RESEARCH ARTICLE

Haploinsufficiency of mechanistic target of rapamycin amelioratesbag3 cardiomyopathy in adult zebrafishYonghe Ding1,2,*, Alexey V. Dvornikov1,2,‡, Xiao Ma1,2,3,4, Hong Zhang1,2,5, Yong Wang1,2,6, Matthew Lowerison7,Rene R. Packard8, Lei Wang9, Jun Chen10, Yuji Zhang9, Tzung Hsiai8, Xueying Lin1,2 and Xiaolei Xu1,2,3,4,*

ABSTRACTThe adult zebrafish is an emerging vertebrate model for studyinghuman cardiomyopathies; however, whether the simple zebrafishheart can model different subtypes of cardiomyopathies, such asdilated cardiomyopathy (DCM), remains elusive. Here, we generatedand characterized an inherited DCM model in adult zebrafish andused this model to search for therapeutic strategies. We employedtranscription activator-like effector nuclease (TALEN) genome editingtechnology to generate frame-shift mutants for the zebrafish orthologof human BCL2-associated athanogene 3 (BAG3), an establishedDCM-causative gene. As in mammals, the zebrafish bag3homozygous mutant (bag3e2/e2) exhibited aberrant proteostasis, asindicated by impaired autophagy flux and elevated ubiquitinatedprotein aggregation. Through comprehensive phenotyping analysisof the mutant, we identified phenotypic traits that resembled DCMphenotypes in mammals, including cardiac chamber enlargement,reduced ejection fraction characterized by increased end-systolicvolume/body weight (ESV/BW), and reduced contractile myofibrilactivation kinetics. Nonbiased transcriptome analysis identified thehyperactivation of the mechanistic target of rapamycin (mTOR)signaling in bag3e2/e2 mutant hearts. Further genetic studies showedthat mtorxu015/+, an mTOR haploinsufficiency mutant, repairedabnormal proteostasis, improved cardiac function and rescued thesurvival of the bag3e2/e2 mutant. This study established the bag3e2/e2

mutant as a DCM model in adult zebrafish and suggested mtor as acandidate therapeutic target gene for BAG3 cardiomyopathy.

KEY WORDS: Dilated cardiomyopathy, BCL2-associatedathanogene 3, mTOR, Danio rerio

INTRODUCTIONDilated cardiomyopathy (DCM) is a heterogeneous group of cardiacdiseases characterized by an enlarged ventricular chamber, thinnedventricular walls and reduced cardiac function (Japp et al., 2016;Merlo et al., 2018). More than 100 genes have been linked to DCM.Mutations in sarcomeric genes, such as titin, account for >25% ofinherited DCMs (Herman et al., 2012). Mutations in many othergenes, which encode cytoskeletal, mitochondrial, desmosomal,nuclear membrane and RNA-binding proteins, have also beenidentified in DCM cases (McNally and Mestroni, 2017). Thus, theprimary damage to the heart must be highly different among DCMsubtypes with distinct etiologies. Nevertheless, the initial damagecan trigger cascades of sequential pathological events that convergeon certain common pathological pathways, which eventually resultin DCM phenotypes. Uncovering common pathological pathwaysof DCM promises the development of shared therapeutic strategiesfor different subtypes of DCMs.

BCL2-associated athanogene 3 (BAG3) mutations were initiallylinked to DCM via two genome-wide association studies (Nortonet al., 2011; Villard et al., 2011). Subsequent human genetic studiesestablished BAG3 as one of the most common DCM causative genes,with its variants contributing to 2.3-6.7% of DCMs (Dominguezet al., 2018; Franaszczyk et al., 2014). BAG3 cardiomyopathy islikely of a loss-of-function nature because truncation mutations inBAG3 are frequently found in DCM patients, and a cardiac-specificBag3-knockout mouse manifests DCM phenotypes (Chami et al.,2014; Dominguez et al., 2018; Fang et al., 2017; Franaszczyk et al.,2014). Multiple functions have been assigned to the BAG3 protein: itis a co-chaperone protein that binds heat shock protein 70 (HSP70)family members and regulates protein aggregation (Meriin et al.,2018; Sturner and Behl, 2017); a BCL2-binding protein that controlsapoptosis (Lee et al., 1999); and a Z-disc protein that is involved insarcomeric protein turnover (Arimura et al., 2011; Hishiya et al.,2010). Because these functions are related to protein quality control(PQC), abnormal protein homeostasis has been postulated as theprimary damage that causes DCM in patients with BAG3 mutations(Myers et al., 2018). Although repairing defective proteostasis couldbe a plausible therapeutic strategy, no target genes have yet beenreported for BAG3 cardiomyopathy.

Mechanistic target of rapamycin (mTOR) is a serine/threonineprotein kinase that plays a pivotal role in regulating proteostasis incardiomyocytes by regulating cardiomyocyte growth, autophagy andsurvival (Saxton and Sabatini, 2017; Sciarretta et al., 2018). mTORsignaling was previously perceived as a pathway involved inphysiological hypertrophy (Maillet et al., 2013). Accumulatingevidence suggests that mTOR signaling can also be manipulated tobenefit pathological cardiomyopathies (Sciarretta et al., 2014; Songet al., 2010). Elevated mTOR activity was detected in cardiachypertrophy and ischemia/reperfusion-induced heart injury(Sciarretta et al., 2018). Partial mTOR inhibition through eitherReceived 16 April 2019; Accepted 28 August 2019

1Department of Biochemistry and Molecular Biology, Mayo Clinic, Rochester,MN 55905, USA. 2Department of Cardiovascular Medicine, Mayo Clinic, Rochester,MN 55905, USA. 3Center for Clinical and Translational Science, Mayo Clinic,Rochester, MN 55905, USA. 4Mayo Graduate School of Biomedical Sciences,Mayo Clinic, Rochester, MN 55905, USA. 5Clinical Center for Gene Diagnosis andTherapy, the Second Xiangya Hospital of Central South University, Changsha,China 410011. 6Institute of Life Science, Beijing University of Chinese Medicine,Beijing, China 100029. 7Department of Urology,MayoClinic, Rochester, MN 55905,USA. 8School of Medicine, University of California Los Angeles, Los Angeles,CA 90073, USA. 9Department of Epidemiology and Public Health, University ofMaryland School of Medicine, Baltimore, MD 21201, USA. 10Division of BiomedicalStatistics and Informatics, Mayo Clinic, Rochester, MN 55905, USA.‡Present address: Department of Cellular and Molecular Medicine, University ofArizona, Tucson, AZ 85721, USA.

*Authors for correspondence (Xu.Xiaolei@mayo.edu; Ding.Yonghe@mayo.edu)

Y.D., 0000-0002-4531-1721; A.V.D., 0000-0003-1514-4747; X.L., 0000-0002-4928-3422; X.X., 0000-0002-4928-3422

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2019. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

pharmacologic or genetic inhibition exerted cardioprotective effectson several subtypes of cardiomyopathies, such as cardiac hypertrophy(Marin et al., 2011; McMullen et al., 2004), lamin A/C-deficientDCM (Ramos et al., 2012), and anemia and doxorubicin-inducedcardiomyopathies (DIC) (Ding et al., 2011). Whether mTORinhibition is effective in ameliorating the BAG3 cardiomyopathysubtype remains untested.Because of the unprecedented opportunities to conduct both

genetic and compound screening, adult zebrafish have recentlybeen developed as an emerging vertebrate model for humancardiomyopathy (Gut et al., 2017; Henke et al., 2017; MacRae andPeterson, 2015). Corresponding orthologs for most known humanDCM genes (96%) have been identified in zebrafish (Shih et al.,2015). Conserved cardiac remodeling responses occur when fishhearts are stressed by either chronic anemia or the chemotherapy drugdoxorubicin (Ding et al., 2011), and a titin truncation mutant inzebrafish exhibits cardiomyopathy-like phenotypes (Huttner et al.,2018). However, owing to its small body size and sponge-like heartstructure, phenotyping cardiomyopathy in adult zebrafish remains achallenging task. As a consequence, the characteristic DCMphenotypes and whether different subtypes of DCM can bediscerned in this simple vertebrate model remain unclear.Here, we report the generation of a zebrafish model of bag3

cardiomyopathy via genome editing technology. Utilizing emergingtechnologies, such as high-frequency echocardiography (HFE) (Wanget al., 2017), our newly developed ex vivo heart pump function assay(Zhang et al., 2018), and biophysical assays at the single-myofibrillevel (Dvornikov et al., 2014), we characterized phenotypic traitscomprehensively in the bag3 mutant. By comparison with otherexisting cardiomyopathy models, we proposed phenotypic traits thatcould be used to define DCM in an adult zebrafish. We show that themTOR pathway is hyperactive in the bag3mutant, and partial mTORinhibition exerts a cardioprotective effect on this particular subtype ofinherited cardiomyopathy.

