Divergent functions of histone acetyltransferases KAT2A and KAT2B in keratinocyte self-renewal and differentiation

The GCN5-related N-acetyltransferase (GNAT) family includes the evolutionarily conserved KAT2A (also known as GCN5), first discovered in yeast (Georgakopoulos and Thireos, 1992), and its vertebrate-specific paralog, KAT2B (also known as PCAF) (Yang et al., 1996). Both were initially described as transcriptional activators conferred by their intrinsic histone acetyltransferase (HAT) activity at promoters (Brownell et al., 1996; Yang et al., 1996). Human KAT2A and KAT2B exhibit high amino acid sequence identity (∼73%) and consist of three distinct domains: (1) a C-terminal catalytic HAT domain that preferentially acetylates histone H3 on lysine 14 (H3K14), lysine 9 (H3K9) and lysine 18 (H3K18) (Kuo and Andrews, 2013); (2) a bromodomain that recognizes acetyl-lysine residues and regulates nucleosome remodeling and substrate specificity (Syntichaki et al., 2000; Cieniewicz et al., 2014); and (3) a metazoan-specific N-terminal extension that may be involved in nucleosome recognition or in facilitating interactions with other proteins (Smith et al., 1998; Xu et al., 1998). Despite their homology, Kat2a and Kat2b are typically expressed mutually exclusively from mouse embryogenesis onwards and are not always functionally redundant, as Kat2a-null mice are developmentally delayed and die by E10.5 whereas Kat2b-null mice live into adulthood (Xu et al., 2000). However, double Kat2a/Kat2b mutants die earlier than single Kat2a-null mice, indicating that both HATs participate in overlapping and distinct roles in vital developmental processes (Yamauchi et al., 2000).

Numerous in vitro studies indicate a role for KAT2A in regulating stem cell pluripotency and differentiation. Kat2a-null mouse embryoid bodies were smaller with disorganized epiblasts and showed significant enhancements in their rate of myogenic differentiation (Lin et al., 2007; Wang et al., 2018). Similarly, chemical inhibition of KAT2A in mouse embryonic stem cells reduced pluripotency and accelerated mesendodermal differentiation (Moris et al., 2018). Furthermore, depletion of KAT2A in mouse neural stem cells and human leukemic cells produced a greater propensity to differentiate toward the oligodendrocyte lineage and myeloid lineage, respectively (Martínez-Cerdeño et al., 2012; Bararia et al., 2016; Tzelepis et al., 2016; Domingues et al., 2017). Interestingly, KAT2A has also been shown to drive cellular differentiation of adipocytes, osteogenic cells and mesenchymal stem cells (Wiper-Bergeron et al., 2007; Xu et al., 2020). In comparison, KAT2B was predominantly associated with cellular differentiation of human embryonic stem cells, mesenchymal stem cells, myoblasts, leukemia cells and primary human keratinocytes (Puri et al., 1997; Pickard et al., 2010; Zhang et al., 2016; Du et al., 2017; Sunami et al., 2017).

The redundant and non-redundant functions of both HATs in balancing self-renewal and cellular differentiation within a single organ system has not been previously defined. The epidermis represents an ideal setting to address these questions as it is a highly regenerative tissue harboring a large pool of stem cells that undergo a well-characterized progressive sequence of differentiation (Fuchs, 2007; Sotiropoulou and Blanpain, 2012; Liu et al., 2013). In vivo studies over the past two decades have revealed an important role for histone demethylases and histone deacetylases in coordinating this process, yet there is a paucity of research implicating HATs (Sen et al., 2008; Ezhkova et al., 2009; LeBoeuf et al., 2010; Liakath-Ali et al., 2014; Botchkarev, 2015). Pickard et al. (2010) have shown that KAT2B promotes in vitro keratinocyte differentiation by acetylating a non-histone substrate. However, the role KAT2A may play in regulating epidermal homeostasis alongside the functions of KAT2B has not been compared.

We report in this study that, unlike KAT2B, KAT2A functions before keratinocyte differentiation to support gene expression, to catalyze global levels of H3K9 acetylation (H3K9ac) and to maintain morphologies associated with the self-renewing cell state. We show that both the acetyltransferase domain and the metazoan-specific N-terminal domain, but not the bromodomain, of KAT2A is essential to mediate these effects in keratinocytes. Finally, we provide the first in vivo reports of an interplay between KAT2A and KAT2B in mouse skin, and show that changes to KAT2A and KAT2B levels lead to significant delays in cutaneous wound healing.

