Kat2a and Kat2b Acetyltransferase Activity Regulates Craniofacial Cartilage and Bone Differentiation in Zebrafish and Mice

Cranial neural crest cells undergo cellular growth, patterning, and differentiation within the branchial arches to form cartilage and bone, resulting in a precise pattern of skeletal elements forming the craniofacial skeleton. However, it is unclear how cranial neural crest cells are regulated to give rise to the different shapes and sizes of the bone and cartilage. Epigenetic regulators are good candidates to be involved in this regulation, since they can exert both broad as well as precise control on pattern formation. Here, we investigated the role of the histone acetyltransferases Kat2a and Kat2b in craniofacial development using TALEN/CRISPR/Cas9 mutagenesis in zebrafish and the Kat2ahat/hat (also called Gcn5) allele in mice. kat2a and kat2b are broadly expressed during embryogenesis within the central nervous system and craniofacial region. Single and double kat2a and kat2b zebrafish mutants have an overall shortening and hypoplastic nature of the cartilage elements and disruption of the posterior ceratobranchial cartilages, likely due to smaller domains of expression of both cartilage- and bone-specific markers, including sox9a and col2a1, and runx2a and runx2b, respectively. Similarly, in mice we observe defects in the craniofacial skeleton, including hypoplastic bone and cartilage and altered expression of Runx2 and cartilage markers (Sox9, Col2a1). In addition, we determined that following the loss of Kat2a activity, overall histone 3 lysine 9 (H3K9) acetylation, the main epigenetic target of Kat2a/Kat2b, was decreased. These results suggest that Kat2a and Kat2b are required for growth and differentiation of craniofacial cartilage and bone in both zebrafish and mice by regulating H3K9 acetylation.


Introduction
Craniofacial bones and cartilage of the head and neck must undergo proper differentiation to form specific skeletal elements in adult vertebrates. These skeletal elements are derived from cranial neural crest cells (cNCC), a population of neural crest cells that originate at the rostral region of the embryo and emigrate from the dorsal part of the neural tube during vertebrate embryogenesis [1][2][3][4][5][6][7][8][9][10]. These migratory and multipotent cells migrate to designated locations in the branchial arches and An important epigenetic modification that promotes transcription is acetylation of lysine 9 residue of histone H3 (H3K9ac) [48,49]. H3K9ac is typically associated with transcriptional activation. A recent study reported that H3K9ac mediates a switch between transcriptional initiation to elongation by releasing the paused RNA polymerase II at the proximal promoter, activating gene expression [50]. In addition, H3K9ac also interacts via cross-talk with other histone modifiers to regulate developmental gene expression [51][52][53][54]. Two epigenetic modifying enzymes known to acetylate H3K9 are Kat2a and Kat2b (also known as GCN5 and PCAF, respectively), and they are both highly conserved from yeast to human. Kat2a and Kat2b share some degree of protein homology, with 73% amino acid sequence identity and conserved domains, including a bromodomain and acetyltransferase domain [55].
Kat2a and Kat2b regulate various physiological phenomenon, including embryonic development, brown adipogenesis, transcription, genome stability, cancer, immune response, and angiogenesis [56][57][58][59][60][61]. In this study, we specifically focused on their role in early developmental stages of the craniofacial skeleton, which is not well-understood. Both Kat2a and Kat2b are implicated in several signaling pathways during development, including retinoic acid (RA), Wnt, and Notch [62][63][64]. In mouse, Kat2a is expressed throughout the embryo except for the allantois and the heart and maintains high expression throughout development, at least until E16.5 [65]. Complete loss of Kat2a in mice results in embryonic lethality by embryonic day (E) 10.5 (E10.5), while Kat2b homozygous mutants are developmentally normal and survive into adulthood with no obvious defects [65]. Double homozygous null mutant mice for Kat2a/Kat2b die in utero by E7.5, earlier than the Kat2a single mutants, suggesting that there is some functional overlap between these two genes early in development [65]. Creation of a catalytically inactive form of Kat2a by creating two point mutations in the histone acetyl transferase domain (HAT) results in a cranial neural tube defect and lethality of Kat2a hat/hat embryos by E16.5, suggesting that the HAT activity of Kat2a is critical for embryonic development [62,66]. Previous work from our group shows that in the mouse forebrain, upon binding of RA to its receptor RARα/γ, Kat2a directly acetylates TACC1, which represses RA-mediated signaling in the absence of RA [62]. However, embryonic in vivo studies of Kat2a mutants are limited, and Kat2a has not been implicated in murine craniofacial development. A recent study in zebrafish implicates kat2a and kat2b in heart and fin development [67]. Tbx5, a transcription factor critical for heart and forelimb development, is directly acetylated by kat2a and kat2b [67]. Knockdown of kat2a and kat2b in zebrafish results in defects in heart looping and fin outgrowth, similar to that in tbx5 mutants, thus connecting the acetylation activity of Kat2a and Kat2b to heart and limb development [67]. However, the role of kat2a and kat2b has not been explored in zebrafish craniofacial development.
In this study, we genetically disrupted kat2a and kat2b using TALEN and CRISPR/Cas9 genome editing technologies, respectively, and analyzed the previously published Kat2a hat/hat mouse model. We observed defects in the neural crest cell-derived craniofacial cartilage and bone formed from both endochondral and intramembranous ossification, including hypoplastic cartilage and bone development in both zebrafish and mice. Our results further show a reduction in posterior ceratobranchial cartilages and a reduced expression of markers for dermal bones in zebrafish, and hypoplastic and ectopic bone growth as well as cartilage defects in mice. These results imply that mutations in kat2a and kat2b affect both pathways of ossification. We also observed differential expression of neural crest cell-specific cartilage and bone differentiation factors, including sox9a, col2a1, and runx2a/2b, as well as a reduction in H3K9ac levels in zebrafish and mouse mutant embryos. Collectively, our study reveals previously unidentified roles for the histone acetyltransferases kat2a and kat2b in craniofacial development.

