Genetic compensation prevents myopathy and heart failure in an in vivo model of Bag3 deficiency

Mutations in the molecular co-chaperone Bcl2-associated athanogene 3 (BAG3) are found to cause dilated cardiomyopathy (DCM), resulting in systolic dysfunction and heart failure, as well as myofibrillar myopathy (MFM), which is characterized by protein aggregation and myofibrillar disintegration in skeletal muscle cells. Here, we generated a CRISPR/Cas9-induced Bag3 knockout zebrafish line and found the complete preservation of heart and skeletal muscle structure and function during embryonic development, in contrast to morpholino-mediated knockdown of Bag3. Intriguingly, genetic compensation, a process of transcriptional adaptation which acts independent of protein feedback loops, was found to prevent heart and skeletal muscle damage in our Bag3 knockout model. Proteomic profiling and quantitative real-time PCR analyses identified Bag2, another member of the Bag protein family, significantly upregulated on a transcript and protein level in bag3-/- mutants. This implied that the decay of bag3 mutant mRNA in homozygous bag3-/- embryos caused the transcriptional upregulation of bag2 expression. We further demonstrated that morpholino-mediated knockdown of Bag2 in bag3-/- embryos evoked severe functional and structural heart and skeletal muscle defects, which are similar to Bag3 morphants. However, Bag2 knockdown in bag3+/+ or bag3+/- embryos did not result in (cardio-)myopathy. Finally, we found that inhibition of the nonsense-mediated mRNA decay (NMD) machinery by knockdown of upf1, an essential NMD factor, caused severe heart and skeletal muscle defects in bag3-/- mutants due to the blockade of transcriptional adaptation of bag2 expression. Our findings provide evidence that genetic compensation might vitally influence the penetrance of disease-causing bag3 mutations in vivo.

Introduction Compensatory transcriptional adaptation of gene expression in response to malignant and harmful gene mutations is a powerful mechanism to warrant genetic robustness. Interestingly, this genetic compensation was recently demonstrated to be triggered by a decay of the mutated mRNA which results in the transcriptional upregulation of one or more related and compensation-competent genes [1][2][3]. Recently, Rossi and coworkers found that CRISPR/Cas9-induced mutation of egfl7, an endothelial ECM gene, did not result in obvious defects in zebrafish. Whereas morpholino-modified antisense oligonucleotide (morpholino, MO)-mediated egfl7 depletion led to severe vascular malformations [3]. Interestingly, Emilins, also important ECM genes, were found to be transcriptionally upregulated in egfl7 mutant zebrafish, but not in morphant embryos, suggesting their compensatory potential to buffer the organism to egfl7 loss. Additionally, Sztal et al. recently found genetic compensation in a zebrafish model of Actin deficiency [4]. Genetic actc1b mutant zebrafish embryos only show very mild muscle defects due to transcriptional upregulation and functional compensation by an α-Actin paralogue. Interestingly, morpholino-mediated actc1b depletion did not result in compensatory upregulation of other Actin family members thereby leading to severe myopathy in vivo [4].
B-cell lymphoma 2 (Bcl2)-associated athanogene 3 (BAG3) is a member of the BAG protein family which acts as a molecular co-chaperone by physically interacting with chaperone molecules such as 14-3-3 proteins, Hsp70, and small heat shock proteins (HSPBs) [5]. One of the main biological functions of BAG3 is the regulation of autophagy and the degradation of misfolded proteins to warrant orchestrated protein homeostasis [5,6]. Mammalian BAG3 is strongly expressed in the heart and skeletal muscle, where it mainly localizes to sarcomeric Z-disks, but is also expressed in the brain and peripheral nervous system [7]. Mutations in the Bag3 gene are described to cause a diverse spectrum of disease phenotypes including striated muscle diseases such as dilated cardiomyopathy (DCM) or myofibrillar myopathy (MFM). Furthermore, decreased BAG3 protein levels were found in failing human hearts [8].
In this study, we uncovered and explored an example of functional genetic compensation in a CRISPR/Cas9-induced zebrafish model of Bag3-deficiency. We and others found that targeted morpholino-mediated knockdown of Bag3 in zebrafish embryos resulted in severe structural and functional heart and skeletal muscle defects [9][10][11]. Interestingly and in contrast to these studies, we found here that the CRISPR/Cas9-induced deletion of 19 nucleotides in exon 2 of the bag3 gene leading to a frame shift and thereby introducing a premature stop codon did not evoke heart and skeletal muscle defects in homozygous mutant embryos. We determined that the degradation of mutated bag3 mRNA triggered the compensatory transcriptional upregulation of the BAG protein family member bag2, which was able to functionally compensate for the loss of Bag3.
Our findings present another example of genetic compensation in vivo, underlining the importance of this exciting biological mechanism in buffering the malignancy of specific gene mutations and, thereby, mediating genetic robustness.

