Targeted disruption of sp7 and myostatin with CRISPR-Cas9 results in severe bone defects and more muscular cells in common carp

The common carp (Cyprinus carpio) as one of the most important aquaculture fishes produces over 3 million metric tones annually, approximately 10% the annual production of the all farmed freshwater fish worldwide. However, the tetraploidy genome and long generation-time of the common carp have made its breeding and genetic studies extremely difficult. Here, TALEN and CRISPR-Cas9, two versatile genome-editing tools, are employed to target common carp bone-related genes sp7, runx2, bmp2a, spp1, opg, and muscle suppressor gene mstn. TALEN were shown to induce mutations in the target coding sites of sp7, runx2, spp1 and mstn. With CRISPR-Cas9, the two common carp sp7 genes, sp7a and sp7b, were mutated individually, all resulting in severe bone defects; while mstnba mutated fish have grown significantly more muscle cells. We also employed CRISPR-Cas9 to generate double mutant fish of sp7a;mstnba with high efficiencies in a single step. These results demonstrate that both TALEN and CRISPR-Cas9 are highly efficient tools for modifying the common carp genome, and open avenues for facilitating common carp genetic studies and breeding.

Here, we have employed both TALEN and CRISPR-Cas9 techniques to modify genes involved in bone formation such as sp7, a zinc-finger-containing transcription factor expressed in osteoblasts 35 , which can activate pre-osteoblast cells to differentiate into mature osteoblast cells and osteocytes 36,37 . In addition, we have used both techniques to modify other genes involved in bone formation including runx2 (runt-related transcription factor 2), an essential transcription factor to regulate osteoblast differentiation and bone development 38 ; spp1 (secreted phosphoprotein 1), a late osteoblast specific marker involved in bone mineralization 39,40 ; opg (osteoprotegerin), a bone-protecting molecule participated in impairing the osteoclast formation 41 ; and bmp2 (encoding for bone morphogenetic protein 2), promoting expression of runx2 and sp7 and then inducing expression of spp1, osteocalcin and other osteogenic genes 42 . We also have used both systems to modify mstnba, a member of the transforming growth factor-β superfamily and a negative regulator of the skeletal muscle growth 43,44 . Mstn knockout mice displayed 2 to 3 fold increase in both myofiber size and number of muscular cells compared to the wild-type littermates 45,46 . Our results showed that mstnba-CRISPR and sp7a-CRISPR mutated common carps display defects in muscle or bone. We also generated double mutants of mstnba;sp7a in common carp with high efficiencies. Together, these results demonstrate that both TALEN and CRISPR-Cas9 systems are effective genome-editing tools for common carp genetic studies and breeding.

Results
Design of TALEN and CRISPR-Cas9 target sites. We designed TALENs and CRISPR-Cas9/gRNAs targeting genes involved in bone and muscle development. In the bone formation pathway, bmp2a is an up-stream gene, runx2 and sp7 are mid-stream genes, and spp1 is down-stream genes 47 . In addition, opg acts to inhibit osteoclast formation 41 . mstn is involved in muscle formation 48 . Among the selected genes, sp7a, runx2 and mstnba were edited using both TALEN and CRISPR-Cas9 systems to examine efficiencies of these two methods.
The cDNA sequences of zebrafish orthologs in ENSEMBL or NCBI were used to interrogate the common carp genome database (http://www.carpbase.org/login.php) 1 . Common carp DNA sequences selected were then blasted against the common carp protein database (http://www.carpbase.org/login.php) to predict the exon-intron structure. In addition, the selected target sites were blasted back against the common carp genome with reciprocal top blast hit for the target regions to minimize off-target effects.
One pair of TALENs each for runx2, sp7a, mstnba and spp1a was designed. Since the spacer length is very important to mutation efficiencies 49 , 3 pairs of the TALENs designed for runx2a, sp7a and mstnba were longer than 22 bp and 1 pair of TALENs for spp1a was shorter than 22 bp for comparison. In addition, there were restriction endonuclease sites located in the spacers of runx2 and mstnba (Supplementary Figs S1 and S2).
