hdac4 mediates perichondral ossification and pharyngeal skeleton development in the zebrafish

Zebrafish husbandry

AB strain wild-type (WT) zebrafish were originally obtained from the Zebrafish International Resource Center (ZIRC, Eugene, OR). Fish were reared and maintained at 28.5 °C on a 14 h on/10 h off light cycle. Fish were fed as previously described (Wasden, Roberts & DeLaurier, 2017). Maintenance and use of zebrafish followed guidelines from ZIRC, the Zebrafish book (Westerfield, 2007), and the Institutional Animal Care and Use Committee (IACUC) of the University of South Carolina Aiken (approval number 010317-BIO-01). Fish were staged using criteria for staging development up to 3 dpf (Kimmel et al., 1995). For 7 dpf samples, at least 95% of fish within a clutch had to show swim bladders in order to be included in analysis.

Generation of CRISPR lines

The CRISPR/Cas9 procedure was based on previously described methodologies (Hwang et al., 2013). CHOPCHOP (http://chopchop.cbu.uib.no/) was used to design a guide RNA (gRNA) sequence targeting exon 5 of hdac4 (ensembl ENSDART00000165238.3), 5′ of the Mef2c binding site (located in exon 6) and histone deacetylase domain, with minimal potential off-target binding (Montague et al., 2014; Labun et al., 2016). An hdac4-specific oligonucleotide was designed containing a 20 bp T7 promoter sequence, 20 bp of target sequence (GGAGCGTCATCGACAGGAGC), followed by a 20 bp scaffold overlap sequence as described (Bassett et al., 2013). This oligonucleotide was annealed to a scaffold oligonucleotide containing the tracrRNA stem loop sequence using Phusion PCR (New England Biolabs, Ipswich, MA, USA) to produce a 120 bp template. Template DNA was column purified (DNA Clean & Concentrator kit; Zymo Research, Irvine, CA) and was used to synthesize RNA using T7 polymerase (MAXIscript T7 Transcription kit; Thermo Fisher, Vitnus, Lithuania). Cas9 mRNA was synthesized from pCS2-nCas9n (Addgene, Cambridge, MA). The plasmid was linearized with NotI-HF (New England Biolabs, Ipswich, MA), column purified (Zyppy Plasmid Miniprep kit; Zymo Research, Irvine, CA, USA), and mRNA was synthesized (mMessenger mMachine SP6 kit; Thermo Fisher, Vitnus, Lithuania). Column-purified hdac4 gRNA and nCas9n mRNA (RNA Clean & Concentrator kit; Zymo Research, Irvine, CA) were co-injected into one cell-stage embryos. Each embryo was injected with approximately 3nl of a 5ul mix containing hdac4 gRNA (∼60 ng/microliter), nCas9n mRNA (∼160 ng/microliter), and phenol red as a marker. Unfertilized or dead embryos were removed from the dish at the end of the first day of injection and on subsequent days.

