Introduction

Hypertrophic cardiomyopathy (HCM) is one of the most common genetic heart diseases and is associated with pathogenic genetic variants in genes that encode sarcomeric proteins. To date, at least 27 HCM-susceptibility genes have been discovered1. However, about half of patients remain genetically elusive with no identified HCM-associated genetic variant2,3. Importantly, there are clear phenotypic differences between sarcomere-positive HCM (HCMSarc) and genotype-negative HCM (HCMNeg). Patients with HCMSarc have a more severe disease phenotype, more severe disease progression, and worse outcomes4,5,6,7. Recent proteomic analysis of HCM have uncovered a wide network of pathways altered in HCM likely responsible for disease development7,8,9,10. Additionally, we recently demonstrated that HCM is characterized by widespread alterations in the proteome7. Despite clear differences between HCM and healthy individuals, the overall proteomic architecture of HCM was similar irrespective of genotype. However, given the clinical differences between genotypes, we hypothesized there may be subtle differences in the (phospho)proteome when comparing the (phospho)proteome between those with a positive genetic test (HCMSarc) versus those with a negative genetic test (HCMNeg).

Methods

We performed a subgroups analysis using our previously published proteomics cohort7. All patients provided written informed consent for this Mayo Clinic Institutional Review Board (IRB 811-98) approved study which abides by the Declaration of Helsinki. Detailed methods are provided in Supplemental Data. Briefly, patients with diagnosed obstructive HCM who underwent clinically indicated surgical myectomy for the relief of outflow tract obstruction were genotyped by genome sequencing followed by variant adjudication using American College of Medical Genetics and Genomics (ACMG) criteria11 and divided into genotype subgroups: HCMSarc (N = 15) defined as patients having an ACMG graded pathogenic (P)/likely pathogenic (LP) variant in a definitive or strong evidence HCM-susceptibility gene encoding a sarcomeric protein, and HCMNeg (N = 8) defined as patients with no variants of any classification in a panel of 54 HCM-associated genes1,7. Any patient with variants in HCM mimicker genes were excluded1. Genome sequencing was used to confirm the HCMNeg did not have deep intronic variants. Additionally, control tissue samples (N = 7) from normal donor hearts for which a suitable heart transplant recipient was not identified, were included in this study. All samples underwent TMT-labeled mass-spectrometry for proteomic and phosphoproteomic analysis. Differential expression analysis was performed to identify differences between subgroups followed by pre-ranked gene set enrichment analysis (GSEA) to identify which Gene Ontology (GO) biological processes were enriched in the proteome and phosphoproteome of HCMSarc compared to HCMNeg. Finally, differentially expressed (phospho)proteins were inputted into Ingenuity Pathway Analysis (IPA) for identifying altered pathways.

Results

Direct comparison of HCMSarc and HCMNeg

HCMSarc patients were diagnosed at a younger age (33 ± 16 vs. 47 ± 19 years; p = 0.04) and had their myectomies done at a younger age (40 ± 17 vs. 52 ± 16 years; p = 0.02) compared with HCMNeg patients (Supplemental Table 1). To test the hypothesis that HCMSarc and HCMNeg have alterations in pathways that may be responsible for observed clinical differences, we directly compared the proteomes and phosphoproteomes of HCMSarc and HCMNeg myectomy samples. Consistent with previous studies7,8, the proteome of HCM was similar regardless of genotype with no clear separation on principal component analysis (PCA) plotting (Supplemental Fig. S1A). Nonetheless, a direct comparison of the proteome of HCMSarc with HCMNeg revealed 243 differentially expressed proteins (DEPs) with 102 up-regulated and 141 down-regulated proteins (Supplemental Fig. S1B).

Overall, 4213 phosphorylated proteins were detected across all samples with PCA analysis showing slight, but not entirely complete separation between HCMSarc and HCMNeg (Supplemental Fig. S1C). Differential analysis identified 257 differentially phosphorylated proteins (DPPs): 134 hypophosphorylated and 123 hyperphosphorylated (Supplemental Fig. S1D). A complete list of DPPs can be found in the Supplemental Data.

