had a great deal of similarity to the RVF prediction set but also that there were a set of genes that were differentially expressed, significantly. Although the SuHx + Carv data were classified as most closely resembling RVF, the result encouraged us to search for pathways that could explain the improvement in RV function that has been reported after carvedilol treatment (3). We hypothesized that a relatively small number of differently expressed genes between the RVF and carvedilol-treated RVs were indeed responsible for the improvement in RV function.

Table 1.

Summary of prediction analysis results

Array ID	Genes in Classifier, n	1-Nearest Neighbor	3-Nearest Neighbors	Nearest Centroid	Linear Discriminant Analysis
237	435	failure	failure	failure	failure
238	435	failure	failure	failure	failure
239	435	failure	failure	failure	failure
526	435	failure	failure	failure	failure
527	435	failure	failure	failure	failure
604	435	failure	failure	failure	failure

Fig. 1.

Dendrogram for clustering experiments using centered correlation and average linkage of the 450 probes that showed 100% agreement across leave-one-out cross-validation (LOOCV). The first 3 major columns on the left, labeled “normal,” “SuHx,” and “SuHxCarv,” denote expression in the normal, SU5416/hypoxia, and SU5416/hypoxia + carvedilol right ventricles (RVs). Red coloration indicates increased expression compared with the average normal RV expression level, while green indicates reduced expression. The small columns represent different animals investigated under the same condition. Each row represents a gene from the prediction set. The last major column to the right of the gray line labeled “-” denotes the difference in expression between the SU5416/hypoxia + carvedilol and SU5416/hypoxia RVs. Red coloration indicates increased expression in the SU5416/hypoxia RV with carvedilol treatment compared with an untreated SU5416/hypoxia RV, and green indicates reduced expression in the SU5416/hypoxia + carvedilol RV compared with the SU5416/hypoxia RV.

Multivariate analysis.

To test this hypothesis, we used a class comparison algorithm, which identified 489 genes as significant given the chosen analysis conditions Of the 43,379 probes compared between SuHx and SuHx + Carv overall gene expression data sets. The probability of selecting at least 489 probes by chance at the threshold P value of 0.001 (if there was no difference between the classes) was 0.2%.

Cardiac hypertrophy, ceramide signaling, and apoptosis.

We found two clearly defined subclusters, which are illustrated in Fig. 2. Subcluster 1 was composed of 282 probes that corresponded to 275 different genes. The top five canonical pathways for genes in this cluster are listed in Table 2. Two of the top canonical pathways, regulation of eukaryotic initiation factor (eIF)4, p70S6k signaling and eIF2 signaling, were also categorized as cellular processes. eIF-2 and 4 are both required for the initiation of translation, and it is known that cardiac hypertrophy requires an increase in protein synthesis by individual cardiomyocytes (43). Thus, cardiac hypertrophy development likely requires greater activity of the eIFs, and indeed in a model of cardiac hypertrophy caused by chronic alcohol intake, eIF2 levels were shown to be increased (24). In our microarray dataset, several of the eIF genes were decreased in expression after carvedilol treatment.

Fig. 2.

Cluster for class comparison analysis between SU5416/hypoxia RV (failure) and SU5416/hypoxia + carvedilol RV microarray data sets. 489 probes were called significant at an α < 0.001 level and multivariate permutation testing gives a false discovery rate < 5%. Red represents greater relative expression than reference RNA, and green represents less relative expression than reference RNA.

Table 2.

Top five canonical pathways of subcluster 1

Pathway Name	P Value
Regulation of eIF4 and p70S6k signaling	1.98 × 10−4
Glucocorticoid receptor signaling	2.61 × 10−4
Ceramide signaling	4.04 × 10−4
EGF signaling	4.94 × 10−4
eIF2 signaling	5.68 × 10−4

eIF, eukaryotic initiation factor.

