Exon- and contraction-dependent functions of titin in sarcomere assembly

The gene that encodes the largest protein, titin (TTN) (Labeit et al., 1992), was recently established as a causative gene for cardiomyopathies and muscular dystrophies (LeWinter and Granzier, 2013; Chauveau et al., 2014). Approximately 25% of patients with end-stage dilated cardiomyopathy (DCM) have truncating mutations in TTN (Herman et al., 2012). However, the clinical translation of this genetic knowledge has been complicated (Norton et al., 2013). While TTN-truncating variants (TTNtvs) in exons encoding A-band domains cause DCM, many TTNtvs in exons encoding Z-disk domains are benign (Herman et al., 2012). Moreover, mutations found in patients with limb-girdle muscular dystrophy type 2J and tibial muscular dystrophy are more common in C-terminal exons that encode M-line domains of TTN (Hackman et al., 2002; Udd et al., 2005). A dominant toxic peptide hypothesis has been proposed to explain the allelic heterogeneity phenotypes, whereby A-band, but not Z-disk, truncated TTN is incorporated into normal sarcomeres to exert toxic effects on heart function (Herman et al., 2012). On the other hand, exon-dependent phenotypes could be explained by the differential use of the 363 exons of the human TTN gene (reviewed by Guo et al., 2010). In line with this loss-of-function hypothesis, a study of patients with DCM identified a correlation between exon usage and severity of the DCM phenotype (Roberts et al., 2015). More recently, the identification of a short TTN isoform, termed Cronos, prompted a new hypothesis. Driven by an internal promoter located in the intron before exon 240 of TTN, the Cronos isoform only encodes A-band and M-line domains and misses all Z-disk and most I-band domains (Zou et al., 2015). As such, the authors posited that mutations in A-band exons are deleterious because of the disrupted Cronos isoform, whereas mutations in Z-disk exons are benign because of the normal Cronos isoform.

One TTN molecule spans half of a sarcomere, and sequences of TTN domains from N-terminus to C-terminus match the different sarcomeric substructures – including the Z-disk, I-band, A-band and M-line (Labeit and Kolmerer, 1995). Therefore, full-length TTN has been proposed to act as a template during de novo sarcomere assembly (Trinick, 1996). Moreover, the A-band region of TTN has been proposed to function as a ‘molecular ruler’ to govern the assembly of thick filaments and define the length of the A-band at 1.6 µm (Whiting et al., 1989). Despite multiple genetic studies of TTN in the past 20 years, however, the template and ruler hypotheses remain inconclusive. Several Ttn mutants have been generated in mice, including a TTNtv affecting the A-band (Gramlich et al., 2009) and internal deletion mutants that remove the N2B exon (Radke et al., 2007) or an M-line exon (Weinert et al., 2006). However, all mutants were not null, which contributes to the inconclusive findings regarding function of TTN in de novo sarcomere assembly.

Zebrafish is a new vertebrate model for studying sarcomere assembly (Schoenebeck and Yelon, 2007; Huang et al., 2009; Sanger et al., 2009; Yang et al., 2014b). From a large-scale screen using N-ethyl-N-nitrosourea as a mutagen, a group of ttn mutants, named pickwick (pik), have been identified (Xu et al., 2002). Because of the gigantic gene size, the precise location of the mutation was only mapped for the pik^m171 allele. Other zebrafish ttn mutant alleles, such as runzel (Steffen et al., 2007) and herzschlag (Myhre et al., 2014), were reported later, but the precise locations of these mutations were not identified. Zebrafish have two ttn homologs, ttna and ttnb (Seeley et al., 2007), which are referred to here as ttn.2 and ttn.1, respectively, according to a new nomenclature (Table S1). We previously used splice-blocking morpholinos to target the N2B and N2A exons of ttn.2 and ttn.1 (Seeley et al., 2007). However, concerns about the specificity of the morpholino technology (Kok et al., 2015; Rossi et al., 2015) and the advent of gene-editing technologies such as transcription activator-like effector nuclease (TALEN)-based methods (Campbell et al., 2013) prompted the generation of stable ttn knockout zebrafish mutants. Whereas a panel of ttn.2 mutants have previously been generated using CRISPR-Cas9 technology (Zou et al., 2015), here, we used TALEN (transcription activator-like effector nuclease) technology to introduce a series of deletion mutations. We detected five phenotypes of different severity from seven different ttn mutants, recapitulating exon-dependent phenotypes. We characterized a near-null ttn mutant and provided the first genetic evidence uncovering functions of titin in de novo sarcomere assembly.