To identify candidate HATs with potential roles in epidermal homeostasis, we compared the expression of key members of the three major HAT families (GNAT, CBP/p300 and MYST) in human N/TERT-1 keratinocytes (herein also referred to as N/TERTs) at subconfluence and after 6 days of Ca²⁺ treatment to induce terminal differentiation (Fig. 1A). The GNAT HATs KAT2A and KAT2B exhibited the largest changes in expression between the undifferentiated (subconfluent) and differentiated (day 6) cell states. The expression changes for the paralogous HATs were distinctly divergent, with KAT2A becoming downregulated while KAT2B was upregulated upon keratinocyte differentiation. The downregulation of KAT2A and upregulation of KAT2B transcripts were detectable as early as 2 days after Ca²⁺ treatment (Fig. 1B). Progressive down- and up-regulation of KAT2A and KAT2B, respectively, were similarly observed upon Ca²⁺-induced differentiation of primary human keratinocytes (NHEK) (Fig. 1B). At the protein level, KAT2A was significantly decreased from day 4 of differentiation and almost undetectable at day 6 in both N/TERT-1 and NHEK cells (Fig. 1C). KAT2B protein was present at low levels before the onset of differentiation, with levels significantly increasing by day 2 of Ca²⁺-induced differentiation.

We asked whether the divergent profiles of KAT2A and KAT2B measured in vitro are consistent with expression in the adult human epidermis. Immunofluorescence staining showed KAT2A in the undifferentiated basal cells and spinous keratin 10 (KRT10)-positive cells in the suprabasal layer, whereas weak to no KAT2A was detected in the most terminally differentiating cells (Fig. 1D; Fig. S1A). All commercial KAT2B antibodies that we tested yielded non-specific signals in immunolabeled tissue sections, so we analyzed KAT2B transcripts in the tissues by in situ hybridization. KAT2B mRNA was rarely detectable in the basal layer of the epidermis but was much more abundant in the suprabasal layers, particularly in terminally differentiating cells (Fig. 1E).

Despite being structurally similar paralogs, we showed that KAT2A and KAT2B exhibit divergent expression profiles across keratinocyte differentiation, suggesting that they may have distinct functions in balancing epidermal homeostasis. This prompted us to pursue a comparative analysis of KAT2A and KAT2B function in keratinocytes.

To characterize the functions of KAT2A and KAT2B in keratinocytes, we generated N/TERT-1 keratinocytes that stably expressed shScramble (shSCR), shKAT2A (sh2A) and shKAT2B (sh2B). The sh2A and sh2B hairpins mediated a specific reduction in the respective KAT2A and KAT2B transcripts and proteins in the keratinocytes (Fig. 2A,B). A second set of keratinocytes expressing shKAT2A (sh2A#2) and shKAT2B (sh2B#2), targeting a different region of the KAT2A and KAT2B transcripts, also had a specific reduction of KAT2A and KAT2B transcripts and proteins (Fig. S1B–D). Of note, KAT2B expression was increased in the undifferentiated sh2A N/TERTs and KAT2A expression was upregulated in the differentiated sh2B N/TERTs, suggestive of a compensatory response upon depletion of either HAT (Fig. 2B,C; Fig. S1B,C). In view of this, cells that expressed both shKAT2A and shKAT2B (sh2A+2B) were generated to investigate potential redundancies between both HATs (Fig. 2A,B).

Depletion of KAT2A or KAT2B, or both, had no overt effects on cell growth (Fig. 2D). Growth morphology of subconfluent keratinocytes depleted for KAT2A, but not for KAT2B, were altered compared with the shSCR control cells. In contrast to the loosely associated shSCR and sh2B cells, the sh2A and sh2A+2B N/TERTs formed compact cell clusters when cultured under subconfluent non-differentiating conditions (Fig. 2E; Fig. S1E). Quantification of isolated cells in the cultures showed significantly fewer cells without neighboring contact upon depletion of KAT2A and of KAT2A and KAT2B (Fig. 2F). The compacted morphology led us to analyze the cells for changes in genes mediating cell–cell and cell–substratum adhesion. The expression of tight junction and desmosomal genes such as claudin 1 (CLDN1), occludin (OCLN), desmoglein 1 (DSG1), desmocollin 1 (DSC1) and periplakin (PPL) but not adherens junction genes E-cadherin (CDH1) and β-catenin (CTNNB1) were increased in sh2A and sh2A+2B keratinocytes compared with control and sh2B cells (Fig. 2G; Fig. S1F). We further analyzed gene expression changes in matrix metalloproteinases (MMPs) that are known to contribute to epithelial cell migration and adhesion by modulating the extracellular matrix (Chen and Parks, 2009). Significant upregulation of the collagenases MMP1 and MMP13, and the stromelysins MMP3 and MMP10 was detected in sh2A N/TERTs but not sh2B N/TERTs, with an even greater change in cells depleted for both HATs (Fig. 2H; Fig. S1F). The expression of vimentin (VIM), which has been shown to regulate cell adhesion and migration in cultured keratinocytes (Velez-delValle et al., 2016), was significantly downregulated in undifferentiated sh2A and sh2A+2B N/TERTs and upregulated in sh2B N/TERTs (Fig. 2I,J). These results show that knockdown of KAT2A, but not KAT2B, in proliferative N/TERTs causes changes in growth morphology, and in genes regulating cell adhesion and migration.

