Research ArticleCardiology Free access | 10.1172/jci.insight.86898
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Burke, M. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Chang, S. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Wakimoto, H. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Gorham, J. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Conner, D. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Christodoulou, D. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Parfenov, M. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by DePalma, S. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Eminaga, S. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Konno, T. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Seidman, J. in: JCI | PubMed | Google Scholar
1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
2Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.
3Department of Cardiology, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.
4Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, USA.
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Find articles by Seidman, C. in: JCI | PubMed | Google Scholar
Authorship note: M.A. Burke and S. Chang contributed equally to this work.
Published May 5, 2016 - More info
Dilated cardiomyopathy (DCM) is defined by progressive functional and structural changes. We performed RNA-seq at different stages of disease to define molecular signaling in the progression from pre-DCM hearts to DCM and overt heart failure (HF) using a genetic model of DCM (phospholamban missense mutation, PLNR9C/+). Pre-DCM hearts were phenotypically normal yet displayed proliferation of nonmyocytes (59% relative increase vs. WT, P = 8 × 10–4) and activation of proinflammatory signaling with notable cardiomyocyte-specific induction of a subset of profibrotic cytokines including TGFβ2 and TGFβ3. These changes progressed through DCM and HF, resulting in substantial fibrosis (17.6% of left ventricle [LV] vs. WT, P = 6 × 10–33). Cardiomyocytes displayed a marked shift in metabolic gene transcription: downregulation of aerobic respiration and subsequent upregulation of glucose utilization, changes coincident with attenuated expression of PPARα and PPARγ coactivators -1α (PGC1α) and -1β, and increased expression of the metabolic regulator T-box transcription factor 15 (Tbx15). Comparing DCM transcriptional profiles with those in hypertrophic cardiomyopathy (HCM) revealed similar and distinct molecular mechanisms. Our data suggest that cardiomyocyte-specific cytokine expression, early fibroblast activation, and the shift in metabolic gene expression are hallmarks of cardiomyopathy progression. Notably, key components of these profibrotic and metabolic networks were disease specific and distinguish DCM from HCM.
Dilated cardiomyopathy (DCM), a common cause of heart failure (HF), is a prevalent disorder that affects up to 1 in 250 individuals (1). DCM may arise as a primary genetic disorder or as a secondary manifestation of other cardiovascular or systemic conditions. Altered myocardial calcium homeostasis is a common feature in genetic and acquired forms of DCM and HF (2) and can perturb cardiac physiology by modulating contractile force, signaling pathways, and gene transcription.
To decipher the temporal expression of key molecular mediators and pathways in the progression to DCM and HF, we capitalized on a genetic mouse model that expresses a missense mutation (p.Arg9Cys) in phospholamban (PLNR9C/+), a transmembrane phosphoprotein that inhibits the cardiac sarcoplasmic/endoplasmic reticulum Ca2+–adenosine triphosphatase (SERCA2a) pump. Calcium cycling is abnormal in young PLNR9C/+ mice before the onset of overt cardiac remodeling (denoted as pre-DCM). However, PLNR9C/+ mice subsequently develop DCM that consistently progresses to fulminant HF and premature death, (3) thus recapitulating the chronologic manifestations in human patients with this mutation (3–5). To delineate the longitudinal consequences of altered calcium homeostasis in cardiac remodeling, we studied the transcriptional changes in cardiomyocytes and nonmyocytes from pre-DCM through DCM to HF.
Hypertrophic cardiomyopathy (HCM) is a genetic disorder caused by mutations in the protein constituents of the sarcomere. HCM shares pathophysiologic and histologic features with DCM, including dysregulated calcium homeostasis, ventricular remodeling, and increases in myocardial fibrosis; however, unlike DCM, progression to systolic dysfunction and end-stage disease is uncommon (6–8). We previously reported transcriptional changes that occur in a mouse model with a human HCM mutation (p.Arg403Gln) in the myosin heavy chain gene MHCR403Q (9, 10). Here, we compared the genes and pathways enriched in PLNR9C/+ mice with DCM to those enriched in MHCR403Q mice with HCM. Our studies identified remarkable similarities and also key differences between these phenotypically distinct cardiomyopathies.
Fibrosis and nonmyocyte proliferation characterize DCM in PLNR9C/+ mice
We defined the temporal onset and progression of disease in PLNR9C/+ mice by echocardiography (Supplemental Table 1; supplemental material available online with this article; doi:10.1172/jci.insight.86898DS1) and cardiac histopathology. At 8 weeks of age, pre-DCM PLNR9C/+ mice had neither ventricular dilatation nor histopathologic findings of myocyte hypertrophy or cardiac fibrosis (Figure 1A and Supplemental Figures 1 and 2). Despite this, hearts from pre-DCM PLNR9C/+ mice displayed increased nonmyocyte BrdU labeling compared with WT controls (2.0% ± 1.5% vs. 1.3% ± 0.9%; 59% relative increase; P = 8 × 10–4, Figure 1B). By 18 weeks of age, overt DCM morphology had emerged, and histopathology showed myocyte enlargement (Supplemental Figure 2) and significant ventricular fibrosis (3.6% ± 1.5% vs. 1.1% ± 0.4% WT, P = 1 × 10–13). At approximately 22 weeks, PLNR9C/+ mice had manifested HF as evidenced by progressive upregulation of natriuretic peptides (Figure 2A) and weight loss (PLNR9C/+ preterminal net weight change –3.6 ± 0.9 g vs. 0.7 ± 0.8 g WT, P = 3 × 10–8), consistent with cardiac cachexia (Supplemental Figure 3). Mice also had notable behavioral changes, including lethargy with labored breathing and reduced peripheral perfusion, suggestive of low cardiac output. HF histopathology showed marked biventricular dilatation, cardiomyocyte hypertrophy, profound fibrosis (17.6% ± 6.9% vs 0.6% ± 0.3% WT, P = 6 × 10–33), and increased nonmyocyte BrdU staining (2.0% ± 1.2% vs 1.4% ± 1.0% WT, P = 4 × 10–3; Figure 1 and Supplemental Figures 1 and 2). Based on BrdU labeling, we deduced that nonmyocyte proliferation contributed in part to increased gene transcription.
PLNR9C/+ mice develop increased cardiac fibrosis and nonmyocyte cell proliferation with disease progression. (A) Light microscopy (scale bars: 100 μm) of Masson’s trichrome–stained LV tissue at 8 weeks (pre-DCM), 18 weeks (DCM), and 22 weeks (overt HF) demonstrates progressive cardiac fibrosis in PLNR9C/+ hearts compared with age-matched WT animals. Data quantitated were individual LV slices (2 per slide) at 10 levels (apex to base) from n = 3 mice per group; 2-tailed Student’s t test. (B) Confocal microscopy (scale bars: 75 μm) and quantification of LV sections from hearts labeled with BrdU demonstrating nonmyocyte proliferation both pre-DCM and with overt HF. BrdU (magenta), wheat germ agglutinin (WGA; green), and nuclear DAPI (blue). Cells were counted from individual LV slices at 10 levels (apex to base) from n = 3 mice per group. Greater than 25,000 nuclei were counted per experiment; 2-tailed Student’s t test.