A reduction in genes encoding proteins involved in the development of cardiac hypertrophy (Fig. 2, subcluster 1, and Fig. 3) is reflected in the experimental data, which show that carvedilol reduces hypertrophy in the SuHx RV (3). Of particular importance to pulmonary hypertension are the genes hypoxia-inducible factor 1 alpha (HIF-1α), and angiopoietin-2 (Ang2). HIF-1α expression was decreased 2.1-fold in the SuHx + Carv RV compared with the SuHx RV. Increased expression of HIF-1a was found in the hearts of patients with PAH (26). Four and a half LIM domains 2 (FHL2) was decreased 2.3-fold in the SuHx RV. A decrease in FHL2 expression occurs in human left ventricular heart failure and is associated with a reduction in activity of phosphofructokinase 2 and adenylate cyclase (4). Ang2 is also known to play a role in PAH. Ang2 has proangiogenic properties when expressed in high concentrations (31) or in combination with VEGF (25, 52). Patients with idiopathic PAH have increased plasma levels of Ang2, and these levels correlate with a low cardiac index and high pulmonary vascular resistance (22). Ang2 expression decreased 2.3-fold in the SuHx + Carv RV compared with the SuHx RV. A reduction in the expression of these genes, as seen in the SuHx + Carv RV, could lead to reduced cardiac hypertrophy and improved myocardial function.

Fig. 3.

Microarray gene expression of genes involved in cardiac hypertrophy functions in subcluster 2. The first 3 major columns on the left labeled “normal,” “SuHx,” and “SuHxCarv” denote expression in the normal, SU5416/hypoxia, and SU5416/hypoxia + carvedilol RVs. Red coloration indicates increased expression compared with the average normal RV expression level, while green indicates reduced expression. The small columns represent different animals treated with the same condition. The last major column to the right of the gray line labeled “-” denotes the difference in expression between the SU5416/hypoxia + carvedilol and SU5416/hypoxia RVs. Red coloration indicates increased expression in the SU5416/hypoxia RV with carvedilol treatment compared with an untreated SU5416/hypoxia RV, and green indicates reduced expression in the SU5416/hypoxia + carvedilol RV compared with the SU5416/hypoxia RV.

We found that the ceramide signaling and glucocorticoid receptor signaling pathways were two of the top five canonical pathways in this subcluster (Table 2). Ceramide is considered injurious to the heart (34), while glucocorticoid receptors have been shown to play a role in postmyocardial infarction heart remodeling (23). Several genes of these pathways were decreased in expression in the SuHx + Carv RV relative to the SuHx RV. To a lesser extent, the sphingosine-1-phosphate signaling pathway was affected by carvedilol treatment as several member genes of this signaling pathway were decreased in expression after carvedilol treatment.

Metabolism, mitochondrial function, and oxidative stress response.

Subcluster 2 was made up of 207 probes that represented 192 different genes (Fig. 2). Figure 4 illustrates the different biological processes, of which cellular, metabolic, and developmental processes, as well as cellular component organization pathways, were the most abundant. The top five canonical pathways for the genes in this cluster are listed in Table 3. Several of these pathways, including the peroxisome proliferator-activator receptor (PPAR) signaling, PPARα/retinoid X receptor-α activation, and nuclear respiratory factor (Nrf)2-mediated oxidative stress response, represent metabolic processes of critical importance for adequate heart function (10). PPAR-α and -γ are both coactivated by the PPAR coactivator-1 alpha (PGC-1α). Both PGC-1α and PPAR-α are preferentially expressed in tissues with high oxidative capacity, such as the heart (10). Through coactivation of PPAR-α and estrogen-related receptor alpha (ERR-α), PGC-1α activates genes involved in the cellular uptake and mitochondrial oxidation of fatty acids. We have previously reported that RVF is characterized by a complex metabolic derangement where downregulation of PGC-1α seems to play a pivotal role (12). Because carvedilol treatment had a significant effect on the gene expression of the PPAR signaling pathways, we measured the gene and protein expression of PGC-1α, as well as different PGC-1α target genes, in RV tissue obtained from carvedilol-treated rats (Fig. 5). Indeed, carvedilol treatment significantly increased the suppressed transcript levels of PGC-1α, ERR-α, and PPAR-α and significantly increased the protein levels of PGC-1α measured by Western blot. Interestingly, treatment with metoprolol, a selective β1-blocker did not modify the expression of PGC-1α, which could suggest a specific effect of carvedilol not explained only by adrenergic blockade. Furthermore, carvedilol significantly affected the expression of numerous PGC-1α target genes involved in fatty acid metabolism (Fig. 5C).

Fig. 4.

Genes in subcluster 2 were classified by biological process by the PANTHER classification system (44, 45). The top 4 processes are cellular, metabolic, developmental, and cellular component organization.

