PRDM16 Is a Compact Myocardium-Enriched Transcription Factor Required to Maintain Compact Myocardial Cardiomyocyte Identity in Left Ventricle

Abstract

Background:

Left ventricular noncompaction cardiomyopathy (LVNC) was discovered half a century ago as a cardiomyopathy with excessive trabeculation and a thin ventricular wall. In the decades since, numerous studies have demonstrated that LVNC primarily impacts left ventricles (LVs), and is often associated with LV dilation and dysfunction. However, owing in part to the lack of suitable mouse models that faithfully mirror the selective LV vulnerability in patients, mechanisms underlying susceptibility of LV to dilation and dysfunction in LVNC remain unknown. Genetic studies have revealed that deletions and mutations in PRDM16 cause LVNC, but previous conditional Prdm16 knockout mouse models do not mirror the LVNC phenotype in patients, and importantly, the underlying molecular mechanisms by which PRDM16 deficiency causes LVNC are still unclear.

Methods:

Prdm16 cardiomyocyte (CM)-specific knockout (Prdm16^cKO) mice were generated and analyzed for cardiac phenotypes. RNA sequencing and ChIP sequencing were performed to identify direct transcriptional targets of PRDM16 in CMs. Single cell RNA sequencing in combination with Spatial Transcriptomics were employed to determine CM identity at single cell level.

Results:

CM-specific ablation of Prdm16 in mice caused LV-specific dilation and dysfunction, as well as biventricular noncompaction, which fully recapitulated LVNC in patients. Mechanistically, PRDM16 functioned as a compact myocardium-enriched transcription factor, which activated compact myocardial genes while repressing trabecular myocardial genes in LV compact myocardium. Consequently, Prdm16^cKO LV compact myocardial CMs shifted from their normal transcriptomic identity to a transcriptional signature resembling trabecular myocardial CMs and/or neurons. Chamber-specific transcriptional regulation by PRDM16 was in part due to its cooperation with LV-enriched transcription factors Tbx5 and Hand1.

Conclusions:

These results demonstrate that disruption of proper specification of compact CM may play a key role in the pathogenesis of LVNC. They also shed light on underlying mechanisms of LV-restricted transcriptional program governing LV chamber growth and maturation, providing a tangible explanation for the susceptibility of LV in a subset of LVNC cardiomyopathies.

Introduction

Left ventricular noncompaction cardiomyopathy (LVNC) is the third most prevalent cardiomyopathy characterized by excessive trabeculation (noncompaction) and thin ventricular wall^1–3, which predominantly impacts left ventricle^4–6. Trabeculation is a meshwork of myocardial protrusions (trabeculae) that extend from the compact ventricular wall into the ventricular lumen in developing heart^{7, 8}. From late embryonic to perinatal stages, trabecular myocardium undergoes compaction, resulting in disappearance of trabeculae and thickening of a tightly-packed ventricular wall, creating a relatively smooth endocardial surface⁹. However, mounting evidence suggests thinning and dilation of compact myocardium can independently predict adverse clinical outcomes^{10, 11}, whereas noncompaction alone without ventricular dilation is usually asymptomatic¹², suggesting that left ventricle (LV) dilation is the most important feature of symptomatic LVNC^{13, 14}.

Classified as a primary genetic cardiomyopathy by AHA¹⁵, LVNC has been linked to mutations in genes encoding sarcomeric and cytoskeletal proteins, nuclear membrane proteins, ion channels and transcription factors^{2, 16}, but the mechanisms by which these mutant proteins lead to LVNC have been poorly defined. This is at least in part due to the lack of suitable mouse models to faithfully mirror disease conditions in patients (i.e., selective LV vulnerability) as mouse models for LVNC constructed to date, including Mib1^{cKO 17}, Nb/Nbl^{cKO 18} and Nae1^{cKO 19}, all manifest biventricular noncompaction and dilation.

LVNC has been associated with 1p36 syndrome, which has been further linked to deletions in PRDM16²⁰, a gene encoding for a member of the PR domain-containing (PRDM) family of transcriptional regulators and histone methyltransferases²¹. Mutations in PRDM16 are also known to cause dilated cardiomyopathy^{20, 22}, which is in line with the localization of PRDM16 in murine and human CM nuclei²⁰. This suggests that PRDM16 may function as an important transcriptional regulator in CM. However, PRDM16 has also been elucidated to play versatile roles in brown adipose tissue determination and differentiation²³, neural stem cell maintenance and migration^{24, 25} and hematopoiesis²⁶. Consequently, Prdm16 global knockout mice develop multi-system defects, including ventricular hypoplasia, and die shortly after birth²⁷. Thus, it remains unclear whether PRDM16 causes LVNC via a cell autonomous mechanism in CM or due to a systematic failure of multiple organs.

To address the potential CM autonomous contribution to LVNC, CM-specific Prdm16 knockout mice have been generated with Myh6-Cre, but unlike the germline null mutation, the mutant mice manifested dilated cardiomyopathy 5 months after birth²⁸. Consistent with this postnatal phenotype, another conditional Prdm16 knockout mouse line was created with Mesp1-Cre, which drives Cre expression in anterior mesoderm, including cardiac progenitor cells, and these mice showed age-dependent onset of hypertrophic cardiomyopathy²⁹, a less severe phenotype than that of Myh6-Cre CM-specific Prdm16 knockout mice. Together, these observations point to the possibility that PRDM16 deficiency-induced LVNC may not solely reflect a CM autonomous requirement.

