Systems Biology Approach Uncovers Candidates for Liver-Heart Interorgan Crosstalk in HFpEF

Heart failure with preserved ejection fraction (HFpEF) accounts for more than half of heart failure cases worldwide. Obesity and metabolic syndrome are major drivers of HFpEF pathophysiology resulting in a complex systemic syndrome with far-reaching clinical implications involving extra-cardiac organs. Owing to their shared metabolic pathogenetic substrate, accumulating evidence support that HFpEF, and metabolic dysfunction-associated steatotic liver disease (MASLD)/nonalcoholic fatty liver disease (NAFLD) do not simply co-exist but are directly linked. Hence, investigating the systemic molecular dynamics of interorgan crosstalk between the liver and the heart might represent a promising way to understand HFpEF pathogenesis in the wider context of cardiometabolic diseases.

We simultaneously conducted bulk RNA sequencing on hearts and livers of 8-week-old, C57BL/6N male mice subjected to HFpEF-inducing regimen for eight weeks, comparing them to mice fed a control diet, as previously described¹ (CHOW versus HFpEF mice, n=5 each). All experiments involving animals were approved by the local authorities (Landesamt für Gesundheit und Soziales, Berlin, Germany; protocol no. G0104/20) and performed in agreement with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines and guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.

To identify tissue-specific, secreted mediators of liver-heart crosstalk in HFpEF, we adopted a bioinformatic approach² (Figure A). This approach estimates the correlation between transcriptomics data across different tissues, where one tissue, serves as a donor and another as an acceptor. The underlying assumption is that if the transcription level of a gene encoding for a secreted protein in the donor tissue correlates with the transcription level of relevant genes in the acceptor tissue, the given protein may serve as a mediator of crosstalk between the two tissues³.

Figure. Identification of circulating mediators of interorgan crosstalk in HFpEF and MASLD/NAFLD.

A. Schematic representation of the research strategy employed for the identification of the liver-heart interactions in HFpEF. B. Top. Volcano plots showing the distribution differential gene expression (DGE) analysis between CHOW and HFpEF hearts (top) and livers (bottom). Bottom. Heatmaps exhibiting the differential gene expression (DGE) between CHOW and HFpEF hearts (top) and livers (bottom), (n=5, each). The liver heatmap displays secretory DEGs only. DEGs are defined as those with |log2FC| > 0.5 and padj.<0.05. RNAseq data are available in the online data repository: https://zenodo.org/records/12794566 C. Top. Heatmaps illustrating the strong, significant biweight midcorrelation (|r|>0.8 for HFpEF, |r|>0.7 for HMDP, padj.<0.05 – FDR method for multiple testing correction was used) values between liver secretory DEGs and cardiac DEGs in the HFpEF dataset (left) the HMDP dataset (right). Non-significant and non-strong correlations are set to zero value. Rows represent the secretory liver DEGs, while columns represent the cardiac DEGs. Liver genes in both heatmaps are ranked based on the total number of their strong significant correlations with cardiac DEGs. Twenty-two of liver secretory DEGs in the HFpEF dataset that were not found in the HMDP dataset are showed at the bottom of the HFpEF dataset heatmap. Cardiac DEGs are clustered according to their similarity (Euclidean distance). The relative expression of each gene in the two tissues is indicated by log2FC data (blue, downregulated; red, upregulated). Bottom. Pie chart displaying the number of liver secretory genes present in the HMDP dataset, categorized as either mouse HFpEF-specific or mouse obese-specific based on the number of correlations with heart DEGs in the HMDP dataset. D. Left. Venn diagram depicting the intersection of three sets: mouse obesity-related genes, mouse HFpEF-specific genes, and differentially regulated proteins in human HFpEF plasma proteome. Seven of the mouse HFpEF-specific genes were also identified as dysregulated in the human HFpEF plasma proteome and defined as mouse/human HFpEF-specific genes. Right. Venn diagram illustrating the intersection of the seven mouse/human HFpEF-specific genes and genes exhibiting a correlation between their liver transcription levels and plasma levels of their encoded protein in human NAFLD subjects. Only two genes, Saa1 and Saa4, appear in both sets. E. Left. Bar plots showing Saa1 (up) and Saa4 (down) mRNA levels in the heart (left), white adipose tissue (WAT, middle), and liver (right) in CHOW and HFpEF mice. Right. Bar plot showing SAA1 protein serum level in CHOW and HFpEF mice. ns: not significant, p: p-value, Mann-Whitney test. F. Sankey diagram representing the top gene ontology (GO) terms significantly enriched (padj.<0.05) among the heart DEGs which are strongly and significantly correlated with liver Saa1/4 in HFpEF mice. The nodes representing heart DEGs are colored according to their log2FC and sorted alphabetically.

Differential gene expression analysis revealed 287 differentially expressed genes (DEGs) in the heart and 729 DEGs in the liver of HFpEF mice compared to CHOW (DEGs; |log2FC|>0.5 and padj.<0.05) Figure B, top). Although our sample size was relatively small, we observed robust differences in DEGs between the two groups. Given that potential mediators of crosstalk in HFpEF are anticipated to be secreted by the liver, we focused on the unique 86 liver DEGs encoding proteins annotated as secreted in the UniProt database – designated as liver secretory genes (Figure B, bottom). We calculated the biweight midcorrelation coefficients between liver secretory DEGs and heart DEGs in HFpEF mice and determined their statistical significance using p-values adjusted for multiple testing by the false discovery rate (FDR) method. We then ranked all 86 liver secreted candidates based on the number of strong (|r|>0.8) and significant (padj.<0.05) correlations they had with cardiac DEGs (Figure C, left).

