Keratinocytes respond to environmental signals by eliciting induction of genes that preserve skin’s integrity. Here we show that the transcriptional response to stress signaling is supported by short-lived epigenetic changes. Comparison of chromatin accessibility and transcriptional changes induced by barrier disruption or by loss of the nucleosome remodeler Mi-2β identified their striking convergence in mouse and human keratinocytes. Mi-2β directly repressed genes induced by barrier disruption by restricting AP1-enriched promoter-distal sites, occupied by Mi-2β and JUNB at steady state and by c-JUN after Mi-2β depletion or stress signaling. Barrier disruption led to a modest reduction in Mi-2β expression and a further selective reduction of Mi-2β localization at stress response genes, possibly through competition with activated c-JUN. Consistent with a repressive role at stress response genes, genetic ablation of Mi-2β did not prevent reestablishment of barrier integrity but was required for return to homeostasis. Thus, a competition between Mi-2β–repressive and activating AP1 complexes may permit rapid transcriptional response to and resolution from stress signaling.

Skin represents the largest interface of a living organism with the outer world and provides both a physical and immunological barrier against incoming threats, including mechanical injury, chemical insults, invading pathogens, and UV exposure (Segre, 2006). These environmental cues elicit stress signaling in keratinocytes that causes their activation and alters interactions with neighboring cells, including those of the immune system (Dainichi et al., 2018; Nestle et al., 2009; Pittelkow, 2005). Swift recovery from a stress response is also important for return to skin homeostasis, with sustained stress signaling contributing to pathogenic conditions that are often characterized by excess inflammation and epidermal barrier dysfunction (Dainichi et al., 2018; Nestle et al., 2009; Pittelkow, 2005).

During keratinocyte differentiation, a large number of genes are in a poised chromatin state with some pre-established interactions reported between promoters and enhancers (Rubin et al., 2017). In addition to transcriptional activators, chromatin modifiers play an important role in providing a poised chromatin landscape. Mi-2β is an ATP-dependent nucleosome remodeler that serves as a core component of the nucleosome remodeling and deacetylase (NuRD) complex (Tong et al., 1998; Xue et al., 1998; Zhang et al., 1998). Two types of enzymes in the NuRD complex, the histone deacetylases (HDACs) HDAC1 and HDAC2 and the nucleosome remodeler Mi-2β, work together to modulate transcription. By altering nucleosome position relative to transcription factor (TF) binding sites (Becker, 2002; Brehm et al., 2000), Mi-2β can either promote or suppress gene expression depending on local context (Williams et al., 2004; Yoshida et al., 2008; Zhang et al., 2011). Mi-2β is expressed in many progenitor cell types, including embryonic stem cells, hematopoietic cells and neuronal, epithelial, and heart muscle precursors where it is required to keep genes supporting alternate cell fates in a poised but repressed state and prevent precocious differentiation (Gómez-Del Arco et al., 2016; Kim et al., 1999; Reynolds et al., 2012; Yamada et al., 2014; Yoshida et al., 2008; Yoshida et al., 2019; Zhang et al., 2011).

During embryonic and fetal development of the epidermis, NuRD complex components regulate cell fate choices by engaging in stage-specific transcriptional networks that control epidermal differentiation (Kashiwagi et al., 2007; LeBoeuf et al., 2010). HDAC1/2 are required for the differentiation of basal epidermal cells into suprabasal and follicular cell fates by providing repressive function to p63, an early epidermal differentiation factor, and for keratinocyte proliferation by inhibiting p53 activity through deacetylation (LeBoeuf et al., 2010; Luo et al., 2000). Mi-2β is required for commitment of the embryonic ectoderm to the epidermal cell lineage as well for self-renewal of the newly generated epidermal precursors; however, it is not required for their stratification (Kashiwagi et al., 2007). After epidermal lineage commitment, loss of Mi-2β interferes with follicular differentiation but not with differentiation of interfollicular epidermis (Kashiwagi et al., 2007). In the adult skin, Mi-2β represses genes supporting basal keratinocyte activation, mobilization, and immune cell function (Kashiwagi et al., 2017). Many of the same genes are also induced by physiological challenge to the skin, such as wounding or barrier disruption (Keyes et al., 2016; Lander et al., 2017; Sextius et al., 2010), suggesting that suppression of Mi-2β activity may be a mechanism by which environmental challenges mobilize the skin’s stress responses.

Activator protein 1 (AP1) is a family of early response transcriptional complexes comprised of homo- and hetero-dimers between c-JUN and c-FOS family members. AP1 complexes bind to a palindromic 12-O-tetradecanoylphorbol-13-acetate response element and regulate gene expression in response to mitogenic, stress, differentiation, and apoptotic signals (Guinea-Viniegra et al., 2009; Mehic et al., 2005; Risse et al., 1989; Shaulian and Karin, 2002; Wurm et al., 2015). In the basal layer of the skin, the polycomb repressor group complex prevents AP1 from binding and activating terminal differentiation genes expressed in the suprabasal layers. Both the differential expression of AP1 constituents and the suppression of polycomb repressor group complex expression allow for activation of AP1-dependent differentiation and acquisition of epidermal barrier function (Ezhkova et al., 2009).

Here, we show that either barrier disruption or genetic ablation of Mi-2β in the mouse epidermis elicits a transcriptional response that involves similar genes and functional pathways. We show that these distinct physiological and genetic challenges cause a similar increase in chromatin accessibility at enhancers associated with stress response genes that frequently encompass binding sites for AP1. We demonstrate that Mi-2β repressed transcription and activation of AP1 complex components, such as c-JUN, and restricted chromatin access at AP1 binding sites. Increasing activation and occupancy of c-JUN with concomitant loss of Mi-2β at these enhancer sites, which were occupied by JUNB at steady state, correlated with induction of gene expression. A moderate reduction in Mi-2β expression was induced by physiological signals such as barrier disruption that was accompanied by a further albeit selective abrogation of Mi-2β at stress response genes. Permanent loss of Mi-2β caused aberrant induction of AP1, interfered with return to skin homeostasis, and caused severe pathologies. Taken together, our studies support a model by which functional interactions between the chromatin remodeler Mi-2β and the early response complex AP1 account for rapid response and recovery from mechanical injury.

Rapid but transient transcriptional responses induced by stress signaling in the epidermis

To examine the molecular mechanisms that control the skin’s response to physical insult, we evaluated the effects of acute stress signaling using a skin barrier disruption assay. Transepidermal water loss (TEWL) was measured before and at different time points after removal of the stratum corneum by repetitive tape stripping (TS; Fig. 1, A and B). Immediately after TS, TEWL was significantly increased, but starting at 1 h after treatment, it began to progressively decrease to reach preinjury levels by 48 h (Fig. 1 B). To evaluate gene expression changes during this acute response, basal keratinocytes (ITGA6+, CD45, CD34) were sorted before and at different time points after TS and processed for RNA sequencing (RNA-seq; Fig. 1 A and Fig. S1 A; before and at 6, 24, and 96 h after TS). A rapid induction of mRNA changes and subsequent recovery was seen within 96 h after treatment, consistent with the phenotypic changes in TEWL (Fig. 1, C and B; and Fig. S1 B). Notably, most of the transcriptional changes were observed by 6 h after TS, with 1,444 up-regulated and 1,415 down-regulated genes detected (Fig. 1 C and Fig. S1 B). Of the genes induced at 6 h, 70% were no longer elevated at 24 h, and 99% were expressed at preinjury levels by 96 h (Fig. 1 C and Fig. S1 B). A similar rapid response and recovery was seen with down-regulated genes (Fig. 1 C and Fig. S1 B). Gene set enrichment analysis (GSEA) showed induction of different functional pathways early versus later in the response (Fig. 1 D). Genes activated by Myc, the unfolded protein response, and TNF signaling peaked at 6 h, whereas genes involved in MTORC1 signaling, oxidative phosphorylation, and the regulation of the G2M checkpoint peaked at 24 h of the response (Fig. 1 D). Further gene ontology (GO) analysis of all deregulated genes at 6 and 24 h after barrier disruption confirmed participation in cell proliferation, keratinocyte differentiation, and immune cell activation (Fig. S1 C).

