The histone acetyltransferase inhibitor Nir regulates epidermis development

The epidermis is a complex tissue formed from embryonic progenitor cells that undergo coordinated changes in gene expression. In mice, epidermal differentiation begins at around embryonic day (E) 8.5, when the ectoderm as a monolayer starts to express keratins (Krt) 8 and 18, and the transcription factor p63 (Trp63), which determines the epidermal lineage (Koster et al., 2007). Markers of basal cells such as Krt5 and Krt14 are expressed from E9.5 (Byrne et al., 1994). Epidermal stratification is initiated at E12.5, when the orientation of cell division becomes perpendicular to the plane of the epidermis (Lechler and Fuchs, 2005). At postnatal day (P) 0, the epidermis is composed of a pluristratified epithelium containing basal, suprabasal, granular, and cornified layers, and constitutes a functional barrier required for survival. Cornified cells from the outer layer are continuously shed off the skin surface and renewed by progenitor mitotic activity. Basal cells can self-renew by symmetric cell division or enter the stratification program (Mikkola and Millar, 2006). The generation of the suprabasal progeny relies on asymmetric cell division, which is dependent on the proper apicobasal orientation of the mitotic spindle in basal cells. To regulate spindle orientation, two key proteins are recruited to polarized apical cortical sites: Inscuteable (Insc) and its partner Leu-Gly-Asn repeat-enriched protein (Lgn; also known as Gpsm2) (Lechler and Fuchs, 2005; Williams et al., 2011). Insc and Lgn drive the orientation of the spindle by capturing cortical microtubules via nuclear mitotic apparatus protein 1 (Numa1) (Poulson and Lechler, 2010). The orientation of cell division is tightly controlled to generate a normal embryonic epidermis, and spindle misorientation can lead to aberrant tissue organization (Poulson and Lechler, 2010).

p63 belongs to the p53 family of transcription factors and is essential for ectodermal appendage specification, epidermal proliferation and asymmetric cell division (Koster, 2010; Laurikkala et al., 2006; Lechler and Fuchs, 2005; Mills et al., 1999; Truong and Khavari, 2007; Yang et al., 1999). p63-deficient epithelia fail to develop beyond ectodermal stage and remain as a monolayer of non-proliferating cells expressing Krt8 and Krt18 (Shalom-Feuerstein et al., 2011). Two p63 transcript starts exist, which encompass or lack an N-terminal transactivation domain sequence to encode TAp63 or ΔNp63, respectively (Blanpain and Fuchs, 2007), which are further diversified by 3′ splicing (α, β, γ). The main p63 isoform in embryonic epidermis, ΔNp63α (McDade and McCance, 2010), activates basal progenitor replication by repressing the expression of genes known to block cell proliferation, such as p21 (Cdkn1a) (Westfall et al., 2003; Truong et al., 2006), and that of cell cycle regulators such as p16Ink4a and p19Arf (alternative products of the Cdkn2a locus) (Su et al., 2009). ΔNp63α also activates the expression of genes involved in epidermal differentiation, such as Krt14 and Gata3 (Chikh et al., 2007; Koster et al., 2007; Lopardo et al., 2008; Marinari et al., 2009; Romano et al., 2009). In addition, to stimulate basal cell proliferation, p63 antagonizes the action of a transcription factor belonging to the same protein family, namely p53 (Trp53). However, some data suggest that p63 drives stratification via a p53-independent pathway (Truong et al., 2006). p63 controls transcription in association with chromatin remodelers that modulate chromatin decompaction at specific DNA-binding sites, and altered expression of epigenetic modifiers has been implicated in epidermal differentiation (Perdigoto et al., 2014). In particular, deletion of the histone deacetylases (HDACs) Hdac1 and Hdac2 results in dramatic failure of hair follicle specification and epidermal proliferation and stratification, similar to loss of p63 (LeBoeuf et al., 2010).

Nir (Noc2l) was discovered as an inhibitor of histone acetyltransferases (INHATs) (Hublitz et al., 2005). INHATs were first described as part of a multiprotein complex capable of inhibiting p300/CBP and PCAF effects on histones (Seo et al., 2001). Nir preferentially associates with hypoacetylated histones and binding is inhibited by acetylation (Hublitz et al., 2005; Schneider et al., 2004; Kutney et al., 2004). In addition to its function in histone acetylation, Nir decreases p53 acetylation in vitro (Hublitz et al., 2005). When acetylated, p53 causes cell growth arrest and apoptosis (Wu et al., 2011). Nir directly interacts with p53 to regulate p53 target genes (Heyne et al., 2010), thereby contributing to the regulation of cell cycle. Previous studies showed that HDACs (e.g. Hdac1) contribute to the removal of acetyl groups from C-terminal lysine residues of p53 (Higashitsuji et al., 2007), in particular in epidermis (LeBoeuf et al., 2010). However, in contrast to the previously identified INHATs, Nir is not part of the HDAC complexes (Hublitz et al., 2005) and contributes to p53 deacetylation via an HDAC-independent mechanism. In addition to its role in p53 post-translational modification, Nir was shown to inhibit TAp63-induced transactivation in HaCaT keratinocytes (Heyne et al., 2010). More recently, Nir was reported to be involved in embryogenesis and especially during early lymphocyte development (Ma et al., 2014) in mice.