To further clarify the roles KAT2A and KAT2B play in keratinocyte self-renewal and differentiation, we performed RNA-seq analysis to assess global transcriptomic changes upon KAT2A and/or KAT2B depletion in N/TERTs at a proliferative timepoint (subconfluency) and a non-proliferative timepoint primed for differentiation (confluency) and after 4 days of differentiation (Fig. 3A; Fig. S2A). Principal component analysis (PCA) revealed distinct clustering of samples according to the cell states at subconfluence, confluence and D4 of differentiation (Fig. 3B). KAT2A and KAT2B depletion produced different effects at different timepoints. At subconfluence, sh2A N/TERTs segregated distinctly from the shSCR control and sh2B cells, while at day 4 of differentiation a strong divergence of sh2B N/TERTs away from the shSCR and sh2A N/TERTs was observed. Notably, the transcriptome profiles of the double sh2A+sh2B cell line were more comparable with sh2A N/TERTs than that of sh2B N/TERTs at all three timepoints (Fig. 3B). Categorization of differentially expressed genes (DEGs) at subconfluency identified a small number of DEGs for sh2B N/TERTs (n=48) and over twice as many for sh2A N/TERTs (n=115; Fig. 3C; Table S5). Double knockdown resulted in larger numbers of DEGs at all timepoints suggesting that compensatory expression of KAT2B upon KAT2A knockdown may regulate a subset of genes redundantly with KAT2A.

Unsupervised hierarchical clustering, gene set enrichment analyses and gene ontology of the DEGs revealed aberrant upregulation of genes involved in epidermis development and keratinocyte differentiation at the subconfluence timepoint in sh2A and sh2A+2B N/TERTs (Fig. 3D,E; Fig. S2B). Conversely, KAT2B-depleted cells exhibited downregulation of these genes at day 4 of differentiation. Increased levels of tight junction and desmosomal genes in sh2A and sh2A+2B N/TERTs at subconfluence were also detected in the RNA-seq analyses, consistent with the clustering phenotype exhibited by these cells (Fig. S2C). Validation of the gene expression changes by qPCR showed that key genes involved in both early–mid differentiation, keratin 10 (KRT10), involucrin (IVL), filaggrin (FLG), as well as late cornification processes, late cornified envelope (LCE1A), transglutamase 3 (TGM3) and kallikrein related peptidase 13 (KLK13), were significantly upregulated in sh2A N/TERTs at subconfluence, whereas mid–late differentiation genes (IVL, FLG, LCE1A and TGM3) were downregulated in sh2B N/TERTs at day 4 post induction of differentiation (Fig. 3F). These results suggest that KAT2A and KAT2B acted in opposing ways in regulating the keratinocyte differentiation gene expression program.

We further characterized the functional impact of transcriptional changes induced by the loss of KAT2A or KAT2B. Control, sh2A and sh2B N/TERTs were cultured to confluence and induced to terminally differentiate in high Ca²⁺ media over 6 days. At confluence (day 0), KAT2B-depleted cells were morphologically similar to control, but the KAT2A-depleted cultures formed patches of stratified growth (Fig. 4A). At day 2, small cell colonies, which continued to enlarge over time, became apparent in the KAT2A-depleted cultures. By contrast, such stratified colonies were observed only in the control and sh2B− cultures at day 4 of differentiation (Fig. 4A). By day 6, cornification of the keratinocytes that resulted in dark-brown patches was observed in all cultures. The number of cornified clusters was consistently higher in the KAT2A-depleted cultures at both day4 and day 6 of differentiation compared with shSCR and sh2B cultures (Fig. 4B). These results showed that KAT2A knockdown keratinocytes acquired more advanced states of differentiation at earlier timepoints compared with the control and sh2B cells.