RESULTSGeneration of bag3 mutations in zebrafishIn zebrafish, there is a single ortholog of the human BAG3 gene onchromosome 13. The bag3 gene encodes a protein that shares 55%similarity with the human BAG3 protein and up to 97% identity infunctional domains, such as the WW domain (Fig. S1). The zebrafishbag3 transcripts are enriched in striatedmuscles during embryogenesisand aremore predominantly expressed in the cardiacmuscle than in thesomites in adults (Fig. S2) (Shih et al., 2015).To model BAG3 cardiomyopathy, we targeted the 2nd exon to

generate bag3 loss-of-function mutants via transcription activator-like effector nuclease (TALEN) technology. Four different truncationalleles predicted to shift the reading frame and lead to a prematurestop codon were obtained, designated bag3e2-M1, -M2, -M3 and -M4(Fig. 1A,B and Fig. S3). No evident phenotypes were detected inthese mutants during embryonic stages (Fig. S4). However, all fouralleles, including both male and female fish, exhibited the samevisually noticeable phenotypes, including smaller body size andelongated Meckel’s cartilage at 3 months of age (Fig. 1C andFig. S5). For simplicity, subsequent experiments focused on thebag3e2-M1 allele that harbored a 10-nucleotide deletion, which wasrenamed bag3e2. The bag3 transcripts were reduced 37% in the hearttissues of bag3 heterozygous fish (bag3e2/+) and 87% in thehomozygous mutant (bag3e2/e2) (Fig. 1D), likely owing to nonsense-mediated RNA decay. We were unable to assess Bag3 proteinsbecause commercially available antibodies for mammalian BAG3did not recognize zebrafish Bag3 (data not shown). Swimming

capacity was significantly reduced in bag3e2/e2 fish (Fig. 1E).Although it is an important clinical index for heart failure in humanpatients, reduced swimming capacity in the bag3e2/e2 mutant couldalso be ascribed to defects in other tissues. Indeed, we noted a skeletalmuscle degeneration phenotype in the bag3e2/e2 homozygous mutantat 6 months and in bag3e/+ heterozygous fish at 12 months (Fig. S6).This result is consistent with its function as a myopathy-causativegene in humans (Selcen et al., 2009). The bag3e2/e2 mutant hadincreased mortality:∼80% of fish died within 1 year of age (Fig. 1F).

The bag3e2/e2 mutant manifests reduced cardiac pumpfunctionNext, we focused on evaluating cardiac pump functions in thebag3e2/e2 mutant. We focused our efforts on 6 months of age as thetime point at which the fish start to die (Fig. 1F) and presentedcombined data from both male and female fish because of theirsimilar phenotypes in our initial studies. We initially conducted apulsed-wave Doppler analysis using the Vevo 2100 imaging systemequipped with a 30 MHz transducer (Packard et al., 2017) anddetected reduced early (E) ventricular filling velocity, normal late (A)ventricular filling velocity, and a significantly reduced E/A ratio in thebag3e2/e2mutant, suggesting a decline in diastolic function (Fig. 2A).The bag3e2/e2 fish also had an extended isovolumic contraction time(IVCT) and a reduced ejection time (ET), which led to an overallelevatedmyocardium performance index (MPI) (Fig. 2B), suggestinga worsening of global cardiac function. More recently, we had accessto the Vevo 3100 image system equipped with a 50 MHz transducer(Wang et al., 2017). The higher resolution of this updated systemenables the determination of chamber borders of a beating adultzebrafish heart. We detected a significantly reduced ejection fraction(EF) and fractional shortening (FS) (Fig. 2C,D; Movies 1 and 2),which can be largely ascribed to the increased end-systolic volume(ESV)/body weight (BW) ratio. The end-diastole volume (EDV)/BWratio remained unchanged.

Beforewe obtained access to the HFE (Vevo 3100) system, we hadbeen developing a Langendorff-like system to quantify cardiacfunction (Zhang et al., 2018). Driven by both electrical pacing andperfusion via fluid flow through the ventricle, the beating heart couldbe documented at high resolution using a digital camera ex vivo.Consistent with echocardiography-based assays, the indices for pumpfunction and contractility, including EF and FS, were significantlydecreased in the bag3e2/e2 mutant, confirming cardiac dysfunction(Fig. 2E;Movie 3). The Langendorff-like ex vivomethod also enablesthe measurement of other functional indices, such as the velocities ofshortening and relaxation, the primary determinant of myocardialpower, and velocity-strain loop area index (an alternative index for thepressure-volume loops to define cardiac performance). All threeindices were significantly reduced in the bag3e2/e2 mutant heart(Fig. 2F-J), confirming a contractility dysfunction phenotype.

The bag3e2/e2 mutant manifests hallmarks ofcardiomyopathyHaving defined cardiac pump dysfunction in the bag3e2/e2mutant, weexamined which phenotypic traits of mammalian cardiomyopathy canbe recapitulated in the zebrafish model. By quantifying the ventricularsurface area (VSA) in dissected hearts, we noted an enlarged heartchamber size, as indicated by the significantly increased VSA/BWindex in the bag3e2/e2 mutant at 6 months of age (Fig. 3A; Fig. S8A).Hematoxylin and eosin (H&E) staining revealed myofibril loss andsignificantly reduced trabecular muscle density in the sectionedbag3e2/e2 fish hearts (Fig. 3C). We confirmed the myofibrildegeneration phenotype by transmission electron microscopy

2

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

(TEM) analysis in 6-month-old bag3e2/e2 fish hearts and notedabnormal mitochondrial swelling and/or vacuolization phenotypes(Fig. 3B). We also noted marginal myofilament disruption andmitochondrial vacuolization phenotypes in the bag3e2/+ heterozygousfish at 12 months but not at 6 months (Fig. S7).To evaluate bag3 cardiomyopathy at the cellular and molecular

levels, we performed a terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) assay. We notedsignificantly increased apoptosis in the bag3e2/e2 mutant at 6 months(Fig. 3E). In contrast, we did not find any significant change inindividual cardiomyocyte cell size (Fig. S8B). The proliferation indexwas not changed, as measured by proliferating cell nuclear antigen(PCNA) staining (Fig. S8C). At the molecular level, we detectedaberrant fetal gene reprogramming using quantitative RT-PCR,including elevated ventricular myosin heavy chain (vmhc, alsoknown as myh7) and natriuretic peptide B (nppb) expression anddecreased natriuretic peptide A (nppa) expression (Fig. 3D).

In contrast to hypertrophic cardiomyopathy (HCM), which oftenmanifests as myofibril hypercontractility, DCM-causative mutationsare associated with myofibril hypocontractility (Spudich et al., 2016;Trivedi et al., 2018). Thus, to further define cardiac phenotypes in thebag3e2/e2 mutant fish, we analyzed contractile function at the single-myofibril level (Fig. 3F-K) (Dvornikov et al., 2014). The kinetics ofmyofibril activation were measured by the activation rate (KACT),which reflects myofibril shortening in response to a fast change incalcium concentration (pCa), and the rate of force redevelopment(KTR), which reflects a release–restretch maneuver (Fig. 3G). Themyofibril relaxation was measured by the duration of the slow linearphase of relaxation (TLIN) and the rate of fast exponential relaxation(KREL).We detected significantly reduced maximal isometric tensionand KTR in bag3e2/e2 mutant myofibrils (Fig. 3H,I), suggestinghypocontractility of myofibrils, a characteristic associated withsarcomeric DCMs (Chuan et al., 2012; Debold et al., 2007). Incontrast, the kinetics of relaxation was not affected (Fig. 3J,K). Next,

Fig. 1. Generation of zebrafishbag3 mutants using TALEN.(A) Schematics of the four bag3mutant alleles generated usingTALEN. The sequences in the 2ndexon that were targeted by theTALEN pairs are underlined. The PstIrestriction enzyme recognition site forgenotyping purposes is boxed.Dashed lines indicate deletednucleotides; nucleotides in blueindicate insertional mutations;nucleotides in red indicate mismatchmutations. Asterisk indicates an earlytranslational stop. fs, frameshift.(B) Chromographs illustrating thesequences of the bag3 wild-type(bag3-WT) and the 10-nucleotidedeletion mutant allele (bag3e2*-M1).Boxed sequence indicates therestriction enzyme PstI cutting sitesin WT that was deleted in the mutant.(C) Representative images andquantification analysis of body lengthand weight in the bag3 mutant andWT control at 6 months. Scale bar:1 cm. n=11 animals, one-wayANOVA. (D) Quantitative RT-PCRdemonstrated transcript reductions inboth the bag3 heterozygous(bag3e2/+) and homozygous(bag3e2/e2) mutants. n=3 biologicalreplicates. (E) Maximum swimmingspeed of the bag3e2/e2 mutantcompared with the WT control at6 months. n=7-8, mixed ANOVA.(F) Kaplan–Meier survival curves ofbag3 mutant fish and WT controls.n=26-27, log-rank test. Data aremean±s.e.m.

3

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

we measured the intracellular calcium transient in isolated ventricularcardiomyocytes loaded with the Indo-1AM radiometric calciumindicator. We found no significant changes in either the magnitude orkinetics of the Ca2+ transients in the bag3e2/e2 mutant heart ventricle(Fig. S9). Together, these comprehensive experiments defined DCM-like phenotypic traits in the bag3e2/e2 mutant.

