Combinatorial genetic replenishments in myocardial and outflow tract tissues restore heart function in tnnt2 mutant zebrafish

ABSTRACT Cardiac muscle troponin T (Tnnt2) mediates muscle contraction in response to calcium ion dynamics, and Tnnt2 mutations are associated with multiple types of cardiomyopathy. Here, we employed a zebrafish model to investigate the genetic replenishment strategies of using conditional and inducible promoters to rescue the deficiencies in the heart. tnnt2a mutations were induced in zebrafish via the CRISPR/Cas9 technique, and the mutants displayed heart arrest and dilated cardiomyopathy-like phenotypes. We first utilized the classic myocardial promoter of the myl7 and TetOn inducible system to restore tnnt2a expression in myocardial tissue in tnnt2a mutant zebrafish. However, this attempt failed to recover normal heart function and circulation, although heart pumping was partially restored. Further analyses via both RNA-seq and immunofluorescence indicated that Tnnt2a, which was also expressed in a novel group of myl7-negative smooth muscle cells on the outflow tract (OFT), was indispensably responsible for the normal mechanical dynamics of OFT. Lastly, tnnt2 expression induced by OFT cells in addition to the myocardial cells successfully rescued heart function and circulation in tnnt2a mutant zebrafish. Together, our results reveal the significance of OFT expression of Tnnt2 for cardiac function and demonstrate zebrafish larva as a powerful and convenient in vivo platform for studying cardiomyopathy and the relevant therapeutic strategies.


INTRODUCTION
Dilated cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM) are two common cardiomyopathies with an estimated prevalence of 1:250 and 1:200 in adults (McKenna et al., 2017;Prondzynski et al., 2018). The prognoses of cardiomyopathies are poor, and the annual mortalities for DCM and HCM are 1.55% and 2.5%, respectively (Pelliccia et al., 2017;Vischer et al., 2009). During the past few decades, genetic screening has identified a list of autosomal inherited mutations as potential 'drivers' or susceptible factors for cardiomyopathy, and approximately 40-50% of all cases can be attributed to genetic mutations (Zou et al., 2013;Paldino et al., 2018). Among these, genes coding sarcomere proteins are frequently found to be dysfunctional in both DCM and HCM, and it is estimated that around 30-40% DCM-associated mutations and 60% HCMassociated mutations occur in sarcomere genes (McNally et al., 2013;Hershberger et al., 2013;Veselka et al., 2017). Besides, mutations in a single sarcomere gene like cardiac troponin T (TNNT2) are sufficient to cause cardiomyopathy, and TNNT2 mutations are the most common 'drivers' of thin filament deficiency in both DCM and HCM (Hershberger et al., 2013;Veselka et al., 2017). Most human TNNT2 mutations are located in central and C-terminal domains of cardiac troponin T and are responsible for both familial cardiomyopathy and sporadic cardiomyopathy (Forissier et al., 1996;Van Driest, 2003;McConnell et al., 2017).
Due to their transparency and small size, zebrafish have been extensively used to investigate the genetics of heart development and to model the pathogenesis of cardiomyopathy in vivo. In zebrafish, a total of four tnnt2a mutations have been previously reported: tnnt2a tc300b , tnnt2a b109 , tnnt2a ex5Δ23 and tnnt2a ex10Δ2 . The former two, tnnt2a tc300b and tnnt2a b109 , were generated by ENU or γ-ray mutagenesis followed by phenotype screening (Sehnert et al., 2002). The tnnt2a tc300b mutant carries a point mutation in the splice site at the intron2-exon3 junction, causing a frameshift and a premature stop in the central domain, while tnnt2a b109 bears a 13-bp deletion of the 5′ regulatory region and displays similar patterns (Sehnert et al., 2002). Contrastingly, the latter two, tnnt2a ex5Δ23 and tnnt2a ex10Δ2 were both generated by our group, carrying a 2-bp deletion in the C-terminal domain and a 23-bp deletion in the central domain, separately . Although both HCM (Gilda et al., 2016) and DCM (Rani et al., 2014) have been found in patients with TNNT2 mutations, all four zebrafish mutants seem to display only DCM-like phenotypes with atrium and ventricle enlargement (Sehnert et al., 2002;Liu et al., 2017). Moreover, electron microscopy in our previous studies as well as in other studies has shown that thin filament formation does not occur in those mutants (Sehnert et al., 2002;Liu et al., 2017). In summary, tnnt2 mutant zebrafish display early onset and highly consistent phenotypes and serve as a specialized but convenient DCM model for assessing potential therapeutic strategies.
In this study, we investigated the transgenic replenishment strategies targeting myocardial and non-myocardial tissues in the zebrafish DCM model to find whether manipulations in these tissues are sufficient for full recovery of heart function. We first demonstrated that tnnt2a mutant zebrafish have typical DCM-like phenotypes. We then designed a myl7 promoter-driven and Tet-On inducible transgenic cascade to ectopically express tnnt2a mRNA only in cardiomyocytes of the mutants. However, we failed to restore mechanical behaviour of the outflow tract (OFT). The results of gene expression and microscopy indicate that tnnt2a was also expressed in a small proportion of the myl7-negative smooth muscle on the OFT, and the additional tnnt2a supplements in these cells significantly rescued the dynamics of the OFT and recovered heart function. To conclude, we have identified a novel non-myocardial Tnnt2+ cell population that is indispensable to a functional OFT and will provide novel insights into therapeutic strategies against TNNT2 for rescuing heart deficiencies.

