Desmoplakin Mutations in Cardiac Fibroblasts Cause TGFβ1-Mediated Pathological Fibrogenesis in Desmoplakin Cardiomyopathy Via Beclin-1 Regulation

Abstract

BACKGROUND:

Pathological fibrosis is a major finding in cardiovascular diseases and can result in arrhythmia and heart failure. Desmosome gene mutations can lead to arrhythmogenic cardiomyopathy. Among arrhythmogenic cardiomyopathies, pathogenic DSP (desmoplakin) variants cause a distinctive cardiomyopathy with excessive cardiac fibrosis that could precede ventricular dysfunction. DSP variants are also linked to other fibrotic diseases. Whether DSP plays any role in pathological fibrosis remains unknown.

METHODS:

Mesenchymal stromal cells (MSCs) are resident fibroblast-like cells that are responsible for fibrogenesis in most organs, including the heart. We first used RNA-seq genome-wide analyses to generate cardiac fibroblast-like, induced pluripotent stem cell-derived MSCs from normal donors and patients with arrhythmogenic cardiomyopathy and DSP mutations. We then studied the fibrogenic responses of cardiac MSCs to TGFβ1 (transforming growth factor β1) using Western/Co-IP, autophagy assays, gene knockdowns/over-expressions, genomic analyses, mouse DSP knockdown models, immunostaining, and qPCR.

RESULTS:

TGFβ1 induced excessive accumulation of VIM (vimentin)/fibrillar collagens and over-activated fibrotic genes in DSP-mutant MSCs when compared with normal MSCs. In normal MSCs, VIMs bind to wild-type DSP during normal fibrogenesis after TGFβ1. DSP-mutant MSCs exhibited a haplo-insufficient phenotype with increased DSP-unbound VIMs that sequestered BECN1 (beclin-1) from activating autophagy and CAV1 (caveolin-1)-mediated endocytosis. Decreased autophagy caused collagen accumulation, and diminished CAV1 endocytosis resulted in abnormal CAV1 plaque formation that over-activated fibrotic genes (COL1A1, COL3A1, and fibronectin [FN]) via heightened p38 activity after TGFβ1. Genome-wide analysis and DSP knockdown in mouse fibroblasts confirmed this novel role of DSP mutations in pathological fibrosis. Overexpression of VIM-binding domains of DSP could suppress pathological fibrosis by increasing collagen autophagic degradation and decreasing fibrotic gene expression.

CONCLUSIONS:

Our data reveal that DSP deficiency in MSCs/fibroblasts leads to exaggerated fibrogenesis in DSP-cardiomyopathy by decreasing BECN1 availability for autophagy and CAV1-endocytosis. Overexpression of VIM binding domains of DSP could be a new strategy to treat pathological fibrosis.

Novelty and Significance.

What Is Known?

• Desmoplakin variants and deficiency are linked to several pathological fibrogenic diseases, including desmoplakin cardiomyopathies.

• Cardiac fibroblasts are known sources of cardiac fibrosis, but cardiac fibroblasts are a heterogeneous group of cells and lack internationally recognized markers for their identification.

What New Information Does This Article Contribute?

• We used genome-wide analyses, rather than selective markers, to generate cardiac fibroblast-like, induced pluripotent stem cell-derived mesenchymal stromal cells to study pathological fibrosis in desmoplakin cardiomyopathy.

• We show for the first time that both desmoplakin and vimentin play active roles in mediating pathological fibrosis by regulating beclin-1 availability.

• Desmoplakin deficiency decreases collagen autophagic degradation and causes abnormal formation of long CAV1 (caveolin-1) plaques that, in turn, overactivate the p38 pathway and profibrogenic genes after TGFβ1 (transforming growth factor β1) in vitro and in vivo.

• A truncated desmoplakin could increase collagen degradation and decrease profibrotic gene activation, which might be used to treat other pathological fibrosis.

Desmosome gene mutations are linked to several fibrotic diseases, including arrhythmogenic cardiomyopathy. A better understanding of how DSP mutations affect pathological fibrosis might lead to novel therapies. Cardiac fibroblasts were usually identified via selective markers without international consensus. We first use genome-wide expression analyses to generate human cardiac fibroblast-like induced pluripotent stem cell-derived mesenchymal stromal cells, which can provide unlimited supplies for pathogenic and therapeutic studies. We then show that DSP-mutant cardiac mesenchymal stromal cells display exaggerated fibrogenic responses to TGFβ1 with excessive accumulations of VIM/fibrillar collagens and over-activated fibrotic genes. We further find that DSP competes with BECN1 (beclin-1) for their associations with VIMs. Pathogenic DSP variants lead to DSP deficiency and increased VIM-BECN1 binding, which sequesters BECN1 from activating autophagy and from inducing CAV1-mediated endocytosis. Decreased autophagy causes collagen accumulation, and diminished CAV1 endocytosis results in abnormal CAV1 plaque formation that over-activates fibrogenic genes via heightened p38 activity after TGFβ1. A truncated VIM-binding DSP protein could free up BECN1 to increase collagen autophagic degradation and downregulate fibrotic gene expression. This truncated DSP transcript might offer a new therapeutic strategy to treat other pathological organ fibrosis.

Arrhythmogenic cardiomyopathy (ACM) is an inherited cardiomyopathy frequently caused by pathogenic variants in 5 desmosomal genes, including plakophilin-2 (PKP2), desmoplakin (DSP), desmoglein 2, desmocollin 2 (DSC2), and plakoglobin. Progressive fibro-fatty replacement of cardiomyocytes with increased cardiomyocyte apoptosis is the pathological hallmark of ACM, which may cause sudden death in the young.^1–4 Among desmosomal mutations, human ACM hearts with DSP mutations are different from other types of desmosomal mutations by displaying more pathological fibrosis than fatty infiltration (adipogenesis) and affecting both ventricles.^5,6 DSP knockout or decreased DSP expression has also been implicated in the pathogenesis of fibrotic diseases, such as idiopathic pulmonary fibrosis⁷ and hypertrophic scars (a form of pathological fibrosis).⁸ Whether and how DSP mutations could cause or mediate pathological fibrosis remains to be determined.

Pathological cardiac fibrosis is usually characterized by excessive production and accumulation of cytosolic VIM (vimentin) and extracellular matrix (ECM, particularly fibrillar collagens) in diseased hearts^9–12 and may lead to lethal ventricular arrhythmia and heart failure. Resident cardiac fibroblasts are generally considered the principal source of cardiac fibrosis.¹³ However, cardiac fibroblasts display heterogeneity¹⁴ and lack commonly accepted markers for identification.¹⁵ In contrast, mesenchymal stromal/stem cells (MSCs), morphologically indistinguishable from fibroblasts, have internationally defined characteristics,¹⁶ can be isolated reproducibly from various organs, and are known adult resident progenitor cells for myofibroblasts (activated fibroblasts)¹³ that produce fibrosis in most organs,¹⁷ including the heart. However, obtaining cardiac MSCs or fibroblasts from cardiac patients is rarely performed. Recent advances in reprogramming somatic cells into induced pluripotent stem cells (iPSCs) have enabled the derivation of MSCs from iPSCs (iPSC-MSCs)^18,19; this process exhibits high purity/yield of MSCs and does not require any invasive procedure. We primarily used an iPSC-MSC derivation protocol¹⁹ that drove pluripotent iPSCs through an early mesoderm stage before the final derivation of MSCs, simulating cardiac development in vivo.

We used TGFβ1 (transforming growth factor beta-1, a well-known fibrosis activator,²⁰ to induce fibrogenic responses in cardiac MSCs because TGFβ1 levels were particularly elevated in ACM mouse hearts²¹ and in human ACM patients.²² In general, fibrogenic responses to TGFβ1 in MSCs could be reparative or maladaptive. In normal reparative responses, TGFβ1 induces transient activation of fibrogenic genes and proteins (eg, collagens) after injury for physiological repair,²³ which processes are controlled and later terminated by TGFβ1-induced endocytosis-lysosome (which terminates TGFβ1 signaling)^23,24 and autophagy (which degrades redundant collagens)²⁵ pathways. In contrast, the balance between activation and termination of the TGFβ1 pathway appears lost in maladaptive responses to TGFβ1, resulting in prolonged and excessive production of ECM that contributes to pathological organ fibrosis.^10,22–26

Here, we used RNA sequencing to show that our iPSC-MSCs are similar to cardiac fibroblasts grown directly from the source human heart by genome-wide profiling. This is the first time that cardiac fibroblast-like iPSC-MSCs are characterized by genome-wide comparisons with fibroblasts directly grown from the same source donor heart rather than using a few highly selective markers. We then generated cardiac iPSC-MSCs from healthy donors and 2 unrelated patients with ACM with pathogenic heterozygous DSP variants to study how DSP mutations affect their fibrogenic potential. We found that DSP-mutant MSCs had no mutant but ≈50% wild-type (WT) DSP protein (a DSP-deficient and haploinsufficiency phenotype). Importantly, TGFβ1 induced excessive accumulations of VIMs/fibrillar collagens and overactivated fibrotic genes (COL1A1, COL3A1, and fibronectin [FN]) in DSP-mutant MSCs. In normal MSCs, VIMs bound to WT DSPs during normal fibrogenesis, yet in DSP-deficient MSCs, decreased WT DSPs led to increased DSP-unbound VIMs that sequestered BECN1 (beclin-1) and decreased its availability for autophagy/lysosome degradation of collagens. Decreased BECN1 availability also reduced CAV1 (caveolin-1)-mediated endocytosis, leading to abnormal formation of long CAV1 plaques that overactivated fibrotic genes via heightened p38 activity. DSP knockdown in normal MSCs and overexpression of DSP in mutant MSCs further support that DSP mutations lead to DSP deficiency, which causes pathological fibrogenesis via deregulated BECN1-VIM interactions. We also generated mice with heterozygous deletion of DSP in activated fibroblasts, which showed aggressive cardiac fibrosis to TGFβ1 in vivo via the same pathogenic mechanisms. Furthermore, genome-wide gene expression analysis validated this novel pathogenic role of DSP deficiency in pathological fibrosis. Most importantly, overexpression of VIM-binding domains of DSP suppressed the pathological fibrosis in DSP-mutant MSCs, which could be a new strategy to treat pathological cardiac fibrosis.