Zebrafish
Zebrafish were maintained as previously described [68]. Wildtype (WT) strains used AB, and transgenic lines include Tg(sox10:EGFP) [69]. Transgenic fish were used alone or crossed with various mutant backgrounds. Developmental staging followed published standards [70]. All experiments utilizing zebrafish were approved by the University of Colorado Denver IACUC protocol #147 and conform to NIH regulatory standards of care.

Genotyping of Zebrafish
Fin clips or single embryos were placed in Lysis Buffer (10 mM TrisHCl (pH 8.0) 50 mM KCl, 0.3% Tween 20, 0.3% NP40, 1 mM EDTA) for 10 min at 95 • C, incubated with 50 µg of Proteinase K at 55 • C for 2-3 h, and incubated again at 95 • C for 10 min. For genotyping kat2a TALEN mutants, the following primers were used, (Fw) 5 -AACAGGATATTGTGAAGAAAG-3 and (Rv) 5 -ACTCCGAGTTTCTCCAGCTT-3 , to generate a 500 bp PCR product. To identify mutant embryos, PCR product was then incubated with SacI-HF (New England BioLabs, Catalog# R3156S), which digests the WT sequence but is unable to cut the mutant sequence. Heterozygotes were identified after SacI-HF digestion, as a 500 bp band for the mutant allele and two lower bands for the WT allele, similar to digestion of WT DNA. kat2b CRISPR mutants were genotyped using the following primers (Fw) 5 -TACTACACGTTTGTGGCCTTT-3 and (Rv) 5 -GCGTTTCTCAGGTGGTAGTT-3 , to generate a 200 bp PCR product. For PCR reactions, REDExtract-N-Amp™ PCR 2× ReadyMix™ (Sigma) was used and subjected to the same genotyping protocol. PCR product was then treated with enzyme T7 Endonuclease I (New England BioLabs, Catalog# M0302L, Ipswich, MA USA), which cleaves double-stranded DNA at the site of mismatch. Heterozygous and homozygous CRISPR mutants were distinguished using primers that were designed with dCAPS Finder 2.0 (helix.wustl.edu/dcaps/dcaps. html). PCR amplification on diluted PCR product from the first PCR with the primer pairs (Fw) 5 -GTTATTGTAACGTGCCCCAGTTCCG-3 and (Rv) 5 -GCGTTTCTCAGGTGGTAGTT-3 , followed by digestion with AciI (New England BioLabs, Catalog# R0551L), yields a digested lower band for mutant, an undigested upper band for WT, and a combination of both for heterozygotes.

Mice and Genotyping
Mice were bred and maintained according to the animal care and use protocols approved by the Institution of Animal Care and Use Committee at the University of Colorado Denver Anschutz Medical Campus protocol #528 and the University of Colorado Boulder protocol #2590. All animal dissections were performed in accordance with NIH guidelines and standards. The Kat2a hat mice (also called Gcn5 hat ) were maintained on a C57/Bl6J background and kept in standard light conditions. Heterozygous males and females were used for breeding and genotyped as previously described [66]. Timed matings were performed, and the morning of the vaginal plug was considered to be embryonic day 0.5. Embryos of matching somite numbers were used for experiments.

Zebrafish kat2a and kat2b Mutant Generation
TALEN pairs were found using (TALE-NT) 2.0 [71], which is a freely accessible web-based tool to find TALEN pairs with editable parameters and the identification of a unique restriction site in the spacer sequence. The TALEN pairs targeted exon 1 of kat2a to create a stop codon and new reading frame close to the start site. TALEN pairs were synthesized using the Golden Gate method [72] using the GoldyTALEN vector [73]. GoldyTALEN constructs were subsequently linearized with SacI, and the mRNA encoding each TALEN was transcribed using the mMESSAGE mMachine T3 kit (Life Technologies). mRNAs encoding TALEN pairs were injected into the cytoplasm of 1-cell stage WT AB zebrafish embryos. F0-injected embryos were outcrossed to Tg(sox10:EGFP) zebrafish to check for germline transmission. F1 fish that were found to carry a mutation were then grown to adulthood, fin-clipped, and sequenced. F1 mutant adults were again outcrossed with Tg(sox10:EGFP) zebrafish. Identified F2 mutant adults were used for crosses to generate the heterozygous and homozygous F3-F5 embryos used in experiments.
CRISPR [74] target sites to mutate kat2b were identified using the ZiFiT Targeter (http://zifit. partners.org/ZiFiT/); the selected target sites were then used to make gene-specific gRNAs. gRNAs were then prepared based on the protocols described in [75]. A Cas9/sgRNA complex was formed by incubating 600 ng/µL Cas9 protein (PNA Bio) with sgRNA at room temperature for 5 min before injection into the cytoplasm of 1-cell stage WT AB zebrafish embryos along with KCl at a final concentration of 200 mM.

