Two Different Functions of Connexin43 Confer Two Different Bone Phenotypes in Zebrafish*

Fish remain nearly the same shape as they grow, but there are two different modes of bone growth. Bones in the tail fin (fin ray segments) are added distally at the tips of the fins and do not elongate once produced. On the other hand, vertebrae enlarge in proportion to body growth. To elucidate how bone growth is controlled, we investigated a zebrafish mutant, steopsel (stptl28d). Vertebrae of stptl28d/+ fish look normal in larvae (∼30 days) but are distinctly shorter (59–81%) than vertebrae of wild type fish in adults. In contrast, the lengths of fin rays are only slightly shorter (∼95%) than those of the wild type in both larvae and adults. Positional cloning revealed that stp encodes Connexin43 (Cx43), a connexin that functions as a gap junction and hemichannel. Interestingly, cx43 was also identified as the gene causing the short-of-fin (sof) phenotype, in which the fin ray segments are shorter but the vertebrae are normal. To identify the cause of this difference between the alleles, we expressed Cx43 exogenously in Xenopus oocytes and performed electrophysiological analysis of the mutant proteins. Gap junction coupling induced by Cx43stp or Cx43sof was reduced compared with Cx43-WT. On the other hand, only Cx43stp induced abnormally high (50× wild type) transmembrane currents through hemichannels. Our results suggest that Cx43 plays critical and diverse roles in zebrafish bone growth.

Vertebrate bones develop during embryonic stages and grow continuously as body size increases. Because bone shape and size determine body shape, the mechanisms that generate and maintain bone shape are of great interest to biologists. Recent studies have increased our knowledge of bone formation mechanisms. Interactions among three types of cells-osteoblasts, osteocytes, and osteoclasts-play major roles in bone formation and bone remodeling (1)(2)(3)(4)(5)(6)(7)(8).
Zebrafish (Danio rerio) is a model organism for developmental and genetic studies, and several mutants related to bone formation have been identified and analyzed (9 -13). Mutant bone formation phenotypes can be categorized into two types: bone development mutants and bone growth mutants. Genes corresponding to bone development mutants are usually expressed at early developmental stages of bone formation (type 1), and bone malformations are detected when the bones first appear in embryo (14 -16). On the other hand there are several zebrafish bone growth mutants, in which bone development looks almost normal at early stages, but mutant phenotypes appear at late stages (type 2) (17)(18)(19).
In this study, we focused on the stoepsel (stp) mutant fish (20) to study bone morphogenesis in adult stages. The stp mutant was originally isolated by a large scale screen of zebrafish mutants induced by ENU 2 (20). This mutant looks normal until ϳ35 days after fertilization (dpf); however, the vertebrae of the mutant fish become shorter than those of wild type fish at later growth stages. The stp allele is dominant and homozygous lethal. We re-examined the stp phenotypes in detail using modern techniques, and we performed positional cloning to identify the stp gene. Surprisingly, we isolated a mutation in the con-nexin43 gene, which is known as the sof (short-of-fin) gene in zebrafish (21). sof is a well studied type 2 zebrafish mutant, in which fin segments are short in the caudal fin (21). Four sof mutant alleles were isolated by mutagenesis screening, and three single mutations that cause amino acid substitutions in Connexin43 protein were identified (22). In the sof b123 allele, no amino acid substitution was identified in the connexin43 gene, but cx43 expression was reduced in the mutant fish (21,23,24).
Connexin proteins are four-pass transmembrane proteins that are subunits of gap junctions (25). Approximately 20 connexin genes are known from the mammalian genome, and ϳ36 connexin genes are predicted in the zebrafish genome (26,27). Six connexin proteins make a hexamer called a connexon; docking of two connexons on adjacent cells creates a gap junction. Gap junctions allow movement of small molecules (Ͻ1000 Da; e.g. ATP, inositol trisphosphate, ions, etc.) between neighboring cells; a connexon acts as a hemichannel on the cytoplasmic membrane (25). connexin43 is one of the most studied connexin genes because it is linked to several human diseases (28 -30). Oculodentodigital dysplasia is one Cx43-related human disease, which causes small eyes, underdeveloped teeth, and malformed fingers (31,32). To identify the cause of the substan-tial difference in phenotype between the sof and stp alleles of cx43, we performed electrophysiological experiments and found that the growth-dependent malformation of stp vertebrae is likely caused by aberrant hemichannel activity of Cx43 stp rather than reduced gap junction intercellular conductance observed in Cx43 sof .