Methods

Data Availability

Data supporting the findings of this study are available from the corresponding author on reasonable request. Detailed descriptions of experimental methods, materials, statistical analysis, and the major resource tables are presented in the Supplemental Materials.

Heart Tissues and Generation of Human iPSCs

The generation and characterization methods of normal iPSC (SW) and desmosomal mutant iPSC lines were published previously.²⁷ We obtained DSP mutant human peripheral blood mononuclear cells from Johns-Hopkins University and the University of California, San Diego, with approved protocols by relevant institutional review boards at each institution. All living human participants gave written informed consent. Frozen human normal ventricular tissues from a 38-year-old donor heart without known heart disease were provided by AnaBios Corp. Frozen ventricular tissues from an ACM patient with a DSP (c.4531G>T, p.Q1511X) mutation were provided by the University of Zurich under approved study protocols. Peripheral blood mononuclear cells were reprogrammed using episomal vectors containing Oct4, Sox2, c-Myc, and Klf4 to generate integration-free iPSC colonies according to published episomal methods (see Supplemental Methods for details). Human iPSCs and H9 human embryonic stem cells were maintained on a feeder-free system using Matrigel-coated culture plates and TeSR-E8 media Kits (STEMCELL Technology Inc).²⁷ Two clones of each mutant or normal iPSC line with the best proliferation capability/rates were used to generate iPSC-MSCs. All culture media and reagents were purchased from Invitrogen (Carlsbad, CA)/Thermo Fisher Scientific (Waltham, MA) unless indicated otherwise.

Generation of Human Normal and DSP-Mutant iPSC-MSCs

We used a well-established protocol to generate iPSC-MSCs via an early mesoderm intermediate stage as previously described¹⁹ to simulate cardiac development in vivo (see Supplemental Methods for details). We plated harvested iPSC-MSCs on non-coated plastic 12-well plates, and only fibroblast-like cells attached to plastic surfaces within 20 minutes after plating were selected (plastic adhesion selection), which enabled the isolation and purification of iPSC-derived fibroblasts that fulfilled published criteria for MSCs.¹⁶ iPSC-MSCs were generated freshly from each iPSC line every 25 passages to avoid cellular senescence.

Culturing Fibroblasts From Human and Mouse Tissues

A small piece (1–2 mm²) of unused and redundant skin or heart tissues from a human subject (SW), or from ventricular tissues of various mouse hearts, was minced and placed into 1 well of a 6-well noncoated plate and cultured in 3 mL MSC medium with Normocin (Invitrogen) and Gentamicin (GIBCO).²⁸ The media was changed daily for 3 days and then every 2 to 3 days. Tissue pieces and nonadherent cells were removed after 7 to 10 days (in 3–4 days for mouse tissues), when fibroblasts appeared in the well. The fibroblasts were cultured to 90% to 95% confluence (passage 0) and were then harvested with trypsin for further expansion or storage. Cardiac fibroblasts at passage ≥3 were used for the experiments when sufficient amounts of fibroblasts were obtained. Human skin fibroblasts required more passages to yield sufficient amounts of fibroblasts for research.

MSC Characterization Using Flow Cytometry

The MSCs were prepared and characterized according to the standard flow cytometry protocol at Indiana University Flow Cytometry Core (see Supplemental Methods for details). Antibodies used are listed in Table S1.

Karyotype Analysis

Karyotype analysis was performed by Cell Line Genetics (Madison, WI).

Total mRNA Isolation and Quantitative Real-Time Polymerase Chain Reaction

Total RNA from cultured cells was prepared using the mirVana miRNA Isolation Kit (Invitrogen) according to the manufacturer’s instructions. For mRNA quantitative polymerase chain reaction (qPCR), about 200 ng of RNA was then reverse transcribed to cDNA using Quantitect Reverse Transcription Kit (Qiagen) according to the manufacturer’s protocol. qPCR was performed with 400 ng cDNA for 40 cycles using SYBR Green I (Roche) and Roche LightCycler 96 according to protocols recommended by the manufacturer. Primer sequences used are listed in Table S2. Quantification for mRNA expression levels was performed by correcting for amplification efficiency of the primer using a standard curve, followed by normalizing transcript levels to the amount of total ubiquitously expressed GAPDH transcripts. Relative expression levels of mRNA were calculated using the 2^−ΔΔCt method. 2 to 3 physical replicates of each sample from ≥3 independent sets of experiments (conducted on different dates) were performed. To construct each histogram, mRNA levels were normalized to the baseline mRNA levels of normal MSCs (as 1 or 100%) in all figures unless indicated otherwise.

Genomic and cDNA Sequence Analysis

Genomic DNA at 200 ng per PCR reaction from undifferentiated iPSCs was collected using QIAamp DNA Mini Kit (Qiagen, catalog no. 51304), and a PCR reaction was set up using DreamTag Green PCR Master Mix (Thermo Scientific, catalog no. K1081) following the manufacturer’s instructions. PCR products were prepared and sent to either GenScript (Piscataway, NJ) or Genewiz (South Plainfield, NJ) for DNA sequencing services according to each company’s standard requirements. Sequences of primers used are listed in Table S3.

Immunoblotting (Western Blotting), Immunoprecipitation, and Co-Immunoprecipitation

Standard techniques were used for Western blots, immunoprecipitation, and co-immunoprecipitation (see Supplemental Methods for details). Antibodies used with indicated dilutions in 5% BSA are listed in Table S4. DSP antibody (Abcam no. ab71690, stains both DSP-I [≈332 kD] and DSP-II [≈260 kD] isoforms)^29,30 was used for all Western blots in this study except the sub-Figure S2E.

Sircol Collagen Assay

To examine collagen released into the cell culture media, normal and mutant MSCs were grown in 12-well plates with or without 1 ng/mL TGFβ1 for 72 hours in the absence or presence of specific drugs. The supernatant was then collected, and collagen content was measured using the Sircol Assay according to the manufacturer’s instructions (Biocolor; catalog no. S1000).

Immunocytochemistry on Cells and Human Heart Tissues

Standard techniques were used for immunocytochemistry on cultured cells and cryo-preserved human heart tissues (see Supplemental Methods for details). Primary and secondary antibodies used are listed in Table S5. Alexa Fluor 488 and Alexa Fluor 594 anti-mouse or anti-rabbit IgG (Invitrogen) were the secondary antibodies used for fluorescence imaging. Samples were imaged using confocal microscopy (Leica confocal). CAV1 plaque lengths, cell sizes, and CAV1-TGFβ1 receptor overlap lengths were measured using ImageJ software.

Imaging Autophagy With Premo Autophagy Tandem Sensor RFP-GFP-LC3B

The MSCs (2×10⁵ cells/well) were cultured in a Nunc Lab-Tek 8 Chamber Slide System and allowed to adhere overnight. Various stages of autophagy were monitored using the Premo Autophagy Tandem Sensor RFP (red fluorescent protein)-GFP (green fluorescent protein) LC3B Kit (Life Technologies) according to the manufacturer’s instructions (see Supplemental Methods for details).

Plasmids and siRNAs

We used PCR and the DSP isoform-I-GFP vector (Addgene plasmid no. 32227) to generate the hemagglutinin (HA)-DSPΔN fragment (corresponding to amino acids 1057–2871) and subcloned this fragment into the pcDNA3.1 vector so as to overexpress hemagglutinin DSPΔN in mutant MSCs (see Supplemental Methods for details). Wild-type full-length beclin-1 cDNA (pcDNA4-beclin-1, Addgene plasmid no. 24388) was used for plasmid transfection and was cloned into the pCDH-CMV-MCS-EF1-GFP lentiviral vector (System Biosciences Inc). The ON-TARGET siRNA-SMART pool L-003551-00-0005 or L-003467-00-0005 was used for VIM and CAV1 knockdown (Horizon Discovery), respectively. A pool of 4 siRNA duplexes against DSP was used (Horizon Discovery) to knock down DSP. We used AllStars Negative Control siRNA (Qiagen) as the control nonsilencing siRNA (siCTR).