In Situ Hybridization, Alcian Blue Staining, and Alizarin Red Staining
Whole-mount RNA in situ hybridization (ISH) on zebrafish embryos was performed as previously described [76,77]. DIG-conjugated antisense probes were synthesized as previously published for the following genes: dlx2a [78] and crestin [79]. The following antisense probes were used: col2a1 (obtained from Tatjana Piotrowski, Kansas City, MO USA) and sox9, runx2a, and runx2b (obtained from Jamie Nichols, Aurora, CO USA). DIG-conjugated antisense probes were synthesized from full-length transcript sequences in the pCS2+ plasmid using primers for the following genes: Following ISH, single-embryo genotyping for kat2a and kat2b mutant alleles was performed. Zebrafish embryos were incubated at 65 • C for 8 h in 300 mM NaCl to remove crosslinking events. Embryos were then processed for single embryo genotyping as above. Alcian blue and Alizarin red staining was performed as previously described [80].
Mouse embryos for ISH were fixed overnight in 4% paraformaldehyde (PFA) at 4 • C. Embryos for whole-mount ISH were then washed in PBS, dehydrated in a series of Methanol:PBS, and then stored in 100% Methanol. Whole-mount ISH on mouse embryos was performed as previously described [81]. Embryos for section in situ were cryopreserved overnight at 4 • C in 30% sucrose and then embedded by flash freezing in Tissue Tek O.C.T., and 10 µm sections were cut on a Leica CM3050S cryostat. Section in situ hybridization was performed as described [82]. In all cases, DIG-labeled RNA probes were used. The following RNA probes were used: mRunx2 (a kind gift from Dr. David Clouthier), mBmp4 (a kind gift from Dr. Liz Robertson), and mSox9 (Fw: 5 -TGAACGAGAGCGAGAAGAA-3 ; Rv: 5 -GATGGTCAGCGTAGTCGTATT-3 ). Mouse skeletons were stained with alcian blue and alizarin red as previously described [83].

Immunohistochemistry and Western Blotting
Zebrafish embryos and larvae were fixed in 4% PFA overnight at 4 • C. Fixed embryos were embedded in 1.5% agar with 5% sucrose and transferred to a 30% sucrose solution in scintillation vials and incubated at 4 • C overnight. The blocks were then frozen, and a cryostat microtome was used to cut the blocks into 10-15 µm sections received on Superfrost Plus slides (12-550-15, Fisher Scientific, USA). The sections were incubated with the following primary antibodies: Anti-Histone H3K9 acetylation (C5B11, Cell Signaling Technology, Danvers, MA USA) and Anti-Histone H3K9me3 (ab8898, Abcam). Slides were processed using the normal Immunofluorescence staining protocol for sections. Slides were washed in PBS several times to remove buffer, then blocked for 30 min at room temperature in 10% serum-blocking buffer and incubated overnight at 4 • C with primary antibody in blocking buffer. The next day, slides were washed several times in PBS then incubated with secondary antibodies (Alexa Fluor 568 goat anti-rabbit, Life Technologies, USA) for 1 h at room temperature. Slides were then washed several times in PBS and cover-glass mounted with Permafluor (Thermo Scientific, USA). Fluorescence was imaged with a Zeiss Axiovert 200 microscope equipped with a PerkinElmer spinning disk confocal system and Volocity software (version 6.0, PerkinElmer, USA), which were imported into Adobe Photoshop. All images for a given figure were processed in the same way.
For Western blotting, 50 embryos of each genotype were collected and deyolked at 48 h post fertilization (hpf). Embryos were then lysed in 40 µL of RIPA buffer plus a protease inhibitor cocktail for 1 h. A Bradford protein assay was performed for protein concentration. Lysates were boiled in SDS sample buffer for 5 min at 95 • C. Samples were loaded into a 10% PAGE gel (BioRad, USA). Western blots were probed with primary and secondary antibodies against H3K9 acetylation (C5B11 and 7074S; Cell Signaling Technology) and histone H3 (9715S, Cell Signaling Technology and A0545, Sigma Aldrich, USA). Imaging was done with a BioRad Chemidoc multiplex imager. Data plotting and statistical analyses were performed in GraphPad Prism software (version 6) (La Jolla, CA USA).

Quantitative PCR
All RT-qPCR experiments on mouse embryos were carried out as two steps with reverse transcription of RNA preceding qPCR. Tissue dissected from embryos was flash frozen in liquid nitrogen, followed by RNA extraction with the Zymo Quick RNA kit (Zymo, Cat #R1050, Irvine, CA USA). cDNA was prepared from 1 µg of RNA with the Superscript III Reverse Transcriptase kit (Thermo Fisher, Cat #18080093). qPCR was carried out using Roche Universal Probe library Master Mix and Roche Light Cycler 480. A combination of GAPDH and GusB were used as a housekeeping control. The following primers and probes were used:

Statistical Analysis
Measurements of zebrafish viscerocranium were completed from the anterior most aspect of Meckel's cartilage at the midline to the base of the hyosymplectic, as previously described [78]. Similar measurements of the neurocranium were made from the ethmoid plate to the base of the trabeculae. All statistical analyses and graphing for experiments were done using either R Studio or PRISM. Statistical significance was determined using Welch's T-Test with p < 0.05 considered significant.

kat2a and kat2b Are Expressed Broadly during Zebrafish Embryogenesis
To determine the expression of kat2a and kat2b within the cranial neural crest region during zebrafish development, we performed in situ hybridization between 24 and 48 h post fertilization (hpf; Figure 1). Both kat2a and kat2b are broadly expressed throughout the anterior head region, including the central nervous system (CNS), the eye, and branchial arches at 24 hpf ( Figure 1A,D). kat2a and kat2b are expressed strongly in the brain region, with kat2b expression extending posteriorly in the spinal cord ( Figure 1A,D). By 40-48 hpf, expression of both genes remains strongly expressed in the head region but is reduced throughout the rest of the embryo ( Figure 1B,C,E,F). A previous analysis in mouse embryos has shown that both Kat2a and Kat2b are similarly ubiquitously expressed [65], and importantly they are expressed in the branchial arch mesenchyme. A cross-sectional analysis confirms that zebrafish kat2a is broadly expressed in the brain, the lens of the eye, and the cranial mesenchyme where neural crest cells are condensing to form cartilage and bone (not shown). Thus, kat2a and kat2b are expressed in the right place and time in both zebrafish and mouse embryos to be involved in craniofacial development.
J. Dev. Biol. 2018, 6, x FOR PEER REVIEW 7 of 20 in the head region but is reduced throughout the rest of the embryo ( Figure 1B,C,E,F). A previous analysis in mouse embryos has shown that both Kat2a and Kat2b are similarly ubiquitously expressed [65], and importantly they are expressed in the branchial arch mesenchyme. A cross-sectional analysis confirms that zebrafish kat2a is broadly expressed in the brain, the lens of the eye, and the cranial mesenchyme where neural crest cells are condensing to form cartilage and bone (not shown). Thus, kat2a and kat2b are expressed in the right place and time in both zebrafish and mouse embryos to be involved in craniofacial development.

