Hypertrophic cardiomyopathy (HCM) is a relatively rare, but debilitating diagnosis in the pediatric population, with an estimated incidence of 2.9 per 100,000.1 Although outcomes vary, mortality is almost 4-fold higher in young patients with HCM than in the general population.2 In particular, children with end-stage HCM are a vulnerable population; approximately a quarter of them experience significant arrhythmias, and waitlist mortality can be as high as 33% in infants.3Single-nucleus RNA sequencing (snRNA-seq) is a powerful tool used to assess gene expression in the heart. Recently, snRNA-seq has been used to study the transcriptome in adult patients with HCM and has shown an enrichment of genes associated with “cardiac stress” in cardiomyocytes, activation and fibrosis in fibroblasts, and macrophage subtype switching.4 However, snRNA-seq studies of cardiac disease in pediatric patients are scarce because of the rarity of samples.5,6 Hence, no studies have focused on pediatric HCM.We performed snRNA-seq to study transcriptome differences between end-stage pediatric HCM patients and control samples. Specifically, we identified pathways found in each of the major cell types in pediatric HCM, the co-expressed gene networks and intercellular communications that were upregulated, and compared these findings with those seen in adult HCM.
Methods
The full methods are available as Supplemental Material. The study conforms to the principles of the Declaration of Helsinki, the protocol was approved by the Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals, and subjects gave informed consent. Data generated is available at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO GSE292754). Data for sample HCM1 (P86) has been previously deposited (GEO GSE203275).
Results
From 8 tissue samples (Figure 1A), we generated 88,915 single-cell transcriptomes and identified 10 clusters representing major cell types (Figure 1C). We focused on cardiomyocytes, fibroblasts, endothelial/endocardial cells, and myeloid cells. After re-clustering, 28,541 cardiomyocytes, 29,512 endocardial/endothelial cells, 13,959 fibroblasts, and 3,106 myeloid cells were identified (Figure 1C). Figure 1D displays the proportion of major cell types. Permutation testing showed increased proportions of endothelial cells (log2FD=1.25,) and decreased proportions of cardiomyocytes (log2FD=-0.77), fibroblasts (log2FD=-0.16), and myeloid cells (log2FD=-1.31) in HCM, all with adjusted P-values of 0.001. Principal component analysis was completed on the pseudo-bulk matrix; PC1 distinguishes control and HCM samples (Figure 1E).Re-clustering of cardiomyocytes identified 5 clusters: CM1-CM5 (Figure 2A), with subject level proportions shown in Figure 2B. Permutation testing exhibited increased proportions of HCM-derived cells in CM1 (log2FD=2.65) and CM4 (log2FD=1.24) and increased proportions of control-derived cells in CM2 (log2FD=0.65), CM3 (log2FD=2.39), and CM5 (log2FD= 2.69), all with adjusted P-values <0.001. This was substantiated in differential abundance (DA) testing (Figure 2C).The differentially expressed genes (DEGs) significantly upregulated (FC >1) in HCM-enriched CM1 included genes associated with cardiomyopathy and cardiac stress, like ANKRD1, NPPB, and PROS1 (Figure 2D, Table S3).7 Other upregulated genes included PLCE1, implicated in ischemia-reperfusion injury, and SLITRK4, a diagnostic HCM marker.8,9 Control-dominant CM3 and CM5 showed upregulation of MYH7, NPPA, and XIRP2, typically upregulated in stressed myocardium. In comparison of HCM to controls, DEGs significantly overlapped with those seen in CM1 and CM4 and identified genes downregulated in end-stage cardiomyopathy, MYH6 and CORIN (Table S3).5,10 Functional enrichment analysis showed CM1 was uniquely enriched in insulin signaling and shared enriched pathways with CM3 and CM5, like focal adhesion, autophagy, Mitogen-Activated Protein Kinase (MAPK), and Hypoxia-InducibleFactor 1 (HIF-1) signaling (Figure 2E, Table S4).To further define cardiac stress of each cluster, we created module scores derived from genes in normal, pre-stressed, and stressed myocardium in adults after infarction.11 CM2 and CM3 had the highest normal expression; CM1 and CM3 had the highest pre-stressed expression; CM1 and CM5 had the highest stressed myocardium (Figure 2F). This identifies HCM-enriched CM1 as the most consistently stressed cluster.We then created co-expression networks to identify interconnected genes enriched in HCM (Supplemental Results). The hdWGCNA pipeline is described in Figure S1B. We identified co-expression of cardiac stress genes and pathways implicated in HCM (Figure S1, Tables S5-S7). Sensitivity analyses showed consistency compared with DEG results and among HCM subjects and