For CRISPR-Cas9 targeting, six genes including sp7a, sp7b, mstnba, runx2, opga and bmp2ab were selected. Most of the gRNAs were designed to target the first exon of the corresponding genes, except for runx2 and sp7b, whose gRNAs target exon 3 and exon 2, respectively (Supplementary Figs S4 and S7). All the target sites include 20-bp sequence and PAM sequence (protospacer adjacent motif NGG). The targeted regions for all these genes have restriction endonucleases sites located either near the PAM sequence or on the PAM for subsequent genotyping (Supplementary Figs S4-S7).
Evaluate mutagenesis efficiencies of TALEN and CRISPR-Cas9 of common carp genes using zebrafish embryos. Phylogenetic analysis showed that common carp have two sp7 genes and four mstn genes, while zebrafish have only one sp7 gene and two mstn genes (Fig. 1A,B), consistent with tetraploidy nature of common carp genome 10 . RT-PCR analysis showed that mstnba is expressed highly in all the tissues/ organs including the maw, liver, gill, eye, heart, brain, gut, testis and muscle, while the other three mstn genes are expressed only in some of these tissues/organs (Fig. 1C-E). We therefore selected mstnba in the following experiments.
Because of the ease of zebrafish embryos to work with, we first used them to evaluate mutagenesis efficiencies of TALENs and CRISPR-Cas9/gRNAs of common carp ( Fig. 2A). Specifically, we examined common carp runx2 and sp7a TALENs and common carp mstnba and sp7a CRISPR-Cas9/gRNAs in zebrafish embryos. For TALEN, 250 ng/ul capped mRNAs of each arm plus 50 ng/ul purified plasmids carrying the common carp genomic DNA fragments containing the runx2-or sp7a-targeted sites were microinjected to one-cell stage zebrafish embryos, respectively. No toxicity was observed for zebrafish embryos at these concentrations of TALENs and purified plasmids (Supplementary Table S1). The common carp DNA fragments containing runx2-or sp7a-targeted sites were PCR amplified from the microinjected zebrafish embryos at 24 hpf. Sequencing analyses revealed that TALENs induced a 6-bp insertion mutation in runx2-targeted site and a 16-bp deletion in sp7a-targeted sites (Supplementary Table S2).
For mstnba-CRISPR-Cas9, 25, 50 and 100 ng/ul gRNAs and 300 ng/ul capped Cas9 RNA plus 100 ng/ul purified common carp plasmid containing the targeted site were microinjected into one-cell zebrafish embryos. No toxicity was observed for zebrafish embryos at these concentrations of Cas9/gRNAs and purified plasmids (Supplementary Table S1). Then PCR products amplified from the microinjected zebrafish embryos at 24 hpf were digested with the StyI enzyme ( Fig. 2A,B). ImageJ analysis revealed that mutation frequencies average approximately 19% (Fig. 2C). Sequencing analysis identified a 6-bp deletion mutation (Fig. 2D). For sp7a-CRISPR-Cas9 Scientific RepoRts | 6:22953 | DOI: 10.1038/srep22953 injected zebrafish embryos, digestion of the PCR product with the HinfI enzyme estimated that the mutation frequency was approximately 2.5% (Supplementary Table S2). Taken together, these results demonstrate that TALEN or CRISPR-Cas9-induced common carp mutagenesis activities can be conveniently evaluated in zebrafish embryos 50 . Mutagenesis efficiencies of TALEN and CRISPR-Cas9 in common carp embryos. TALEN pairs each for runx2, sp7a, mstnba and spp1a were microinjected into the one-cell common carp embryos at a concentration of 250 ng/ul per arm, respectively (Fig. 3A,B). DNA fragments harboring targeted sites were PCR amplified from approximately 20 embryos at 72 hpf from each injected group, and digested with T7E1 enzyme. ImageJ analysis revealed that mutagenesis frequencies are 15.2% for runx2, 36.8% for sp7a, 29.1% for mstnba and 81.5% for spp1a ( Fig. 3C and Supplementary Table S3). Digestion of the PCR products containing the runx2-targeting site or the mstnba -targeting site with HinfI or XbaI also estimated that mutation efficiencies are 1.23% for runx2, and 13.2% for mstnba (Supplementary Figs S1, S2 and Supplementary Table S3). Alternatively, these PCR-amplified fragments were cloned into the PMD-19T vector, respectively. DNA sequencing analysis showed that mutation efficiencies were 5% (1 out of the 20 single clones) for runx2, 20% (4 out of the 20 clones) for sp7a, 27.3% (3 out of the 11 clones) for mstnba and 75% (3 out of the 4 clones) for spp1a ( Fig. 3D; Supplementary Figs S1-S3 and Supplementary Table S3), consistent with the enzymatic analyses. For TALEN-induced mutation efficiencies in common carp, the shorter the spacer lengths lead to the higher mutation efficiencies (Supplementary Table S4), as reported in other species previously 49 .