Experimental Procedures
Animal Care-All experiments in this study were conducted in accordance with the guidelines and approved protocols for animal care and use (approval number FBS-14-002-1) at Osaka University (Osaka, Japan). We used the Tu and AB strains as wild type zebrafish lines. stoepsel (stp tl28d ) mutant was obtained from the Tubingen Zebrafish Stock Center.
Clearing and Measuring of Vertebral Length and Height-Fish samples were fixed with 3.7% formaldehyde and treated with ethanol to remove lipids. Then they were transferred back to H 2 O. Next, samples were incubated in a saturated aqueous solution of sodium tetraborate for 1 day. Then they were incubated overnight in clearing solution (30% saturated aqueous solution of sodium tetraborate with 3 mg/ml trypsin) at 48°C to clear the samples. Transparent samples were dyed with bone dye solution (90 mM KOH, 50 g/ml alizarin red S) for 1 day and then incubated in 90 mM KOH for 1 day. Finally, samples were immersed in glycerol with thymol and preserved.
CT Scanning-A LaTheta LCT-100 (Hitachi Aloka Medical, Ltd.) x-ray micro-CT scanner at Osaka Bioscience Institute and an inspeXio SMX-100CT microfocus x-ray CT system at Shimadzu Corporation were used for taking CT images of bone structure. Fish samples were fixed with 3.7% formaldehyde and stored in PBS. Samples were scanned in air.
Positional Cloning and Sequence Analyses-The stp tl28d/ϩ (TE background) mutant was crossed with the wild type AB strain. TE is a mixture of Tu and TL lines. Here, we used Tu instead of TE. F1 heterozygous fish were backcrossed with wild type parents, and F2 fish were used for genetic mapping. stp was mapped to chromosome 20 by checking simple sequence length polymorphism markers. Then stp was mapped between single nucleotide polymorphisms listed in Table 1, which were identified within chromosome 20 (40.55-41.01 Mb). Gene predictions were derived from Ensemble (Sanger Institute). Candidate gene sequencing was performed using a 3130 Genetic Analyzer (Applied Biosystems).
TALEN-mediated cx43 Knock-out-Transcription activatorlike effector nucleases (TALEN) pairs for cx43 knock-out were designed using the TALEN Targeter program (available at the Cornell University website). TAL repeats were assembled using a Golden Gate TALEN and TAL Effector kit 2.0 (34) with some modifications (35) and cloned into expression vectors pCS2TAL3DD or pCS2TAL3RR (36). The TALEN expression plasmids were linearized by NotI digestion, and then mRNAs were in vitro transcribed using an mMESSAGE mMACHINE SP6 transcription kit (Ambion). Synthetic TALEN mRNAs were injected into one-cell stage embryos (100 pg/embryo each). To test the efficacy of mutagenesis, TALEN mRNA-injected embryos were harvested 1 day after fertilization. The genomic DNA was extracted, and the loci containing the TALEN target sequence were amplified by PCR with a forward primer (5Ј-TTCAAGTGTCACCAAAGTGTCT-3Ј) and a reverse primer (5Ј-CTGCTGGGTATTGCACTTGA-3Ј). TALEN-induced mutations were detected by T7 endonuclease I assay. The most effective TALEN pair was used for cx43 knock-out. The target sequence is 5Ј-TCTGGCTCTCTGTG-CTCTtcatcttccggatccTTGTTCTGGGAACAGCA-3Ј (TALEN binding sequences are shown in uppercase, and spacer sequence is in lowercase).
Measurement of mRNA Expression for cx43 Alleles-To precisely compare the expression levels of the wild type and mutated alleles, which differ at only one nucleotide, we generated cDNA clones from the heterozygous fish and then sequenced them. mRNAs were collected from tail fins of stp tl28d/ϩ fish and stp tl28d /sof b123 fish by RNeasy Protect mini kit (Qiagen), respectively. Then mRNAs were reverse-transcribed using SuperScript III reverse transcriptase (Invitrogen). The cx43 sequence around the stp mutation site was amplified from the cDNAs, and the amplified fragments were cloned with plasmid vector. Then Escherichia coli colonies were picked up randomly, and DNA fragments in the plasmid were sequenced. The frequencies of DNA fragments derived from stp allele or wild type allele in stp tl28d/ϩ fish and DNA fragments from the stp allele and sof allele in stp tl28d /sof b123 were compared. Locations of SNP Markers are on the Zv9 assembly. Both forward (F) and reverse (R) primers were used for sequencing. All SNPs were assessed by sequencing reads.