Generation and phenotype analyses of tnnt2 mutant zebrafish
Tnnt2a mutant zebrafish were successfully generated via the CRISPR/Cas9 technique Wang et al., 2018). Genotyping indicated that two bases of exon ten of tnnt2a were deleted as previously reported (Fig. 1A,B). The frame shift mutation induced a protein-truncating variant, and the homologous tnnt2a ex10Δ2 mutant (tnnt2a −/− ) displayed dilated cardiomyopathylike phenotypes, such as heart arrest as well as atrium and ventricle enlargement 24 h post fertilization (hpf) (Fig. 1C-F). Meanwhile, mutant zebrafish also demonstrated significant pericardial effusion, a phenotype of heart failure, at approximately 72 hpf (Fig. 1C,G).
To determine whether the cardiomyocytes in mutant zebrafish maintained normal electrocardiac rhythm, we introduced the Tg(bactin2:GCamp6s) transgenic background that expresses a slower version of fluorescent calcium sensor ubiquitously . In the completely arrested mutant hearts at 72 hpf, a fluorescent wave of Ca 2+ influx into cardiomyocytes from the venous to arterial end could still be observed (Fig. 1H,I; Movies 1 and 2). This confirmed that tnnt2a mutation blocks the electrical-tomechanical signaling transduction in cardiomyocytes. In addition, the heterozygous tnnt2a ex10Δ2 (tnnt2a +/− ) did not display any defects in heart function at all stages as previously reported .
Heart function is not rescued in Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− zebrafish To perform myocardial-specific gene replenishment, we selected the promoter of myosin light 7 (myl7), which is shorter than the reported tnnt2a promoter (Verma and Parnaik, 2017). Myl7 promoter has also been extensively employed to drive gene expression in atrial and ventricular myocardium in zebrafish hearts at all stages (Thisse et al., 2004;Hsiao et al., 2003). We generated myl7 promotor-driven and dosage-induced (Tet-On system) transgenic cassettes (Yao et al., 2018) for delicate regulations of tnnt2a expression ( Fig. 2A,B). Afterwards, tnnt2a −/− mutant background was introduced into the transgenic lineage by outcrossing twice to generate Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− embryos. Doxytetracycline (DOX) was added before or after the initial time of cardiac contraction (21-22 hpf) to determine whether the genetic rescue was time sensitive. Real-time RT-PCR on dissected hearts demonstrated that DOX administrated before cardiac contraction produced a higher level of tnnt2a mRNA, equivalent to the level in the wild-type hearts (Fig. 2C). Fluorescent images also showed that homozygous mutant with enlarged atria and ventricle relieved abnormal heart phenotypes after adding DOX before the staining point of heartbeat (Fig. 2D). However, phenotyping and statistical analyses indicated that, although the heart beating was restored, and the dilated atrium/ ventricle recovered partially ( Fig. 2E-G), pericardium effusion remained, and pumping of the heart had barely improved blood circulation ( Fig. 2H; Movie 3).