Identification of founders and generation of mutant lines

At 36 hpf, approximately 20% of injected embryos (abnormal and normal-looking) were pooled into groups of 5 fish per tube, lysed using HotShot (Truett et al., 2000), and PCR was performed using genomic primers flanking the site of potential mutation. PCR products were gel-purified and digested using T7 endonuclease (New England Biolabs, Ipswich, MA, USA) to identify mismatched DNA indicating potential founder lines (Hwang et al., 2013). Siblings of fish (approximately 40 fish) with positive T7 results were reared to adulthood and used as founder (F₀) lines for subsequent experiments. Three F₀ fish demonstrated germ line transmission to offspring, and F₁ lines were generated from these founders by out-crossing founders to AB wild-types. Adult F₁ fish were identified as heterozygous carriers of potential mutations using PCR and T7 endonuclease digest on amputated tail DNA. Heterozygous F₁ fish were outcrossed to generate F₂ lines, and F₂ lines were intercrossed to produce homozygous mutants. PCR products from potential mutants and wild-type siblings were sequenced (Eurofins, Louisville, KY, USA) and genomic sequences were compared to wild-type siblings to identify mutations (Geneious, version 8, http://www.geneious.com) (Kearse et al., 2012). Mutations were confirmed by synthesis of cDNA from mRNA (RevertAid First Strand cDNA Synthesis kit; Thermo Fisher, Vitnus, Lithuania) and then amplified by PCR using an exon 3 forward primer 5′-gccactggaacttctcaagc-3′ and an exon 6 reverse primer 5′-gcagtggttgagactcctct-3′ (Tm = 58 °C × 40 cycles or Touchdown PCR, Tm = 72–65 °C × 15 cycles followed by Tm = 64.5 °C × 20 cycles). PCR products were column purified as described above and sequenced to confirm the mutation. Heterozygous F₂ and F₃ carriers of mutant alleles were intercrossed to produce wild-type, heterozygote, and mutant offspring. In order to test for the influence of maternal hdac4 on development, maternal-zygotic mutants were generated using the hdac4^aik3 line by crossing homozygote mutant females with heterozygote males. Female mutants could not be generated using the hdac4^aik2 line, although mutant males survived.

Genotyping adults and larvae

Amputated tails from adult fish, amputated tails from stained larvae, or whole larvae were genotyped using Hotshot lysis as described above. An 822 bp region of exon 5 spanning the site of mutation in hdac4 was PCR-amplified using an intron 4–5 forward primer 5′-atgttctccctgtgttggtg-3′ and an intron 5–6 reverse primer 5′-gctgtatttccgctcatgtg-3′ (Tm = 58 °C, ×40 cycles). PCR products were run on a 2% agarose gel at 60V for 5–6 h to produce band separation sufficient to distinguish heterozygote (2 bands) fish from wild-type and mutant fish (both 1 lower and upper band, respectively).

Alcian Blue and Alizarin Red histological stain

Alcian Blue and Alizarin Red staining to label cartilage and bone was performed as described (Walker & Kimmel, 2007). Briefly, 7 dpf zebrafish were fixed for 1 h in 2% paraformaldehyde/1X PBS, washed in 50% ethanol/H₂O, and stained overnight rocking at room temperature in 0.04% Alcian Blue (Anatech, Battle Creek, MI, USA)/0.01% Alizarin Red S (Sigma-Aldrich, St. Louis, MO, USA)/10 mm MgCl₂/80% ethanol. On day 2, fish were rinsed in 80% ethanol/10mM MgCl₂/H₂O for several hours, then washed in 50% and 25% ethanol/H₂O, bleached using 3% H₂O₂/0.5% KOH for 10 min, then cleared in 25% glycerol/0.1% KOH, and incubated in 50% glycerol/0.1% KOH overnight, rocking at room temperature. Skeletal preparations were stored covered at 4 degrees Celcius.

mRNA in situ hybridization

Single non-fluorescent and double fluorescent mRNA in situ hybridizations were performed as described (Talbot, Johnson & Kimmel, 2010; Thisse & Thisse, 2014), using probes for hdac4, runx2a, runx2b, sp7, and sox9a (DeLaurier et al., 2010; Huycke, Eames & Kimmel, 2012) using larvae from the hdac4^aik3 line. Fish were genotyped after staining. All mutant and wild-type larvae were used for confocal imaging and image analysis. No larvae were excluded from analysis with the exception of larvae that were damaged during processing or imaging. For mRNA in situ hybridizations of embryos from 2-cell to 90% epiboly stages, all embryos were collected from the same clutch of fertilized eggs on the same day, and fixed in 4% paraformaldehyde at specific stages. All embryos were pooled for in situ hybridization and developed for staining simultaneously. The reaction was stopped at the same point for all specimens. Specimens used for in situ hybridization were stored covered, in 1X PBS, at 4 degrees Celcius.