Table 3.

Top five canonical pathways of subcluster 2

Pathway Name	P Value
PPAR signaling	2.59 × 10−5
PPARα/RXRα activation	5.75 × 10−4
Agrin interactions at neuromuscular junction	8.10 × 10−4
NRF2-mediated oxidative stress response	8.76 × 10−4
Prostate cancer signaling	1.56 × 10−3

PPAR, peroxisome proliferator-activator receptor; RXR, retinoid X receptor; NRF, nuclear respiratory factor.

Fig. 5.

A: quantitative real-time PCR analysis of PPAR coactivator (PGC)-1α, estrogen-related receptor (ERR)-α, and peroxisome proliferator-activator receptor (PPAR)-α in control, failing (SuHx), carvedilol-treated (carvedilol), and metoprolol-treated (metoprolol) RVs (*P < 0.01, #P < 0.001). B: Top: quantitative real-time PCR analysis of PGC-1α in control, failing (SuHx), and carvedilol-treated (carvedilol) RVs (*P < 0.01) Bottom: Western blot analysis of PGC-1α in failing (SuHx) and carvedilol-treated (carvedilol) RVs. C: quantitative real-time PCR analysis of the PGC-1α downstream genes CD36, ACSL1, CPT1α, CPT1β, CPT2, and ACADM in control, failing (SuHx), and carvedilol-treated (carvedilol) RVs (*P < 0.01, #P < 0.001). D: mitochondrial efficiency measured as ADP/O ratio in control, failing (SuHx), and carvedilol-treated (carvedilol) RVs (*P < 0.01).

Reactive oxygen species created during cardiac stress increase levels of nuclear factor kappa B (NF-κB), which subsequently increases levels of proinflammatory cytokines such as IL-1, IL-6, and TNF-α (38). Increased expression of NF-κB inhibitor alpha (NF-κBIA) suppresses the transcriptional activity of NF-κB. In our microarray data set, expression of NF-κBIA was increased 2.1-fold in the SuHx + Carv RV compared with the SuHx RV. This is perhaps one of several mechanisms whereby carvedilol affects inflammation.

In addition to those pathways listed in Tables 2 and 3, carvedilol treatment significantly affected mitochondrial function (Fig. 6). Indeed, we found an increase in expression of genes encoding mitochondrial enzymes (Fig. 6) after carvedilol treatment. Cytochrome c oxidase (complex IV of the electron transport chain) is the final enzyme in the electron transport chain, which catalyzes the oxidation of cytochrome c and reduction of oxygen to water. We found four subunits of cytochrome c oxidase encoding genes increased in their expression when we examined the microarray data set, and their expression changes are listed in Table 4. Because it is likely that these expression changes could be beneficial to the heart and could reflect a trend toward normalization in RV mitochondrial function, we evaluated mitochondrial efficiency in isolated mitochondria from failing RV tissue obtained from rats treated with carvedilol. Figure 5D demonstrates that mitochondria obtained from SuHx RVs had a significantly reduced ADP/O ratio, which measures mitochondrial respiratory efficiency. A lower ADP/O ratio represents high oxygen consumption per unit of ADP provided. In addition, carvedilol treatment partially normalized the ADP/O ratio.

Fig. 6.

A: canonical pathway analysis identified the pathways from the Ingenuity Pathway Analysis library of canonical pathways that were most significant to the dataset. The P value for the mitochondrial dysfunction canonical pathway is 5.16 × 10⁻³. Genes are colored based on fold-change comparison between the SU5416/hypoxia + carvedilol (treated) RV and the SU5416/Hypoxia (failing) RV. Red indicates a relatively higher expression in the treated RV than the failing RV, and green indicates a relatively higher expression in the failing RV than the treated RV. B: microarray gene expression of genes encoding proteins involved in mitochondrial dysfunction in subcluster 2. The first 3 major columns on the left labeled “normal,” “SuHx,” and “SuHxCarv” denote expression in the normal, SU5416/hypoxia, and SU5416/hypoxia + carvedilol RVs. Red coloration indicates increased expression compared with the average normal RV expression level, while green indicates reduced expression. The small columns represent different animals treated with the same condition. The last major column to the right of the gray line labeled “-” denotes the difference in expression between the SU5416/hypoxia + carvedilol and SU5416/hypoxia RVs. Red coloration indicates increased expression in the SU5416/hypoxia RV with carvedilol treatment compared with an untreated SU5416/hypoxia RV, and green indicates reduced expression in the SU5416/hypoxia + carvedilol RV compared with the SU5416/hypoxia RV.