To examine the CRISPR-Cas9-induced mutagenesis efficiencies, Cas9 mRNAs plus the six individual gRNAs for corresponding genes sp7a, sp7b, mstnba, runx2, opga and bmp2ab were co-microinjected into one-cell common carp embryos, respectively ( Fig. 2A). For sp7a-CRISPR, enzymatic digestion and sequencing analyses showed that 18 out of the 20 clones tested harbor mutations ( Fig. 4A-C), which were reconfirmed with T7E1 enzyme analysis (Supplementary Table S5). For mstnba-CRISPR, the DNA fragment containing the targeted site was PCR amplified from 72 hpf embryos, and digested with the Styl enzyme. The results showed that the mutation efficiency was more than 70% (Fig. 4D,E), consistent with T7E1 enzymatic digestion and DNA sequencing analyses ( Fig. 4F and Supplementary Table S5). In addition, runx2-, sp7b-and opga-CRISPR-Cas9 also produced gene mutations with efficient efficiencies more than 50%, but bmp2ab-CRISPR had relatively low mutation efficiency More muscular cells in mstnba mutant common carps. To generate sp7a-CRISPR and mstnba-CRISPR targeted common carp fish, more than 200 embryos were microinjected for each group. Most of the microinjected and uninjected control fish were put into the ponds when they were 30 days old. In addition, 20 carps each for wild type, sp7a-CRISPR and mstnba-CRISPR were raised separately in the laboratory under the same condition. All the phenotypes were examined from these 20 carps for each group. PCR products amplified with DNAs extracted from caudal fins of one-month-old common carps were digested with restriction enzyme StyI for mstnba-CRISPR. Results showed mstnba-CRISPR had high mutation efficiency in the somatic cells of one-month-old carp, averaging at 56.6% for the 20 carps, similar to the sequencing results in carp embryos (Fig. 5A). The body weight and body length of these microinjected (all the 20 carps in the group including 3# in Fig. 5A with no detectable mutations) and uninjected control common carps were measured at one mpf (month postfertilization), two mpf and three mpf. Unsurprisingly, mstnba-CRISPR common carps grew heavier than uninjected controls in all three time points and longer than uninjected controls in 1 mpf and 3 mpf (Fig. 5B,C). Body weights (Fig. 5B) were significantly increased at 1 mpf (P < 0.03), 2 mpf (P = 0.005) and 3 mpf (P < 0.049) in mstnba-CRISPR carp compared with wild types. Body lengths (Fig. 5C) were also significantly increased in mstnba-CRISPR carp compared with wild types at 1 mpf (P = 0.011) and 3 mpf (P = 0.047). We selected 7 mstnba-CRISPR carps with more than 90% mutation efficiency as the high-rate subgroup, 7 mstnba-CRISPR carps with 10-90% mutated efficiencies as the middle-rate subgroup and 6 mstnba-CRISPR carps with less than 10% mutation efficiency as the low-rate subgroup. Body weights were significantly increased (P = 0.007) in high-rate subgroup compared with wild types, and in high-rate subgroup (P = 0.049) compared with low-rate subgroup, implicating that the body weight exhibits a phenotype-genotype correlation, i.e., the higher the mutation rate (the high-rate subgroup), the heavier the fish compared with the low-rate subgroup (Fig. 5D). We also examined the level of phosphorylated Smad2, which is an essential intracellular transducer for the TGF-β signaling pathway 51 , in dorsal muscles of five-month-old mstnba-CRISPR and wild-type common carps with high mutation rate by Western blotting. Results showed approximately 25% reduced phosphorylated Smad2 in mstnba-CRISPR common carps (P = 0.024). (Fig. 5E,F), implicating that CRISPR-Cas9-induced mutations   impair the Mstn signal pathway of the skeletal muscle of common carp. Moreover, H&E (hematoxylin and eosin) staining of the dorsal muscle of 7# carp in Fig. 5A showed significantly higher numbers of