Imaging, image analysis, and statistics: skeletal preparations

Alcian Blue and Alizarin Red stained specimens were dissected and flat mounted on microscope slides and imaged on a compound microscope. The right pharyngeal skeleton was flat mounted for each specimen unless it was damaged or defective, in which case the left side was used. Among all genotyped fish, the extent of ossification of the ceratohyal was scored after removing sample identification and genotype information to ensure unbiased analysis. The region of ossification was scored for the evenness of the border of Alizarin red stain (regular border, irregular border), and for the approximate proportion of total area of the element that was stained (small area = less than approximately 20% or less of the total area, large area = greater than approximately 20% of the total area). In total, 94 fish were scored for the hdac4^aik2 line (39 wild-type, 23 heterozygotes, and 32 mutants), and 96 fish were scored for the hdac4^aik3 line (22 wild-type, 44 heterozygotes, and 30 mutants).

In order to quantitate the levels of ossification in wild-type, heterozygous, and mutant fish, the areas of flat mounted ceratohyal and hyosymplectic cartilages were measured using pixel area of cartilage and bone using ImageJ 1.51m9 (Schneider, Rasband & Eliceiri, 2012). In total, 93 fish were measured for the hdac4^aik2 line (38 wild-type, 23 heterozygote, 32 mutant) and 61 fish were measured for the hdac4^aik3 line (17 wild-type, 28 heterozygote, 16 mutant). The area of bone as a ratio of cartilage area was used a measure of ossification in statistical analysis (SAS 9.4; SAS Institute, Carey, NC, USA). Total area of the element was used as a method to control for differences in ossification due to subtle growth/stage differences between individual fish. For the hdac4^aik2 line, for the ceratohyal, and for the hdac4^aik3 line, for the hyosymplectic, measurements of “bone area” and “total area” met assumptions of normality and homogeneity of slope assumptions for analysis of covariance (ANCOVA). For the hdac4^aik2 line, for the hyosymplectic, and for the hdac4^aik3 line for the ceratohyal, there was no significant relationship between measurements of “bone area” and “total area” and so there was no reason to apply ANCOVA, and conventional ANOVA was applied. In the case of application of ANOVA, datasets met assumptions of normality and equal variance between comparison groups (Shapiro–Wilk normality test, Hartley’s F_max-test, SAS 9.4; SAS Institute, Carey, NC, USA).

Imaging, image analysis, and statistics: mRNA in situ hybridizations

Specimens used for fluorescent in situ hybridization were mounted ventral-side down on glass coverslips and imaged using an inverted Leica SPEII confocal microscope (Leica Microsystems, Buffalo Grove, IL, USA). In total, 21 4 dpf larvae stained using the runx2a probe were imaged (10 wild-type, 11 mutant), 17 4 dpf larvae stained using the runx2b probe were imaged (8 wild-type, 9 mutant), and 22 4 dpf larvae stained using the sp7 probe were imaged (12 wild-type, 10 mutant). For runx2a, larvae from two stage-matched clutches were combined for analysis. For runx2b and sp7, only siblings from a single clutch were analyzed. Each sample was scanned to produce an approximately 80 z-slice stack. Confocal settings were kept exactly the same between imaging sessions. After scanning of all samples was complete, stacks were renamed (removing sample identification and genotype information) and randomized by a participant not involved in this study to ensure unbiased analysis of data. For all image stacks, gene expression was measured as the maximum length of expression (in microns) on the anterior and posterior margins of the right and left ceratohyal cartilages (Leica Application Suite X (LAS X) software 1.8.0.13370; Leica Microsystems). Measurements were recorded twice, on separate days, and data from the two sessions were averaged to produce a final measurement for each region (i.e., left posterior, left anterior, right posterior, right anterior) of the ceratohyal for each sample. All four regions were averaged to produce a final average length of expression of each gene for each sample. Measurements for each genotype within each gene study were assessed for normality and equal variance (Shapiro–Wilk normality test, Hartley’s F_max-test, SAS 9.4; SAS Institute, Carey, NC). All datasets met assumptions of normality and groups that were compared had equal variance, so ANOVA was performed.