Preparation of Connexin mRNA and Injection into Xenopus
Oocytes-Connexin cDNAs were amplified using PCR and cloned into pGEM-HeFx plasmid (37). The plasmids were linearized using restriction enzyme and then applied to the mRNA preparation kit (T7 mMessage mMachine; Ambion) according to the manufacturer's protocol. An adult Xenopus laevis female was anesthetized with MS-222, and the ovarian lobes were collected using a surgical knife and forceps. The eggs were treated with collagenase solution (2 mg/ml collagenase I (Sigma) and 2 mg/ml hyaluronidase (Sigma) in OR2 buffer: 82.5 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES, adjusted to pH 7.5 using NaOH) at 18°C for 2 h. Stage V and VI oocytes were collected manually and used for mRNA injection. Then mRNA (0.05 ng for gap junction recording; 5 ng for hemichannel recording) was co-injected with 10 ng of antisense oligonucleotide DNA for Xenopus cx38 into Xenopus oocytes. For the negative control, only antisense oligonucleotide was injected. Oocytes injected with mRNA were incubated overnight at 18°C in 1/2ϫ diluted L15 medium (Sigma) with 2 mM CaCl 2 (HEPES-buffered and adjusted to pH 7.5 with NaOH).
Voltage Clamp Recording-Vitelline membranes of mRNAinjected oocytes were removed manually using forceps in a hypertonic solution (200 mM aspartic acid, 1 mM MgCl 2 , 10 mM EGTA, 20 mM KCl, and 10 mM HEPES, pH 7.5). The oocytes were manually paired, touching at the vegetal poles, while in ND96(ϩ) solution (93.5 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES). The transjunctional current was measured using the dual whole cell voltage clamp technique. Current and voltage electrodes were prepared with a micropipette puller P-1000 (Sutter Instruments) to obtain a resistance of 0.5-1.0 M⍀. The pipette was filled with solution containing 3 M KCl, 10 mM EGTA, and 10 mM HEPES (pH 7.4). Voltage clamp experiments were performed using two Multi Electrode Clamp Amplifiers iTEV90 (HEKA). To measure the transjunctional current, both cells were initially clamped at Ϫ40 mV, and one cell was then subjected to alternating pulses of from Ϫ160 to ϩ80 mV in 20-mV increments. Currents detected in the second oocyte were recorded, and junctional conductance was calculated using the current value at the end of each 3-s pulse (steady state). The conductance was obtained by G j ϭ I j /(V 1 Ϫ V 0 ), and normalized G j values were compared.
Hemichannel Recording-Hemichannel currents were recorded from a single oocyte using a Multi Electrode Clamp Amplifier iTEV90. Modified ND96(ϩ) solution with/without 2 mM CaCl 2 was used as a bath solution. To obtain the hemichannel current, the cells were initially clamped at Ϫ40 mV and then subjected to 5-s voltage steps from Ϫ30 to ϩ60 mV in 10-mV increments.
Western Blotting and Immunofluorescence Staining-Xenopus oocytes, in which the antisense oligonucleotide for Xenopus cx38 and mRNA for cx43, cx43 stp , or cx43 sof were injected, were harvested after overnight incubation. Then each oocyte was used for Western blotting and immunofluorescence staining. Western blotting was performed using anti-Cx43 antibody (Santa Cruz Biotechnology). Anti-␤-tubulin antibody (Cell Signaling) was used for the positive controls. Signals were detected using ECL Western blotting detection kit (GE Healthcare). Im-munofluorescence staining was performed using the anti-Cx43 antibody and Alexa 488 conjugated antibody (Invitrogen). Pictures were taken using a BZ-X700 microscope (KEYENCE) with OP-66835 BZ filter GFP.