Transcriptome analysis identifies mTOR as one of the topsignaling pathways altered in the bag3e2/e2 mutant heartTo elucidate the molecular basis of bag3 pathogenesis, we carried outtranscriptome analysis using 6-month-old fish hearts. Approximately

30 million reads per sample were obtained with three biologicalrepeats, with an average of >92% coverage of the whole annotatedzebrafish transcriptome (Zv9). Good reproducibility was observedamong independent sample replicates, with >95% Pearson correlationcoefficients (Fig. S10). Based on a cut-off of adjusted P<0.05, 5361genes were differentially expressed (DE) in the mutants comparedwith wild-type (WT) controls (Fig. 4A). A molecular signatureanalysis of the DE genes distinguished the bag3e2/e2 mutant heartsfrom the WT control (Fig. 4B). Ingenuity pathway analysis (IPA)ranked DE genes by signaling pathways and identified mitochondrialdysfunction and oxidative phosphorylation as the top two canonical

Fig. 2. The bag3e2/e2 mutant manifests a cardiac dysfunction phenotype. (A,B) Blood flow indices measured using a custom 30 MHz ultrasound probe.n=8-10, Student’s t-test. (C) Examples of echocardiography images extracted from movies of beating hearts in WT controls and bag3e2/e2 mutants atsystole (upper panel) and diastole (lower panel). (D) Quantification of cardiac function indices measured by echocardiography in the bag3e2/e2 mutant and WTcontrol at 6 months. n=8, Student’s t-test. (E-J) Cardiac pump function indices as measured by ex vivo heart pump function assay. n=12-13, Student’s t-test. Dataare mean±s.e.m.

4

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

pathways affected, confirming the heart failure phenotypes in thebag3e2/e2 mutant (Fig. 4C). Interestingly, we noted that the genes inthe mTOR pathway were ranked the fourth most DE genes. BecausemTOR signaling is vital for proteostasis, and mTOR is a knowntherapeutic target for several types of cardiomyopathies (Sciarrettaet al., 2018), we focused on mTOR signaling experimentally. Wedetected significant hyperphosphorylation of ribosomal S6 protein, akey downstream target of mTOR signaling, in the bag3e2/e2 mutanthearts (Fig. 4D) (Sciarretta et al., 2014). In addition, two negativedownstream effectors of mTOR signaling, phosphor-4E-BP1 andLC3-II, were significantly reduced after normalization by total

4E-BP1 or Actin, respectively, as controls. These experiments werecarried out in the presence of bafilomycin A1, a specific inhibitor ofthe autophagosome-lysosome, to reveal autophagic flux and staticautophagy. Together, our unbiased transcriptome analysis andsubsequent experimental validation suggested mTOR activation inthe bag3e2/e2 mutant.

mtorxu015+/− mitigates bag3 cardiomyopathyLast, we tested whether genetic inhibition ofmTOR is cardioprotectivefor bag3 cardiomyopathy. mtorxu015, a hypomorphic zebrafish mutant(Ding et al., 2011), was bred into the bag3e2/e2 mutant to generate

Fig. 3. bag3e2/e2 mutants manifest hallmarks of cardiomyopathy resembling DCM in mammals. (A) Representative images of isolated hearts andquantification of the ventricular surface area (VSA) normalized to body weight (BW) in the bag3e2/e2mutants andWT controls at 6 months. n=11, Student’s t-test.(B) TEM images confirmed the myofibril degeneration phenotype (yellow asterisks) and identified abnormal mitochondrial swelling (red arrows) in thebag3e2/e2 mutant fish heart at 6 months. (C) Representative images of H&E staining in the apex area and quantification of trabecular muscle density in thebag3e2/e2 mutants and WT controls at 6 months. n=6, Student’s t-test. (D) Quantitative RT-PCR analysis of cardiomyopathy molecular markers in bag3e2/e2

mutant hearts. n=3 biological replicates, Student’s t-test. (E) Representative images of the TUNEL assay and quantification of the percentage of TUNEL-positivenuclei (red arrows) in the bag3e2/e2 mutant and WT control at 6 months. n=4, Student’s t-test. (F) A representative image of a single myofibril isolated from thebag3e2/e2 mutant fish heart attached to glass microtools. (G) Example of a myofibril activation trace (pCa 10→4.5) with force-redevelopment during the release–restretch maneuver and relaxation when pCa was changed back from 4.5 to 10. Activation (pCa=4.5) and the fast release–restretch maneuver were used tomeasure kTR in myofibrils fromWT and the bag3e2/e2 mutant fish heart at 6 months. (H) Quantification of maximal isometric tension in activated single myofibrils.(I-K) Rates of force redevelopment (KTR) (I), fast relaxation (KREL) (J) and the time of the linear phase of relaxation (TLIN) (K) in the bag3e2/e2mutant andWT controlat the single-myofibril level. H-K, n=10-13, Student’s t-test. Data are mean±s.e.m. Scale bars: 1 mm in A; 2 µm in B; 100 µm in C; 20 µm in E; 50 µm in F.

5

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

bag3e2/+;mtorxu015+/− mutants, which were sequentially incrossed togenerate bag3e2/e2;mtorxu015+/− double mutants. By analyzing thesedouble mutants, compared with their siblings harboring either a singlemutant or WT, we found that partial mTOR inhibition through themtorxu015+/− mutant largely normalized the hyperphosphorylation ofribosomal S6, rescued the impaired autophagic flux as revealed by therestored LC3-II protein, and reduced the ubiquitinated proteinaggregations in the bag3e2/e2 mutant (Fig. 5A). The cardiomyopathyhallmarks present in the bag3e2/e2 mutant, including myofibril loss,myofibril degeneration and mitochondrial swelling, were all largelyrescued in the bag3e2/e2;mtorxu015+/− double mutants (Fig. 5B,C).Importantly, the introduction of the mtorxu015 allele into bag3e2/e2 fishsignificantly restored cardiac pump function, as measured using an ex

vivo assay (Fig. 5D). As a consequence, fish survival was significantlyimproved (Fig. 5E). Together, these data provide strong geneticevidence to support mtor as a candidate therapeutic target for bag3cardiomyopathy in zebrafish.

DISCUSSIONThe bag3e2/e2 mutant is likely a DCMmodel in adult zebrafishIn this study, we report the generation of bag3e2/e2 mutants in adultzebrafish to model BAG3 cardiomyopathy. The reading-frame-shiftgenotype of bag3e2/e2 mutants would most likely confer a loss offunction, recapitulating genetic lesions in mammals that lead to BAG3cardiomyopathy (Fang et al., 2017;Hommaet al., 2006).Thebag3e2/e2

mutant fish developed hallmarks of mammalian cardiomyopathy at

Fig. 4. Transcriptomeanalysis identifiesmTORasoneof the top signalingpathways affected in the bag3e2/e2mutant. (A)Numbers of differentially expressed(DE) genes between bag3e2/e2 mutant hearts andWT control with a cutoff of adjusted P<0.05. (B) Heat map presenting the expression of 5361 DE genes betweenthe bag3e2/e2 mutant and WT control. Gene expression levels are shown in log10 [reads per kilobase per million reads (RPKM)]. (C) The top 10 DE geneenriched pathways with significant P-values suggested by IPA. The red column shows that the mTOR pathway was one of the top pathways altered and wassubsequently experimentally tested. (D) Representative western blot images and quantification analysis of the expression levels of mTOR downstream proteins.LC3-II protein was examined in heart tissues from fish treated with 50 nM bafilomycin A1 for 4 h. n=4 biological replicates, Student’s t-test. Data are mean±s.e.m.

6

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

Fig. 5. Genetic testing reveals the therapeutic effects of mtor+/− on bag3 cardiomyopathy. (A) Western blot and quantification analysis of ribosomal S6protein (p-S6), LC3-II and ubiquitinated proteins in the bag3e2/e2;mtor+/− double mutants compared with their corresponding single mutants and theWT control at6 months. LC3-II protein was examined in heart tissues dissected from fish treated with 50 nM bafilomycin A1 for 4 h. n=4 biological replicates, one-way ANOVA.(B) Representative H&E staining images from the apex area and quantification of the trabecular muscle density from the bag3e2/e2;mtor+/− double mutants andtheir corresponding single mutants and WT controls at 6 months. n=4-10, one-way ANOVA. (C) TEM images and quantification analysis of bag3e2/e2;mtor+/−

mutants at 6 months compared with their siblings harboring either single mutants or the WT control. Yellow asterisks indicate regions of myofibril degeneration.Red arrows indicate mitochondria with abnormal swelling. (D) Percent EF and FS of bag3e2/e2;mtor+/− double mutant fish at 6 months compared with theircorresponding single mutants and the WT control. n=12-20, one-way ANOVA. (E) Kaplan–Meier survival curves of the bag3e2/e2;mtor+/− double mutantscompared with their corresponding single mutants and the WT control. n=18-44, log-rank test. Scale bars: 100 µm in B; 2 µm in C. Data are mean±s.e.m.