Table 4.

Genes differentially expressed in the mitochondrial dysfunction pathway

Gene Name	Fold Change	q Value, %	Description
CytB	1.638	0.000	cytochrome b
Ndufa3	1.494	1.230	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 3
Cox6a2	1.641	0.316	cytochrome c oxidase subunit VIa polypeptide 2
Atp6	2.174	0.000	ATP synthase 6, mitochondrial
BQ782568	1.489	1.230	BQ782568 UI-R-FF0-cpk-h-15-0-UI.s1 NCI_CGAP_FF0 Rattus norvegicus cDNA clone UI-R-FF0-cpk-h-15-0-UI 3¢, mRNA sequence
Nctsn	1.675	0.422	nicastrin
Cox2	2.562	0.000	cytochrome c oxidase II, mitochondrial
Cox1	2.207	0.000	cytochrome c oxidase I, mitochondrial
Cox8b	2.173	0.180	cytochrome c oxidase, subunit VIIIb

As shown in Fig. 2, genes of the PPAR signaling and NRF2-mediated oxidative stress response pathways increase in expression, suggesting a return toward normal expression levels after carvedilol treatment. Furthermore, carvedilol affects a large number of genes in the mitochondrial function pathway (Fig. 6). In the aggregate, these results could suggest that a change in energy substrate utilization is associated with the reversal of RVF and that carvedilol might be able to reverse the metabolic remodeling in failing RVs.

Protein ubiquitination.

We found a decrease in the expression in the SuHx + Carv RV of a large number of genes representing the protein ubiquitination pathway (Fig. 7). The protein ubiquitination system has been shown to be cardioprotective because of its ability to remove proapoptotic signaling molecules from the cell (1). A decreased expression of ubiquitination proteins leads to reduced proteasomal activities (28) and allows accumulation of proapoptotic proteins in the cell (1). A list of gene expression changes is provided in Table 5. We list genes representing several members of the DNAJ homolog family, several proteasome subunits, ubiquitin-conjugating enzymes, and ubiquitin-specific peptidases. Because this expression pattern suggested to us that the ubiquitin proteasome system is highly activated in RVF and that carvedilol treatment returns activity towards normal, we measured the levels of ubiquitinated proteins and the activity of the 20S proteasome. In support of our microarray data, we found a threefold increase in the levels of total ubiquitinated protein and evidence of heightened 20S proteasome activity in the SuHx RVs, while the SuHx + Carv RVs were characterized by normal or near normal activities.

Fig. 7.

A: canonical pathway analysis identified the pathways from the Ingenuity Pathway Analysis library of canonical pathways that were most significant to the dataset. The P value for the protein ubiquitination pathway is 6.59 × 10⁻¹¹. Genes are colored based on fold-change comparison between the SU5416/hypoxia + carvedilol (treated) RV and the SU5416/Hypoxia (failing) RV. Red indicates a relatively higher expression in the treated RV than the failing RV, and green indicates a relatively higher expression in the failing RV than the treated RV. B: zoom-in of the same figure showing the 20S proteasome. C: 20S proteasome activity in control, failing (SuHx), and carvedilol-treated (SuHx + Carv) RV (*P < 0.05). D: total ubiquitinated protein in control, failing (SuHx), and carvedilol-treated (SuHx + Carv) RV (**P < 0.01, ***P < 0.001).

Table 5.