Transfection of Plasmids

Plasmid cDNAs of wild-type BECN1, BECN1-GFP, HA-DSPΔN, or empty vector were transfected using Lipofectamine 2000 in accordance with the manufacturer’s instructions.

RNA Interference/Transfection of Synthetic siRNAs

MSCs were cultured in antibiotic-free MSC medium to 70% to 80% confluency, and then synthetic siRNA targeting human VIM, CAV1, or DSP, and AllStars Negative Control siRNA at 50 nmol/L were transfected into MSCs using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific) following the manufacturer’s instructions. At 48 hours after the first transfection, a second transfection of siRNA targeting each gene at 50 nmol/L was performed to obtain the best knockdown efficiency. After incubation for an additional 48 hours, total protein was extracted to confirm the decreased protein levels of each target gene by Western blots.

Electroporation of MSCs

To evaluate the effects of BECN1 or HA-DSPΔN overexpression on mRNA and protein levels, electroporation was performed using the Neon Transfection System 100 µL Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Electroporation was performed using parameters to obtain the best cell survival and transfection efficiency: voltage 990 to 1200 V; interval 20 to 40 ms; 1 to 2 pulses. After electroporation, the cell suspension was transferred and cultured in antibiotic-free MSC media.

Generation and Study of Mice With Heterozygous DSP Deletion in Myofibroblasts

In-house mice with tamoxifen-inducible, periostin-specific Cre recombinase (POSTN^MRM-Cre+) genotypes³¹ were intercrossed with mice with heterozygous DSP-floxed mice (DSP^w/f, Jax #026665, The Jackson Lab). Mice with POSTN^MCM-Cre+; DSP^w/f and POSTN^MCM-Cre-; DSP^w/f genotypes (from both male and female mice) were selected. The ethical use of mice for research was approved by Indiana University with protocol no. 22118. We injected tamoxifen intraperitoneally for 5 days starting at postnatal 3 weeks of age to all mice, but tamoxifen would only induce myofibroblast-specific heterozygous deletion of DSP in POSTN^MCM-Cre+ mice. We then injected 1 µg TGFβ1 daily intraperitoneally for 7 to 8 days into all 6-week-old mice, and mouse hearts were harvested for research after completing TGFβ1 injection. Sirius Red-fast green staining for collagens was performed as described previously²¹ (see Supplemental Methods for details).

Illumina Microarray Data and RNA Sequencing

Illumina microarray data and RNA sequencing data were performed at the Microarray Core facility at Sanford-Burnham Prebys Medical Discovery Institute (SBP) and the Genomic Core facility at Indiana University, respectively. The standard analyses for microarray²⁷ and RNA-seq data were performed by the Core facility at each institution (see Supplemental Methods for details).

Statistical Analysis

Data were presented as the mean±SD. All experiments were performed with ≥3 independent experiments conducted on different dates (termed biological replicates here), and N represents the number of biological replicates except for cell imaging analysis. For cell imaging analysis, the total number of counted cells from 3 to 5 biological replicates was shown unless indicated otherwise. Using the StatView/JMP program (SAS Institute, Cary, NC) or Origin (OriginLab Corp), statistical differences were analyzed by an ANOVA procedure with a post hoc Tukey/Kramer test for multiple comparisons (labeled as ANOVA). A difference with P<0.05 was considered statistically significant in all figures. The precise P values were obtained by the unpaired Student t (with Welch corrections) in Origin software, or nonparametric (Mann-Whitney U test or Kruskal-Wallis) tests in StatView/JMP software according to the data distribution for pairwise comparisons. The normality of each data set was evaluated by the Shapiro-Wilk test in Origin. All data with a sample size <4 were analyzed by a nonparametric test.

Results

Characterization of iPSCs From ACM Patients With DSP Mutations

DSP is a cytoskeletal linker protein with 3 distinct domains (Figure 1A): (1) the N-terminal plakin domain containing 6 spectrin-like repeats that mediate desmosome formation; (2) the middle coiled-coil rod domain required for DSP homodimerization; and (3) the C-terminal region containing 3 plakin repeat domains that interact with intermediate filaments such as VIM in MSCs.^29,30 We first generated patient-specific iPSCs from peripheral blood mononuclear cells of 2 unrelated ACM patients with a heterozygous nonsense mutation (DSP c.478 C>T [p. Arg160X]) or a heterozygous insertion of an adenine nucleotide that results in a frame-shifted C-terminus with premature stop codons (DSP c.3474_3475insA [p. Glu1159ArgfsX3]). We used the genome nonintegrating episomal vectors²⁷ containing Oct4, Sox2, Klf4, and cMyc to generate 2 groups of DSP-mutant iPSC lines, and termed here R160X and E1159R lines, respectively, for brevity. R160X and E1159R lines are color-coded in blue and green, respectively, in all figures to facilitate the readability of this article. Characterization of these iPSCs (Figure S1) shows that all iPSC lines (1) express pluripotent markers including TRA-1-60, TRA-1-81, Nanog, Oct4, and SSEA4 by immunohistochemistry, (2) can differentiate into cells of 3 germ layers (including ectoderm [Tuj1+], mesoderm [GATA4+], and endoderm [Sox17+]), (3) contain corresponding DSP mutations (by genomic DNA sequencing), (4) have normal karyotypes, and (5) exhibit no integration of episomal vectors (verified by qPCR for the exogenous episomal EBNA-1 and endogenous FBXO15 expression ratios). These results demonstrate that we have successfully generated 2 sets of integration-free iPSC lines with DSP R160X or E1159R truncation mutations (with 3–4 different subclones for each DSP-mutant iPSC line; Figure S1F).

Figure 1.

Characterization of induced pluripotent stem cell (iPSC)-derived mesenchymal stromal cells (MSCs). A, A simple diagram of 3 structural domains in DSP (desmoplakin). B, All MSCs express MSC phenotypes by flow cytometry. C, A clustering analysis of global transcriptome mRNA expression profiles (a heatmap) of iPSC-MSCs, cardiac FBs, and skin FBs from the same donor (SW) by RNA-seq. Increased and decreased normalized gene expression are shown in red and green, respectively. D and E, DSP-mutant MSCs show <50% wild-type (WT) DSP mRNA and DSP protein levels when compared with normal MSCs. F, All MSCs expressed similar immunostaining intensities and mRNA levels of αSMA (alpha-smooth muscle actin) at baseline and after TGFβ1 (transforming growth factor β1). The number of independent biological replicates for each data point is shown in all Figures. Data were analyzed by ANOVA and Student t except F by Kruskal-Wallis tests. Values are means±SD. Scale bars are 50 µm. Raw data are in the Source Data Files.

Characterization of MSCs Derived From Normal and DSP-Mutant iPSCs

After successfully generating 2 different DSP-mutant iPSC lines, we used an established protocol to derive MSCs from human iPSCs (iPSC-MSCs) via an early mesoderm intermediate stage as previously described¹⁹ to simulate the mesodermal origin of cardiac MSCs/fibroblasts.³¹ Because fibroblasts lack internationally recognized markers for identification^14,15 and to ensure reproducibility, we used cardiac fibroblast-like iPSC-MSCs as the cardiac fibroblast source for our study. All iPSC-MSCs displayed morphologies indistinguishable from fibroblasts and adhered to standard plastic culture surfaces (Figure S2A). Based on the MSC identification criteria published by the International Society for Cellular Therapy,¹⁶ flow cytometric analysis revealed high expression of MSC markers (CD73+=99.9%, CD90+ ≥98.9%, and CD105+ ≥92.9%), and low expression of an endothelial marker (CD144+ ≤0.91%) in all iPSC-MSCs (Figure 1B and cytometry plots in Figure S2B). A previously published normal iPSC line (SW line)²⁷ was used to derive normal iPSC-MSCs. We used the term Normal fibroblasts/MSCs throughout this study to describe fibroblasts/MSCs derived or obtained from human donors without any cardiac disease phenotype. To determine if our iPSC-derived MSCs resemble cardiac fibroblasts directly sampled from the same iPSC donor in global genomic expression profiles, we obtained fibroblasts directly grown from discarded heart tissues and skin samples of the same normal donor (SW) for comparison. Using RNA sequencing and unsupervised hierarchical clustering analysis (see Supplemental Methods), the whole genome-wide expression profiles of our iPSC-MSCs are clustered together with cardiac fibroblasts but not skin fibroblasts (at passages 7, 8, and 16; Figure 1C). Thus, instead of using a selected but limited set of markers, we verify the cardiac fibroblast-like nature of our iPSC-MSCs by their close similarity in the whole genome-wide expression profiles to those in cardiac fibroblasts directly sampled from the same donor heart. Thus, we have derived cardiac fibroblast-like MSCs from iPSCs with the advantage that iPSCs could provide limitless supplies of cardiac fibroblast-like MSCs.