muscle cells and muscle fibers, and significantly increased size of muscle fibers in mstnba-CRISPR carp displays in comparison with uninjected controls (Fig. 5G-K). We also analyzed dorsal muscles of five-month-old mstnba-CRISPR with high mutation rate and wild-type common carps using qRT-PCR. Results showed significant up-regulation of myogenic regulatory factors myf5a (a hyperplasia marker) 51 (P < 0.01) and myogenina (a hypertrophy marker) 51 (P < 0.001) but no difference for myoda (a hyperplasia marker) 51 (P = 0.353) in mutant carps (Fig. 5L). Together, these results indicate that mstnba-CRISPR mutated F 0 carps showed hyperplasia as well as hypertrophy 51 . Sequencing analysis of 7# carp in Fig. 5A showed that six out of the 10 single clones were mutated, and most of the mutation types are frameshifters resulting in truncated peptides, even though mutations are likely mosaic (Fig. 5M). The phenotypes of increasing the body size and muscles of mstnba-CRISPR carp are consistent with those of the Mstn-mutated animals 45 .
Severe bone defects in sp7a mutant common carps. PCR products amplified with DNAs extracted from caudal fins of one-month-old common carps were digested with restriction enzyme HinfI for sp7a-CRISPR. Results showed that 16 out of the 18 sp7a-CRISPR injected carp were mutated, suggesting that almost all the somatic cells of the fin were mutated in the sp7a target site of common carp (Fig. 6A).
It usually takes 3 to 4 years for common carp to grow sexually mature. To examine the possibility of germline transmission, DNAs were extracted from the testis of 5# mstnba-CRISPR carp that has nearly 100% mutation rate based upon analysis of fin-clipped DNAs. Sequencing data show 10 out of the 10 single clones (100%) were mutated ( Supplementary Fig. S12), whereby implicating that these mutations are likely transmitted to the subsequent generation.

Multiplex gene-editing in a single common carp.
Common carp is a tetraploidy species with 100 chromosomes and the high DNA content 2 . The ability to modify duplicated genes or multiple genes of common carp is important for its molecular genetic studies and breeding. To address this issue, sp7a-CRISPR and mstnba-CRISPR were co-microinjected into common carp embryos. PCR products amplified with sp7a and mstnba primers and DNAs extracted from caudal fins of one-month-old injected and uninjected control carps were digested with HinfI or StyI, respectively. The results showed that 16 out of the 21 carps had both gene mutations. Mutation efficiencies was 63.4% for sp7a gene, and 60.1% for mstnba gene, and among them, eight carps had more than 90% efficiencies in these two genes (Fig. 7A,B). These results indicated that double-gene-editing are feasible in common carp with CRISPR-Cas9 in a single step. mstnba mutant carps (H), shown by H&E staining. Scale bar: 0.1 mm. Numbers of muscle fibers (I) and the average area of the muscle fibers (K) of mstnba mutant and wild-type carps, quantified H&E staining images (see G,H) with ImageJ. Approximately 24 stained images each were quantified. (J) Distribution analysis of the fiber area. Values are means ± SEM. (L) Quantitative RT-PCR analysis of myoda, myf5a and myogenina in mstnba-CRISPR (red) and wild-type (blue) carps. RNAs were extracted from dorsal muscles of wild-type (3 samples) and mstnba-CRISPR carps (2#, 6#, 10#) at 5 mpf. Values are means ± SD. n = 3. (M) Mutations in mstnba CRISPR-Cas9 7# carp, revealed by sequencing analysis. The targeted sequence is in green, and the PAM sequences are in blue. Electrophoretic and Western blotting results were cropped from the original images shown in Supplementary Fig. S13I,L. Two-tailed Student's t-test or one-way ANOVA with LSD were conducted, *P < 0.05, **P < 0.01 ***, P < 0.001, and N.S., no significant change.