Next pre-ranked, GSEA identified 128 biological processes altered at the proteome level with 97 being up-regulated and 31 down-regulated in HCMSarc compared with HCMNeg (Supplement). The top up-regulated processes were involved in either cell adhesion, extracellular matrix formation, and activation of fibrosis or regulation of cytoskeletal processes (Fig. 1A). In contrast, the top down-regulated processes were entirely metabolic processes with predominant down-regulation of aerobic respiration and mitochondrial function and catabolic processes, especially of fatty acids (Fig. 1B). At the phosphoproteome level, 57 biological processes were altered with 3 up-regulated and 54 down-regulated (Fig. 1C) with the main up-regulated processes involved in chromatin organization and gene expression while the majority of down-regulated processes were metabolic pathways akin to what was observed in the proteome.

Figure 1
figure 1

Comparison of gene ontology (GO) biological processes altered between HCMSarc and HCMNeg. (A) Most up-regulated gene ontology (GO) biological processes in the proteome using gene set enrichment analysis (GSEA). (B) Most down-regulated GO biological processes in proteome using GSEA. (C) Most altered GO biological processes in phosphoproteome using GSEA.

Full size image

Finally, DEPs and DPPs were inputted into Ingenuity Pathway Analysis (IPA) to identify alterations in canonical pathways. Overall, 72 pathways were altered at the proteome level with 7 up-regulated (z-score ≥ 1), 6 down-regulated (z-score ≤ -1), and 59 pathways for which directionality was indeterminate (Supplement). The valine degradation I, methylmalonyl pathway, actin cytoskeletal signaling, 2-oxobutanoate degradation I, and dilated cardiomyopathy signaling pathways were the most statistically altered pathways (largest -log [BH p-value]) (Fig. 2A). RhoA signaling (z-score = 2.5) and signaling by Rho Family GTPases (z-score = 2.1) were the two most up-regulated pathways in HCMSarc. In addition, oxytocin signaling pathway (z-score = 1.89), integrin-linked kinase (ILK) pathway (z-score = 1.6), integrin signaling (z-score = 1.3), coronavirus replication pathway (z-score = 1.3), and BAG2 signaling pathway (z-score = 1.0) were moderately up-regulated (Fig. 2B).

Figure 2
figure 2

Comparison of pathways altered between HCMSarc and HCMNeg. (A) Most statistically altered pathways (largest-log [BH p-value]) in proteome. (B) Top up- and down-regulated pathways in proteome. (C) Most statistically altered phosphorylation (largest-log [BH p-value]) of pathways. (D) Top activated and inactivated pathways based on phosphorylation (z-score ≥|1|). *Regulation of the epithelial mesenchymal transition by growth factors pathway was shortened.

Full size image

In addition, ketolysis, ketogenesis, mitochondrial dysfunction, glycerol-3-phosphate shuttle, fatty acid β-oxidation, acetate conversion to acetyl-CoA, branched-chain α-keto acid dehydrogenase complex were altered significantly. Oxidative phosphorylation was the most down-regulated pathway (z-score = − 2.7; Fig. 2B). Collectively, HCMSarc showed widespread down-regulation of aerobic respiration, mitochondrial function, and catabolic pathways compared with HCMNeg.

Pathway analysis of the DPPs demonstrated that 109 pathways had significant alterations in protein phosphorylation with 14 pathways predicted to be activated (z-score ≥ 1) and 8 inactivated (z-score ≤ − 1); for 87 pathways the directionality was unclear (Supplement). The pathways with the most alteration in phosphorylation status (largest − log [BH p-value]) are presented in Fig. 2C and included actin cytoskeleton signaling, RhoA signaling, estrogen receptor signaling, glycolysis I, ILK signaling, and signaling by Rho family GTPases. The pathways predicted to be activated or inactivated due to changes in phosphorylation are shown in Fig. 2D with activation of integrin signaling, actin cytoskeleton signaling, epithelial adherens junction signaling, signaling by Rho family GTPases, oxytocin signaling pathway, G-protein coupled receptor signaling, phagosome formation, calcium signaling, ILK signaling, coronavirus pathogenesis pathway, synaptic long-term depression, and epithelial mesenchymal transition by growth factors. Many metabolic pathways, including oxidative phosphorylation, glycolysis, gluconeogenesis, and dilated cardiomyopathy signaling pathways were predicted to be inactivated based on changes in phosphorylation (Fig. 2D). Overall, these results show significant concordance between proteomic and phosphoproteomic alterations supporting distinct differential regulation between HCMSarc and HCMNeg.