Further analysis of differentially expressed genes by K-means clustering revealed induction and suppression of distinct genes and pathways at different time points after barrier disruption (Fig. 1 E and Fig. S1 D). Of the four groups of temporally dysregulated genes identified, genes whose induction peaked as early as 6 h fell within the first two clusters (C1 and C2) and encoded factors required for lipid biosynthesis and keratinocyte terminal differentiation, such as Abca12, Alox12b, Aqp3, Cldn4, and Tgm1; inflammatory mediators, such as Cxcl1, Il24, S100a8, and S100a9; and cell proliferation–promoting signals, such as members of the EGF family, Ereg, Areg, Tgfa, Hbegf, and Vegfa (Fig. 1, E and F, 6 h C1 and C2; Fig. S1 D; and Table S1). Genes induced at a later time point fell into the third cluster (C3) and contained many cell cycle genes, such as Mki67, Ccna2, Ccnb1, Cdk1, and Cdkn3 (Fig. 1 E, 6 h; Fig. S1 D; and Table S1). Notably, genes in this cluster (C3) showed an initial decrease in expression at 6 h, but by 24 h were increased to a higher level relative to steady state, which declined by 96 h. Finally, genes down-regulated early in the response were contained within the fourth cluster (C4) and were associated with circadian rhythm regulation, such as Per3 and Cry2, and lipid metabolism, such as Acaa2, Echs1, and Acss2 (Fig. 1, E and F; Fig. S1 D; and Table S1). These genes were rapidly down-regulated early in the repair process but became reexpressed during the recovery phase. The decrease in the circadian transcriptional regulators Per3 and Cry2 may reflect the physiological response to barrier disruption. Whereas reduction of these genes, seen early in the response, correlates with an increase in keratinocyte proliferation, the increase toward normality, seen later in the response, correlates with epidermal stem cell differentiation. This is in line with previous studies that have shown that changes in core clock genes, such as Per3 and Cry2, predispose epidermal stem cells to environmental cues that regulate proliferation or differentiation (Janich et al., 2013). The down-regulation of lipid metabolism/catabolism genes follows a requirement of keratinocytes to synthesize and secrete lipids into the intercellular space of the cornified layer for reestablishment of the barrier. These genes are also rhythmically regulated, and their expression may reflect changes in Per3 and Cry2 expression (Gnocchi et al., 2015).

Thus, barrier disruption triggers the rapid induction of genes involved in barrier reestablishment, cell proliferation, immune cell activation, and repression of genes involved in lipid metabolism and circadian rhythm regulation. Such rapid transcriptional responses appear to be critical for repairing the compromised structural barrier and fortifying the local immune cell defenses. The observed prompt recovery from these transcriptional changes is likely critical for preventing excessive barrier formation and inflammation that can lead to skin pathologies.

Similar transcriptional response elicited by stress signaling or Mi-2β depletion

Aberrant keratinocyte activation and cell proliferation demarcated by expression of keratin 6 (K6) and Ki67, respectively, were evident shortly after conditional inactivation of the chromatin remodeler Mi-2β in the epidermis (Fig. S2 A; Kashiwagi et al., 2017). Moreover, a highly significant overlap and similar trend in gene expression changes were seen either by barrier disruption or by acute loss of Mi-2β in the epidermis (Fig. 2, A–C; P < 0.0001 by Fisher’s exact test; and Table S2). Among the significantly up-regulated genes at 6 h after barrier disruption, 36% were also significantly up-regulated by loss of Mi-2β (Fig. 2, A and B; 524/1,444). By 24 h the number of up-regulated genes declined; however, the relative number shared by Mi-2β loss was further increased (Fig. 2, A and B; i.e., 65%: 439/673). Further examination of the shared significantly up-regulated genes at 6 h and in Mi2Δ keratinocytes showed that the majority was still significantly induced at 24 h compared with nonshared genes that were not (Fig. S2 B). These shared genes enriched functional categories such as keratinocyte activation, barrier reestablishment, immune cell activation, and cell growth (Fig. S2 C and Fig. S1 C). Similarly, among the significantly down-regulated genes after 6 h of barrier disruption, 38% (532/1,415) were also significantly down-regulated by loss of Mi-2β, and by 24 h their relative number was further increased (i.e., 54%: 170/315; Fig. 2, A–C).

GSEA comparing significantly up- and down-regulated genes in TS keratinocytes with the transcriptome of Mi-2β–depleted keratinocytes confirmed a similar trend in regulation (Fig. 2 D, false discovery rate [FDR] = 0). Genes up- and down-regulated at both 6 and 24 h strongly correlated with genes up- and down-regulated in Mi-2β–depleted keratinocytes (Fig. 2 D, enrichment score [ES]: 0.752–0.810). The reverse comparison of significantly induced genes upon loss of Mi-2β with the transcriptome of TS keratinocytes showed a weaker correlation (Fig. S2 D, ES: 0.605–0.710), which is consistent with participation of Mi-2β in the repression of genes that are involved in other biological processes in keratinocytes.

Barrier disruption induces rapid changes in chromatin accessibility

We next examined whether the response to stress signaling was associated with changes in chromatin accessibility. Chromatin accessibility was evaluated genome-wide in sorted keratinocytes before and after barrier disruption or after Mi-2β depletion using the Tn5 transposase assay (assay for transposase-accessible chromatin [ATAC] sequencing [ATAC-seq]; Fig. 3 A). ATAC peaks in TS or Mi-2β–depleted keratinocytes were compared with WT steady-state keratinocytes; if induced, they were classified as differentially induced ATAC (dATACi), and if lost as differentially lost ATAC (dATACl). Similar to the gene expression changes, a rapid gain in dATACi peaks was observed at 6 h after TS that declined by 24 h (Fig. 1 and Fig. 3 B; dATACi peaks: 9,178 at 6 h and 6,262 at 24 h). In sharp contrast to the dATACi peaks, an insignificant number of dATACl peaks was seen at 6–24 h (Fig. 3 B; dATACl peaks: 61 at 6 h and 32 at 24 h; and Tables S3 and S4). The gain in dATACi and loss in dATACl peaks in Mi-2β–depleted keratinocytes by far exceeded what was seen by barrier disruption (Fig. 3 B; dATACi peaks [27,068] and dATACl peaks [1,943]; and Table S5). Nonetheless, a larger number of dATACi compared with dATACl peaks was also seen in Mi-2β–depleted keratinocytes (i.e., >10 fold) as in keratinocytes after barrier disruption.

The majority (∼80%) of dATACi peaks observed after Mi-2β depletion or barrier disruption were distributed over intergenic and intronic regions, with only ∼10–15% associated with promoters (Fig. 3 C). Moreover, the magnitude of increase in dATACi enrichment was greater at intergenic and intronic regions compared to promoters or exonic regions (Fig. 3 D). These data indicate that barrier disruption or Mi-2β depletion causes a significant reprogramming of regulatory elements that are distal to promoters and may include transcriptional enhancers. It also shows that the detected increase in chromatin accessibility is not just a consequence of a transcriptional increase that should affect promoters as well. Nonetheless, the number and magnitude of change at dATACi peaks was greater in Mi-2β–depleted keratinocytes compared with keratinocytes after barrier disruption (Fig. 3, B and D).