As part of our long-term interest in understanding mechanisms underlying LVNC, we have embarked on our own efforts to create Prdm16 CM-specific knockout mice by using two independent Cre lines: Xmlc2-Cre and cTnT-Cre, which are activated in CM as early as E7.5 with high efficiency^{30, 31}. Contrary to the aforementioned mouse models, we found that both of these lines of CM-specific ablation of Prdm16 (Prdm16^cKO) developed LV-specific dilation and biventricular noncompaction, typical of LVNC phenotypes in patients. This is reminiscent of the observation that different CM-specific Cre lines show distinct cardiac phenotypes due to variable recombination efficiencies in different conditional knockout genes³². More importantly, we seized this opportunity to dissect the disease mechanism autonomous to CM, revealing that PRDM16 functions as a compact myocardium-enriched transcription factor required to maintain compact myocardial CM identity. This is accomplished by activating a subset of essential compact myocardial genes (herein referred to as ‘compact genes’) while repressing a subset of trabecular myocardial genes (herein referred to as ‘trabecular genes’). Consequently, the loss of PRDM16 leads to a switch in gene signature associated with compact CM identity to that characteristic of trabecular CM and/or neurons in LV compact myocardium. Our data thus shed critical lights on an LV-specific transcriptional program instituted by PRDM16 and underscore the role of defective compact myocardium in the pathogenesis of LVNC.

Methods

All high-throughput sequencing data have been made publicly available at the Gene Expression Omnibus (GEO, accession No. GSE179393). Other data and study materials are available from the corresponding author upon reasonable request. The sources of reagents and detailed methods are provided in the online-only Data Supplement. All animal procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of the University of California San Diego with approved protocol # S01049.

Results

Cardiomyocyte-specific deletion of PRDM16 recapitulates left ventricular noncompaction

To study the function of PRDM16 in the heart, we generated cardiomyocyte (CM)-specific Prdm16 knockout (KO) mice (Prdm16^fl/fl::Xmlc2-Cre^+/−, hereafter Prdm16^cKO) by crossing Prdm16^fl/fl mice³³ with Xmlc2-Cre mice, which takes advantage of the Xenopus myosin light chain 2 promoter with dramatically enhanced activity and specificity in mouse CM^{31, 34}. Fluorescence in situ hybridization (FISH) revealed Prdm16 expression overlapped with that of Hey2, a well-established marker for compact myocardium^{35, 36} (Figure 1A). Immunofluorescence (IF) further demonstrated PRDM16 was predominantly localized to CM nuclei of compact myocardium and interventricular septum (IVS) (Figure 1B). In contrast, cardiac transcription factor Nkx2–5 was expressed evenly between compact and trabecular CMs (Figure 1B). The mRNA and protein levels of PRDM16 were greatly reduced in Prdm16^cKO mice, confirming specificity of Prdm16 FISH probes and PRDM16 antibody (Figure S1A–S1B). Western blot further confirmed the absence of PRDM16 in whole heart (Figure S1C).

Figure 1. Cardiomyocyte-specific deletion of PRDM16 recapitulates left ventricular noncompaction.

(A) Representative images of Prdm16 and Hey2 in situ hybridization (ISH) performed on E13.5 mouse embryonic heart cryosections. DNA was stained with DAPI. LV, left ventricle. RV, right ventricle. IVS, interventricular septum. LA, left atrium. RA, right atrium. Scale bars, 0.2 mm (whole heart view) and 0.1 mm (magnified view). See also Figure S1A.

(B) Representative images of PRDM16, Nkx2–5 and CD31 immunofluorescences (IF) performed on E13.5 mouse embryonic heart cryosections. Compact myocardium (Com) or IVS CMs (Nkx2–5-positive) with high expression of PRDM16 are indicated by magenta arrows. Trabecular myocardium (Tra) CMs with low expression of PRDM16 are indicated by magenta arrowheads. Endothelial cells (CD31-positive) without expression of PRDM16 are indicated by cyan arrows. The approximal boundaries between compact myocardium and trabecular myocardium are depicted by yellow dashed lines. Scale bars, 50 μm. See also Figure S1B.

(C) Representative wholemount and H&E stained cryosection images of P1 control and Prdm16^cKO mouse hearts. White arrowhead indicates a cleft at the apex of Prdm16^cKO heart. White arrow indicates the enlarged left ventricle of Prdm16^cKO heart. Scale bars, 0.5 mm (whole heart view of wholemount and H&E) and 0.2 mm (magnified view on H&E stained cryosections). See also Figure S1D–G.

(D) Measurements of left ventricular compact myocardium (LV-Com), IVS and right ventricular compact myocardium (RV-Com) thicknesses on control and Prdm16^cKO mice (n = 3–5 per group) heart sections (n = 4 per heart) from E12.5 to P3 (IVS thicknesses were measured from E15.5 to P3). Data are represented as mean ± SEM. Statistical significance was determined with two-tailed Student’s t test (ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001).

(E-H) Echocardiographic parameters fraction shortening (FS) (E), left ventricle internal dimension, systolic (LVIDs) (F), left ventricle internal dimension, diastolic (LVIDd) (G) and interventricular septum thickness, diastolic (IVSd) (H) of control and Prdm16^cKO neonatal mice (n = 4–6 per group) from P0 to P3. Data are represented as mean ± SEM. Statistical significance was determined with two-tailed Student’s t test (ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

(I-J) Representative images (I) and quantitative analyses (J) of EdU positive (EdU⁺) CMs from Nkx2–5 and EdU IF performed on E13.5 (left panels) and E15.5 (right panels) control and Prdm16^cKO hearts (E13.5: n = 7 hearts per group, n = 3 sections per heart; E15.5: n = 6 hearts per group, n = 3 sections per heart). Arrows indicate examples of proliferating CMs positive for both Nkx2–5 and EdU. White dashed lines indicate the approximal boundaries between compact myocardium and trabecular myocardium. Scale bars, 50 μm. Data are represented as mean ± SEM. Statistical significance was determined with two-tailed Student’s t test (ns, not significant; **p < 0.01, ****p < 0.0001).

See also Figure S1 and Figure S2.