To investigate Nir function in vivo, we selectively deleted Nir in epidermis during mouse embryo development. Nir ablation results in perinatal lethality caused by an absence of pluristratified epidermis. In Nir-deficient embryonic epidermis, expression of the key epidermal transcription factor p63 is barely detectable. Nir is recruited to the p63 promoter where it limits acetylation of lysine 18 of histone H3 (H3K18ac). In addition, the epidermis of Nir-deficient fetuses shows p53 hyperacetylation and increased apoptosis, providing evidence that Nir prevents epidermal cells from entering apoptosis by limiting p53 acetylation. Interestingly, we found that Nir is required for initiation of asymmetric division to generate a pluristratified epithelium. Together, our data show that Nir is indispensable for epidermis development via a dual mechanism involving p53 and p63 functions.

Previous studies showed that Nir is ubiquitously expressed throughout mouse embryonic development and at adulthood (Hublitz et al., 2005). To investigate the function of Nir in skin development in vivo, we first analyzed its expression pattern during murine fetal development. Immunofluorescent detection of Nir at different embryonic stages revealed that the protein is expressed from E12.5 to P0 in all epidermal cells (Fig. 1A).

To delineate whether it is required for epidermis development, we selectively ablated Nir in mouse epidermis before hair follicle development and epidermis stratification (Liu et al., 2007) by crossing mice harboring conditional Nir alleles (Nir^fl/fl) with the Krt14-Cre deleter strain (Hafner et al., 2004) to produce Nir^fl/fl; Krt14-Cre mice, hereafter referred to as Nir^cKO mice (Fig. S1A). Immunostaining of Nir^cKO mouse skin revealed the absence of Nir protein in surface epithelia by E12.5 (Fig. 1A). Whereas Nir^cKO/+ heterozygous mice displayed no gross abnormalities (Fig. S1B), Nir^cKO mice died at birth and displayed multiple and dramatic epidermal defects (Fig. 1B,C, Fig. S1C). Skin of Nir^cKO embryos was thin and smooth, and eyelid fusion did not occur (Fig. 1B). Histological analysis by Hematoxylin and Eosin (H&E) staining showed that, instead of stratifying, epidermis remained as a single layer throughout embryogenesis and lacked hair follicle development (Fig. 1C, Fig. S1C). Of note, epidermis partially detached between E17.5 and P0 and was almost absent at birth (Fig. 1C).

To determine the molecular underpinnings of these abnormalities of Nir^cKO embryos, we first analyzed the expression of epidermal stratification markers. The basal cell marker Krt14 was expressed similarly in the epidermis of control and Nir^cKO mice at E12.5 (Fig. S1D). At later stages, Krt14 expression was restricted to basal cells in control embryos and was expressed at low levels in Nir^cKO embryos (Fig. S1D). In addition, whereas control mice initiated stratification at E14.5, in the absence of Nir the suprabasal layer marker Krt10 colocalized with the basal cell marker Krt5 at E14.5 and E16.5, indicating that Nir is required for proper stratification (Fig. 1D). In accordance, cornified layers were not detectable in the epidermis of Nir^cKO mice as shown by H&E staining and immunodetection of loricrin (Lor), a marker for the granular layer (Fig. 1E).

Similar to epidermal stratification, initiation of hair follicle development requires a global switch in the ectodermal differentiation program. H&E staining revealed an absence of hair follicles in Nir-depleted embryos (Fig. 1C, Fig. S1C). In accordance, immunostaining for β-catenin, an early marker for hair follicle placode initiation (Zhang et al., 2009), showed intense fluorescence in control hair follicle placodes at E14.5, whereas the epidermis of Nir^cKO mice displayed uniform, low-level expression of β-catenin (Fig. S1E). Together, our data show that Nir is required for stratification and hair follicle initiation in epidermis.