Additionally, the DEPs and DPPs were analyzed to identify enrichment for altered diseases and processes with the top 20 most statistically altered shown in Fig. 3. Of note, the disease processes altered at the proteome level were relevant to HCM such as hereditary myopathy, fibrogenesis, hypertrophy of heart, enlargement of heart. While growth and cancer processes were altered in both the proteome and phosphoproteome, they were more prominent at the phosphoproteome level. Thus, the proteomic and phosphoproteomic differences are predicted to impact cardiac function and cardiac hypertrophy demonstrating the changes between these genotypes are in disease relevant pathways.

Figure 3
figure 3

Altered diseases and functions in proteome and phosphoproteome identified using ingenuity pathway analysis.

Full size image

Comparisons of HCM with controls

In addition to HCMSarc and HCMNeg comparisons, we subsequently compared the phospho(proteomes) of HCMSarc and HCMNeg to non-hypertrophied normal cardiac tissue (controls) to identify disease-specific pathways. As expected, there was prominent dysregulation in the proteome and phosphoproteome of both HCMSarc and HCMNeg compared with controls. Overall, HCM was characterized by up-regulation of cytoskeletal and hypertrophy pathways and down-regulation of metabolic pathways. These findings are summarized in the Supplemental Results and Supplemental Figs. S2–S5.

Commonalities and differences in the proteome and phosphoproteome of HCMSarc and HCMNeg

Next, the three comparisons: HCMSarc versus HCMNeg, HCMSarc versus controls, and HCMNeg versus controls were juxtaposed to identify genotype-specific differences and similarities. Venn diagrams were generated to summarize the findings. Changes found in the HCMSarc versus controls and HCMNeg versus control comparisons are considered disease-relevant changes. Overall, there was significant overlap between the proteins and phosphoproteins altered in HCMSarc versus HCMNeg and those altered in disease (Fig. 4A,B; red circle). In fact, 232/243 (95%) DEPs in HCMSarc versus HCMNeg were also altered when comparing disease to controls (Fig. 4A). Integrating the DPPs across the three phosphoproteome comparisons revealed most of the direct differences between HCMSarc and HCMNeg (245/257 [95%]) were in phosphoproteins also observed to be altered in disease compared with controls (Fig. 4B; red circle). Thus, both at the protein and phosphoprotein level, most of the changes between subtypes were in disease-altered proteins (proteins altered in HCM compared to controls).

Figure 4
figure 4figure 4

Commonalities and differences in the proteome and phosphoproteome of HCMSarc and HCMNeg. (A) Venn diagram comparing differentially expressed proteins (DEPs) between different comparisons. (B) Venn diagram comparing differentially phosphorylated proteins (DPPs) between different comparisons. (C) Heatmap showing directionality of differentially expressed proteins (DEPs) altered in both HCMSarc and HCMNeg compared with controls. (D) Heatmap showing directionality of differentially phosphorylated proteins (DPPs) altered in both HCMSarc and HCMNeg compared with controls. Log2fc, log2 fold change.

Full size image

Overall, there were 376/664 (57%) DEPs found in HCMSarc compared with controls that were not seen when comparing to HCMNeg (Fig. 4A; orange circle). These HCMSarc specific DEPs were used to generate a protein–protein interaction network followed by enrichment analysis for biological processes which demonstrated predominant enrichment for metabolic gene sets (Supplemental Fig. 6). Since there were only 3 DEPs unique to HCMNeg versus controls, no genotype specific network was generated (Fig. 4A; blue circle).