In line with PRDM16’s prominent expression in heart, we found that Prdm16^cKO mice exhibited cardiac abnormalities, including clefts at the apices of the hearts and enlarged left ventricles as early as embryonic day (E) 15.5 (Figures 1C and S1D), with progressive cardiac dysfunction leading to death of all Prdm16^cKO mice before postnatal (P) day 7 (Figure S1I). This fully recapitulates the abnormal heart morphology observed in Prdm16 global KO mice²⁷, indicating that the lost cell autonomous function of PRDM16 in CM is sufficient to account for the morphological defects. By H&E staining, we noticed that Prdm16^cKO mice developed biventricular noncompaction (Figures 1C and S1E–S1H), as seen in previous LVNC mouse models^17–19. However, unlike the previous models, we also observed dramatic dilation that occurred only in LV of Prdm16^cKO mice (Figures 1C and S1E–S1G). Corroborating these observations, measurements of compact myocardium and interventricular septum (IVS) thickness on heart sections demonstrated thinner LV compact myocardium and IVS in Prdm16^cKO mice, starting from E15.5, whereas the thickness of RV compact myocardium was not significantly different between Prdm16^cKO mice and littermate controls (Figure 1D). As LVNC accompanied by LV dilation and dysfunction in human patients leads to far worse clinical outcomes compared with isolated LVNC¹², we then sought to determine the functional consequences of observed abnormal heart morphology in Prdm16^cKO mice. To this end, we performed echocardiography on postnatal day (P) 0 to P3 neonatal Prdm16^cKO mice and littermate controls, and detected severe cardiac contractile dysfunction (Figure 1E), increased left ventricular chamber size (Figure 1F–1G), and thinned IVS (Figure 1H). These data demonstrated that our model largely mirrored the LVNC phenotype in patients.

Role of PRDM16 in embryonic versus adult cardiomyocyte

Thinned LV compact myocardium and IVS may result from decreased cell proliferation or increased apoptosis in Prdm16^cKO mice. To differentiate between these possibilities, we used EdU labeling³⁷ to measure the percentage of proliferating CMs in Prdm16^cKO mice and littermate controls. Concordant with specific decreases in thicknesses of LV compact myocardium and IVS, we found significant reductions in EdU-positive CM percentages only in LV compact myocardium and IVS of Prdm16^cKO mice at both E13.5 and E15.5 (Figure 1I–1J), developmental stages respectively preceding and following overt morphological abnormalities observed in Prdm16^cKO hearts (Figure S1D–S1G). On the other hand, we did not observe any differences in CM apoptosis as measured by cleaved caspase 3 (cCSP3) immunofluorescence³⁸ (Figure S1J–S1K), indicating that thinning of LV compact myocardium and IVS result largely from the reduction in CM proliferation in the critical regions coincident with PRDM16 expression. This is in direct contrast to the lack of impact on CM proliferation when conditional Prdm16^fl/fl mice were bred with Mesp1-Cre mice²⁹, suggesting that the observed late onset of the disease phenotype in this mouse model may result from inefficient depletion of the gene during development. To further establish the CM autonomous effect of Prdm16 KO, we used an independent cTnT-Cre line³⁰ to generate Prdm16 CM-specific KO, which again displayed postnatal lethality before P7 and dramatically dilated left ventricle (Figure S1L), indicating that cardiac phenotypes we observed in Prdm16^cKO mice were not Cre driver line dependent. Taken together, these findings suggested that PRDM16 was essential to ventricle growth and maturation during embryonic and neonatal heart developmental stages.

Because our data clearly show developmental defects in early embryos, we next sought to elaborate the function of PRDM16 in adult CMs. To this end, we generated tamoxifen-inducible CM-specific Prdm16 knockout mice (Prdm16^fl/fl::αMHC-MerCreMer^+/−, Prdm16^icKO)³⁹ to circumvent perinatal lethality of Prdm16^cKO mice. To our surprise, Prdm16^icKO mice did not show overt cardiac phenotypes, even 36 weeks following induction (Figure S2A), despite efficient deletion of PRDM16 (Figures S2B). We noticed that PRDM16 protein levels were much lower in adult hearts than postnatal hearts (Figure S2B). To investigate the developmental expression pattern of PRDM16 in the heart, we performed western blot analyses using antibody to PRDM16 with wild type heart samples from embryonic to adult stages, and found that PRDM16 protein levels dramatically decreased after birth (Figure S2C–S2D). This observation is consistent with recently published human and mouse heart RNA-seq data across multiple developmental and adult stages (Figure S2E–S2F)⁴⁰. Collectively, our findings indicate that PRDM16 plays a negligible role in adult CMs, consistent with the developmental origin of LVNC⁴¹.

Loss of PRDM16 leads to widespread gene misregulation in developing heart

The requirement of PRDM16 for ventricle growth and maturation in developing heart suggests that its regulated gene expression program is pivotal to these processes. To determine gene expression differences between control and Prdm16^cKO mice, we collected embryonic heart ventricles at E13.5, a developmental stage before Prdm16^cKO hearts showed overt cardiac phenotypes, for RNA-seq analysis. In addition, to investigate the underlying mechanism for the selective impact of Prdm16 ablation on LV relative to RV, we performed RNA-seq on dissected LV and RV from E13.5 Prdm16^cKO and littermate control mice. Our RNA-seq data verified chamber-specific expression of LV genes (e.g., Hand1 and Tbx5)^{42, 43} and RV genes (e.g., Pcsk6)⁴⁴ (Table S1), indicating our data faithfully reflected gene expression differences between LV and RV. Principal component analysis (PCA) revealed a clear separation between Prdm16^cKO and control on LV samples but less so on RV samples (Figure 2A), indicating the transcriptomic differences were most pronounced in LV upon loss of PRDM16. Using false discovery rate (FDR) < 0.05, we determined differentially expressed genes (DEGs) in LV with 1,448 downregulated and 1,508 upregulated, which greatly outnumbered those in RV or with whole ventricles (Figure 2B and Table S2). We further noted that commonly altered gene expression events in both chambers tended to be more pronounced in LV than those detected in RV (Figure 2C), suggesting that PRDM16 is involved in a chamber-specific gene expression program.

Figure 2. Loss of PRDM16 leads to widespread gene misregulation in developing heart.

(A) Principal component analysis of RNA-seq data from whole ventricle, left ventricle (LV) and right ventricle (RV) samples of E13.5 Prdm16^cKO and littermate control mice.

(B) Diagram showing the overlap of upregulated or downregulated differentially expressed genes (DEGs) from whole ventricle, left ventricle (LV) and right ventricle (RV) samples of E13.5 Prdm16^cKO mice compared with littermate controls (FDR < 0.05). See also Table S2.