Discussion
Due to the tetraploidy genome of common carp, it is extremely difficult to obtain the homozygous mutants using the traditional methods such as ENU, TILLING, and retroviral insertion. However, using TALEN and CRISPR-Cas9 methods we have successfully targeted carp genes of interest and observed the specific phenotypes in F 0 generation. We used TALEN to target four genes, sp7a, runx2, mstnba and spp1a and CRISPR-Cas9 to target six genes, sp7a, sp7b, runx2, mstnba, opga and bmp2ab with high efficiencies. With TALEN, we have achieved higher mutagenesis efficiencies with average of 31.83% in common carp than previously reported 3.0% to 12.4% in zebrafish 12 . We also observed that the shorter the spacer length results in the higher mutagenesis in common carp with TALEN, consistent with previous studies 49 . The highest efficiency of common carp TALEN is spp1a-TALEN that has the shortest spacer length of 15 bp. Even though TALEN and CRISPR-targeted F 0 fish were genotypically complex, the highly induced mutagenesis frequencies of somatic mutations still can result in obvious phenotypes in F 0 founders, as reported in mice and zebrafish 24,33 . Hence CRISPR-Cas9 can have the similar effect like Morpholinos 33 to alter the gene functions in common carp. It is applicable to the genes that function in later development or in adulthood such as mature osteoblast genes, circadian clock genes, and ageing genes.
Although sp7a-targeted site with CRISPR-Cas9 has a nearly 100% mutation efficiency in embryos and somatic cells, its mutation types are different among the different cells (Fig. 6J). CRISPR-Cas9-induced mutations are complex in F 0 founder carps. First, these mutations could be mosaic, meaning that the same mutations may be located in different tissues or organs 33,52 , and in some cases the targeted tissues or organs may not carry the mutations and thus display no expected phenotypes 33,52 . Second, several different types of mutations could be in different tissues or organs, for example, the mutated sequences of mstnba, some causing truncated peptides and the other resulting in only deletion or insertion of a few amino acids, are different among different types of cells, as shown in Fig. 5M and Supplementary Fig. S12. Therefore the phenotypes of the body weight vary in the same group of carps (Fig. 5D). The complex nature of CRISPR-Cas9-induced mutations is also responsible for variations of Alizarin Red staining of sp7a-CRISPR F 0 carps (Supplementary Figs S8 and S9) and only 25% reduced levels of phosphorylated Smad2 (Fig. 5E,F). Another aspect of the complexity of phenotypes observed in F 0 founder carps is that there are multiple copies of genes under study, for instance, four mstn genes in common carp, and usually only one gene is mutated (Fig. 1B). A possible way to obtain a consistent phenotype in the F 0 founder carp is to target the important functional domains such as the Z-F (Zinc-Finger) motif of sp7 with high mutation rate, which would induce biallelic mutations.
