Dysferlin Enables Tubular Membrane Proliferation in Cardiac Hypertrophy

Proteomic data were deposited to the ProteomeXchange Consortium via the PRIDE²¹ partner repository with the dataset identifiers PXD043108, PXD053051, PXD045738, and PXD053037.

For more detailed methods description and the Major Resources Table please refer to the Supplemental Material.

The generation of constitutive dysferlin-KO and cardiomyocyte-specific JP2 overexpression mice was described previously.^1,22 Sex-mixed mouse cohorts in the C57BL/6J background aged 8 to 16 weeks and 30 to 32 weeks were used for experiments. Mouse cardiomyocytes were isolated using collagenase type II (Worthington) and modified Langendorff-perfusion.^20,23 Mice were anesthetized with 2% isoflurane before cervical dislocation. TAC was induced with a 27 G spacer in 9- to 15-week-old mice after IP anesthesia. Blinded transthoracic echocardiographic measurements were performed in parasternal views with ECG monitoring under isoflurane anesthesia. Animal procedures were approved by the veterinarian state authority (LAVES, Oldenburg, Germany; animal protocols no. T11.2, 33.9-42502-04-18/2975, 33.9-42502-04-16/2102, and 33.9-42502-04-21/3785).

LV biopsies were obtained from patients suffering from severe symptomatic aortic valve stenosis (AS) who underwent transfemoral transcatheter aortic valve replacement according to current guidelines. Myocardial biopsies from nonfailing (NF) donor hearts rejected for transplantation served as controls. The institutional ethics committee approved the study (no. 10/5/16), and written informed consent was obtained from all patients.

For live-cell membrane STED imaging, VMs were incubated with 5 µmol/L of the custom-made fluorescent cholesterol analog Chol-PEG-KK114,²⁴ or 40 µmol/L of di-8-ANEPPS (Molecular Probes) diluted in nominally Ca²⁺-free perfusion buffer. Refer to Table S1 for information on physiological buffer compositions. For multicolor immunofluorescence STED imaging, isolated cardiomyocytes were fixed with 4% PFA on laminin-coated coverslips and permeabilized with 0.2% Triton in PBS. Human myocardial biopsies were fixed in 4% PFA, paraffin-embedded, and cut into 4-µm thick sections. Following deparaffinization and rehydration, epitopes were unmasked in sodium citrate buffer, and samples permeabilized with 0.1% Triton. For information on primary antibodies, see Table S2 and the Major Resources Table. Primary antibody specificity was verified by experiments including knockout mice (dysferlin), overexpression and knockdown mice (JP2), and knockin mice (ryanodine receptor type 2 [RyR2]). In addition, secondary antibody-only controls were used to distinguish genuine target staining from background. STED images were acquired with a Leica TCS SP8 laser-scanning microscope with an HC-PL-APO-C2S 100×/1.40 oil objective. Images were processed in ImageJ/Fiji.

VMs were high-pressure frozen in perfusion buffer plus 8% BSA using a Leica HPM100 system as described.²⁰ Freeze-substitution was performed in an EM-AFS2 (Leica Microsystems) according to Wong et al.²⁵ For ET, semi-thick (300 nm) sections were prepared and imaged using 300 kV Tecnai TF30 (Thermo Fisher Scientific). Tilt series were aligned, reconstructed, and combined using IMOD as described.²⁶

LV tissue and isolated left-VMs from wild-type (WT) versus dysferlin-KO mice at an age of 16 and 32 weeks, and WT sham- versus TAC-operated animals were subjected to data-independent acquisition mass spectrometry (DIA-MS) and quantitative analysis using a hybrid ion mobility/time-of-flight mass spectrometer hyphenated to nanoflow chromatography.²⁷ Using a custom prefractionated MS/MS spectral library, precursors were extracted at 1% FDR, which were then combined to peptide and protein levels. Protein profiles were log-transformed and quantile-normalized before proteome analyses. The raw data have been deposited to the ProteomeXchange Consortium via the PRIDE²¹ partner repository.