(C) Correlation analysis of shared DEGs between LV and RV of E13.5 Prdm16^cKO mice. Upregulated or downregulated DEGs with bigger fold changes in either LV or RV are indicated by dots with different colors. Gene number of each category is indicated.

(D-F) Gene ontology analysis of LV-RV shared DEGs (D), LV-specific DEGs (E) and RV-specific DEGs (F), which was divided into four categories: biological process (BP), cellular component (CC), molecular function (MF) and KEGG pathway. Notable GO terms or pathways are highlighted in red (upregulated) or blue (downregulated).

See also Figure S3.

Gene ontology (GO) analysis revealed some of the most enriched GO terms in both whole- ventricle-downregulated DEGs and LV-RV-shared downregulated DEGs were associated with mitochondrial biogenesis and function (Figures S3A and 2D), as observed earlier in heart of Prdm16^fl/fl::Mesp1-Cre mice²⁹. We further noted that while specifically downregulated genes in LV were mostly enriched in mitochondrial pathways (Figure 2E), those specifically downregulated in RV did not show any significant GO enrichment (Figure 2F). Since mitochondria accounts for nearly all ATP production supporting heart contraction⁴⁵, impaired mitochondrial function likely contributed to decreased heart contractile function in Prdm16^cKO mice (see Figure 1E). We also found that genes involved in fatty acid metabolism were downregulated in both LV and RV (Figure 2D), suggesting that PRDM16 shares some common transcriptional regulatory functions between CM and adipocyte⁴⁶.

Compared to downregulated genes, the most enriched GO terms in whole-ventricle-upregulated and LV-RV-shared upregulated genes were related to neuronal development pathways (Figures S3A and 2D), indicating that PRDM16 may repress alternative cell fates such as neurons in CMs, similar to the observations made with Nkx2–5⁴⁷ and Tbx20⁴⁸. We confirmed that Sox3, a gene encoding a transcription factor expressed in neural progenitor cells and critical for neural stem cell maintenance⁴⁹, was upregulated in E13.5 Prdm16^cKO ventricles by immunofluorescence microscopy. Interestingly, similar to PRDM16, Sox3 was predominantly localized to CM nuclei in compact myocardium and its upregulation in Prdm16^cKO was more prominent in LV than RV (Figure S3B). Last, but not least, the TGF-β signaling pathway was specifically upregulated in LV compared to RV (Figure 2E), and given the negative role of TGF-β signaling in CM proliferation⁵⁰, this may partially account for the LV-restricted decrease in CM proliferation in Prdm16^cKO mice.

PRDM16 functions as a bifunctional transcription regulator in the genome

DEGs in Prdm16^cKO mice may be directly or indirectly regulated by PRDM16. To identify direct targets for PRDM16, we performed PRDM16 chromatin immunoprecipitation deep sequencing (ChIP-seq) on E13.5 mouse heart ventricles. We identified 17,587 peaks in control samples (Figure 3A), compared to only 330 peaks in Prdm16^cKO samples, indicative of the high specificity with our ChIP-seq experiments (Figure S4A). Among PRDM16-specific ChIP-seq peaks, 35.7% reside in intergenic regions, 45% in introns, 2.9% in exons and 12.3% at promoters (Figure 3A). Intersection of the PRDM16 ChIP-seq peaks with publicly available histone modification ChIP-seq profiles⁵¹ revealed co-occupation of PRDM16 with H3K27ac, an active enhancer marker, H3K4me1, a poised and active enhancer marker, H3K4me3, a marker for promoters, as well as other histone modifications associated with enhancers and promoters (Figure 3B). This is further illustrated on the Nppb-Nppa locus (Figure 3C), which encodes for atrial natriuretic factor and brain natriuretic peptide critical for heart development⁵².

Figure 3. PRDM16 functions as a bifunctional transcription regulator in the genome.

(A) Genomic distributions of PRDM16 ChIP-seq peaks.

(B) H3K4me2, H3K4me3, H3K27ac, H3K9ac, H3K4me1, H3K27me3, H3K9me3 and H3K36me3 profiles within a 12 kb window centered on PRDM16 binding sites.

(C) Integrative genomics viewer (IGV) view showing histone modifications and PRDM16 binding at Nppb-Nppa locus. Enhancers and promoters are indicated by green and purple boxes, respectively. Direction of transcription is indicated by black arrows.

(D) De novo (left panel) and known (right panel) motif analyses of PRDM16 ChIP-seq peaks. See also Figure S4B.

(E) Pie chart showing percentages of whole ventricle DEGs with PRDM16 binding sites within 2 kb of their transcription start sites (TSSs) (proximal), with PRDM16 binding sites within 2 kb-200 kb of their TSSs (distal), with proximal and distal PRDM16 binding sites (proximal+distal) and DEGs without any PRDM16 binding sites within 200 kb of their TSSs (None). See also Figure S4C–D.

(F) PRDM16 binding profiles of control and Prdm16^cKO ventricles within a 4 kb window centered on TSSs of whole ventricle DEGs.

(G) Gene ontology analysis of LV-specific DEGs with promoter PRDM16 binding divided into four categories: biological process (BP), cellular component (CC), molecular function (MF) and KEGG pathway. Notable GO terms or pathways are highlighted in red (upregulated) or blue (downregulated).

See also Figure S4.

Motif analysis of PRDM16 ChIP-seq peaks using Homer⁵³ revealed binding motifs for several cardiac transcription factors (TFs), such as MEF2 family TFs, Tbx20, TEAD and GATA family TFs (Figure 3D), pivotal for heart development^{54, 55}. Of note, we did not retrieve the putative PRDM16 binding motif⁵⁶, which echoes other studies suggesting that PRDM16 functions as a cofactor with other DNA-binding TFs^{57, 58}. Indeed, we observed extensive co-occupation of PRDM16 with Mef2c, Tbx5, Nkx2–5⁵⁹ and Gata4⁶⁰ in embryonic mouse heart (Figure S4B).