There were 106 DEPs found when comparing both HCMSarc versus controls and HCMNeg versus controls (Fig. 4A; black circle). Figure 4C shows that the direction of change for all the DEPs is the same between both comparisons. There were 304 DPPs altered in both HCMSarc versus controls and HCMNeg versus controls (Fig. 4B; black circle). Figure 4D shows that the direction of change for all the DPPs is the same between both comparisons.

Considering GO biological processes, there was a similar, prominent overlap across the different genotype comparisons (Supplemental Fig. S7A). Since GSEA uses the entire protein list and not only differentially expressed proteins, more differences were observed. Still, 84/128 (66%) biological processes altered in HCMSarc versus HCMNeg were also altered in disease versus controls (red circle). There were 44/128 (34%) biological processes altered specifically in HCMSarc compared with HCMNeg (green circle). Notably, across the comparisons, the biological processes were similar throughout and were mostly extracellular matrix and cytoskeletal processes or metabolic processes. Interestingly, GO biological processes altered due to phosphorylation only had 24/57 (42%) of those altered in HCMSarc compared with HCMNeg also altered in disease compared with controls (Supplemental Fig. S7B; yellow circle). Specifically, there were epigenetic processes (i.e., chromatin organization) processes along with metabolic processes altered uniquely between the genotypes but not in disease compared with controls.

At the pathway level, the overlap between the genotype comparisons was very clear with 59/72 (82%) pathways altered between HCMSarc and HCMNeg also altered between disease and controls (Fig. 5A; red circle). There were 43 pathways altered at the proteome level both when comparing HCMSarc with controls and HCMNeg with controls suggesting these pathways likely serve a central role in maladaptive cardiac hypertrophy regardless of genetic background (black circle).Twenty-four pathways were common to all proteomic comparisons including HCMSarc versus HCMNeg such as oxidative phosphorylation, mitochondrial dysfunction, sirtuin signaling pathway, actin cytoskeleton, gluconeogenesis I, remodeling of epithelial adherens junctions, ILK signaling, integrin signaling, dilated cardiomyopathy pathway, signaling by Rho Family GTPases, and calcium signaling (Fig. 5A; gray circle and Supplement).

Figure 5
figure 5figure 5

Commonalities and differences in the proteome and phosphoproteome of HCMSarc and HCMNeg continued. (A) Venn diagram comparing pathways altered in proteome between different comparisons. (B) Venn diagram comparing pathways with altered phosphorylation between different comparisons. (C) Heatmap showing directionality of pathways altered in both HCMSarc and HCMNeg compared with controls. (D) Heatmap showing directionality of phosphorylated pathways altered in both HCMSarc and HCMNeg compared with controls.

Full size image

There were even greater numbers of pathways impacted due to phosphorylation with 89/109 (82%) pathways altered in both HCMSarc compared with HCMNeg and disease compared with controls (Fig. 5B; red circle). Fifty-one pathways were altered due to phosphorylation in all comparisons including aldosterone signaling in epithelial cells, estrogen receptor signaling, protein kinase A signaling, and actin cytoskeleton signaling (Fig. 5B; gray circle and Supplement).

There were 43 pathways altered when comparing both HCMSarc and HCMNeg compared with controls (Fig. 5A; black circle). Of these 43 pathways the directionality of all changes were concordant between both comparisons (Fig. 5C; Supplemental Table S2). These included pathways such as actin cytoskeleton signaling, BAG2 signaling, calcium signaling, dilated cardiomyopathy signaling, gluconeogenesis, glycolysis, ILK signaling, integrin signaling, oxidative phosphorylation, RhoGDI signaling, and signaling by Rho Family GTPases. There were 148 phosphorylated pathways common between both comparisons (Fig. 5B; black circle). Interestingly, for three pathways, actin cytoskeleton signaling, ERK/MAPK signaling, and G-protein coupled receptor signaling, it is predicted that these particular pathways are activated due to phosphorylation in HCMSarc versus control samples but inactivated in HCMNeg compared to controls (Fig. 5D; Supplemental Table S3).