Quantitative SDS-PAGE was performed as described.²⁰ Twenty micrograms of protein per lane were resolved by SDS-PAGE on 4% to 20% Tris-HCl protein gradient gels. Full immunoblot scans are provided in the supplemental figures. For coimmunoprecipitation experiments, we used a solubilization buffer containing 0.15% CHAPS and incubated 500 µg protein with 4 µg anti-dysferlin antibody (ab124684) or unspecific rabbit IgG antibody. Dynabeads protein G (Thermo Fisher Scientific) was used to precipitate dysferlin. For cell surface protein crosslinking, isolated VMs were incubated for 30 minutes with 10 mmol/L of the membrane-impermeable crosslinker bis-sulfosuccinimidyl suberate (BS3, Thermo Fisher Scientific).

Membrane pellets (100 000 g) from isolated mouse VMs were solubilized, and 100 µg of protein per lane were separated by blue native gel electrophoresis as described.²⁸ Gel lanes were cut into 60 equally sized slices and subjected to in-gel tryptic digestion. Peptides were analyzed by mass spectrometry (LC-MS/MS) in replicates and processed using MaxQuant software.

Data are presented as mean±SEM. P<0.05 was accepted to indicate statistical difference. Statistical analyses were performed with GraphPad Prism not older than 9.4.0, SPSS Statistics 28.0.1.1, and R versions 4.2.2 and 4.3.1. Details regarding experimental design, sample sizes, normalization procedures, tests establishing normality, named statistical tests, post hoc correction for multiple comparisons, and precise P values are provided in the Supplemental Table: statistical details and the corresponding figure legends. Experiment-wide multiple test correction was not applied. Representative images were selected to visualize the average in each group.

Among the group of 6 ferlin proteins, only dysferlin is substantially expressed in isolated mouse VMs on the RNA transcript level, with myoferlin transcripts at a 9 times lower level (Figure 1A). All other ferlin isoforms including otoferlin, which is essential for hearing as a mediator of the Ca²⁺-triggered exocytosis in hair cells,²⁹ are not expressed at detectable levels in mouse VMs (Figure 1A). Quantitative analysis of immunoblots from mouse myocardial tissue lysates revealed the highest expression of dysferlin protein in the ventricles, compared with 59% relative dysferlin expression in the atria (Figure 1B; Figure S1). Atrial and ventricular tissue lysates from heterozygous dysferlin knockout mice showed ≈25% dysferlin protein expression in comparison to WT myocardium; homozygous dysferlin knockout (hereafter referred to as KO) lysates confirmed the antibody specificity (Figure 1B). Of note, only full-length dysferlin and no cleavage products were detected in mouse myocardium (Figure 1B; Figure S1).³⁰

To localize dysferlin in isolated VMs in situ, we established coimmunostaining protocols for dysferlin and the membrane/tubule marker caveolin-3 for STED nanoscopy (see Supplemental Material for detailed information). Relatively small dysferlin immunofluorescent spots in groups of up to 10 signals accumulated near caveolin-3-positive Z-line striations, apparently associated with endomembrane TAT network components and the lateral surface membrane, where dysferlin signals alternated with caveolin-3-positive regions and did not directly overlap with caveolin-3 at the level of STED resolution (Figure 1C, left). Dysferlin-KO VMs were used to exclude unspecific antibody signals (Figure 1C, right), and Figure S2A documents the superior resolution of STED nanoscopy versus confocal microscopy using the same optical setup. While the majority of dysferlin signals were localized to the TAT network or the lateral surface membrane in VMs, we noted additional dysferlin signals with a high local density at the terminal intercalated disc (ICD) membrane folds (Figure S2A). To further estimate the quantitative distribution of dysferlin at specific membrane domains in the whole cell volume, we acquired confocal Z-stacks of dysferlin-immunostained VMs and applied image segmentation for quantitative immunofluorescence analyses (Figure S2B and S2C). Notably, compared with the predominant signal intensity-weighted dysferlin area fraction associated with the TAT network, only 16% and 7% were attributed to the lateral surface membrane or ICD membrane folds, respectively (Figure S2C). Immunohistological stainings and STED imaging of intact mouse ventricular myocardium confirmed a similar dysferlin and caveolin-3 signal structure and distribution at TAT membranes and ICD membrane folds (Figure S3A and S3B) as shown in isolated VMs. Confocal imaging and quantitative immunofluorescence analyses, however, demonstrated a signal intensity-weighted dysferlin area fraction of 15% at the ICD membrane folds compared with TAT-associated signals, twice as high as seen in isolated VMs (Figure S3C and S3D). This result is explained by 2 neighboring myocytes at the lateral ICD cell poles, whose dysferlin signals cannot be analyzed separately due to highly interdigitated ICD membrane folds.

