Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 Sep 2;15(9):3009-28.
doi: 10.1021/acs.jproteome.6b00149. Epub 2016 Aug 3.

Integrated Omic Analysis of a Guinea Pig Model of Heart Failure and Sudden Cardiac Death

Affiliations
Free PMC article

Integrated Omic Analysis of a Guinea Pig Model of Heart Failure and Sudden Cardiac Death

D Brian Foster et al. J Proteome Res. .
Free PMC article

Abstract

Here, we examine key regulatory pathways underlying the transition from compensated hypertrophy (HYP) to decompensated heart failure (HF) and sudden cardiac death (SCD) in a guinea pig pressure-overload model by integrated multiome analysis. Relative protein abundances from sham-operated HYP and HF hearts were assessed by iTRAQ LC-MS/MS. Metabolites were quantified by LC-MS/MS or GC-MS. Transcriptome profiles were obtained using mRNA microarrays. The guinea pig HF proteome exhibited classic biosignatures of cardiac HYP, left ventricular dysfunction, fibrosis, inflammation, and extravasation. Fatty acid metabolism, mitochondrial transcription/translation factors, antioxidant enzymes, and other mitochondrial procsses, were downregulated in HF but not HYP. Proteins upregulated in HF implicate extracellular matrix remodeling, cytoskeletal remodeling, and acute phase inflammation markers. Among metabolites, acylcarnitines were downregulated in HYP and fatty acids accumulated in HF. The correlation of transcript and protein changes in HF was weak (R(2) = 0.23), suggesting post-transcriptional gene regulation in HF. Proteome/metabolome integration indicated metabolic bottlenecks in fatty acyl-CoA processing by carnitine palmitoyl transferase (CPT1B) as well as TCA cycle inhibition. On the basis of these findings, we present a model of cardiac decompensation involving impaired nuclear integration of Ca(2+) and cyclic nucleotide signals that are coupled to mitochondrial metabolic and antioxidant defects through the CREB/PGC1α transcriptional axis.

Keywords: heart failure; hypertrophy; mass spectrometry; median sweep; metabolic bottleneck; metabolome; pathway analysis; proteome; proteomics; transcriptome.

Conflict of interest statement

Notes

The authors declare no competing financial interest.

The microarray data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (GEO) and are accessible through GEO series accession number GSE78077 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE78077). Proteomic data sets, including spectral data files (.RAW), peak lists (.mgf), search results (.msf), and the Cavia porcellus protein sequence database (.FASTA), have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD003980