To identify Nir-regulated gene expression, we performed global transcriptome analysis by RNA-seq on the epidermis of control and Nir^cKO mice at E15.5, when control mice initiate stratification and mutant mice show a monostratified epithelium, and later on at E17.5. At E15.5, we found 4393 differentially expressed genes, with 2068 upregulated and 2325 downregulated (Figs 2A, Fig. S2A). At E17.5, 5673 genes were differentially expressed, 1914 being upregulated and 3759 downregulated (Fig. 2A, Fig. S2A).

To unravel the gene networks responsible for the observed aberrations, we performed phenotype pathway analysis on the upregulated and downregulated genes. Analysis of the 1411 transcripts upregulated at both E15.5 and E17.5 (Fig. S2A) identified genes related to inflammation, as the term ‘immune system phenotype’ was the most enriched (P=8.35×10⁻¹⁷). A similar analysis on downregulated genes common to E15.5 and E17.5 (Fig. S2A) revealed ‘integument phenotype’ (P=5.9×10⁻¹⁷) as one of the most prominent pathways, including the subterm ‘impaired skin barrier function’ (P=1.06×10⁻¹³).

To corroborate these transcriptome data, we performed staining of the whole embryo with Toluidine Blue. In contrast to control mice, Nir^cKO mice absorbed the dye indicating that the barrier function of skin was indeed lost (Fig. 2B). In addition, the pathway analysis highlighted subterms related to epidermis and hair follicle development and maintenance, including ‘abnormal skin morphology’ (P=1.24×10⁻²⁰) and ‘abnormal skin adnexa morphology’ (P=1.70×10⁻¹³), corroborating our histological observations. The analysis confirmed that transcript levels of the epidermal marker Cdh1 (also known as E-cadherin) and up to 20 different types of keratin were decreased upon loss of Nir (Table S1), including the basal and suprabasal markers Krt14, Krt5, Krt1 and Krt10, as well as Lor and β-catenin (Ctnnb1). Our data also revealed that levels of the ectodermal keratin markers Krt8 and Krt18 were strongly increased in Nir^cKO mice (Table S2), suggesting that Nir-deficient epidermal cells did not shift their transcriptional repertoire from ectoderm to pluristratified epidermis.

To determine whether Nir-deficient epidermal cells conserved the expression of ectodermal markers over time, we analyzed Krt8 expression by immunostaining at different time points of epidermis development. At E10.5, Krt8 was similarly expressed in control and Nir^cKO epidermis (Fig. 2C). However, in contrast to control mice, in which Krt8 was no longer expressed in epidermis at E16.5 and E17.5, Nir^cKO mice showed prominent Krt8 expression from E10.5 to P0 (Fig. 2C). In addition, we evaluated the expression of laminins and integrins, which are elements of the basement membrane. Importantly, the RNA-seq data revealed that the expression of laminin 5 subunits (Lama3, Lamb3, Lamc2) and of the transmembrane hemidesmosomal integrins α6 (Itga6) and β4 (Itgb4) was strongly downregulated (Table S1), providing an explanation for the epidermis detachment observed between E17.5 and P0 (Fig. 2C, arrows). Together, these data show that basal epidermis development is initiated in Nir^cKO embryos, but that epidermal differentiation and stratification are impaired, whereas the expression of ectodermal markers is maintained until P0.

Considering that the ectodermal defects observed in Nir^cKO embryos were reminiscent of the phenotype described in embryos lacking p63 (Mills et al., 1999; Yang et al., 1999), we examined whether the phenotype of Nir^cKO embryos could be due, at least in part, to p63 misregulation. First, we investigated whether the expression of p63 was maintained following deletion of Nir. RNA-seq data showed that all p63 transcript variants were more than 90% downregulated in Nir^cKO mice (Table S3). Immunostaining using an antibody that recognizes all the different p63 isoforms revealed expression of p63 in basal cells and in some suprabasal cells in control embryos from E12.5 to P0 (Fig. 2D). In E12.5 Nir^cKO embryos, no major difference in p63 expression was observed (Fig. 2D). However, at E14.5 and E15.5, p63 expression was strongly decreased in epidermal cells and some were even devoid of p63 (Fig. 2D, Fig. S2B, blue arrows). At E16.5 and E17.5, only a few faintly p63-positive cells remained (Fig. 2D, Fig. S2B, green arrows), and p63 was no longer expressed at E18.5 and P0 (Fig. 2D, Fig. S2B). These findings were confirmed by immunodetection of ΔNp63 isoforms (Fig. 2E). Consistent with the absence of stratification, Gata3, a positive target of p63 in suprabasal cells (Chikh et al., 2007), was barely expressed in the epidermis of E14.5 Nir^cKO mice (Fig. 2F).