Consistent with the gene expression evidence, the mechanical contraction of the OFT in Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− was dysfunctional during the cardiac cycle (Movie 4), and it was observed that the maximum diameter of OFT in Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− was significantly reduced with restricted movement range compared with that of the normal OFT ( Fig. 4A,B). Furthermore, similar to tnnt2a −/− , blood cells in the heart chambers of Tg(myl7:tetOn; tre: tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− were mostly stacked and heart beating was restored (Fig. 4C). These results suggest that OFT dysfunction may be responsible for the unsuccessful rescue of heart function and blood circulation in Tg(myl7:tetOn; tre:tnnt2a-p2-mKate2); tnnt2a −/− . Additional replenishment of tnnt2a in the smooth muscle cells of OFT recovered the heart function of Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2); tnnt2a −/− zebrafish To further confirm whether Tnnt2 was expressed in certain noncardiomyocyte tissues of OFT, we first conducted in situ hybridization with a tnnt2a anti-sense probe and it was verified that Tnnt2 was expressed in OFT (Fig. 5A). Then we performed immunofluorescence studies and revealed that a cluster of myl7negative, Tnnt2-positive cells in cardiac OFT were stained by F-actin and α-smooth muscle actin (α-SMA) (Fig. 5B-E; Fig. S2), indicating that Tnnt2 is expressed in not only myocardial tissue, but also in some smooth muscles in the OFT. Smooth muscle cells are also present in the aortic valves of mammalian OFTs (Bairati and DeBiasi, 1981;Rogers et al., 2003), suggesting that these noncardiomyocyte Tnnt2-positive smooth muscle cells may be evolutionally conserved between fish and mammals. Next, to validate that Tnnt2 expression in the smooth muscle cell group was indispensable for normal functional OFT, we performed a double-rescue experiment on tnnt2a −/− zebrafish by injecting tol2 mRNA and the elf1α:tetOn; tre:tnnt2a-p2a-mKate2 vector flanked by tol2 repeats into Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− embryos at the single-cell stage (Fig. 6A,B). The universal promoter elf1α was employed to induce mosaic expression in tissues including the OFT smooth muscle cell group because there is no available promoter for those OFT smooth muscle cells. The injected Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7:eGFP); tnnt2a −/− larvae carrying additional red fluorescence of mKate2 (and Tnnt2a) around the OFT (without overlapping green fluorescence) were regarded as the doublerescued samples (Fig. 6C). Representative samples with and without additional mKate2 in the OFT of those injected larvae are shown in Fig. 6C, which indicates the successful/unsuccessful replenishment of tnnt2a in the specific smooth muscle cell group of OFT. In a typical larva with replenishment of tnnt2a in both cardiomyocyte and OFT, the heart contraction and OFT dynamics were both restored (labelled in red fluorescence), and blood circulation was smoothly recovered (Movies 5 and 6). The maximum diameter of the OFT and the size of the pericardium were also measured and statistical analysis showed that the heart of the injected Tg(myl7:tetOn; tre:tnnt2a-p2a-mKate2; myl7: eGFP); tnnt2a −/− zebrafish with tnnt2a-2A-mKate2 expressed in OFT performed better than those without tnnt2a-2A-mKate2 expression in terms of pericardium size and OFT mechanic behaviour (Fig. 6D,E).
Together, our results reveal the importance of OFT expression of Tnnt2 for cardiac function and indicate the complexity of assessing therapeutic strategies for cardiomyopathy caused by sarcomerebased mutations (Fig. 7). Our study also demonstrates zebrafish larva as a powerful and convenient in vivo platform for modelling and investigating cardiomyopathy.