By examining the positions of PRDM16 ChIP-seq peaks relative to transcription start sites (TSSs) of E13.5 whole ventricle DEGs, we found that approximately one-third of DEGs had PRDM16 peaks at their promoters and for the 62.6% of DEGs without PRDM16 binding at their promoters, we were able to identify at least one PRDM16 peak within 200 kb of their TSSs (Figure 3E), indicating that PRDM16 may regulate a subset of DEGs by occupying their enhancers. Corroborating this idea, PRDM16 peaks were found on two in vivo validated enhancers of Hey2⁶⁰, a downregulated DEG encoding a transcription factor that has important roles in ventricular wall morphogenesis⁶¹ (Figure S4C), as well as on two in vivo validated enhancers of Tbx3⁶², a upregulated DEG which encodes a transcription factor regulating a myriad of gene programs in the cardiac conduction system⁶³ (Figure S4D). These observations implied that PRDM16 is a bifunctional transcriptional regulator, acting to activate or repress gene expression in a context-dependent manner, likely through cooperating with different sets of TFs in different action sites in the genome.

To further establish such context-dependent action of PRDM16, we chose to focus on DEGs that showed PRDM16 binding at their promoters (Figure 3F), thus likely to be its direct targets (Table S3). Ontology analysis on LV-specific, RV-specific or LV-RV common DEGs with PRDM16 binding at their promoters revealed LV-specific downregulated target genes were enriched in GO terms directly related to heart development and function (Figure 3G), whereas similar analysis performed on RV-specific or LV-RV-shared downregulated target genes did not recover any of these GO terms (Figure S4E). On the other hand, both LV-specific and LV-RV-shared upregulated target genes were enriched for neuronal development pathways (Figures 3G and S4E), suggesting that PRDM16 directly suppresses the expression of neural genes in both ventricles. Together, these findings suggest that PRMD16 may act as a transcription activator to drive cardiogenesis, but as a transcription repressor to restrict neurogenesis in developing heart.

Cooperation of PRDM16 with LV-enriched TFs

While PRDM16 depletion clearly was more impactful on LV than RV, our FISH, immunostaining, and RNA-seq studies found little significant difference in PRDM16 expression between LV and RV (see Figures 1A, 1B and Table S1). To further confirm this, we dissected LVs and RVs from E13.5 mouse embryos and performed western blots to compare their PRDM16 protein levels. Concordant with our previous data, western blot analyses of PRDM16 levels showed that LV had the same amount of PRDM16 as RV. In comparison, using Tbx5 as a known LV-enriched TF⁴², we confirmed that LV had more than twice the amount of Tbx5 than RV (Figure 4A–4B).

Figure 4. Cooperation of PRDM16 with LV-enriched TFs.

(A-B) PRDM16 and Tbx5 western blots (A) and their densitometry quantification results (B) normalized to GAPDH using protein samples extracted from LV or RV of E13.5 wild type embryos (n = 3 per group). Samples from LV or RV of Prdm16^cKO embryos were used as negative controls. Data are represented as mean ± SEM. Statistical significance was determined with two-tailed Student’s t test (ns, not significant; ***p < 0.001).

(C) Correlation analysis of LV and RV PRDM16 binding intensity at PRDM16 ChIP-seq peaks calculated from whole ventricle PRDM16 ChIP-seq data. The threshold for defining LV- or RV-specific peaks was set at 1.5 folds.

(D) Ratio of PRDM16 binding density between LV and RV at genomic regions as indicated. Statistical significance was determined with permutation test with 1,000,000 iterations.

(E-F) IGV view showing histone modifications and LV-, RV-PRDM16 binding at Hey2 (E) and Mylpf (F) loci. Promoters are indicated by purple boxes.

(G) Relative transcription factor motif enrichment at TSS of LV-specific or RV-specific downregulated DEGs. Notable TF motifs were highlighted in red. See also Figure S5.

(H) Ratio of PRDM16 binding density between LV and RV at promoters of LV-specific downregulated DEGs, with or without the presence of Tbx5 motif. Statistical significance was determined with permutation test with 1,000,000 iterations.

Since the LV-restricted regulation of cardiac genes cannot be explained by chamber-specific expression of PRDM16 in LV, we then asked if PRDM16 had differential affinity for key target genes in LV versus RV. To this end, we performed PRDM16 ChIP-seq separately on LV or RV dissected from E13.5 mouse embryos. A majority of LV and RV PRDM16 peaks had similar binding intensities (Figure 4C), indicating that PRDM16 generally occupies similar genomic loci in both ventricles. However, we identified 2,643 LV-specific peaks and 2,753 RV-specific peaks (Figure 4C), underscoring the chamber-specific binding preference of PRDM16 at these loci. We then compared LV and RV PRDM16 binding intensities at the promoters of LV-specific or RV-specific DEGs, and found PRDM16 binding in LV was significantly stronger than that observed in RV at promoters of LV-specific downregulated DEGs (Figure 4D). In contrast, PRDM16 binding at promoters of RV-specific downregulated DEGs or LV-, RV-specific upregulated DEGs were not significantly different between two chambers (Figure 4D). For example, the promoter of Mycn, which was specifically downregulated in LV of Prdm16^cKO mice, showed much greater PRDM16 binding in LV than RV (Figure 4E). On the contrary, Mylpf, which was specifically downregulated in RV of Prdm16^cKO mice (Table S2), had nearly identical PRDM16 binding at its promoter between LV and RV (Figure 4F). These findings indicated that PRDM16 had higher affinity for LV-specific downregulated genes in LV than RV, which may explain why the loss of PRDM16 in Prdm16^cKO mice has a much larger impact on the expression of these genes in LV.