To determine the Nir binding profile on the promoter region of p63, we performed chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) in mouse epidermis at P0. A ChIP walk on the promoters of the p63 isoforms TAp63 and ΔNp63 revealed that Nir was recruited at the transcription start site (TSS) of TAp63 (−20 bp to 130 bp) and at the ATG regions of both TAp63 (354 bp to 566 bp) and ΔNp63 (116 bp to 305 bp) (Fig. 2G). To investigate further how Nir modulates p63 expression, we analyzed human keratinocyte HaCaT cells as a tissue culture model. As shown in Fig. S2C,D, siRNA-mediated NIR knockdown impaired the expression of ΔNP63, KRT5 and CDH1, corroborating our in vivo data. Moreover, ChIP-qPCR analysis revealed that NIR was enriched at both TAP63 and ΔNP63 promoters in HaCaT cells (Fig. 2H). In particular, NIR was recruited at a region ranging from −884 bp to −665 bp but not at the TSS of TAP63, and at the TSS of ΔNP63 (Fig. 2H), indicating that Nir binds to TAp63 and ΔNp63 promoters in mouse and human, with different binding regions in the two species. Of note, the enrichment at the p63 promoter was not observed in cells treated with siRNA directed against Nir, nor when IgG was used as a control (Fig. 2G,H). Nir was also not recruited at the promoter region of genes unrelated to epidermis development, such as myogenin (MYOG) (Fig. S2E).

Since Nir has been reported (Hublitz et al., 2005) to inhibit histone acetyltransferases such as p300 (Ep300), we investigated the role of Nir in histone acetylation. Immunofluorescence staining indicated that global acetylation levels of histone H3 were not enhanced in Nir^cKO mice at E16.5 compared with control littermates (Fig. S2F). Since acetylation of H3K18 is involved in the regulation of P63 in human (Stefanowicz et al., 2015), we questioned whether the levels of H3K18ac on the P63 promoter are affected by the loss of NIR in human keratinocytes at the −884 bp to −665 bp promoter region. Importantly, acetylation of H3K18 was increased in NIR-depleted cells at the P63 promoter (Fig. 2H). No alteration in H3K18 acetylation levels was observed at the TSS of P63 or at an unrelated promoter (Fig. 2H, Fig. S2E). Taken together, our data indicate that Nir is recruited at the p63 promoter and is required for limiting the acetylation of H3K18.

Among the 1752 genes that were downregulated at E17.5 but not at E15.5 (Fig. S2A), pathway analysis revealed the term ‘cellular phenotype’ (P=5.24×10⁻¹⁴), in which the subterms ‘abnormal cell death’ (P=8.35×10⁻⁸) and ‘abnormal cell cycle’ (P=2.53×10⁻²⁰) were found. Since Nir decreases p53 acetylation (Ac-p53) in vitro (Hublitz et al., 2005) to promote p53-dependent apoptosis and cell growth arrest (Wu et al., 2011), we next examined whether loss of Nir in the epidermis would affect Ac-p53 levels. At E12.5, p53 was acetylated at the C-terminal lysine in both control and Nir^cKO mice. At E14.5 and E16.5, whereas levels of Ac-p53 strongly decreased during the development of stratified epidermis in control mice, they remained remarkably high in Nir^cKO embryos as revealed by immunofluorescence (Fig. 3A). In addition, at E12.5, we observed a nuclear signal in the dermis of both control and Nir^cKO mice that was lost at later stages, when it resembled the background (Fig. S3A), raising the question of a potential cross-talk between epidermis and dermis during stratification.

An increase in Ac-p53 may be attributed to either reduced function of INHATs such as Nir or of HDACs. Immunostainings of Nir^cKO mice at E14.5 showed that Hdac1 protein levels were dramatically decreased (Fig. 3B), which may contribute to p53 hyperacetylation. Hdac2 levels were unchanged (Fig. 3C). Taken together, our data show that the phenotype of Nir^cKO embryos is mediated, at least in part, by hyperacetylation of p53 in embryonic epidermis.

To determine whether programmed cell death contributes to the phenotype of Nir mutant embryos, we performed a TUNEL assay (Fig. 3D). At E16.5, an ∼45-fold increase in the rate of basal cell apoptosis was observed in Nir^cKO compared with control epidermis (Fig. 3D,E), indicating that the defects observed in Nir^cKO mutant skin were, at least in part, due to increased apoptosis.

To identify whether cell division was altered by Nir deletion in skin, we analyzed the percentage of basal cells positive for the mitosis-specific modification H3S10ph (Teperek-Tkacz et al., 2010) in mutant and control epidermis. At E16.5, mitotic cells were decreased in Nir^cKO mice to ∼68% of the levels observed in controls (Fig. 3F,G). We corroborated these data by staining for the proliferation marker Ki67, which showed a 77% decrease (Fig. S3B,C). In addition, cell division evaluated from DAPI staining decreased by 82% at E12.5, 75% at E14.5, and 78% at E16.5 in basal cells of Nir^cKO mutant mice (Fig. S3D). These data indicate that loss of Nir in epidermis impairs cell proliferation and promotes apoptosis.