RNA-Seq data

We used the RNA-seq expression atlas data for zebrafish (https://www.ebi.ac.uk/gxa/experiments/E-ERAD-475/Results) to establish expression of hdac4 transcripts per million transcripts (TPM) between zygote (1-cell) and 5dpf-stage larvae.

Hdac4 mutants have a frameshift lesion

Using CRISPR/Cas9 to target exon 5 of hdac4 (Fig. 1A), we induced frameshift mutations in two individual fish that were used to generate mutant lines. The hdac4^aik2 allele has a 19 bp insertion three bases upstream of the protospacer adjacent motif (PAM) site associated with Cas9 binding and cleaving of DNA. The hdac4^aik3 allele has a 2 bp insertion, followed by retention of 7 bp of the wild-type sequence, followed by a 27 bp insertion one base pair upstream of the PAM site (Fig. 1B). In both cases, frameshifts were induced by insertion of nucleotides into exon 5, resulting in aberrant amino acids being added to the protein sequence (Fig. 1C). In both mutant lines the frameshift is predicted to cause the loss of the Mef2c binding domain and premature stop codons resulting in truncated proteins 174 aa (hdac4^aik2) and 181 aa (hdac4^aik3) in length (Fig. 1D). Frameshifts were detected in mutant cDNA compared to wild-type cDNA using primers spanning exons 3–6, and there was no evidence of splice variants or exon skipping detected on agarose gels for either mutant (Figs. 1E and 1F). In the case of hdac4^aik3, a larger band was detected along with the band of expected size (Fig. 1F, indicated by asterisk). This band was excised and sequenced and was found to have an identical sequence to the band of the expected size. We interpret that this larger band is the product of heterodimers of our PCR product or a slower running single-stranded DNA product and not a splice variant or other genomic feature within the mutant. In both the hdac4^aik2 and hdac4^aik3 lines, adult fish and embryos were genotyped using intronic primers spanning exon 5. The larger mutant band (841 bp hdac4^aik2, 858 bp hdac4^aik3) can be distinguished from the wild-type band (822 bp), and heterozygotes show both bands (Figs. 1G and 1H). At 7 dpf and earlier, hdac4^aik3 mutants have no discernable external abnormalities compared to wild-type siblings (Figs. 1I and 1J). Mutant fish form swim bladders, feed, and grow normally into adult fish. Mutant fish from the hdac4^aik2 line also showed no overt external abnormalities compared to wild-type siblings (data not shown).

Mutants have increased ossification of pharyngeal cartilage

Examination of Alcian Blue and Alizarin Red-stained specimens reveal that mutants from the hdac4^aik2 and hdac4^aik3 lines showed a greater extent of ossification of the ceratohyal cartilage compared to wild-type siblings (Figs. 2B–2G, hdac4^aik2 only shown in Figs. 2D–2G, hdac4^aik3 not shown). Wild-type fish had a smaller area of bone stain localized to the mid-shaft of the ceratohyal, usually appearing first at the dorsal margin of the cartilage and spreading ventrally, and this region of bone had even, regular, borders (Figs. 2D and 2E, see Figs. 2B and 2C for frequencies). Among mutants, a larger area of bone was observed at the mid-shaft of the ceratohyal, and this area of bone showed irregular borders (Figs. 2F and 2G, see Figs. 2B and 2C for frequencies). Among heterozygotes, the amount of bone area on the ceratohyal appeared as a mixture of the small and large areas observed in wild-types and mutants, and showed both regular and irregular borders (not shown, see Figs. 2B and 2C for frequencies). No other defects were detected in the pharyngeal skeleton or neurocrania in larvae analyzed at this stage.