Genes differentially expressed in the protein ubiquitination pathway

Gene ID	Fold Change	Entrez Gene Name
AMFR	0.899	autocrine motility factor receptor, E3 ubiquitin protein ligase
B2M	0.546	beta-2-microglobulin
CRYAB	1.120	crystallin, alpha B
DNAJA1	0.481	DnaJ (Hsp40) homolog, subfamily A, member 1
DNAJB1	0.534	DnaJ (Hsp40) homolog, subfamily B, member 1
DNAJB5	0.856	DnaJ (Hsp40) homolog, subfamily B, member 5
DNAJB6	1.083	DnaJ (Hsp40) homolog, subfamily B, member 6
DNAJC7	0.865	DnaJ (Hsp40) homolog, subfamily C, member 7
DNAJC9	0.550	DnaJ (Hsp40) homolog, subfamily C, member 9
DNAJC10	0.531	DnaJ (Hsp40) homolog, subfamily C, member 10
DNAJC11	1.119	DnaJ (Hsp40) homolog, subfamily C, member 11
DNAJC15	0.373	DnaJ (Hsp40) homolog, subfamily C, member 15
HLA-C	0.541
HSCB	0.537	HscB iron-sulfur cluster co-chaperone homolog (E. coli)
HSP90AA1	1.075	heat shock protein 90 kDa alpha (cytosolic), class A member 1
HSP90AB1	0.506	heat shock protein 90 kDa alpha (cytosolic), class B member 1
HSPA4	0.856	heat shock 70 kDa protein 4
HSPA5	0.418	heat shock 70 kDa protein 5 (glucose-regulated protein, 78 kDa)
HSPA8	0.436	heat shock 70 kDa protein 8
HSPA9	0.454	heat shock 70 kDa protein 9 (mortalin)
HSPA14	0.431	heat shock 70 kDa protein 14
HSPA12B	1.487	heat shock 70kD protein 12B
HSPA1A/HSPA1B	0.395	heat shock 70 kDa protein 1A
HSPB2	0.871	heat shock 27 kDa protein 2
HSPB6	1.131	heat shock protein, alpha-crystallin-related, B6
HSPB7	1.460	heat shock 27 kDa protein family, member 7 (cardiovascular)
HSPD1	0.453	heat shock 60 kDa protein 1 (chaperonin)
HSPE1	0.402	heat shock 10 kDa protein 1 (chaperonin 10)
NEDD4	0.503	neural precursor cell expressed, developmentally down-regulated 4
PSMA1	0.470	proteasome (prosome, macropain) subunit, alpha type, 1
PSMA2	0.565	proteasome (prosome, macropain) subunit, alpha type, 2
PSMA3	0.482	proteasome (prosome, macropain) subunit, alpha type, 3
PSMA4	0.452	proteasome (prosome, macropain) subunit, alpha type, 4
PSMA5	0.478	proteasome (prosome, macropain) subunit, alpha type, 5
PSMA6 (includes EG:246582)	0.408	proteasome (prosome, macropain) subunit, alpha type, 6
PSMB2	0.480	proteasome (prosome, macropain) subunit, beta type, 2
PSMB3	0.507	proteasome (prosome, macropain) subunit, beta type, 3
PSMB4	0.494	proteasome (prosome, macropain) subunit, beta type, 4
PSMB5	0.516	proteasome (prosome, macropain) subunit, beta type, 5
PSMB6	0.420	proteasome (prosome, macropain) subunit, beta type, 6
PSMB9	0.528	proteasome (prosome, macropain) subunit, beta type, 9 (large multifunctional peptidase 2)
PSMC1	0.571	proteasome (prosome, macropain) 26S subunit, ATPase, 1
PSMC2	0.524	proteasome (prosome, macropain) 26S subunit, ATPase, 2
PSMC4	0.492	proteasome (prosome, macropain) 26S subunit, ATPase, 4
PSMC6	0.398	proteasome (prosome, macropain) 26S subunit, ATPase, 6
PSMD1	0.468	proteasome (prosome, macropain) 26S subunit, nonATPase, 1
PSMD2	0.543	proteasome (prosome, macropain) 26S subunit, nonATPase, 2
PSMD7	0.566	proteasome (prosome, macropain) 26S subunit, nonATPase, 7
PSMD10	0.486	proteasome (prosome, macropain) 26S subunit, nonATPase, 10
PSMD14	0.398	proteasome (prosome, macropain) 26S subunit, nonATPase, 14
PSME2	0.520	proteasome (prosome, macropain) activator subunit 2 (PA28 beta)
SMURF1	0.570	SMAD specific E3 ubiquitin protein ligase 1
SMURF2	1.137	SMAD specific E3 ubiquitin protein ligase 2
SUGT1	0.501	SGT1, suppressor of G2 allele of SKP1 (S. cerevisiae)
TCEB1	0.478
TCEB2	0.421
THOP1	0.534	thimet oligopeptidase 1
UBA1	0.516	ubiquitin-like modifier activating enzyme 1
UBE2A	0.453	ubiquitin-conjugating enzyme E2A
UBE2D1	1.118	ubiquitin-conjugating enzyme E2D 1
UBE2D4	0.539
UBE2E2	0.499	ubiquitin-conjugating enzyme E2E 2
UBE2E3	0.382	ubiquitin-conjugating enzyme E2E 3
UBE2F	0.468	ubiquitin-conjugating enzyme E2F (putative)
UBE2G2	0.963	ubiquitin-conjugating enzyme E2G 2
UBE2I	0.456	ubiquitin-conjugating enzyme E2I
UBE2J2	0.835	ubiquitin-conjugating enzyme E2, J2
UBE2M	0.492	ubiquitin-conjugating enzyme E2M
UBE2N	0.453	ubiquitin-conjugating enzyme E2N
UBE2S	0.496	ubiquitin-conjugating enzyme E2S
UBE2V1	0.523	ubiquitin-conjugating enzyme E2 variant 1
UBE4A	0.514	ubiquitination factor E4A
UBE4B	1.016	ubiquitination factor E4B
UCHL3	0.464	ubiquitin carboxyl-terminal esterase L3 (ubiquitin thiolesterase)
UCHL5	0.515	ubiquitin carboxyl-terminal hydrolase L5
USO1	0.524	USO1 vesicle docking protein homolog (yeast)
USP7	1.198	ubiquitin specific peptidase 7 (herpes virus-associated)
USP14	0.441	ubiquitin specific peptidase 14 (tRNA-guanine transglycosylase)
USP15	0.863	ubiquitin specific peptidase 15
USP20	0.893	ubiquitin specific peptidase 20
USP32	1.284	ubiquitin specific peptidase 32
USP40	1.106	ubiquitin specific peptidase 40
USP42	1.309	ubiquitin specific peptidase 42
USP46	0.894	ubiquitin specific peptidase 46
USP47	0.937	ubiquitin specific peptidase 47
USP9X	0.929	ubiquitin specific peptidase 9, X-linked