Sp7/Osterix plays an essential role in differentiation and maturation of osteoblasts and formation of osteocytes but not in cartilage development in mice and medaka 35,53 . We observed that body weight is lighter in sp7a-CRISPR mutated F 0 carps than wild-type control, maybe because the delayed and abnormal osteoblast development; some mutant carps have curved spines, the other mutant carps have abnormal craniofacial bones (Fig. 6D-I and Supplementary Figs S8 and S9). Further, micro-CT analysis showed the bone volume (BV), bone surface (BS), bone volume/tissue volume (BV/TV) and bone surface/tissue volume (BS/TV) are significantly reduced in sp7a-CRISPR carps (Fig. 6N). Together, these results strongly indicate that the observed phenotypes were resulted from the disruption of sp7a gene, consistent with previous studies 35,53,54 . In contrast to the mouse model died after birth 35 , sp7 mutant carps can survive at least 3 months, providing a unique opportunity for investigating roles of Sp7 in bone development. In particular, the inter-muscular bones are shorter in of sp7a-CRISPR carps in comparison with wild types (Fig. 6H) but their number are not changed, demonstrating that Sp7a plays a role in common carp inter-muscular bone development. Even though sp7a-CRISPR mutated carps are not good aquaculture lines due to the abnormal bones and lighter body weight, they are certainly useful for finding carps without or with fewer inter-muscular bones and investigating teleost osteoblast formation.
We have used the CRISPR-Cas9 method to generate mutant carps for multiple genes with high efficiencies. Many phenotypes are derived from mutations in multiple genes rather than a single gene. There are a lot of genes with multiple copies in common carp due to additional round of genome dupolication 55 . Hence mutagenizing double genes or multiple genes with high efficiencies would facilitate generation of multiple-gene-mutated carps. In addition, we can mutagenize multiple copies simultaneously in a single carp and study their functions. Future work will use single gRNA to modify multiple homologous genes as described previously in Xenopus Laevis 56 .
For a large aquatic species such as common carp, it is important to generate mutants with high germline transmission frequencies to reduce the cultivation time, pool space and labor force. Many studies have been focused on increasing the targeting efficiency 18,20 . In this study, we have achieved mutagenesis efficiencies up to 75% for TALEN and up to 100% for CRISPR-Cas9, not only in embryos and juveniles examined, but also in the testis. Our examination of the testis indicated that almost all the testis cells including spermatogonia or spermatocytes are biallelically mutated, implicating that these CRISPR-Cas9-induced mutations are likely transmitted to the subsequent generation. A study of the F 0 monkey placenta showed that the germline cells likely would be modified by CRISPR-Cas9 57 . Zebrafish studies showed that 87.3% mutated efficiency for gol in F 0 somatic cells 33 and 52.7% mutated efficiency for fh in F 0 somatic cells 58 with CRISPR-Cas9 both led to 100% germline transmission rate. Another zebrafish study showed that 12.5% mutated rate for tnikb in F 0 somatic cells with TALEN led to 33.3% germline transmission rate 12 . Similar to these previous studies, we should be able to obtain germline-transmitted carp lines with as high as 100% of somatic mutation rate in TALEN-and CRISPR-Cas9-targeted genes.
In summary, our studies demonstrate that TALEN and CRISPR-Cas9 technologies are effective tools for genetics studies in common carp, which will greatly promote genetic engineering in aquaculture and have the potentials for improving their qualities and economic value in the future. To our knowledge, this is the first report on targeted disruption of endogenous genes in common carp using TALEN and CRISPR-Cas9.

Methods
All animal care and experiments were performed in accordance with the institutional ethical guidelines for animal experiments, and all fish experimental procedures were approved by the Soochow University Committee on Animal Use and Care.
Design of TALEN and CRISPR target sites. The TALEN sites were selected by "TAL Effector Nucleotide Targeter 2.0" to target on the first exon of most genes 59 . An enzyme site in the spacer was selected if it is available for genotyping. Blasting the target sequences against the whole genome DNA database (http://www.carpbase.org/ login.phpweb address) of common carp was performed to avoid the off-target sites.