7

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

6 months, including reduced cardiac pump function, increasedventricular chamber size, myofibril degeneration, abnormalmitochondria, reduced exercise capacity and aberrant expression offetal genes that are molecularmarkers of cardiomyopathy (Harvey andLeinwand, 2011; Taegtmeyer et al., 2010). Despite the robustexpression of bag3 during the early embryonic stage, no evidentcardiac or somite phenotypes were detected, suggesting thatdevelopmental defects are less likely the root cause of cardiacphenotypes at 6 months. Similar to DCM in mammals, which ischaracterized by ventricular chamber enlargement, systolicdysfunction and myofibril hypocontractility (Merlo et al., 2018), thefollowing three phenotypic traits in the bag3e2/e2 mutants furthersuggest a DCM-like model in zebrafish. First, at the cardiac systolicfunction level, we detected reduced EF that was ascribed to increasedESV/BW, resembling the eccentric hypertrophy that is a characteristicof DCM inmammals (Japp et al., 2016;McNally andMestroni, 2017;Merlo et al., 2018). Second, at the single-myofibril level, we detecteddecreased maximal isometric tension and the reduced activation ofmyofibril kinetics, suggesting ‘hypocontractility’, another key featureof DCMhearts in mammals (Spudich et al., 2016). Third, we detectedventricular chamber enlargement, as indicated by the increased VSA/BW, which can serve as a surrogate DCM index for the unchangedEDV from our HFE analysis. Notably, both reduced ventricularsystolic function and enlarged ventricular chamber size have beenreported in the ttnatv/+ model (Huttner et al., 2018). Whether thettnatv/+model alsomanifests ‘hypocontractility’ at the single-myofibrillevel remains to be examined.Because human BAG3 cardiomyopathy is an autosomal dominant

disease, we acknowledge the limitation of using a homozygousbag3e2/e2 mutant to model this subtype DCM. As exemplified by arecent study of Bag3 knockouts in mice (Fang et al., 2017),homozygous Bag3 mutants can be a highly efficient surrogate formechanistic studies, as long as the key phenotypic traits can berecapitulated. As a future direction, much milder phenotypes in theaged bag3e2/+ heterozygous mutant will be analyzed in more detail,which might be a more faithful model recapitulating human BAG3cardiomyopathy. This effort is feasible in zebrafish because we wereable to detect cardiac phenotypes in bag3e2/+ heterozygous fish at12 months. Notably, caution must be taken when interpreting cardiacphenotypes in these compound knockout mutants because thenoncardiac expression of bag3, such as skeletal muscle, couldindirectly affect cardiac-related indices, such as reduced swimmingcapacity phenotype and increased mortality. We also acknowledgethe limitations of using zebrafish as a lower vertebrate tomodel DCM.For example, ventricular wall thickness, a major phenotypic trait thatdiscerns DCM from HCM, is difficult to measure in a highlytrabeculated zebrafish heart (Wang et al., 2011). We suggest that thereduced trabecular muscle density in the bag3e2/e2mutant heart mightserve as a surrogate index of thinner walls in mammals. In addition,certain phenotypes in mammalian Bag3 DCM models, such as areduced Ca2+ response (Feldman et al., 2016), cannot be recapitulatedin the zebrafish bag3 model. These intrinsic weaknesses need to beconsidered when developing therapies for human diseases using thisnew vertebrate model.

Integration of new phenotyping tools enables thediscernment of different subtypes of cardiomyopathies inadult zebrafishThe paucity of cardiac phenotyping tools in adult zebrafish with asmall heart size has been amajor hurdle that curtails the surgical poweroffered by this efficient vertebrate model. Here, we integrated newcardiac phenotyping tools developed recently and demonstrated that

enriched cardiac function details can be obtained, enabling thediscrimination of different DCM subtypes. For example, by employingultrasound-based Doppler imaging (Packard et al., 2017), we detectedprolonged IVCT and normal isovolumic relaxation time (IVRT) in ourbag3e2/e2 mutant. In contrast, prolonged IVRT was reported in thettnatv/+ model (Huttner et al., 2018), but IVCT was not reported. Wealso noted significantly reduced E, unchanged A, and as aconsequence, a reduced E/A ratio in our bag3e2/e2 mutant. Incontrast, only reduced A was noted in male fish of the ttnatv/+

model, whereas both the E and E/A ratios remained unchanged.Moreover, increased E and E/A ratios were reported in the DIC model(Packard et al., 2017), which is different from both the bag3e2/e2 andttnatv/+ models.

The improved phenotyping tools justify future efforts to comparemore cardiomyopathy models in zebrafish. Additional inheritedzebrafish models harboring a different DCM-causative gene and/ora known HCM-causative gene should be generated. Commonphenotypes between the bag3e2/e2 mutant and other DCM modelsand/or different phenotypes between DCM and HCM models willlead to optimized guidelines to define cardiomyopathies of differentetiologies. On the other hand, different phenotypes among differentsubtypes of DCM will justify this simple vertebrate model forindividualized medicine for cardiomyopathies.

The zebrafish bag3 cardiomyopathy model can be used tosearch for therapeutic targets, such as mtorThe improved phenotyping tools enabled us to determine the effects oftherapy on the bag3e2/e2 cardiomyopathy model more reliably.Through a combination of unbiased RNA-seq analysis with targetedgenetic validation, we found the activation of mTOR signaling in thebag3 mutant and showed the therapeutic capacity of the geneticinhibition of mTOR signaling. Mechanistically, we noted that bothimpaired autophagic flux and elevated ubiquitinated proteinaggregation in the bag3e2/e2 mutant hearts can be largely rescued bythe genetic inhibition ofmTOR, suggesting that the therapeutic effectsof mTOR inhibition are ascribed to repairing the primary damage ofprotein homeostasis caused by bag3 deficiency. Consistent with thishypothesis, a recent biochemical and cell culture study showed thatBag3 balances protein synthesis and degradation by spatiallyregulating mTOR complex 1 (mTORC1) via direct interaction withtuberous sclerosis protein 1 (TSC1) (Kathage et al., 2017).

On the other hand, it is also possible that mTOR inhibitionattenuates a common pathological event that is shared among differenttypes of cardiomyopathy, as suggested by its cardioprotective effectson several subtypes of cardiomyopathies in both zebrafish and mouse(Ding et al., 2011; Sciarretta et al., 2018). Consistent with thishypothesis, impaired cardiomyocyte autophagy was reported in ratmodels of Titin-truncating variants (TTNtv), which can be rescued byrapamycin, a specific pharmacologic inhibitor of mTOR signaling(Zhou et al., 2019). Dysregulated proteostasis might be part of thecommon pathological event, and the accumulation of toxic proteinsmight further drive the pathogenesis into irreversible heart failure.Further mechanistic studies and drug development efforts of mTOR-based therapy are needed, which could benefit a broad spectrum ofcardiomyopathies with different etiologies.

In summary, this study established the bag3e2/e2 mutant as aDCM model in adult zebrafish and demonstrated the feasibility ofidentifying therapeutic strategies, such as the genetic inhibition ofmTOR signaling. To fully leverage the potential of zebrafish as anew animal model, one future research direction is to conductmutagenesis screens for systematically identifying gene modifiersand therapeutic strategies. The feasibility of such a forward-genetic

8

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

strategy has been recently demonstrated for the DIC model (Dinget al., 2016). It will be interesting to explore whether the geneticmodifiers and related therapeutic targets identified from the DICmodel can be extrapolated to an inherited cardiomyopathy model,such as bag3e2/e2.

MATERIALS AND METHODSExperimental animalsZebrafish (Danio rerio; WIK strain) were maintained under a 14 h light/10 hdark cycle at 28.5°C and handled with care. All animal study procedureswere performed in accordance with the Guide for the Care and Use ofLaboratory Animals published by the US National Institutes of Health (NIHPublication No. 85-23, revised 1996). The anesthetic agent used in the studyis tricaine (0.02%) (Argent Chemical Laboratories), through incubation infish system water for 5-10 min. Overdose of a chemical was used as themeans of euthanasia for the zebrafish: 1% 3-aminobenzoic acid ethyl ester(Tricaine S) was added to the water as the final concentration. Animal studyprotocols were approved by the Mayo Clinic Institutional Animal Care andUse Committee (IACUC), protocol number A3531.

Generation of bag3 mutants via TALENTALEN techniques were employed to generate bag3mutants, according to ourpreviously published approaches (Shih et al., 2016). Briefly, TALEN primerpairs targeting the 2nd exon of the bag3 gene were designed using Zifit (http://zifit.partners.org/ZiFiT/ChoiceMenu.aspx). The TALEN left-arm bindingsequence 5′-TGTCATGAAAACCCTGAA-3′ and right-arm bindingsequence 5′-ATCCTAGTTTCTCCTACA-3′ were used. Both TALENs werethen assembled using a Golden Gate kit (Addgene) (Cermak et al., 2011).Capped mRNAs were synthesized using a mMESSAGE mMachine T3 kit(Ambion). Approximately 25 pg capped mRNA was injected into one-cell-stage embryos. Founder fish (F0) were raised to adulthood and outcrossed togenerate F1 embryos. Individual F1 embryos were used for genotyping PCR toidentify mutant alleles (forward primer: 5′-CGGCGTATAAAGAATTGCT-GG-3′; reverse primer: 5′-GTGAAGTAGGTGAGCAAGAC-3′). Theresulting PCR products were digested with the restriction enzyme PstI toidentify the WT or mutant genotype. The uncut PCR products were Sanger-sequenced to determine the precise genomic lesions. Four different bag3mutant alleles that presumably resulted in different shifts of the reading framefor each mutant locus were selected for continuous outcrosses up to the 5thgeneration and subsequent phenotypic analysis.

