TIP55, a splice isoform of the KAT5 acetyltransferase, is essential for developmental gene regulation and organogenesis

Regulation of chromatin structure is critical for cell type-specific gene expression. Many chromatin regulatory complexes exist in several different forms, due to alternative splicing and differential incorporation of accessory subunits. However, in vivo studies often utilize mutations that eliminate multiple forms of complexes, preventing assessment of the specific roles of each. Here we examined the developmental roles of the TIP55 isoform of the KAT5 histone acetyltransferase. In contrast to the pre-implantation lethal phenotype of mice lacking all four Kat5 transcripts, mice specifically deficient for Tip55 die around embryonic day 11.5 (E11.5). Prior to developmental arrest, defects in heart and neural tube were evident in Tip55 mutant embryos. Specification of cardiac and neural cell fates appeared normal in Tip55 mutants. However, cell division and survival were impaired in heart and neural tube, respectively, revealing a role for TIP55 in cellular proliferation. Consistent with these findings, transcriptome profiling revealed perturbations in genes that function in multiple cell types and developmental pathways. These findings show that Tip55 is dispensable for the pre- and early post-implantation roles of Kat5, but is essential during organogenesis. Our results raise the possibility that isoform-specific functions of other chromatin regulatory proteins may play important roles in development.

Regulation of chromatin structure is necessary to establish and maintain cell type-specific gene expression patterns during development. Generation of active or repressive chromatin architecture at gene enhancers and promoters is necessary for the binding and activities of lineage specific transcription factors 1,2 . On a broader scale, large chromosomal domains of active or repressive chromatin structure help to enforce gene expression patterns particular to each cell lineage 3 . Consequently, mutations in subunits of chromatin remodeling factors-multisubunit protein complexes that regulate chromatin architecture through a range of enzymatic activities-frequently result in developmental abnormalities 4 .
The lysine acetyltransferase KAT5 (originally named TIP60) is conserved throughout eukaryotes and activates gene expression through acetylation of histones H2A (and H2A variants), H4, and numerous transcription factors [5][6][7][8][9] . The KAT5 acetyltransferase activity is also necessary for the cellular response to DNA damage, in part by remodeling chromatin structure near sites of DNA damage 7,8,[10][11][12] . Interestingly, KAT5 also has an essential non-catalytic role in regulation of chromatin accessibility and gene expression in embryonic stem cells and pre-implantation embryos 13 . KAT5 is a component of the 17-subunit TIP60-P400 complex, which remodels chromatin architecture not only through acetylation of histone tails, but also through incorporation of the H2A.Z histone variant into nucleosomes. Previously, individual knockdown of multiple subunits of TIP60-P400 resulted in loss of embryonic stem cell self-renewal, as well as a defect in differentiation 14,15 . Consistent with these phenotypes, mice with homozygous null mutations of the Kat5 gene arrest development at E3.5, prior to implantation, and Kat5 −/− blastocysts fail to hatch from the zona pellucida when cultured in vitro 16 .
Although most studies of Kat5 functions have focused on a single isoform, denoted Tip60α, several isoforms are expressed in mouse and human by virtue of alternative splicing. Four isoforms are expressed in mouse, Tip60α, Tip60β, LTip60, and Tip55, which differ by alternative inclusion of intron 1, exon 5, or a portion of intron 11 combined with early termination of the reading frame (Fig. 1A) [17][18][19] . The amino acid sequences of LTIP60 and TIP55 differ from TIP60α and TIP60β near or within the critical chromodomain and MYST domain, respectively, raising the possibility that these sequence alterations affect their functions. The Kat5 gene is broadly expressed throughout development, although it is unclear which isoforms are expressed in each tissue, owing to a lack of reagents for distinguishing the different isoproteins. Therefore, the extent to which each Kat5 isoform functions specifically or redundantly with other isoforms is currently unknown.
Here we test the hypothesis that Tip55 acts non-redundantly in development by specifically mutating this isoform within the mouse Kat5 gene. We show that Tip55 is necessary for embryonic development, with Tip55 homozygous mutant animals dying at or around E11.5. Prior to the appearance of overt developmental phenotypes, we observe defects in both heart and neural tube, which manifest as defects in cellular proliferation and increased cell death. Transcriptomic analyses of embryos prior to the appearance of developmental defects reveal alterations in developmental pathways that contribute to numerous tissues and organ systems. These data reveal a critical function of Tip55 in organogenesis that is distinct from the early, pre-implantation defect observed when all four Kat5 isoforms are lost and raise the possibility that additional isoforms of Kat5 have unique, non-redundant functions during development.