Next, we measured WT and mutant DSP mRNA and DSP protein levels to functionally characterize these 2 DSP-mutant MSC lines. We used 2 qPCR primer sets targeting either N-terminal (113-217bp) or C-terminal (5202-5386bp) regions of WT DSP mRNA and found that WT DSP mRNA levels in both DSP-mutant MSC lines are <50% of those in normal (SW) iPSC-MSCs (Figure 1D). We used 2 different DSP antibodies to detect truncated mutant DSP peptides if they were theoretically translated from R160X or E1159R mutations: (1) The DSP antibody 1 to 100 recognizes the first 1 to 100 amino acids of DSP and it should detect WT DSP and shorter R160X/E1159R peptides; (2) The DSP 400 to 500 antibody recognizes the 400^th to 500th amino acids of DSP and it should detect WT DSP and the short E1159R but not R160X peptides (Figure S2C). Western blot (Western) analyses with these 2 DSP antibodies showed that neither DSP-mutant MSCs expressed truncated mutant DSP proteins (Figure S2D), but both mutant lines expressed ≈50% of WT DSP protein levels when compared with normal MSCs (Figure 1E). All MSCs expressed similar levels of αSMA (alpha-smooth muscle actin) by immunostaining or by qPCR (Figure 1F) at baseline or after TGFβ1 treatment. These results support that both pathogenic DSP variants result in no mutant DSP proteins and a ≈50% reduction of WT DSP protein levels (a common haploinsufficiency phenotype) with similar αSMA expression levels when compared with normal MSCs.

Increased Collagen and VIM Levels in DSP-Mutant MSCs by TGFβ1

Because TGFβ1 levels are elevated in hearts of an ACM mouse model²¹ and in ACM patients,²² we first tested fibrogenic responses to 1 ng/mL TGFβ1 in all iPSC-MSCs. In normal MSCs, TGFβ1 should activate fibrogenic gene/protein production within hours and initiate autophagic degradation of collagens later (at ≈24 hours) to regulate the fibrogenic response and avoid excessive collagen accumulation^23–25 (Figure 2A). In culture media, 72 hours of TGFβ1 increased collagen content by 1.52 µg/mL (between black dotted and dashed lines) in normal MSCs with the Sircol Assay, yet 72-hour TGFβ1 increased collagen content by 3.86 to 4.93 µg/mL in DSP-mutant MSCs (between red dotted and black dash lines, representing 2- to 3-fold increases when compared with the 1.52 µg/mL increase in normal MSCs, Figure 2B). 72-hour TGFβ1 also induced intense intracellular immunostaining of fibrillar collagens (Figure 2C, left images) in DSP-mutant MSCs but not in normal MSCs. This excessive accumulation of collagens induced by TGFβ1 could be suppressed by various autophagy activators,^32,33 including 10 nmol/L rapamycin (via mTOR [mammalian target of rapamycin] inhibition), 2.5 µmol/L MK-2206 (via inhibition of Akt1/2/3), and 10 µmol/L trifluoperazine (an mTOR-independent autophagy inducer; Figure 2B and 2C; Figure S3A and S3B), suggesting that the common autophagic pathway for collagen degradation might be downregulated in DSP-mutant MSCs. The increases in COL1 (collagen 1 protein) levels by Western blot were also about 2-fold higher in mutant than normal MSCs after 72-hour TGFβ1 (Figure 2D). For COL3, the levels were upregulated at baseline and after 72-hour TGFβ1 in both DSP mutant iPSC-MSCs when compared with normal MSCs. VIM protein levels (a mesenchymal marker for activated fibroblasts or myofibroblasts)³⁴ by Western blot or immunofluorescence staining were also increased by> 2-fold in DSP-mutant MSCs after 72-hour TGFβ1 when compared with normal MSCs (Figure 2E and 2F and images in Figure S3C). Collectively, these data indicate that DSP-mutant MSCs display >2-fold increases in fibrogenic responses to TGFβ1 versus normal MSCs.

Figure 2.

Exaggerated fibrogenic responses to TGFβ1 (transforming growth factor β1) in DSP (desmoplakin)-mutant induced pluripotent stem cell (iPSC)-derived mesenchymal stromal cells (MSCs). A, A controlled reparative TGFβ1 response scheme in normal MSCs. B, After 72-hour TGFβ1, the increases in collagen content in culture media were >2-fold higher in mutant MSCs than in normal MSCs. C, DSP mutant MSCs displayed intense fibrillar collagen immunostaining (white arrows) immunostaining (green) vs normal MSCs after 72-hour TGFβ1, which could be inhibited by 3 different autophagy activators. D, Western blots (normalized to normal MSCs at baseline) of fibrotic markers showed higher levels of COL1 (left) and COL3 (right) in mutant than normal MSCs. E, VIM (vimentin) protein levels by Western or (F) by immunofluorescent assays also increased by>2-fold in DSP mutant MSCs when compared with normal MSCs after 72-hour TGFβ1. Data were analyzed by ANOVA and Student t tests. Scale bars are 50 µm.

Defective Autophagy Pathway in DSP-Mutant MSCs

To further determine if the overall degree of autophagy is decreased in DSP-mutant MSCs, we first performed an autophagy fluorescent flux assay using the Premo Autophagy Tandem Sensor RFP-GFP-LC3B Kit (Figure 3A)^35–37 to screen whether the capacity to form autophagosomes and autophagolysosomes in DSP-mutant MSCs is decreased before and after TGFβ1 exposure. In this autophagy monitoring system, RFP-GFP-LC3B sensors are activated by the BECN1-containing PI3 kinase complexes and bind to phagophores to form the autophagosomes (positive for both RFP and GFP). Then, autophagosomes continue to fuse with lysosomes to form autophagolysosomes, where the acidic environment in lysosomes quenches the green GFP fluorescence and the sensor turns red only. Autophagic flux assays indicated that both autophagosome (both RFP+ and GFP+) and autophagolysosome (red dot only) counts were decreased by 40% to 50% in both DSP-mutant MSCs versus normal MSCs at baseline and after 72-hour TGFβ1 (Figure 3B and 3C), and these results are consistent with a decreased overall autophagy flux in DSP-mutant MSCs. We then tracked back along the autophagy pathway to find the earliest defective autophagy step in DSP-mutant MSCs. We first investigated the LC3 lipidation step that drove phagophores to form autophagosomes. We found that LC3B-I to -II conversion (= LC3B-II/I ratio) was also decreased by ≈50% at baseline and after exposure to TGFβ1 for 24 to 72 hours in DSP-mutant MSCs (Figure 3D), consistent with the results detected by the RFP-GFP-LC3B sensors. During autophagy initiation, BECN1 forms a protein complex with VPS34/15 (subunits of a class III PI3 kinase) and ATG14L (Autophagy Related Gene 14 L; Figure 3A, the color-coded structure at left) that drives LC3B-I to -II conversion via a series of ATG proteins to initiate phagophore elongation and transformation to the autophagosome in mammalian cells.^35–37 Thus, we examined whether BECN1 protein levels were lower in DSP-mutant MSCs. Interestingly, we found that total BECN1 protein levels were similar in normal and DSP-mutant MSCs at baseline and after exposure to TGFβ1 for 24 to 72 hours (Figure 3E). Taken together, these results narrowed down the earliest impaired autophagic step(s) in DSP-mutant MSCs to processes between BECN1-ATG14L-VPS34/15 PI3 kinase complexes and LC3B-I to -II conversion. Of note, combining all data in Figure 3B through 3D (and in Figure 4), results from the autophagy fluorescent flux assay are not due to any significant difference in cell sizes before and after TGFβ1 (also see Figure S4A).

Figure 3.

The autophagy deficit in DSP (desmoplakin)-mutant mesenchymal stromal cells (MSCs) is located at steps between BECN1 (beclin-1) and LC3B-I to –II conversions. A, A diagram of the autophagy fluorescent assay using LC3 tandem plasmids to detect autophagosomes (positive for both RFP [red fluorescent protein] and GFP [green fluorescent protein]) and autophagolysosomes (red dots only, yellow circles in B). The BECN1/ATG14L-VPS34/VPS15 complex is color-coded for each component in the model. B and C, DSP-mutant MSCs showed decreased autophagosomes and autophagolysosomes at baseline and after 72-hour TGFβ1 (transforming growth factor β1) vs normal MSCs. D, LC3B-II to -I ratios at any time point were lower in DSP-mutant MSCs, but E, total BECN1 protein levels were similar at any time point. Data were analyzed by ANOVA/Student t except D and E by Kruskal-Wallis.

Figure 4.