Activation of the cardiac stress response in PLNR9C/+ mice with disease progression. (A) Progressive and marked increase in atrial natriuretic peptide (ANP) and brain natriuretic peptide (BNP) mRNA levels with disease progression in PLNR9C/+ mice (P < 1 × 10–300 vs. WT at all disease stages). (B) Progressive increase in 4-and-a-half LIM domains protein 1 (Fhl1) mRNA levels with disease progression in PLNR9C/+ mice (P < 1× 10–300 vs. WT at all disease stages). (C) Reduction in the ratio of adult (α-MHC/Myh6) to fetal (β-MHC/Myh7) myosin heavy chain gene expression with disease progression in PLNR9C/+ mice. RNA from n = 3 mice was pooled prior to RNA-seq. Bayesian P value corrected for multiple hypothesis testing (10).
RNA-seq reveals differences between myocyte and nonmyocyte gene expression
Transcripts from left ventricle (LV) tissues derived from age-matched WT and PLNR9C/+ mice with pre-DCM, DCM, or HF were sequenced (RNA-seq) to identify genes with significantly altered expression (Supplemental Figure 4). We also performed RNA-seq on isolated cardiomyocytes and nonmyocytes from hearts of DCM and age-matched WT mice to categorize transcripts as predominantly expressed in either or both cell compartments. Separation of isolated cell populations was excellent, and RNA-seq data showed less than 4% cross-cell contamination of prototypic lineage–specific genes (Supplemental Table 2).
DCM mice had characteristic transcriptional changes indicating activation of the fetal gene program, which typifies HF. PLNR9C/+ mice displayed markedly increased expression of prototypical cardiomyocyte stress-response genes, including the natriuretic peptides (Nppa and Nppb) and 4-and-a-half LIM domains protein-1 (Fhl1), as well as a shift in myosin heavy chain gene expression from the adult (Myh6) to fetal (Myh7) isoforms (Figure 2).
Nonmyocyte transcriptomes showed over 2,000 genes that were differentially expressed in pre-DCM compared with WT, and this gene set increased with the transition to DCM and HF (Table 1, Supplemental Table 3). Fully, 24% (n = 1,209) of differentially expressed nonmyocyte genes were common to all stages (Figure 3). Prominent increases in expression of profibrotic cytokines such as TGFβ and connective tissue growth factor were observed, as were markers of extracellular matrix remodeling, including numerous collagen precursor genes, periostin, osteopontin, osteonectin (Sparc), tenascin-c, thrombospondin, matrix metalloproteinases, and metalloproteinase inhibitors (Supplemental Table 3). Notably, nonmyocytes had significantly increased expression of gene family members encoding complement factors, chemokines, and immune signaling receptors.
Differential gene expression at distinct stages of disease in nonmyocytes and myocytes from PLNR9C/+ hearts. (A) Differences in nonmyocyte gene expression progress steadily with worsening disease in LV tissue, with nearly a quarter of differentially expressed genes common to all stages. (B) Few differentially expressed genes were unique to cardiomyocytes before DCM, and only 12% of differentially expressed genes were common to all stages. More significant changes were noted with onset of phenotypic disease.
The cardiomyocyte transcriptome displayed a different dynamic pattern from that of nonmyocytes (Table 1 and Supplemental Table 3). Only 348 genes were differentially expressed in pre-DCM cardiomyocytes compared with WT, and only 12% of differentially expressed genes (n = 178; P = 2 × 10–16) were common throughout all stages of disease (Figure 3). The majority (69%) of pre-DCM differentially expressed cardiomyocyte genes were upregulated. Among these were genes that encode molecules involved in modulating calcium, including sarcolipin, sarcolemmal membrane–associated protein, calsequestrin, calreticulin, and the calcium-binding mitochondrial carrier. The onset of overt DCM marked a substantial shift and predominant downregulation in cardiomyocyte gene expression, including the expression of many genes encoding proteins that regulate cellular metabolic processes. With progression to HF, the overall cardiomyocyte transcriptome profile was largely unchanged, and while some transcripts varied, 54% of differentially expressed cardiomyocyte genes were common to both DCM and HF (Figure 3).
Divergent pathway activation in nonmyocytes vs. myocytes
To provide a survey of the molecular processes altered in PLNR9C/+ nonmyocytes and cardiomyocytes, we used the bioinformatics tools Ingenuity pathway analysis (IPA) and gene ontology sequencing (GO-seq) to define canonical biological pathways, GO terms and Kyoto Encyclopedia of Genes and Genomics (KEGG) pathways at 3 different stages of disease (Supplemental Tables 4–6). Additionally, the Ingenuity upstream regulator analysis tool was used to predict the activation state of molecules associated with the pattern of differential gene expression at each stage of disease (see Methods for details).
Nonmyocytes. Enriched pathways in DCM nonmyocytes included those involved in translation, acute phase responses, integrin signaling, and cytokine signaling (Supplemental Tables 4–6) — processes that would promote inflammation and fibrosis. The pattern of differentially expressed nonmyocyte genes predicted the activation of all 3 TGFβ isoforms (Figure 4A). Expression of Tgfb2 (10.4-fold, P = 3 × 10–277) and Tgfb3 (4.7-fold, P = 5 × 10–177) was strongly increased in DCM hearts compared with WT (Figure 4B). By contrast, Tgfb1 expression was only modestly increased (1.9-fold, P = 2 × 10–17), a finding likely related at least in part to the increased proliferation of nonmyocytes in PLNR9C/+ mice. RNA-seq of isolated DCM nonmyocytes revealed no change in Tgfb1 expression (1.05-fold, P = 0.06). Quantitative PCR (qPCR) of TGFβ isoforms in a separate cohort (n = 24 DCM and age-matched WT mice) confirmed selective activation of Tgfb2 (5.2-fold, P = 2 × 10–8) and Tgfb3 (3.7-fold, P = 0.02) but not Tgfb1 (1.4-fold, P = 0.10). Consistent with these gene-expression data, all bioinformatics platforms identified enrichment of TGFβ signaling in PLNR9C/+ nonmyocytes.