We then analyzed TF binding motifs at promoters of LV- or RV-specific downregulated DEGs. Interestingly, we found motifs of LV-enriched TF Hand1 and Tbx5 were among the highest-ranked TF motifs for LV-specific downregulated DEGs (Figure 4G). In fact, PRDM16 peaks harboring the Tbx5 motif generally had higher density in LV than RV at the promoters of LV-specific downregulated DEGs (Figure 4H), suggesting that PRDM16 may cooperate with LV-enriched TFs, such as Hand1 and Tbx5, to activate the expression of its target genes in LV. On the other hand, Hand1 and Tbx5 were less enriched at promoters of LV-specific upregulated DEGs (Figure S5), indicating these LV-enriched TFs mainly function as activators when cooperating with PRDM16 in LV. Interestingly, another high-rank TF motif associated with LV-specific downregulated DEGs corresponded to the Mecom/PRDM3 motif (Figure 4G). As the PRDM16 DNA binding motif is almost identical to the PRDM3 motif⁵⁶, this result suggested that PRDM16 may act as a direct DNA-binding TF at a subset of LV-specific downregulated DEGs.

PRDM16 activates compact genes while repressing trabecular genes in LV compact myocardium and IVS

To gain further mechanistic insight into LV- and IVS-specific dilation in Prdm16^cKO mice, we focused on LV-specific downregulated target genes Hey2 and Mycn as they encode compact myocardium-restricted transcription factors^{35, 64} essential for ventricle wall morphogenesis^{61, 65}. We confirmed LV-specific downregulation of Hey2 and Mycn in Prdm16^cKO mice using qRT-PCR (Figure S6A). Strikingly, by FISH, we found that the expression of both Hey2 and Mycn was drastically decreased in LV compact myocardium and IVS, but remained unchanged in RV myocardium of Prdm16^cKO mice (Figure 5A–5B). Notably, the differential FISH signals appeared more pronounced than qRT-PCR signals, particularly with Hey2 (Figures 5A, 5B and S6A). Consequently, the expression of Mlc2a/Myl7 and Tbx5, which are normally repressed by Hey2 in compact myocardium^{35, 61}, were increased in IVS and LV compact myocardium (Figure S6B). Furthermore, we observed the ectopic expression of the trabecular myocardium marker gene Nppa³⁵ in the IVS and some regions of the LV compact myocardium of Prdm16^cKO mice (Figure 5A–5B). Consistent with this, the trabecular-compact expression gradient, where the expression of trabecular genes (e.g., Nppa) gradually decreases from trabecular myocardium to compact myocardium⁴⁴, disappeared in LVs of Prdm16^cKO mice, as the expression of Nppa expanded to the base of the trabeculae (Figure 5A–5B). This suggests that the gradient of gene expression from trabecular myocardium to compact myocardium was disrupted. Moreover, the chemokine ligand gene Cxcl12 was also dramatically decreased in LV compact myocardium but not in LV trabeculae of Prdm16^cKO mice (Figure 5C). Collectively, these findings indicated that PRDM16 helps maintain compact myocardial CM identity by activating expression of compact myocardial genes (e.g., Hey2 and Mycn), while repressing trabecular myocardial genes (e.g., Nppa) in LV compact myocardium and IVS (Figure 5D).

Figure 5. PRDM16 activates compact genes while repressing trabecular genes in LV compact myocardium and IVS.

(A-B) Representative images of Hey2 (A) or Mycn (B) and Nppa in situ hybridization (ISH) performed on E13.5 mouse embryonic heart cryosections. DNA was stained with DAPI. LV, left ventricle. RV, right ventricle. IVS, interventricular septum. LA, left atrium. RA, right atrium. White arrows indicate ectopic expression of Nppa in LV compact myocardium. Green arrow indicates the expression of Nppa expands to base of trabeculae in LV. White dashed lines indicate the approximal boundaries between compact myocardium and trabecular myocardium. Scale bars, 0.1 mm.

(C) Representative images of Cxcl12 and Tbx20 (compact myocardium marker) in situ hybridization (ISH) performed on E13.5 mouse embryonic heart cryosections. LV compact myocardium where Cxcl12 was specifically reduced in Prdm16^cKO is indicated by white arrows. Scale bars, 0.1 mm.

(D) Illustration depicting the expression of compact myocardium marker genes (compact genes) were reduced in compact myocardium and IVS of Prdm16^cKO mice, whereas the expression of trabecular myocardium marker genes (trabecular genes) expanded from trabeculation to compact myocardium and IVS.

See also Figure S6.

Single-cell RNA-seq uncovers a PRDM16-anchored transcription network in LV

Reduced expression of a subset of compact CM marker genes and ectopic expression of a subset of trabecular CM marker genes suggest that PRDM16 deficiency may cause improper specification of compact CM identity in LV compact myocardium. To test this hypothesis, we performed single-cell RNA-seq (scRNA-seq), which provides much greater resolution than bulk RNA-seq⁶⁶, to compare between Prdm16^cKO and control mice. To focus on CMs in scRNA-seq, we crossed a Rosa26-tdTomato reporter transgene⁶⁷ into the Prdm16^cKO background to label CMs with tdTomato, which enabled fluorescence-activated cell sorting (FACS) to obtain pure single CM preparations (Figures 6A and S7A). We obtained 7,783 high quality scRNA-seq profiles from two control and two Prdm16^cKO samples. Unsupervised clustering separated CMs (Nkx2–5+), epicardial cells (Tcf21+)⁶⁸ and atrioventricular cushion mesenchyme cells (Postn+)⁶⁹ (Figure 6B). Owing to the FACS enrichment of CMs prior to scRNA-seq, our dataset predominantly consists of CMs (Figure 6B). We also identified clusters of outflow tract (OFT) CMs that expressed the OFT marker gene Rspo3⁴⁴ (Figure S7B).

Figure 6. Single-cell RNA-seq uncovers a PRDM16-anchored transcription network in LV.

(A) Overview of single-cell RNA-seq (scRNA-seq) and spatial transcriptomics (ST) experimental design.

(B) UMAP plots of FACS-sorted E13.5 CMs, showing the expression of CM marker gene Nkx2–5, epicardial cell maker Tcf21 and atrioventricular cushion mesenchyme cell marker Postn. See also Figure S7B.

(C) ST expression profiles and representative FISH images of selected anatomical marker genes in E13.5 control and Prdm16^cKO hearts. Scale bars, 0.1 mm. See also Figure S7I and Figure S7J.