The potential contribution of TFs to the barrier disruption response was tested by de novo motif analysis of all dATACi peaks obtained at 6 or 24 h after TS (Fig. 3 E). AP1 complex motifs were most significantly associated with induced regulatory sites (Fig. 3, E and F). Other factors associated with dATACi at 6 or 24 h were CEBP and KLF5, albeit at a lower frequency and P value compared with AP1 (Fig. 3 E). Enrichment of dATACi peaks with STAT3, a key regulator of keratinocyte proliferation and migration, was seen at 24 h, consistent with the later increase in epidermal proliferation (Fig. S2 A and Fig. 1 D; Sano et al., 2005). The dATACi peaks in Mi-2β–depleted keratinocytes were also significantly enriched for AP1 motifs (Fig. 3, E and F). In addition to AP1, a highly significant enrichment of NFY, OCT, and KLF5 motifs was seen at dATACi peaks from Mi-2β–depleted keratinocytes (Fig. 3 E). The association of OCT and KLF5, TFs essential for maintaining keratinocyte cell identity (Ge et al., 2017; Racila et al., 2011), with Mi-2β–regulated chromatin sites suggest that Mi-2β participates in a wider transcriptional network involved in both stress responses and keratinocyte differentiation.

Taken together, our studies show that barrier disruption causes a rapid and highly significant gain in chromatin accessibility that correlates with the rapid induction of stress response genes. The very small number of sites with reduced chromatin accessibility detected does not correlate with the large number of down-regulated genes, indicating that the loss in expression was not due to a change in chromatin structure. Finally, the similar genome-wide distribution of sites with gained chromatin accessibility and AP1 enrichment detected by either Mi-2β loss or barrier disruption strongly supports a role for this chromatin remodeler in the skin’s response to damage.

Stress signaling–induced chromatin changes are regulated by Mi-2β

We further evaluated the mechanism that controls induction of chromatin accessibility after barrier disruption or Mi-2β loss. We examined the temporal nature of chromatin changes induced at 6 or 24 h after TS and after Mi-2β loss. K-means clustering was performed with ATAC-seq datasets centered on dATACi peaks obtained at 6 or 24 h after TS (Fig. 4, A–D). In all five clusters, the increase in chromatin accessibility seen at 6 h after TS was reduced by 24 h (Fig. 4, A and B). These chromatin changes strongly correlated with the rapid induction of gene expression observed in response to barrier disruption (Fig. 1). On the other hand, three of the dATACi peak clusters (Fig. 4 C; C1, C2, and C4) seen at 24 h displayed a similar increase in chromatin accessibility at 6 h, indicating rapid induction but also maintenance of accessibility. This is consistent with the maintenance of induced expression for a subset of stress response genes (Fig. 2 C). Two of the dATACi clusters that were specifically induced at 24 h (Fig. 4, C and D; C3 and C5), were associated with cell cycle genes such as Ccnh, Ccnl1, and Cdk14 expressed later in the response, as shown in Fig. 1, E and F.

Notably, in four of the dATACi clusters detected at 6 h after barrier disruption (C1, C3, C4, and C5) that displayed little chromatin accessibility at steady state, ≥50% of the peaks were associated with AP1 motifs (Fig. 4 E). In contrast, peaks of the C2 cluster that were already accessible at steady state showed a much lower association with AP1. Furthermore, whereas ∼1/3 of the C2 peaks were associated with promoters, very little promoter association (on average <5%) was observed for the other four clusters that were strongly associated with AP1 motifs (Fig. 4 E). Three of the five dATACi clusters detected at 6 or 24 h showed overlap with dATACi peaks in Mi-2β–depleted keratinocytes (Fig. 4, A–D). Moreover, ∼65–80% of the genes associated with dATACi peaks upon barrier disruption also displayed either overlapping or associated dATACi peaks upon Mi-2β loss (Fig. 4 F and Fig. S3 A). Thus barrier disruption or Mi-2β loss caused similar changes in chromatin accessibility at regulatory sites located in the vicinity of stress response genes.

Genes associated with an increase in chromatin accessibility displayed a significant and progressive increase in expression from 6 to 24 h (Fig. 4 G). Genes that were up-regulated at 6 h after barrier disruption were also tested for their association with dATACi peaks (Fig. 4 H and Table S4). 33% of the induced genes at 6 h after TS were associated with dATACi peaks at this time point. An additional 28% of these genes were also associated with dATACi peaks induced by genetic ablation of Mi-2β (Fig. 4 H). A majority (84%) of the genes induced at 6 h that were associated with dATACi peaks after TS were also associated with dATACi peaks after Mi-2β deletion. Furthermore, genes that were commonly induced by barrier disruption or Mi-2β depletion showed an even greater association (i.e., 89%) with dATACi peaks caused by Mi-2β deletion (Fig. 4 H). A very strong enrichment with dATACi peaks caused by Mi-2β loss or TS (i.e., 91–95%) was also observed for up-regulated genes at 24 h after barrier disruption (Fig. S3 B, 24 h).

Il24, an activator of MAPK and STAT3 signaling in keratinocytes, Cldn4, Krt17, and Ngf, are examples of stress response genes whose expression is normally repressed by Mi-2β in the steady state (Fig. 4 I and Fig. S3 C). In these cases, a rapid increase in chromatin accessibility at upstream or intragenic sites was detected upon barrier disruption or Mi-2β loss that correlated with rapid induction in gene expression. In some cases, the promoters were already in an accessible state that was not greatly increased by either barrier disruption or Mi-2β loss.

Thus, changes in chromatin accessibility occur rapidly after stress response signaling is induced in keratinocytes. Some of these chromatin changes are dissipated rapidly while others are maintained. A highly significant number of these stress response–induced regulatory sites also respond to Mi-2β depletion in keratinocytes and are associated with genes induced during both processes, implicating this chromatin remodeler as a negative effector in their regulation.

Mi-2β–dependent mechanism of regulation is conserved between mouse and human

Mi-2β, an ATP-dependent chromatin remodeler, represses expression of genes normally induced by stress response signaling in keratinocytes, in part by restricting access at promoter-distal regulatory sites associated with putative AP1 binding sites. A potential direct role for Mi-2β in this molecular process was tested by establishing its chromatin distribution in both mouse and human cultured primary keratinocytes and in an immortalized human keratinocyte cell line (HaCaT) by chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq).

A highly significant overlap was seen for genes associated with Mi-2β chromatin enrichment sites (P < 2.77 × 10−68) in mouse and human keratinocytes (Fig. 5 A and Tables S6 and S7). Shared genes contributed to biological pathways involved in immune cell activation, epithelial cell proliferation and differentiation, cell adhesion, and EGFR and MAPK signaling (Fig. 5 B). Genomic annotation of Mi-2β chromatin enrichment sites revealed that >70% were distributed at non-promoter sites in both mouse and human keratinocytes (Fig. 5 C). De novo motif analysis of Mi-2β enrichment sites in chromatin from mouse or human keratinocytes again implicated AP1 as a major Mi-2β functional partner (Fig. S4 A). Furthermore, a majority of genes induced at 6–24 h (i.e., 64–65%) after barrier disruption in the mouse epidermis were directly bound by Mi-2β (Fig. 5 D). An even greater association with Mi-2β (82%) was seen for genes commonly induced by barrier disruption (6–24 h) and Mi-2β loss in mouse epidermis (Fig. 5 D). Additionally, 73% of these commonly repressed genes in mouse keratinocytes were also bound by Mi-2β in human keratinocytes (Fig. 5 D).