Increased expression of TGFβ in PLNR9C/+ mice. (A) All TGFβ isoforms were predicted to be activated at all 3 stages of disease by upstream regulator analysis. Z-score reflects both the confidence and direction of the inferred activation state (P < 1 × 10–6 for all analyses). (B) TGFβ gene expression was induced in PLNR9C/+ LV tissue at all stages of disease (n = 3 mice pooled prior to RNA-seq). (C) Isolated nonmyocyte and cardiomyocyte cells (n = 6 mice pooled prior to RNA-seq) at 18 weeks (DCM) showed that TGFβ genes were predominantly expressed in WT nonmyocytes, with modest increases in Tgfb2 and Tgfb3 expression in PLNR9C/+ nonmyocytes. In contrast, Tgfb2 and Tgfb3 were strongly induced in cardiomyocytes of PLNR9C/+ mice with DCM. Tgfb1 levels do not change significantly in either cell compartment. Bayesian P value corrected for multiple hypothesis testing (10).
We also observed increased expression of signaling molecules in PLNR9C/+ mice that would amplify TGFβ-mediated fibrosis and inflammation. TGFβ receptors 1 and 2 (Tgfbr1, 2.7-fold, P = 3 × 10–29; Tgfbr2, 2.4-fold, P = 7 × 10–57) were significantly upregulated in DCM hearts, and with progression to HF, Tgfbr3 (2.0-fold, 1 × 10–29) was also increased. The TGFβ1-responsive NADPH oxidase 4 (Nox4) gene, which mediates conversion of quiescent fibroblasts to active myofibroblasts, was also dramatically induced (56.5-fold, 2 × 10–33) in DCM hearts (11). Additionally, a host of profibrotic transcription factors were upregulated in DCM nonmyocytes. Runt-related transcription factor 1 (Runx1, 4.8-fold, P = 4 × 10–17) controls TGFβ-dependent myofibroblast differentiation (12). The related molecule, Runx2 (4.3-fold, P = 2 × 10–9), is a known regulator of collagen type-1 gene expression and was recently identified as a key mediator of vascular fibrosis (13). PU.1 (Sfpi1/Spi1, 3.3-fold, P = 3 × 10–11) is a master transcription factor that regulates the cell-specific response to TGFβ signaling (14). Transcription factor 21 (Tcf21, 3.2-fold, P = 3 × 10–31) is highly expressed in regions of myocardial fibrosis (15). Though upstream control of this molecule is unclear, some evidence suggests Tcf21 works in concert with the AP-1 transcription factor (16). JunB, a TGFβ-responsive member of the AP-1 transcription factor family (17), is also upregulated in DCM hearts (2.4-fold, P = 1 × 10–26). Also differentially expressed is E2F transcription factor 1 (E2f1, 2.9-fold, P = 3 × 10–5), which is known to play a role in the cardiac stress response (18) and is an important downstream effector of TGFβ in cancer and various fibrotic diseases (19, 20). The cyclic-AMP dependent transcription factor Atf3 (7.4-fold, P = 2 × 10–44), which promotes maladaptive cardiac remodeling and regulates cardiac fibrosis, was also upregulated in PLNR9C/+ (21, 22).
PLNR9C/+ mice also displayed striking upregulation of the stress-response proto-oncogenes c-myc (Myc, 8.7-fold, P = 2 × 10–52) and c-fos (Fos, 11.2-fold, P = 8 × 10–18) with DCM. Other developmental cardiac transcription factors, including Meox1 (6.1-fold, P = 1 × 10–131), Sox17 (2.8-fold, P = 3 × 10–18), (23) and Tbx20 (1.4-fold, P = 1.3 × 10–8), (24) also had increased expression in PLNR9C/+ nonmyocytes, indicating that DCM incited a broad shift in transcriptional control of nonmyocytes.
Nonmyocytes from DCM hearts also had markedly augmented transcription of a number of components of the Wnt signaling network, which plays important roles in cardiac embryogenesis as well as remodeling in the adult heart after myocardial infarction and with HF (25). Notable among these were the frizzled receptors Fzd1 (2.2-fold, P = 1 × 10–17), Fzd2 (3.3-fold, P = 3 × 10–32), and Fzd7 (2.6-fold, P = 1 × 10–19); the Wnt-signaling inhibitors–secreted frizzled-related protein-1 (Sfrp1, 7.3-fold, P = 3 × 10–132); Sfrp2 (50.8-fold, P = 2 × 10–174); and the Dickkopf Wnt signaling pathway inhibitor-3 (Dkk3, 12.1-fold, P = 2 × 10–77). There was increased expression of the downstream effector WNT1-inducible signaling pathway protein-2 gene (Wisp2, 58.4-fold, P = 3 × 10–253). Consistent with these data, pathway analyses identify enrichment of the canonical Wnt signaling pathway (Ingenuity pathway Wnt/β-catenin pathway, P = 0.005; GO term canonical Wnt receptor signaling pathway [GO:0060070], P = 0.0004) in the DCM nonmyocyte compartment. Both the increased proliferation of nonmyocytes and activation of inflammatory gene programs could account for the increased burden of myocardial fibrosis that accompanied progressive ventricular remodeling in PLNR9C/+ mice.
Cardiomyocytes. DCM and HF cardiomyocytes exhibited transcriptional changes indicating altered cell signaling and cell metabolism. Although WT cardiomyocytes had little expression of TGFβ isoforms in comparison with nonmyocytes, isolated PLNR9C/+ cardiomyocytes had substantially increased expression of Tgfb2 (13.0-fold, P ≤ 1 × 10–300) and Tgfb3 (4.5-fold, P = 1 × 10–58). Tgfb1 levels (–1.3-fold, P = 0.001) were unchanged in DCM cardiomyocytes (Figure 4C). Together, these data suggest there are distinct roles for the different TGFβ isoforms in cardiac remodeling. Interestingly, isolated cardiomyocytes from DCM mice also had increased expression of other proinflammatory cytokines, including growth/differentiation factor 15 (Gdf15, 21.6-fold, P < 1 × 10–300) and connective tissue growth factor (Ctgf, 5.3-fold, P < 1 × 10–300). Together, these findings suggest that cardiomyocytes have an active and dynamic role in profibrotic signaling in HF.
With the onset of DCM, PLNR9C/+ mice displayed a nearly uniform reduction in expression of genes controlling aerobic metabolism, and as mice developed overt HF, there were continued gene expression changes, with progressive activation of genes regulating glucose metabolism (Figure 5). Consistent with these changes in gene expression, pathway analyses identified prominent reductions in aerobic metabolism (fatty acid oxidation, citric acid cycle, oxidative phosphorylation) and increased glucose metabolism (pyruvate metabolism, glycolysis, glycogen metabolism, gluconeogenesis; Supplemental Tables 4–6 and Table 2). Data were confirmed by qPCR for selected nuclear-encoded mitochondrial oxidative phosphorylation genes in a separate cohort of 12 PLNR9C/+ mice with DCM or age-matched WT mice (Ndufs7 –1.9-fold, 5 × 10–5; Cox7a1 –2.2-fold, 2 × 10–4; Atp5a1 –1.8-fold, 2 × 10–4).