DSP (desmoplakin)-mutant mesenchymal stromal cells (MSCs) have less DSP-VIM (vimentin) but enhanced BECN1 (beclin-1)-VIM associations, leading to decreased BECN1 availability. A and B, Using co-IP, only ≈50% of VIMs were associated with DSP in DSP-mutant MSCs when compared with normal MSCs (see Figure S4E for 24-hour gels). C, At 24 hours after TGFβ1 (transforming growth factor β1), BECN1-VIM associations were ≈2-fold higher in mutant than in normal MSCs. D, At 24 hours after TGFβ1, BECN1-ATG14 and BECN1-UVRAG associations in DSP-mutant MSCs were ≈50% of those in normal MSCs. E, DSP KD by siDSP, but not by siCTR, in normal MSCs enhanced VIM-BECN1 binding and decreased BECN1-ATG14 and BECN1-UVRAG associations after 24-hour TGFβ1. F, VIM KD in mutant MSCs decreased VIM and COL1 levels after 72-hour TGFβ1. G, VIM KD in both DSP-mutant MSCs decreased collagen content in the media after 72-hour TGFβ1. H, VIM KD in mutant MSCs increased BECN1-ATG14 and BECN1-UVRAG associations after 24-hour TGFβ1. I, VIM KD in mutant MSCs increased LC3B-II/-I ratios after 24-hour TGFβ1. *P values by Mann-Whitney U test.

Increased DSP-Unbound VIM Restricted BECN1 Availability for Autophagy in Mutant MSCs

Next, we sought to determine whether DSP haploinsufficiency led to autophagy defects in stages between BECN1-PI3 kinase complexes and LC3B-I to -II conversion. There are 3 major known protein complexes between BECN1 and LC3B-I to -II conversion:^35–37 BECN1-PI3 kinase, a ULK1 (Unc-51 like autophagy activating kinase 1) complex, and a series of ATG protein complexes (Figure 3A). Since (1) MSCs do not form desmosomes, (2) DSP proteins are not known to interact with ATG proteins, and (3) both DSP³⁰ and BECN1³⁸ have been shown to bind VIM (the main type of intermediate filaments in MSCs), we hypothesized that DSP might compete with BECN1 for VIM binding. Using co-immunoprecipitation, we found that VIM-DSP interaction was also reduced by ≈50% at baseline and after 24- or 72-hour TGFβ1 (Figure 4A and 4B) in DSP-mutant MSCs. By VIM and DSP immunostaining overlap measurement (Figure S4B through S4D), the percentage of colocalization of VIM- and DSP-stained areas decreased, resulting in ≈2-fold increase in non-DSP-overlapped VIMs in DSP-mutant MSCs at baseline and after 24-hour TGFβ1 (Figure S4D). We then checked the VIM-BECN1 association after 24-hour TGFβ1 because total BECN1 level reached its peak at 24 hours after TGFβ1 (Figure 3E), which would provide the highest BECN1 protein levels for Western blots. With ≈50% less WT DSP and ≈2-fold increased DSP-unbound VIM, the association of BECN1 with VIM by co-immunoprecipitation increased by ≈2-fold (Figure 4C) after 24 hours of TGFβ1 in mutant MSCs. To further support the decreased availability of BECN1 caused by excessive DSP-unbound VIM in DSP-mutant MSCs, we checked the levels of association of BECN1 with 2 known partners in the autophagy and lysosomal degradation pathway, ATG14L and UVRAG (UV radiation resistance associated gene proteins; Figure 4D)³⁷ by co-immunoprecipitation, respectively. Both ATG14L and UVRAG protein associations with BECN1 were also decreased by ≈50%. These data in Figure 4A through 4D strongly support that excessive DSP-unbound VIMs actively sequester BECN1 from participating in multiple BECN1-mediated signaling pathways in DSP-mutant MSCs (see more in Figures 5 and 6). To further support the notion that DSP competes with BECN1 for VIM binding, we knocked down DSP in normal MSCs with specific siRNA against DSP and found that BECN1-VIM association increased, and BECN1 associations with ATG14 and UVRAG decreased accordingly (Figure 4E). The DSP knockdown protocol is shown in Figure S5A. Both DSP mRNA and protein levels were decreased significantly by siRNA against DSP when compared with normal MSCs treated with control nonsilencing siRNA (Figure S5B). To further show that VIM plays an active role in pathological fibrogenesis in DSP-deficient MSCs, we knocked down VIM in DSP-mutant MSCs with 4 known siRNAs against VIM and found that collagen 1 levels by Western blot (Figure 4F) and collagen content in media (Figure 4G) decreased to the levels of 72-hour TGFβ1-treated normal MSCs. After VIM knockdown in DSP-mutant MSCs, BECN1 associations with ATG14L and UVRAG increased by almost 2-fold (Figure 4H), and LC3B-II to -I ratios also almost doubled (Figure 4I) when compared with MSCs treated with control nonsilencing siRNA and TGFβ1 at 24 hours. Taken together, data in Figure 4 strongly support that increased DSP-unbound VIMs sequester BECN1 from actively participating in autophagic degradation of collagens in DSP mutant MSCs after TGFβ1. VIM-DSP interactions regulate BECN availability and mediate TGFβ1-induced pathological fibrosis in DSP-mutant MSCs. Withaferin A (WFA) is presumed to be an inhibitor of VIM assembly that could decrease fibrotic lung injury at 0.25 to 1 µmol/L.³⁹ Therefore, we explored the potential anti-fibrotic effect of WFA on our MSCs, but WFA, when co-applied with TGFβ1, killed all human MSCs at concentrations as low as 10 nmol/L in 1 day and could not be tested for its efficacy in reducing pathological fibrosis in DSP-mutant MSCs.

Figure 5.

TGFβ1 (transforming growth factor β1) induced higher fibrotic gene expression via long CAV1 (caveolin-1) plaques and p-p38 overactivation in DSP (desmoplakin)-mutant mesenchymal stromal cells (MSCs). A, COL1A1, COL3A1, and FN expression was higher in mutant MSCs (n=3) than in normal MSCs (n=6) after TGFβ1 for the indicated durations. B, Protein phosphorylation screens showed that only the p-p38 pathway was enhanced after TGFβ1 in mutant MSCs (n=5). C, Western blots showed (n=3) that total p38 levels (top) did not change with time after TGFβ1, yet p-p38 levels increased in mutant MSCs (middle), which could be quickly screened by measuring p-p38 intensity using immunocytochemistry (bottom, n=5). Time scales are for all plots in (C). D, DSP KD in normal MSCs enhanced the p-p38 response to TGFβ1 at 2 hours. E, Only the increased fibrotic gene expression in mutant MSCs by TGFβ1 could be effectively inhibited by SB2 and SB5 (n=6–12). F and G, In human DSP mutant heart tissues (yellow circles) and DSP-mutant MSCs (arrows), CAV1 formed long plaques when compared with normal counterparts. H, DSP KD in normal MSCs with siDSP, but not siCTR (arrowheads), induced long CAV1 plaques (arrows). I, CAV1 KD in mutant MSCs reduced p-p38 responses to 2-hour TGFβ1. *P by Student t except C by Mann-Whitney U test. Scales are shown in images. Number of counted cells from 3 Bio-replicates in G and H.

Figure 6.

BECN1 overexpression (OE) reduced long CAV1 (caveolin-1) plaques, p-p38 overactivation, fibrotic gene expression, autophagy defects, and collagen accumulations in DSP (desmoplakin)-mutant mesenchymal stromal cells (MSCs). A, In DSP-mutant MSCs, OE of BECN1 (yellow circles), but not nontransfected cells (NT, arrowheads) or GFP (green fluorescent protein) vectors, reduced long CAV1 plaques at baseline and (B) p-p38 overactivation by 2-hour TGFβ1 (transforming growth factor β1), which could be prevented by 1 µmol/L dynasore (Dyn, arrows). C, BECN1 OE, but not NT or vector transfection (+vector), decreased fibrotic gene expression by TGFβ1. D, BECN1 OE, but not +vector, enhanced LC3B-I to II conversions. E and F, BECN1 OE but not +vector or NT (arrowheads), decreased collagen in culture media and collagen staining by 72-hour TGFβ1. * P<0.05 (ANOVA). The total numbers of counted cells from 3 biological replicates are shown in (A), (B), and (F). For C through E, the numbers of biological replicates are shown. Scale bars are 50 µm. *P values by Student t except C and D, by Mann-Whitney U test.

Increased Fibrotic Gene Expressions in DSP-Mutant MSCs Via p38 Pathway

In addition to activating autophagy, TGFβ1 activates profibrotic/fibrogenic genes during normal controlled reparative fibrosis.^23,24 We therefore examined whether DSP haploinsufficiency in DSP-mutant MSCs also modified expressions of ECM mRNA, such as collagen types 1a1 and 3a1 (COL1A1 and COL3A1), and fibronectin (FN). We collected mRNAs at baseline, 24, 48, and 72 hours after treating normal (SW line) and DSP-mutant MSCs with TGFβ1, and then performed qPCR to evaluate their fibrogenic gene expression. We found that levels of COL1A1, COL3A1, and FN mRNA were significantly higher in DSP-mutant MSCs than those in normal MSCs after 72-hour TGFβ1 treatment (Figure 5A). To further support the hypothesis that DSP plays a key role in increased fibrogenic gene expression, we showed that DSP knockdown in normal MSCs by siRNA against DSP, but not by control nonsilencing siRNA, increased expression of COL1A1, COL3A1, and FN as well as collagen content in culture media after 72-hour TGFβ1 (Figure S5C and S5D). Thus, the increased fibrotic gene expression and collagen content by TGFβ1 are specific to DSP-deficient MSCs.