Targeted knockout of zebrafish bag3 by CRISPR/Cas9 does not provoke heart and skeletal muscle dysfunction
In humans, mutations in BAG3 (Bcl2-associated athanogene 3) were found to cause dilated cardiomyopathy (DCM) and myofibrillar myopathies (MFM) [9,12]; however, the molecular BAG3-associated underpinnings are poorly understood. To study the in vivo roles of BAG3 in more detail, we generated a Bag3-deficient zebrafish line by using the CRISPR/Cas9 genomeediting technology.
Although teleosts went through whole genome duplication during evolution [13,14], only one single bag3 ortholog of the human BAG3 gene can be found on zebrafish chromosome 13. Zebrafish Bag3 shows 41% overall amino acid homology to human BAG3, with up to 77% sequence identity in distinct functional domains such as the WW domain (S1 Fig). As in humans, zebrafish bag3 transcripts are enriched in the heart and skeletal muscle during embryogenesis and are more predominantly expressed in the adult cardiac muscle [10,15].
To generate a loss-of-function zebrafish model by introducing CRISPR/Cas9-induced mutations in the bag3 gene, we first designed a CRISPR-RNA (crRNA) targeting exon 2 of the bag3 genomic sequence. By co-injecting synthetic tracrRNA, bag3-specific crRNA, and Cas9 protein, a frameshift mutation in the bag3 gene was generated, leading to an alternative stop codon and premature termination of protein translation. Specifically, a mutant allele with a 19-nucleotide deletion was identified among F1 fish and the mutant line was propagated ( Fig  1A). The putative mutation leads to the loss of the crucial Bag3 protein domains IPV and PXXP as well as the BAG domain ( Fig 1A). To confirm Bag3 deficiency, we performed immunoblot assays and found the complete loss of wild-type Bag3 proteins in homozygous mutant bag3 -/embryos (N = 3, P<0.0001) (Fig 1B and 1C). We also performed quantitative RT-PCR analysis using homozygous mutant bag3 -/embryos and found bag3 mRNA levels significantly downregulated (N = 3, P<0.0001) (Fig 1D). This suggested that nonsense-mediated mRNA decay (NMD) was triggered by the CRISPR/Cas9-induced gene mutation.
Since Bag3 loss did not result in phenotypic alterations during development under normal physiological conditions, we aimed to mechanically overload the skeletal muscle in Bag3-deficient zebrafish embryos. Mutant bag3 -/embryos were incubated in 1% methylcellulose, leading to increased viscosity of the media, which is known to significantly enhance workload on skeletal muscle [10]. Interestingly, increased muscle workload resulted in no significant elevation of myopathic phenotypes in bag3 -/embryos compared to bag3 +/and bag3 +/+ clutchmates (N = 4, n = 50, bag3 +/+ and bag3 +/-P = 0.6050; bag3 +/+ and bag3 -/-P = 0.8320; bag3 +/and bag3 -/-P = 0.8582). (Fig 2H). Similar to our observations on unstressed bag3 mutant embryos, methylcellulose-incubated bag3 +/+ , bag3 +/-, and bag3 -/embryos did not display structural failure of myofibers following treatment ( Fig 2J). This demonstrates that loss of Bag3 in our CRISPR/Cas9-induced zebrafish line does not cause cardiomyopathy and skeletal muscle dysfunction under physiological or stressed conditions. Additionally, we found that homozygous mutant bag3 -/fish can be raised to adulthood and they are fertile. However, we found a significant increase in the mortality rate of bag3 -/zebrafish starting from 9 month of age as depicted in a Kaplan-Meier survival curve (S2 Fig), which is consistent with the recently published findings by Ding et al [17].