Heat maps of metabolic genes demonstrate marked dysregulation of myocyte metabolism. (A) Mitochondrial genes controlling oxidative phosphorylation were significantly downregulated with the onset of DCM. (B) There was also reduced expression of genes governing fatty acid β oxidation and the citric acid cycle with the onset of DCM. (C) By contrast, genes associated with glucose utilization became dysregulated with the onset of DCM, and their expression was progressively increased with overt HF. Data plotted are fold-change values versus age-matched WT controls (n = 3 mice pooled prior to RNA-seq).
In addition to a shift from aerobic metabolism to glucose utilization, there was evidence for an increase in nonoxidative glucose metabolism. Key genes in the production of N-acetyl-glucosamine by the hexosamine biosynthetic pathway were upregulated in isolated cardiomyocytes (Gfpt1 1.7-fold, P = 2 × 10–7; Gnpnat1 1.5-fold, P = 3 × 10–4; Pgm3 1.7-fold, P = 1 × 10–3), though it is noted that expression of the genes encoding the enzymes controlling protein O-glycosylation (Ogt, Mgea5) were not altered with the development of DCM or HF. Also enriched were the pentose-phosphate pathway and glycogen biosynthesis (Table 2).
The cardiomyocyte transcriptional changes in DCM and HF hearts predicted substantive changes in metabolic regulators, such as inhibition of signaling by the PPAR pathway (Figure 6A), including the master metabolic regulator Pparγ coactivator 1α (Ppargc1a). Indeed, the expression of many PPAR genes and key cofactors were significantly reduced with development of DCM and HF (Figure 6B). Data were confirmed by qPCR for selected PPAR family members in a separate cohort of 12 PLNR9C/+ mice with DCM or age-matched WT mice (Ppara –3-fold, P = 8 × 10–4; Ppargc1a –3.9-fold, P = 4 × 10–3; Ppargc1b –2.0-fold, P = 9 × 10–4). Metabolic pathways affected by PPAR signaling match those that are enriched in PLNR9C/+ mice with DCM and HF (Figure 6C).
Downregulation of PPAR signaling in PLNR9C/+ with development of DCM. (A) Upstream regulator analysis predicted transcription factors that were activated (gray, Z-score > 2) or inhibited (black, Z-score < –2) in PLNR9C/+ cardiomyocytes with DCM. (B) PPAR pathway genes and key cofactors were downregulated in PLNR9C/+ mice with development of DCM (*P < 0.001). Data plotted as natural log (ln) of fold-change (n = 3 mice pooled prior to RNA-seq). Bayesian P value corrected for multiple hypothesis testing (10). (C) PPARα and RXRα downregulation is predicted to affect a number of downstream metabolic pathways. Genes and pathways regulated by PPARα/RXRα that are downregulated (red) or upregulated (green) in PLNR9C/+ mice with DCM are highlighted. Dashed arrows indicate signaling networks not shown in detail. Adapted from Ingenuity canonical pathway PPARα/RXRα activation. †, representative cell surface receptors that signal to PPARα/RXRα.
In addition to these well-described metabolic regulators, we also identified increased expression of T-box transcription factor 15 (Tbx15) in DCM hearts (8.5-fold, P = 1 × 10–15). Interestingly, induction was exclusive to cardiomyocytes (11.7-fold, P = 8 × 10–27 in isolated cardiomyocytes, with no detectable expression in isolated nonmyocytes). Data were confirmed by qPCR in a separate cohort of 12 PLNR9C/+ mice with DCM or age-matched WT mice (8.5-fold, P = 1 × 10–3). In addition to its developmental functions, Tbx15 has been implicated in the regulation of glycolytic metabolism in skeletal muscle fibers (26). We suggest that the complementary activation of Tbx15 drives the increase in glycolytic metabolism in conjunction with the inhibition of PPAR signaling; together, this results in the profound metabolic changes that characterize DCM and HF.
Common and distinct pathways emerge in DCM and HCM
Using previously reported differential gene expression in a genetic mouse model of HCM (MHCR403Q) (10), we performed pathway analysis using the bioinformatics platforms described herein to define canonical pathways, GO terms, KEGG pathways, and upstream regulators in HCM nonmyocytes and cardiomyocytes, and we compared these with DCM datasets from PLNR9C/+ mice. HCM nonmyocytes were enriched in profibrotic and inflammatory/immune pathways while HCM cardiomyocytes were notable for substantial changes in metabolism (Figure 7A and Supplemental Table 7), patterns that were remarkably similar to those observed in DCM mice. Despite these similarities, the majority of differentially expressed genes were unique to one or the other cardiomyopathy (Figure 7B). Among nonmyocyte genes with differential expression, 39% were common and 61% were unique to DCM or HCM; only 28% of differentially expressed cardiomyocyte genes were shared, while 72% were unique to either DCM or HCM mice.
Transcriptional comparisons of DCM and HCM. (A) Enriched (FDR < 0.01) Ingenuity canonical pathways showed a predominance of inflammatory and cell remodeling pathways in nonmyocytes with perturbed cardiomyocyte metabolic pathways in both DCM and HCM. (B) Few differentially expressed genes in nonmyocytes and cardiomyocytes were common to both DCM and HCM. (C) TLR genes were upregulated in DCM but not in HCM. (D) Genes controlling mitochondrial oxidative phosphorylation and (E) those controlling fatty acid oxidation and the citric acid cycle were significantly downregulated, while (F) genes governing glucose metabolism were upregulated with the onset of DCM or HCM. Data plotted in heat maps are fold-change values versus age-matched WT controls (n = 3 mice pooled prior to RNA-seq).
To identify potentially distinctive pathways in these cardiomyopathies, we performed additional restricted pathway analyses that included only genes with (i) cardiomyopathy-specific differential expression (e.g., significantly changed only in DCM or HCM), or (ii) cardiomyopathy-selective differential expression (differential expression > 2.5-fold higher in one cardiomyopathy versus the other). Genes with specific or selective differential expression in nonmyocytes identified both a common and a unique suite of canonical pathways (Supplemental Table 8), transcriptional regulators, growth factors, and kinases (Supplemental Table 9) in DCM and HCM.
Nonmyocytes. Immune signaling pathways were specifically enriched in nonmyocytes from DCM but not HCM mice. Activation of TLR-dependent processes (P = 2 × 10–3) was inferred from DCM-specific upregulation of key TLR genes (Figure 7C and Table 3). DCM nonmyocytes also showed evidence for increased activation of ILs (IL-4 signaling, P = 5 × 10–5; IL-8 signaling, P = 4 × 10–4; IL-3 signaling, P = 2 × 10–3) and NFκ-B signaling (P = 7 × 10–5). These genes are generally not expressed in the heart (Supplemental Table 10); thus, pathway enrichment is either reflective of systemic inflammation or overlap with other inflammatory pathways, such as TGFβ. Inferred upstream regulators in DCM nonmyocytes included multiple interferon regulatory factor (IRF) family members and transcription factors involved in innate and adaptive immunity.