To further evaluate the Mi-2β–based mechanisms, we knocked-down Mi-2β mRNA with shRNA lentiviral vectors in cultured normal human keratinocytes (NHKs; Fig. 5 E and Fig. S4 B). Cultured keratinocytes replicate aspects of in vivo basal keratinocytes and are amenable to mechanistic studies. ChIP-seq studies for histone modifications and TFs were performed with both control (CTR) and Mi-2β–depleted (Mi2KD) primary human keratinocytes (Fig. 5 E and Fig. S4 C). Regulatory elements associated with Mi-2β were established by K-means clustering of histone modifications at Mi-2β binding sites in NHK chromatin (Fig. 5 F). Mi-2β distribution was detected at promoters (C2: marked by both H3K27Ac and H3K4me3), in the vicinity of promoters (C1 and C5: H3K27Ac and H3K4me3), and at promoter-distal enhancers (C3: marked only by H3K27Ac; Fig. 5 F). Upon Mi-2β depletion, enhancers (C3) associated with Mi-2β exhibited an increase in enrichment for H3K27Ac and RNA polymerase II (RNApII), indicating a potential increase in activity (Fig. 5, F and G). The preferential increase in transcriptionally permissive histone modifications at Mi-2β–associated enhancers compared with promoters in NHKs is reminiscent of the significant increase in Mi-2β–associated dATACi peaks at promoter-distal sites in mouse keratinocytes, indicating a conserved mechanism of regulation (Fig. 3 C). No change in the already strong H3K4me3 enrichment was seen at the Mi-2β–associated promoters (Fig. 5, F and G, C2: promoter). The Mi-2β–associated C3 enhancers were significantly enriched for AP1 motifs, although enrichment for p63, TEAD, and CEBP binding was also observed at lower frequency (Fig. 5 H). Genes associated with the Mi-2β C3 enhancers were involved in leukocyte activation, establishment of endothelial barrier, epidermal differentiation, and MAPK signaling and were up-regulated upon Mi-2β loss (Fig. 5 I). About 1/3 of the genes associated with Mi-2β–bound enhancers displayed binding of Mi-2β at their promoter regions (Fig. S4 D).

Thus, Mi-2β serves as the central node of a conserved epigenetic mechanism that directly controls expression of gene networks that are involved in both the response to stress signaling and keratinocyte differentiation.

Antagonism between Mi-2β and stress response signaling in the regulation of AP1

Normally repressed enhancers associated with Mi-2β in NHKs were enriched for AP1 motifs, suggesting a potential antagonism between Mi-2β and an activated AP1 complex (Fig. 5 H). AP1 motifs were also highly enriched at enhancers induced by barrier disruption in mouse keratinocytes and upon Mi-2β depletion (Figs. 3 and 4). In addition, among the genes directly bound by Mi-2β in NHKs was c-JUN and in mouse keratinocytes Jun and JunB (Fig. 6, A and B).

Upon Mi-2β knockdown (Mi2KD) in NHKs, a rapid increase of c-JUN mRNA, protein, and protein phosphorylation were seen from early to later time points after Mi-2β depletion (Fig. 6, C and D; c-JUN and phosphorylated-c-JUN [p-c-JUN] at days 3 and 5 in Mi-2βKD). In contrast to c-JUN, JUNB was expressed in NHKs and displayed a small increase at the protein level at 3 d after Mi-2β depletion and a small reduction at 5 d (Fig. 6 D; JUNB day 3 versus day 5). However, in mouse ex vivo keratinocytes, induction of JunB mRNA and protein were reproducibly seen both after Mi-2β depletion and TS (Fig. 6, E and F). Although an increase in c-Jun mRNA was not reproducibly seen after Mi-2β depletion or TS in mouse keratinocytes, a strong increase at the protein and phosphorylation level were detected (Fig. 6 F). Mi2Δ mouse epidermis was also tested for c-Jun protein expression by immunostaining. c-Jun expression was augmented at the basal and suprabasal layer of the epidermis as well as at the granular layer, where AP1 is normally involved in terminal differentiation (Fig. 6 G). Thus, loss of Mi-2β or mechanical stress causes rapid induction of AP1 factors in basal keratinocytes in support of stress signaling.

Functional interactions between Mi-2β and AP1 in human keratinocytes

We next tested the chromatin distribution of JUNB and c-JUN before and after Mi-2β depletion in NHKs (Fig. 7, A–C). Whereas JUNB enrichment peaks were readily detected in NHKs, a very small number of c-JUN peaks were present, consistent with the very low level of c-JUN expression in these cells (Fig. 7 A, Fig. 6, C and D; and Table S8). De novo c-JUN peaks dramatically increased upon Mi2KD in NHKs and overlapped with the majority of JUNB peaks observed in CTR NHKs (Fig. 7, B and C). Motif analysis of JUNB and c-JUN peaks in NHKs before and after Mi2KD confirmed binding through AP1 motifs that were distributed predominantly at promoter distal sites (Fig. S5, A and B). Co-clustering of JUNB (NHK/CTR), c-JUN (NHK Mi2KD), and Mi-2β (NHK and HaCaT) ChIP-seq datasets at de novo c-JUN peaks detected after Mi2KD revealed co-occupancy of JUNB and Mi-2β at these sites in NHKs (Fig. 7, B and C). Although co-occupancy by JUNB and Mi-2β was detected at the chromatin level in NHKs, stable interaction between these factors was not readily seen off chromatin (Fig. S5 C). Upon Mi2KD, an overall increase in H3K27ac was detected in the vicinity of JUNB and c-JUN enrichment peaks, a majority of which were bound by Mi-2β in NHKs and HaCaTs (Fig. 7, B and C; and Fig. S5 D). Up-regulated genes associated with de novo c-JUN peaks in Mi2KD NHKs supported pathways of cell adhesion, migration, signaling, and proliferation that are also induced by stress responses (Fig. 7 D).

The functional contribution of c-JUN to genes induced by loss of Mi-2β in NHKs was tested. Gene expression changes between NHKs transfected with a CTR shRNA vector and NHKs transfected with shRNAs for CHD4 (Mi-2β), c-JUN, or CHD4 and c-JUN were compared (Fig. S5 E, Fig. 7 E, and Table S9). About 1/4 (419/1,611) of the genes up-regulated upon Mi2KD showed a significant reduction in expression by concomitant knockdown of c-JUN (Fig. S5 E and Fig. 7 E). These reverted genes showed a highly significant association with c-JUN peaks compared with non reverted genes as calculated by a χ2 test (Fig. 7, F and G; P = 0.001). Furthermore, Mi-2β was more significantly enriched in the proximity of genes that were induced by its loss and bound by c-JUN compared with genes that were induced but not bound by c-JUN (Fig. 7 H; P = 0.0001). COL7A1, RARA, and PLB1 are examples of direct targets of Mi-2β and c-JUN that are regulated by the two factors in a reciprocal fashion (Fig. 7 I and Fig. S5, F and G).

In conclusion, these studies reveal a dynamic interaction between Mi-2β and the heterogeneous collection of protein heterodimers collectively referred to as the AP1 complex. Under homeostatic conditions in human keratinocytes, an AP1 complex containing JUNB co-localizes with Mi-2β on repressed genes that include those encoding alternative AP1 complex components. Stress signaling or Mi-2β depletion relieve the repression and allow an activating AP1 complex to promote gene expression. The antagonism between Mi-2β and this activating AP1 complex contributes to timely abatement of stress responses.