We designed TALEN pairs targeting sequences in exons 5 and 201 of ttn.2, which encode domains in the Z-disk and the A-band, respectively (xu064 and xu065; Fig. 1A), and in exons 5 and 165 of ttn.1, which encode the corresponding domains in the Z-disk and the A-band (xu066 and xu067; Fig. 1A). For each TALEN, at least two deletion mutant alleles with two different frameshifts were obtained (data not shown). Because the two alleles exhibited the same phenotypes, suggesting no off-target effects, we focused on one allele for each mutant locus.

To determine the protein level of titin in these mutants, we used vertical agarose gels followed by silver staining (Warren et al., 2003) and detected two Ttn bands in larvae at 2 days post-fertilization (dpf). Surprisingly, all four mutants showed the same two bands as in wild-type (WT) larvae (Fig. 1B,C). No truncated Ttn bands of smaller size were detected in the ttn.2 mutants (Fig. 1B), which might be partially ascribed to nonsense-mediated mRNA decay since ttn.2 mRNA expression was decreased by more than half in both ttn.2 mutants (Fig. 1D). We detected a shorter Ttn band at ≈2.11 MDa in ttn.1^xu067 (Fig. 1C). The ttn.1 transcript was decreased in ttn.1^xu066 but remained unchanged in the ttn.1^xu067 mutant (Fig. 1E). At the mRNA level, ttn.2 and ttn.1 seemed to regulate each other, as indicated by the decrease in ttn.1 mRNA expression in ttn.2^xu064 (Fig. 1D) and the 2- to 3-fold increase in ttn.2 mRNA expression in both ttn.1^xu066 and ttn.1^xu067 mutants (Fig. 1E).

Both ttn.2 homozygous mutants showed pericardial edema with reduced ventricle size (Fig. S1A), slightly decreased heart rates (Fig. S1B), and nearly depleted shortening fraction at 2 dpf (Fig. S1C). Their hearts appeared as a string at 5 dpf, the bodies exhibited severe abdominal edema at 7 dpf, and most embryos died at ∼10 dpf (Fig. S1D,E). Compared with ttn.2^xu064 mutants, which still swam at 3 dpf, the ttn.2^xu065 mutants were completely paralyzed (Fig. S1D). Compared with well-assembled myofibrils in somites of WT fish at 2 dpf, myofibrils in ttn.2^xu064 appeared slightly misaligned, with reduced sarcomere length (mean±s.d. length, 1.89±0.23 µm in ttn.2^xu064 versus 2.21±0.16 µm in WT; Fig. 2A); myofibrils in ttn.2^xu065 were severely disrupted, with only residue-striated myofibrils consisting of much shorter sarcomeres (1.32±0.20 µm; Fig. 2A).

Unlike ttn.2, the ttn.1 mutants appeared normal until 5 dpf, when an inflated swim bladder failed to develop (Fig. S1D), and both mutants exhibited decreased mobility upon touch stimulation with a needle. ttn.1^xu066 mutants died at ∼12 dpf, whereas ttn.1^xu067 mutants survived up to 17 dpf (Fig. S1E). Consistently, disarranged myofibrils were noted in ttn.1^xu066 somites at 9 dpf when myofibrils in ttn.1^xu067 were normal (Fig. 2B). Together, our data revealed four distinct phenotypes resulting from deletion mutations in four different exons of ttn.2 or ttn.1, recapitulating the allelic heterogeneity of TTNtvs in DCM (Herman et al., 2012).