Morpholino-mediated ablation of Bag3 specifically results in (cardio)myopathy
Recently, different morpholino-mediated knockdown studies in zebrafish showed that loss of Bag3 led to severe heart and skeletal muscle defects and (cardio-)myopathy [9,10,18]. Similarly, we found severe cardiac and skeletal muscle defects (Fig 3A and 3B) associated with myofibrillar disorganization (Fig 3A-3C), diminished cardiac contractile function (Fig 3D and 3E; S9 Movie), and significantly reduced motility ( Fig 3F) in zebrafish embryos injected with a morpholino targeting the splice donor site of exon 2 in bag3 [9]. By contrast, injection of a specific 5bp mismatch control morpholino (MO-bag3 5bp mm) unable to target bag3 did not evoke any cardiac and skeletal muscle defects (Fig 3A-3F).

Bag2 mediated genetic compensation triggered by CRISPR/Cas9-induced Bag3 deletion prevents heart and skeletal muscle dysfunction
Next, to assess whether genetic compensation is activated in our CRISPR/Cas9-induced bag3 mutants, we performed a mass spectrometry (MS)-based proteomic analysis on muscle tissue derived from adult homozygous bag3 -/mutants compared to homozygous wild-type bag3 +/+ clutchmates ( Fig 5A). The obtained proteomics profiles revealed 68 down-and 126 up-regulated proteins (log fold change >2 and <-2) (Fig 5B; S2 Table). As expected, Bag3 was not detected in homozygous bag3 -/mutant muscle tissue (Fig 5B), confirming Bag3 deficiency in muscle tissue of our CRISPR/Cas9-induced mutant line. Interestingly, among the most upregulated proteins, we found levels of Bag2, a member of the Bag protein family, significantly elevated (Log Fold Change = 2.2). This suggested Bag2 could be a potential functional compensator for the loss of Bag3.
Bag2 (Bcl2-associated athanogene 2) is a molecular co-chaperone and a member of the Bag protein family that shows high structural similarity to Bag3. In zebrafish, Bag2 was found to be expressed in the heart and skeletal muscle [19]. To first investigate the role of Bag2 in vivo, we inactivated zebrafish bag2 by injecting a morpholino against the splice donor site of exon 2 (MO-bag2) into one-cell stage zebrafish embryos. When injected with either MO-bag2 or the 5bp mismatch control morpholino (MO-bag2 5bp mm), injected embryos (N = 3, n = 30/50, P<0.5010) did not develop (cardio-)myopathy (Fig 6A-6F)). We found normal cardiac and skeletal muscle structure and function (Fig 6A and 6B), unconfined cardiac contractile function (Fig 6D and 6E), and undisturbed motility (Fig 6F) in bag2 morphant embryos. This demonstrates that Bag2 is dispensable for early heart and skeletal muscle development and function.