The proliferation defects could be due to impaired self-renewal of progenitor cells. Since p63 is required for proper cell division orientation (Sen et al., 2010; Senoo et al., 2007), we investigated the expression levels of the genes involved in this process. The Par complex, which comprises Par3 (Pard3), Par6 (Pard6a, Pard6b and Pard6g) and atypical protein kinase c (Prkcz and Prkci), determines epithelial cell polarity. We found that expression of Prkci, Pard3, Pard6b and Pard6g was decreased by 50% in Nir^cKO epidermis (Table S4). Gα_i (Gnai1, 2, 3), Pins [Gpsm1 and Lgn (Gpsm2)], Numa1 (Williams et al., 2011; Zhu et al., 2011) and Dctn1 regulate orientation of the mitotic spindle. Insc links these factors to the Par complex by binding to both Par3 and Pins. Most of these spindle orientation determinants were unaltered in Nir mutant mice (Table S4).

Expression of Numa1, which is localized at the spindle apparatus during cell division, was unaffected in Nir^cKO epidermis (Table S4). We therefore used Numa1 as a marker to determine the percentage of symmetric (SCD) versus asymmetric (ACD) cell division (Fig. 4A, Fig. S4A). In epidermis of control mice, the ACD rate strongly increased from E12.5, when stratification starts, to E16.5, reflecting the generation of suprabasal cell layers. By contrast, Nir^cKO mice failed to initiate proper ACD and most of the division remained symmetric at E16.5 (Fig. 4B). A detailed analysis of the spindle orientation of mitotic cells corroborated these findings. Indeed, at E14.5, control cell divisions occurred essentially on angles of 0°-10°, which corresponds to SCD, and on 60°-90°, reflecting ACD (Fig. S4B). By contrast, in Nir^cKO mice, most of the divisions were parallel to the basal lamina at angles of 0°-30°, and perpendicular divisions were nearly absent. In agreement, at E16.5, cell division of Nir-depleted epidermis was essentially symmetric, whereas 70% of control basal cells divided in an asymmetric manner (Fig. 4C).

In addition, whereas Lgn was localized at the mitotic poles in E16.5 control fetuses, it was more diffuse in Nir^cKO epidermal cells (Fig. S4C). Of note, our RNA-seq data revealed a reduction of Cdh1 to 0.24-fold at E15.5 and to 0.12-fold at E17.5 in Nir^cKO mice compared with Ctrl mice (Table S1). At E14.5 in control basal cells, Cdh1 staining was concentrated at the apical pole and absent at the basal pole (Fig. 4D, blue arrows). By contrast, in Nir^cKO embryos Cdh1 appeared concentrated in the basal area (Fig. 4D, white arrows). Since Cdh1 has been shown to be required for cell division orientation (Gloerich et al., 2017; Maretzky et al., 2005; Rubsam et al., 2017), the spindle misorientation phenotype observed in Nir^cKO epidermis could at least in part be due to disturbed localization and function of Cdh1. Together, these data show that the stratification defects in Nir^cKO mice are due to their incapacity to enter ACD.

In conclusion, our data support a model in which Nir regulates epithelial stratification and cell death by control of p63 expression and restriction of p53 acetylation, respectively (Fig. 4E). Thus, Nir is a key factor for establishing the epidermis.

In this study, we investigated the function of Nir during epidermis development in vivo. Previous data reported that Nir knockout and overexpression both led to cell death (Heyne et al., 2010). Here we show that ablation of Nir in the forming epidermis causes perinatal death due to skin defects that destroy its barrier function. We found that Nir is required for differentiation of basal epidermal cells, since Nir depletion prevents the formation of suprabasal layers and hair follicle placodes. Our data indicate that Nir controls the expression of essential epidermal components including keratins (e.g. Krt14, Krt5, Krt10), loricrin, laminins, and of β-catenin, which is essential for the initiation of hair placodes. In addition, Nir is required to restrict ectodermal development at later stages of embryogenesis, as demonstrated by the repression of ectodermal markers Krt8 and Krt18. Nir directly targets p63, a master regulator of epidermis development, and embryos lacking Nir progressively lose p63 expression from E14.5 onwards. Interestingly, loss of Nir induces a phenotype that is reminiscent of that of p63-null mice (Shalom-Feuerstein et al., 2011), including decreased proliferation, absence of epidermal stratification and hair follicle development, faults in eyelid fusion, and ectopic Krt8 expression (Laurikkala et al., 2006; Lechler and Fuchs, 2005; Mills et al., 1999; Yang et al., 1999). Similar to what was observed in skin devoid of p63, Nir-deficient epidermis showed strong defects in cell proliferation and differentiation. It was previously shown that Nir negatively regulates TAp63 activity (Heyne et al., 2010), binding to its transactivation and C-terminal oligomerization domains, which inhibits TAp63-dependent p21 transactivation. Conversely, in the presence of TAp63, siRNA-mediated ablation of Nir enhanced p21 transcription. Despite dramatically reduced expression of the various p63 isoforms, expression of the p63 target gene p21 was unaffected, suggesting other compensatory mechanisms that maintain p21 levels.