The observation of more severe sarcomere phenotypes in ttn.2^xu065 than ttn.2^xu064 seems consistent with findings in patients with DCM – that is, a TTNtv in an A-band exon, but not a Z-disk exon, results in DCM (Herman et al., 2012). To directly test the toxic peptide hypothesis, we generated a ttn.2^xu068 cis-compound mutant by introducing a Z-disk nonsense mutation into ttn.2^xu065/+ (Fig. 2C). The ttn.2^xu068 cis-compound homozygous mutant exhibited comparable defects to those in the ttn.2^xu065 mutant, which indicates that introducing a Z-disk nonsense mutation fails to rescue the phenotype (Fig. 2D). Together with the lack of truncated proteins in our vertical agarose gels (Fig. 1B), our data argue against the toxic peptide hypothesis.

To assess the Cronos hypothesis (Zou et al., 2015), we tested for the presence of Cronos by protein gel analysis but detected no bands in the region corresponding to the predicted size for the Cronos protein in either WT embryos or ttn mutants (Fig. 1B,C).

To test the exon usage hypothesis (Roberts et al., 2015), we used absolute quantitative reverse transcriptase polymerase chain reaction (qPCR) to compare the expression level of these four exons, with the N2B and N2A exons as controls. In 2 dpf WT larvae, the expression level of exon 5 in ttn.2 was half that of exon 201 (Fig. 3A) and expression of exon 5 in ttn.1 was 6-fold higher than that of exon 165 (Fig. 3B). The different expression levels of the two exons in ttn.2 remain at 5 dpf (Fig. S3B,C) and 9 dpf, respectively (Fig. S3D,E), whereas expression levels of the two exons in ttn.1 become similar. We posited that the different exon usage might result from alternative splicing events. Using qPCR with primers targeting exon 2 of ttn.2 and exon 7 of ttn.1, we detected alternative splicing events around exon 5 of ttn.2 but not around exon 5 of ttn.1 (Fig. 3C). Three alternative splicing isoforms of ttn.2 were found, all of which skipped exon 5 (Fig. 3D). In contrast, we found no alternative splicing events around exon 199 to exon 203 of ttn.2 (Fig. S2A,B) but did note exon-skipping events on exon 165 and exon 166 of ttn.1 (Fig. 3E,F). These findings are consistent with the exon usage data and correlate with the severity of phenotypes in these four ttn.2 or ttn.1 mutants.

The presence of two full-size Ttn bands in all four ttn mutants containing a single deletion mutation suggested their identity as hypomorphic mutants. To test this hypothesis, we generated the ttn^xu069 (double-Z) mutant that contains nonsense mutations in Z-disk exons of both ttn homologs and the ttn^xu070 (double-A) mutant that contains deletion mutations in A-band exons in both ttn homologs (Fig. 4A). The higher Ttn band disappeared in both double-Z and double-A mutants, but the lower band remained (Fig. 4B,C). The compensation of the Ttn protein level was at least partially removed at the transcriptional level, as indicated by the suppression of ttn.2 mRNA upregulation induced by ttn.1 mutations in these double mutants (Fig. 4D). Compared with ttn.2^xu064 and ttn.1^xu067 mutants that could swim, further reduction of Ttn.1 in ttn.2^xu064 or reduction of Ttn.2 in ttn.1^xu067 paralyzed the fish (data not shown). Myofibrils were severely disrupted in both double-Z and double-A mutants (Fig. 4E).

Because the lower Ttn band still existed in these cis double mutants, we targeted two highly used exons of ttn.2 and ttn.1, and generated ttn^xu071 (Fig. 5A). Both Ttn bands were undetectable in ttn^xu071 (Fig. 5B), and both ttn.2 and ttn.1 transcripts were reduced (Fig. 5C). Sarcomeres were disrupted to a similar degree as in the double-A and double-Z mutants (Fig. 5D). Only residual thick-filament-like structures can be detected in the electron micrographs (Fig. 5E).

The functional redundancy between ttn.2 and ttn.1 was further underscored by our data on ttn.2^xu071/+;ttn.1^xu066, a mutant generated by crossing ttn.1^xu066 into ttn^xu071/+. In contrast to the ttn.2^xu065/+ and ttn.1^xu066 mutants with normal striated Z-disks at 2 dpf, ttn.2^xu071/+;ttn.1^xu066 larvae showed myofibril disarray (Fig. 5F).