Blockade of bag3 mRNA decay inhibits transcriptional adaptation and induces cardiomyopathy and skeletal muscle dysfunction in Bag3-deficient zebrafish
To investigate whether the inhibition of nonsense-mediated decay (NMD) of mRNA and the subsequent blockade of the transcriptional adaptation machinery [1][2][3] can induce heart and skeletal muscle defects in bag3 -/embryos, we inactivated upf1, a factor known to be crucial for regular NMD function [20,21]. We found that morpholino-mediated knockdown of upf1 (MO-upf1) in bag3 +/+ and bag3 +/embryos did not result in heart and skeletal muscle defects, whereas knockdown of upf1 in bag3 -/embryos led to severe structural and functional heart and skeletal muscle defects (Fig 8A-8F). Muscle structure was disrupted in bag3 -/embryos injected with MO-upf1 (Fig 8C), while sarcomeric organization in MO-upf1-injected bag3 +/+ and bag3 +/embryos was preserved (Fig 8C). Furthermore, MO-upf1 injection into bag3 -/-   embryos caused severe functional heart and skeletal muscle defects, however, MO-upf1injected bag3 +/+ and bag3 +/embryos showed no phenotypic alterations (Fig 8D-8F; S6, S7 and S12 Movies). High efficiency of the morpholino-mediated upf1 knockdown was confirmed by RT-PCR-based mRNA splicing assays (S6A Fig). Finally, to assess whether inhibiting NMD in bag3 -/embryos indeed resulted in the preservation of bag3 mRNA levels and reduction of bag2 mRNA levels, we performed qRT-PCR analyses in MO-upf1-injected bag3 -/embryos. Indeed, we found the preservation of mutant bag3 mRNA levels as well as the diminution of bag2 mRNA levels to physiological levels in bag3 -/embryos injected with MO-upf1, (Fig 8G and 8H), indicating that blocking the NMD of bag3 mRNA effectively inhibits the transcriptional adaptation of bag2 expression.
In summary, our findings demonstrate that transcriptional adaptation of bag2 expression in a model of CRISPR/Cas9-induced Bag3 ablation is involved in the process of genetic compensation to preserve heart and skeletal muscle structure and function in vivo.