Finally, when looking at the pathways that were either up- or down- regulated when directly comparing HCMSarc and HCMNeg (Fig. 3B) those pathways were not only altered when comparing disease with controls (Fig. 3 and Supplemental Table S2), but the direction also matched that observed in disease. Thus HCMSarc and HCMNeg have alterations in disease specific pathways and the dysregulation at least for several pathways is more severe in HCMSarc (Fig. 6).

Figure 6
figure 6

Proteome of sarcomere-positive HCM and sarcomere-negative HCM have differences in disease specific pathways with most differences between them being more severe in sarcomere-positive HCM.

Full size image

Comparing the two most common genotypes of sarcomeric HCM: MYBPC3-HCM and MYH7-HCM

The proteomes and phosphoproteomes for the two most common HCM-susceptibility genes: MYBPC3 and MYH7, were compared. Details are provided in the Supplemental Data. The proteome of the two different HCMSarc genotypes did not separate on PCA plotting and no DEPs or DPPs were identified between the two genotypes. Thus, there is little difference in the proteomes of MYBPC3-HCM and MYH7-HCM.

Discussion

While treatments and outcomes for HCM are improving, its vast genotypic and phenotypic heterogeneity continues to make optimal treatments and sudden cardiac death prevention challenging. Furthermore, there are enormous alterations that occur in the proteome of HCM compared with healthy controls8,9,10,12 making it hard to identify therapeutic targets. Additionally, the mono/oligo/polygenic basis or non-genetic basis for the patients classified as HCMNeg remains unknown, but there is growing evidence for some within this subset to be driven by either oligogenics, epigenetics, or environmental factors, or a combination of the above13,14. Previous studies have shown that HCMSarc is characterized by earlier onset of disease and more severe hypertrophy4,5. Given these differences, there is a need for a deeper understanding of the underlying mechanisms responsible as this could help in the identification of novel therapeutic targets that modify disease progression and may help with prioritization of therapeutic targets.

Given the clinical differences across HCM genotypes15, there have been several studies aimed at understanding mechanistic and structural differences across genotypes which have identified nuanced differences between different genotypes8,16,17,18. While some mechanisms underlying these genotype-based differences have been elucidated, genotype comparisons of the (phospho)proteome of HCM have been limited. Using unsupervised machine learning methods, prior studies have demonstrated the proteome of HCM is similar regardless of genotype7,8,19 suggesting that regardless of underlying genetic etiology, HCM converges on a final common pathway. However, one of these studies, by Schuldt et al.8 identified that some proteins were in fact altered in HCMSarc compared with controls but not in HCMNeg and vice-versa, suggesting there may in fact be subtle genotype-based differences in the proteome of HCM. However, this study did not identify proteins directly altered when comparing the two predominant HCM subtypes: those with a positive genetic test (HCMSarc) and those with a negative genetic test (HCMNeg) and did not identify changes in hypertrophy pathways. Additionally, a second study looking at post-translational modifications and alternative splicing of sarcomeric proteins found minor genotype differences but an overall similar proteomic profile19. In the current study, consistent with previous proteomic studies, we found the proteomes of HCM genotypes are not distinct enough to be separated by unsupervised clustering analyses, however, our analysis did reveal direct differences in protein levels between HCMSarc and HCMNeg. These proteomic changes were predicted to have significant effects on cardiac hypertrophy, cardiac function, and development of heart failure. Interestingly, most of the proteins and pathways altered between HCMSarc and HCMNeg were also altered when comparing HCM to controls, and in fact at the protein level, the alterations were generally more severe in HCMSarc suggesting genotype differences between HCM genotypes may be due to degree of dysregulation of the disease associated pathways.