In a zebrafish embryo at the 18-somite stage, different stages of sarcomere assembly in the slow muscle can be observed simultaneously (Sanger et al., 2009). Z-bodies formed dotted structures with a mean regular distance of 0.89±0.19 µm (mean±s.d.) at the 15th somite, became better striated with an expanded sarcomere length of 1.35±0.27 µm at the 12th somite, and further expanded to 2.14±0.2 µm at the 6th somite (Fig. 6A,C). At the 6th somite, the Z-bodies completed the initial lateral growth process and formed the wider Z-disk with a width of 0.86±0.18 µm (Fig. 6A,B). Ttn started to be expressed as random dots at the 18th somite but became striated at the 12th to 13th somites (Fig. S4).

Because ttn^xu071 is the first near-null ttn mutant allele in a vertebrate (Fig. S4), we interrogated the functions of ttn in de novo sarcomere assembly. The earlier steps of Z-body assembly appeared normal in ttn^xu071, as indicated by a regular dotted pattern of Z-bodies at an average distance of 0.80±0.18 µm at the 15th somite and the expansion of the striated Z-bodies to 1.18±0.21 µm at the 12th somite (Fig. 6A,C). However, these Z-bodies failed to grow further in either width (0.45±0.09 µm; Fig. 6B) or length (1.29±0.16 µm) at the 6th somite (Fig. 6A,C).

Similarly, the assembly process of the thick filament in the slow muscle can also be revealed in a single embryo at the 18-somite stage using an F59 antibody. We found that myosin filaments in the WT zebrafish were partially bundled along the long axis of the somite, with an average thick filament length of 0.8±0.15 µm at the 15th somite, became better organized and striated with expanded length (1.46±0.2 µm) at the 12th somite, and finally formed mature A-bands with a length of 1.64±0.13 µm at the 6th somite (Fig. 7A,C). Similar results were obtained using an A4.1025 antibody, which also detects myosin heavy chain isoforms (data not shown). Resulting from lateral growth, the width of myosin filaments also expanded to 0.92±0.14 µm at the 6th somite (Fig. 7B). In contrast, only random myosin dots with a length of 0.83±0.13 µm could be detected in ttn^xu071 at the 15th somite, which remained unorganized and were 0.87±0.18 µm long at the 12th somite (Fig. 7A,C). Myosin filaments with a width of 0.6±0.11 µm and length of 1.1±0.21 µm did emerge at the 6th somite (Fig. 7A-C). Different from the WT embryos, these thick filaments did not form mature A-bands, which indicates an instructive role for titin in thick filament assembly.

In ttn.2^xu064, sarcomere structures and Z-disks remained relatively normal at 2 dpf (Fig. 2A). Interestingly, thick and thin filaments were undetectable in the middle part of each myofibril, and the blanked middle section was enclosed by two rod-like structures that connected to the remaining myofibrils in each somite segment (Fig. 8A). The rod-like structures could be stained by both Phalloidin and an anti-myosin antibody. This unusual phenotype only occurred in the slow muscles located superficially under the skin and not in the fast muscles located deeper and oblique to the anterior-posterior axis (Devoto et al., 1996), as indicated by both Phalloidin staining and immunostaining using antibodies against fast-muscle tropomyosin and myomesin (Fig. 8A,B).

Given that a developing zebrafish larva starts to twitch at 27 hpf, when slow muscle is mainly used (McKeown et al., 2009), we posited that slow-muscle-specific phenotypes in ttn.2^xu064 may be due to the mechanical stress induced by muscle usage in the setting of reduced sarcomere stability. To test this hypothesis, we inhibited muscle contraction by anesthetizing ttn.2^xu064 mutant larvae with 1% tricaine at 24 hpf, right before the first twitch, for 24 h. Indeed, the disrupted myofibrils were significantly rescued (Fig. 8C and Fig. S5). Similar rescuing effects were noted in embryos treated with 10 µM blebbistatin (data not shown), a myosin inhibitor that inhibits muscle contraction via a different mechanism (Straight et al., 2003). By contrast, the disrupted myofibrils were further exaggerated by increasing muscle contraction via either bleaching the chorion (Zhang et al., 2009) or treatment with galanthamine (Otten and Abdelilah-Seyfried, 2013), an acetylcholine esterase inhibitor (Fig. 8C, Fig. S5, and data not shown). Together, our data suggested that partially decreased expression of Ttn affects the stability of sarcomeres in the slow muscle, resulting in higher susceptibility to damage from mechanical stress. Alternatively, this result might reflect different titin isoform usage in slow and fast muscles.