In AMs, which express a TAT network primarily composed of AT membranes interconnected by sparse TT components,^20,24 STED imaging also identified punctate dysferlin signals accumulating close to the TAT network and alternating with caveolin-3 positive membrane regions (Figure 1D, left). Of note, the lower expression level of dysferlin in atrial myocardium correlated with a significantly smaller TAT network density in AMs versus VMs.²⁰ While dysferlin-KO AMs confirmed the specificity of the immunolabeling approach (Figure 1D, right), immunostainings of VMs and AMs from dysferlin-KO hearts did not show obvious differences in the organization of the caveolin-3 signals (Figure 1C and 1D, right).

At regular intervals, TAT membranes are known to form junctional membrane complexes with the SR, a.k.a. cardiac dyads, nanodomains regulating local intracellular Ca²⁺ release in cardiomyocytes.¹⁶ Hence, we wondered if the dysferlin signals border on SR Ca²⁺ release sites, and thus co-immunostained dysferlin and the cardiac RyR2 Ca²⁺ release channel in VMs. STED imaging in fact confirmed plenty of dysferlin signals in nanometric proximity to RyR2 channel clusters (Figure 1E, left). Accordingly, ET of mouse VMs identified numerous small membrane vesicles in the vicinity of tubule-SR junctions with an average diameter of 76.9±3.6 nm (Figure S4A and S4B). Moreover, immunoelectron microscopy confirmed dysferlin signals in vesicular membranes next to tubule-SR junctions (Figure S4C), suggestive of a nearby dysferlin-positive vesicular compartment.

To further characterize dysferlin and RyR2 signals spatially, immunofluorescence in both STED channels was subjected to image segmentation protocols³¹ before quantitative signal analysis (Figure 1E, right). Intracellular dysferlin immunofluorescent spots were 1 order of magnitude smaller than RyR2 clusters (average cluster size 0.006 versus 0.05 µm²) but exhibited a significantly higher cluster density (Figure 1F and 1G). Importantly, the dysferlin cluster frequency increased exponentially with smaller distances to RyR2 (Figure 1H), while the average nearest neighbor distance of dysferlin to RyR2 clusters was 0.11 µm small (Figure 1I). Accordingly, co-immunostaining of dysferlin and the L-type Ca²⁺ channel (LTCC) Ca_V1.2, which controls voltage-dependent Ca²⁺ influx through membrane tubules into cardiac dyads, confirmed a similar arrangement of dysferlin signals immediately next to Ca_V1.2 clusters (Figure S5A). However, our attempts to resolve the specific membrane compartment of dysferlin signals by co-immunostaining with vesicle markers have, to date, failed. For example, dysferlin colocalized neither with the vesicular coating protein clathrin (Figure S5B) nor with the synaptic vesicle protein synaptotagmin-1 (Figure S5C). Together, these data establish dysferlin as the predominant ferlin expressed in VMs that localizes in close proximity to Ca²⁺ release units, where it exists in a previously unknown vesicular compartment.

As dysferlin is the only ferlin expressed at substantial protein levels in VMs, we aimed to reassess the cardiac phenotype in 10 to 16 versus 30 to 32 weeks old dysferlin-KO mice. In total, we observed 202 animals per genotype and did not identify an increased mortality in the overall cohort (Figure 2A). However, more detailed analyses revealed that dysferlin-KO male mice showed a significantly higher mortality of ≈10% in the 32-week observation period (Figure 2B) in contrast to unimpaired dysferlin-KO female mice (Figure S6A). Importantly, none of the observed KO animals presented obvious symptoms related to muscular dystrophy in routine animal observation. In addition, body weight and heart weight were not changed between WT and KO animals at any time point (Figure S6B). However, investigator-blinded transthoracic echocardiography showed that the LV end-diastolic diameter increased significantly, whereas LV anterior wall thickness and LV ejection fraction decreased at higher ages in dysferlin-KO animals (Figure S6B; Figure 2C; Tables S3 and S4). The echocardiographic changes indicated a progressive but mild form of dilated cardiomyopathy in 30-week-old dysferlin-KO mice, which may contribute to the slightly increased mortality in male mice. Interestingly, the heart rate during echocardiography was reduced in young and older dysferlin-KO mice (Figure 2C, right), implicating a role of dysferlin for cardiac pacemaking and conduction propagation, which may be the subject of future studies.