Results
Stoepsel (stp tl28d ) Phenotype-The stp mutation was isolated in a large scale screen of zebrafish mutants induced by ENU (20). In the original paper, stp was described as a mutant associated with abnormalities of the axial skeleton (Fig. 1, A and AЈ). stp homozygous (stp tl28d/tl28d ) fish die from edema of the eye and abdomen at ϳ10 dpf (Fig. 2, A-C). Because wild type zebrafish start to form vertebrae along the notochord at ϳ14 dpf, we could not observe the bone phenotype in stp tl28d/tl28d fish. Heterozygous mutants (stp tl28d/ϩ ) had similar body lengths to the wild type at early stages, but during later growth they gradually became shorter relative to the wild type along the anterior-posterior (AP) axis (Fig. 1, A and AЈ). The number of vertebrae appeared similar between the mutant and wild type fish. The body height along the dorsoventral (DV) axis was not affected by the stp mutation (Fig. 1, B and BЈ).
Next, we examined the bone phenotype in detail by microCT scanning (Figs. 1, B-F and BЈ-FЈ, and 3). AP length (Fig. 3A), DV height (Fig. 3B), and the proportions (Fig. 3C) of corresponding vertebrae (identification numbers from anterior to posterior) were compared between wild type and stp mutant fish. Vertebrae in stp tl28d/ϩ fish were obviously shorter along the AP axis than those of wild type fish (Fig. 3A), whereas the head and fins did not differ notably (Fig. 1, B and BЈ).
Each vertebra is composed of a centrum and appendixes, which include a ventral arch ( Fig. 1, C, CЈ, D, and DЈ, red arrowheads), hemal arch ( Fig. 1, C, CЈ, D, and DЈ, green arrowheads), and complex beam structures (Fig. 1, C and CЈ, blue arrowheads). The centrum appears at ϳ14 dpf. The appendixes form after centrum growth. In a wild type centrum, an hourglass-like structure was seen along the AP axis (Fig. 1E). On the other hand, in stp tl28d/ϩ mutants, the hourglass-like structure was distorted and sometimes filled with clots ( Fig. 1, DЈ and EЈ). Although the centrum was malformed in the stp tl28d/ϩ mutant, the hemal arch (Fig. 1, CЈ and DЈ, green arrowheads), neural arch (Fig. 1, CЈ and DЈ, red arrowheads), and beam structures (Fig. 1CЈ, blue arrowhead) were approximately normal. These data indicate that the stp mutation mainly affected the formation of the centrum of the vertebrae.
Next, to reveal when vertebral malformation occurred, we investigated the time course of centrum growth (Fig. 1, G-I).
We measured the AP length (Fig. 1G) and DV height (Fig. 1H) of the centrums of caudal (No. 1-5) vertebrae (Fig. 3). Graphs in Fig. 1 (G-I) show that early vertebral development (35 dpf) was not affected by the stp mutation; however, after 35 dpf, centrum growth was abnormally low along the AP axis (Fig. 1G), but not along the DV axis (Fig. 1H). This suggests that the stp mutation affects the mechanism that maintains appropriate vertebral shape during growth.
Cloning the Gene Affected by the stp Mutation-To identify the gene affected by the stp mutation, we next performed positional cloning and identified a critical region of 460 kb in chromosome 20 (Fig. 4A). This region contains eight genes, of which six genes code for connexins (cx). The coding sequences of the stp allele were sequenced, and their amino acid sequences were compared with the wild type. Among the eight genes, only the cx43 gene had a point mutation (G233T) specific to the stp allele, which altered the amino acid sequence (tryptophan 3 leucine: W78L) (Fig. 4B). The amino acid substitution occurs in the second transmembrane domain of the Cx43 protein (Fig.  4C). A BLAST search revealed that the region around the mutation was highly conserved in vertebrates (Fig. 4D). To confirm cx43 expression in vertebrae, we generated transgenic fish that expressed mCherry under the control of the cx43 promoter (Fig. 4, E, EЈ, F, and FЈ). This transgenic fish showed mCherry expression at the growing edges of the centrum, where the defect resulting in the stp phenotype was expected to occur (Fig.  4, E and EЈ), and at the junctions between fin bones (Fig. 4, F  and FЈ).