Mi-2β response to and recovery from barrier disruption

Our studies indicate that Mi-2β holds the genes of the stress response in a repressed state and that abrogation of Mi-2β activity induces their expression by activating AP1. However, stress response genes are only a subset of the genes held repressed by Mi-2β in the epidermis. Thus, if interference with Mi-2β is the mechanism by which stress response genes are activated, there must be a selective reduction of Mi-2β activity at these loci.

To test this hypothesis, we first examined Mi-2β expression during the stress response. A consistent, albeit, small reduction in Mi-2β mRNA was detected at 6 h after TS (Fig. 8 A). Immunostaining for Mi-2β protein in the skin also showed a decrease in the nuclei of the basal epidermis (Fig. 8 B). A small reduction in Mi-2β activity can sensitize its direct gene targets to additional mechanisms that antagonize its activity. Indeed, when we tested whether Mi-2β chromatin occupancy was altered during the stress response by ChIP–quantitative PCR (qPCR) at candidate Mi-2β gene targets that are induced during the response, a significant reduction in Mi-2β occupancy at 6 h after barrier disruption was seen that correlated with an increase in expression (Fig. 8 C). JunB, Aloxe3, Mapk6, and Ppid were genes induced by barrier disruption in vivo that showed Mi-2β occupancy in cultured and ex vivo mouse keratinocytes. All exhibited a significant reduction in Mi-2β chromatin occupancy at 6 h after TS (Fig. 8 C). In contrast, Mi-2β occupancy at candidate loci whose expression was not affected by TS (e.g., Krt14 and Myod1) remained unaltered (Fig. 8 C).

Thus, induction of the response to barrier disruption may involve selective abrogation of Mi-2β activity at a subset of responder genes. Consistent with this model, Mi-2β activity was not required for an initial recovery of barrier function after TS (Fig. 8 D; mice treated 2× with 4-OHT before TS). Prior to barrier disruption, TEWL was at a similar low level in both WT and Mi-2β mutant skin (Fig. 8 E). After barrier disruption, a similar rapid recovery from TEWL was observed under both conditions (Fig. 8 E; 1–6 h). However, if Mi-2β is a critical repressor of these genes, then the resolution of the response should depend on Mi-2β activity. Since mice with efficient deletion of Mi-2β in the epidermis did not survive the additional stress of the TS procedure, mice with mosaic deletion of Mi-2β in the epidermis that tolerate the stress response were used for evaluation of the resolution phase (Fig. 8 D; mice treated once with 4-OHT before TS). A clear difference in skin phenotype was observed at 4–5 d after TS, with severe desquamation observed in the tape-stripped compared with noninjured areas in Mi-2β mutant skin (Fig. 8 G). Histological analysis showed mild thickening of WT epidermis at 24 h that returned to normal thickness by 5 d (Fig. S2 A and Fig. 8 F). In agreement with histology, expression of K6 and Ki67 in WT epidermis was induced at 6–24 h after TS but returned to a preinjury level by 5 d after treatment (Fig. S2 A and Fig. 8 G; WT tape-stripped day 5). In sharp contrast, Mi-2β mutant epidermis showed a severe and persistent increase in thickness and keratinization after TS compared with noninjured areas (Fig. 8 F). Further increase in Ki67-positive cells and strong K6 expression was seen in Mi2β mutant epidermis at 5 d after TS (Fig. 8 G; Mi2Δ mosaic tape-stripped day 5).

In summary, a selective reduction of Mi-2β activity at rapidly responding genes appears to be a key early event in stress signaling. Nonetheless, reduction in Mi-2β activity is short-lived, and Mi-2β repression is rapidly reinstated for skin to return to homeostasis and normal differentiation. Upon constitutive, instead of a normally transient, loss of Mi-2β, epidermis appears to be over-repairing the disrupted barrier by hyperproliferation and hyperkeratinization, manifestations of aberrant skin differentiation.

Skin, as our interface to the outside world, is constantly exposed to environmental challenges to which it responds by inducing signaling and transcriptional programs that mend the structural barrier while placing the immune system on high alert. Both rapid induction and subsequent repression of stress responses are critical for preserving normal skin function by engaging signaling and TFs as well as chromatin regulators to work in concert toward this goal. Here, we show that Mi-2β, the nucleosome remodeler of the NuRD complex, is a key regulator of this process that functions by restricting chromatin access at enhancers associated with stress response genes, many of which contain AP1 binding sites. Our data support a model by which a transcriptionally repressive AP1 complex based on JUNB acts with Mi-2β to locally repress at these enhancer sites. Stress signaling displaces Mi-2β from these loci and may be sufficient to initiate stress response gene expression, de-repression and activation of c-JUN, and replacement of a repressive by an activating AP1 complex that may amplify the response (Fig. 9).

Previous studies have shown that loss of Mi-2β in the epidermis causes keratinocyte activation and skin inflammation in the absence of an overt differentiation defect, suggesting that Mi-2β and the NuRD complex play a critical role in regulating the skin’s stress responses (Kashiwagi et al., 2017). To test this hypothesis, we characterized both the transcriptional and chromatin response to skin barrier disruption induced by TS and compared it with that of Mi-2β loss in keratinocytes in vivo. A highly significant overlap in the induction of genes involved in epidermal barrier reestablishment, EGFR signaling, keratinocyte proliferation, and immune cell activation was seen under both conditions in mouse keratinocytes. A rapid induction of chromatin accessibility sites and gene expression changes observed upon barrier disruption indicated a mechanism of transcription regulation that incorporates active chromatin remodeling. The majority of newly accessible regulatory sites were located at promoter-distal areas and were also induced by loss of Mi-2β in the epidermis, thereby supporting a role of this chromatin remodeler in their regulation. These putative stress response enhancers were significantly enriched for AP1 binding sites, suggesting that Mi-2β holds the stress response genes in a poised state by restricting their access (Fig. 9). This hypothesis was examined in detail using cultured keratinocytes from both human and mouse skin. Mi-2β chromatin enrichment sites were associated with transcriptional enhancers encountered at both differentiation and stress response genes before their activation.

AP1 is a transcriptional activator that is comprised of dimers formed between members of the JUN and FOS families that are differentially expressed in the basal and suprabasal layers of the skin (Welter and Eckert, 1995). AP1 family members regulate keratinocyte proliferation and differentiation and are also involved in the wound-healing process (Adhikary et al., 2004; Efimova et al., 2004; Takahashi et al., 2002). Aberrant activation of AP1 has been implicated in skin pathological conditions such as carcinogenesis and disease development caused by chronic inflammation (Kolev et al., 2008; Zenz et al., 2005). Mi-2β directly repressed expression of c-JUN and JUNB mRNAs in human and mouse keratinocytes, respectively. In addition, an increase in JunB and Jun protein and phosphorylation were detected in the basal layer of mouse epidermis after Mi-2β depletion or barrier disruption. Nonetheless, in NHKs, JUNB but not c-JUN was expressed and occupied many enhancers that were also enriched for Mi-2β. Upon loss of Mi-2β, these enhancers became occupied by activated c-JUN and displayed an increase in chromatin accessibility and transcription of associated genes. The coexistence of JUNB with Mi-2β-NuRD at AP1 sites suggests the existence of an AP1 transcription-repressive complex that upon stress signaling is replaced by an AP1 transcription-activating complex (Fig. 9). This model is supported by a previous study that demonstrated that repression of gene expression was mediated by c-JUN/NuRD complex interaction in colon cancer epithelial cells, and that this interaction and the gene repression were disrupted by JNK signaling (Aguilera et al., 2011). The transient nature of a transcription-activating AP1 complex may permit rapid reactivation of Mi-2β and rerepression of stress response genes.