Discussion
Genetic compensation by transcriptional adaptation was recently described to contribute to genetic robustness that guarantees viability and fitness of an organism in the presence of malignant and harmful gene variations and mutations [1,2]. Here, we describe compensatory transcriptional adaptation in a CRISPR/Cas9-induced zebrafish model of Bag3-deficiency. Whereas morpholino-mediated ablation of Bag3 in zebrafish embryos led to severe heart failure and myopathic phenotypes, the knockout of Bag3 by CRISPR/Cas9 technology was not accompanied by phenotypic alterations in the developing embryo. We found that in our CRISPR/Cas9-induced bag3 mutant embryos, the expression of the Bag protein family member Bag2 was significantly upregulated.
El-Brolosy et al. proposed an explanatory model for the discrepancy in transcriptional adaptation in mutants and morphants. It was found that degradation of the mutated mRNA by mRNA degradation pathways is the molecular trigger for the upregulation of genes that exhibit sequence similarity thereby enabling functional compensation [1,22,23]. Pharmacological but also genetic inhibition of NMD (e.g. by the inactivation of Upf1, a factor pivotal for NMD) led to a significant reduction of mutated mRNA and the loss compensatory transcriptional adaptation [2,23,24].
https://doi.org/10.1371/journal.pgen.1009088.g008 these results and introduced BAG3 as a major DCM causing gene with an incidence of up to 6.7% of all DCM cases [17,25,26]. To date, several different human BAG3 mutations are described to be associated with DCM. Remarkably, DCM-associated BAG3 mutations are frequently found to result in the truncation of the BAG3 protein and are assumed to cause the complete loss of BAG3 function [25,27]. These findings are supported by analyses in conditional, cardiomyocyte-specific Bag3 knockout mice that also develop DCM shortly after birth [28]. To the best of our knowledge, DCM-associated truncating human BAG3 mutations, predominately frame shift and nonsense mutations [25], were never analyzed for the induction of the NMD of the mutated bag3 mRNA and thereby the activation of putative compensatory mechanisms.
Global disruption of the Bag3 gene in homozygous mice led to lethality after 4 weeks of age. These BAG3 knockout mice were characterized by severe myopathic phenotypes including disruption of Z-disk architecture and myofibrillar degeneration [15]. Cardiac-specific knockout of BAG3 by the use of cardiomyocyte-specific α-myosin heavy chain Cre-transgenic mice (αMHC-Cre mice) significantly impaired structure and function of the contractile apparatus in cardiomyocytes resulting in DCM and premature death of these conditional Bag3 knockout mice during their first year of life [28]. Fang and co-workers did not observe increased expression of other BAG family members such as Bag 1 or Bag2 in their conditional Bag3 knockout model [28], implying that their knockout strategy might not activate NMD-related genetic compensation. Knockdown of Bag3 in the zebrafish model using morpholino-modified antisense oligonucleotides resulted in the development of cardiac contractile dysfunction and cardiomyopathy [9] as well as a contraction-dependent myofibrillar disintegration and skeletal muscle dysfunction, two of the characteristics of BAG3-related myofibrillar myopathies (MFM) [10,11]. Very recently, Ding et al. described that TALEN-mediated ablation of Bag3 led to cardiomyopathy in adult zebrafish most likely due to interference with protein degradation and homeostasis [17].
Here, we describe a CRISPR/Cas9-induced zebrafish model of Bag3 deficiency and we show that bag3 mutant zebrafish embryos, in contrast to bag3 morphants, do not develop heart and skeletal muscle defects. Expression of Rpl10, the 60S ribosomal protein L10 and Bag2 were found to be strongly induced in bag3 -/embryos. Loss of Rpl10 was shown to cause neurodevelopmental defects in zebrafish [29], but does not show any sequence or domain similarities to Bag3 making Rpl10 unlikely to functionally compensate for the loss of Bag3. By contrast, Bag2, a member of the Bag protein family, shows high sequence and domain similarities to Bag3 turning Bag2 into a perfect candidate to compensate for the loss of Bag3 function. In humans, BAG2 is expressed in several tissues, including brown adipose and the lung but also the heart [30]. Unfortunately, not much is known about the in vivo function of Bag2. To date, no bag2-null animal models are described. Here, we knocked down Bag2 by morpholinos in the zebrafish embryo and found no phenotypic alterations in the morphants. By contrast, injection of MO-bag2 into homozygous mutant bag3 -/zebrafish embryos led to severe heart and skeletal muscle defects, resembling the phenotypic alterations observed in bag3 morphants. This suggests that Bag2 functionally compensates for the loss of Bag3 in our genetic model of Bag3 deficiency. El-Brolosy and colleagues found that blocking NMD in hbegfa Δ7 , vegfaa and vcla Δ13 zebrafish mutants by genetically inactivating Upf1 led to the preservation of the mutated mRNA levels and, subsequently, to the loss of transcriptional adaptation [1]. Similar to these studies, we found that NMD of the mutated bag3 mRNA triggers the transcriptional adaptation of bag2 expression since blocking NMD by the injection of upf1-specific morpholinos was able to preserve mutant bag3 mRNA levels and to reduce the bag2 mRNA levels to normal levels in homozygous mutant bag3 -/embryos. These findings imply that bag3 mRNA degradation is the main trigger of transcriptional adaptation in bag3 -/zebrafish embryos. Interestingly, bag3 -/zebrafish display increased mortality starting from 9 month of age. We found that bag2 mRNA expression was still increased in the adult skeletal muscle in bag3 -/zebrafish, suggesting that transcriptional adaptation of bag2 expression is able to compensate for the loss of bag3 during embryogenesis but not in the adult zebrafish. Whether this is due to insufficient compensatory bag2 levels or specific Bag2-independent molecular pathomechanisms in the adult zebrafish is unknown but should be investigated in future studies.
Recent findings from zebrafish and mouse studies suggest that genetic compensation is a widespread phenomenon which plays a vital role in the phenotypic peculiarity in response to malignant and harmful gene mutations. Human genetics studies also suggest that genetic compensation might be an active mechanism to influence clinical disease manifestation in response to severely mutated alleles [31][32][33]. Further elucidation of the molecular mechanisms that control genetic compensation might pave the way to develop therapeutic interventions to foster genetic robustness to a malignant mutation.