(D) Dot plot of CM clusters that were grouped into LV compact myocardial CMs (LV-Com), RV compact myocardial CMs (RV-Com), LV trabecular myocardial CMs (LV-Tra) and RV trabecular myocardial CMs (RV-Tra) based on marker gene expression. FPM, fragments per million reads.

(E) UMAP plots of CM populations assigned to their original anatomical locations (upper panel) or genotypes (lower panel).

(F) UMAP plots of selected anatomical marker gene expression in CM populations. See also Figure S8A.

(G) UMAP plots of LV compact myocardial CMs assigned to subclusters (upper panel) or genotypes (lower panel). See also Figure S8B.

(H) Heatmap showing most-enriched genes in subclusters of LV compact myocardial CMs. See also Figure S8C.

(I) UMAP plots of LV compact myocardial CMs showing selected subcluster marker genes. Red dashed lines indicate the approximate positions of Prdm16^cKO LV compact myocardial CMs in UMAP plot. See also Figure S8D.

To facilitate mapping CM clusters to their original locations in the heart, we also performed Spatial Transcriptomics (ST)⁷⁰ on E13.5 and E15.5 control and Prdm16^cKO hearts (Figure 6A). Unsupervised clustering of each sample generated four clusters that mapped back to their original locations (Figure S7C–S7D). To identify marker genes for trabecular myocardium and compact myocardium, we selected their corresponding ST spots and calculated their top enriched genes that were consistent across all samples (Figure S7E–S7F) and validated some of them by FISH (Figure 6C). We also discovered Cttnbp2, a gene encoding a neural-specific F-actin regulator⁷¹, as a specific marker for LV compact myocardium of Prdm16^cKO hearts and validated it with FISH (Figure S7I). We then used similar approach, combining with LV and RV bulk RNA-seq data (Table S1), to identify LV and RV marker genes (Figure S7G–S7H), some of which were validated by FISH (Figures 6C and S7J).

The assignment of anatomical marker genes enabled us to assign scRNA-seq CM clusters to their original locations (Figures 6D–6F and S8A). As PRDM16 is predominantly expressed in compact myocardium (see Figure 1A–1B), we extracted and reclustered LV-Com CMs and RV-Com CMs and found that control and Prdm16^cKO RV-Com CMs were transcriptionally indistinguishable (Figure S8B–S8C). By contrast, control and Prdm16^cKO CMs of LV-Com were segregated into different clusters, underscoring their significant transcriptomic differences (Figure 6G). Corroborating our findings by bulk RNA-seq that LV compact CMs in Prdm16^cKO ectopically expressed trabecular genes and neural genes, we found that the Prdm16^cKO LV-Com CMs formed a cluster enriched with trabecular genes, such as Nppa, Cited1, Nppb and Mest (Cluster 7) as well as clusters enriched with neural genes, such as Cttnbp2 and Spon1 (Clusters 1, 5, 7 ) (Figures 6H- 6I and S8D), thus further establishing the role of PRDM16 in maintaining the compact CM identity by repressing alternative CM (trabeculation) and other cell fate (neuron) genes. Conversely, control LV-Com CMs showed two clusters (Clusters 0, 6) that were enriched with compact genes, such as Hey2 and Mb (Figure 6H), which were markedly decreased in Prdm16^cKO LV-Com CMs (Figures 6I and S8D). These data confirmed that PRDM16 activates compact genes in LV compact myocardium. Interestingly, Cluster 4 enriched for Cxcl12 was exclusively included in control CMs (Figure 6G–6H), consistent with its specific downregulation in LV compact myocardium of Prdm16^cKO mice (see Figure 5C).

To investigate the underlying gene regulatory network (GRN) in compact myocardial CMs, we applied SCENIC⁷² to our control and Prdm16^cKO compact CMs single-cell data and found twenty TF regulatory modules in all compact CMs (Figure S8E). Interestingly, the activity of five TFs was specifically altered in LV-Com CMs but not in RV-Com CMs upon loss of PRDM16 (Figure S8F). Among the TFs whose activity decreased were Pgc-1α (Ppargc1a), Hey2 and Gata6. On the other hand, the activity of neural TF Sox3 and cardiac conduction system TF Tbx3 were increased. These findings thus provide evidence that PRDM16 plays a key role in orchestrating a transcriptional regulatory network in LV compact myocardium to promote compact CM gene expression while inhibiting inappropriate gene expression of alternative cell fate.

Discussion

The most intriguing question of LVNC is the susceptibility of the LV to disease, in particular dilation and dysfunction. This phenomenon has been poorly recapitulated in previous LVNC animal models in which noncompaction and dilation seem to occur at similar severity in both ventricles^17–19. Our current work demonstrated that the Prdm16^cKO LVNC model is inherently advantageous in studying LV susceptibility in LVNC as the dilation was restricted to LV. This enabled us to use RV as internal control for identifying critical molecular changes responsible for LV-specific phenotypes. Through RNA-seq and PRDM16 ChIP-seq, we identified compact myocardial marker genes Hey2 and Mycn as PRDM16 direct target genes, both of which are essential for CM proliferation in compact myocardium^{61, 65}. In line with the specific reduction of CM proliferation in LV and IVS of Prdm16^cKO mice, these genes were only reduced in these specific areas of the heart, suggesting that their regulation by PRDM16 is spatially restricted.

Although the biventricular noncompaction phenotype might be attributed to misregulation of mitochondrial genes and neural genes that was observed in both ventricles (Figure 2D), the LV-specific dilation indicated that PRDM16 might specifically regulate genes in LV, not in RV, that are essential to ventricular wall morphogenesis. As we did not find any significant differences in PRDM16 protein levels between the two ventricles (Figure 4A), how would PRDM16 selectively regulate these genes in LV? To investigate this, we performed RNA-seq and PRDM16 ChIP-seq independently on each ventricle and found that PRDM16 appeared to be ‘left-handed’, as loss of PRDM16 resulted in significantly decreased expression of a subset of critical target genes in LV, that were barely affected in RV. Motif analyses suggested that this preferential effect of PRDM16 on this subset of LV targets might occur through cooperation with LV-enriched TFs, like Tbx5 and Hand1. Our findings thus provide a tangible explanation for LV susceptibility in PRDM16-related LVNC and open a new avenue for future mechanistic studies on the etiology of LVNC.