Results
Tip55 is essential for embryonic development. The TIP55 protein lacks the C-terminal 124 amino acids of the other three KAT5 isoforms, TIP60α, TIP60β, and LTip60. In its place, TIP55 has a unique 103 amino acid C-terminus encoded by a portion of intron 11 17 (Fig. 1A). This unique C-terminal region lacks amino acid sequence features of any well-established protein domain. We previously generated a multifunctional Kat5 mutant mouse that specifically eliminates the Tip55 isoform and can be converted to a catalytically inactive Kat5 allele upon Cre-mediated recombination. To generate the Tip55 null mutation, we deleted introns 11, 12, and 13, fusing exons 11-14, and in the process removing the region encoding the unique 103 amino acids of TIP55 (Fig. 1B). This allele eliminates Tip55 expression, but does not alter the coding sequence or expression of the remaining isoforms (Fig. 1A, Fig. S1A-C), which are able to form normal TIP60-P400 complexes 13 . We previously examined the role of the Kat5 lysine acetyltransferase (KAT) activity after Cre-mediated recombination of the Tip55 KO cassette, and found that these KAT-deficient mutant mice progress beyond pre-implantation stages but exhibit defects at or before gastrulation 13 . However, the question of whether Tip55 is required at any stage of development has not been addressed.
Tip55 is required for development of the heart and neural tube. Key early steps in organogenesis, including formation of the heart tube and closure of the neural tube, occur at approximately embryonic day 8 of mouse development, shortly before developmental defects are apparent in Tip55 Δ/Δ embryos. To investigate the possibility that Tip55 Δ/Δ embryos are defective in some aspects of organ formation, we first tested whether Tip55 was expressed in normal embryos at the initial stages of organogenesis. We performed RT-qPCR for Tip55, as well as for two additional Tip60 isoforms for which specific primers could be designed, LTip60 and Tip60β, using RNA isolated from E7.5-E11.5 embryos. We observed Tip55 expression at all stages, albeit to lower levels than observed for the other isoforms (Fig. S2). Furthermore, using whole-mount in situ hybridization and an intron 11 probe specific for the Tip55 isoform (dark gray boxes in Fig. 1A), we observed that expression of Tip55 was widespread in E8.5-E9.0 embryos (Fig. 2C).
Next, we examined several features of Tip55 +/+ and Tip55 Δ/Δ embryos at E8.5 (prior to the appearance of developmental defects in Tip55 Δ/Δ mutants) in embryo sections. We observed no obvious differences in organization of the body plan in sagittal sections of multiple embryos of each genotype (Fig. 3A). We performed immunohistochemistry with antibodies against Histone H3 phosphorylated at serine 10 (H3S10P) or cleaved caspase 3 (CC3) to test for proliferating or apoptotic cells, respectively. Although most regions of Tip55 Δ/Δ embryos were normal, the proportion of H3S10P positive cells was significantly reduced in heart, whereas cleaved caspase 3 staining was significantly elevated in neural tube (Fig. 3B-D). (Cleaved caspase 3 staining also appeared to be higher in heart in Tip55 Δ/Δ embryos, but the differences were not statistically significant.) These data suggest that loss of Tip55 results in defects in proliferation and/or increased apoptosis in multiple organs and further suggest that embryonic lethality caused by this mutation is due to defects in organ development.
We next examined whether the observed phenotypes were associated with a failure to specify the heart and neural lineages. To test this possibility, we performed immunofluorescence staining of wild type and mutant embryo sections for markers of neural lineage (SOX2) and cardiac muscle (cardiac troponin T, cTNT). As above, we stained sections from E8.5 embryos to minimize indirect effects of developmental arrest. However, we observed no difference in staining of either marker in Tip55 mutant embryos relative to wild type (Fig. S3). These data suggest that Tip55 loss impairs proliferation and viability of cardiac and neural progenitor cells downstream of cell type specification in both organs. Tip55 is required for proliferation of MEFs ex vivo. The reduced H3S10P staining of Tip55 Δ/Δ embryos in E8.5 heart suggested that TIP55 may promote cellular proliferation. However, a reduction in the number of proliferating cells could also result indirectly from increased cell death or other defects. Therefore, to directly test whether TIP55 was necessary for proliferation, we measured the proliferation rate and morphology of cells isolated from embryos with and without Tip55 mutations. Fibroblasts (MEFs) were isolated from Tip55 +/+ , Tip55 +/Δ , and Tip55 Δ/Δ embryos and cultured during a four-day time course. We found that Tip55 Δ/Δ MEFs proliferated minimally after isolation at E9.5, followed by growth arrest shortly thereafter (Fig. 4A). Furthermore, Tip55 Δ/Δ MEFs exhibited flattened and elongated cell morphology reminiscent of cells undergoing senescence, in contrast to Tip55 +/+ and Tip55 Δ/+ cells (Fig. 4B). Indeed, Tip55 Δ/Δ MEFs exhibited modest but reproducibly elevated staining for senescence-associated β-galactosidase activity compared to Tip55 +/+ and