Another important result of this study is that Nir is required for proper ACD. Indeed, in Nir^cKO embryos, most of the divisions occur symmetrically. This could be due to the absence of essential components of the lamina such as laminin 5 and integrins, or to reduced expression of Par3 and Par6 determinants of cell division polarity. The lack of ACD in Nir-depleted epidermis differs from what was reported for p63-null embryos, which display random division orientation (Poulson and Lechler, 2010). Therefore, loss of p63 contributes to the phenotype of Nir^cKO embryos by regulating the expression of essential components of the stratification process, but is not the only factor responsible for ACD. We also considered the possibility that Nir deficiency acts by disturbing Cdh1 functions. Cdh1 is a cell-cell adhesion protein that promotes adhesion to cytoskeletal proteins by formation of adherence junctions at the plasma membrane, with higher concentrations in the zona adherence in the vicinity of the apical side (Stemmler, 2008). Recent evidence showed that Cdh1 is also a master regulator of tissue polarity in stratifying epithelia (Rubsam et al., 2017; Stemmler, 2008) and plays a prominent role in establishing cell division orientation (Gloerich et al., 2017; Lázaro-Diéguez and Müsch, 2017; Rubsam et al., 2017). Our transcriptomic and histological data revealed a strong reduction of Cdh1 in Nir^cKO embryos, which, combined with mislocalization of the protein, could contribute to misallocation of the spindle apparatus.

Our study also revealed that the absence of Nir in epidermal cells blocks proliferation and promotes programmed cell death. In a previous study, we identified the tumor suppressor p53 as a Nir interaction partner (Hublitz et al., 2005). Upon recruitment by p53, Nir represses transcription of p53 target genes, and p53-dependent apoptosis is robustly increased upon Nir depletion (Hublitz et al., 2005). Several studies have reported that Nir represses p53 activity by decreasing its acetylation levels in vitro (Heyne et al., 2014; Wu et al., 2012). We showed that in vivo, loss of Nir leads to an increase in the levels of acetylated p53, providing a potential explanation for impaired cell proliferation and increased apoptosis. Post-translational modifications of p53 are also altered by acetyltransferases and deacetylases including HDACs (Higashitsuji et al., 2007). We found that Hdac1, but not Hdac2, levels were strongly decreased upon Nir loss. Interestingly, embryos lacking Hdac1/2 specifically in epidermis show a phenotype that resembles that of Nir^cKO mice (LeBoeuf et al., 2010). Hdac1/2 epidermal mutants display multiple severe ectodermal defects, including aberrant suprabasal epidermal differentiation and initiation of hair follicle development, but also defects in proliferation and apoptosis. However, contrary to what we observed, LeBoeuf et al. (2010) showed an increase in several negatively regulated p63 target genes such as p21. The differences between the phenotype of Nir-deficient and Hdac1/2-deficient mice could be explained by the fact that Hdac2 levels are unaffected in our model.

We found that Nir limits H3K18 acetylation. This is interesting in view of recent studies carried out in asthmatic patients (Stefanowicz et al., 2015). Indeed, such subjects show elevated H3K18 acetylation in airway epithelial cells, which could not be rescued by HDAC inhibitors (Stefanowicz et al., 2015). We could therefore postulate that a depletion of Nir in these patients might be the cause of their disease. However, from a search of public databases we could not identify any human skin pathology in which NIR expression was altered. In addition, no mutation of NIR was reported in the Human Gene Mutation Database (www.hgmd.cf.ac.uk). Further investigations are required to determine whether NIR is linked to epithelial diseases in humans.