Interestingly, at the protein level, cytoskeletal processes and extra-cellular matrix processes were significantly up-regulated when comparing HCMSarc with HCMNeg. This is consistent with the observation that HCMSarc shows increased fibrosis compared with HCMNeg correlating with late gadolinium enhancement (LGE) which is associated with increased fibrosis on cardiac MRI20,21. Additionally, we observed activation of RhoA signaling and found it to be even more activated in HCMSarc compared to HCMNeg. RhoA signaling activates ROCK signaling which effects several cardiac cell types leading to altered cardiomyocyte contraction, cardiac gene expression, and protein phosphorylation and subsequently to cardiac hypertrophy and fibrosis22,23,24. Up-regulation of RhoA signaling and integrin signaling influences stress induced hypertrophy and is protective against heart failure; however, activation can also result in fibrosis25. Rho A signaling has a complex role in cardiac function and whether it is beneficial or deleterious depending on temporal and contextual factors and of cell type25,26. Our observations suggest that Rho A signaling has a central role in HCM as a mechanism for disease and possible cause for genotype-based differences, although further studies are necessary to parse out Rho A signaling’s precise role.

Other studies of this kind have shown HCM is characterized by widespread down-regulation of metabolic pathways, especially aerobic respiration and fatty acid oxidation27,28. In this study we not only replicated these findings, but also demonstrated genotype-specific differences with a more severe down-regulation of metabolic pathways observed in HCMSarc. Metabolic reprogramming is observed in other forms of pathologic cardiac hypertrophy such as pressure overload hypertrophy and can serve as a marker for early onset of heart failure29. Consistent with our findings, down-regulation of amino acid metabolism and fatty acid metabolism and aerobic respiration have been observed as early markers towards heart failure. Thus, the more severe derangement in the proteome of HCMSarc reflects a more severe underlying metabolic reprogramming likely playing a role in genotype specific clinical differences. Additionally, we identified almost four hundred proteins altered exclusively in HCMSarc which allowed us to generate a HCMSarc protein–protein network revealing enrichment in metabolic processes. The findings suggest specific metabolic pathways may explain genotype specific differences observed in HCM and warrant further investigation.

In addition to the proteomic differences, this study showed a phosphorylation-mediated activation of integrin and cytoskeletal signaling and phosphorylation-mediated down-regulation of metabolic pathways in HCMSarc. As changes in the same pathways were also observed on the protein level, this would be a strong indication that phosphorylation is regulating the function of these pathways.

Since predominant down-regulation of metabolic pathways appears central to HCM both in the proteome and phosphoproteome counteracting these changes could be a therapeutic strategy. For instance, a kinase inhibitor screen has shown promise in identifying novel therapeutics for dilated cardiomyopathy which is also characterized by metabolic reprogramming and so a similar strategy manipulating phosphorylation of metabolic pathways altered in HCM could be successful30. Consequently, the functional and temporal effects of protein phosphorylation warrant further investigation and could open the door to identifying novel strategies to modifying disease progression.

The phosphoproteomic differences were more complex than the proteomic changes. In some cases, phosphorylation changes were more severe in HCMNeg with predominant phosphorylation-mediated inactivation. In addition, for some pathways, there were differences in the direction of change between genotypes. In contrast with HCMSarc, HCMNeg showed inactivation of growth pathways such as ERK/MAPK signaling and cardiac hypertrophy signaling due to changes in phosphorylation. As we identified previously, this signaling cascade appears central to many hypertrophy pathways up-regulated in HCM7. This phosphorylation-mediated inactivation could contribute to the less severe progression observed in HCMNeg.

Limitations

This study was performed on myectomy tissues and thus provides a snapshot view of the myocardium at the time of myectomy, representing obstructive HCM generally in a later stage of disease progression. Therefore, these identified perturbations in the (phospho)proteomic architecture of obstructive HCM may not apply to the other morphologic subtypes of HCM such as non-obstructive HCM or apical HCM.

Conclusion

Although the proteome and phosphoproteome of obstructive HCM is similar regardless of genetic etiology, there were important differences between HCMSarc and HCMNeg with a more severe dysregulation of disease relevant pathways observed in HCMSarc. Further studies are necessary to determine whether these changes underlie the clinical differences between HCMSarc and HCMNeg.