cells derived from left or right ventricles (Figure S4-S5, Table S8).Re-clustering of fibroblasts identified 4 clusters: FB1-FB4 (Figure 3A). Figure 3B shows their proportions by subject. Permutation testing identified increased proportions of HCM-derived cells in FB2 (log2FD=0.45), FB3 (log2FD=3.27), and FB4 (log2FD=0.59), and increased proportions of control-derived cells in FB1 (log2FD=2.06), all with adjusted P-values <0.001 and supported by DA testing (Figure 3C).DEGs and functional pathway enrichment revealed distinct biology across clusters (Figure 3D/E; Table S9-S10). FB3 expressed markers of fibroblast activation such as POSTN, PALLD, FAM155A, and FGF14, along with enrichment of integrin, extracellular matrix (ECM), and calmodulin signaling pathways, consistent with a matrifibrocyte phenotype.4 FB2 exhibited DEGs associated with collagen deposition (COL3A1 and COL1A2), lipogenic activity (APOE and APOD), and anti-apoptosis (TIMP1), accompanied by enrichment in energy metabolism, ECM, actin, fibronectin, mitophagy, and insulin-like growth factor pathways.12,13 This supports a metabolically active cell state with myofibroblast features. The FB4 cluster showed pro-inflammatory and fibrotic DEGs such as OSMR.12 In comparing HCM to controls, we found significant overlap in DEGs involving genes associated with fibroblast activation and collagen deposition (Table S9).To explore the collective expression of activated fibroblast genes, we created a module score based on genes identified in activated fibroblast populations.4 HCM-derived cells and HCM-enriched FB3, and to a lesser extent FB2, exhibited significantly higher module expression, consistent with our previous findings (Figure 3F). This was seen across all HCM subjects (Figure S5A). We then created FB2 and FB3 module scores and compared our data to a pediatric dilated cardiomyopathy (DCM) dataset, and found enrichment in HCM (Figure 3G).14 This indicates that our findings represent a genetic signature in HCM.To support this, we compared human FB clusters to cell states identified in a curated snRNA-seq dataset from injury murine models, using the Jaccard index to quantify transcriptional overlap.15 FB2 overlapped with the myofibroblast (MYO) and F-CI states, indicating a transitional, injury-responsive phenotype and FB3 overlapped with the matrifibrocyte (MFC) state (Figure 3H). Patients with preserved cardiac function (H2/H3) were enriched for FB2, while subject H1 with mildly reduced function showed a higher proportion of FB3, consistent with a more advanced fibrotic state. These findings suggest a continuum of fibroblast activation linked to disease progression.As described in the Supplemental Results, our hdWGCNA analysis identified co-expressed genes corroborating our earlier findings (Figure S2, Tables S11-S13). Sensitivity analyses showed consistency with our DEG analyses and among HCM subjects and cells derived from left or right ventricles (Figure S4-S5, Table S14).Re-clustering identified five clusters denoted as EC1-EC3, EndoC, and LEC (Figure 4A). With permutation testing of proportions (Figure 4B), LECs (log2FD=1.29, adjusted P-value=0.002) and EC1 (log2FD=0.19, adjusted P-value=0.002) were HCM dominant, whereas EndoC was control dominant (log2FD=1.95, adjusted P-value=0.002). Similarly, DA testing showed the greatest HCM enrichment in EC1-EC3 (Figure 4C).DEG analysis (Table S15, Figure 4D) identified markers for venous, capillary, and arterial endothelial cells; endocardium and lymphatic endothelial cells were also identified. EC1 represented capillary endothelial cells (ie, CA4 and RGCC); EC2, arterial endothelial cells (ie, GJA5, DLL4, and SEMA3G); EC3, venous endothelial cells (ie, NR2F2, SELP, and ACKR1); EndoC, endocardium (ie, SPOCK1 and NPR3); and LECs, lymphatic endothelial cells (ie, CCL21).16 EC1 and EC2 were enriched with genes associated with cell fate, angiogenesis, cell junction, and vascular remodeling (NOTCH4, FLT1, ADAMTS6, FGFR1, RGCC, COL15A1; Figure 4E).12,17,18 This was validated when comparing HCM to controls showing overlapping genes such as FLT1, FGFR1, and RGCC (Table S15). Functional pathway analysis indicated enrichment of cell adhesion, adherens junction, cadherin binding, NOTCH signaling/binding, and vascular endothelial growth factor (VEGF) receptor binding pathways (Table S16; Figure 4F). These findings support the role of endothelial cells in vascular remodeling in HCM.In the hdWGCNA analysis, we found coexpression of genes associated with cell fate and angiogenesis (Supplemental Results; Figures S3 and Tables S17–19). These results are consistent across our previous DEG analyses, HCM subjects, and cells derived from left or right ventricles (Figure S4-S5, Table S20).After re-clustering, we identified 3 