stp Gene Knock-out by TALEN-To confirm that the mutation in cx43 is responsible for the stp phenotype, we knocked out the stp allele of cx43 using TALEN. The TALEN construct was designed to target the middle region of the first extracellular loop of Cx43, just at the 5Ј side of the stp mutation (Fig. 5A). TALEN mRNA was injected into embryos of stp tl28d/ϩ fish, and we isolated the F1 fish that had a frameshift in the stp allele (Fig. 5B). Because stp (Fig. 5, D and DЈ) is a dominant mutation, knock-out of the stp allele in stp tl28d/ϩ fish should rescue the phenotype. The heterozygous fish, which had the frameshifted stp allele and the wild type cx43 allele (KO/ϩ; Fig. 5, E and EЈ), did not show any difference in centrum structure compared with wild type siblings (Fig. 5, C, CЈ, and   H), indicating that knock-out of the stp allele rescued vertebral structure. Vertebrae in stp mutants were 12% shorter than in the wild type, but vertebral lengths were similar in KO/ϩ fish and the wild type (Fig. 5H). This result confirmed that the mutation in cx43 (Cx43W78L) is responsible for the stp phenotype.
We also examined the lethality of the stp homozygote. If the lethality was caused by the loss of normal Cx43 protein, survival of stp tl28d/tl28d larvae should be rescued by introducing exogenous wild type cx43. We generated Tg(cx43pro-cx43) stp tl28d/tl28d fish by introducing a wild type cx43 gene. Tg(cx43pro-cx43)stp tl28d/tl28d fish did not die at larval stages and grew into adults (Fig. 2, D and E). Interestingly the rescued fish showed a unique phenotype (Fig. 2E): they had very short body lengths compared with the wild type fish (Fig. 2D) and stp/ϩ heterozygous fish (Fig. 1AЈ). These results suggest that the lethality of stp tl28d/tl28d results from the loss of cx43 function. In addition this transgenic fish also suggested that Cx43W78L has a dominant negative effect on wild type Cx43 because the transgene, cx43pro-cx43, did not rescue the stp phenotype perfectly (Fig. 2E).
Another cx43 Mutant, Short Fin-The zebrafish mutant, sof, which exhibits short fin bone segments, also involves cx43 (21). There are four reported sof alleles. All of these alleles show similar recessive phenotypes. Three alleles have amino acid substitutions in cx43, and one allele causes low expression of cx43. It is very puzzling that these alleles of the same gene do not cause shortening of vertebrae. We first examined the amount of cx43 mRNA between stp and sof b123 (low expression allele) and found that the expression of mRNA from the stp allele was comparable with that from the wild type cx43 allele (Fig. 6A); however, expression of mRNA from the sof 123 allele was less than one-tenth that of the stp allele (Fig. 6B).
We then investigated the shape of sof b123 mutant fish to determine whether they had morphologically normal vertebrae. As shown in Fig. 5 (F, FЈ, and H), vertebrae of both sof b123/b123 and sof b123/ϩ fish are normal in size and shape. These data indicated that a simple decrease of cx43 mRNA level did not affect vertebral growth. However, the vertebrae of stp/sof b123 fish are drastically shorter than those of stp tl28d/ϩ fish (Fig. 5, G, GЈ, and  H), suggesting that the wild type Cx43 protein compensates for the dominant effect of the stp allele. When the amount of normal Cx43 protein was decreased, the stp phenotype (short vertebrae) became more evident.
Next, we measured the length of fin ray segments for fish carrying the different cx43 alleles (Fig. 5I) to see whether the stp allele causes defects in fin bones. Segment lengths of sof b123/b123 and sof b123/ϩ fish were 64 and 87% of wild type, respectively. This showed that fin ray length is sensitive to the amount of  Cx43 protein. stp tl28d/ϩ fish also had shorter fin ray segments than wild type fish, but the difference was only 5% (Fig. 5I,  arrowhead 1). The difference of the fin ray segment length between stp/sof b123 and sof b123/ϩ fish was 7% (Fig. 5I, arrowhead 2). Therefore, the stp allele also causes the short fin bone phenotype, but the severity is less than that of the sof alleles.
Functional Analysis of Cx43 stp -As described above, the influence of the stp allele is severe in vertebrae but mild in fin bones. On the other hand, the influence of the sof b123 allele is undetectable in vertebrae (Fig. 5H) and severe in fin bones. This apparent discrepancy suggests that the Cx43 protein has more than one distinct function. It is known that there are two distinct functions of connexins: formation of gap junctions or hemichannels. We examined both functions of Cx43 sof and Cx43 stp electrophysiologically. There are three known sof missense mutations: Cx43F30V, Cx43P191S, and Cx43F209I. A previous paper reported that all of them have similar hemichannel and gap junction activity (22). Here, we used cx43 j7e2 (which codes for Cx43P191S) as a representative sof mutant.