Cardiac functional phenotyping via Doppler imagingThe initial cardiac phenotypic analysis was performed on bag3 mutantsusing our reported Doppler imaging technique (Packard et al., 2017).Briefly, adult zebrafish at designated stages were anesthetized in tricaine(0.02%) for 10 min and placed ventral side facing upward. A 30 MHztransducer was placed ∼6 mm above the ventral side of the zebrafishvertically to acquire an ultrasound signal. Under the guidance of B-modeimaging, a Doppler gate (window) was positioned downstream from theatrioventricular (AV) valve in the ventricular inflow region to interrogateinflow velocities. The pulse repetition frequency (PRF) for pulsed-wave(PW) Doppler was set to 9.5 kHz, and the estimated Doppler angle was ∼0°because the blood flow of the zebrafish cardiac chambers is in the dorsal-ventral direction. PW Doppler signals were recorded in the control anddoxorubicin groups for∼3 s and stored for offline analysis usingMATLAB.To interrogate cardiac hemodynamics, we analyzed Pulse Wave Dopplersignals of passive (E wave velocity) and active (A wave velocity)ventricular filling during diastole. These measures – IVCT, ET across theventriculobulbar valve, and IVRT – determined the MPI as follows:MPI=(IVCT+IVRT)/ET. The MPI thus constitutes an integrated measure ofsystolic and diastolic function, with increases in MPI values indicating aworsening of cardiac function.

Cardiac functional phenotyping via the Vevo 3100echocardiography systemCardiac functional phenotypes were subsequently measured and analyzedusing the Vevo 3100 high-frequency imaging system equipped with a

50 MHz linear array transducer (Fujifilm VisualSonics). Acoustic gel(Aquasonic® 100, Parker Laboratories) was applied over the surface of thetransducer to provide adequate coupling with the tissue interface. Adultzebrafish at 6 months old were anesthetized in tricaine (0.02%) for 5 min,placed ventral side up, and held in place with a soft sponge stage. The50 MHz (MX700) transducer was placed above the zebrafish to provide asagittal imaging plane of the heart. B-mode images were acquired with animaging field of view of 9.00 mm in the axial direction and 5.73 mm in thelateral direction, a frame rate of 123 Hz, medium persistence and a transmitfocus at the center of the heart. Image quantification was performed usingthe VevoLABworkstation. Data were acquired and processed according to arecent report (Wang et al., 2017). Ventricular chamber dimensions weremeasured from B-mode images using the following two indices: FS=(EDD-ESD)/EDD and fractional area change (FAC)=(EDA-ESA)/EDA. EDD andESD are the perpendicular distances from the ventricular apex to theventricular basal line at the end-diastolic stage and end-systolic stage,respectively; EDA and ESA are defined as the areas of the ventricularchamber at the end-diastolic stage and end-systolic stage, respectively. Foreach index in individual fish, measurements were performed on three to fiveindependent cardiac cycles to acquire average values.

Cardiac functional phenotyping via an ex vivo assayCardiac functional phenotypes were also assessed using our recentlydeveloped ex vivo assay (Zhang et al., 2018). Briefly, at the designatedstages, adult zebrafish hearts were isolated, cannulated by 34-G ultrathincatheters through the atrioventricular canal under a stereo-microscope (LeicaM165C), perfused with Ca2+-containing Tyrode solution [132 mM NaCl,2.5 mM KCl, 4 mM NaHCO3, 0.33 mM NaH2PO4, 1 mM CaCl2, 1.6 mMMgCl2, 10 mM HEPES, 5 mM glucose and 5 mM sodium pyruvate(pH 7.5)] using a peristaltic pump EP-1 Econo Pump (Bio-Rad), and pacedusing a stimulator MyoPacer (Ionoptix; ∼15 V, 10 ms, 2 Hz). Movies ofpaced heart beats were recorded using a 14-megapixel Amscope MU1403camera (66 fps). Images in the perpendicular plane of the heart were alsoobtained using a 45-degree right angle aluminum mirror (Thorlabs). TheEDV and ESV were calculated using the biplane area-length formula: V= 2/3AAL×LAL, where AAL denotes the ventricle area in the transverse plane(short axis), LAL denotes the ventricle length in the longitudinal plane (longaxis), E=(EDV–ESV)/EDV. FS=(Ld–Ls)/Ld, where Ld and Ls denote theventricle lengths at diastole and systole, respectively. Three cardiac cycleswere analyzed and averaged to obtain each parameter. Image series wereanalyzed using ImageJ software (NIH) and a custom-written MATLABcode. The code performed edge detection of all four heart images (twohearts, two planes); then, radial strain was measured by averaging thenumber of ventricle radii (set by hand; from center of mass to periphery) on aframe-to-frame basis. In three cardiac cycles, we analyzed this radial strainand the maximal contraction and relaxation velocities (by differentiation ofthe deformation parameter and finding the extrema of the function). Weplotted velocity against strain, connecting points frame-to-frame to obtain avelocity-strain loop that was a good representation of cardiac contractility.

Single-myofibril assaysCardiac phenotypes were further defined at the single-myofibril level. Singlemyofibrils were prepared as previously described (de Tombe and Stienen,2007; Dvornikov et al., 2014; Walker et al., 2011). Briefly, hearts werepermeabilized (skinned) in 1% (v/v) Triton X-100 in a relaxing buffersupplemented with protease inhibitors at 4°C overnight, followed byhomogenization in ice-cold relaxing solution at 20,000 rpm for 10 s(homogenizer MDT500, 5 mm probe, MicroDisTec). The pellet was thenresuspended and placed into the tissue bath for subsequent mounting to fire-polished glass microtools in relaxing solution containing 10 mM EGTA. Theleft probewas stiff for fiber stretching; the right probewas compliant, serving asa force transducer. The tip of the right probe was ∼1 µm, and its stiffness wascalibrated to be 10-50 nN/µm. The attachedmyofibril was then superfused withalternating solution streams of relaxing and activating solutions (1 mMEGTA)emanating from a double-barreled pipette (∼180 µl/min) that wasmounted on atranslation stage capable of rapid (<5 ms) solution switches.

The activation of the myofibril led to its shortening and, therefore,displacement of the right probe. The positions of the left and right probes and

9

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

the length of the sarcomere of a singlemyofibril were detected optically with afast camera (Teledyne Dalsa Genie HM640) and HVSL software (AuroraScientific). The release–restretch maneuver (release is ∼20% of the fiberlength) was performed using a piezo controller (Thorlabs) driving the leftprobe. All motors, the perfusion system (Warner Instruments) and recordingswere fully controlled by custom-written Labview software (courtesy of Pieterde Tombe, Loyola University Chicago, IL, USA). A sarcomere length of2.0 µmwas used for force-pCa experiments. To generate the force-pCa curve,myofibrils were subjected to various pCa-activating buffers (in a range of 5.5-6). The passive stiffness test was performed by releasing the fiber at graduallyincreasing sarcomere/fiber lengths (SL) (from slack position to SL=2.4 µm).This approach allows the steady-state [Ca2+] saturated force and the rates ofCa2+-induced force activation and relaxation to be measured. Experimentswere performed at SL=2.3 µm. Rates of activation (kCa), rapid release–restretch force redevelopment (ktr), and biphasic relaxation (klin and kexp),together with the duration of the slow linear relaxation phase (Tlin), wereanalyzed by linear and exponential curve fitting using offline custom in-housewritten software (Labview). The compositions of the bath, relaxing andactivating solutions were as previously described (Dvornikov et al., 2014).The exclusion criterion for fibers was a release of the developed tension ofmore than 20% per set of contractions. All zebrafish single-myofibrilmechanics experiments were performed at 10°C.

Single-cardiomyocyte cell size measurementsCardiomyocytes from dissected ventricles of bag3e2/e2mutants andWT controlfish were dissociated as reported (Warren et al., 2001). Dissociatedcardiomyocytes were then resuspended in L-15 medium containing 10%fetal bovine serum (Invitrogen) and placed in Lab-Tek eight-well chambers(Thermo Fisher Scientific). Healthy dissociated cardiomyocytes were usuallyattached to the chamber within 1 h, and the attached cardiomyocytes were thencultured at 28.5°C for 12 h, followed by α-actinin antibody immunostaining toconfirm their cardiomyocyte identity (Yang and Xu, 2012a). Images ofα-actinin-stained cardiomyocytes were captured, and the cardiomyocyte areawas measured by outlining each individual cardiomyocyte using ImageJsoftware.

Measurement of ventricular surface-area-to-body weight indexOwing to the small size of an adult zebrafish heart, we previously calculatedVSA normalized by BW as an index to assess heart size in adult zebrafish(Ding et al., 2011). To measure VSA, individual zebrafish hearts weredissected and imaged next to a millimeter ruler under a Leica MZ FLI IIImicroscope. The largest projection of a ventricle was outlined using ImageJsoftware. To measure BW, fish were anesthetized in tricaine (0.02%)solution for 3 min, semidried on a paper towel and weighed on a scale. VSA/BW was then determined by the largest projection area of the ventricle (inmm2) divided by body weight (in g).