Loss of kat2a and kat2b in Zebrafish, and Kat2a hat/hat in Mice, Results in Craniofacial Cartilage and Bone Defects
To determine the function of kat2a and kat2b in cNCC development, we created mutants in zebrafish and analyzed the existing Kat2a hat/hat mouse mutant. For the zebrafish mutations, we used both TALEN-and CRISPR/Cas9-based mutagenesis. kat2a zebrafish mutants were created with TALEN gene-targeted mutagenesis to generate two mutant kat2a alleles (Supplemental Figure S1A-C). The TALEN targeted exon 1 to create a stop codon and a new reading frame close to the start site (Schematic in Supplemental Figure S1A). We then outcrossed the injected F0 adult zebrafish, sequenced the locus in the resulting embryos, and identified two independent kat2a alleles that are predicted to create severe loss-of-function mutations with short open reading frames (ORFs; 303-353 base pairs). These lines were maintained for further analysis and designated kat2a CO1007 and kat2a CO1010 , with kat2a CO1007 used for the remainder of this study (referred to as kat2a −/− mutants from this point forward). kat2b zebrafish mutants were generated using CRISPR/Cas9 (Supplemental Figure S1D-F). Guide RNAs were designed to target a Protospacer adjacent motif (PAM) site in exon 6 and we identified three alleles by sequencing: kat2b CO1008 , kat2b CO1011 , and kat2b CO1012 with an ORF for the mutations between 858 and 990 base pairs. kat2b CO1008 was used for the remainder of this study (referred to as kat2b −/− mutants from this point forward). To determine whether the expression of kat2a and kat2b was reduced in the respective zebrafish mutants, in situ hybridization was performed. Comparing wildtype kat2a and kat2b expression in kat2a −/− and kat2b −/− embryos, respectively, there is a clear reduction in expression in the mutant embryos (Supplemental Figure S2). Hence, although the