On a single-cell basis, isolated dysferlin-KO VMs exhibited an increased cell area and surface perimeter, mainly reflecting an enlarged cell length compared with WT cells (Figure 2D), contributing to the mild dilated cardiomyopathy phenotype at 30 weeks. We also sought to analyze the TAT network in 30-week-old dysferlin-KO mice using the bright, photostable fluorescent cholesterol analog Chol-PEG-KK114²⁴ for live confocal membrane imaging. Surprisingly, we could not identify changes in TAT network density or individual TAT component orientations at weeks 30 to 32 (Figure 2E) in contrast to a previous report for 50- to 60-week-old mice.¹⁴

To investigate proteomic changes at higher ages, we used ventricular tissue from WT versus dysferlin-KO hearts at 16 versus 32 weeks of age for DIA-MS analyses after tissue lysis by pressure-cycling technology.³² Based on 2 technical replicate injections from each n=5 WT versus KO hearts of both ages, DIA-MS reproducibly quantified 3746 proteins across all samples, of which 1265 proteins showed statistically significant abundance changes. Unsupervised hierarchical clustering clearly segregated WT from dysferlin-KO and 16 from 32 weeks old ventricular tissue samples as presented in the heat map in Figure 2F. Most interestingly, the pairwise comparison of 16-week-old mouse hearts commonly revealed 34 downregulated versus 28 upregulated proteins, while 69 downregulated versus 107 upregulated proteins were identified by the 32-week pairwise comparison (Figure 2G). Differentially abundant proteins from the 32-week pairwise comparison were subjected to gene ontology term enrichment analysis. Figure 2H highlights the top-10 most regulated gene ontology term biological pathways, 7 of which are directly linked to the regulation of lipid or phospholipid metabolism, and 3 to changes in respiratory electron chain transport, corroborating a progressive phenotype of dilated cardiomyopathy before echocardiographic changes.³³ Additional information on gene ontology term enrichment and gene-concept-networks from the proteomic readout can be found in Figures S7 and S8. To characterize cardiomyocyte-specific proteomic changes in the dysferlin-KO, we additionally studied isolated VMs from each n=5 WT versus KO hearts at 16 weeks of age (Figure S9). Altogether, DIA-MS quantified 3178 proteins across all samples. The principal component analysis of the DIA-MS readout and the unsupervised hierarchical clustering of the differentially abundant proteins segregated WT from KO samples, thereby confirming a myocyte-specific dysferlin-KO proteotype (Figure S9A and S9B). The pairwise comparison of KO and WT VMs revealed 412 upregulated versus 957 downregulated proteins (Figure S9C), while the top 3 enriched gene ontology terms for biological pathways highlighted proteomic changes in Golgi vesicle transport, protein folding and establishment of protein localization to organelle, indicative of a disturbed vesicular transport in KO left-VMs (Figure S9D and S9E). Taken together, dysferlin-KO mice show a progressive phenotype of dilated cardiomyopathy based on and reflected by proteomic changes occurring before full expression of dilated cardiomyopathy in ventricular tissue and isolated VMs.

STED nanoscopy of coimmunostained VMs and immunofluorescence intensity line profiling underlined the close spatial relationship of dysferlin and RyR2 clusters (Figure 3A). Accordingly, co-immunostaining with JP2, a RyR2 and LTCC Ca_V1.2 binding protein,^16,22,34 confirmed dysferlin signals in close proximity to JP2 clusters (Figure 3B). Indeed, image segmentation and detailed cluster signal analysis established similar characteristics for dysferlin/JP2 compared with dysferlin/RyR2 co-immunostainings with regard to cluster size and density (Figure S10, compare with Figure 1F and 1G). Hence, we wondered if dysferlin may interact with RyR2 and JP2 in the heart, although quantitative analysis of immunoblots from ventricular WT versus dysferlin-KO tissue lysates did not reveal changes in protein expression of RyR2, JP2 or caveolin-3 (Figures S11 and S12).