Swimming tunnel assayThe swimming tunnel assay was conducted using a swim tunnel respirometer(Mini Swim 170, Loligo Systems) to measure the swimming capacity of bag3mutants. This protocol was derived from previous reports with modifications(Sun et al., 2015; Wang et al., 2011). Briefly, bag3 heterozygous andhomozygous mutants were raised together with age-matched WT controls.All fish were fasted for 24 h before the first swimming capacity measurement.To evaluate swimming capacity, adult fish were transferred, four to eight fishper group, into the swim tunnel respirometer with an initial water speed of9 cm/s for a 20-min acclimation. Water flow was then increased in stages of8.66 cm/s (Ti) every 150 s (Tii) until all fishwere exhausted. The speeds at thelast stage (Uii) and the previous stage (Ui) were recorded for each individualfish. The critical swimming capacity (Ucrit) was calculated with the followingformula: Ucrit=Ui+[Uii×(Ti/Tii)]. Ucrit was then normalized to the bodylength (BL) of the corresponding individual. The same batches of fish weretested at 48 h and 96 h later for validation.

Western blottingWestern blotting was performed as described previously (Ding et al., 2016).Embryos from 6 days postfertilization (dpf ) or fresh hearts isolated from

adult fish were transferred immediately to RIPA buffer (Sigma-Aldrich)supplemented with complete protease inhibitor cocktail (Roche) andhomogenized using a Bullet Blender tissue homogenizer (Next Advance).The resultant protein lysates were subjected to western blotting using astandard protocol. The following primary antibodies were used: anti-actin(1:8000, Santa Cruz Biotechnology, sc-1615), anti-phospho-mTOR(Ser2448) (1:2000, Cell Signaling Technology, 2971), anti-phospho-S6ribosomal protein (Ser240/244) (1:5000, Cell Signaling Technology, 2215),anti-S6 ribosomal protein (1:8000, Cell Signaling Technology, 2217), anti-phospho-4E-BP1 (Thr37/46) (1:1000, Cell Signaling Technology, 2855),anti-4E-BP1 (1:2000, Cell Signaling Technology, 9644), anti-ubiquitin(1:1000, Thermo Fisher Scientific, PA5-17067) and anti-LC3 (1:3000,Novus Biologicals, NBP100-2331).

HistologyEmbryos at 6 dpf, heart and somite tissues were harvested from adult fish atthe designated stages after euthanization by incubation with 0.032% tricainefor 10 min. Embryonic somite or dissected tissues were immediately fixedin 4% PBS-buffered formaldehyde and sent to the Mayo Clinic HistologyCore Laboratory for subsequent sample processing and H&E staining.Images for somite tissues at 6 dpf, and heart tissues from the apex regionwere captured using the EVOS FL Auto Imaging System (Thermo FisherScientific). The density of the trabecular muscle was quantified usingImageJ software. For PCNA immunostaining, 8 µm cryostat-cut frozensections (Leica CM3050S) were subjected to immunostaining usingpreviously described methods (Yang and Xu, 2012b). The primary anti-PCNA antibody (1:3000, Sigma-Aldrich, p8825) was used. For the TUNELassay, cryostat-sectioned fish ventricles (8 μm) were stained with the In-SituCell Death Detection Kit, Fluorescein (Roche Applied Science) according tothe manufacturer’s protocol. Both the anti-PCNA antibody and TUNEL-stained tissue samples were mounted using Vectashield mounting mediumwith DAPI (Vector Laboratories, H-1200). All images were captured using aZeiss Axioplan II microscope equipped with ApoTome and AxioVisionsoftware (Carl Zeiss Microscopy).

Transmission electron microscopyFor TEM analysis, a single zebrafish adult heart was dissected and fixedimmediately in Trump’s fixative solution [4% paraformaldehyde and 1%glutaraldehyde in 0.1M phosphate buffer (pH 7.2)] at room temperature for1 h, followed by overnight incubation at 4°C. The fixed samples weresubsequently processed and imaged at the Mayo Clinic Electron MicroscopyCore Facility using a Philips CM10 transmission electron microscope.Quantification of the area of intact myofibril was largely performed based on apublished report (Huttner et al., 2018). Briefly, TEM images from 2500- to8000-fold magnification from three areas, including the apex, anterior wallnear the base/bulbo-aortic valve and posterior wall near the AV-valve, werecaptured. The area of intact myofibril from each image was outlined andexpressed to the total area of the image using ImageJ.

RNA-seq data collection and analysisTotal RNA was extracted from dissected ventricular tissue of 6-month-oldbag3e2/e2 mutants and WT siblings. Five ventricles were pooled as onesample. Six total samples (three biological replicates for each genotype)were sequenced using the HiSeq 2000 platform (Illumina) with a 50-bppaired-end sequencing protocol in the Mayo Clinic DNA Sequencing CoreFacility. Raw RNA-seq reads for each sample were aligned with TopHat(Version 2.0.12) (Kim et al., 2013) to the zebrafish genome assembly (Zv9)using the Ensembl annotation Zv9 (Danio_rerio.Zv9.79.gtf ). Each genewasassembled with Cufflinks (Version 2.2.1) (Trapnell et al., 2010). Genes wereconsidered to be differentially expressed across different groups if theyexhibited a greater than twofold change and a false discovery rate of lessthan 0.05 according to the Cuffdiff script from Cufflinks. Unsupervisedhierarchical clustering was performed with Pearson correlation and scaledbased on the fragments per kilobase of transcript per million mapped readsvalue using the pheatmap R package (https://github.com/raivokolde/pheatmap). The gene lists of interest were annotated by IPA (Qiagen)(http://www.ingenuity.com/). We queried the IPA with the gene list of

10

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

interest to map and generate putative biological processes/functions,networks and pathways based on the manually curated knowledgedatabase of molecular interactions extracted from the public literature. Theenriched pathways and gene networks were generated using both direct andindirect relationships/connectivity. These pathways and networks wereranked by their enrichment score, which measures the probability that thegenes were included in a network by chance. Both the primary RNA-seq rawand processed datasets have been deposited in GEO under accession numberGSE135823.

StatisticsSurvival curve, echocardiography and ex vivo heart function analyses arefrom cumulative data. The unpaired two-tailed Student’s t-test was used tocompare two groups. One-way or mixed ANOVA was used to assessdifferences among multiple groups, as appropriate. The log-rank test wasused to determine the difference in animal survival. All quantitative data arepresented as the mean±s.e.m. Sample size (n) represents animal number,otherwise specifically designated as biological replicates. P-values <0.05were considered to be significant. All statistical analyses were performedusing GraphPad Prism 7 and/or R Statistical Software Version 3.6.1. For thepost hoc analysis, we employed Tukey’s test to confirm our findings.

AcknowledgementsWe thank Kashia Stragey and Beninio Gore for maintaining the zebrafish facility atthe Mayo Clinic and for technical support in crossing the fish and performinggenotyping PCR.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: Y.D., X.X.; Methodology: Y.D., A.V.D., X.M., H.Z., Y.W., M.L.,R.R.P., L.W., J.C., Y.Z., X.X.; Software: L.W., J.C., Y.Z.; Validation: Y.D.; Formalanalysis: Y.D.; Investigation: Y.D., A.V.D., X.M., H.Z.; Data curation: Y.D.; Writing -original draft: Y.D., X.X.; Writing - review & editing: Y.D., X.L., X.X.; Supervision:Y.D., Y.Z., T.H., X.L., X.X.; Project administration: Y.D., X.X.; Funding acquisition:Y.D., X.X.

FundingThis work was supported in part by the Scientist Development Grant from theAmerican Heart Association (14SDG18160021) to Y.D., the Ted and Loretta RogersCardiovascular Career Development Award Honoring Hugh C. Smith (from theMayo Clinic) to Y.D., the National Institutes of Health (HL81753, HL107304,HL111437 and GM63904) to X.X., and the Mayo Foundation for Medical Educationand Research to X.X.

Data availabilityRaw and processed primary RNA-seq datasets have been deposited in GEO underaccession number GSE135823.