Loss of kat2a and kat2b in Zebrafish, and Kat2a hat/hat in Mice, Results in Craniofacial Cartilage and Bone Defects
To determine the function of kat2a and kat2b in cNCC development, we created mutants in zebrafish and analyzed the existing Kat2a hat/hat mouse mutant. For the zebrafish mutations, we used both TALEN-and CRISPR/Cas9-based mutagenesis. kat2a zebrafish mutants were created with TALEN gene-targeted mutagenesis to generate two mutant kat2a alleles (Supplemental Figure S1A-C). The TALEN targeted exon 1 to create a stop codon and a new reading frame close to the start site (Schematic in Supplemental Figure S1A). We then outcrossed the injected F0 adult zebrafish, sequenced the locus in the resulting embryos, and identified two independent kat2a alleles that are predicted to create severe loss-of-function mutations with short open reading frames (ORFs; 303-353 base pairs). These lines were maintained for further analysis and designated kat2a CO1007 and kat2a CO1010 , with kat2a CO1007 used for the remainder of this study (referred to as kat2a −/− mutants from this point forward). kat2b zebrafish mutants were generated using CRISPR/Cas9 (Supplemental Figure  S1D-F). Guide RNAs were designed to target a Protospacer adjacent motif (PAM) site in exon 6 and we identified three alleles by sequencing: kat2b CO1008 , kat2b CO1011 , and kat2b CO1012 with an ORF for the mutations between 858 and 990 base pairs. kat2b CO1008 was used for the remainder of this study (referred to as kat2b −/− mutants from this point forward). To determine whether the expression of kat2a and kat2b was reduced in the respective zebrafish mutants, in situ hybridization was performed. Comparing wildtype kat2a and kat2b expression in kat2a −/− and kat2b −/− embryos, respectively, there is a clear reduction in expression in the mutant embryos (Supplemental Figure S2). Hence, although the kat2b alleles have mutations in a later exon, the resulting mutations are predicted to be loss-of-function alleles.
To test the hypothesis that kat2a and kat2b play a role in cartilage development, we assessed cartilage formation in the kat2a −/− and kat2b −/− zebrafish embryos by performing Alcian blue staining at 4 days post fertilization (dpf). Both zebrafish mutants display severe craniofacial defects (Figure 2A-F,K-N). kat2a −/− embryos display overall hypoplastic or loss of cartilage elements, including a reduced palatoquadrate, a reduced pterygoid process of the palatoquadrate and hyosymplectic of the viscerocranium, shortening of the Meckel's cartilage, and widening of the angle of the ceratohyal of the viscerocranium. kat2a −/− embryos also have a smaller hypoplastic ethmoid plate, a deformed trabeculae of the neurocranium, and ceratobranchials that are reduced in number with the anterior elements hypoplastic if present. kat2b −/− embryos have a similar but more severe hypoplasia, including a complete lack of staining of the posterior ceratobranchial cartilages, and a more significant gap in the anterior edge of the ethmoid plate that forms a "cleft" (Figure 2E,F). Hence, several cartilage elements of the craniofacial skeleton, including the neurocranium, ceratohyal, and ceratobranchials, which are formed via endochondral ossification, are affected in kat2a −/− and kat2b −/− embryos. Quantification of the length from Meckel's cartilage to the palatoquadrate and from the ethmoid plate to the tribecule indicates a significant reduction by more than 50% in both mutants when compared to the wildtype ( Figure 2K-N). Interestingly, kat2a heterozygous embryos also display a mild phenotype, which suggests that the created mutation is a dominant mutation (~19% of heterozygous embryos). Both kat2a −/− and kat2b −/− were viable at 4 dpf, presenting at~25% of the total number of embryos in each clutch. However, for both single zebrafish mutants, the embryos did not survive to 6-7 dpf, and, therefore, we were unable to assay for bone formation.
To determine if kat2a and kat2b function together or redundantly in craniofacial cartilage development, we created double kat2a −/− ;kat2b −/− mutant zebrafish embryos and examined the craniofacial phenotype at 4 dpf ( Figure 2G-J). Alcian staining of double kat2a −/− ;kat2b −/− mutant embryos showed a more severe craniofacial phenotype compared to the individual kat2a −/− and kat2b −/− mutants. In fact, because of the reduction in cartilage formation, it was not possible to dissect the viscerocranium from the neurocranium. The whole-mount stains of double kat2a −/− ;kat2b −/− mutants suggest that the cartilage elements start to differentiate but then differentiation stalls, since only a small portion of the neurocranium and the early developing viscerocranium has begun to form ( Figure 2G-J). It has been described previously that zebrafish cartilage develops in a specific order, and the pattern we observe is consistent with that analysis, as shown by morphology [28] as well as gene expression [84]. These data indicate that, in addition to sharing redundant roles, kat2a and kat2b also contribute individually to craniofacial development.
To determine whether mouse Kat2a also functions in craniofacial development, we analyzed skeletal development in the Kat2a hat/hat mouse homozygous mutants. This also allowed us to extend our study to craniofacial bone formation. Previous work from our lab and others has identified that the HAT activity of Kat2a is critical for development; however, a role for Kat2a in craniofacial development has not been previously identified. Interestingly, previous in vitro studies using mesenchymal stem cells that were differentiated into osteoblasts and treated with shRNA to knock down Kat2a showed that disruption of Kat2a resulted in a loss of Runx2 expression [85], raising the question of whether Kat2a regulates bone growth in vivo. Therefore, to test the hypothesis that Kat2a is required for skeletal development, we evaluated bone and cartilage formation in E16.5 wildtype and Kat2a hat/hat mouse mutants using Alizarin Red and Alcian blue staining. This revealed several craniofacial phenotypes in the Kat2a hat/hat mouse mutants. Of these, the most severe defects were observed in derivatives of the mandibular prominence of the first branchial arch, affecting both endochondral and intramembranous ossification of the craniofacial skeleton. Intramembranous bone defects included ectopic bone growth on the mandible or dentary bone ( Figure 2T, black arrows), as well as a hypoplastic squamosal ( Figure 2R). Endochondral ossification defects included loss of mandibular condyles, hypoplastic malleus and incus, as well as missing tympanic rings of the ear ( Figure 2R, arrow heads). In the Kat2a hat/hat mutants, we did not observe a cleft lip nor a palate. We also did not observe any skeletal defects in the axial skeleton (data not shown). Thus, the kat2a/kat2b acetyltransferase genes are required in both zebrafish and mouse for proper craniofacial development.
the anterior edge of the ethmoid plate that forms a "cleft" (Figure 2E,F). Hence, several cartilage elements of the craniofacial skeleton, including the neurocranium, ceratohyal, and ceratobranchials, which are formed via endochondral ossification, are affected in kat2a −/− and kat2b −/− embryos. Quantification of the length from Meckel's cartilage to the palatoquadrate and from the ethmoid plate to the tribecule indicates a significant reduction by more than 50% in both mutants when compared to the wildtype ( Figure 2K-N). Interestingly, kat2a heterozygous embryos also display a mild phenotype, which suggests that the created mutation is a dominant mutation (~19% of heterozygous embryos). Both kat2a −/− and kat2b −/− were viable at 4 dpf, presenting at ~25% of the total number of embryos in each clutch. However, for both single zebrafish mutants, the embryos did not survive to 6-7 dpf, and, therefore, we were unable to assay for bone formation.