DISCUSSION
Use of zebrafish tnnt2 mutants as highly specialized DCM models Unlike the severe DCM-like phenotypes in the homozygous tnnt2 mutant zebrafish, DCM caused by heterozygous (dominant) mutations in Tnnt2 is a progressive disease in human patients and usually occurs in adults (Pelliccia et al., 2017;Vischer et al., 2009). Hypothetically, the tnnt2 heterozygous mutant and other non-shift mutants may better mimic adult patients. However, we did not observe any heart failure-like phenotypes in heterozygous mutants and other non-shift mutants up to 12 mpf (months post fertilization) (data not shown). The tnnt2 homozygous mutant used in this study has a very early-onset phenotype and can only be considered as a highly specialized embryonic/larval model of DCM. In fact, the original purpose of our project was to integrate the inducible transgene into the mutant background to delay/adjust the onset of the DCM-like phenotypes at different stages (embryonic, larval, juvenile and adult stages). For that purpose, the tnnt2 promoter instead of the myl7 promoter was chosen. Furthermore, the consistency of the phenotypes in the homozygous tnnt2 mutant zebrafish at early developing stages also make this model ideal for quantitative analysis.
Identification of a non-cardiomyocyte Tnnt2-positive smooth muscle cell population in zebrafish OFT Both tnnt2a and myl7 are typical cardiomyocyte-specific genes with slight differences in their expression patterns. In our study, we identified a small population of Tnnt2-positive but myl7-negative cells, which are indispensable to normal heart function in zebrafish. In other teleost, a smooth muscle organ called bulbous arteriosus (BA) has been previously described in the OFT, which displays contraction at each pumping (Moriyama et al., 2016). In mammals, molecular and histological studies have identified the presence of a similar structure, which is an α-SMA-positive smooth muscle system that is differentiated from heart interstitial cells (Ayoub et al., 2016;Latif et al., 2015) in the aortic valve leaflets (AVL) (Bairati and DeBiasi, 1981;Taylor et al., 2000;Rogers et al., 2003). Teleost BA and mammalian AVL in the OFT may be evolutionally conserved in function and cellular composition (Moriyama et al., 2016).
The hints for gene therapy exclusively targeting cardiomyopathy Current therapeutic strategies, as well as patient outcomes associated with cardiomyopathy, remain disappointing. Over 40% of children with DCM die within 2 years without a heart transplantation (Lee et al., 2017). In comparison to compoundbased therapies, gene therapy holds promise for an ultimate cure for cardiomyopathy. In preclinical experiments, Jiang et al. (2013) and Mearini et al. (2014) delivered an adeno-associated virus (AAV9) carrying Myh6 siRNA or Mybpc3 cDNA into the thoracic cavity or temporal vein of Myh6 or Mybpc3 mutant mice, respectively, relieving cardiomyopathy symptoms in both models. Although the virus-based delivery techniques used in most clinical gene therapies are not commonly applied to zebrafish, the convenient transgenesis in zebrafish via zygote injection can easily simulate or mimic the fundamental rescue strategies of those gene therapies.
Theoretically, cell/tissue-specific promoters restrain the expression of transgenes at an appropriate time and space and may be utilized to optimize gene therapy. In cases of heart diseases, cardiomyocytes have been regarded as the most preferred target cells in most current gene/cell therapies. In our study, we highlighted an important warning sign for the transgenic strategy of cardiomyocyte-exclusive rescue. The myl7 promoter-driven tnnt2a replenishment failed to restore heart function in tnnt2a mutant zebrafish, while the additional expression of tnnt2a in a small group of smooth muscle cells in the OFT made up for the deficiency. In successful rescue, it was more likely that tnnt2a supplement expression in OFT smooth muscle cells was responsible, although it was also possible that in the embryos selected for OFT expression, expression elsewhere is also induced. Our results demonstrated how slightly different expression patterns between two promoters may significantly affect the rescue efficiency of genetic replenishment, and the selection of non-endogenous cell/ tissue-specific promoters for developing treatment for patients with cardiomyopathy should be carefully assessed. (D-G) Statistics for the heart rate, atrial and ventricular diameter, and pericardium size in wild-type and tnnt2a −/− zebrafish; ****P<0.0001; n=15 per group. Statistical differences were evaluated using two-tailed unpaired t-tests (D-F) or unpaired t-test with Welch's correction (G). (H,I) Fluorescent images and pseudo-colours of electrocardiac rhythm in Tg(bactin2: GCamp6s; tnnt2a +/+ ) and Tg(bactin2:GCamp6s; tnnt2a −/− ) zebrafish; arrows represent increased Ca 2+ flux in cardiomyocytes. The direction of Ca 2+ flux was from venous to arterial end both in tnnt2a +/+ and tnnt2a −/− . Scale bars: 100 µm.
Additionally, pDestTol2CG2 was used as LR vector. NotI digestion was used to linearize pCS2FA-transposase (Tol2kit) and transposon RNA was synthesized using mMESSAGE mMACHINE ® SP6 Transcription Kit (Invitrogen) and purified using PureLink ® RNA Mini Kit (Invitrogen). A mixture of knockin plasmid (30 ng µl −1 ) and Tol2 transposase RNA (50 ng µl −1 ) was directly injected into the one-cell-stage zygote. Successful generation of knockin zebrafish was confirmed via heart-specific green or red fluorescence with/ without DOX induction. Knockin zebrafish larvae were nurtured into adulthood and outcrossed with wild-type AB or mutant zebrafish.