Figures

Figure 1
Figure 1
Study design.
Figure 2
Figure 2
Quantitative proteomic data set. (A) Venn analysis of protein identification across three independent iTRAQ experiments. Data-dependent sampling of peptides for MS2 spectral acquisition yielded distinct but overlapping proteomes. Empirical Bayesian statistical methods can accommodate sampling-related absence and, therefore, analysis of relative abundance, provided that the protein is observed in two or more independent experiments (denoted with *). (B) Principal component analysis of protein abundances across the data set. Component 1 differentiates distinct global protein abundance biosignatures of sham, HYP, and HF hearts. (C) Volcano plot providing a visual representation of differential protein regulation in HF, for which a detailed table is found in panel 2 in the Supporting Information. Dark-red circles represent significantly regulated (p < 0.05) proteins identified in all three experiments. Light-red circles denote those identified in two out of three experiments. The use of EB-modified p values obviates arbitrary fold-change thresholds that undermine FDR assessment. (D) Corroboration of fold changes (FCs) determined by mass spectrometry with those determined by immunoblot analysis.
Figure 3
Figure 3
Top regulated proteins and transcripts. The proteome (A–C) and the transcriptome (D–F) were parsed into three categories, those regulated (p < 0.05) in HYP only, in HYP and HF, and in HF only. Red symbols and traces denote entities upregulated in the specified condition. Blue symbols and traces denote entities that are downregulated in the specified condition. Transcripts and proteins were mapped to their respective genes, and the top genes in each category are listed in the right-hand panel.
Figure 4
Figure 4
Pathway analysis. (A) Pathway analysis of transcripts and proteins significantly regulated in both HYP and HF. (B) Analysis of transcripts and proteins significantly regulated in HF only. Where sufficient information exists, pathways are colored according to inferred activation or inhibition on the basis of their z scores (see the Supporting Information). Warm colors (orange, brown) denote inferred pathway activation. Cool colors (blue shades) denote inferred pathway inhibition.
Figure 5
Figure 5
Upstream signaling. (A) Upstream signaling analysis of transcripts and proteins significantly regulated in both HYP and HF. Exemplar signaling networks are shown. The TNF network was inferred from the transcriptome, whereas the TGFB1 network was inferred from the proteome. (B) Analysis of transcripts and proteins significantly regulated in HF only. The INFG network was inferred from the proteome and the BDNF network was inferred from the proteome. Where sufficient information exists, pathways are colored according to inferred activation or inhibition on the basis of their z scores (see the Supporting Information). The peripheral nodes represent the HF/sham ratios from the Supporting Information. The colors of the peripheral nodes are given by the red/green spectrum, where red denotes upregulation and green denotes downregulation. The central node indicates the inferred upstream signal that would explain the ratios represented in the peripheral nodes. The dotted lines linking the central node to the peripheral nodes indicate indirect relationships between the signal and the ultimate transcript or protein ratios. The colors of the dotted lines indicate inferred activation or inhibition of signaling. Yellow dotted lines indicate a transcript or protein ratio at odds with the predicted effect of signaling.
Figure 6
Figure 6
Upstream transcriptional programs. (A) Upstream transcriptional program analysis of transcripts and proteins significantly regulated in both HYP and HF. The top three programs in the transcriptome (according to log(p value)), KLF2, FOS, and NEUROG1, are shown as a network. The top three transcriptional programs inferred from proteomic data are HIF1A, ATF4, and DDIT3. (B) Transcriptional programs inferred from transcripts and proteins significantly regulated in HF only. Where sufficient information exists, pathways are colored according to inferred activation or inhibition on the basis of their z scores (see the Supporting Information). The top 11 activated transcriptional programs (from z scores) based on transcriptome data are shown to illustrate the degree of redundancy between programs. The top six activated programs based on the proteome are shown. It should be noted that the cases of ING1 and NFKB1A illustrate how activation of the program is consistent with repression of downregulation of the protein (blue lines). The network specifications are the same as in Figure 5, with the exception that the solid lines between the central and peripheral nodes indicate direct regulatory relationships.
Figure 7
Figure 7
Emerging network underlying guinea pig HF. Pathway analysis and upstream activator analysis predicted cAMP and retinoic acid deficiencies in guinea pig HF. Here we indicate the inferred impact on cAMP- and RA-responsive transcriptional programs, including Creb, CREBBP, RXRA, PGC1a, BDNF, and MEF2C. Inhibition of the programs (top; blue) is consistent with the proteomic profile observed in HF (downregulation in green, upregulation in red). CP denotes canonical pathway involvement. Tx denotes inferred toxic or pathological sequelae.
Figure 8
Figure 8
Concordance and divergence between the transcriptome and proteome in HF. (A) Fold changes in HF for the proteome and transcriptome exhibit a weak correlation (Pearson’s R = 0.48). (B) Concordant/divergent directions of fold changes. One-third of genes display divergent fold-change directions in HF (see panel 18 in the Supporting Information).
Figure 9
Figure 9
Characterization of the metabolome in HYP and HF. (A) Partial least-squares discriminant analysis reveals the distinctive nature of the metabolome in HF. The metabolome in HYP was less differentiated relative to the sham group. (B) Box plots representing the metabolite trajectories in HYP and HF, including metabolites up- or downregulated in HF only and metabolites progressively downregulated in HYP and HF. (C) Metabolite set enrichment analysis (MSEA) of metabolite regulation in HF. The x axis shows fold enrichment, and the color spectrum represents the span of Holm’s FWER-corrected p values. (D) Metabolite pathway analysis (MetPA) incorporating MSEA (y axis) and pathway topology (x axis). The size of the circles is proportional to the number of compounds identified in a given pathway, and the color reffects the Holm’s FWER-corrected p value. Selected pathways are highlighted. The entire set of pathways identified by MSEA and MetPA can be found in panels 23 and 24 in the Supporting Information.
Figure 10
Figure 10
Integration of the proteome and metabolome. (A) Combined GSEA/MSEA analysis of protein (mapped to the gene) and metabolite ratios in HF relative to sham. Dark gray bars indicate relative fold enrichment, and light gray bars indicate the impact of pathway topology. The composite score is simply the sum of the enrichment score (−log(p value)) and the topology score. (B) Schematic representation of proteins and metabolites underlying the fatty acid metabolism pathway. Squares denote metabolites, and ovals indicate gene names of proteins implicated in HF. Red indicates upregulation, and green denotes downregulation. The combined data suggest a bottleneck immediately prior to mitochondrial fatty-acylcarnitine import. (C) Schematic representation of proteins and metabolites of the citrate (TCA) cycle. ** denotes p < 0.05, and * denotes p < 0.075. The combined protein and metabolite data imply TCA cycle flux and/or anaplerosis may be inhibited in guinea pig HF.
Figure 11
Figure 11
Model for excitation–transcription coupling responsible for maladaptive metabolic and antioxidant remodeling in HF. Deficits in β-adrenergic signaling and Ca2+ handling are hallmarks of HF that are recapitulated in the guinea pig model. These deficits are expected to diminish the activity of protein kinase A (PKA) and Ca2+ calmodulin-dependent kinases. Pathway analysis suggests an analogy to neuronal signaling, in which PKA and CAMK activity have been shown to govern the phosphorylation state and activation status of the cAMP-responsive element (CRE) binding protein (CREB), a transcriptional coactivator. Since CREB is a documented transcriptional coactivator of PGC1α (itself a coactivator) and other genes (e.g., RYR2, BDNF), impaired CREB activation by CaMK/PKA would impinge on PGC1α-dependent gene programs, among them PPAR/RXR-mediated activation of genes for fatty acid metabolism (e.g., CPT1B, β-oxidation enzymes). PGC1α also participates in NRF 1- and 2-mediated activation of genes responsible for mitochondrial homeostasis (e.g., TFAM) and antioxidant function (e.g., SOD2 and PRDX3). Finally, antioxidant defenses suffer a double blow. Not only are antioxidant proteins downregulated, but also, acute Ca2+/Na2+ dysregulation abrogates mitochondrial TCA-cycle-dependent NADPH production, which is required to sustain the activity of thiol-bearing antioxidant enzymes.

Similar articles

See all similar articles

Cited by 13 articles

See all "Cited by" articles

Publication types

Feedback