The gRNAs for CRISPR-Cas9 were designed by "seqbuilder" software according to the 5′ -GGNNNNNNNNNNNNNNNNNNNGG-3′ roles 13 . The first two G are necessary for the T7 RNA polymerase and the end NGG is the PAM. The minmal number of nucleotides (N) is 19 bp depending on the sequence for each gene. A restrictive enzyme near the PAM was also selected if it is available.

Construction of TALEN and synthesis of Cas9 and gRNAs.
The two arms of TALEN were constructed using the "Unit Assembly" method 12 . Simply, customized TALE repeats were inserted into the 5′ -end of the backbone vector (PMD-19T) by double-digesting with either NheI and HindIII or SpeI and HindIII. The final pMD-TALE repeats were confirmed by sequencing with the M13 primer. Then, the DNA fragments digested with NheI and SpeI, were ligated into the TALEN expression vectors-PCS2-PEAS and PCS2-PERR. The final constructs were linearized with NotI and used as templates for TALEN mRNA synthesis with SP6 mMESSAGE mMACHINE Kit (Ambion).
The Cas9 mRNA and gRNAs were synthesized as described previously 29,33 with modifications. Briefly, the Cas9 mRNA was synthesized by in vitro transcription using T7 mMESSAGE mMACHINE Kit (Ambion). The DNA templates of gRNA were generated by PCR with a pair of primers (Supplementary Table S6), and then purified by phenol and chloroform. gRNAs were in vitro transcribed with SP6 Riboprobe Systems (Promega), and purified with lithium chloride and ethanol precipitation. For the injection of the TALEN arms, 250 pg TALEN mRNAs of each arm were co-microinjected into one-cell carp embryos; For CRISPR-Cas9, 300 pg Cas9 mRNA and 25-200 pg gRNA were co-injected into the embryos. Microinjected embryos were grown at room temperature and collected for DNA extraction at 48-72 hpf.
Common carp larval feeding and maintenance. Four dpf-12 dpf larvae were fed with paramecia; then fed with brine shrimps until 25 dpf. 20-90 dpf juveniles were fed with the eel powder for two days, and then were transferred into pools. Those fish to be examined were taken to the laboratory. Each fish was raised in a 3-L tank at 22 °C under 14 h light:10 h dark (14:10 h LD) cycles and fed with brine shrimp and eel powder. The food was fed equivalently in all the groups.

Mutagenesis efficiency analyses in injected embryos. Restriction digestion assay. TALEN-and
CRISPR-Cas9/gRNA-induced mutation efficiencies were examined by the restriction digestion assay as described previously 18 . PCR products were digested with corresponding restriction endonuclease enzymes for 2 h at 37 °C. The intensities of cleaved and uncleaved bands were quantified with the ImageJ software.
T7 endonuclease I (T7EI) assay. 300 ng of purified PCR products were denatured and slowly re-annealed to facilitate heteroduplex formation. The re-annealing procedure includes a 5-min denaturing step at 95 °C, followed by cooling to 85 °C at − 2 °C/sec and further to 25 °C at − 0.1 °C/sec. The re-annealed amplicon was then digested with 10 U of T7E1 endonuclease I (New England Biolabs) at 37 °C for 90 min, and digested amplicon was electrophoresed on a 2% agarose gel. Band intensities were quantified with the ImageJ software.
DNA sequencing. For DNA sequencing, DNA fragments containing the targeted sites were PCR amplified with DNAs extracted from caudal fins of 1-month old carps, and cloned into the PMD-19T vector. Single clone was picked up, verified by PCR and sequenced by Sanger sequencing (GENEWIZ, Inc.).

RNA isolation, RT-PCR and quantitative real-time PCR analyses.