By contrast, nonmyocyte genes from HCM hearts predicted activation of a distinct suite of signaling pathways, including noncanonical Wnt signaling (Wnt-calcium signaling, P = 1 × 10–4), Her-2 signaling (P = 2 × 10–4), and protein kinase A signaling (P = 7 × 10–4), as well as others. Among the inferred regulators in HCM nonmyocytes are a host of signaling molecules including those governing Wnt signaling (e.g., Wnt3A, Wnt5A), cardiac hypertrophy (e.g., Erbb2, Akt, Src) and the cellular response to hypoxia (Arnt2).
To provide biologic evidence for distinct inflammatory signals in nonmyocytes from DCM and HCM mice, we assessed the effects of the angiotensin II receptor blocker, losartan, in DCM. Our prior studies showed that losartan reduced cardiac fibrosis in MHCR403Q mice by inhibiting angiotensin II–induced TGFβ-Smad signaling (27). PLNR9C/+ mice received continuous losartan treatment from weaning until death with longitudinal echocardiography and terminal assessment of fibrosis. Losartan treatment had no salutary effects on the age at onset of DCM, severity of LV dilation, nor on systolic dysfunction, progression to HF, or age at death. The degree of myocardial fibrosis in treated and untreated PLNR9C/+ mice was identical (Supplemental Figure 5).
Cardiomyocytes. While altered metabolic gene expression was identified in both DCM and HCM, there were some notable differences (Figure 7, D–F). First, though many key PPARs and cofactors were similarly downregulated in HCM (Ppara –2.5-fold, P = 6 × 10–16; Ppargc1a –1.7-fold, P = 1 × 10–6; Esrrg –1.5-fold, P = 0.005; Rxrg –2.6-fold, P = 9 × 10–15), others were not (Ppargc1b 1.0-fold, P = 0.96; Esrra 1.0-fold, P = 0.64; Rxra 1.1-fold, P = 0.41), suggesting differences in PPAR signaling between DCM and HCM. Second, a subset of aerobic metabolism genes remained unchanged in HCM (Figure 7E). Analysis of these genes showed enrichment for genes controlling peroxisomal metabolism (Hsd17b4, Hsd17b8/H2-Ke6, Peci, Abcd2, Ehhadh, Abcd3, Acsl4) and consistent with this, pathway analyses showed differential enrichment for peroxisome function between DCM and HCM at baseline (GO term peroxisome [GO:0005777]: DCM P = 4 × 10–5, HCM P = 0.012; KEGG pathway peroxisome [04146]: DCM P = 2 × 10–5, HCM P = 0.014). Finally, Tbx15 was not expressed in HCM; despite this, changes in expression of genes governing glucose metabolism were similar between DCM and HCM (Figure 7F).
Analysis of genes selectively or specifically expressed in either DCM or HCM cardiomyocytes revealed no pathways or upstream regulators that were selectively enriched in HCM. DCM cardiomyocytes showed predicted enrichment for dopamine receptor (P = 1 × 10–4) and β-adrenergic (P = 3 × 10–4) signaling, an observation that is consistent with profound myocardial dysfunction and impending HF. There was also evidence for the extensive metabolic changes in DCM cardiomyocytes with predicted inhibition of PPAR-pathway signaling via PGC1α (Z-score: –2.9, P = 9 × 10–7) and retinoid X receptor-α (Rxra, Z-score: –2.6, P = 3 × 10–3), a critical cofactor regulating PPARα function. In addition, this restricted set of DCM genes predicted inhibition of MED1 (Z-score: –2.4, P = 7 × 10–4), a subunit of the large TRAP/Mediator transcriptional complex that suppresses gene transcription programs associated with energy expenditure in muscle tissue (28). Notably, while Med1 is predominantly expressed in nonmyocytes at baseline (3.4-fold higher than isolated cardiomyocytes, P = 1 × 10–115), significant upregulation occurred in isolated DCM cardiomyocytes (1.5-fold, P = 2 × 10–14). As MED1 inhibition augments mitochondrial biogenesis, (28) this may be a compensatory mechanism in response to energy demands.
Using longitudinal transcriptome analyses, we defined the marked temporal changes in gene expression as PLNR9C/+ mice transition from preclinical disease to DCM and HF. By parsing gene expression into nonmyocyte and cardiomyocyte compartments, we identified lineage-specific changes. Nonmyocytes in PLNR9C/+ mice were proliferating and displayed increased transcription of genes that function in innate immunity and inflammation, with notable selective increases in TGFβ isoforms. Concurrently, cardiomyocytes showed profound dysregulation of cellular metabolism with reduced expression of key PPAR-pathway genes and a marked increase in Tbx15 transcription. By comparison of genes and pathways activated in DCM and HCM hearts, we identified cardiomyopathy-specific gene programs.
Although the PLNR9C/+ mutation is selectively expressed in cardiomyocytes, transcriptional analyses of nonmyocytes highlighted their early and central role in ventricular remodeling. In pre-DCM hearts, contractile performance was normal, yet we identified increased proliferation of nonmyocytes and altered gene expression, an observation that suggests early dysregulation of calcium homeostasis in mutant cardiomyocytes may promote paracrine signaling to nonmyocytes, possibly via cardiomyocyte-specific expression of cytokines. With the onset of DCM, bioinformatics programs deduced the activation of many inflammatory pathways in nonmyocytes. As genes involved in these pathways (e.g., IRF family members, matricellular proteins, TLRs, etc.) remodel the extracellular matrix of the heart in response to disease, these pathways are likely to increase myocardial fibrosis in DCM and HF (29–31). The TLR pathway was activated in DCM mice and has been implicated in ischemic and nonischemic HF in human patients (30, 32). TLR activation can also promote decreased expression of PGC1α and PGC1β (33), suggesting a possible link between immune signaling networks and metabolic dysregulation in HF. Notably, the TLR pathway was not activated in HCM, implying different remodeling and profibrosis mechanisms in these cardiomyopathies.
Collagen gene expression was markedly upregulated with progression to DCM and HF, translating into significant extracellular matrix remodeling that ultimately results in massive fibrosis. Stromal remodeling and the synthesis of collagen is a common late-phase step in both wound healing and cancer progression in response to activation of the immune system (34). Thus, immune system activation has overlapping responses between HF and other fibrotic diseases. However, the systemic inflammatory response observed in HF is not associated with significant persistence of infiltrating immune cells but rather results in activation of myofibroblasts that play a key role in fibrosis and stromal remodeling (35). Thus, the extensive immune signaling activation observed in cardiac nonmyocytes of PLNR9C/+ and MHC403 mice suggests that both paracrine and autocrine signaling mechanisms are involved.