DISCUSSION

Right heart failure remains the main cause of death in patients with chronic severe PAH, and, although treatment of patients with vasodilator drugs may reduce the pulmonary vascular resistance, this effect is not necessarily accompanied by improvement of RV function (48, 49). We have proposed that the failing RV should be a treatment target in its own right (48), and we have begun to investigate the effects of drugs that target the heart and improve RV function in the setting of persistently elevated RV afterload (3, 47). In our SuHx model of chronic and progressive RVF (33), we could show that a 4 wk treatment with carvedilol was sufficient to reverse the established RVF and that this reversal was associated with an increased RV capillary density and improved hemodynamics (3). Here, we sought to examine whether carvedilol treatment affected the RVF gene expression “signature” (6). By applying the tool of a relatively small set of genes, the RVF prediction set, we found that carvedilol treatment of rats with RVF did not categorically alter the RVF gene expression signature (Fig. 1). Indeed, the overall structure of gene expression pattern of the carvedilol-treated RVs reflects the RVF gene expression pattern. However, additional data analysis reveals that carvedilol treatment altered the expression of subsets of genes that were not contained within the RVF prediction set (Fig. 2). This is of interest because this finding suggests that a limited number of genes and/or signaling pathways may be critical to RVF reversal. We concluded that a prediction set analysis is useful for distinguishing between failing and nonfailing RVs but that a more penetrating and expanded analysis is required to detect carvedilol treatment-associated gene expression alterations that may provide mechanistic clues that explain an improvement in RV function.

The major finding of our study is that carvedilol is associated with directional changes in the expression of a set of genes which are likely controlling the development of RV hypertrophy and RVF. Because carvedilol treatment brought about an improvement of RV performance, a decrease in RV hypertrophy, a reduction in myocardial fibrosis, and a normalization of the density of the myocardial capillaries (3), it appears to be justified to consider that the carvedilol treatment-associated gene expression changes reflect a reduction in cardiac stress and that carvedilol improves mechanisms of cardiac repair.