To further understand how DSP deficiency can increase fibrogenic gene expression, we looked into 3 major TGFβ1-mediated signaling pathways. For fibrogenesis, TGFβ1activates SMAD2/3 and non-SMAD-dependent signaling cascades (p38 and ERK MAPK [mitogen-activated protein kinase] pathways) to induce fibrotic gene expression in various tissues/organs.²³

Thus, we explored how DSP haploinsufficiency in MSCs affected the activation of SMAD2/3 and non-SMAD-dependent signaling (p38 and ERK) pathways. Because TGFβ1 usually activates these signaling pathways on timescales of minutes to hours,⁴⁰ we checked the ratios of phosphorylated to total protein levels of SMAD2/3, p38, and ERK1/2 at baseline, 1, 2, and 4 hours after treating MSCs with 1 ng/mL TGFβ1. Western blot analyses showed that only phospho-p38 (p-p38)/p38 ratios in both DSP-mutant MSCs were higher than those in normal MSCs at 2 and 4 hours after TGFβ1 (Figure 5B). The phospho-SMAD2/3 (p-SMAD2/3)/SMAD2/3 or phospho-ERK1/2 (p-ERK1/2) to ERK1/2 ratios were similar between normal and DSP-mutant MSCs after TGFβ1. We then examined the p-p38 and p38 levels in a longer time frame, since fibrotic genes are significantly elevated at 72 hours after TGFβ1 in DSP-mutant MSCs. We found that total p38 levels are relatively unchanged between different MSC lines from baseline to 72 hours after TGFβ1 (Figure 5C, top panel). Changes in p-p38/p38 ratios depended mainly on the p-p38 levels in all MSC lines (Figure 5C, middle panel). Furthermore, to quickly screen the potential mechanisms regarding how DSP deficiency enhanced the p38 pathway, we measured the intensities of p-p38 immunofluorescence by immunocytochemistry in MSCs and showed that the changes in p-p38 fluorescent intensities in MSCs reliably mirrored the changes measured by Western blots, especially in the first 16 hours after TGFβ1 exposure (Figure 5C, lower panel). Moreover, DSP knockdown in normal MSCs enhanced p-p38 levels (Figure 5D), fibrotic gene expression, and collagen secretion (Figure S5C and S5D) after TGFβ1, supporting that decreased WT DSP levels are responsible for the enhanced p-p38 and fibrotic responses to TGFβ1. We also used a TGFβ1 receptor blocker (1 µmol/L SB525334 [SB5]) and a p38 inhibitor (1 µmol/L SB202190 [SB2]) to study their effects on the TGFβ1-induced fibrotic gene activation.⁴¹ As predicted, SB2 inhibited fibrotic gene activation as well as SB5 in DSP-mutant MSCs, but the TGFβ1 receptor blocker (SB5) is more effective in blocking TGFβ1-induced activation of fibrotic genes in normal MSCs (Figure 5E). These results indicate that DSP deficits lead to specifically enhanced p-p38 activities and subsequently increased fibrotic gene expressions in pathological fibrosis.

DSP Deficiency Induced Long CAV1 Plaques and p38 Overactivation

CAV1, an integral membrane protein, is known to form caveolae to terminate non-SMAD signaling pathways after TGFβ1 stimulation in MSCs.^23,24 Since our data showed that the p38 pathway is abnormally activated in DSP-mutant MSCs, we hypothesized that CAV1-mediated termination of TGFβ1-induced p-p38 responses might be downregulated in DSP-mutant MSCs. We first immuno-stained a human ACM heart sample with a heterozygous DSP c.4531G>T (p.Gln1511X or Q1511X) mutation, and both DSP-mutant MSC lines with a CAV1 antibody. CAV1 is expressed in MSCs but not in cardiomyocytes.⁴² Interestingly, we found that CAV1 formed abnormally long CAV1 clusters on the cell membranes (termed CAV1 plaques here) in the nonmyocardial regions (negative for a cardiomyocyte-specific marker, α-actinin immunostainings) of human ACM heart samples with the DSP Q1511X mutation (Figure 5F, left panels) and in 2 DSP-mutant MSC lines (Figure 5F, right panels) at baseline when compared with normal counterparts (Figure 5F through 5G). Knocking down DSP in normal MSCs with siRNA against DSP also resulted in abnormal formation of long CAV1 plaques (Figure 5F and 5H). More importantly, CAV1 knockdown in DSP-mutant MSCs eliminated the abnormal activation of the p38 pathway after 2-hour TGFβ1, indicating that the abnormally long CAV1 plaques mediate the abnormal p38 overactivation after TGFβ1 in DSP-mutant MSCs (Figure 5I). The CAV1 knockdown efficiency is shown in Figure S5E.

Decreased BECN1 Availability Raises Fibrotic Genes Via Long CAV1 Plaques and p-p38

Since our data indicate that decreased BECN1 availability is the main reason for decreased autophagy/collagen degradation and rapamycin (an autophagy activator) could suppress collagen accumulation in DSP-mutant MSCs (Figures 2 through 4), we tested whether rapamycin could decrease CAV1 plaques. Treatment of DSP-mutant MSCs with 2 days of 10 nmol/L rapamycin minimally changed the length of CAV1 plaques in DSP-mutant MSCs (Figure S5F). However, BECN1 has been shown to affect various types of endocytosis and endocytic trafficking.⁴³ We, therefore, hypothesized that BECN1 availability might also affect the endocytosis of CAV1 and tested whether overexpression of BECN1 could mitigate the abnormal CAV1 plaque formation on cell membranes and p-p38 over-activation in DSP-mutant MSCs. We found that overexpression of BECN1 for 2 days eliminated the abnormally long CAV1 plaques (Figure 6A) and significantly decreased fluorescent intensities of p-p38 at 2 hours after TGFβ1 (Figure 6B) in DSP-mutant MSCs. These effects of BECN1 overexpression on CAV1 plaques and p-p38 intensity could be blocked by cotreatment with an endocytosis/dynamin inhibitor, Dynasore.⁴⁴ Overexpression of BECN1 (with 50% to 70% transfection efficiency by electroporation) also decreased fibrotic gene expression levels (COL1A1, COL3A1, and FN) at 72 hours after TGFβ1 in DSP-mutant MSCs when compared with nontransfected or vector-transfected MSCs (+vector; Figure 6C). Taken together, these data support that DSP deficiency in DSP-mutant MSCs reduces BECN1 availability for CAV1-mediated endocytosis, leading to long CAV1 plaque formations and subsequent p-p38/fibrotic gene over-activation after TGFβ1. Overexpression of BECN1 restored LC3B-II/-I ratios in autophagy (Figure 6D), reduced fibrillar collagen accumulation/secretion (Figure 6E), and decreased collagen immune-staining in DSP-mutant MSCs after 72-hour TGFβ1 (Figure 6F).

Over-Expressing VIM-Binding Domains of DSP Reduces Pathological Fibrosis

Exogenous expression of PKP2 proteins has been proposed to treat ACM patients with PKP2 mutation.⁴⁵ Since DSP proteins exist and interact with VIM in all fibroblasts, over-expression of DSP might suppress all pathological fibrosis by increasing BECN1 availability for greater collagen autophagic degradation and CAV1-p38 pathway termination. However, desmoplakins are large-size proteins and have 2 isoforms created by alternative splicing: DSP-I (≈332 kD) and DSP-II (≈260 kD).^29,30 Expressing the full-length DSP might not be efficient in vivo and would be hard to use for clinical therapy. We thus subcloned a hemagglutinin-tagged DSP fragment with a deleted N-terminal domain (named hemagglutinin DSPΔN, containing the minimal required domains for VIM binding)⁴⁶ to perform rescue experiments in MSCs. The hemagglutinin tag would allow us to identify exogenous-expressed DSPΔN proteins by hemagglutinin antibody but not by our DSP antibodies against the DSP N-terminal domains. Over-expressing hemagglutinin-DSPΔN in DSP mutant MSCs restored their VIM-DSP associations after TGFβ1 (Figure 7A), LC3B-II/-I ratios at baseline (Figure 7B), collagen content in culture media (Figure 7C), and collagen immunocytochemistry staining (Figure 7D) after TGFβ1 to the levels of normal MSCs. Over-expressing hemagglutinin-DSPΔN in mutant MSCs also decreased BECN1-VIM associations and increased associations of BECN1 with UVRAG and ATG14 (complexes mediating autophagy, Figure 7E). CAV1 plaque lengths, phospho-p38 immunocytochemistry staining levels, and subsequent fibrotic gene activations (COL1A1, COL3A1, and FN) after TGFβ1 are all decreased by overexpressing HA-DSPΔN in mutant MSCs (Figure 7F through 7H). Overall, our results strongly support that DSP mutations lead to DSP deficiency, which mediates the pathological fibrogenesis of mutant cardiac MSCs by decreasing BECN1 availability via increased BECN1-VIM interactions. Over-expressing DSPΔN might be a new strategy to combat pathological fibrosis in other diseased organs by increasing autophagic degradation of collagens and decreasing CAV1-p38-mediated fibrotic gene activation.