myeloid cell clusters (Figure 5A): MP1, a monocyte-derived macrophage cluster enriched with HLA class II genes and NAMPT; MP2, a tissue-resident macrophage cluster enriched with LYVE1; and Mono, a monocyte cluster enriched with VCAN. Permutation testing exhibited heightened HCM proportions in MP1 (log2FD=1.21, adjusted P-value<0.001) and heightened control proportions in MP2 (log2FD=0.38, adjusted P-value<0.001) and Mono (log2FD=2.08, adjusted P-value<0.001; Figure 5B). DA testing showed enrichment of MP2 and Mono clusters in controls and MP1 in HCM (Figure 5C).We completed DEG analysis (Table S21, Figure 5D) and created a module score based on markers of recruited (ie, CCR2+) and resident (ie, CCR2-) macrophages (Figure 5E and 5F).19 HCM-enriched MP1 differentially expressed NAMPT, which plays a role in myocardial adaption to pressure overload, and genes involved in NAD+ nucleosidase activity (eg, IL18R1).20 MP1 also differentially expressed MHC class I and II genes (ie, HLA-DRA, HLA-DRB1, HLA-DQB1, HLA-DPB1). This is consistent with expression of recruited myeloid markers suggesting an inflammatory role and is substantiated by functional enrichment analysis showing enrichment of antigen processing and presentation, immune activity, and MHC protein pathways (Table S22). MP2 expressed resident myeloid markers and DEGs such as SCN9A (a sodium channel gene), FGF13 (a sodium channel modulator), and F13A1 (gene encoding the coagulation factor XIII A subunit).19 When comparing HCM to control, we noted DEGs associated with macrophage activation and differentiation, such as ANKRD22 (Table S21).21In studying cell-to-cell communication, we found that the differential number and weight of interactions were higher in HCM than in control myocardium across most interactions (Figure 6A). Fibroblasts had the highest strength of outgoing interactions; endocardium, endothelial cells, and cardiomyocytes had the highest incoming interaction strengths (Figure 6B). The strongest differential outgoing interactions from fibroblasts involved LAMININ and COLLAGEN pathways (Figure 6C). To identify patterns in diseased tissue, we further assessed cell-to-cell communication within HCM clusters and identified FB2 and FB3 as the highest contributors to outgoing interactions (Supplemental Results; Figures S6). The predicted ligand and receptors are available in Tables S23-S24.To substantiate our RNA findings at the protein-level, we mapped DEGs in HCM to their corresponding proteins and intersected them with a large-scale proteomic dataset from adult HCM patients undergoing myectomy.22 This identified 191 overlapping proteins, several of which were fibroblast-derived ligands also detected in our CellChat analysis, with their corresponding receptors (Figure 6D). While limited by differences in patient age and disease stage, this cross-reference provides supporting evidence for key fibroblast ligand-receptor pairs for future study.We integrated data from 5 adult women with end-stage HCM (Figure 7A).4 Since mitochondrial and ribosomal DEGs may relate to technical differences between datasets, we removed these genes from downstream analyses. In total, 1580 DEGs were identified (Table S25). Upregulated DEGs (ie, FC>1) included collagen and laminin genes such as COL21A1, COL4A1, LAMA4, and LAMB1. Some fibroblast activation genes were upregulated in pediatric HCM (eg, PALLD, POSTN, and AKT3), but others were upregulated in adult HCM (eg, FAP and TSHZ2). Consistent with these results, we found enrichment of pathways related to fibrosis (ie, ECM protein pathways), metabolic stress (ie, antioxidant, glutathione peroxidase activity), and cell turnover (ie, apoptosis, autophagy, ferroptosis, FoxO signaling) in pediatric HCM (Figure 7B, Table S26). Because our primary analysis showed heightened fibroblast outgoing communication, we created module scores from CellChat COLLAGEN and LAMININ ligand genes and found higher expression of these modules in pediatric HCM (Figure 7C).We identified four fibroblast clusters: cFB1-cFB4 (Figure 7D, left). Two clusters of interest were cFB3, an activated fibroblast population, and cFB2, the only cluster enriched with pediatric HCM neighborhoods in our DA analysis. In cFB2, functional enrichment analysis identified pathways associated with fibrosis, energy metabolism (ie, NADH/NADPH, lipid, electron transport chain pathways), metabolic stress, and cell turnover (Figure 7D, right; Table S26). Importantly, genes identified in apoptotic pathways were upregulated in FB2, such as IGFBP6 and GPX4, and can have anti-apoptotic properties.23,24 cFB2 shared 81.50% (1,132 genes) of its DEGs with our original FB2, compared to 2–6% in the remaining clusters, suggesting they are biologically similar, pediatric-specific clusters.