Consistent with the morphological phenotypes, expression levels of key early and late markers of keratinocyte differentiation and cornification, such as KRT10, IVL, FLG, KLK13 and LCE1A, were elevated in sh2A N/TERTs under self-renewing conditions and at confluence (Fig. 4C). Western blotting for KRT10 and IVL revealed that both proteins were readily detectable in subconfluent cultures of sh2A N/TERTs but not sh2B N/TERTs or control cells (Fig. 4D). Profilagrrin and filaggrin monomers were detected in sh2A N/TERTs by day 2 of differentiation, in contrast to day 4 of shSCR control cells (Fig. 4E). The sh2A N/TERTs also had a greater than threefold increase in IVL proteins at 4 days post-differentiation compared with control cells, whereas sh2B N/TERTS exhibited lower levels of both FLG and IVL. Of note, the levels of the early differentiation marker KRT10 at mid to late differentiation timepoints were similar in all three cell lines, indicating that depletion of KAT2A had a predominant effect on proliferative and early differentiating keratinocytes. Subconfluent cultures of primary human keratinocytes (NHEKs) exhibited similar upregulation of differentiation marker gene expression and significant increases in the levels of IVL and TGM1 upon knockdown of KAT2A (Fig. 4F,G; Fig. S1G).

We further generated organotypic cultures of the N/TERT cell lines to assess their ability to stratify and terminally differentiate in 3D, which more faithfully recapitulates differentiation in vivo (Fig. 4H). After eight days of air-liquid interface culture, KAT2A-depleted keratinocytes formed epidermal constructs that had more stratified layers and cornified cells, relative to control and sh2B N/TERTs (Fig. 4H,I). Co-immunostaining of the constructs for KRT14 and KRT10 also revealed more layers of KRT10-positive cells in the sh2A epidermal constructs compared with the shSCR control (Fig. 4H). By contrast, the sh2B constructs exhibited few KRT10-positive layers.

Taken together, these results indicate that KAT2A normally acts to inhibit differentiation in proliferative keratinocytes such that depletion of KAT2A leads to the cells acquiring a differentiation-primed state in the absence of inducing signals and thus respond more rapidly to differentiation initiation cues. KAT2B, conversely, supports keratinocyte differentiation upon induction, a finding that is consistent with previous reports (Pickard et al., 2010).

As the function of KAT2B in epidermal differentiation has been previously reported, we focused our attention on understanding the mechanisms by which KAT2A blocks premature differentiation in proliferative keratinocytes. We first assessed the functional contributions of the different domains of KAT2A to the changes in growth morphology and gene expression. This was achieved by performing experiments in sh2A N/TERTs transduced with empty vector, shRNA-resistant full-length KAT2A (FL), bromodomain-deleted KAT2A (delBr), N-terminus-deleted KAT2A (delN) or catalytically dead KAT2A (mAT_E575R, referred to herein as mAT) (Fig. 5A). Western blotting showed the overexpression of the FLAG-tagged wild-type and mutant KAT2A proteins in transduced keratinocytes compared with endogenous KAT2A levels (Fig. 5B).

The aberrant growth morphologies observed in KAT2A-depleted N/TERTs were reversed to normal by overexpressing FL- and delBr- but not by mAT- or delN-KAT2A (Fig. 5C; Fig. S3A). In addition to morphological changes, premature expression of differentiation markers, KRT10, IVL, TGM1 and LCE3D, were also found to be effectively reduced in sh2A N/TERTs overexpressing FL- and delBr-KAT2A but remained high in mAT- and delN-KAT2A-expressing cells (Fig. 5D,E). In normal keratinocytes, endogenous KAT2A exhibited a predominant nuclear localization before and over the course of differentiation, indicating that KAT2A functions in these cells by primarily targeting nuclear substrates (Fig. S4A,B). As deletion of the N-terminal domain of KAT2A resulted in the removal of a putative nuclear localization signal, we confirmed by immunostaining that the delN-KAT2A constructs were localized to the nucleus (Fig. 5F; Fig. S3B). Thus, the ineffective reversal of the KAT2A depletion phenotype by the delN mutant was not the result of mislocalization but reflected a key functional contribution of the N-terminal domain to KAT2A activity in keratinocytes.