Imaging and image processing
Live zebrafish larvae narcotized with 200 mg l −1 tricaine (Sigma-Aldrich, A5040) were mounted in 3% methylcellulose for imaging. Fixed samples were balanced and located in 80% glycerol before imaging. Images of whole zebrafish or dissected heart, and movies of the beating heart were obtained via either a fluorescent dissecting microscope (Olympus, DP73) or a confocal microscope system (Leica, TCS-SP8). All videos were taken at 10 frames per second. Confocal images are shown as single optical sections or maximum intensity Z projections, respectively, as indicated. Heart atrium and ventricle diameter measurements were performed at the end of atrial and ventricular diastole using straight measurement in ImageJ. Pericardium size was measured using area measurement in ImageJ. The maximum diameter of OFT was used to calculate the difference between diastolic and contractile diameters of OFT.

DOX induction
Doxycycline (Sigma-Aldrich, D9891) was added to E3 medium to produce a final concentration of 80 µg ml −1 . Zygotes were treated with DOX at 16 hpf or 48 hpf. E3 medium containing doxycycline was changed every 12 h.

Transcriptome resequencing analysis (RNA-seq)
Zygotes were obtained by crossing Tg(myl7:TetOn; tre:tnnt2a-mKate2; myl7:eGFP); tnnt2a +/− with tnnt2a +/− . DOX was added at 16 hpf to ensure full rescue of Tg(myl7:TetOn;tre:tnnt2a-mKate2;myl7:eGFP); tnnt2a −/− . At 48 hpf, transgene group zebrafish was screened via fluorescence. At 72 hpf, tails were cut to identify genotypes, and the body was placed in a 96well plate filled with Trizol. Zebrafish (N=50 per group) with identical mutants were collected to extract RNA using the RNeasy Plus Mini Kit (Qiagen 74132). Library construction, sequencing and bioinformatics analyses were performed as previously described (Fu et al., 2013). In brief, cDNA libraries were constructed according to the manufacturer's standard protocol (Illumina, Inc.). Oligo (dT) isolated mRNA was fragmented, and cDNA was synthesized using these mRNA fragments as templates. Short fragments were purified and resolved with EB buffer for end reparation and single nucleotide adenine addition. Next, the short fragments were connected with adapters, and suitable fragments were selected for PCR amplification. During the QC steps, the Agilent 2100 Bioanaylzer and ABI StepOnePlus Real-Time PCR System were used in quantification and qualification of sample libraries. Libraries were sequenced using the Illumina HiSeq 4000 platform. After sequencing, we first filtered low-quality, adapter-polluted and high content of unknown base (N) reads to separate clean reads. The clean reads were mapped to the reference genome using HISAT. All coding transcripts were then merged with reference transcripts to get complete references. Next, we performed gene expression analysis using these references via StringTie (Kim et al., 2015) and Biowtie (Shen et al., 2014). We then detected DEGs via DESeq2 (parameters: Fold Change≥2.00 and Adjusted P≤0.05) and performed Gene Ontology (GO) classification and functional enrichment, as well as KEGG pathway classification and functional enrichment for DEGs using phyper, a function of R (r-project.org). Sequencing was performed once for each sample as an initial screening. The raw data may be downloaded from SRA: SRP7896793, SRP7896794, SRP7896795, and SRP7896796 (Table S3).

Real time RT-PCR
Zebrafish larvae at 72 hpf (N=30 per sample, 3 samples per group) were fixed in 4% PFA at 4°C for 1 h before the hearts were dissected. Total RNA was extracted from whole zebrafish larvae or dissected hearts by using the RNeasy FFPE Kit (QIAGEN 73054). PrimeScript RT reagent kit (Takara RR037A) was used to synthesize cDNA. Real time RT-PCR was conducted using SYBR Premix Ex Taq II (Takara RR420D) in the StepONEPlus Real-Time PCR system (Applied Biosystems). The mRNA levels were normalized to an average of β-actin as an internal control. The primers used were listed in Table S5.

Quantification and statistical analysis
Diameter measurements, statistical calculations and graphs were generated using ImageJ and GraphPad Prism 6. Data are expressed as mean±s.d. (N is listed in each figure legend). Statistical differences were evaluated using two-tailed unpaired t-test, unpaired t-test with Welch's correction, Kruskal-Wallis test, one-way ANOVA, and two-way ANOVA.