We identified activation of canonical Wnt signaling (Wnt/β-catenin pathway) in both DCM and HCM. Wnt signaling (both canonical and noncanonical) plays a key role in cardiac development but is largely quiescent in the normal adult heart. However, Wnt signaling is notably altered with various cardiac pathologies including cardiac hypertrophy, myocardial infarction, and HF (25). Importantly, canonical Wnt signaling is necessary for TGFβ-induced fibrosis; hence, it is not surprising that this pathway is activated in both DCM and HCM mice (36). We also identified activation of noncanonical Wnt signaling (Wnt/calcium pathway) only in HCM. The role of noncanonical Wnt signaling pathways in cardiac disease are poorly defined, and it is possible this finding reflects differences in intracellular calcium homeostasis induced by these distinct genetic disorders. But TGFβ may also activate noncanonical signaling in a context-dependent fashion (37); thus, noncanonical Wnt signaling in HCM mice could be related to differences in TGFβ signaling, such as may be induced by different TGFβ isoforms.
Selective activation of TGFβ family members also distinguished fibrotic pathways in DCM and HCM. Mice with either DCM or HCM have increased expression of Tgfb2 and Tgfb3. However, HCM mice also have a significant, 4-fold increase in Tgfb1, while this isoform is not significantly elevated in DCM nonmyocytes. Consistent with possible selective activation of TGFβ isoforms, losartan-mediated inhibition of TGFβ1 was ineffective in reducing myocardial fibrosis in PLNR9C/+ mice, although it markedly attenuated fibrosis in HCM hearts (27). Increased expression of another TGFβ family member, Gdf-15, may also have contributed to the lack of effect of losartan in DCM hearts. Gdf-15 is secreted into the extracellular matrix, where it can translocate into cell nuclei, interrupt Smad complexes, and inhibit TGFβ-dependent Smad signaling pathways (38). Gdf15 was 47-fold increased in DCM (P < 1 × 10–300) but was not significantly differentially expressed in HCM (2.2-fold, P = 0.005), providing further support that profibrotic signaling is distinct in DCM and HCM.
PLNR9C/+ cardiomyocytes had minimal transcriptional changes until overt DCM was evident. But with onset of morphologic abnormalities, RNA-seq and subsequent qPCR revealed a marked shift in cardiomyocyte metabolic gene levels with decreased expression of genes associated with aerobic respiration and increased expression of those regulating glucose metabolism. Reduced expression of aerobic respiratory genes directly correlates with mitochondrial oxidative capacity in the heart; thus, the overwhelming number of aerobic respiratory genes downregulated with DCM and HF are presumed to be indicative of reduced mitochondrial respiration (39–41). Similarly, expression levels of genes regulating glucose metabolism correlate with metabolic flux (42, 43). Thus, we anticipate that the progressive increase in glycolytic gene expression is reflective of increased glucose utilization, possibly through increased nonoxidative glucose metabolism via glycogen synthesis, the hexosamine biosynthetic pathway, and the pentose phosphate shunt. Together, these data support the hypothesis that HF in PLNR9C/+ cardiomyocytes is associated with aerobic glycolysis, a phenomenon known as the “Warburg effect” that is common to many forms of cancer (44). Further studies are required to determine the degree of activation of these mechanisms, the usage and role of alternative metabolic substrates, and their collective contribution to DCM progression.
PPAR family member gene expression was depressed with onset of DCM, most notably Ppara, Ppargc1a, and Ppargc1b. PGC1α is a master regulator of mitochondrial biogenesis and regulates expression of genes governing oxidative phosphorylation in the heart (45). Overexpression (45, 46) or ablation of PGC1α in mice (39, 40) causes DCM with diminished cardiac performance and poor stress response — findings that imply tight regulation of PGC1α is critical to prevent maladaptive responses. PPARα is a key regulator of myocardial fatty acid oxidation (47), and mice deficient in PPARα have impaired contractile performance in response to hemodynamic stress (48, 49). Expression levels of key PPAR cofactors were also reduced, including RXRα (Rxra), which heterodimerizes with PPARα to evoke PPARα-induced transcriptional regulation, and estrogen-related receptor α (Esrra), which is induced by and coactivated by PGC1α (50, 51).
Concurrent with the inhibition of PPAR signaling, DCM cardiomyocytes exhibited markedly increased levels of Tbx15. This T-box transcription factor was recently found to regulate the composition of fast-twitch glycolytic fibers in skeletal muscle (26). Ablation of Tbx15 in mice reduced the numbers of glycolytic skeletal muscle fibers and whole body oxygen consumption, and it altered AMPK and IGF signaling. By extrapolation to our data, 8-fold increased levels of Tbx15 in DCM would substantially increase glycolysis in DCM and HF cardiomyocytes. We conclude that the stress-induced reciprocal activation of Tbx15 and attenuation of PPAR signaling accounts for the altered metabolic gene expression that characterized DCM and HF.
While these transcriptional data defined multiple molecules that are excellent candidates for propagating cardiomyopathic pathophysiology, our study has several limitations. Transcript levels do not always correlate with protein levels or activity and are thus an indirect measure of the activation state of transcription factors and regulatory molecules. Hence, these data are associative and would require further testing to perturb molecules and pathways of interest to derive conclusive mechanistic links. Despite normalization of RNA libraries, we recognize that proliferation of nonmyocytes confounds the precise interpretation of increased transcription among nonmyocyte genes. Further, the use of cyclosporine to induce robust hypertrophy in MHC403 mice alters basal expression of some genes and introduces bias when compared with PLNR9C/+, as does the difference in genetic background of these 2 mouse models. Bioinformatics tools also have limitations, as pathway analyses are predicated on previously reported findings that may not be directly relevant to the heart or these cardiomyopathies or may not include newly identified critical genes and mediators of disease. Finally, we recognize the need for extensive additional studies to validate the many new hypotheses generated by these data.
Mindful of these caveats, we conclude that a genetic model of DCM with altered calcium signaling incited a myriad of transcriptional changes in the heart that promote myocardial fibrosis and likely metabolic reprogramming. As calcium dysregulation is common to both ischemic and nonischemic HF, we suggest that the genes and pathways activated in PLNR9C/+ mice define potential therapeutic targets that are relevant to other forms of advanced heart disease. Based on comparative transcriptional analyses of DCM and HCM models, we suggest that these cardiomyopathies share some common pathways but also have distinct molecular programs. Strategies to selectively probe these unique genes and pathways should improve mechanistic insights into pathologic remodeling and clinical phenotypes that characterize DCM and HCM.
Mouse models
Transgenic mice overexpressing PLNR9C/+ under control of the α-cardiac myosin heavy chain promoter have been previously characterized (3). Male PLNR9C/+ mice were compared with WT male FVB control mice. Mice expressing the MHC R403Q/+ knock-in transgene and comparable age-matched 129sv/ev control mice have been previously characterized (9). Hypertrophy was induced in MHCR403Q/+ mice using oral cyclosporine fed to the mice in the chow as previously described (52).