Zebrafish strains and injection procedures
The present study was performed after appropriate institutional approvals (Tierforschungszentrum (TFZ) Ulm University, No. 0183), which conform to EU Directive 2010/63/EU. Care and breeding of zebrafish, Danio rerio, was conducted as described previously [34,35] Morpholino-modified antisense oligonucleotides (MOs; Gene Tools, LLC, Oregon, USA) were directed against the splice-acceptor/donor site of zebrafish bag3, bag2, bag1 and upf1. As For CRISPR/Cas9 injections, 400 ng/μl recombinant Cas9 protein (Eupheria GmbH, Germany) was mixed with synthetic tracrRNA (100 ng/μl) and a gene-specific crRNA (Eurofins Genomics, Germany) against bag3 (50 ng/μl) in 200 mM KCl. Sequences of CRISPR RNA oligonucleotides and Morpholinos are summarized in S1 Table. RNA extraction and quantitative real-time PCR For RNA extraction, 30 wild-type embryos and 30 bag3 -/embryos were collected at 72 hpf. To extract the RNA the RNeasy1 Mini Kit was used (Qiagen) according to the manufacturer's instructions. Reverse transcription was performed by using SuperScript1 III Reverse Transcriptase (Life Technologies), 1μg total RNA and oligo(dT) primer. Quantitative real-time PCR was carried out according to standard protocols using SYBR-Green master mix (Roche) and a Roche LightCycler 480 II. cDNA was generated as described above from 72 hpf embryos. To correct for sample to sample variation, housekeeping genes β-actin and rpl13 were used for normalization [36]. Sequences of primers used are summarized in S1 Table. Immunoblotting For immunoblot analysis, from 72 hpf wild-type and bag3 -/embryos, proteins lysates were prepared. 30μg of protein lysate was boiled in 5x Laemmli Buffer, separated on an SDS-PAGE (8-16%), and blotted onto polyvinylidene difluoride (PVDF) membranes. Then, the membranes were blocked using 5% milk powder in TBST for 1h at RT and incubated with primary antibodies overnight at 4˚C. The membranes were washed and incubated with polyclonal antirabbit-IgG antibody conjugated to horseradish peroxidase. All membranes were developed using the Pierce ECL Western Blotting Substrate (Thermo Scientific) and a luminescent image analyzer (Image Quant Las4000 mini) [36].

Mass spectrometry
Protein lysates derived from muscle tissue of adult homozygous bag3 -/mutant and homozygous wild-type (bag3 +/+ ) controls (age: 3 month) were separated using SDS-PAGE and the entire lane was subsequently processed as described previously [37] for MS analysis.
Samples were analyzed as previously described [37] by employing an U3000 RSLCnano (Thermo Fisher Scientific, Idstein, Germany) for peptide separation online coupled to an LTQ Orbitrap Velos Pro (Thermo Fisher Scientific, Bremen, Germany) mass spectrometry.
A database search was performed using MaxQuant Ver. 1.6.3.4 (www.maxquant.org) [38]. Using the built-in Andromeda search engine [39], MS/MS spectra were correlated with the Uni-Prot zebrafish database (www.uniprot.org) for peptide identification. Carbamidomethylated cysteine was considered as a fixed modification along with oxidation (M), and acetylated protein Ntermini as variable modifications. False Discovery rates were set on both, peptide and protein levels, to 0.01. For quantitation, LFQ was enabled within MaxQuant using default parameters.

Birefringence analysis
Images were taken with an Olympus SZX 16 microscope and movies were recorded with a Leica DM IL LED microscope. The functional assessment of cardiac contractility was carried out as previously described [34,40].
Regarding the birefringence analysis, 72 hpf zebrafish embryos were anesthetized with tricaine and embedded in 2.5% methylcellulose, in a glass petri dish, making sure that they were lying as flat as possible. Images were acquired between two polarizing filters on an Olympus SZX16 with a DP72 camera and the Olympus Stream software. After acquisition, the inverted images were analysed using the Fiji software and the mean grey value of the selected area of the skeletal muscles was measured. Data analysis was performed normalizing the mean intensity with the selected area as previously described [16,41]. For the publication, the analysed images where optimized from black and white to colored heat map using the tool display "LUT" and selecting the "physics" option in Fiji software.

Functional assessment
For the touch evoked assay, zebrafish embryos at 72 hpf were touched by a needle tip and their flight response was analysed. An immediate and straightforward flight was regarded as "adequate" while no movement or a delayed/incomplete response were regarded as "inadequate". Movies of 10-30 seconds were acquired for each sample to show the larval response to the needle touch. For each experiment the number of embryos with adequate response to the touch was converted to a percentage, analysed, and plotted in a bar graph using GraphPad Prism software [11,42].
To increase the workload on the skeletal muscles, embryos at 120 hpf were incubated in 1% methylcellulose for 2 hours as previously described [10].

Statistical analysis
Statistical analyses were performed using GraphPad Prism software. All the experiments were conducted in triplicate using different biological samples.