Cartilage and Bone Markers are Differentially Expressed in kat2a and kat2b Zebrafish and Kat2a Hat/Hat Mouse Embryos
The cartilage and bone defects in Kat2a hat/hat mutant mouse embryos led us to first examine whether these phenotypes were due to a defect in cNCC specification or branchial arch patterning. To do so, we examined the expression of AP2, Twist1, and Dlx1 in E9.5 and E10.5 mouse embryos. Ap2 is expressed in the ectoderm and migratory cNCCs and acts to specify cNCC identity, while Twist1 and Dlx1 are expressed in the branchial endoderm and regulate branchial arch patterning [86]. At both E9.5 and E10.5, AP2 positive migratory cNCCs can be seen in the branchial arches of Kat2a hat/hat and control mouse embryos, suggesting that cNCC specification is not affected by loss of Kat2a HAT activity in mice (Supplemental Figure S3A-H). Similarly, the expression of Dlx1 was not significantly changed in E9.5 and E10.5 Kat2 hat/hat mutants compared to control mouse embryos (Supplemental Figure S3J,K). We did observe a significant change in the expression of Twist1 in E9.5 Kat2a hat/hat mutant mouse embryos; however, this change was not observed in E10.5 mutants compared to the control (Supplemental Figure S3J,K). Likewise, in kat2a −/− and kat2b −/− zebrafish embryos, expression of dlx2 and crestin was not significantly reduced at early stages of neural crest cells development (Supplemental Figure S4). Together, these results suggest that Kat2a/2b is required after cNCC specification and migration.
To determine whether Kat2a/Kat2b are involved more directly in regulating craniofacial bone and cartilage development, we investigated the temporal pattern of gene expression for cartilage and bone in the zebrafish and mouse mutants. To assess pre-chondrocyte differentiation, we used in situ hybridization to examine the expression of sox9 and col2a1 between 48 and 72 hpf in zebrafish embryos. In zebrafish, the overall level of sox9a was reduced in both the kat2a −/− and kat2b −/− mutants ( Figure 3A-L). Moreover, the domains of sox9a expression were also smaller in accordance with the overall smaller head size as compared to wildtype or heterozygous siblings at the same time point (Figure 3A-L). sox9a expression at 48 hpf marks the early forming cartilage of the Meckel's cartilage and the palatoquadrate, whereas in the kat2a and kat2b mutant embryos, the anterior Meckel's cartilage expression appears to be diffused, and the expression in the posterior elements is significantly reduced compared to the corresponding wildtypes ( Figure 3E,F,I,J compared to WT in Figure 3A,B). In 72 hpf wildtype zebrafish, when the cartilage elements are beginning to form, sox9a and col2a1 are expressed within the forming Meckel's cartilage, the ceratohyal, and the ceratobranchial cartilages 1-5 ( Figure 3C,D,M). However, in the kat2a and kat2b mutant zebrafish embryos, we observed a delay in formation of the cartilages and sox9a and col2a1 expression, with sox9a ( Figure 3G,H,K,L) and col2a1 ( Figure 3N,O) expression present in the ceratohyal as well as the first ceratobranchial. We were unable to assay for hypertrophic chondrocyte markers that are expressed at later developmental stages because the zebrafish mutants do not survive up to those stages. To determine the effects on bone formation in the absence of kat2a and kat2b, we performed in situ hybridization with the markers of bone precursors runx2a and runx2b at 72 hpf. In kat2a and kat2b mutant embryos, runx2a and runx2b expression is reduced (Figure 4). While each element is present in the kat2a and kat2b mutants, the expression is significantly decreased in the maxilla, ceratohyal, ceratobranchials, and cleithrum ( Figure 4B,C,F,G) as compared to wildtype embryos at the same stage ( Figure 4A,D,E). There is one exception, where we observe an increase in runx2a expression in the forming dentary bone ( Figure 4B,C as compared to Figure 4A). The decrease in the expression of bone precursors in the mutants in the maxilla, branchiostegal ray, and opercle, which are formed via intramembranous ossification, indicate that kat2a and kat2b mutations affect this pathway, in addition to the endochondral ossification described previously (Figure 2A-F). Combined, these results suggest that kat2a and kat2b is required for cartilage and bone differentiation in zebrafish.  In E12.5 mouse embryos following condensation of the cNCC-derived mesenchyme, Sox9 RNA expression within the facial prominences was visibly decreased in the frontonasal prominence of Kat2a hat/hat embryos compared to controls ( Figure 3P,Q), although Sox9 expression was detected in Meckel's cartilage ( Figure 3R,S). Quantification of the expression of Sox9 and Col2a1 in Kat2a hat/hat and control mouse embryos by RT-qPCR using RNA extracted from E12.5 facial prominences showed that both Sox9 and Col2a1 were significantly decreased in expression in mutant mouse embryos ( Figure 3U).
We also assayed the expression of the osteoblast progenitors Bmp4 and Runx2 by whole-mount in situ hybridization of E12.5 mouse embryos. Kat2a hat/hat mutant mouse embryos showed no obvious loss of expression of Bmp4 or Runx2 compared to controls ( Figure 4H-K). We quantified the expression of these markers using RNA extracted from the facial prominences of E12.5 Kat2a hat/hat and control mouse embryos followed with RT-qPCR quantification. This showed a significant increase in the expression of Runx2, but also a significant reduction in Bmp4 in Kat2a hat/hat mouse mutants ( Figure 4L). This may suggest that a subpopulation of osteoblasts is increased in Kat2 hat/hat mice, perhaps related to ectopic bone formed through intramembranous ossification ( Figure 2T) and similar to the increased dentary bone expression of runx2a/2b in zebrafish. Taken together, our data support a role for Kat2a in both mouse and zebrafish chondrocyte and osteoblast development. embryos. The yellow outline in (H) marks the area dissected for RNA extraction. Gene expression was normalized to GusB, n = 4. The Y-axis shows the Log2 fold change of the relative gene expression, following the ddCT method; * denotes statistical significance (p < 0.05) using Welch's t-Test. The error bars show the standard error of the mean. Neural tube closure defects are evident in Kat2a hat/hat mutants. bs, branchiostegal ray; cb-5, ceratobranchial-5; ch, ceratohyal; cl, cleithrum; de, dentary; ep, ethmoid plate; hs, hyosymplectic; md, mandible; mx, maxilla; op, opercle. The scale bars represent 100 μm.
In E12.5 mouse embryos following condensation of the cNCC-derived mesenchyme, Sox9 RNA expression within the facial prominences was visibly decreased in the frontonasal prominence of Kat2a hat/hat embryos compared to controls ( Figure 3P,Q), although Sox9 expression was detected in Meckel's cartilage ( Figure 3R,S). Quantification of the expression of Sox9 and Col2a1 in Kat2a hat/hat and control mouse embryos by RT-qPCR using RNA extracted from E12.5 facial prominences showed that both Sox9 and Col2a1 were significantly decreased in expression in mutant mouse embryos ( Figure 3U). We also assayed the expression of the osteoblast progenitors Bmp4 and Runx2 by wholemount in situ hybridization of E12.5 mouse embryos. Kat2a hat/hat mutant mouse embryos showed no obvious loss of expression of Bmp4 or Runx2 compared to controls ( Figure 4H-K). We quantified the expression of these markers using RNA extracted from the facial prominences of E12.5 Kat2a hat/hat and control mouse embryos followed with RT-qPCR quantification. This showed a significant increase in the expression of Runx2, but also a significant reduction in Bmp4 in Kat2a hat/hat mouse mutants ( Figure  4L). This may suggest that a subpopulation of osteoblasts is increased in Kat2 hat/hat mice, perhaps and mutant (Kat2a hat/hat ; Mut on graph) embryos. The yellow outline in (H) marks the area dissected for RNA extraction. Gene expression was normalized to GusB, n = 4. The Y-axis shows the Log2 fold change of the relative gene expression, following the ddCT method; * denotes statistical significance (p < 0.05) using Welch's t-Test. The error bars show the standard error of the mean. Neural tube closure defects are evident in Kat2a hat/hat mutants. bs, branchiostegal ray; cb-5, ceratobranchial-5; ch, ceratohyal; cl, cleithrum; de, dentary; ep, ethmoid plate; hs, hyosymplectic; md, mandible; mx, maxilla; op, opercle. The scale bars represent 100 µm.