Tip55 Δ/+ MEFs (Fig. 4B). Together, these data suggest that Tip55 is required for the proliferation of some embryonic cell types and suppression of cellular senescence. Drivers of organ and tissue development are misregulated in Tip55 Δ/Δ mice. KAT5 isoforms act in part by acetylation of the N-terminal tails of histones H2A and H4 near gene regulatory regions, which promotes transcription 10 . To test whether genes required for development, cellular proliferation, or other processes were misregulated upon loss of the Tip55 isoform, we isolated RNA from Tip55 +/+ or Tip55 Δ/Δ embryos at E8.5, prior to the appearance of morphological defects in Tip55 Δ/Δ embryos (Fig. 2B), and performed RNA-seq. We observed strong concordance among three biological replicates for Tip55 +/+ and Tip55 Δ/Δ (Fig. 5A,B). Next, we used EBseq analysis 20 to identify genes that are significantly differentially expressed in Tip55 Δ/Δ embryos. We identified 2507 genes that were significantly misregulated in Tip55 Δ/Δ embryos (posterior probability of differential expression, PPDE > 0.95), with 278 of those genes misregulated more than two-fold ( Fig. 5C and Table S1). We validated several differentially expressed genes by RT-qPCR and observed results consistent with the RNA-seq data (Fig. 5D).
To uncover classes of genes significantly misregulated in Tip55 Δ/Δ embryos, we identified Gene Ontology (GO) terms significantly enriched among the differentially expressed genes. These analyses demonstrated that the differentially expressed genes were highly enriched for developmental regulators and genes required for development of multiple tissue types, including blood, muscle, and brain (Fig. 5E,F). Kat5 is necessary for proper silencing of a number of developmental regulators in embryonic stem cells 13 . Therefore, genes upregulated in Tip55 Δ/Δ embryos may include some that are repressed by TIP55. In addition, developmental delays in Tip55 Δ/Δ embryos may cause expression of genes that peak early in wild type embryos (and normally exhibit reduced expression by E8.5) to be more highly expressed in Tip55 Δ/Δ mutants at E8.5 as a result of their delayed or extended window of expression.
Among numerous developmentally regulated genes that were downregulated in Tip55 Δ/Δ embryos were factors that promote blood cell development -including key erythroid factors Gata1 and Klf1 -suggesting Tip55 loss leads to a failure to initiate erythropoiesis. However, these data do not distinguish whether the effect of Tip55 Δ/Δ on erythroid development is due to a direct role for TIP55 at the regulatory regions of key erythroid regulators or an indirect effect on gene expression due to other developmental abnormalities (such as in heart). In sum, we conclude that Tip55 is necessary in multiple cell and tissue types for normal expression of differentiation genes that are critical for organ development.

Discussion
Here we have shown an essential, non-redundant role of the Tip55 isoform of Kat5 during post-implantation embryonic development. These findings contrast with the phenotype of a mouse mutant that lacks all four Kat5 transcripts, which causes pre-implantation lethality at the blastocyst stage 16 , as well as a KAT-deficient mutant that exhibits defects in gastrulation 13 . The defects of Tip55 Δ/Δ embryos in formation of heart and neural tube suggest this mutation impairs development of multiple cell lineages, a finding that is verified by our transcriptomic data indicating mis-expression of developmental regulators of multiple tissue types. On a cellular level, Tip55 Δ/Δ results in a combination of reduced proliferation and increased apoptosis in several different tissues and cell  types. Therefore, widespread developmental defects may result from poor proliferation or survival of cells that comprise multiple tissues. Alternatively, these cellular phenotypes may result from more complex developmental impairments owing to defects in the gene regulatory networks specific to each lineage.
The finding that Tip55, a poorly-studied splice variant of Kat5, plays specific roles in development raises the possibility that each of the other Kat5 isoforms play non-redundant roles in development that have yet to be discovered. This raises a larger question regarding regulation of developmental gene expression in mammals -do additional chromatin regulatory enzymes that are expressed in multiple isoforms have distinct functions that are specific for each isoform? Although the functions of splice variants of a few mammalian chromatin regulatory enzymes have been addressed in vitro 21,22 , the generation of isoform specific mutant animals will be necessary to assess their potential functions in vivo. Such studies have been performed in Drosophila for the SWI/SNF family ATPase, domino, in which two splice isoforms were found to function in distinct cell types 23 . Although multiple isoforms of one of two mammalian homologs of domino, Ep400, have been identified, it remains to be tested whether they exhibit similar cell type specificity. Therefore, an effort to make and characterize isoform-specific alleles of numerous mammalian chromatin remodeling enzymes will likely provide important insights into developmental gene regulation.