The polycomb-PRC2 complex is another epigenetic regulator of keratinocyte differentiation that functions in part by antagonizing AP1 activity (Ezhkova et al., 2009; Wurm et al., 2015). PRC2 is highly active in basal keratinocytes, where it restricts access to AP1 sites required for expression of keratinocyte differentiation genes, such as those of the epidermal differentiation complex (Ezhkova et al., 2009; Wurm et al., 2015). In suprabasal cells, PRC2 expression is diminished, allowing AP1 to access enhancers supporting epidermal differentiation complex expression and stratification. While these chromatin-based mechanisms of restricting AP1 activity are similar, there is a major mechanistic difference. Whereas the loss of PRC2 associated with terminal keratinocyte differentiation is permanent, Mi-2β activity is transiently overridden in the basal layer to allow a rapid stress response and subsequent recovery.

The strong overlap in gene expression and chromatin changes observed during barrier disruption and Mi-2β depletion establish the importance of this chromatin remodeler in holding genes poised to respond to challenge and suggest that abrogation of Mi-2β may be a mechanism by which these genes are activated. However, not all genes regulated by Mi-2β in the epidermis are activated in response to barrier disruption. In addition to a decrease in Mi-2β transcription and protein after TS, we observed a selective decrease in Mi-2β occupancy at genes induced by stress response in vivo. The general reduction in Mi-2β expression may cooperate with enhanced expression or activation of TFs like AP1 to tip the balance of competition between repression and activation selectively at gene loci occupied by these factors. Other genes dependent on Mi-2β for repression may remain silent if the TFs responsible for their induction are not activated. In that context, it is perhaps noteworthy that the subset of stress response genes that also respond to Mi-2β deletion remain expressed longer than the subset that does not. The incorporation of a positive feedback loop (that attenuates Mi-2β expression) to tip this balance may take longer to reverse than a mechanism supporting induction of stress response genes that are not repressed by Mi-2β. Consistent with a repressive role of Mi-2β in the reestablishment of skin barrier function, its loss did not affect this process upon mechanical injury. However, even mosaic depletion of Mi-2β in the epidermis caused too much repair and regeneration of the disrupted barrier, with excessive proliferation and keratin production that resulted in hyperkeratosis and acanthosis. Thus, rapid restoration of Mi-2β activity is necessary for transient activation of stress response signaling and normal keratinocyte differentiation.

In conclusion, our studies provide a new working model for the mechanisms that permit rapid response to skin challenge and return to epidermal homeostasis (Fig. 9). Of the genes induced by mechanical barrier disruption, roughly half depend on Mi-2β activity to remain repressed in normal epidermis. For this Mi-2β–repressed gene subset, the challenge results in a transient and local reduction in Mi-2β activity, which, in concert with augmented expression of transcription-activating factors such as c-JUN, elicits a rapid response. These chromatin changes and the transcriptional response to stress are short-lived, as is the signaling that supports c-JUN activation; Mi-2β activity is rapidly reinstated, and the balance is shifted back to gene repression. The antagonistic interplay between epigenetic and transcriptional regulators was well established here, but the mechanism by which Mi-2β repressive activity is reinstated in a timely fashion at the end of the stress response remains to be elucidated. While the changes in AP1 phosphorylation and composition summarized in Fig. 9 are formally sufficient, interactions between Mi-2β and other transcriptional regulators bound nearby may also contribute to the reestablishment of repression. The discovery of key nodes in the regulatory network that controls keratinocyte homeostasis presented in these studies provides access to these questions. Our detailed characterization of the genomic response to barrier disruption also identifies genes that may be regulated by distinct mechanisms.

While our effort to confirm the relevance of this work in mice to human skin was confined to studies in vitro, it is noteworthy that a study of newly accessible chromatin sites in xenografted human skin equivalents after TS also identified AP1 as a likely regulator of the barrier disruption response (Lander et al., 2017). In addition, microarray analysis of gene expression changes in human skin after barrier disruption is consistent with the results reported here (Sextius et al., 2010). This suggests that our work, which provides a mechanism by which an AP1-regulated response may be mobilized in response to stress, may be translated to new approaches to target chronic skin conditions in which a failure to repress the stress response leads to a range of skin disease and cancer.

Mice

Mi-2βloxF/loxF mice were generated in the Georgopoulos laboratory (Williams et al., 2004) and bred to the Krt14-CreERT2 transgenic mice as previously described (Kashiwagi et al., 2017). Mi-2βloxF/loxFKrt14-CreERT2 transgenic mice used in this study were backcrossed to C57BL/6 mice more than six times. C57BL/6 mice were used for TS. Mice were 8–12 wk old at the time of analyses. Gene inactivation was achieved by two i.p. injections of mice with 0.5 mg of 4-OHT (Sigma-Aldrich) with a 3-d interval. This protocol was used to generate keratinocyte samples for RNA-seq and ATAC-seq studies and to study the role of Mi-2β in the initial recovery phase of barrier function after TS. To examine the role of Mi-2β in the return to homeostasis after barrier disruption, mice were delivered a single i.p. injection of 4-OHT to achieve mosaic deletion of Mi-2β before TS. All animal experiments were performed according to protocols approved by the Subcommittee on Research Animal Care at Massachusetts General Hospital (Charlestown, MA) and in accordance with the guidelines set forth by the US National Institutes of Health.

Mechanical injury by barrier disruption

Barrier disruption by TS was performed with 20–50 strokes of transparent tape on the shaved back skin until TEWL reached 40 grams per square meter per hour (g/m2 ⋅ h). For evaluation of barrier recovery, TEWL was measured before (x) and at 0 (y), 1 (z1), 3 (z2), and 6 (z3) h after barrier disruption. The recovery rate was calculated by yz (z1, z2, or z3)/yx × 100 (%). Skin was harvested for RNA-seq and ATAC-seq at the indicated time points (before and at 6, 24, 96, and 120 h after TS).

Cell isolation and sorting

Dorsal epidermis was separated from the dermis by digestion with 0.25% trypsin overnight at 4°C to obtain a single-cell suspension. Cells were stained with anti-CD45 (30-F11), anti-CD34 (RAM34), and anti-ITGA6 (eBioGoH3; eBioscience or BD PharMingen). ITGA6+, CD45, and CD34 cells were sorted with a FACS Aria (BD Biosciences) and used for further analyses.

Keratinocyte cell cultures

Primary human epidermal keratinocytes were isolated from three independent de-identified neonatal foreskin samples (a gift from Drs. Rachael A. Clark and Jessica E. Teague, Brigham and Women’s Hospital, Boston, MA). Briefly, a single-cell suspension was obtained from foreskin by incubation with 2 U/ml Dispase II (Roche) overnight at 4°C, followed by incubation with 0.25% trypsin (Gibco) for 10 min at 37°C. For primary mouse keratinocytes, cells were prepared from newborn skin by digestion with 0.25% trypsin (Gibco) overnight at 4°C. Primary keratinocytes were cultured under serum-free conditions with keratinocyte growth medium (Gibco). HaCaT cells were cultured in DMEM supplemented with 10% FCS.

Lentiviral gene transduction

The shRNA lentiviral constructs for Mi-2β, c-JUN, and CTR were obtained from Mission shRNA (Sigma-Aldrich). Transfection into HEK293T cells and lentiviral production were performed as described previously ((Hu et al., 2016). For transduction into primary NHKs, cells were incubated with the infection medium containing lentiviral particles for 6 h and then replaced with fresh medium containing 1 µg/ml puromycin (Sigma-Aldrich). Cells were then selected for 3–5 d in puromycin and used for further studies.