H3K9ac Is Reduced in Kat2a and Kat2b Mutants
We next tested the hypothesis that Kat2a/Kat2b regulates the expression of bone and cartilage specifiers through H3K9 acetylation, which is associated with the activation of transcription. To begin to address this, we analyzed the H3K9ac expression in both zebrafish and mouse during craniofacial development. In zebrafish embryos at 48 hpf, when neural crest cells have populated the branchial arches and begun to differentiate, we analyzed H3K9ac both by a Western blot analysis and immunofluorescence staining ( Figure 5). In kat2a −/− and kat2b −/− zebrafish at 48 hpf, neural crest cells expressing sox10:GFP, while present in the branchial arches, are reduced in number and highly disorganized such that each arch is not defined (Figure 5A -C ). To assay for H3K9 acetylation (AC), we stained these embryos with an antibody that recognizes this specific H3K9 acetyl mark in vivo. H3K9Ac levels were reduced in branchial arch regions where migrating cranial neural crest cells populate and differentiate into craniofacial derivatives ( Figure 5A-C"). This is in particular observed in kat2a mutant zebrafish embryos ( Figure 5B). In addition, this reduction is observed in the more anterior migrating cells that migrate to a dorsal position of the yolk ball, likely forming the cardiac neural crest populations (Supplemental Figure S5). To confirm and quantify the levels of H3K9Ac in kat2a −/− and kat2b −/− whole zebrafish embryos, we used a Western blot analysis. Similarly to what we observed in whole-mount immunofluorescence staining, H3K9Ac was significantly reduced at 48 hpf when compared to the wildtype ( Figure 5D-G; quantified as relative intensity compared to total H3 levels). These data suggest that kat2a and kat2b function as histone acetyltransferases during craniofacial development in zebrafish. to address this, we analyzed the H3K9ac expression in both zebrafish and mouse during craniofacial development. In zebrafish embryos at 48 hpf, when neural crest cells have populated the branchial arches and begun to differentiate, we analyzed H3K9ac both by a Western blot analysis and immunofluorescence staining ( Figure 5). In kat2a −/− and kat2b −/− zebrafish at 48 hpf, neural crest cells expressing sox10:GFP, while present in the branchial arches, are reduced in number and highly disorganized such that each arch is not defined ( Figure 5A'-C'). To assay for H3K9 acetylation (AC), we stained these embryos with an antibody that recognizes this specific H3K9 acetyl mark in vivo. H3K9Ac levels were reduced in branchial arch regions where migrating cranial neural crest cells populate and differentiate into craniofacial derivatives ( Figure 5A-C"). This is in particular observed in kat2a mutant zebrafish embryos ( Figure 5B). In addition, this reduction is observed in the more anterior migrating cells that migrate to a dorsal position of the yolk ball, likely forming the cardiac neural crest populations (Supplemental Figure S5). To confirm and quantify the levels of H3K9Ac in kat2a −/− and kat2b −/− whole zebrafish embryos, we used a Western blot analysis. Similarly to what we observed in whole-mount immunofluorescence staining, H3K9Ac was significantly reduced at 48 hpf when compared to the wildtype ( Figure 5D-G; quantified as relative intensity compared to total H3 levels). These data suggest that kat2a and kat2b function as histone acetyltransferases during craniofacial development in zebrafish.  We also tested H3K9ac expression in mouse using immunofluorescence on cryosections through the craniofacial region in E12.5 mouse embryos, when bone and cartilage specification is taking place. At E12.5, we did not observe a significant change in the overall staining in Kat2a hat/hat mutant mouse embryos as compared to Kat2a Wt/hat control embryos in the frontonasal prominence ( Figure 5H-K). Quantification of total cell fluorescence, corrected for background and area (Corrected Total Cell Fluorescence, CTCF) also did not show a significant difference in the CTCF means of E12.5 Kat2a hat/hat mouse mutants compared to controls ( Figure 5L). Our data in zebrafish suggest that the craniofacial defects could be due to broad epigenetic regulation; however, in mouse global H3K9ac, levels were not significantly changed, suggesting that Kat2a may act on more specific epigenetic targets that are not detectable when viewed in whole cells and tissues.