Using TALEN-based genome-editing technology, we generated four zebrafish ttn mutants with mutations affecting single exons in either ttn.2 or ttn.1 and three double mutants. All mutations were small deletions that purposely shift the reading frame and presumably result in stop codons. However, both Ttn bands could be identified by protein gel analysis of the four single mutants, indicating that they are hypomorphic mutants. No truncated Ttn band was detected in any of the mutants except ttn.1^xu067. Strong compensational effects must exist, which might include a scavenger system that removes the truncated proteins.

This prediction was validated by our follow-up studies with double mutants. Whereas double-A and double-Z mutants lack the upper band, the ttn^xu071 double mutant lacks both bands, resulting in a near-null mutant. We did not term the mutant as a null because of possible residual Ttn protein that is beyond the detection limit of silver staining. Our data challenge the concept that the upper band represents the product of ttn.2 and that the lower band denotes the product of ttn.1 (Steffen et al., 2007). Instead, each Ttn band likely consists of a mixture of ttn.2 and ttn.1 products because neither upper nor lower band can be completely depleted in single mutants. Based on results from mammalian Ttn gels (Granzier and Labeit, 2004), we predict that the two bands might represent N2B- and N2BA-based Ttn isoforms. Given that ttn.2 consists of 234 exons with a 176,965 bp coding sequence, which is longer than ttn.1 (201 exons with 137,988 bp coding sequence) (Seeley et al., 2007), our observation further suggests that there is a length-control mechanism for maintaining the size of Ttn proteins.

Our qPCR studies suggest that the length-control mechanisms and the scavenger mechanism might involve regulation at the transcriptional level. We detected significant downregulation of ttn.2 or ttn.1 transcripts in all mutants except ttn.1^xu067, which indicates that nonsense-mediated decay is involved. We detected increased ttn.2 activation in two ttn.1 mutants (Fig. 1E), which suggests genetic compensation induced by the loss of ttn.1. Of note, the changes at the protein level do not always match the changes at the RNA level, as indicated by the decreased protein bands in ttn.1^xu066 (Fig. 1C), suggesting additional regulation at the translational level. Further detailed studies are warranted using these zebrafish mutants to uncover compensational mechanisms for regulating the length and isoforms of Ttn.

By depleting Ttn proteins to different degrees, we noticed five sarcomeric phenotypes of different severity among the seven ttn mutants (Fig. S6). First, de novo sarcomere assembly is completely arrested in the double-Z, double-A and near-null mutants (Fig. 4E and Fig. 5D). The phenotypes are indistinguishable among this group using immunostaining assays, despite the fact that Ttn proteins are disrupted to different degrees, ranging from loss of the upper band to loss of both bands. Second, de novo sarcomere assembly was also arrested in ttn.2^xu065, but residue-striated myofibrils could be transiently seen. The ttn.2^xu068 double mutant also belongs to this group. Third, sarcomeres can be successfully assembled in ttn.2^xu064, despite the disarray of myofibrils shown at 2 dpf. Fourth, sarcomere structure appears normal in ttn.1^xu066 at 2 dpf, but myofibril disarray appears at 9 dpf, and the larvae die at 12 dpf. Fifth, sarcomere assembly appears normal in ttn.1^xu067 at 9 dpf, but the larvae die at 17 dpf.

This panel of different sarcomeric phenotypes clearly demonstrates a crucial function of Ttn in maintaining the integrity of the sarcomere (Fig. S6). A slight decrease in total Ttn protein expression is not sufficient to affect de novo sarcomere assembly; however, myofibrils become unstable and exhibit phenotypes under biomechanical stress. It remains to be determined whether the detachment of thin and thick filaments in the center of a myofibril is a common initial subcellular form of damage occurring in patients with a TTN mutation.