In recent years evidence has accumulated that histone acetylation can be present at some genes together with putative repressive histone marks (De Gobbi et al., 2011; Zhang et al., 2015). For instance, it was recently shown that H3K4 and H3K27 trimethylation and H3K27ac contribute to the transcriptional repression of Slc47a2 in renal cell carcinoma (Yu et al., 2017). In addition, Stefanowicz et al. (2015) found increased H3K18ac associated with enhanced H3K9 trimethylation (Stefanowicz et al., 2015). Since Nir was shown not to be present with HDACs, it is highly possible that Nir is part of an activating complex, where it prevents the acetylation of one of the key transcriptional activation factors, thereby promoting gene transcription, and deacetylation of H3K18 might only be a side effect. This was previously observed for the INHAT SET/TAF-Iβ, which inhibits Foxo1 acetylation and activates its transcriptional activity toward p21 (Chae et al., 2014). Similarly, hyperacetylation of Foxp3, which is regulated by the histone acetyltransferase p300 and the HDAC SIRT1, increases Foxp3 protein stability (van Loosdregt et al., 2010). We postulate that the effect of Nir depends on local networks of transcription factors in combination with histone marks and their regulation by acetylation. This interesting aspect remains to be elucidated.

In summary, our data reveal essential roles for Nir in proliferation, stratification and cell fate decisions in the embryonic epidermis. Our work also indicates that this phenotype is due, at least in part, to the INHAT function of Nir, which limits H3K18ac on the p63 promoter to control p63 expression and prevents p53 hyperacetylation. Together, our results uncover the essential function of Nir in the control of epidermis development.

Mice were housed in a pathogen-free barrier facility of the University Medical Center Freiburg in accordance with institutional guidelines and approved by the regional board. Mice were maintained in a temperature- and humidity-controlled animal facility with a 12 h light/dark cycle, free access to water, and a standard rodent chow (Kliba, breeding, 3807). Animals were sacrificed by cervical dislocation. Tissues were immediately collected, frozen in liquid nitrogen or processed for further analyses.

We generated a floxed conditional Nir allele by inserting a selectable PGK neomycin resistance (Neo) marker downstream of exon 4 and two loxP sites flanking Nir exons 5 to 12 (Fig. S1A). Two FRT sequences were added to each end of the PGK Neo cassette to facilitate removal by FLP recombinase. Embryonic stem cells in which Frt sites were recombined by Flp were injected into blastocysts to generate mice heterozygous for the floxed Nir allele (Nir^+/fl). Nir^+/fl mice were mated with Krt14-Cre mice (Hafner et al., 2004) to selectively ablate Nir in the epidermis (Nir^cKO mice, Fig. S1A). Homozygous conditional mice (Nir^fl/fl) were used as controls. Mice were genotyped with primers for detection of conditional Nir alleles and Cre recombinase (Table S5).

Epidermis was prepared from dorsal skin as described (Surjit et al., 2011). Briefly, the inner side of dorsal skin was incubated in a 0.8% trypsin solution for 30 min at 37°C. Epidermal sheets were recovered in cell culture medium [EMEM (Lonza, BE12-125F) without calcium, supplemented with 10% FCS], incubated at 37°C for 15 min, and filtered through a 70 μm nylon cell strainer (BD Falcon). Epidermal cells were then processed for RNA extraction and ChIP experiments.

RNA was isolated with TRIzol Reagent (Invitrogen) from isolated mouse dorsal epidermis at E15.5 and E17.5 and processed as described (Duteil et al., 2014). RNA with a RIN above 8.0 was sequenced at the DKFZ core facility (Heidelberg, Germany) using the standard Illumina protocol. Paired-end reads were mapped to Ensemble annotation NCBI m38/mm10 with TopHat version 2 (Trapnell et al., 2012) using default parameters. The aligned reads were counted using Homer software (analyze RNA) and differentially expressed genes were identified using EdgeR (Robinson et al., 2010). Differentially regulated genes (reads >50, P<0.001, fold change >1.3 or <0.77) were further used for pathway analysis in WebGestalt (Heinz et al., 2010; Wang et al., 2013).