Discussion
The epigenetic modifiers Kat2a and Kat2b are highly conserved and participate in a variety of developmental pathways. Here, we describe a previously unknown role for Kat2a and Kat2b in craniofacial development. These two acetyltransferase enzymes share a high degree of protein homology, both at the level of amino acid sequence and protein structure, and they are both broadly expressed during development. Despite this, their functions in vivo are not necessarily redundant. Murine embryos that are homozygous null for Kat2a die early in embryonic development; however, Kat2b null mice survive into adulthood. Although Kat2b mice have not been reported to have a craniofacial defect, here we show that both Kat2a and Kat2b are required for zebrafish craniofacial development, and the phenotype increases in severity in double Kat2a/2b mutants. Craniofacial skeletal elements belonging to both endochondral and intramembranous ossification pathways are affected in kat2a and kat2b mutants. The differences in Kat2b loss-of-function phenotypes between zebrafish and mouse could be a result of divergent evolution. Here, we also show that the HAT activity of at least Kat2a is required for murine craniofacial development, using an allele in which Kat2a protein is still made and can be incorporated into multi-protein complexes but lacks enzymatic activity. Finally, our results suggest that Kat2a acetyltransferase activity is evolutionarily conserved in craniofacial development. Simultaneously, we acknowledge that owing to their high degree of conservation across species, Kat2a and Kat2b are likely to regulate other skeletal structures, in addition to the craniofacial development reported here. Their epigenetic acetyltransferase roles may also be exerted again later in life with the onset of age-related skeletal diseases, which is yet to be determined.
The craniofacial defects that result from loss of function of Kat2a/b acetyltransferases affect the development of craniofacial elements that are derived from the first and second branchial arches in both mouse and zebrafish. These include hypoplastic Meckel's cartilage and loss of cartilaginous elements. Our analysis of Kat2a hat/hat mutant mouse embryos at early stages of cNCC specification, migration, and arch patterning did not detect an obvious change, whereas we did observe a reduction in the expression of the chondrocyte markers Sox9 and Col2a1 in both zebrafish and mouse mutants. Thus, our data point to a requirement for Kat2a/Kat2b during craniofacial chondrocyte development subsequent to NCC migration and arrival in the branchial arches. Our analysis of the bone markers Runx2 and Bmp4 also showed a significant expression change in mouse E12.5 Kat2a hat/hat mutant embryos, suggesting that Kat2a could also have a role in both mechanisms of bone formation, requiring further studies. Previous in vitro studies using mesenchymal stem cells that were induced to differentiate into osteoblasts that were treated with shRNA to knockdown Kat2a function suggested that Kat2a is required for pre-osteoblast differentiation, supported by a reduction in Runx2 expression [85]. In this study, we show that Runx2 expression was not decreased in mouse mutants but instead increased in the craniofacial region, suggesting that pre-osteoblast differentiation is not lost. Furthermore, we did not observe abnormalities in axial skeleton development in the mouse Kat2a hat/hat mutants. One possible explanation is that Kat2b may be able to compensate for a loss of Kat2a in the axial skeleton. It is also possible that Kat2a has different targets in the craniofacial region compared to long bone formation in the limbs and that these targets in the craniofacial region are required for proper skeletal patterning. Nevertheless, future investigations should identify the direct targets of Kat2a during intramembranous and endochondral ossification.
Our findings in zebrafish and mouse are largely similar, but the differential requirement for Kat2b in zebrafish craniofacial development may be explained by a loss of conservation in the Kat2b function. As the genome of zebrafish has undergone a genome duplication event, the redundant functions of Kat2a and Kat2b could be more critical for zebrafish embryonic development. Interestingly, in zebrafish, kat2b has a more severe cartilage phenotype as compared to kat2a mutants, even though kat2a mutants have a more severe reduction in H3K9 acetylation. This suggests that, although kat2a has more enzymatic functional activity, kat2b may function independently of this activity for proper craniofacial development. In mouse, Kat2b expression is low throughout development while Kat2a expression is high through most of the embryonic development and then significantly drops at E16.5, with low levels detected in the adult. It has also been suggested that Kat2b expression is higher in the adult than in embryos, supporting the hypothesis that Kat2b may have critical functions in post-natal mammals [65]. Taken together, our results suggest that Kat2b is more critical for zebrafish development than for mouse development.
In addition to H3K9 acetylation, both Kat2a and Kat2b acetylate other non-histone substrates in various aspects of development and disease, including brain development and cancer [62,87]. However, Kat2b targets have been linked to cell differentiation, cell proliferation, and apoptosis, which may explain the severity in the cartilage phenotype in kat2b −/− versus kat2a −/− (Figure 2A-F), even though the H3K9 acetylation levels of kat2b −/− were higher (53.6%) than those of kat2a −/− embryos (50.3%) ( Figure 5D,F). One promising substrate of Kat2b in this process is Gli1, which has been identified as a Kat2b binding and acetylation target in hepatocellular carcinoma cells [88]. Gli1 is a transcription factor that is required for differentiation of neural crest cells [89,90], and the acetylation of Gli1 by Kat2b may be one of the potential contributors to the kat2b −/− cartilage phenotype. Another substrate of Kat2b, Akt1, has been found to be correlated to the regulation of neural crest cells specification and migration [91] and embryonic skeletal development [92], and mutated in patients with craniofacial defects, including mandible and frontal bone deformities [93]. In addition, acetylation of Akt1 by Kat2b has been shown to regulate cell proliferation and survival in vitro [61,94,95]. Thus, it is clear that Kat2b regulates other factors directly, and not via the modification of H3K9. It remains to be tested whether the acetylation by Kat2b regulates the survival of chondrocytes and osteoblasts in a similar way to regulate craniofacial development.
Nonetheless, Kat2a and Kat2b are highly conserved epigenetic modifiers that preferentially acetylate H3K9. This epigenetic mark is correlated with active transcription, a critical step during cell differentiation. In the kat2a/b zebrafish mutants, we observed a significant reduction in the expression of H3K9ac. Although we did not observe a global change in the levels of H3K9ac in E12.5 Kat2a hat/hat mutant mouse embryos, we cannot rule out the possibility that there are changes at a single cell resolution or at specific target loci. While it should be noted that both Kat2a and Kat2b can acetylate non-histone proteins, the loss of H3K9ac, as well as the loss of expression of bone and cartilage genes, suggests that Kat2a/Kat2b are required for active transcription of key craniofacial skeletal genes. Our results support the hypothesis that Kat2a/Kat2b directly target the loci of bone-and cartilage-specific genes to activate their transcription in the craniofacial-derived mesenchyme.