The generation and characterization of seven ttn mutants with defined mutation locations help to interpret other reported zebrafish ttn mutants. On the basis of their cardiac phenotypes at 2 dpf, it is reasonable to predict that all pik mutants affect ttn.2 (Xu et al., 2002). By contrast, the late-onset phenotypes in both ttn.1 mutants are similar to those described in runzel (Steffen et al., 2007), which suggests that the runzel mutant affects ttn.1 but not ttn.2.

As to the mechanism underlying exon-dependent functions of Ttn.2 or Ttn.1 in sarcomere assembly, our data do not favor the toxic peptide hypothesis for three reasons. First, we did not detect truncated Ttn protein in most ttn mutants. Second, although we did detect a truncated Ttn band in the ttn.1^xu067 mutant (but not the ttn.1^xu066 mutant), the phenotype was more severe in the Z-disk ttn.1^xu066 mutant than the A-band ttn.1^xu067 mutant (Fig. 2B). Third, the phenotype of the ttn.2^xu068 cis double mutant was similar to that of ttn.2^xu065 but not ttn.2^xu064 (Fig. 2D).

It remains to be determined whether our conclusions based on fish embryos can be extrapolated to TTN-based adult human diseases. Truncated TTN protein was not detected in samples of skeletal muscle (Gerull et al., 2006) or left ventricles (Roberts et al., 2015) from patients with cardiomyopathy. A recent study using engineered induced pluripotent stem cell-derived cardiomyocytes mimicking different A-band truncation mutations from human patients reported a smaller TTN fragment in only one of the three cell lines (Hinson et al., 2015). In contrast, truncated TTN was found in skeletal muscle biopsy specimens from patients with DCM (Gerull et al., 2002) and in a mouse model (Gramlich et al., 2009). Obviously, a much larger sample size is needed to clarify whether truncated Ttn results from TTNtvs and whether these truncated proteins contribute to pathogenesis.

Our data also do not support the Cronos hypothesis for several reasons: (1) our protein gel analysis failed to detect the Cronos protein, (2) ttn.1^xu066 exhibits more severe phenotypes than ttn.1^xu067 (Fig. 2B) and (3) the double-Z mutant exhibits phenotypes similar to the double-A mutant (Fig. 4E).

Among the three current hypotheses for explaining exon-dependent phenotypes of ttn.2/1 mutants, our data agree only with the exon usage hypothesis. We demonstrated a correlation among exon usage, alternative splicing patterns (Fig. 3) and severity of phenotypes in the somites (Fig. 2A,B). Consistent with this finding, a recent study showed a strong correlation between ‘percentage spliced in’ of nonsense mutation-containing exons and DCM severity (Roberts et al., 2015), and another study using engineered induced pluripotent stem cells suggested that nonsense mutations in exons of TTN with higher levels of percentage spliced in can cause significant contractile deficits via haploinsufficiency of full-length TTN (Hinson et al., 2015). Our generation of a near-null ttn mutant allele by targeting highly used exons in ttn.2 and ttn.1 gives more supporting evidence that exon usage is the underlying mechanism (Fig. 5).

Generation of the first near-null ttn mutant in a vertebrate enables us to genetically test the function of Ttn, which was proposed 20 years ago (Whiting et al., 1989; Trinick, 1996). We found that depletion of Ttn did not affect the alignment of Z-bodies to form a striated pattern at the 12th to 15th somites in an 18-somite embryo (Fig. 6A,C), which indicates that Ttn is not required for assembly of the premyofibrils. Ttn is needed, however, for the formation of nascent myofibrils from premyofibrils, which is characterized as length expansion of the sarcomere from 0.8 to 2.2 µm and widening of the Z-bodies to form Z-disks.

The ttn^xu071 mutant also uncovered two functions of Ttn in thick filament assembly. First, Ttn facilitates the incorporation of the random myosin dots into thin filament networks. When Ttn is absent, myosin filaments remain scattered in the 12th and 15th somites in an 18-somite embryo. Incorporation of the thick filaments into the thin filament networks is delayed until a later developmental stage, as represented by the 6th somite in an 18-somite embryo. Second, Ttn is required for the expansion of the rod-like thick filaments, with a length of ∼0.8 µm, to a uniform A-band with a length of 1.6 µm. This assembly step of the thick filaments is part of the sarcomere assembly step when premyofibrils transform into nascent myofibrils.