Discussion
In this study, we identified transcriptional changes in ventricular tissue derived from pediatric patients with HCM compared to structurally normal hearts. In addition, we described cell states and cellular processes unique to pediatric HCM. A marked fibroblast response was seen and a disrupted micro-tissue environment was largely mediated by signaling originating from fibroblasts. Furthermore, compared to adult HCM, we saw downstream upregulation of fibrotic pathways and identified a unique pediatric cluster. These data add to our understanding of pediatric end-stage HCM.We identified genes and pathways upregulated in pediatric HCM cardiomyocytes. Importantly, we noted upregulation of genes associated with cardiac stress, such as NPPB and ANKRD1.4,12 Other cardiac stress genes (eg, MYH7 and XIRP2) were upregulated in some control-enriched clusters. This may be associated with the organ procurement process and illustrates that cardiac stress can occur outside of structural heart disease.5 We also identified genes associated with maladaptive and inflammatory responses to stress, cardiac remodeling, and hypertrophy. For example, PLCE1 and FKBP5 were identified in multiple analyses; these genes promote inflammation via NF-κB signaling pathway activation.8,25Our results show endothelial cell heterogeneity and depletion of tissue-resident macrophages in HCM myocardium. In this setting, endothelial cells likely promote angiogenesis in response to chronic hypoxia, and pathways such as those involving NOTCH and VEGF receptor binding drive angiogenesis.26,27 Although the mechanisms are not fully understood, angiogenesis-induced hypertrophy can occur without hypertrophic stimulation.27 In myeloid cells, we found reduced tissue-resident macrophages and slightly enriched recruited macrophages, exhibiting a shift to a pro-inflammatory myeloid state in HCM. When compared with recruited macrophages, resident macrophages are more protective, secrete fewer inflammatory mediators, and inhibit leukocyte recruitment.19 This shift in myeloid populations has been noted in adult HCM, suggesting shared therapeutic targets in children and adults.4,7Cardiac fibroblasts are the main mediators of fibrosis in heart disease, and their functions increase in response to injury.28 Acutely, fibroblasts are reparative but may become deleterious in the chronic setting. This is represented by an initial response of activation and proliferation of ECM-producing myofibroblasts, that transition into longer-lived matrifibrocytes, specialized in maintaining collagen rich environments.28 Fibroblast activation is seen in adult HCM and DCM and is found almost exclusively in pathologic samples.4,10 We noted a large population of activated, matrifibrocyte fibroblasts in pediatric HCM, with similar enrichment compared to adult HCM. However, one cluster we identified was specific to pediatric HCM. This cluster represented a metabolically stressed and injury-responsive myofibroblast-like state that orchestrates more ECM deposition, adjusts energy metabolism, and may have anti-apoptotic properties. Similar to previous studies, Patrick et al. confirmed that injury-responsive (F-CI) and myofibroblast states eventually progress to matrifibrocytes through trajectory analysis and our data suggests a similar progression phenotypically.15 The presence of this pediatric-specific cluster may represent a targetable window to halt the progression to matrifibrocytes. Finally, we identified upregulation of cell-to-cell communication in HCM, which is consistent with a disrupted microtissue environment and identified ligand and receptor pairs substantiated by proteomics.5Our study is limited by sample size and lack of functional validation in pediatric samples. This limitation is inherent to most pediatric studies because of the rarity of samples and disease. In addition, we could not control for biases such as race, ethnicity, surgical sampling, medication, and epi-genetic modifications, which may have affected results. Finally, our samples came from patients with end-stage disease and may not represent processes seen in non-end-stage disease. Regardless, our study is the first to use snRNA-seq to examine genetic signatures in pediatric HCM.In conclusion, we have described the underlying cellular processes in pediatric HCM. Fibroblast-mediated processes are the most prominent among the major cardiac cell types analyzed. Moreover, pediatric patients have more downstream processes associated with fibrosis than do adult patients as well as a unique myofibroblast-like cluster with increased energy metabolism and anti-apoptotic properties.
HT
Dr. Hanna Tadros
Pediatric Cardiology · Pediatric Heart Failure & TransplantationBaylor College of Medicinehanna