Echocardiography
Mice were anesthetized with an isoflurane vaporizer (VetEquip) and attached to ECG leads on a Vevo Mouse Handling Table (VisualSonics Inc.). Chest hair was removed with depilatory cream. Transthoracic echocardiography was performed with heart rate approximately 500–550 bpm using a Vevo 770 High-Resolution In Vivo Micro-Imaging System and RMV 707B scan-head (VisualSonics Inc.). A single, experienced echocardiographer blinded to genotype acquired the images. Parasternal 2D images and M-mode images were acquired to assess LV chamber size and fractional shortening (calculated as FS = [LVEDD-LVESD]/LVEDD).
Quantification of myocardial fibrosis
Mice were anesthetized with isoflurane, and hearts were exposed by midline thoracotomy. After excision, serial sections of fixed, paraffin-embedded LV — as well as fresh whole mount heart tissue — were stained with Masson trichrome. Fibrosis was quantified as the ratio of fibrotic area to total ventricular area using Image J (NIH) as previously described (27).
IHC
PLNR9C/+ mice and age-matched controls were injected i.p. on 3 consecutive days with BrdU (Sigma-Aldrich) at 100 mg/kg body weight (BW) in PBS at week 8 or week 22. Mice were sacrificed 1 hour after the last injection. Hearts were harvested and then fixed in 4% paraformaldehyde, were paraffin-embedded, and were sectioned.
Sections were deparaffinized with xylenes and ethanol, rehydrated in deionized water and PBS, and then permeabilized with 0.1% Tween (BioRad) in PBS. Antigen retrieval was performed by heating slides in target retrieval solution (Dako) at 125°C for 30 minutes followed by enzymatic digestion with EcoRI and ExoIII. Sections were blocked in goat serum for 1 hour at room temperature (RT), incubated with anti-BrdU antibody (Dako, 1:100) at 4°C overnight, washed in PBS, and then incubated with anti–IgG Alexa Fluor 488 (Invitrogen) and anti–wheat germ agglutinin (WGA) Alexa Fluor 594 (Invitrogen, 1:500) at RT for 1 hour. After washing with PBS, sections were incubated for 5 minutes at RT with DAPI (Sigma-Aldrich, 1:10,000), then washed in PBS and mounted using Prolong Gold anti-fade reagent (Invitrogen). Microscopy was performed using a Leica SP2 confocal microscope.
BrdU-labeled cells were counted at ×40 magnification from 4 fields/section per animal; > 13,000 nuclei were counted per genotype. DAPI-labeled cells were counted at ×40 magnification using Image J. The percent of proliferating cells was calculated as the number of BrdU-labeled cells divided by total DAPI-stained nuclei. Myocyte cell size was measured using Image J. WGA stains highly glycosylated collagen, identifying extracellular matrix, and, therefore, myocyte cell margins. P values were calculated using χ2.
Tissue and cell harvesting for RNA analysis
Tissue isolation. For collection of whole LV tissue, hearts from PLNR9C/+ mice pre-DCM, with DCM and with overt HF, and age-matched controls were exposed by midline thoracotomy and immediately excised. Hearts were bathed in cold 1× PBS to remove blood and then subjected to left ventriculectomy followed immediately by RNA extraction (below).
Cell isolation. Myocyte and nonmyocyte cell populations were isolated from PLNR9C/+ mice with echocardiographically confirmed DCM and age-matched controls (n ≥ 6). First, hearts were exposed by midline thoracotomy, excised, and placed in cold calcium-free tyrode solution (140 mM NaCl, 5.4 mM KCl, 1 mM MgCl2, 0.33 mM NaH2PO4, 10 mM glucose, and 5 mM HEPES buffer pH 7.4). Retrograde coronary perfusion was established via aortic cannulation as described previously (53). The heart was perfused with enzyme buffer (in calcium-free tyrode solution: 0.3 mg/g BW collagenase D [Roche Diagnostics], 0.4 mg/g BW collagenase B [Roche Diagnostics], 0.05 mg/g BW proteinase XIV [Sigma-Aldrich]) for 15 minutes. The atria and right ventricle were removed, and the LV was minced into small pieces in transfer buffer (calcium-free tyrode solution + 5 mg/ml BSA) and then passed several times through a sterile pipette. The resulting cell suspension was passed through a mesh filter into a 50 ml centrifuge tube and incubated for 15 minutes at RT to allow myocytes to pellet by gravity. The pellet was serially resuspended every 10 minutes using a calcium gradient (tyrode solution + 5.5 mM glucose [pH 7.4] + serial calcium concentrations of 0.06 mM, 0.24 mM, 0.6 mM, 1.2 mM) to derive a cell fraction enriched in myocytes.
The supernatants from the filtered cell solution, as well as the supernatant from the 4 calcium gradient buffers, were pooled and centrifuged at 1,000 g for 10 minutes at 4°C. The pellet was resuspended in DMEM + 10% FBS, plated on sterile petri dishes, and incubated for 2 hours at 37°C in a humidified, 5% CO2 incubator. Plated cells were washed in sterile PBS to remove debris and nonadherent cells, resulting in a nonmyocyte cell population enriched in cardiac fibroblasts.
RNA isolation. RNA was extracted from excised LV or isolated myocyte and nonmyocyte cell populations using TRIzol (Invitrogen). To minimize biologic variation, RNA was pooled from n = 3 mice for whole LV tissue samples and n = 6 mice from isolated cell populations. cDNA libraries were constructed, and 5′ RNA-seq was performed as previously described (10). Libraries were sequenced on the Illumina platform and then aligned to the mouse reference sequence mm9 using Tophat (version 1.4.0). Of note, myosin heavy chain family genes (Myh1-Myh15) were selectively realigned using the STAR aligner because of frequent misalignment with Tophat secondary to the high degree of homology among these genes. The total number of reads of each transcript was normalized to 1 million, with P values of gene fold-changes determined as previously described (10).
qPCR. Total RNA (1 μg) from excised LV samples was treated with DNase I and then converted to cDNA (SuperScript III First-strand Synthesis System, Invitrogen). Target genes underwent qPCR with SYBR green (Tgfb1, Tgfb2, Tgfb3, Gdf15) on a 7500 Fast RT-PCR System (Applied Biosystems) or TAQ-man probes (Ppara, Ppargc1a, Ppargc1b, Tbx15, Ndufs7, Cox7a1, Atp5a1, Polr3h) using a BioRad QX200 Droplet Digital PCR System. Gapdh was progressively upregulated in PLNR9C/+ mice (pre-DCM, 1.2-fold, P = 1 × 10–5; DCM, 1.3-fold, P = 2 × 10–10; HF, 1.6-fold, P = 2 × 10–28). Hence, Polr3h was chosen as a control for normalization because of robust cardiac expression and minimal biologic variability between PLNR9C/+ and WT mice at all time points. See Supplemental Table 11 for probe and primer details.