First, to study proteins in their functional high molecular weight complexes, we employed mass spectrometry-based complexome profiling, a global interaction analysis approach based on nondenaturating blue native gel electrophoresis separation followed by LC-MS/MS analysis (Figure S13A).²⁸ Interestingly, the blue native gel electrophoresis migration pattern of dysferlin presented 3 local maxima: the first and highest maximum presumably indicates dysferlin monomers/dimers in smaller protein complexes, whereas the third maximum at ≈5 MDa represents the functional high molecular weight protein complex including dysferlin and RyR2 (Figure 3C). Interestingly, the JP2 migration profile showed a total of seven local maxima, 3 of which overlapped clearly with the 3 dysferlin maxima (Figure 3D). In contrast, the caveolin-3 pattern did not show any high molecular weight maximum overlapping with dysferlin and neither did we detect protein subunits of the LTCC Ca_V1.2 (CACNA1C, CACNB2, and CACNA2D1) in the high molecular weight range (Figure S13B). However, lower molecular weight maxima for caveolin-3 and CACNA1C overlapped with the low molecular weight complex for dysferlin.

Second, to further explore the proteomic interactome of dysferlin in the heart, we used recombinant expression of N-terminally Twin-Strep II-tagged full-length human dysferlin to establish an in vitro binding assay. Specifically, isolated mouse heart membrane fractions were incubated with dysferlin-coupled beads under 4 different detergent conditions. Dysferlin pull-down eluates were analyzed by DIA-MS, revealing altogether 412 potential protein interaction partners (disregarding dysferlin) as highlighted in Figure 3E, among which JP2 was identified. To provide a subcellular synopsis, the percentage of dysferlin interactors was mapped on VM organelles and compartments in a cartoon presentation (Figure 3F). Top dysferlin interactor protein functions included the endosomal and secretory pathway as well as (tubular) membrane repair and biogenesis (Figure 3G; Supplemental Data S2).

Finally, as dysferlin is thought to function as a Ca²⁺-binding protein,³⁵ activated upon local plasma membrane rupture, we wondered if the above described protein interactions depend on increased Ca²⁺ concentrations. Coimmunoprecipitation experiments of ventricular tissue membrane fractions in absence and presence of Ca²⁺ (1 mmol/L) showed that the protein-protein interactions of dysferlin with RyR2 and JP2 are independent of the Ca²⁺ concentration (Figure 3H; Figure S14), including dysferlin-KO and IgG controls to rule out unspecific protein binding. In summary, 3 independent biochemical approaches demonstrated protein-protein interactions of dysferlin with the cardiac SR proteins RyR2 and JP2, which may be crucial to stabilize the functionally important junctional membrane complexes during VM contractions and increased biomechanical stress.

Previous studies have shown that JP2 overexpression induces the formation of complex polyadic tubule-SR junctions in VMs and AMs.^31,36 Hence, in light of the interaction of dysferlin with proteins of the cardiac dyad, we sought to study dysferlin clustering in VMs isolated from cardiomyocyte-restricted transgenic JP2 overexpressing mice.²¹ Live-cell membrane STED imaging using Chol-PEG-KK114 confirmed stacked tubular superstructures (Figure 4A) and parallelized ATs stacked on top of each other in JP2 overexpressing VMs (Figure S15A and S15B), providing a profoundly increased tubular membrane surface for gain-of-function polyadic tubule-SR junctions as demonstrated by ET and Ca²⁺ imaging previously.³¹ Stacked tubular superstructures in JP2 overexpressing VMs showed an average diameter of 600 nm, which is approximately thrice as large as TTs and ATs in WT and JP2 overexpressing VMs (Figure 4B).

Coimmunostaining confirmed the stacked tubular membrane structures as caveolin-3-positive membrane signals (Figure 4C; Figure S15C). Most interestingly, tubular superstructures were accompanied by a markedly increased local density of dysferlin clusters (Figure 4C, red). To quantify the changes of dysferlin expression in JP2 overexpression, cluster analysis was performed based on segmented STED images. Although the histogram distribution showed no obvious changes in the dysferlin cluster size in WT versus JP2 overexpressing VMs (Figure 4D left; Figure S16A), a mild but significant increase of the average dysferlin cluster size was documented (Figure 4D, middle). This increase in dysferlin cluster size may have resulted from overlapping but distinct dysferlin signals, which could be observed in context of an increased dysferlin cluster density (Figure 4D, right) and area fraction in JP2 overexpressing VMs (Figure S16B). Although the dysferlin cluster density grew, the nearest distance of dysferlin to adjacent RyR2 clusters remained unchanged, indicating a specific localization of dysferlin at Ca²⁺ release units (Figure S16C).