Figure 7.

Overexpression (OE) of DSPΔN ameliorates the pathological fibrogenesis in mutant mesenchymal stromal cells (MSCs). A, DSPΔN OE increased VIM (vimentin)-DSPΔN association (also see Figure S6A) and (B) LC3B-II/I ratios in mutant MSCs. C, DSPΔN OE decreased collagen content and (D) collagen staining in mutant and normal MSCs vs no transfection (NT) or vector transfection (+vector). E, DSPΔN OE decreased VIM-BECN1 but increased BECN1-UVRAG and BECN1-ATG14 associations in mutant MSCs. F, DSPΔN OE (yellow circles) decreased CAV1 (caveolin-1) plaque length at baseline (base) and (G) p-p38 intensities by TGFβ1 (transforming growth factor β1) in mutant MSCs. H, DSPΔN OE decreased fibrotic gene levels by TGFβ1 when compared with NT cells. *P values by Student t (except B, C, and E by Mann-Whitney U test; H by Kruskal-Wallis). Scale bars=50 µm.

Heterozygous DSP Deletion in Mouse Myofibroblasts In Vivo

To obtain in vivo evidence that DSP deficiency in activated fibroblasts (myofibroblasts) enhances fibrogenic responses to TGFβ1 in vivo, we used tamoxifen-controlled, periostin-specific Cre recombinase (POSTN^MRM-Cre+)³¹ to heterozygously delete DSP in activated fibroblasts of DSP-floxed mice (DSP^w/f) with 5-day tamoxifen intraperitoneal injections at 3-week-old mice. After tamoxifen injection, 6-week-old mice were intraperitoneally injected with 1 µg TGFβ1 daily for 7 to 8 days, and mouse hearts were then harvested for comparisons at the seventh week. When compared with POSTN^MCM-Cre-negative; DSP^w/f mice, POSTN^MCM-Cre-positive; DSP^w/f mice showed increased cardiac fibrosis (stained by Sirius red, Figure 8A), decreased autophagy (lower LC3B-II/I ratios and higher p62 levels), higher p-p38 and fibrotic gene levels, and lower DSP mRNA/protein levels (Figure 8B through 8D). Of note, SQSTM1 (sequestosome 1; or p62) binds LC3B to deliver ubiquitinated cargo proteins for autophagic degradation. As such, p62 accumulations with low LC3B-II/I ratios are indicators of suppressed autophagy.³⁸ Additionally, when compared with 3 normal human donor hearts, the human ACM heart with a heterozygous Q1511X mutation (from Figure 5F) also showed increased fibrotic gene levels, higher p-p38/p38 ratios, and decreased autophagy markers (lower LC3B-II/I ratios with elevated p62 levels; Figure S6B and S6C). Importantly, cardiac fibroblasts directly isolated from POSTN^MCM-Cre-positive mouse hearts showed higher BECN1-VIM associations and lower BECN1-ATG14/BECN1-UVRAG associations (Figure 8E) after TGFβ1. The Cre-positive mouse cardiac fibroblasts also showed long CAV1 plaques at baseline (Figure 8F) and higher collagen secretion and p-p38 levels after TGFβ1 (Figure 8G-H). Taken together, these data strongly support that DSP deficiency led to increased cardiac fibrosis after TGFβ1 in vivo via the same pathogenic mechanisms as we showed in Figures 1 through 7. Most importantly, we co-immunostained human MSCs with CAV1 and TGFBR1 (type 1 TGFβ receptor) antibodies. We found that CAV1 plaque-TGFBR1 colocalizations on cell membranes in DSP mutant MSCs were higher at baseline, which increased significantly after 2-hour TGFβ1 when compared with normal MSCs (Figure 8I; Figure S6D), strongly supporting that CAV1-mediated TGFBR1 endocytosis^23,24 was downregulated in mutant MSCs. Figure 8J showed a diagram summarizing the pathogenic network specific to deficient VIM-DSP interactions, which led to decreased BECN1 availability and enhanced fibrogenic responses to TGFβ1 in DSP-deficient MSCs.

Figure 8.

Hearts and cardiac fibroblasts from DSP^w/f mice with POSTN^MRM-Cre-positive or -negative genotypes. DSP^w/f mouse hearts with POSTN^MRM-Cre (Cre+) showed (A) enhanced fibrosis to TGFβ1 (transforming growth factor β1) in vivo by Sirius red-Fast green staining; (B) decreased autophagy (low LC3B-II/I ratios and high p62 levels), overactivated p-p38; (C) higher fibrotic gene expression; and (D) lower DSP (desmoplakin) levels when compared with Cre-negative mice. DSP^w/f mouse cardiac fibroblasts with Cre+ showed (E) higher BECN1-VIM (vimentin) binding but lower BECN1-UVRAG/ BECN1-ATG14 associations; (F) long CAV1 (caveolin-1) plaques at baseline; (G) higher collagen secretion by TGFβ1; and (H) higher p-p38 levels by TGFβ1. I, Using co-immunostaining, CAV1-TGFBR1 colocalizations at the cell membrane (yellow circles) were higher at baseline and increased significantly after 2-hour TGFβ1 in human DSP mutant mesenchymal stromal cells (MSCs; Figure S6D) but not in normal MSCs. J, A simple diagram for TGFβ1-mediated pathological fibrosis due to reduced BECN1 availability in DSP mutant MSCs. Scale bars are 50 µm. All data were from 3 independent experiments. *P values by Mann-Whitney U test (except F, H, and I by Student t).

DSP Mutations and Fibrogenic Pathways by Gene Expression Analysis

To obtain a global view of how DSP deficiency affects the gene expression networks (Figures S6E and S6F and S7), we used Illumina mRNA expression microarrays to profile iPSC-MSCs from DSP R160X proband, DSP R160X proband’s mother (carrying the same heterozygous DSP mutation but with no clinical ACM), DSP E1159R proband, and the normal (SW) line at baseline to find common deregulated pathways. Notably, iPSC-MSCs derived from asymptomatic R160X proband’s mother (a mutation carrier) had ≈40% of WT DSP mRNA levels but ≈82% of WT DSP protein levels (versus normal MSCs), did not have long CAV1 plaques, did not show increased production of collagens (1 and 3) and VIM, nor increased fibrotic gene expression after 72-hour TGFβ1 (Figure S7A through S7E). To find the common differentially expressed genes between 2 patients with ACM MSC lines and 2 MSC lines without clinical ACM, we first generated 3 sets of pairwise comparisons from the mRNA expression profiles: (1) DSP R160X proband versus his mother (Mother) with no clinical ACM, (2) DSP R160X proband versus a normal donor without ACM (SW), and (3) DSP E1159R proband versus the normal SW donor. We then identified the consistently upregulated or downregulated (by 1.3-fold) genes common to these 3 pairwise comparisons (total 1186 genes, Figure S6E and S6F). Using the DAVID 6.7 KEGG pathway analyses, we identified the top 15 dysregulated pathways for the 2 DSP-mutant MSCs with clinical ACM, which showed altered expressions of genes in ECM, focal adhesion, several cancer pathways, axon guidance (axon guide), vascular smooth muscle contraction, hematopoietic cell lineage, cytokine-cytokine receptor interaction (cytokine interact), viral myocarditis, and ACM (termed ARVC in the KEGG pathway), as well as signaling pathways of TGFβ, GnRH (gonadotropin-releasing hormone), and Jak-STAT pathways. These results further support our findings that DSP deficiency plays a key role in pathological fibrogenic responses to TGFβ1 in DSP-mutant MSCs because ECM, focal adhesion pathway, TGFβ pathway, and ACM are known pathways that are related to pathological cardiac fibrosis in cardiomyopathic hearts.

Of note, we also studied TGFβ1 response of a plakophilin 2 (PKP2) haplo-insufficient ACM MSC line without DSP deficiency (Figure S7A and S7B, the delC line with a PKP2 c.2013delC mutation that was previously characterized in reference 27). When compared with DSP mutant MSCs, delC PKP2-deficient MSCs did not have excessive productions of collagens 1/3 or VIM after 72-hour TGFβ1 (Figure S7B and S7C), have no abnormally long CAV1 plaques (Figure S7D) nor excessive fibrotic gene expressions (Figure S7E). Thus, the pathogenic mechanisms presented here for increased fibrogenic responses to TGFβ1 are specific for DSP-deficient MSCs but not for PKP2-deficient MSCs.