Bioinformatics pipeline for pathway analysis. We defined genes as being differentially expressed if they met the following 3 criteria: (i) normalized read count > 2 copies/million reads in the RNA-seq library, (ii) fold-change > 1.33 (upregulated) or < –1.33 (downregulated), and (iii) P < 0.001 for the comparison of normalized read counts between mutant and control animals. Our bioinformatics pipeline is described in Supplemental Figure 4 and was executed using R (version 3.0.1). Culled lists representing those genes differentially expressed in nonmyocytes or myocytes were then subjected to pathway analysis using 2 bioinformatics platforms: IPA (Qiagen) and GO-seq (Bioconductor package, “Goseq”).
IPA. IPA uses proprietary algorithms and a proprietary database of gene information to generate a number of different analyses based on a set of user-identified genes (http://www.ingenuity.com/products/ipa). We used IPA to identify enriched canonical pathways and to predict possible upstream regulatory molecules from our RNA-seq data.
Only genes identified as differentially expressed (Supplemental Figure 4) were uploaded into IPA. We used the following IPA core analysis settings: general settings, reference set = Ingenuity knowledge base (genes only), relationships to consider = direct and indirect relationships; data sources = all; confidence = experimentally observed only; species = all. Some gene symbols in our mm9 database were not identifiable by IPA and were consequently not used in the analysis (range 12%–19% for different datasets).
We first reviewed Ingenuity canonical pathways to study known biological pathways and processes that were enriched in our differentially expressed genes. Specific canonical pathways of interest were assessed for significance based on the nominal P value as determined by Ingenuity software, which is ascertained by a right-tailed Fisher exact test. This P value is a measure of the likelihood that the association between a particular canonical pathway and our differentially expressed genes is due to chance. A nominal P value of < 0.05 was considered significant, when considering specific pathways of interest. To control for the rate of false discoveries when analyzing all significantly enriched pathways for a given set of genes, significance was based on the P value corrected for multiple hypothesis testing (Benjamini-Hochberg). In this case, the P value represents the FDR. We used a FDR < 0.01 as significant. The canonical pathways were generated through the use of QIAGEN’s IPA (http://www.qiagen.com/ingenuity). For more details on calculation of P values by IPA, the interested reader is referred to the white paper, “Calculating and Interpreting the P values for Functions, Pathways and Lists in IPA” (http://www.ingenuity.com/wp-content/themes/ingenuity-qiagen/pdf/ipa/functions-pathways-pval-whitepaper.pdf).
We also used the IPA upstream regulator tool. This attempts to predict the functional status of molecules (upstream regulators) such as transcription factors, kinases, and growth factors based on known downstream targets (the input set of differentially expressed genes). Two metrics of significance were used. The overlap P value represents the significance in the overlap between the identified differentially expressed genes (the known downstream targets) and all genes associated with a particular upstream regulator in the Ingenuity database. We used a P value threshold of < 0.01 as significant. The Z-score also accounts for the direction of activation of the differentially expressed genes in relation to published literature with respect to a particular upstream regulator, thereby allowing a prediction of whether the upstream regulator is activated or inhibited. Four possibilities exist: (i) a gene known to be activated by an upstream regulator that is upregulated in the dataset predicts activation of the upstream regulator; (ii) a gene known to be activated by an upstream regulator that is downregulated in the dataset predicts inhibition of the upstream regulator; (iii) a gene known to be inhibited by an upstream regulator that is upregulated in the dataset predicts inhibition of the upstream regulator; and (iv) a gene known to be inhibited by an upstream regulator that is downregulated in the dataset predicts activation of the upstream regulator. The Z-score accounts for both the total number of genes differentially expressed and the direction of activation; a Z-score ≥ 2 (activated) or ≤ –2 (inhibited) was considered significant. Additional details can be found on the IPA website (http://pages.ingenuity.com/rs/ingenuity/images/0812 upstream_regulator_analysis_whitepaper.pdf).
Goseq. Goseq is a free, R-software–based program that allows for identification of GO terms and KEGG pathways from a set of user-identified genes set against a customizable, user-identified background dataset as described previously (54). R packages Goseq (v. 1.12.0), GO.db (v. 2.9.0), and org.Mm.eg.db (v. 2.9.0) were used. In brief, differentially expressed genes were identified within a vector of all genes expressed in LV tissue (normalized read count > 2 copies/million reads in either WT or PLNR9C/+ RNA-seq libraries), which served as the background dataset, at a given stage of disease. Goseq was then used to identify significantly enriched GO terms and KEGG pathways. P values were adjusted for multiple hypothesis testing (Bonferroni) and a FDR < 0.05 was considered significant.
Statistics
Data are presented as mean ± SD. Between groups, differences were calculated using a 2-tailed Student’s t test or a Bayesian P value as described previously for RNA-seq data (10). Statistical significance for pathway analysis tools were calculated as described in the above referenced links; P values were corrected for multiple hypothesis testing where appropriate and are represented as the FDR. P values < 0.05 were considered significant except as noted (P < 0.001 for RNA-seq expression levels, FDR<0.01 for IPA canonical pathways).
Study approval
Studies were approved by the Animal Care and Use Committee of Harvard Medical School. All care and procedures were performed as per these guidelines.
MAB designed analyses, conducted experiments, acquired and analyzed data, and drafted the manuscript. SC designed analyses, conducted experiments, and acquired and analyzed data. HW conducted experiments and analyzed data. JMG and DCC conducted experiments. DAC provided reagents and assisted with animal care. MGP and SRD processed and analyzed data. SE and TK assisted with experimental design and execution. JGS and CES designed analyses, analyzed data, and drafted the manuscript.
This work was supported in part by the Clinical Skills Development Core Training Grant (NHLBI U10HL110337; to M.A. Burke), the John S. LaDue Memorial Fellowship in Cardiology at Harvard Medical School (to M.A. Burke), the NIH (R01 HL080494-08, R01 HL084553-06A1, to J.G. Seidman) the Howard Hughes Medical Institute (to C.E. Seidman), and the LeDucq Foundation (to J.G. Seidman and C.E. Seidman).
Address correspondence to: Michael A. Burke, WMB 322, Cardiology Division, Emory University School of Medicine, 101 Woodruff Circle, Atlanta, Georgia 30322, USA. Phone: 404.712.2690; E-mail: michael.burke@emory.edu. Or to: Christine E. Seidman, NRB 256, Department of Genetics, Harvard Medical School, 77 Ave Louis Pasteur, Boston, Massachusetts 02115, USA. Phone: 617.432.7871; E-mail: cseidman@genetics.med.havard.edu.
M.A. Burke’s present address is: Cardiology Division, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia, USA.
Conflict of interest: J.G. Seidman and C.E. Seidman are founders and owns shares in Myokardia Inc., a startup company that is developing therapeutics that target the sarcomere.
Reference information: JCI Insight. 2016;1(6):e86898. doi:10.1172/jci.insight.86898.