Coimmunostaining of dysferlin and JP2 confirmed large JP2 clusters in JP2 overexpressing VMs in vicinity of numerous dysferlin clusters (Figure 4E). Of note, the average JP2 cluster size and area fraction was increased by 27% in JP2 overexpression, whereas the JP2 cluster density did not change significantly (Figure 4F; Figure S16D and S16E). In addition, JP2 overexpression exhibited apparently larger RyR2 clusters compared with WT VMs, which were surrounded by a high local density of dysferlin clusters (Figure 4G). This was confirmed by the histogram distribution of the RyR2 cluster size suggesting a small shift toward larger RyR2 clusters in JP2 overexpression (Figure 4H; Figure S16F and S16G). In summary, polyadic tubule-SR junctions induced by JP2 overexpression resulted in an increased local density of dysferlin clusters in VMs.

Dysferlin was previously shown to enable tubular membrane proliferation in heterologous cell systems.¹⁸ Hence, we hypothesized that dysferlin may impact cellular hypertrophic remodeling induced by LV pressure overload through the biogenesis of TAT network components. Therefore, we applied TAC (Figure 5A) in 9- to 15-week-old WT and dysferlin-KO mice to assess potential differences in LV hypertrophy 4 weeks post-surgery (4pTAC; Figure 5B). The TAC-induced transaortic pressure gradients were equally elevated in the first 3 days post-surgery in WT and KO animals (Figure S17A). Unexpectedly, despite a mild tendency toward dilated cardiomyopathy per se (Figure 2), dysferlin-KO mice did not show a significant higher mortality 4pTAC (Figure 5C). However, the heart weights of dysferlin-KO mice were considerably lower compared with WT mice 4 weeks post-TAC (−25%; Figure 5D), while the body weight was unchanged (Figure S17B). Investigator-blinded echocardiographic measurements revealed that the LV end-diastolic diameter had not changed, while the LV anterior and posterior wall thickness, both indicators of hypertrophy, increased to a lesser extent in KO versus WT hearts 4pTAC (Figure 5E; Figure S17C; Table S5). To understand the reduced LV hypertrophy development in KO hearts, we studied isolated left-VMs. Dysferlin-KO left-VMs were significantly smaller in regard to cell area and width 4pTAC compared with WT cells (Figure 5F; Figure S17D). These results were further underlined by 3D volumetry of isolated left-VMs, confirming a decrease in volume of −29% in dysferlin-KO compared with WT cells 4 weeks post-TAC (Figure 5G; Figure S17E). Hence, we conclude that dysferlin is necessary for the hypertrophic growth of left-VMs following TAC.

Interestingly, immunoblots of LV tissue lysates identified an elevation of dysferlin expression up to 143% 1pTAC and 192% 4pTAC, while JP2 expression was slightly reduced (−24%, Figures S12 and S17F; Figure 5H), further highlighting the role of dysferlin in post-TAC cardiac hypertrophy. To differentiate between dysferlin expression on the cell surface versus intracellular membrane vesicles, we employed VM surface protein crosslinking using the membrane-impermeable crosslinker BS3 (Figure S17G). BS3 protein crosslinking requires extracellular lysines, of which surface-expressed dysferlin possesses 2 in the extracellular domain. Under baseline conditions, 30-minute BS3 treatment of VMs resulted in at least 4 additional crosslinked dysferlin bands as shown by immunoblotting (Figure 5I). In the crosslinking experiments, the Na⁺-K⁺-ATPase (NaK) served as positive and the intracellular membrane protein caveolin-3 as negative control. Quantitative analysis of the BS3 crosslinking revealed 22% of dysferlin versus 84% of NaK protein crosslinking as normalized to the total protein (Figure 5I). In VMs 4pTAC, crosslinked dysferlin products were significantly elevated up to 47% (Figure 5J), implying an active role of dysferlin for membrane stabilization and remodeling 4pTAC.