To examine gap junction activity of these Cx43 mutants, we performed voltage clamp current recording using Xenopus oocytes. In this study, we injected only 0.05 ng of mRNAs for Connexin proteins into each oocyte because we detected very large transjunctional currents when we injected 5 ng of mRNA for wild type Cx43 (22). Fig. 7A shows examples of junctional currents induced by wild type Cx43 (Cx43-WT) and Cx43W78L (Cx43 stp ). Cx43W78L (Cx43 stp ) showed reduced junctional currents between the oocytes compared with Cx43-WT (Fig. 7B). On the other hand Cx43P191S (Cx43 sof ) did not show significant junctional current in this experimental condition (0.05 ng of mRNA/oocyte).
We next examined the hemichannel activities of wild type and mutant Cx43s. Panels A and B of Fig. 8 show current traces of Cx43-WT and Cx43W78L (Cx43 stp ), and panels C-G summarizes the current-voltage relationship of these hemichannels. As reported previously, Cx43-WT and Cx43P191S (Cx43 sof ) did not increase current in comparison with the negative control (Fig. 8, C, E, and G). On the other hand Cx43W78L (Cx43 stp ) induced extraordinarily high hemichannel activity (Fig. 8D), which was more than 50-fold greater than that of Cx43-WT. We examined the hemichannel activity of Cx43W78L (Cx43 stp ) co-expressed with Cx43-WT, mimicking the situation in stp tl28d/ϩ fish in vivo. The hemichannel activity of Cx43W78L (Cx43 stp ) co-expressed with Cx43-WT was also very high (Fig. 8F). Assuming that wild type and mutant connexins interact heteromerically, these results suggest that heteromeric hemichannels of Cx43-WT and Cx43W78L (Cx43 stp ) in stp tl28d/ϩ fish would have increased activity. These results suggest that the hemichannel activity of Cx43P191S (Cx43 sof ) is similar to that of Cx43-WT, whereas that of Cx43W78L (Cx43 stp ) is abnormally increased.
To determine the protein expression and functional changes of connexin proteins in Xenopus oocytes, we performed Western blotting and immunofluorescence staining. The bracket in Fig. 9A indicated that Cx43 and its mutants were expressed in Xenopus oocyte, although several bands were detected in the Cx43W78L (Cx43 stp ) lane. The bands detected in Cx43W78L imply that the W78L mutation causes a protein modification defect in Cx43. Further study is needed to determine whether this modification of Cx43 is related to the stp phenotype. 50-kDa bands in Fig. 9A are due to nonspecific binding of the anti-Cx43 antibody used in this study. The arrow in Fig. 9AЈ indicates the expression of ␤-tubulin (positive control). Protein localization of Cx43 and its mutants were also examined, and membrane localization was detected by immunofluorescence staining (Fig. 9, B-E and BЈ-EЈ). Compared with the wild type Cx43 (Fig. 9B), the signals of Cx43W78L (Cx43 stp ) (Fig. 9D) and Cx43P191S (Cx43 sof ) (Fig. 9E) detected at the cell membranes looked weak (22). For the negative controls (NC), only antisense oligonucleotide for Xenopus cx38 was injected into Xenopus oocytes (Fig. 9, A, AЈ, and C).

Discussion
In this paper we analyzed a zebrafish mutation that disrupts normal vertebra formation. Homozygous (stp tl28d/tl28d ) fish die; heterozygous (stp tl28d/ϩ ) fish grow to adults but show a specific phenotype of shorter vertebrae along only the AP axis. Because

Connexin43 Functions in Vertebra and Fin of Zebrafish
this phenotype is not seen in young fish (ϳ35 dpf) and becomes evident only after ϳ60 dpf, we deduced that the mutation caused some defect in the control of bone growth. To identify the gene altered in the stp mutant, we performed positional cloning and TALEN knock-out of the genomic sequence and found that a single base mutation in cx43 causes the phenotype.