Our approach, using an RVF gene expression prediction set as a data filter, allowed us to uncover clusters of genes that encode proteins involved in angiogenesis and cell growth. For example, EGFR signaling is known to play a role in cardiac hypertrophy and in mediating β-arrestin-dependent cytoprotective effects via an interaction with β-adrenergic receptors (32). A decrease in the expression of eIF genes could suggest, at least on a transcriptional level, a decreased protein synthesis and partially explain the reduction in RV hypertrophy associated with carvedilol treatment (3). As shown in Table 2, genes encoding proteins in the eIF2-, eIF4-, and EGF pathways were increased in expression in the SuHx model when compared with the SuHx + Carv RV tissues. This suggests that these pathways have been affected by carvedilol to reduce RV hypertrophy.

That carvedilol affected the expression of genes encoding enzymes of sphingolipid pathways was unexpected. Karliner recently reviewed sphingosine kinase and spingosine-1 phosphate and their cardioprotective role in acute ischemia/reperfusion injury (17). Ceramide is considered injurious, causing heart dysfunction (34), and it has been acknowledged that carnitine palmitoyltransferase 1 (CPT1) deficiency (CPT1 oxidation) is associated with cardiac hypertrophy and increased heart tissue ceramide content (14). We also found that carvedilol affected the gene expression of members of the glucocorticoid receptor pathway. A study by Kuster et al. (23) implies that glucocorticoid receptors play an important part in post myocardial infarction heart remodeling.

Carvedilol treatment decreases the expression of many genes associated with cardiac hypertrophy. The higher than normal expression of FHL2 (Fig. 3) is reduced by carvedilol treatment. As FHL2 directly interacts with HIF-1α and represses HIF-1α-dependent transcription (16), overexpression of FHL2 represses pathological cardiac growth (15). We speculate that the high expression of FHL2 in the SuHx failing RVs could compromise HIF-1a-dependent angiogenesis, as would the overexpression of homeodomain-only protein X (Fig. 3) (18).

Finally, we detected that, in the failing RV, there was a highly significant overexpression of a very large number of genes encoding proteins of the ubiquitin proteasome pathways. Carvedilol treatment returned the expression of these genes and the heightened activity of these protein degrading systems toward normal (Fig. 7). These findings support the concept that, in RVF, the ubiquitin proteasome system is highly activated and, therefore, that RVF is associated with proteotoxicity. Ubiquitination is a multistep process involved in the regulation of a number of signal transduction pathways in the heart, among them p53, Fox0, calcineurin/NFAT, and estrogen receptor signaling (35). Indeed, treatment of mice with the proteasome inhibitor bortezomib causes cardiac hypertrophy, and proteasome inhibition of mice with experimentally induced transaortic constriction generates a maladaptive left ventricular dilation (5, 20). Based on such data, we suggest that the heart ubiquitin proteasome system is required for the adaptation of the RV to stress including: cell growth, apoptosis, and autophagy. The dampening effect of carvedilol on proteasome gene expression is consistent with the carvedilol-induced improved and more compensated RV function, which perhaps requires less ubiquitin proteasome activity. To reiterate, the activation of the ubiquitin proteasome system in the failing RV is also consistent with a response to a proteotoxic stress of the cardiomyocytes generated by accumulation of misfolded and aggregated proteins and defective cell organelles.

Conclusion

In conclusion, we show that microarray gene expression studies are useful to characterize RVF and to show that carvedilol alters the “RVF gene expression signature.” When the carvedilol-induced expression changes are viewed in the context of RVF reversal, then these pattern changes begin to explain molecular mechanisms of RVF and begin to characterize the components of the failure program that are reversible.

GRANTS

This work was supported by funds from the Victoria Johnson Center for Obstructive Lung Research. The chemiluminescence microplate reader is located in the VCU Nucleic Acids Research Facility core laboratory, which is supported, in part, by funding by NIH-CI Cancer Center Support Grant P30 CA-016059.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: J.I.D., R.N., H.J.B., P.F., and N.F.V. conception and design of research; J.I.D., D.K., and R.N. performed experiments; J.I.D., C.I.D., R.N., and H.J.B. analyzed data; J.I.D., J.G.-A., R.N., H.J.B., P.F., and N.F.V. interpreted results of experiments; J.I.D., J.G.-A., C.I.D., and P.F. prepared figures; J.I.D., J.G.-A., C.I.D., and N.F.V. drafted manuscript; J.I.D., J.G.-A., H.J.B., P.F., and N.F.V. edited and revised manuscript; J.I.D., J.G.-A., C.I.D., R.N., H.J.B., P.F., and N.F.V. approved final version of manuscript.