For both the ceratohyal and hyosymplectic, the area of bone and the area of cartilage were measured, and the amount of ossification was calculated as the ratio of bone to cartilage present (Figs. 2H–2K). For the hdac4^aik2 line, ANCOVA revealed that the effect of genotype on the area of ossification of the ceratohyal was significant (F = 4.01; df = 2,89; p = 0.0215). Tukey’s multiple comparisons showed that mutants had significantly more bone than wild-types (p = 0.0057), but heterozygotes were not significantly different from either mutants (p = 0.1751) or wild-types (p = 0.2488) (Fig. 2H). For the hdac4^aik3 line, ANOVA revealed that the effect of genotype on the area of ossification of the ceratohyal was also significant (F = 7.77; df = 2,58; p = 0.001). Tukey’s multiple comparisons showed that mutants had significantly more bone than wild-types (p = 0.0002), that heterozygotes also had significantly more bone than wild-types (p = 0.0134), and that there was no significant difference between heterozygotes and mutants (p = 0.0697) (Fig. 2I). For the hdac4^aik2 line, ANOVA revealed no significant effect of genotype on area of ossification of the hyosymplectic (F = 2.3; df = 2,90; p = 0.106) (Fig. 2J). For the hdac4^aik3 line, ANCOVA also revealed no significant effect of genotype on area of ossification of the hyosymplectic (F = 2.48; df = 1,56; p = 0.0928) (Fig. 2K).

hdac4 is expressed in regions of the pharyngeal skeleton consistent with a role in cartilage maturation

mRNA in situ hybridization was used to detect hdac4 transcripts in larvae at 72 hpf. Previously, we described the expression of hdac4 in the ventral region of the developing pharyngeal skeleton at 72 hpf (DeLaurier et al., 2012), and here we show how specific regions of expression are associated with sites of ossification of cartilage. By 72 hpf, expression is localized to regions of sox9a-expressing cartilage as well as in tissue surrounding cartilage and dermal bone elements. Co-expression of hdac4 and sox9a was detected in the hyosymplectic, ceratohyal, and palatoquadrate cartilages (Figs. 3A–3H, indicated by arrows). In the case of the hyosymplectic and ceratohyal, co-expression of hdac4 and sox9a was in regions that undergo ossification at later stages. hdac4 was strongly expressed in the posterior pharyngeal arches, overlapping in the mid-region of each arch with a domain of sox9a expression in the ceratobranchial cartilage within each arch (Figs. 3I–3L).

Mutants have increased expression of runx2 factors

MEF2C is known to activate transcription of Runx2, a transcription factor that activates chondrocyte maturation (Arnold et al., 2007). Runx2 in turn activates Sp7, a transcription factor associated with osteoblast differentiation (Nishio et al., 2006). We examined how loss of hdac4, which would lead to over-activity of Mef2c, affects levels of mRNA expression of these factors. Among teleost fish, a whole genome duplication event generated two copies of vertebrate runx2, runx2a and runx2b (Van der Meulen et al., 2005). As in previous studies, we observed differential patterns of expression of both genes, indicating that these genes have distinct functions in skeletal development in zebrafish (Figs. 4A–4H). At 4 dpf, runx2a and runx2b were expressed in the mid-shaft region of the ceratohyal cartilage and branchiostegal ray (Figs. 4B–4F), and also in the opercle (not shown). In general, runx2a expression was broader than runx2b expression at 4 dpf, encompassing more of the anterior and posterior lengths of the ceratohyal (Figs. 4B–4F). runx2b was expressed in the hyosymplectic at 4 dpf, but runx2a was not (not shown). Among mutants, the average length of runx2a expression in the ceratohyal was significantly increased compared to wild-types (Figs. 4A–4D, indicated by arrows, Fig. 4O; F = 6.49; df = 1,19; p = 0.0196). Expression of runx2b was also significantly increased in mutants compared to wild-types (Figs. 4E–4H, indicated by arrows, Fig. 4P; F = 11.94; df = 1,15; p = 0.0035). At 4 dpf, in both wild-types and mutants, sp7 was expressed in a small patch of cells on the posterior margin, and less frequently also on the anterior margin, of the mid-shaft of the ceratohyal in both wild-type and mutant larvae (Figs. 4I–4L, indicated by arrows). Analysis of the average length of sp7 expression revealed no significant differences between mutants and wild-types (Fig. 4Q, F = 0.0001; df = 1,20; p = 0.99). Schematic diagrams showing the differences in patterns of expression of runx2a, runx2b, and sp7 in wild-type vs. zygotic mutant larvae at 4 dpf are shown in Figs. 4M–4N.