To further tackle the spatially restricted function of PRDM16, we performed scRNA-seq in combination with spatial transcriptomic analysis. Using these advanced functional genomics technologies, we now systematically identified anatomical marker genes for E13.5 and E15.5 mouse hearts. We then applied them to our scRNA-seq data and successfully mapped CM clusters back to their anatomical locations, together illustrating how the identity of compact myocardial CMs was compromised in Prdm16^cKO mice. To our knowledge, this is the first integration of scRNA-seq with Spatial Transcriptomics on embryonic mouse heart, which provides a rich resource for studying spatial gene expression at single-cell level during mammalian heart development.

Spatially resolved scRNA-seq data led to the discovery that some LV compact myocardial CMs of Prdm16^cKO transcriptionally resembled trabecular myocardial CMs or even neurons, suggesting PRDM16 is instrumental in maintaining the identity of compact CM. Since compact myocardial CMs are generally more proliferative than trabecular CMs^{36, 73}, the loss of compact CM identity in LV compact myocardium and IVS of Prdm16^cKO animals thereby resulted in the reduction of CM proliferation that ultimately led to LV dilation and dysfunction according to our working model (Figure 7). This observation is reminiscent of PRDM16’s role in brown adipose tissue (BAT) where it activates BAT genes while repressing white adipose tissue genes⁴⁶, and similarly in common progenitor cells of BAT and skeletal muscle, PRDM16 activates BAT genes while repressing skeletal muscle genes⁷⁴. Together these findings all point to a general role of PRDM16 in determining cell fate by promoting a set of lineage-specific genes while inhibiting other sets of genes critical for inappropriate cell fates.

Figure 7. Working model of PRDM16’s transcriptional regulation in the heart.

PRDM16 (purple ovals) expression is restricted to left ventricle (LV)-, right ventricle (RV)-compact myocardium and interventricular septum (IVS). PRDM16 is recruited by direct DNA-binding cardiac transcription factors (grey ovals) to the promoters of its target genes. In LV compact myocardium and IVS, PRDM16 cooperates with LV-enriched Tbx5 or Hand1 to activate the transcription of compact genes and has stronger binding at their promoters (PRDM16 binding intensity is depicted by the size of purple peaks). PRDM16 also suppresses the transcription of trabecular and neural genes in LV compact myocardium and IVS, as well as the transcription of neural genes in RV compact myocardium. The loss of PRDM16 in CMs led to dramatic downregulation of compact genes and upregulation of trabecular genes specifically in LV compact myocardium and IVS, whereas the expression of compact genes and trabecular genes in RV compact myocardium did not change. Despite neural genes were upregulated in both LV and RV compact myocardium, the upregulation tended to be smaller in RV. The compromised compact gene program in LV compact myocardium and IVS results in decreased CM proliferation and ultimately causes LVNC with LV-specific dilation in Prdm16^cKO mice. The upregulation of trabecular genes in both LV and RV trabecular myocardium, along with other commonly misregulated genes in LV and RV, may account for the biventricular noncompaction observed in Prdm16^cKO mice.

A major unsolved puzzle in pathogenesis of LVNC is the relationship between noncompaction and ventricular dilation. A recent study found that decreased CM proliferation in compact myocardium alone was sufficient to cause noncompaction³⁶. Our findings seemed to validate this observation as we saw decreased CM proliferation in LV compact myocardium of Prdm16^cKO mice, and noncompaction was indeed evident in LV. However, we also found noncompaction in RV without any changes in CM proliferation. More importantly, we only saw ventricular dilation in LV. These observations strongly suggest that noncompaction and ventricular dilation are independent phenomena, and noncompaction is not necessarily caused by defects in CM proliferation.

Our findings that LV compact CM of Prdm16^cKO mice expressed decreased levels of compact CM marker genes but ectopically expressed trabecular CM marker genes (Figure 5D), are opposite to observations in previously characterized LVNC-mimicking Mib1^cKO mice¹⁷, which barely expressed trabecular CM marker genes but ectopically expressed compact CM marker genes in their trabecular CMs. Although the trabecular/compact marker gene misregulation of Prdm16^cKO and Mib1^cKO largely took place in compact or trabecular myocardium, both models developed noncompaction and ventricular dilation in LV. Considered together, we propose that proper specification of either compact CM or trabecular CM is important for normal morphogenesis of the ventricular wall, and that disruption of specification of either leads to LVNC.

Clinical Perspective.

What is new?

• Cardiomyocyte specific knockout of Prdm16 leads to left ventricular noncompaction cardiomyopathy with LV-specific dilation and dysfunction.

• PRDM16 maintains compact myocardial cardiomyocyte transcriptomic identity by activating compact myocardial genes while repressing trabecular myocardial genes in LV.

• PRDM16 cooperates with LV-enriched transcription factors to specifically regulate transcription in LV.

What are the clinical implications?

• This study provides a unique LVNC mouse model for developing therapeutic interventions for LV dilation and dysfunction, which emerge as the most important disease features in LVNC.

• Improper specification of compact or trabecular cardiomyocytes is likely a common mechanism in pathogenesis of LVNC.

• Spatial single-cell gene expression profiles in normal or diseased conditions facilitate future studies for identifying new drug targets for LVNC.

Non-standard Abbreviations and Acronyms

LVNC

left ventricular noncompaction cardiomyopathy

LV

left ventricle

RV

right ventricle

IVS

interventricular septum

PRDM16

PR domain-containing 16

HEY2

Hes related family bHLH transcription factor with YRPW motif 2

MYCN

N-Myc proto-oncogene, bHLH transcription factor

NPPA

natriuretic peptide A

CM

cardiomyocyte

DEG

differentially expressed genes

ChIP-seq

chromatin immunoprecipitation deep sequencing

TF

transcription factor

TSS

transcription start site

FACS

fluorescence-activated cell sorting

scRNA-seq

single-cell RNA sequencing

ST

spatial transcriptomics

GRN

gene regulatory network