The 360-residue long N-terminal domain in KAT2A is conserved in higher vertebrates but is not present in yeast GCN5. As this domain does not contain specific functional motifs, we sought to determine specific regions in the N terminus that supported KAT2A activity in the keratinocytes. To this end, we generated and tested constructs encoding HA-tagged KAT2A with N-terminal truncations of amino acids 1–97 (delN₉₇), 1–259 (delN₂₅₉) and 1–361 (delN) for their ability to reverse the KAT2A-knockdown phenotypes (Fig.S3B,C). Additionally, we included a construct encoding only the N-terminal domain in the assay to investigate trans- regulatory effects of the domain. Expression of either delN₉₇ or delN₂₅₉ in KAT2A-knockdown N/TERT1 cells resulted in partial reversal of the growth and gene expression phenotypes, with the delN₉₇ construct having a stronger effect on the cellular morphology (Fig. S3D–F). Expression of the N-terminal domain by itself had no effects. These results suggest that the region between residues 260–362, which is computationally predicted to be structurally localized near to the bromodomain (Varadi et al., 2022), is important in regulating KAT2A function in keratinocytes (Fig.S3G).

KAT2A and KAT2B have been shown to act on both histone and non-histone substrates. We assessed the global levels of histone modifications previously associated with KAT2A and KAT2B activity, and observed that in self-renewing N/TERTs, only H3K9ac was significantly decreased in KAT2A-depleted cells (Fig. 5G). A further reduction in the levels of H3K9ac was observed in cells depleted for both HATs, suggesting that the upregulation of KAT2B upon KAT2A knockdown contributed to H3K9ac in the absence of KAT2A (Figs 2B,C and 5G). Of note, depletion of KAT2B in differentiating keratinocytes with a concomitant increase in KAT2A did not significantly alter the levels of H3K9ac in these cells (Fig. 2B,C; Fig. S4C). We examined the functional contribution of each KAT2A domain in regulating H3K9ac and found that only the HAT domain was required for this activity (Fig. 5H). The N-terminal domain, although important for KAT2A function, did not affect the acetylation of histone substrates. Taken together, these results revealed distinct requirements for the acetyltransferase catalytic activity and N-terminal domain in mediating the function of KAT2A in self-renewing keratinocytes.

Our data suggest that KAT2A and KAT2B act distinctly to balance epidermal differentiation, but these conclusions are derived from cultured human keratinocytes and may not be compatible under in vivo conditions. To address this issue, we turned to using mouse models as the murine epidermis suitably represents many cellular and molecular events that occur during human epidermal homeostasis.

Constitutive knockout (KO) of Kat2a is embryonically lethal, but Kat2b-null mice are viable and develop normally to adulthood. Therefore, we focused on assessing the in vivo functions of KAT2B in the epidermis. Primary mouse keratinocytes and fibroblasts isolated from age-matched adult mice demonstrated the absence of KAT2B protein in the Kat2b KO mice, but heterozygotes contained similar KAT2B levels as wild-type mice (Fig. 6A). Notably, KAT2A levels were significantly increased in both Kat2b^−/− keratinocytes and fibroblasts, consistent with the compensatory upregulation observed upon knockdown in the human keratinocytes (Fig. 2B,C and Fig. 6A). Histological analysis of neonatal and adult skin revealed no overt epidermal defects in the Kat2b KO tissues at homeostasis (Fig. 6B; Fig. S5). However, distinct defects were observed upon disruption of homeostatic epidermal regeneration. In wild-type animals, dorsal excisional wounds of ∼1 cm diameter gradually closed over a 12-day time course. By contrast, a notable delay in healing was observed in the Kat2b KO mice in which wounds started to constrict only 2 days after wounding and complete closure was not observed by day 12 (Fig. 6C,D). Histological analysis further showed that re-epithelialization and generation of the epidermis in KAT2B KO wounds was significantly reduced compared with wild type (Fig. 6E,F).

These results demonstrate the involvement of KAT2B in the tissue healing processes, particularly at early timepoints after injury, suggestive of a role for the KAT2A- and KAT2B-mediated regulation of keratinocyte proliferation and differentiation in the regeneration of the epidermis upon cutaneous wounding.

In this study, we analyzed the interplay of the mammalian GCN5 homologs KAT2A and KAT2B in regulating epidermal homeostasis. Our work shows that KAT2A plays a key function in supporting the self-renewing state of keratinocytes, in contrast to its paralog KAT2B, which acts to promote epidermal differentiation. Notably, KAT2A and KAT2B are inversely expressed in self-renewal and differentiating keratinocytes, and although the reciprocal paralog become upregulated in response to depletion of the dominant KAT, complete functional redundancy between the two acetyltransferases is not observed both in vitro and in vivo.