Maternal-zygotic mutants have increased ossification of the pharyngeal skeleton and defects in the anterior facial region

In order to examine the role of a maternal contribution of hdac4 to development, we performed analyses of skeletal phenotypes in mutants and heterozygotes generated from maternal mutants crossed with heterozygote males. At 7 dpf, a subset of maternal-zygotic mutants and heterozygotes showed a prominent increase in cartilage ossification compared to wild-types (non-sibling controls) (Figs. 5A, 5C–5E, 5G–5I). In total, 40.0% (16/40) of maternal-zygotic mutants and 40.0% (16/40) heterozygotes had evidence of premature or excessive ossification (Fig. 5F), and resembled ossification patterns present in 12 dpf wild-type fish (Fig. 5B). Increased ossification in mutants and heterozygotes included formation of an anguloarticular bone associated with the Meckel’s cartilage (Figs. 5B–5D), an enlarged quadrate (Figs. 5C and 5E), a ventral hypohyal element associated with the ceratohyal (Figs. 5C and 5E), and ossification of the symplectic cartilage (Fig. 5C, indicated by asterisk). Among these maternal-zygotic mutants and heterozygotes, ossification of the mid-shaft of the first or first and second ceratobranchial cartilages was detected, which is also normally observed in wild-type larvae at 12 dpf (Figs. 5D, 5E, 5G–5I).

Maternal-zygotic mutants have defects in the development of first pharyngeal arch skeletal elements and the anterior neurocranium

Among maternal-zygotic mutants, 25.9% (15/58) had defects of the first pharyngeal arch skeleton, including a shortened face (Figs. 6A, 6B and 6G) and loss of one or both Meckel’s cartilages and the anterior portion of the palatoquadrate cartilage, with retention of a small remnant of the entopterygoid and quadrate (Fig. 6C, see Figs. 2A and 2D for reference). Among maternal-zygotic heterozygotes, 10.2% (6/59, Fig. 6G) also had loss of first pharyngeal arch structures. In the case of larvae with a loss of first arch cartilages, posterior second arch and more posterior arch structures including the ceratohyal, branchiostegal ray, and opercle were present (Fig. 6C). Some maternal-zygotic mutants and heterozygotes had defects in the neurocranium cartilage (the primary palate in fish), including clefts, holes, and shortening of the anterior portion of the element (Fig. 6D representative heterozygote with normal neurocranium, Figs. 6E and 6F heterozygote and mutant with neurocranium defects). In total, 30.8% (16/52) of maternal-zygotic mutants, and 15.2% (7/46) of maternal-zygotic heterozygotes had neurocranium defects (Fig. 6G). Most neurocranium defects occurred in fish that also had loss of first arch structures (11/16, 68.8% of maternal-zygotic mutants, and 5/7, 71.4% of maternal-zygotic heterozygotes).

hdac4 transcripts are detected in embryos younger than 50% epiboly

Based on the RNA-seq expression atlas data for zebrafish, the highest level of hdac4 mRNA expression (relative to total transcripts) is between cleavage (2 cell) and blastula dome (over 1K cell stage), followed by relatively low expression of transcripts up to 3 dpf (larval protruding mouth), after which transcript levels increase (Fig. 7A). mRNA in situ hybridization experiments detected high levels of hdac4 transcripts in wild-type embryos 512 cells or younger (Figs. 7B–7D, stages in between those presented also showed comparable expression). By 75–90% epiboly, hdac4 expression was either not detectable, or expressed at very low levels compared to younger stages (Fig. 7E).