To assess early versus late cardiomyocyte-specific proteomic changes in LV hypertrophy, we isolated left-VMs from WT hearts 1 and 4 weeks post-TAC for DIA-MS analysis (Figure S18). Based on 1814 proteins reproducibly quantified across all samples, 515 proteins showed significant abundance changes between the sham- versus TAC-operated experimental groups. The principal component analysis of the DIA-MS readout and the unsupervised hierarchical clustering of differentially abundant proteins clearly segregated left-VM proteotypes depending on intervention and time points (Figure S18A and S18B). Most interestingly, the 1pTAC proteotype showed the greatest Euclidean distance from sham controls based on proteomic differences observed in early left-VM remodeling post-TAC. More specifically, the pairwise comparison of 1pTAC and 1pSham samples identified 487 upregulated versus 96 downregulated proteins (Figure S18C), of which dysferlin was confirmed as a significantly upregulated protein 1pTAC in line with our immunoblot quantifications. In the 4pTAC versus sham comparison we detected 314 up- versus 83 downregulated proteins, highlighting persistent proteomic differences in late left-VM hypertrophic remodeling (Figure S18D). Finally, Figure S18E directly compares 1pTAC and 4pTAC samples to point out 274 versus 45 proteins in early and late left-VM remodeling, respectively. The top-10 most regulated proteins in the pairwise comparisons are listed in Figure S18F. In addition, differentially abundant proteins from individual group comparisons were subjected to gene ontology term enrichment analysis. Notably, most of the top-10 overrepresented gene ontology terms molecular function in the pairwise comparisons can be linked to regulation of protein translation, protein folding, and actin binding (Figure S19). Figure S20 highlights the most significantly enriched gene ontology terms related to cardiac development and hypertrophy, visualizing an increase in the myosin-7/myosin-6 ratio among numerous other changes on the protein level in isolated left-VMs post-TAC.

Additionally, we analyzed the transcriptome of WT versus dysferlin-KO LV tissue from mouse hearts 4 weeks after sham versus TAC surgery (Figure S21; Supplemental Data S1). While we confirmed genotype-specific differences in all group comparisons, most of the transcriptomic changes were identified in WT sham versus TAC samples (1476 significantly regulated genes, Figure S21D). Some of the differentially regulated genes in the WT sham versus TAC comparison can be linked to the enriched gene ontology terms cardiac cell/muscle development (Figure S21D and S21F). In contrast, none of these genes, nor previously established transcription factors involved in cardiac development or reactivation of the fetal gene program during hypertrophy (like Gata4, Mef2a, Nkx2-5, Tbx, etc)^37,38 were regulated in WT versus KO TAC samples (Figure S21E). Likewise, no significant changes in the myosin heavy chain isoforms Myh6 and Myh7 or the natriuretic peptides Nppa and Nppb were detected in WT versus KO TAC samples on the mRNA level (Figure S21E). Overall, we did not observe any significantly enriched gene ontology terms for biological pathways in the WT versus KO TAC comparison. Altogether, the attenuated hypertrophic remodeling observed in the dysferlin-KO mice was not associated with prominent changes in cardiac gene regulation compared with WT controls post-TAC.

To study the consequences of VM hypertrophy on the TAT network architecture, we applied live-cell membrane STED imaging on isolated left-VMs post-TAC, uncovering dramatic changes in the network composition after signal skeletonization (Figure 6A). Most prominently, the TAT network density increased 1 and 4pTAC (Figure 6B), mainly driven by the proliferation of AT membranes along the longitudinal direction of the cell axis (Figure 6C; Figure S22A). Consequently, the AT density 1 and 4pTAC was increased (Figure 6D), contributing to an elevated number of network junctions (Figure S22B). In sharp contrast, left-VMs from dysferlin-KO hearts lacked a comparable proliferation of AT membranes 4pTAC, while the TT components were preserved (Figure 6E and 6G).

ET of isolated left-VMs did not only confirm the proliferation of AT components in cross-sectional views but also identified a markedly increased AT cross-sectional area 4pTAC (Figure 6H and 6I), together with enlarged tubule-SR contact sites as visualized by 3D reconstruction of an AT and its surrounding SR in Figure 6H. Additional 3D ET reconstructions of longitudinal left-VM sections demonstrated changes in tubule geometry with membrane extensions and protrusions not seen in sham left-VMs (Figure 6J). These tubular extensions may indicate active membrane remodeling and branching mechanisms necessary for the TAT network proliferation. Next to the TAT system, ET allowed the detection of membrane vesicles 4pTAC in a high local density that could not be found in sham controls and may additionally contribute to TAT component proliferation (Figure 6J; Figure S23). In conclusion, AT membrane proliferation together with AT dilation seem to be hallmarks of subcellular membrane remodeling in left-VM hypertrophy induced by pressure overload but are significantly attenuated in the dysferlin-KO heart.