KAT2A has been shown to support proliferation and regulate differentiation processes in stem and progenitor cells, including mouse embryonic stem cells, neural progenitors and periodontal stem cells (Lin et al., 2007; Martínez-Cerdeño et al., 2012; Li et al., 2016). Using unbiased RNA-seq analyses, we demonstrated that keratinocyte differentiation genes were dysregulated and prematurely expressed before differentiation induction in the absence of KAT2A. This result mirrored findings in leukemic cells depleted for KAT2A, which have a higher propensity to exit self-renewal and enter differentiation (Domingues et al., 2020). We also detected an enhancement in the rate of differentiation of KAT2A-depleted N/TERTs, similar to that observed upon skeletal muscle differentiation of Kat2a^−/− mESCs (Lin et al., 2007). Our results thus provide further evidence that KAT2A is a major regulator of the stem cell state.

KAT2A may regulate key components of self-renewal and differentiation pathways in two ways: histone acetylation at the promoters to facilitate gene expression of target genes; or acetylation of non-histone substrates, including transcriptional regulators such as MYC and CEBPA (Lin et al., 2007; Martínez-Cerdeño et al., 2012; Li et al., 2016). Our data show that KAT2A is a major effector of H3K9ac in keratinocytes, suggesting that KAT2A may act predominantly by regulating the promoter activities of target genes in these cells. It is, however, possible that KAT2A also acetylates nuclear non-histone proteins to exert its influence on self-renewal in basal keratinocytes. It will be interesting, in future experiments, to elucidate and compare the full KAT2A and KAT2B acetylomes in self-renewing and differentiating keratinocytes using high-throughput proteomics analyses (Fournier et al., 2016).

Rather than direct regulation of specific transcription factors, the results in this study point in favor of a histone acetylation-mediated mechanism underlying the role of KAT2A in regulating self-renewal and differentiation in epidermal keratinocytes. The expression of a large subset of the epidermal differentiation genes before inductive signaling, coupled with a strong decrease in H3K9ac and pan-histone H3 acetylation levels, which is observed upon KAT2A knockdown, are highly consistent with the model of transcriptional noise buffering in cell fate transitions (Raser and O'Shea, 2005; Domingues et al., 2020; Arede and Pina, 2021). In this model, KAT2A promotes the basal keratinocyte self-renewing state by facilitating consistent transcriptional output through H3K9ac at promoters. Loss of KAT2A and concomitant reduction of H3K9ac result in irregular promoter activities that give rise to high transcriptional variability and to consequential loss of cellular identity and stem-like functions. A limitation of the present study is the lack of genome level insight into the levels of H3K9ac and occupancy of KAT2A in control and KAT2A-depleted N/TERTs. Future work to address these issues using ChIP-seq would reveal where changes in H3K9ac occur across the genome and whether they are a direct consequence of KAT2A loss.

Our study gives unique insight into the contribution of the N-terminal domain in regulating KAT2A activity in mammalian cells. The N-terminal domain was found to have a strong effect on regulating KAT2A activity in a manner that was apparently distinct from the bromodomain at a subset of target genes. Given that this domain arose with the emergence of metazoans, it would be of interest to explore whether the N-terminal domain specifically aids KAT2A in regulating genes and cellular processes that are important for multicellular development and homeostasis.

It is also intriguing to observe perturbed keratinocyte behavior during cutaneous wound healing in Kat2b KO mice, which did not exhibit aberrant skin phenotype under homeostatic conditions. Our in vitro data demonstrate KAT2B as a strong driver of the epidermal differentiation gene expression program that is not essential for the completion of terminal differentiation. This conclusion is consistent with experiments in primary human keratinocytes performed by Pickard et al. (2010) and is in line with the lack of epidermal phenotype we observed in Kat2b KO mice. Together, these results imply that loss of KAT2B yields differentiated keratinocytes with an improper molecular composition, which could compromise the integrity of the epidermis and its ability to respond to external stress factors. This may account for the delays in cutaneous wound healing we observed in Kat2b KO mice. Alternatively, KAT2B may play a more direct role in coordinating the dramatic transcriptional changes that occur soon after wounding. Although such a role was not apparent in our in vitro characterization of KAT2B function, we did find many genes important for epithelial migration and adhesion to be regulated by KAT2A, which seemed to affect growth morphology. Therefore, the functions of KAT2A and KAT2B may be more important during cutaneous wound healing than normal epidermal homeostasis. Indeed, histone deacetylase (HDAC) inhibitors have been shown to accelerate wound healing in mice (Spallotta et al., 2013a). More specifically, chemical activation of KAT2B has been shown to enhance lysine acetylation in keratinocytes and promote wound repair in mice (Spallotta et al., 2013b). Both studies strongly support that the delayed wound healing phenotype we observed in Kat2b KO mice is due to changes in keratinocyte function rather than to any other cell type. It is also important to note that the compensatory upregulation of KAT2A observed in cultured keratinocytes was also detected in vivo in the absence of KAT2B. This compensatory increase in KAT2A expression in Kat2b KO skin highlighted the closely interlinked, yet non-redundant, relationship between the two acetyltransferases in mediating epidermal homeostasis and regeneration. Embryonic lethality in Kat2a KO mice had precluded the analysis of KAT2A function in adult skin. The use of K14-Cre transgenic mice to conditionally delete Kat2a singly or in combination with Kat2b KO, would be valuable to further dissect the in vivo contribution of these enzymes to skin regeneration and healing.