As obesity significantly influences the HFpEF phenotype, we aimed to distinguish liver-heart crosstalk mediators specific to HFpEF, rather than those solely attributable to obesity. To achieve this, we applied the same bioinformatic pipeline on genetic variation within the Hybrid Mouse Diversity Panel (HMDP) which encompasses liver and heart transcriptomic data from approximately 100 well-characterized strains of mice, fed either CHOW or high fat-high sucrose (HFHS) diet for eight weeks³. Liver secretory genes that correlated with cardiac DEGs in HFpEF were tested in the HMDP dataset (Figure C, right). Among these, 22 genes were not found in the HMDP liver transcriptomic data. 26 genes (Angptl4, Cd9, Cpxm1, Enpp1, Esm1, Fgb, Fgf2, Fgg, Hbegf, Il6ra, Lect2, Loxl4, Mmp12, Nid2, Orm1, Papln, Prg4, Ptgds, S100a9, Shh, Snca, St6gal1, Tgfbi, Trem2, Vcam1, Wfdc2) exhibited strong (|r|> 0.7) and significant (padj.<0.05) correlation with 90 or more of the 287 cardiac DEGs in the HMDP dataset, suggesting that changes in their expression/secretion are linked to obesity status – these are termed as mouse obesity-related genes (Figure C, bottom). 38 genes (Adamts10, Apcs, Apoa4, Apom, Ccl9, Ces4a, Col15a1, Creg1, Creld2, Crlf2, Cspg5, Cxcl13, Dkk4, Enpp2, Fga, Fgf1, Gpc1, Gpx6, Hamp2, Il15ra, Itih5, Lcn2, Manf, Omd, Plin4, Pltp, Prss8, Qsox1, Saa1, Saa2, Saa3, Saa4, Serpina3m, Spon2, Tnxb, Vwa8, Wif1, Wnt5b), showed less than 90, and most often only a few, significant correlations with cardiac DEGs in the HMDP dataset, indicating that they are uniquely dysregulated in HFpEF, independent of obesity – these are termed mouse HFpEF-specific genes (Figure C, bottom).

To explore the potential translational relevance of liver secretory genes identified as dysregulated in an HFpEF preclinical model, we analyzed an available plasma proteomic dataset of human subjects with HFpEF⁴. Among the 130 dysregulated plasma proteins in human HFpEF, 7 of them (Apcs, Apoa4, Apom, Fga, Qsox1, Saa1, Saa4) were among the 38 mouse HFpEF-specific liver secretory genes – these are termed mouse/human HFpEF-specific genes (Figure D, left). Collectively, these analyses indicated unique, obesity-independent, liver-heart mediators in HFpEF across species.

We next sought to explore the conserved relationships between HFpEF and MASLD/NAFLD, two tightly linked cardiometabolic outcomes using a proteo-transcriptomic dataset from NAFLD patients⁵. In addition, using this approach, we tested the assumption that increased tissue transcription corresponds to increased plasma secretion of the respective protein. Among the 4,584 plasma proteins detected in the entire cohort, 190 liver gene transcripts exhibited significant correlation (p<0.01) with their plasma protein levels. Of the 7 mouse/human HFpEF-specific liver secretory genes identified in both preclinical and clinical HFpEF datasets, only Saa1 and Saa4, showed a significant correlation between their transcription levels and plasma proteome levels (Figure D, right).

Serum amyloid A proteins (SAA1/4) are acute-phase proteins produced in response to inflammation. To explore the specificity of their liver origin in HFpEF, we evaluated transcript levels of Saa1/4 in the liver, heart, and white adipose tissue (WAT) of CHOW and HFpEF mice. Saa1/4 mRNA levels were significantly increased only in the liver of HFpEF mice, with no significant changes observed in the heart and WAT compared to CHOW controls (Figure E, left). Furthermore, serum SAA1 protein levels were elevated in HFpEF mice, confirming its role as a secreted, circulating factor in HFpEF (Figure E, right). Interestingly, gene ontology (GO) enrichment analysis of heart DEGs in HFpEF, which are strongly (|r| > 0.8) and significantly (padj.<0.05) correlated with Saa1/4, revealed that Saa1/4 is significantly (padj.< 0.05) associated with cardiac genes involved in extracellular matrix deposition and remodeling (Figure F). These results suggest a potential role of SAA1/4 as inflammatory and fibrotic mediators between liver and heart in HFpEF, possibly contributing to the metainflammation – i.e. inflammation stemming from metabolic cues – observed in this syndrome.

In conclusion, we conducted the first, exploratory, bioinformatic-based correlation analysis of liver-heart interorgan crosstalk in preclinical HFpEF, which was subsequently supported by validation in human HFpEF and NAFLD subjects. Our findings highlight the identification of novel mediators of cardiometabolic dysregulation in HFpEF and potentially in MASLD/NAFLD, such as Saa1/Saa4. These findings warrant further exploration through in-depth mechanistic and clinical studies in cardiometabolic HFpEF and other HFpEF subtypes.