A combined bioinformatics, experimental and clinical approach to identify novel cardiac-specific heart failure biomarkers: is Dickkopf-3 (DKK3) a possible candidate?

Abstract

Aims: Cardiac specificity provides an advantage in correlating heart failure (HF) biomarker plasma levels with indices of cardiac function and remodelling, as shown for natriuretic peptides. Using bioinformatics, we explored the cardiac specificity of secreted proteins and investigated in more detail the relationship of Dickkopf-3 (DKK3) gene expression and DKK3 plasma concentrations with cardiac function and remodelling in (pre)clinical studies.

Methods and results: The cardiac specificity of secreted proteins was determined using RNAseq data for a large panel of organs and tissues. This showed that natriuretic peptides (NPPA and NPPB) are highly cardiac-specific (>99%), whereas other HF biomarkers, including galectin-3 (Gal-3, LGALS3) and growth differentiation factor-15 (GDF-15), lack cardiac specificity (<4%). DKK3 was cardiac-enriched (44%), warranting further investigation. In three different HF mouse models, cardiac Dkk3 expression was altered, but DKK3 plasma concentrations were not. In humans, DKK3 plasma concentrations were higher in HF patients (n = 2090) in comparison with age- and sex-matched controls without HF (n = 240) (46.4 ng/mL vs. 36.3 ng/mL; P < 0.001). Multivariate regression analysis revealed that DKK3 was strongly associated with HF risk factors and comorbidities, including age, kidney function and atrial fibrillation. After correction for existing prediction models, DKK3 did not independently predict HF outcome [all-cause mortality/HF hospitalization, hazard ratio 1.13 (0.79-1.61) per DKK3 doubling; P = 0.503].

Conclusions: Of actively secreted HF biomarkers, only natriuretic peptides showed high cardiac specificity. Despite a cardiac specificity of 44%, secreted DKK3 had limited additional diagnostic and prognostic value.

Keywords: Biomarker; Cardiac specificity; DKK3; Heart failure; Natriuretic peptides.

Figure 1

Cardiac specificity of the human genome and heart failure (HF) biomarkers. (A) Schematic depiction of the bioinformatics approach. Gene expression data as measured by RNAseq in all organs and tissues of healthy humans were used. ¹³ Included are adipose tissue, adrenal gland, appendix, bone marrow, brain, colon, duodenum, gall bladder, heart, kidney, liver, lung, lymph node, oesophagus, pancreas, salivary gland, skin, small intestine, spleen, stomach and thyroid gland. Sex‐specific organs or tissues were the prostate and testis for males, and endometrium and ovary for females. The cardiac specificity of genes was calculated by the formula shown. Finally, genes assumed to encode for secreted proteins were selected. (B) The cardiac specificity of the human genome (top) and of secreted protein encoding genes (bottom, showing the numbers of included genes. Coloured areas represent the amount of genes with the corresponding degree of specificity. Data include those for male tissues. (C) Cardiac specificity of several HF biomarkers. Coloured areas represent organ specificity. Data include those for male tissues. Gal‐3 (LGALS3), galectin‐3; GDF‐15, growth differentiation factor 15; GI tract, gastrointestinal tract, including appendix, colon, duodenum, oesophagus, small intestine and stomach; NPPB, natriuretic peptide precursor type B; TIMP‐1, tissue inhibitor of metalloproteinase 1; Sex organs, prostate and testis. The bioinformatics analysis was based on publicly available RNAseq data previously published by Fagerberg et al. ¹³

Figure 2

The cardiac specificity of secreted protein encoding genes. The total cardiac expression of secreted protein encoding genes was plotted against their calculated cardiac specificity. All genes with cardiac specificity of >10% are shown. Gene name abbreviations are used for readability. Organs and tissues include adipose tissue, adrenal gland, appendix, bone marrow, brain, colon, duodenum, gall bladder, heart, kidney, liver, lung, lymph node, oesophagus, pancreas, salivary gland, skin, small intestine, spleen, stomach and thyroid gland, and sex‐specific organs or tissues for males (prostate and testis). The bioinformatics analysis was based on publicly available RNAseq data previously published by Fagerberg et al. ¹³

Figure 3

Cardiac gene expression and plasma concentrations of ANP and DKK3 in three different heart failure (HF) mouse models. Included HF models are transverse aortic constriction (TAC) (top row), myocardial infarction (MI) (middle row) and an obese/hypertension mouse model with HF with preserved ejection fraction characteristics (HFD + AngII, bottom row). (A) Natriuretic peptide precursor type A (Nppa) left ventricular (LV) gene expression (left) and N‐terminal pro‐atrial natriuretic peptide (NT‐proANP) plasma concentrations (right). Data on Nppa and NT‐proANP have been published previously, ⁹ but are presented here for convenience. (B) Dickkopf‐3 (Dkk3) gene expression (left) and DKK3 plasma concentrations (right). Gene expression values are corrected for 36b4 gene expression and presented as fold change. Bars represent means. Error bars represent standard errors of the mean. *P < 0.05 vs. sham or low fat diet. #P < 0.05 vs. HFD. AngII, angiotensin‐II; HFD, high fat diet; LFD, low fat diet.