Generation of Tip55 knockout mice. All animal experiments were approved by the Institutional Animal
Care and Use Committee of University of Massachusetts Medical School (approval number A-2165). All animal procedures were performed in accordance with UMMS and NIH guidelines on animal care. The mutant mouse line with the Tip55 deletion is equivalent to the parental line in which the Tip60 ci allele was generated by Cre mediated recombination of fused exons 11-14, placing a catalytically inactive version of exon 11 into the Kat5 gene. This parental allele (prior to Cre mediated recombination) specifically lacks the Tip55 isoform 13,14 . Mice were genotyped by PCR with primers listed in Table S2. Tip55 Δ/+ mice were maintained as heterozygotes on an inbred FVB/N background and intercrossed to generate Tip55 +/+ , Tip55 Δ/+ and Tip55 Δ/Δ embryos.
RNA in situ hybridization. Whole mount in situ hybridization of wild type embryos were performed as previously described 24 . A unique region of Tip55 cDNA (293 bases) was used to generate Tip55 sense and anti-sense probes.
Derivation of MEFs. MEFs were generated from E9.5 embryos as previously described 25 . Briefly, E9.5 embryos were dissected, and trypsin (0.05%) digested for 12 minutes at 37 °C. Embryos were pipetted to obtain single cells and cultured in a 12 well plate in DMEM with 10% FBS.
Cell proliferation assays. MEFs from two independent embryos from each genotype (Tip55 +/+ , Tip55 Δ/+ and Tip55 Δ/Δ ) were seeded at approximately 50% confluency and cultured in wells of a 12-well plate. The total number of cells were counted and re-plated into new 12-well plates every 48 hours for a total of four days to measure proliferation rate. β-galactosidase staining. MEFs were isolated from E9.5 embryos and cultured in a 12-well plate. Three days after plating, cells were sub-cultured for an additional two days followed by staining for β-galactosidase activity using a kit (Millipore KAA002) according to manufacturer's protocol.
Hematoxylin and Eosin (H&E) Staining. E8.5 embryos were collected and sectioned at 8μm thickness for morphological analysis as previously described 26 . Hematoxylin-and eosin-staining was performed by de-paraffinizing sections in xylene, rehydrating slides through an ethanol gradient, staining for 30 s with 30% Harris modified hematoxylin and a 30 s counterstain with eosin Y. Slides were rinsed and dehydrated with ethanol, cleared with xylene, and mounted using Vectashield mounting media. Immunohistochemistry and Immunofluorescence. Sections from E8.5 embryos were examined for proliferation and apoptosis defects by immunohistochemistry, following protocols described previously 26 . Briefly, sections were rehydrated through an ethanol gradient, followed by heat antigen retrieval (Buffer A, Electron Microscopy Sciences). Immunostaining was conducted using the Vectastain Elite ABC and DAB Peroxidase Substrate kit according to manufacturer guidelines. Sections were incubated with H3S10P (1:100; Cell signaling, CSG 9706) or cleaved caspase 3 (1:100; Cell signaling, CSG 9661) biotinylated primary antibodies overnight at 4 °C 26 . For counterstaining, slides were rinsed and then incubated with 30% hematoxylin for 30 s after developing staining with 3,3′ diaminobenzimidine. All slides were ethanol-dehydrated, cleared with xylene, and mounted with Vectashield mounting medium.
Expression of cardiac and neural markers in embryonic sections was examined by immunofluorescence. E8.5 embryo sections were rehydrated, then subjected to heat antigen retrieval as described above. Sections were incu- Proliferation and Apoptosis Quantification. For quantification of H3S10P or cleaved caspase 3 immunostaining, digital images were taken using a Nikon Eclipse 80i microscope and the NIS-Elements 4.00.03 software. Positively stained cells were counted manually using ImageJ (v 1.6.0_65). The percentage of H3S10P positive or cleaved caspase-3 positive cells relative to the total number of nuclei was calculated for a minimum of three embryos per genotype. RNA sequencing. RNA was isolated from E8.5 embryos using a Direct-zol RNA MicroPrep kit (Zymo research). Enrichment of mRNA, library preparation, and sequencing were performed at BGI, using the BGI-seq format. Reads were mapped to the mm10 genome and quantified using RSEM 27 . Identification of differentially expressed genes was performed using a combined RSEM-EBseq pipeline 20 . Gene ontology enrichment was performed using Metascape (http://metascape.org) 28 .

Data Availability
Sequence reads for all RNA-seq libraries have been deposited at Gene Expression Omnibus (GEO) and are available with accession number: GSE111691.