Histology, immunofluorescence, and immunohistochemistry

Samples were fixed in 4% paraformaldehyde overnight at 4°C and then embedded in paraffin. Tissues were sectioned at 5 µm and stained with hematoxylin and eosin after de-paraffinization. For immunofluorescence, sections were de-paraffinized and then used for antigen retrieval using a citrate-based buffer (Vector). The primary antibodies used for staining were rabbit polyclonal Ki67 (Abcam) and rabbit polyclonal K6 (BabCo). Alexa Flour 488 conjugated anti-rabbit antibodies (Cell Signaling) were used for secondary antibodies, and DAPI (Vector) was used for counterstaining nuclei. For immunohistochemistry, a mouse monoclonal Mi-2β (16G4; Millipore; Kim et al., 1999) was used as a primary antibody. The MOM Fluorescent kit (Vector) was used for staining. Sections were incubated with the avidin-biotin-peroxidase complex and then stained with 3,3′-Diaminobenzidine (Vector). Images were taken with a Nikon A1 confocal microscope or a Zeiss Axio Scan.

RNA-seq and data analysis

RNA-seq datasets were generated from two independent experiments using 106 cells for mouse sorting keratinocytes (Figs. 1, 2, and 4) and 2 × 106 cells for human primary keratinocytes expanded in culture (Figs. 5 and 7). Total RNA was extracted with a Direct-zol RNA kit (Zymo Research). The Truseq stranded mRNA sample prep kit (Illumina) was used for the construction of cDNA libraries for RNA-seq. The libraries were single-end sequenced using the Illumina Hiseq 2000 platform at the Bauer Center Systems Biology Core at Harvard University or the Illumina NextSeq 550 at the Cutaneous Biology Research Center. Read alignment was performed on the mm10 or hg19 assembly of the mouse or human genome respectively with the STAR genome alignment algorithm (Dobin et al., 2013). Read normalization, reads per kilobase of transcript per million (RPKM), and differential gene expression were performed using the HOMER scripts analyzeRepeats.pl and getDiffExpression.pl with implementation of DESeq2 through R (Love et al., 2014). Heat maps of RPKM values for differentially expressed genes were generated with Cluster 3.0 (open source software by Michael Eisen) and visualized with Java TreeView (open source software by Alok J. Saldanha). The PANTHER pathway was used for GO analysis (Mi and Thomas, 2009). GSEA was used to compare gene expression changes detected upon TS or in Mi2Δ relative to WT keratinocytes. Significantly up- and down-regulated genes (DESeq2-log2 fold change > 1 or < –1, FDR < 0.05) detected after TS were compared with all genes in the Mi2Δ keratinocyte RNA-seq dataset using the preranked GSEA parameter for analysis. The all-genes Mi2Δ keratinocyte RNA-seq dataset was ranked by log2 fold change relative to the WT keratinocyte dataset.

ATAC-seq and data analysis

5 × 104 freshly sorted cells were used for ATAC-seq. The construction of cDNA libraries for ATAC-seq was performed as described (Buenrostro et al., 2013). Briefly, cells were subjected to transposition reaction with 3 µl of the Tagment DNA enzyme at 37°C for 30 min (Nextera DNA sample preparation kit; Illumina). Transposed DNA was amplified and then purified using the Qiagen PCR Purification kit. The quality of libraries was tested by the Bioanalyzer (Agilent). Libraries were sequenced using a Hiseq 2000 platform at the Bauer Center Systems Biology Core at Harvard University. Read alignment was performed on the mm10 mouse assembly using the Bowtie2 genome alignment algorithm 1.4.1 with the sensitive option (Langmead et al., 2009). Duplicate reads were removed by Picard Tools (Broad Institute). Peak calling for ATAC-seq was performed with MACS2 using the following parameters for peak calling: --keep-dup all --nomodel --shift 37 --extsize 73 (Zhang et al., 2008). Differentially enriched peaks in-between conditions were obtained by setting one condition as test and the other as input control.

ChIP, ChIP-seq, and data analysis

ChIP was performed as previously described, with some modifications (Zhang et al., 2011; Hu et al., 2016). Cells were fixed with 1% fresh formaldehyde for 8 min at room temperature and then quenched by 2.5 M glycine. Cells were lysed in swelling buffer (25 mM Hepes-KOH at pH 7.8, 1.5 mM MgCl2, 10 mM KCl, 1% igepal, 1 mM 1,4-dithiothreitol, and 1× protease inhibitor cocktail) for 10 min on ice, followed by Dounce homogenization. Nuclei were pelleted and resuspended in radioimmunoprecipitation assay buffer (10 mM Tris at pH 8.0, 140 mM NaCl, 1% Triton X-100, 0.1% Na-deoxycholate, 0.1% SDS, 1 mM EDTA, and 1× protease inhibitor cocktail). Chromatin was sonicated to an average size of 300–500 bp with the Diagenode Bioruptor. Chromatin was cleared by centrifugation at 20,000 g for 10 min at 4°C followed by overnight incubation at 4°C with primary antibodies that were prebound to Dynabeads protein G (Life Technologies). The primary antibodies used were H3K27Ac (Abcam, ab4729), H3K4me3 (Active motif, 39159), RNApII (Abcam, ab817), c-JUN (Abcam, ab31419), JUNB (CST, C37F9), and Mi-2β (Abcam, ab72418; and CST, 12011). The number of cells used for each antibody in ChIP were ∼2 × 106 cells for H3K27Ac, 4 × 106 cells for H3K4me3 and RNApII, 107 cells for c-JUN, 107 cells for JUNB, and 2 × 107 cells for Mi-2β. Chromatin was washed five times with immunoprecipitation wash buffer (10 mM Tris at pH 8.0, 500 mM NaCl, 1% Triton X-100, 0.1% Na-deoxycholate, and 0.1% SDS), followed by a wash with TE buffer (50 mM Tris at pH 8.0, 1 mM EDTA, and 50 mM NaCl). Chromatin was eluted and then de-crosslinked overnight at 65°C. DNA was purified using the DNA Clean and Concentration 5 kit (Zymo Research). Fold enrichment for ChIP-qPCR was calculated as the ratio of ratios obtained from Mi-2β ChIP and input control for Mi-2β target regulatory regions/negative control region. For ChIP-seq, the construction of cDNA libraries was performed as previously described (Hu et al., 2016). In brief, 2–50 ng of DNA was end-repaired, end-adenylated, and then ligated with Illumina TruSeq indexed adaptors. The ligated DNA was amplified, size-selected on a 2% agarose gel, and purified using a gel DNA recovery kit (Zymo Research). The libraries were sequenced using a Hiseq 2000 platform at the Bauer Center Systems Biology Core at Harvard University. Read alignment was performed on the hg19 assembly of the human genome using either Bowtie2 or the STAR genome mapper with the option to disable spliced alignments and prohibit gaps (Dobin et al., 2013). Peak calling for ChIP-seq was performed using the Homer findPeaks algorithm (Heinz et al., 2010). Differentially enriched peaks between two conditions were obtained by setting one condition as test and the other as input control. Heat maps of K-means clustering of ChIP-seq or ATAC-seq enriched peaks were generated using NGS.plot (Shen et al., 2014). Histograms of read densities for histone modifications and TFs at regulatory sites were plotted with NGS.plot (Shen et al., 2014). HOMER de novo motif discovery algorithm was used for motif discovery analysis of TF and dATAC peaks. Venn diagrams were generated with the online-based Venny Tool.