In summary, our results show that KAT2A is enriched in undifferentiated keratinocytes where it functions via its HAT activity to maintain the self-renewal state and balance the pro-differentiation functions of KAT2B (Fig. 7A). The correct balance in KAT2A and KAT2B activities may be important in promoting effective re-epithelization during cutaneous wound healing (Fig. 7B).

Immortalized keratinocytes (NTERT1, Dickson et al., 2000) and Primary Neonatal Normal Human Epidermal Keratinocytes (NHEK-Neo, Lonza) were grown in T25 flasks (Thermo Fisher Scientific) under feeder-free conditions in complete Keratinocyte Serum-Free Media (KSFM, Gibco) or complete Keratinocyte Growth Medium Gold (Lonza). To induce differentiation, NTERTs and NHEKs were cultured to 70% and 100% confluency, respectively, then grown in low-Ca²⁺ differentiation media DFK1 overnight (day 0 of differentiation) before switching to high-Ca²⁺ differentiation media DFK2. See Table S1 for media formulations. All cell lines were tested monthly for mycoplasma contamination using the MycoAlert Mycoplasma detection kit (Lonza).

Stable knockdown cell lines were generated in the presence of 8 µg/ml polybrene by lentiviral transduction with pLenti constructs constitutively expressing shRNA sequences provided by TransOMIC Technologies (Table S2). NTERT transductants were selected with 1 µg/ml puromycin or 4 µg/ml blasticidin for 1 week, while NHEK-Neo transductants were selected by flow sorting cells strongly expressing the ZsGreen fluorescent marker. KAT2A expression constructs were generated from the full-length wild-type KAT2A-coding sequence (CDS) purified from NTERT-derived cDNA and cloned into the pCR BluntII-TOPO vector (Thermo). Silent mutations at the target regions of shKAT2A were generated by using the Site-Directed Mutagenesis kit (NEB). The resultant shRNA-resistant full-length KAT2A CDS was PCR amplified and subcloned into the pLenti-P2A-Bsd vector. The mAT KAT2A was generated by swapping guanine and adenine at bases 1723–1724 using the QuickChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies). The delBr KAT2A and the delN KAT2A were generated by PCR amplifying bases 1–2169 and 1084–2514 of KAT2A CDS, respectively. A FLAG tag sequence was introduced N-terminally of KAT2A for all constructs.

Whole-cell protein lysates were generated using RIPA buffer supplemented with 1 mM DTT, 1×cOmplete protease inhibitor (Roche), 1 mM PMSF, 1 mM sodium butyrate and 0.9% SDS. Subcellular fractions were generated as previously described (Gagnon et al., 2014). For western blotting, 10 µg of denatured total protein was subjected to SDS–PAGE before wet transferring resolved proteins onto nitrocellulose membranes (GE Healthcare). Membranes were blocked with 5% nonfat dried milk powder and then incubated overnight at 4°C in primary antibody. After washing in TBST, membranes were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature then detected using Clarity Western ECL substrate (Bio Rad) and X-ray film. Details of antibodies used are provided in Table S3. Images of uncropped blots are provided in Fig. S6.

Samples for qPCR analysis were prepared from total RNA that was extracted and purified using Trizol and RNeasy Mini Spin Columns (Qiagen), and converted to cDNA using the High-Capacity cDNA reverse transcription kit (Thermo). qPCR reactions were prepared using PowerUp SYBR Green Master Mix (Applied Biosystems) and analyzed using the QuantStudio 7 Flex Real-time PCR system (Applied Biosystems). Target Ct values were normalized to RPL13A and relative fold change was calculated using the 2^−ΔΔCt method. Primer sequences are provided in Table S4.