Histological analyses were performed on dorsal skin. Embryos were fixed in 10% buffered formalin and embedded in paraffin. 5 µm paraffin sections were deparaffinized and rehydrated. H&E staining and immunofluorescence analyses were performed as described (Duteil et al., 2014), using antibodies directed against Nir [R.S. laboratory (Hublitz et al., 2005); rabbit polyconal, 1/500], Krt14 (Covance, PRB-155P; 1/500), Krt5 (DAKO, M7237; 1/2000), Krt10 (Covance, PRB-159P; 1/500), loricrin (Covance, PRB-145P; 1/500), Krt8 (R. Kemler laboratory, MPI, Freiburg, Germany; 1/200), β-catenin (BD Transduction Laboratories, 610154; 1/100), β-tubulin (Sigma, T6074; 1/1000), H3S10-phospho (Upstate, 06-570; 1/1000), Ki67 (Novus Biologicals, NB110-89717; 1/200), Numa1 (Abcam, ab97585; 1/50), p63 (DAKO, M7247; 1/50), ΔNp63 (S. Sinha laboratory, Jacobs School of Medicine & Biomedical Sciences, Buffalo, NY, USA; 1/50), Gata3 (Proteintech Group, 10417-1-ap; 1/100), Hdac1 (Abcam, ab7028-50; 1/200), Hdac2 (Santa Cruz, sc-7899; 1/200), acetylated histone H3 (Upstate, 06-599; 1/200), acetylated p53 (Abcam, ab52172; 1/100), Lgn (Abcam, ab84571; 1/100) and Cdh1 (E-cadherin, BD Transduction Laboratories, 610182; 1/50). Slides were then incubated with anti-mouse or anti-rabbit secondary antibody coupled to Alexa546 (Invitrogen A11010 or A11030; 1/400) or Alexa488 (Invitrogen A11029 or A11034; 1/400) and mounted in aqueous medium (Fluoromount-G, SouthernBiotech, 0100-01) with DAPI (Sigma, D-9542, 1/1000).

The TUNEL assay was performed using the In situ Cell Death Detection Kit (Roche, 11684795910) according to the manufacturer's recommendations.

Whole-embryo staining with Toluidine Blue was performed as previously described (Hardman et al., 1998).

Human HaCaT keratinocytes were cultured in DMEM (Lonza, BE12-125F) supplemented with L-glutamine (2 mM) and 10% FCS. Cells were transfected with 1 mM siRNA against Nir, or unrelated control (Invitrogen), using Lipofectamine RNAimax (Invitrogen) according to the manufacturer's instructions. siRNA oligonucleotide sequences (5′-3′) were: Nir siRNA, CUGGAAGACCUGAACUUCCCUGAGA; control siRNA, CUGCAGAAGUCUUCAUCCCGAGAGA. Cells were either harvested and snap frozen for mRNA and protein analysis, or fixed for 20 min at 4°C with 1% paraformaldehyde for ChIP experiments.

ChIP experiments were performed using anti-Nir [R.S. laboratory (Hublitz et al., 2005)] and anti-H3K18ac (Abcam, ab1191) antibodies, or a rabbit IgG negative control on protein G-Sepharose 4B (GE Healthcare) beads as described (Metzger et al., 2008). ChIPed DNA was used for qPCR analyses with the primers described in Table S6.

Western blot analysis, co-immunoprecipitation assays, gel filtration, and mass spectrometry experiments were performed as described (Duteil et al., 2014). Western blot membranes were probed using primary antibodies against Nir [R.S. laboratory (Hublitz et al., 2005); rabbit polyclonal, 1/1000], Krt5 (DAKO, M7237; 1/2000), Cdh1 (BD Transduction Laboratories, 610182; 1/50) and β-actin (Sigma, A1978; 1/10,000). Secondary antibodies conjugated to horseradish peroxidase (GE Healthcare) were detected using an enhanced chemiluminescence detection system (GE Healthcare).

Spindle orientation was determined by measuring the angle between the centrosomal axis and the basement membrane in late prophase, metaphase, and anaphase cells, when two centrosomes were observed at opposite sides of the cell (in early prophase the centrosomal pair is localized apically). Stages of mitosis were defined as follows: early prophase cells had condensed chromatin lacking a clearly defined pair of centrosomes; late prophase cells had a pair of centrosomes positioned at opposing poles; metaphase cells resembled late prophase cells but displayed aligned chromosomes characteristic of the metaphase plate. All cells were positive for phospho-histone H3 and Numa1.

Data were analyzed and statistics performed in Prism 6 (GraphPad). For determination of axis of cell division, the number of cells analyzed (n) is indicated in the radial histograms, and included cells from three or more embryos of the same age. Radial histograms of angle of division were plotted in Microsoft Excel from raw data binned in 10° increments. All other graphs were prepared in Prism. Data are represented as mean+s.e.m. Significance was calculated by two-tailed Student's t-test for Fig. 3E,G, Fig. 4B, Fig. S2C and Fig. S3C,D, by one-way ANOVA for Fig. 2G and by two-way ANOVA for Fig. 2H and Fig. S2E.

We are obliged to F. Pfefferle, L. Walz and J. M. Bornert for providing excellent technical assistance and to G. Hua for help with performing ChIP experiments in keratinocytes. ΔNp63 antibody was a gift from the laboratory of Dr Sinha. Krt8 antibody was provided by the laboratory of Dr Kemler. HaCaT cells were provided by Prof. Dr Walter Wahli. The pGL-promoter vector was provided by Dr D. Metzger.