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.040154.supplemental

ReferencesArimura, T., Ishikawa, T., Nunoda, S., Kawai, S. and Kimura, A. (2011). Dilatedcardiomyopathy-associated BAG3 mutations impair Z-disc assembly andenhance sensitivity to apoptosis in cardiomyocytes. Hum. Mutat. 32,1481-1491. doi:10.1002/humu.21603

Cermak, T., Doyle, E. L., Christian, M., Wang, L., Zhang, Y., Schmidt, C., Baller,J. A., Somia, N. V., Bogdanove, A. J. and Voytas, D. F. (2011). Efficient designand assembly of custom TALEN and other TAL effector-based constructs for DNAtargeting. Nucleic Acids Res. 39, e82. doi:10.1093/nar/gkr218

Chami, N., Tadros, R., Lemarbre, F., Lo, K. S., Beaudoin, M., Robb, L., Labuda,D., Tardif, J. C., Racine, N., Talajic, M. et al. (2014). Nonsense mutations inBAG3 are associated with early-onset dilated cardiomyopathy in FrenchCanadians. Can. J. Cardiol. 30, 1655-1661. doi:10.1016/j.cjca.2014.09.030

Chuan, P., Sivaramakrishnan, S., Ashley, E. A. and Spudich, J. A. (2012). Cell-intrinsic functional effects of the alpha-cardiac myosin Arg-403-Gln mutation infamilial hypertrophic cardiomyopathy. Biophys. J. 102, 2782-2790. doi:10.1016/j.bpj.2012.04.049

Debold, E. P., Schmitt, J. P., Patlak, J. B., Beck, S. E., Moore, J. R., Seidman,J. G., Seidman, C. and Warshaw, D. M. (2007). Hypertrophic and dilated

cardiomyopathy mutations differentially affect the molecular force generation ofmouse alpha-cardiac myosin in the laser trap assay. Am. J. Physiol. Heart Circ.Physiol. 293, H284-H291. doi:10.1152/ajpheart.00128.2007

de Tombe, P. P. and Stienen, G. J. M. (2007). Impact of temperature on cross-bridge cycling kinetics in rat myocardium. J. Physiol. 584, 591-600. doi:10.1113/jphysiol.2007.138693

Ding, Y., Sun, X., Huang, W., Hoage, T., Redfield, M., Kushwaha, S., Sivasubbu,S., Lin, X., Ekker, S. and Xu, X. (2011). Haploinsufficiency of target of rapamycinattenuates cardiomyopathies in adult zebrafish. Circ. Res. 109, 658-669. doi:10.1161/CIRCRESAHA.111.248260

Ding, Y., Long, P. A., Bos, J. M., Shih, Y. H., Ma, X., Sundsbak, R. S., Chen, J.,Jiang, Y., Zhao, L., Hu, X. et al. (2016). A modifier screen identifies DNAJB6 as acardiomyopathy susceptibility gene. JCI Insight 1, e88797. doi:10.1172/jci.insight.88797

Dominguez, F., Cuenca, S., Bilinska, Z., Toro, R., Villard, E., Barriales-Villa, R.,Ochoa, J. P., Asselbergs, F., Sammani, A., Franaszczyk, M. et al. (2018).Dilated cardiomyopathy due to BLC2-associated athanogene 3 (BAG3)mutations. J. Am. Coll. Cardiol. 72, 2471-2481. doi:10.1016/j.jacc.2018.08.2181

Dvornikov, A. V., Dewan, S., Alekhina, O. V., Pickett, F. B. and de Tombe, P. P.(2014). Novel approaches to determine contractile function of the isolated adultzebrafish ventricular cardiac myocyte. J. Physiol. 592, 1949-1956. doi:10.1113/jphysiol.2014.270678

Fang, X., Bogomolovas, J., Wu, T., Zhang, W., Liu, C., Veevers, J., Stroud, M. J.,Zhang, Z., Ma, X., Mu, Y. et al. (2017). Loss-of-function mutations in co-chaperone BAG3 destabilize small HSPs and cause cardiomyopathy. J. Clin.Invest. 127, 3189-3200. doi:10.1172/JCI94310

Feldman, A. M., Gordon, J.,Wang, J., Song, J., Zhang, X.-Q., Myers, V. D., Tilley,D. G., Gao, E., Hoffman, N. E., Tomar, D. et al. (2016). BAG3 regulatescontractility and Ca(2+) homeostasis in adult mouse ventricular myocytes. J. Mol.Cell. Cardiol. 92, 10-20. doi:10.1016/j.yjmcc.2016.01.015

Franaszczyk, M., Bilinska, Z. T., Sobieszczanska-Małek, M., Michalak, E.,Sleszycka, J., Sioma, A., Małek, Ł. A., Kaczmarska, D., Walczak, E.,Włodarski, P. et al. (2014). The BAG3 gene variants in Polish patients withdilated cardiomyopathy: four novel mutations and a genotype-phenotypecorrelation. J. Transl. Med. 12, 192. doi:10.1186/1479-5876-12-192

Gut, P., Reischauer, S., Stainier, D. Y. R. and Arnaout, R. (2017). Little fish, bigdata: zebrafish as a model for cardiovascular and metabolic disease. Physiol.Rev. 97, 889-938. doi:10.1152/physrev.00038.2016

Harvey, P. A. and Leinwand, L. A. (2011). The cell biology of disease: cellularmechanisms of cardiomyopathy. J. Cell Biol. 194, 355-365. doi:10.1083/jcb.201101100

Henke, K., Daane, J. M., Hawkins, M. B., Dooley, C. M., Busch-Nentwich, E. M.,Stemple, D. L. and Harris, M. P. (2017). Genetic screen for postembryonicdevelopment in the zebrafish (Danio rerio): dominant mutations affecting adultform. Genetics 207, 609-623. doi:10.1534/genetics.117.300187

Herman, D. S., Lam, L., Taylor, M. R. G., Wang, L., Teekakirikul, P.,Christodoulou, D., Conner, L., DePalma, S. R., McDonough, B., Sparks, E.et al. (2012). Truncations of titin causing dilated cardiomyopathy. N. Engl. J. Med.366, 619-628. doi:10.1056/NEJMoa1110186

Hishiya, A., Kitazawa, T. and Takayama, S. (2010). BAG3 and Hsc70 interact withactin capping protein CapZ to maintain myofibrillar integrity under mechanicalstress. Circ. Res. 107, 1220-1231. doi:10.1161/CIRCRESAHA.110.225649

Homma, S., Iwasaki, M., Shelton, G. D., Engvall, E., Reed, J. C. and Takayama,S. (2006). BAG3 deficiency results in fulminant myopathy and early lethality.Am. J. Pathol. 169, 761-773. doi:10.2353/ajpath.2006.060250

Huttner, I. G., Wang, L. W., Santiago, C. F., Horvat, C., Johnson, R., Cheng, D.,von Frieling-Salewsky, M., Hillcoat, K., Bemand, T. J., Trivedi, G. et al. (2018).A-band titin truncation in zebrafish causes dilated cardiomyopathy andhemodynamic stress intolerance. Circ. Genom. Precis Med. 11, e002135.doi:10.1161/CIRCGEN.118.002135

Japp, A. G., Gulati, A., Cook, S. A., Cowie, M. R. and Prasad, S. K. (2016). Thediagnosis and evaluation of dilated cardiomyopathy. J. Am. Coll. Cardiol. 67,2996-3010. doi:10.1016/j.jacc.2016.03.590

Kathage, B., Gehlert, S., Ulbricht, A., Ludecke, L., Tapia, V. E., Orfanos, Z.,Wenzel, D., Bloch, W., Volkmer, R., Fleischmann, B. K. et al. (2017). Thecochaperone BAG3 coordinates protein synthesis and autophagy undermechanical strain through spatial regulation of mTORC1. Biochim. Biophys.Acta 1864, 62-75. doi:10.1016/j.bbamcr.2016.10.007

Kim, D., Pertea, G., Trapnell, C., Pimentel, H., Kelley, R. and Salzberg, S. L.(2013). TopHat2: accurate alignment of transcriptomes in the presence ofinsertions, deletions and gene fusions. Genome Biol. 14, R36. doi:10.1186/gb-2013-14-4-r36

Lee, J.-H., Takahashi, T., Yasuhara, N., Inazawa, J., Kamada, S. and Tsujimoto,Y. (1999). Bis, a Bcl-2-binding protein that synergizes with Bcl-2 in preventing celldeath. Oncogene 18, 6183-6190. doi:10.1038/sj.onc.1203043

MacRae, C. A. and Peterson, R. T. (2015). Zebrafish as tools for drug discovery.Nat. Rev. Drug Discov. 14, 721-731. doi:10.1038/nrd4627

Maillet, M., van Berlo, J. H. and Molkentin, J. D. (2013). Molecular basis ofphysiological heart growth: fundamental concepts and new players.Nat. Rev. Mol.Cell Biol. 14, 38-48. doi:10.1038/nrm3495

11

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms

Marin, T. M., Keith, K., Davies, B., Conner, D. A., Guha, P., Kalaitzidis, D., Wu, X.,Lauriol, J., Wang, B., Bauer, M. et al. (2011). Rapamycin reverses hypertrophiccardiomyopathy in a mouse model of LEOPARD syndrome-associated PTPN11mutation. J. Clin. Invest. 121, 1026-1043. doi:10.1172/JCI44972

McMullen, J. R., Sherwood, M. C., Tarnavski, O., Zhang, L., Dorfman, A. L.,Shioi, T. and Izumo, S. (2004). Inhibition of mTOR signaling with rapamycinregresses established cardiac hypertrophy induced by pressure overload.Circulation 109, 3050-3055. doi:10.1161/01.CIR.0000130641.08705.45

McNally, E. M. and Mestroni, L. (2017). Dilated cardiomyopathy: geneticdeterminants and mechanisms. Circ. Res. 121, 731-748. doi:10.1161/CIRCRESAHA.116.309396