Total RNAs were extracted from the maw, liver, gill, eye, heart, brain, gut, testis and muscle of 3-month-old common carp fishes with TriZol (Invitrogen, CA, USA), and reverse transcribed into cDNAs with Superscript III Reverse Transcriptase (Invitrogen, CA, USA), respectively. RT-PCRs were carried out with the following thermal profile of 94 °C for 5 min and 94 °C, 20 sec; 58 or 60 °C, 20 sec, 72 °C, 15 sec for 38 cycles. Quantitative real-time PCR (qRT-PCR) was performed in an ABI StepOnePlus instrument with the SYBR green detection system (Invitrogen) using a PCR thermal profile with 40 cycles of 10 s at 95 °C and 30 s at 60 °C. All results were normalized to the expression level of the housekeeping gene β -actin. qRT-PCR results are shown as a relative expression level calculated using the 2 −△△CT method 60 . Primers for mstnaa, mstnab, mstnba, mstnbb, myoda, myf5a and myogenina are listed in Supplementary Table S6. Each PCR assay was done with three biological samples. P values were calculated with Student's t test.
Alizarin red staining. Mineralization of calcium deposits was assessed by Alizarin Red S (Sigma) staining conducted as previously described 61 . Three-month-old carps from mutant and control groups were anesthetized with 0.03% Tricaine (Sigma, USA). The stained matrix was examined using a Leica stereomicroscope and was photographed using an Olympus 35-mm camera (Olympus, Tokyo, Japan).

H&E (Hematoxylin and eosin) staining.
Three-month-old carps from mutant and control groups were anesthetized with 0.03% Tricaine (Sigma, USA). Subsequently, carp dorsal muscles were dissected out and fixed in 4% paraformaldehyde at room temperature for 24 hours, and then sectioned and stained with hematoxylin/eosin as described previously 51,62 . Dorsal muscles of carp pelvic fins were counted. Cell number was calculated as the number of fibers per cross-sectional muscle area. The muscle fiber numbers and area were determined with the ImageJ program. The muscle numbers were calculated by the "Cell Counter" features of ImageJ.
Micro computed tomography (Micro-CT). Two-month-old common carp with approximately 3.5 cm of body length were anesthetized with 0.03% Tricaine (Sigma, USA) and gently put into 5 ml centrifuge tubes. The whole body was imaged in a SkyScan 1176 high-resolution micro-CT scanner (Skyscan, Kontich, Belgium) using a 8.8 μm voxel size. The applied X-ray voltage was 45 kV and current was 550 uA. Scans were over 180° with a 0.7° rotation step. Images were reconstructed and binarised with global thresholding using SkyScan CT Analyser software (Version: 1.10.11.0), as described 63 . A region of interest was traced around individual whole body or head models prepared using the "Double Time Cubes" 3D reconstruction method. Cortical bone mineral density (BMD) was estimated by comparing bone density with calibration phantoms of known BMD, scanned at the same time as the head.
Western blotting analysis. Proteins extracted from dorsal muscles of five-month-old common carps were washed with fish water and homogenized in lysis buffer as described previously 60 . Protein samples were separated with sodium dodecyl sulfate-poly-acrylamide gel electrophoresis (SDS-PAGE) using 5% acrylamide stacking gel followed by 10% gradient separation gel and transferred to NC membranes. After blocking with 5% nonfat milk (blocking solution) for 1 h at room temperature. The blocking solution was also used as dilution solutions for antibodies. NC membranes were incubated with primary antibody overnight at 4 °C; anti-phospho Smad2 (1:500, Millipore Corp., MA, USA), anti-Tubulin (1:1000, Cell Signaling Technology, Inc., MA, USA). The next day, membranes were washed in TBST (0.5% Tween 20) four times, each for 5 min, and then incubated for 2 h with goat anti-rabbit HRP-conjugated secondary antibody (1:10000; Santa Cruz Biotechnology, Inc., CA, USA) at room temperature. After washing with TBST four times, 10 min each, membranes were detected by Amersham ECL prime (GE Healthcare). Equal loading of the proteins was confirmed with an anti-Tubulin antibody.
Statistical analysis. Statistical analyses were performed with the unpaired, two-tailed Student's t-test or one-way ANOVA with post hoc LSD method. All statistical analyses were executed with SPSS 16.0 software and P < 0.05 was regarded as a statistically significant difference, *P < 0.05, **P < 0.01, ***P < 0.001.