Discussion

Pathogenic variants or single nucleotide polymorphisms in the DSP gene have been implicated in the pathogenesis of ACM,^5,6 IPF,⁷ hypertrophic scars,⁸ and cancer metastasis⁴⁷ with unknown mechanisms. Here, we used patient-specific iPSC-MSCs derived from ACM patients with DSP mutations to elucidate the novel pathogenic roles of DSP deficiency in pathological fibrosis. We first showed that 2 different DSP truncation mutations in MSCs resulted in no new mutant DSP protein but ≈50% reduction of WT DSP (a DSP haplo-insufficient phenotype, Figure 1). DSP-deficient MSCs displayed exaggerated collagen production/secretion and overactivated fibrotic gene expression (Figures 2 and 5). Importantly, DSP deficiency resulted in decreased DSP-VIM associations and increased VIM-BECN1 interactions, which restricted BECN1 availability and led to decreased autophagic degradation of collagens due to decreased BECN1-ATG14 and BECN1-UVRAG complexes (Figures 3 and 4).

Decreased BECN1 availability also induced abnormal formation of long CAV1 plaques that over-activated the p38 pathway after TGFβ1 and increased fibrotic genes and protein levels. These fibrogenic pathologies can be reproduced by DSP knockdown in normal MSCs in vitro (Figures 5 and 6; Figure S5) and in mouse activated fibroblasts in vivo (Figure 8). As predicted, overexpression of BECN1 (Figure 6) or HA-DSPΔN (Figure 7) rescued all pathological fibrotic responses to TGFβ1 in DSP-deficient MSCs. Genome-wide comparisons by mRNA expression array analysis further supported the role of DSP in pathological cardiac fibrosis in DSP-deficient MSCs (Figure S6E and S6F).

VIM is a type III cytoskeletal intermediate filament that has been shown to maintain cell structural integrity, organize cytosolic organelles, and participate in cell adhesion, migration, invasion, and signaling.⁴⁸ Structurally, VIM is composed of a head domain that can bind BECN1,³⁹ a central coil-coil domain responsible for VIM dimer/tetramer formation that can interact with the C-terminal plakin repeat domains of DSP,⁴⁹ and a C-terminal tail domain. VIM is the only cytoplasmic intermediate-filament system in fibroblasts⁵⁰ and has been associated with inflammatory and autoimmune diseases, cancers, and various fibrotic diseases⁵⁰ via unknown mechanisms. Of note, VIM is frequently used as a classic marker for cells of mesenchymal origin, cells undergoing epithelial-to-mesenchymal transition, and activated fibroblasts (myofibroblasts).³⁴ Whether VIM actively participates in fibrogenesis remains to be determined. Although VIM has been shown to bind BECN1 and decrease autophagy,³⁸ the mechanisms that regulate VIM-BECN1 interactions for fibrogenesis remain unclear. Our study indicates that DSP is the main regulator of BECN1-VIM interaction during fibrogenesis. In DSP-deficient MSCs, diminished amounts of WT DSP protein increase the levels of VIM-BECN1 associations.

Increased VIM-BECN1 binding in turn restricts BECN1 availability for initiating autophagy (by decreased BECN1/ATG14L-PI3K complexes) and for autophagosome-lysosome fusion (by decreased BECN1/UVRAG complexes), leading to decreased collagen degradation and increased cytosolic fibrillar collagen and collagen secretion.

Interestingly, VIM has been shown to stabilize COL1A1 mRNA between 6 and 24 hours after TGFβ without affecting fibronectin levels.⁵¹ This particular mechanism could only explain the COL1A1 mRNA expression pattern at 24 hours but not at 72 hours after TGFβ1 in our experiments (Figure 5A) because COL1A1, COL3A1, and FN mRNA levels were all significantly upregulated at 72 hours after TGFβ1, which could be effectively inhibited by a p38 blocker (Figure 5E).

Furthermore, knockdown or knockout of DSP in mouse cardiomyocytes and epidermal cells was associated with the abnormal upregulation of TGFβ1/p38 signaling via unknown mechanisms.⁵² Here, we show a novel finding that DSP knockdown or DSP deficiency in MSCs induces abnormal formation of CAV1 plaques due to defective endocytosis of CAV1.

Importantly, abnormal CAV1 plaques are the main mechanism for over-activating p-p38 and subsequent fibrotic gene expression (COL1A1, COL3A1, and FN) by TGFβ1. Of note, MSCs derived from an ACM patient with PKP2 haploinsufficiency (the delC line) did not have DSP deficiency, did not show any abnormal CAV1 plaques, nor did they show enhanced fibrogenic responses to TGFβ1 (Figure S7). These novel findings uncover the missing link between DSP deficiency and exaggerated p38 pathway activation in cardiac fibroblasts/MSCs.Our data also have significant therapeutic implications for treating pathological fibrosis. Silencing VIM in fibroblasts may lead to defects in wound healing and cellular polarity.³⁹ Nonspecific autophagy activators could degrade other cellular proteins, leading to unwanted side effects. Downregulating CAV1 remains an uncertain antifibrotic strategy because CAV1 has dynamic roles in normal fibrogenic processes.⁵³ Drugs that inhibit VIM assembly, such as WFA, have been shown to reduce invasiveness and fibrogenic potential of fibroblasts in IPF at 0.25 to 1 µmol/L.³⁹ However, WFA has pleiotropic effects on various types of cells, including reduction of inflammatory cytokines and attenuating profibrotic gene expressions (including COL1A2, COL3A1, and FN),⁵⁴ but can cause cancer cell apoptosis.⁵⁵ These widespread effects of WFA might indicate that WFA could affect fibrosis via mechanisms in addition to just increasing VIM disassembly. In our DSP mutant MSC cultures, coapplication of WFA even at 10 nmol/L with TGFβ1 killed most cells in 1 day, and thus we could not further study the anti-fibrotic effect of WFA. Importantly, efficient overexpression of a VIM-binding domain of DSP protein may specifically tackle any pathological fibrosis mediated by over-activated VIM-BECN1 interactions. Thus, overexpressing VIM-binding domains of DSP potentially might be used to treat DSP cardiomyopathy and other types of pathological fibrosis with exaggerated VIM expression. Also, a p38 inhibitor, ARRY-371797, has been used to treat cardiomyopathy from Lamin A/C mutations (NCT03439514 in ClinicalTrials.gov)⁵⁶ and potentially could be used to treat cardiomyopathy with DSP mutations or other organ fibrosis due to DSP deficiency. Finally, reduced DSP levels are associated with carcinogenesis.⁵⁷ Our novel findings may instigate research for using VIM-binding domains of DSP as a novel therapy against cancer invasion or metastasis.

ARTICLE INFORMATION

Acknowledgments

The authors thank all patients for their participation, Microarray and Genomic Core facilities at SBP Medical Discovery Institute and at Indiana University for their support, Johnson Wong, Ed Simpsons and Yunlong Liu for RNA seq analysis, Tiara Tirasawasdichai for conducting some preliminary experiments during her graduate studies, Zhenhui Chen for his reanalysis of immunostaining and Western blot images in a blinded fashion, and AnaBios Inc. for normal donor heart tissues.

Author Contributions

C. Wei, S.-F. Chan, and H-S.V. Chen designed experiments and wrote the article; C. Wei, W. Shou, S.-F. Chan, and H-S.V. Chen performed experiments and data analysis. A.M. Saguner, C. Brunckhorst, F. Duru, J.E. Marine, C.A. James, H. Calkins, W. Shou, and D.P. Judge provided clinical assessment and patients’ fibroblasts/heart tissues as well as helped in the preparation of this article. All authors read and approved the article, and H.-

S.V. Chen supervised the entire research project.

Disclosures

None.

Supplemental Material

Supplemental Methods

Figures S1–S7

Tables S1–S5

Source data file for Figures 1–8 and S1–S7

Source data file for Figures 1C and S6E and S6F

Source Figures With Unprocessed Blots

References 58–68

Supplementary Material

res-139-e325512-s001.pdf ^{(6.4MB, pdf)}

res-139-e325512-s002.pdf ^{(3MB, pdf)}

res-139-e325512-s003.pdf ^{(10.1MB, pdf)}

res-139-e325512-s004.xlsx ^{(117.9KB, xlsx)}

res-139-e325512-s005.xlsx ^{(1.1MB, xlsx)}

Funding Statement

This work was supported (1) by grants to The Zurich ARVC Program from the Georg and Bertha Schwyzer-Winiker, the Baugarten, the USZ Foundation (Wild grant), the Swiss Heart, and the Swiss National Science Foundations (A.M. Saguner, C. Brunckhorst, and F. Duru); (2) by California Institute of Regenerative Medicine (CIRM) Grants RB4-06276 (H-S.V. Chen); (3) by National Institutes of Health grants RO1 HL105194 and R56 HL161741 (H-S.V. Chen); and (4) by Krannert Institute of Cardiology start-up funds (H-S.V. Chen).