Curiously, a previous paper reported that another mutation in cx43 is responsible for the sof phenotype, in which fins are short but vertebrae are normal. To determine why these alleles of the same gene differed in their effects, we investigated two different functions of connexins: as gap junctions and hemichannels. Both Cx43 stp and Cx43 sof showed reduced gap junction currents relative to wild type, which confirmed that the sof phenotype was caused by the defect in gap junction activity, as suggested by Hoptak-Solga et al. (22). On the other hand, Cx43 stp hemichannels showed abnormally high (50-fold higher than wild type) conductance, whereas Cx43 sof hemichannel conductance was almost the same as Cx43-WT. These results are consistent with the hypothesis that extremely high hemichannel activity causes the vertebral phenotype of the stp allele.
Although the mechanism by which gap junctions and hemichannels affect different bones is unknown, we assume that it is related to the difference in the bone forming process between vertebrae and fin bones. Fins are mechanically supported by fin rays arrayed in parallel. Each fin ray is composed of short segments of bone called lepidotrichia. Each segment is almost the same size along the AP axis (for tail fins). When the fin grows, a new fin ray segment is made at the tip of each fin ray. The newly made segments are nearly as long as the old segments along the AP axis and do not grow longer with time or with increasing fin length. On the other hand, vertebrae initially appear at ϳ14 days post fertilization and continuously enlarge during the growth of the fish. Therefore, we infer that Cx43 gap junctions are involved in initial bone formation, and Cx43 hemichannels are involved in later bone growth.
One proposed mechanism of mechanical stress sensing by osteocytes is as follows. When bones are exposed to mechanical stress, they deform slightly, causing movement of the fluid surrounding the osteocytes. This movement of fluid causes shear stress that induces effusion of prostaglandin E 2 through hemichannels (38). Prostaglandin E 2 promotes osteoclast differentiation by down-regulating Osteoprotegerin (39), a protein that blocks RANKL-RANK interaction and promotes osteoclast differentiation.  (G233T). B, Mutated sequences produced by TALEN. Three F1 lines had frame-shifting mutations, which were (Ϫ15 bp/ϩ10 bp), (Ϫ6 bp/ϩ7 bp), and (Ϫ7 bp). The (Ϫ15 bp/ϩ10 bp) line was used for the stp gene knock-out investigation. C-G, vertebral structure of wild type (C), stp/ϩ (D), KO/ϩ (stp allele knock-out) (E), sof b123/b123 (F), and stp/sof b123 (G) fish. Cleared, alizarin red S-stained specimens were made from 70 dpf fishes. KO/ϩ fish had a wild type allele and a knock-out stp allele. sof b123/b123 fish showed normal vertebra structure, but stp/sof b123 fish had markedly shorter vertebrae. H and I, six mutant fish were randomly chosen and compared with six of their wild type siblings at 70 dpf. H, length to height ratio of vertebral centrums. I, length of fin ray segment. The segments second from the tail bud were measured, and average length was calculated. Arrowheads 1 and 2 indicate pairs of fish genotypes that differ in the presence or absence of the stp allele. Error bars, S.E. Scale bars, 500 m in CЈ-GЈ. stp/ϩ, stp tl28d/ϩ ; sof/sof, sof b123/b123 ; stp/sof, stp/sof b123 . FIGURE 6. cx43 mRNA expression ratios in mutant fish. A, cx43 mRNA expression ratio in stp tl28d/ϩ fish, normalized to WT fish. 39 -48 colonies were sequenced for each fish (n ϭ 6). B, cx43 mRNA expression ratio in stp/sof b123 fish, normalized to stp fish. 28 -48 colonies were sequenced for each fish (n ϭ 5). Error bars, S.E. **, p Ͻ 0.01. stp/ϩ ϭ stp tl28d/ϩ ; stp/sof ϭ stp/sof b123 . We hypothesize that the stp allele reduces vertebral elongation by mimicking high mechanical stress. We found that cx43 expression in zebrafish vertebrae is highest in the growing edges (Fig. 4, E and EЈ). Because these parts of the vertebra contact the adjacent vertebra, we expect that mechanical stress is highest there. Therefore, one possible explanation for the effect of the stp mutation on vertebral shape is that it enhances hemichannel activity, allowing more prostaglandin E 2 effusion. This would mimic the effects of high mechanical stress and lead to increased osteoclast differentiation and, hence, increased rates of bone breakdown at the edges of the centrum.
The major finding of our study is that there are two distinct functions of Cx43 in bone formation, one of which affects early development only and one of which affects subsequent growth only. These two functions correlate with differences in the electrophysiological properties of gap junctions and hemichannels among cx43 alleles. We hypothesize that the differences in gap