RNA-seq libraries were prepared from 100 ng of polyA-selected RNA using the NEBNext Ultra II Directional RNA Library Prep Kit (New England Biolabs). All cDNA libraries were multiplexed and loaded onto the Illumina MiSeq platform for single-end 76 bp sequencing. Sequencing depth was over 29 million mappable sequencing reads. Post-run Fastq files were parsed and checked for sequencing quality through FastQC. Primary sequencing reads were aligned to the hg19 human reference genome using BowTie2 (Langmead and Salzberg, 2012). Reads were counted for each exon and summarized at the gene level using featureCounts from the Rsubread Bioconductor package and annotated using Ensembl transcripts release 75. Only protein-coding genes were included in the final read count matrix. The read count matrix was then exported and analyzed in the statistical environment R 3.6.1. Differentially expressed genes were identified using DESeq2 (Love et al., 2014). Post-hoc filtering [log₂ fold change (LFC)>1 and false discovery rate (FDR)<0.05] of the pairwise comparison gene list produced the final list of differentially expressed genes (Table S5). Gene Ontology enrichment (GO) and Gene Set Enrichment Analysis (GSEA) were performed on each set of DEGs from the pairwise comparisons (Ashburner et al., 2000; Subramanian et al., 2005).

Histological staining was performed on fresh frozen OCT-embedded human skin biopsies consensually sourced from the National University Hospital and Singapore General Hospital. All participants gave written, informed consent. The study was done in accordance with the Declaration of Helsinki and was approved by A*STAR Human Biomedical Research Office, Institutional Review Board. Cryosections were fixed in 4% paraformaldehyde (PFA) and immunolabeled using indirect immunofluorescence or in situ hybridized using RNAscope 2.5 HD Assay – RED according to the manufacturer's instructions (ACD). Immunocytochemistry was performed on PFA-fixed cells cultured on Lab-Tek II Chamber Slides (Thermo Fisher Scientific), blocked with 1% BSA and 5% goat serum, and stained with Alexa Fluor 488 phalloidin (Thermo Fisher Scientific) before mounting with ProLong Gold Antifade (Invitrogen). Antibodies used are provided in Table S3.

3D epidermal constructs were generated by seeding 5×10⁵ NTERTs into a six-well Greiner Thincert with 1 µm pore size pre-coated with 100 µg/ml rat-tailed type I collagen. After 3 h of incubation at 37°C, KSFM was added to the outer well. Two days later, all media were changed to DFK-1 media and the Thincert was transferred to air-lifted phase culture in a six-well plate and cultured with modified DFK2 media the following day. After 8 days, the cultures were fixed in 4% PFA overnight and subjected to tissue processing and paraffin wax embedding. Histological examination was performed on sections stained with Hematoxylin and Eosin using an Olympus BX51 brightfield and widefield microscope. The thickness of the epidermis was measured using scaled micrographs in ImageJ.

All procedures involving animals were approved by the A*STAR IACUC (A*STAR Institutional Use and Care of Animals Commitee; approved protocols #171238, 191495 and 201519). Mice were derived from Kat2b^{tm1a(KOMP)Wtsi} targeted ES cells obtained from the KOMP Repository at UC Davis. Kat2b^−/− KO mice were obtained by subsequent breeding of Kat2b^{tm1a(KOMP)Wtsi} mice with Zp3-cre mice. For wounding experiments, 8-to 10-week-old adult Kat2b^−/− mice were anesthetized and the dorsal skin was shaved and cleaned before creating two 1 cm full thickness excisional wounds using curved scissors. Wounded mice were allowed to recover in a warmed cage for 1 h and monitored daily for signs of infection, pain and discomfort. Wounds were imaged every 2 d to assess wound closure kinetics.

We thank Drs Kimberley Mace (School of Biological Sciences) and Matthew Ronshaugen (School of Medical Sciences) at the University of Manchester for providing additional supervision to B.W.W. We thank Dr Birgit Lane and Declan Lunny for kind provision of antibodies, John Lim and the A*STAR Microscopy Platform for assistance in image analyses, Eden Lin Shengxin for technical assistance, and the Genome Institute of Singapore Sequencing Platform for assisting with RNA-sequencing experiments.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260723.reviewer-comments.pdf.