To further explore the contribution of dysferlin to post-TAC tubule proliferation, we studied co-immunostainings of dysferlin and caveolin-3 in left-VMs 4pTAC. The confocal overviews in Figure 7A highlight the increased expression of dysferlin as visualized by the red signal intensity in hypertrophied left-VMs 4pTAC, which was particularly prominent in the center of the cells between the 2 nuclei. As dysferlin clusters are associated tightly with TAT network structures, confocal images were used for skeletonization before network analysis to identify the contribution of dysferlin to specific TAT network changes (Figure 7A). Of note, the dysferlin skeleton density increased by 28% in favor of axial network components (Figure 7B). Consequently, the frequency of axial skeleton components was elevated in left-VMs 4pTAC (Figure 7B), consistent with a role of dysferlin in AT membrane proliferation (Figure 6).

Parallel analysis of the caveolin-3 signal showed an unchanged caveolin-3 TAT network density 4pTAC, and a minor change toward more axial network components but markedly reduced TT elements (Figure 7C). These differences in dysferlin versus caveolin-3 relative to the TAT network composition were substantiated by frequency subtraction histograms (Figure 7D), showing a pronounced axial component proliferation in the dysferlin versus prominent transverse component reduction in the caveolin-3 network signal. In line with a tentatively reduced caveolin-3 protein expression 4pTAC facing an elevated TAT network density (Figure 5H), caveolin-3 expression on tubular membranes was generally limited in left-VMs 4pTAC but mainly reduced on TT elements.

STED nanoscopy of dysferlin and caveolin-3 coimmunostained left-VMs confirmed the presence of dysferlin clusters at regular intervals at the TAT network in sham cells (Figure 7E). In left-VMs 4pTAC, however, more numerous dysferlin clusters highly decorated newly shaped AT membranes in string-of-pearls arrangements (Figure 7E, right). In light of an increased dysferlin surface crosslinking 4pTAC, these data suggest a direct role for dysferlin in the proliferation of AT membranes during VM hypertrophy.

LV biopsies from human hearts were used to extend our imaging data to a highly prevalent human disease context.³² In healthy nonfailing LV sections fluorescently stained for dysferlin and the continuous membrane marker wheat germ agglutinin,²⁴ punctate dysferlin signals were localized in close association to TT membranes of left-VMs, further highlighted by the line profiles of the wheat germ agglutinin and dysferlin signals (Figure 8A). Dysferlin signals were also found in high local density at ICD membrane folds (Figure 8B), thereby fully validating our mouse VM imaging data in human LV biopsy samples.

To study dysferlin in human LV pressure overload in situ, we obtained LV biopsies from patients with 2 different hemodynamic subtypes of severe aortic valve stenosis, who underwent transfemoral aortic valve replacement: classic normal ejection fraction and high gradient aortic stenosis with compensated LV hypertrophy versus low ejection fraction and low gradient aortic stenosis representing decompensated LV hypertrophy that progressed into systolic heart failure (Figure 8C).³² Please refer to Supplemental Material for detailed patient information. While confocal overviews of the dysferlin and wheat germ agglutinin signals in coimmunostained biopsy sections confirmed substantially hypertrophic left-VMs in both aortic stenosis subtypes in contrast to nonfailing myocardium, TAT network composition and dysferlin clustering showed clear differences. In normal ejection fraction and high gradient aortic stenosis, left-VMs contained AT membrane structures clearly associated with many dysferlin clusters, as highlighted by STED magnifications (Figure 8C, middle). Left-VMs from low ejection fraction and low gradient aortic stenosis biopsies, in contrast, hardly contained any AT structures and only rare residual TT elements (Figure 8C, right). Interestingly, dysferlin clusters in low ejection fraction and low gradient aortic stenosis mainly decorated the lateral surface membrane and the ICD membrane folds but could rarely be found on TAT structures (Figure 8C, STED magnification). Thus, dysferlin clusters are located at different subcellular membrane compartments in left-VMs of distinct aortic stenosis subtypes and may be regulated differentially in the context of compensated versus decompensated LV hypertrophy and acute murine versus chronic human LV pressure overload.

Supplemental Methods

Figures S1–S23

Tables S1–S5

Data S1 and S2

Supplemental Table: Statistical details

Major Resources Table

References 45–62