Together, our genetic evidence supports the following hypothesis regarding functions of Ttn (Fig. 9). At the premyofibril stage, the Ttn N-terminus binds to the Z-bodies in the existing myofibril network, and its flexible C-terminus binds the thick filaments that are assembled independently and still scattered in the cell (Hill et al., 1986; Wang et al., 1988; Komiyama et al., 1990). Ttn facilitates the incorporation of the thick filaments into the thin filament networks and then guides the lateral alignment of thick filaments and Z-bodies to form mature nascent myofibrils (Fig. 9). Of note, our studies were carried out at the 18 somite stage to eliminate interference from the fast muscle, as its myofibrillogenesis occurs several hours later. Whether our conclusion based on studies in slow muscle is applicable to fast muscle requires future experimental validation. Moreover, our genetic data cannot be used to support the idea that titin functions as a ‘molecular ruler’ to define the length of the A-band – a hypothesis that has also been disproved by a study of a mouse mutant harboring an internal deletion of titin at its I-band/A-band junction (Granzier et al., 2014).

Our data seem to contradict a previous publication stating that the A-band rod domain of Ttn.2 is not essential for thick filament assembly (Myhre et al., 2014). The conclusion was based on herzschlag, a ttn.2 mutant harboring an unidentified mutation mapped to the I-band region of ttn.2. It was assumed that the unidentified mutation produced a truncated Ttn.2 protein that loses its thick-filament binding domains, which might not be correct. It is possible that herzschlag is a hypomorphic allele but not a null allele.

In summary, our studies of a panel of ttn mutants uncover functions of Ttn in both de novo sarcomere assembly and sarcomere integrity after assembly. Importantly, our data show that a TTNtv cannot be assumed to be a truncating mutation without supporting evidence from a protein gel. Thus, the nomenclature for these nonsense variants as TTNtvs can be misleading. A slight decrease in the Ttn pool is likely to be sufficient to result in a wide spectrum of phenotypes. The weakened myofibrils have much higher susceptibility to damage incurred by stresses such as muscle contraction.

Caution must be taken when extrapolating these mechanisms uncovered in sarcomere assembly during somitogenesis to cardiomyopathy and muscular dystrophy, which mainly occur in adults. An obvious future direction that is enabled by our study is to raise this panel of ttn mutants to adulthood and then assess phenotypes in both heart and muscle. With the convenience of introducing mutations precisely into any exon of ttn, it is anticipated that future studies using the zebrafish model will considerably facilitate our understanding of titinopathy and its related therapies.

Zebrafish were handled following the guidelines of the Mayo Clinic Institutional Animal Care and Use Committee (IACUC protocol A17814). The WIK line was used.

TALEN pairs in ttn.2 and ttn.1 (Seeley et al., 2007) were designed using Zifit (http://zifit.partners.org/ZiFiT/ChoiceMenu.aspx) and assembled using a Golden Gate kit (Addgene) (Cermak et al., 2011). Capped mRNAs were synthesized using mMESSAGE mMachine T3 (Ambion) and injected into 1-cell-stage embryos. Successful mutagenesis was confirmed by PCR followed by digestion with restriction enzymes (Table S2). Two mutant alleles that result in two different shifts of reading frame for each mutant locus were selected. Because ttn.2 and ttn.1 are located on the same chromosome, double ttn mutants were generated by injecting a second TALEN pair into an existing ttn mutant.

