Skeletal muscle transcriptional dysregulation of genes involved in senescence is associated with prognosis in severe heart failure

Figures

• 

  Fig. 1. Flowsheet of patient and control-subject recruitment, characterization, and follow-up. a Heart failure patients with HFrEF, clinically stable with severe exercise intolerance (NYHA III) but with no life-threatening comorbidities were asked to participate in the study at the outpatient clinic of the Karolinska University Hospital. In total, 84 patients were willing to give up a skeletal muscle biopsy and undergo all diagnostic procedures associated with the study and were followed for 60 months (about 5 years) regarding all-cause mortality. Of those, 12 patients did not fulfill echocardiographic inclusion criteria, 14 patients dropped-out due to not being able to undergo all clinical characterization or issues with skeletal muscle sampling, RNA-extraction, or quality. b A control group was recruited in parallel with heart failure patients from individuals admitted to the clinic with signs and symptoms indicative of heart failure but did not meet criteria of heart failure based on the echocardiography and normal NT-proBNP levels (LVEF > 55% and NT-proBNP <300 ng/ml). Control group underwent the same clinical characterization as the heart failure patients with HFrEF. The external…

• 

  Fig. 2. Skeletal muscle transcriptional analysis workflow. Transcripts detected in the muscle samples originating from the heart failure patients (a) and where skeletal muscle expression could be confirmed in either the STRRIDE or GTEx cohorts (b) constituted the starting point for the co-expression analysis. Initial network was thereafter constructed based on weighted co-expression of transcripts in the heart failure cohort and the network was organized into communities based on the connectivity (i.e., number of significant correlations) between transcripts across patients (c). To ensure external and biological validity, and to minimize the risk of false discoveries, the initial network was pruned from edges (i.e., gene-gene correlations) that could not be confirmed in the independent skeletal muscle datasets from STRRIDE and GTEx. The validity of the communities in the network was also confirmed by calculation of module-preservation statistics and communities of the network that was not preserved in the independent cohorts were discarded from further analysis (d). The communities of the final network were characterized through gene-ontology, pathway and genome-scale metabolic modelling…

• 

  Fig. 3. Final transcriptional network and network characterization. The final co-expression network derived from human skeletal muscle samples of patients with severe heart failure consisted of 1481 transcripts (dots). The network was divided into four branches with a total of 14 distinct communities (denoted by color) of transcripts with strong mutual correlations (lines) and with confirmed significant external validity in at least one of the two validation cohorts representing transcriptionally profiled skeletal muscle of >600 individuals. Each network community was characterized with regards to biological processes, canonical pathways and metabolites through enrichment analyses and the communities were labeled based on the most highly enriched term. The communities of the final network were thereafter tested for enrichment of genes established to be regulated in clinical conditions with relevance in the context of heart failure: Physical deconditioning through prolonged bed rest, regular exercise training, cancer cachexia and cellular senescence (denoted with icons). Created with icons from BioRender (2025; https://BioRender.com/4pw1fl6).

• 

  Fig. 4. Associations between network communities and clinical characteristics in patients with heart failure. Community-level gene expression (eigengene) was compared between patients and controls using a two-sided Welch’s two-sample t-test (outer ring, denoted by squares). Five communities differed significantly, of which all but one (p53-signalling) had lower expression in patients compared with controls. Interindividual variation in gene expression of the network communities were also correlated with clinical traits in heart failure using Pearson correlation analysis (inner ring, denoted by circles). Characteristics of interest were plasma NT-proBNP, cardiac systolic function in the form of LVEF, exercise capacity in the form of peak VO2, daily physical activity assessed though accelerometer measurements and time to death (TTD) in the form of time from baseline characterization to mortality event. Three communities, p53-signalling, mitochondrial FA beta-oxidation and ECM remodeling were both differentially regulated in patients compared to controls and showed significant associations with≥2 clinical traits. Created with icons from BioRender (2025; https://BioRender.com/5sel492).

• 

  Fig. 5. Skeletal muscle gene expression in heart failure with regards to clinical traits. In depth analysis of community-level gene expression (eigengene) in relation to clinical traits in heart failure cohort. a The ECM community was down-regulated in heart failure patients compared with controls (P = 0.004) and GTEx subjects (P = 0.014) and was positively associated with higher levels of daily physical activity. b The mitochondrial FA beta-oxidation community was down-regulated in patients compared with controls (P = 0.006) and GTEx subjects (P = 0.017) and was successively higher expression was associated with lower levels of NT-proBNP. c The p53 signalling community had significantly higher (P = 1.30 × 10−4) transcriptional expression in heart failure patients compared with controls and in relation to propensity-matched subjects from the GTEx cohort (P = 6.84 × 10−7). Transcriptional expression was associated with lower LVEF and successively higher levels of NT-proBNP. All tests are two-sided Welch’s two-sample t-tests. Created with icons from BioRender (2025; https://BioRender.com/4pw1fl6).

• 

  Fig. 6. Prognostic impact of skeletal muscle gene-expression in heart failure with regard to mortality. Community-level gene expression (eigengene) was divided into terciles and assessed for prognostic value using Cox regression; a No significant association was observed for ECM remodelling and mortality (HR = 0.59, P = 0.148) however, higher expression was related to better prognosis. b In contrast, significant association with prognosis was observed for mitochondrial FA beta-oxidation community where higher expression was associated with better prognosis (HR = 0.35, P = 0.007). c Expression of the p53 signalling community was also associated with prognosis (HR = 3.44, P = 8.81 × 10−4) where higher gene-expression was associated with worsening prognosis. Wald tests were used to evaluate the Cox-models.

• 

  Fig. 7. External validation of the prognostic value of skeletal muscle gene-expression in relation to cause of death. a–c Gene-expression of the p53 network community in relation to health status and circumstances of death: The GTEx dataset provides information the prevalence of cardiac disease and the time foregoing death on all samples so-called hardy-scale which divides patients into categories as follows; Sudden unexpected death where subject can be assumed to be previously healthy (Violent). Sudden unexpected deaths of people who had been healthy (Fast), after a terminal phase estimated at <1 hr. Death after a terminal phase of 1 to 24 hrs, patients who were ill, but death was unexpected (Intermediate). Death after prolonged illness, with a terminal phase longer than 1 day i.e., deaths that are not unexpected (Slow). d–f Skeletal muscle gene-expression of network communities in patients >60 years of age with or without established cardiovascular disease prior to death in the GTEx cohort (n = 252). Statistical inference was done using one-way ANOVA.