Meriin, A. B., Narayanan, A., Meng, L., Alexandrov, I., Varelas, X., Cisse, I. I. andSherman, M. Y. (2018). Hsp70-Bag3 complex is a hub for proteotoxicity-inducedsignaling that controls protein aggregation. Proc. Natl. Acad. Sci. USA 115,E7043-E7052. doi:10.1073/pnas.1803130115

Merlo, M., Cannata, A., Gobbo, M., Stolfo, D., Elliott, P. M. and Sinagra, G.(2018). Evolving concepts in dilated cardiomyopathy. Eur. J. Heart Fail 20,228-239. doi:10.1002/ejhf.1103

Myers, V. D., McClung, J. M., Wang, J., Tahrir, F. G., Gupta, M. K., Gordon, J.,Kontos, C. H., Khalili, K., Cheung, J. Y. and Feldman, A. M. (2018). Themultifunctional protein BAG3: a novel therapeutic target in cardiovasculardisease. JACC Basic Transl. Sci. 3, 122-131. doi:10.1016/j.jacbts.2017.09.009

Norton, N., Li, D., Rieder, M. J., Siegfried, J. D., Rampersaud, E., Zuchner, S.,Mangos, S., Gonzalez-Quintana, J., Wang, L., McGee, S. et al. (2011).Genome-wide studies of copy number variation and exome sequencing identifyrare variants in BAG3 as a cause of dilated cardiomyopathy. Am. J. Hum. Genet.88, 273-282. doi:10.1016/j.ajhg.2011.01.016

Packard, R. R. S., Baek, K. I., Beebe, T., Jen, N., Ding, Y., Shi, F., Fei, P., Kang,B. J., Chen, P.-H., Gau, J. et al. (2017). Automated segmentation of light-sheetfluorescent imaging to characterize experimental doxorubicin-induced cardiacinjury and repair. Sci. Rep. 7, 8603. doi:10.1038/s41598-017-09152-x

Ramos, F. J., Chen, S. C., Garelick, M. G., Dai, D. F., Liao, C. Y., Schreiber, K. H.,MacKay, V. L., An, E. H., Strong, R., Ladiges, W. C. et al. (2012). Rapamycinreverses elevated mTORC1 signaling in lamin A/C-deficient mice, rescuescardiac and skeletal muscle function, and extends survival. Sci. Transl. Med. 4,144ra103. doi:10.1126/scitranslmed.3003802

Saxton, R. A. and Sabatini, D. M. (2017). mTOR signaling in growth, metabolism,and disease. Cell 169, 361-371. doi:10.1016/j.cell.2017.03.035

Sciarretta, S., Volpe, M. and Sadoshima, J. (2014). Mammalian target ofrapamycin signaling in cardiac physiology and disease. Circ. Res. 114,549-564. doi:10.1161/CIRCRESAHA.114.302022

Sciarretta, S., Forte, M., Frati, G. and Sadoshima, J. (2018). New insights into therole of mTOR signaling in the cardiovascular system. Circ. Res. 122, 489-505.doi:10.1161/CIRCRESAHA.117.311147

Selcen, D., Muntoni, F., Burton, B. K., Pegoraro, E., Sewry, C., Bite, A. V. andEngel, A. G. (2009). Mutation in BAG3 causes severe dominant childhoodmuscular dystrophy. Ann. Neurol. 65, 83-89. doi:10.1002/ana.21553

Shih, Y.-H., Zhang, Y., Ding, Y., Ross, C. A., Li, H., Olson, T. M. and Xu, X. (2015).Cardiac transcriptome and dilated cardiomyopathy genes in zebrafish. Circ.Cardiovasc. Genet. 8, 261-269. doi:10.1161/CIRCGENETICS.114.000702

Shih, Y.-H., Dvornikov, A. V., Zhu, P., Ma, X., Kim, M., Ding, Y. and Xu, X. (2016).Exon- and contraction-dependent functions of titin in sarcomere assembly.Development 143, 4713-4722. doi:10.1242/dev.139246

Song, X., Kusakari, Y., Xiao, C.-Y., Kinsella, S. D., Rosenberg, M. A., Scherrer-Crosbie, M., Hara, K., Rosenzweig, A. and Matsui, T. (2010). mTOR attenuatesthe inflammatory response in cardiomyocytes and prevents cardiac dysfunction in

pathological hypertrophy.Am. J. Physiol. Cell Physiol. 299, C1256-C1266. doi:10.1152/ajpcell.00338.2010

Spudich, J. A., Aksel, T., Bartholomew, S. R., Nag, S., Kawana, M., Yu, E. C.,Sarkar, S. S., Sung, J., Sommese, R. F., Sutton, S. et al. (2016). Effects ofhypertrophic and dilated cardiomyopathy mutations on power output by humanbeta-cardiac myosin. J. Exp. Biol. 219, 161-167. doi:10.1242/jeb.125930

Sturner, E. and Behl, C. (2017). The role of the multifunctional BAG3 protein incellular protein quality control and in disease. Front. Mol. Neurosci. 10, 177.doi:10.3389/fnmol.2017.00177

Sun, Y., Fang, Y., Xu, X., Lu, G. and Chen, Z. (2015). Evidence of an associationbetween age-related functional modifications and pathophysiological changes inzebrafish heart. Gerontology 61, 435-447. doi:10.1159/000369094

Taegtmeyer, H., Sen, S. and Vela, D. (2010). Return to the fetal gene program: asuggested metabolic link to gene expression in the heart. Ann. N. Y. Acad. Sci.1188, 191-198. doi:10.1111/j.1749-6632.2009.05100.x

Trapnell, C., Williams, B. A., Pertea, G., Mortazavi, A., Kwan, G., van Baren,M. J., Salzberg, S. L., Wold, B. J. and Pachter, L. (2010). Transcript assemblyand quantification by RNA-Seq reveals unannotated transcripts and isoformswitching during cell differentiation. Nat. Biotechnol. 28, 511-515. doi:10.1038/nbt.1621

Trivedi, D. V., Adhikari, A. S., Sarkar, S. S., Ruppel, K. M. and Spudich, J. A.(2018). Hypertrophic cardiomyopathy and the myosin mesa: viewing an olddisease in a new light. Biophys. Rev. 10, 27-48. doi:10.1007/s12551-017-0274-6

Villard, E., Perret, C., Gary, F., Proust, C., Dilanian, G., Hengstenberg, C.,Ruppert, V., Arbustini, E., Wichter, T., Germain, M. et al. (2011). A genome-wide association study identifies two loci associated with heart failure due todilated cardiomyopathy. Eur. Heart J. 32, 1065-1076. doi:10.1093/eurheartj/ehr105

Walker, J. S., Walker, L. A., Margulies, K., Buttrick, P. and de Tombe, P. (2011).Protein kinase A changes calcium sensitivity but not crossbridge kinetics inhuman cardiac myofibrils. Am. J. Physiol. Heart Circ. Physiol. 301, H138-H146.doi:10.1152/ajpheart.00838.2010

Wang, J., Panakova, D., Kikuchi, K., Holdway, J. E., Gemberling, M., Burris,J. S., Singh, S. P., Dickson, A. L., Lin, Y.-F., Sabeh, M. K. et al. (2011). Theregenerative capacity of zebrafish reverses cardiac failure caused by geneticcardiomyocyte depletion. Development 138, 3421-3430. doi:10.1242/dev.068601

Wang, L. W., Huttner, I. G., Santiago, C. F., Kesteven, S. H., Yu, Z.-Y., Feneley,M. P. and Fatkin, D. (2017). Standardized echocardiographic assessment ofcardiac function in normal adult zebrafish and heart disease models. Dis. Model.Mech. 10, 63-76. doi:10.1242/dmm.026989

Warren, K. S., Baker, K. and Fishman, M. C. (2001). The slow mo mutationreduces pacemaker current and heart rate in adult zebrafish.Am. J. Physiol. HeartCirc. Physiol. 281, H1711-H1719. doi:10.1152/ajpheart.2001.281.4.H1711

Yang, J. and Xu, X. (2012a). alpha-Actinin2 is required for the lateral alignment of Zdiscs and ventricular chamber enlargement during zebrafish cardiogenesis.FASEB J. 26, 4230-4242. doi:10.1096/fj.12-207969

Yang, J. and Xu, X. (2012b). Immunostaining of dissected zebrafish embryonicheart. J. Vis. Exp. 59, e3510. doi:10.3791/3510

Zhang, H., Dvornikov, A. V., Huttner, I. G., Ma, X., Santiago, C. F., Fatkin, D. andXu, X. (2018). A Langendorff-like system to quantify cardiac pump function in adultzebrafish. Dis. Model. Mech. 11, dmm034819. doi:10.1242/dmm.034819

Zhou, J., Ng, B., Ko, N. S. J., Fiedler, L. R., Khin, E., Lim, A., Sahib, N. E., Wu, Y.,Chothani, S. P., Schafer, S. et al. (2019). Titin truncations lead to impairedcardiomyocyte autophagy and mitochondrial function in vivo. Hum. Mol. Genet.28, 1971-1981. doi:10.1093/hmg/ddz033

12

RESEARCH ARTICLE Disease Models & Mechanisms (2019) 12, dmm040154. doi:10.1242/dmm.040154

Disea

seModels&Mechan

isms