Protein samples from zebrafish larvae were extracted by homogenization using a mortar and pestle (Thermo Fisher Scientific) in sample buffer [8 M urea, 3% v/v SDS, 2 M thiourea, 0.05 M Tris-HCl (pH 8.0), 0.03% w/v Bromophenol Blue, 75 mM DTT and 0.01% protease inhibitors (Roche)]. After the samples were incubated in 65°C prewarmed water for 10 min and centrifuged at 4°C for 5 min, the samples were separated on a 2% SDS/agarose gel (30% v/v glycerol, 0.05 M Tris base, 0.384 M glycine, 0.1% w/v SDS and 2% w/v SeaKem Gold agarose) using a vertical agarose gel electrophoresis system (Warren et al., 2003); 1 kb Plus DNA Ladder (Invitrogen) was used for size markers. The gel was fixed in prefixing solution (50% v/v methanol, 12% v/v glacial acetic acid and 5% w/v glycerol), dried in an oven at 50°C overnight and visualized by silver staining (Bio-Rad).

Total RNA from 2 dpf zebrafish larvae was extracted using TRIzol (Sigma). Purified RNA (500 ng) was used to generate cDNA using the Superscript III First-Strand Synthesis System (Invitrogen). Relative qPCR was carried out in 96-well QPCR plates (Roche) using a Roche LightCycler 480 QPCR apparatus with the primers listed in Table S2. Gene expression was normalized to the expression level of glyceraldehyde-3-phosphate dehydrogenase (gapdh) using –ΔΔCt (cycle threshold) values. The efficiency of the primers and absolute qPCR was performed using an external standard curve with calibrated 10-fold serial dilutions of the RT-PCR product. Each measurement contained at least three biological replicates and data were analyzed with JMP statistical software.

Dechorionated embryos at the desired stages were fixed in 4% paraformaldehyde in PBS for 20 min at room temperature, permeabilized with 0.1% Triton X-100 in PBS (PBST), blocked in 10% normal sheep serum (Jackson ImmunoResearch) in PBS for 1 h, and stained with primary antibody for 1 h at room temperature. After the embryos were washed in PBST three times for 10 min each, they were incubated for 1 h with secondary antibodies conjugated with Alexa Fluor 488 or 568 (1:1000; Invitrogen). After the embryos were washed in PBST for 10 min three times, they were incubated with the mounting medium with DAPI (Vector Laboratories). Antibodies against the following targets were used: α-Actinin (1:1000; Sigma, A7811), Myosin heavy chain (F59 and A4.1025, both at 1:10; DSHB), Myomesin B4 (mMAC, 1:10; DSHB), Tropomyosin (CH1, 1:200; Sigma, T9283) and Ttn (MIR, 1:200; a gift from the Siegfried Labeit laboratory). Actin filaments were visualized using Alexa Fluor 568 Phalloidin (1:40, Thermo Fisher Scientific). After image acquisition, the genotypes of all embryos used were determined by PCR and restriction enzyme digestion. For each experiment, more than ten mutant embryos were analyzed to confirm the phenotypes.

After immunostaining, embryos were imaged using a Zeiss Axioplan 2 microscope with Apotome (Zeiss) and an Axiocam Camera (Carl Zeiss) using 20× (Plan-Apochromat) and 63× (Plan-Neofluar) objective lenses combined with a 10× ocular lens. The images were taken and analyzed using AxioVision SE64 4.8.1 software (Zeiss).

Embryos at 2 dpf were fixed in Trump's solution (4% paraformaldehyde and 1% glutaraldehyde in 0.1 M phosphate buffer, pH 7.2) at room temperature for 1 h and then overnight at 4°C. Fixed tissues were then embedded in 2% agarose and subsequently processed and imaged at the Mayo Clinic Electron Microscopy Core Facility using a Philips CM10 transmission electron microscope.

To stop the movement of embryonic somites, embryos at 24 hpf were incubated in E3 water containing either 1% tricaine (Argent Labs) or 10 µM blebbistatin (Sigma) for 24 h (Yang and Xu, 2012; Yang et al., 2014a). To activate muscle contraction, embryos at 24 pdf were incubated in 0.07% bleach to harden the chorion (Zhang et al., 2009) or treated with 0.5 mM galanthamine (Otten and Abdelilah-Seyfried, 2013).

Data were analyzed by JMP software using one-way ANOVA to compare groups. P<0.05 was considered significant. *P<0.05, ***P<0.01 and ***P<0.00001. Data are presented as the mean±s.d.

We thank Beninio Gore and Kashia Stragey for zebrafish maintenance.