Co-immunoprecipitation analysis

Cells were lysed with lysis buffer (25 mM Tris at pH 7.5, 10% glycerol, 150 mM NaCl, 1.5 mM MgCl2, and protease inhibitors) containing 1% Triton X-100. The protein extracts were precleared with Dynabeads protein G (Invitrogen). The precleared extracts were incubated with the anti-Mi2β (Abcam, ab72418), JUNB (CST, C37F9), and cJUN (Abcam, ab31419) or the relevant isotype control in the presence of Dynabeads protein G and rotated overnight. Beads were then collected, washed at least five times with lysis buffer, and resuspended in SDS sample buffer. Eluents were subjected to SDS-PAGE and transferred to polyvinylidene difluoride membranes, probed with anti-Mi2β (CST, 12011), JUNB (CST, C37F9), cJUN (BD Bioscience, 610327), HDAC2 (CST, D6S5P), and MTA3 (Bethyl Laboratories, A300-160), and examined by autoradiography by Enhance Chemical Luminescence.

Data availability

The sequencing datasets generated by this study have been deposited to National Center for Biotechnology Information and are accessible through Gene Expression Omnibus series accession no. GSE139685.

Online supplemental material

Fig. S1 shows gene expression changes in mouse basal epidermis upon barrier disruption. Fig. S2 shows common transcription changes induced by barrier disruption and Mi-2β depletion in mouse basal epidermis. Fig. S3 shows changes in chromosome accessibility and associated transcriptional changes induced by barrier disruption and Mi-2β depletion in mouse basal epidermis. Fig. S4 provides evidence for a conserved Mi-2β–dependent mechanism of transcriptional regulation in mouse and human keratinocytes. Fig. S5 provides evidence for functional antagonism between AP1 complex and Mi-2β in human keratinocytes by changes in gene expression in single and double knockdown keratinocytes and by ChIP-seq analysis in these cells. Table S1 contains the datasets used to draw box-whisker plots in Fig. 1 E. Table S2 contains the IDs of genes that were commonly de-regulated by barrier disruption and Mi-2β depletion in mouse basal epidermis. Table S3 contains dATACi and dATACl peak files at 6 h after TS. Table S4 contains dATACi and dATACl peak files at 24 h after TS. Table S5 contains dATACi and dATACl peak files in Mi-2β–depleted compared with WT control keratinocytes. Table S6 contains Mi-2β ChIP-seq peaks identified in mouse keratinocytes. Table S7 contains Mi-2β ChIP-seq peaks in human keratinocytes. Table S8 contains cJUN and JUNB ChIP-seq peaks in human keratinocytes. Table S9 contains two datasets: one dataset of up-regulated genes upon Mi2KD in human keratinocytes and a second dataset of genes whose expression was reverted by concomitant knockdown of Mi-2β and c-JUN in human keratinocytes.

We thank Drs. Rachael A. Clark and Jessica E. Teague and the Human Skin Disease Resource Center at Brigham and Women’s Hospital for providing us with discarded de-identified foreskin from human newborns. We thank Dr. Jin Mo Park for critical review of the manuscript, and Eleanor Wu and Robert Czyzewski for mouse husbandry.

Research was supported by National Institutes of Health grant R01AR069132 to K. Georgopoulos and M. Kashiwagi and by National Institutes of Health grant R21AR072976 to B.A. Morgan. K. Georgopoulos is a Massachusetts General Hospital scholar supported by Dr. Jean de Gunzburg. S. Shibata has been supported by fellowships from the Japanese Society for the Promotion of Science and Uehara Memorial Foundation. High-throughput RNA-seq was performed at the Bauer Center for Genomic Research at Harvard University and at Cutaneous Biology Research Center.

The authors declare no competing financial interests.

Author contributions: S. Shibata, M. Kashiwagi, K. Georgopoulos, and B.A. Morgan designed the experiments. S. Shibata and M. Kashiwagi performed the experiments. S. Shibata, M. Kashiwagi, and K. Georgopoulos analyzed the data. S. Shibata and K. Georgopoulos mainly wrote the manuscript. K. Georgopoulos and B.A. Morgan supervised the study.

Adhikary
,
G.
,
J.
Crish
,
J.
Lass
, and
R.L.
Eckert
.
2004
.
Regulation of involucrin expression in normal human corneal epithelial cells: a role for activator protein one
.
Invest. Ophthalmol. Vis. Sci.
45
:
1080
1087
.
Aguilera
,
C.
,
K.
Nakagawa
,
R.
Sancho
,
A.
Chakraborty
,
B.
Hendrich
, and
A.
Behrens
.
2011
.
c-Jun N-terminal phosphorylation antagonises recruitment of the Mbd3/NuRD repressor complex
.
Nature.
469
:
231
235
.
Becker
,
P.B.
2002
.
Nucleosome sliding: facts and fiction
.
EMBO J.
21
:
4749
4753
.
Brehm
,
A.
,
G.
Längst
,
J.
Kehle
,
C.R.
Clapier
,
A.
Imhof
,
A.
Eberharter
,
J.
Müller
, and
P.B.
Becker
.
2000
.
dMi-2 and ISWI chromatin remodelling factors have distinct nucleosome binding and mobilization properties
.
EMBO J.
19
:
4332
4341
.
Buenrostro
,
J.D.
,
P.G.
Giresi
,
L.C.
Zaba
,
H.Y.
Chang
, and
W.J.
Greenleaf
.
2013
.
Transposition of native chromatin for fast and sensitive epigenomic profiling of open chromatin, DNA-binding proteins and nucleosome position
.
Nat. Methods.
10
:
1213
1218
.
Dainichi
,
T.
,
A.
Kitoh
,
A.
Otsuka
,
S.
Nakajima
,
T.
Nomura
,
D.H.
Kaplan
, and
K.
Kabashima
.
2018
.
The epithelial immune microenvironment (EIME) in atopic dermatitis and psoriasis
.
Nat. Immunol.
19
:
1286
1298
.
Dobin
,
A.
,
C.A.
Davis
,
F.
Schlesinger
,
J.
Drenkow
,
C.
Zaleski
,
S.
Jha
,
P.
Batut
,
M.
Chaisson
, and
T.R.
Gingeras
.
2013
.
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics.
29
:
15
21
.
Efimova
,
T.
,
A.M.
Broome
, and
R.L.
Eckert
.
2004
.
Protein kinase Cdelta regulates keratinocyte death and survival by regulating activity and subcellular localization of a p38δ-extracellular signal-regulated kinase 1/2 complex
.
Mol. Cell. Biol.
24
:
8167
8183
.
Ezhkova
,
E.
,
H.A.
Pasolli
,
J.S.
Parker
,
N.
Stokes
,
I.H.
Su
,
G.
Hannon
,
A.
Tarakhovsky
, and
E.
Fuchs
.
2009
.
Ezh2 orchestrates gene expression for the stepwise differentiation of tissue-specific stem cells
.
Cell.
136
:
1122
1135
.
Ge
,
Y.
,
N.C.
Gomez
,
R.C.
Adam
,
M.
Nikolova
,
H.
Yang
,
A.
Verma
,
C.P.
Lu
,
L.
Polak
,
S.
Yuan
,
O.
Elemento
, and
E.
Fuchs
.
2017
.
Stem cell lineage infidelity drives wound repair and cancer
.
Cell.
169
:
636
650.e14
.
Gnocchi
,
D.
,
M.
Pedrelli
,
E.
Hurt-Camejo
, and
P.
Parini
.
2015
.
Lipids around the clock: focus on circadian rhythms and lipid metabolism
.
Biology (Basel).
4
:
104
132
.
Gómez-Del Arco
,
P.
,
E.
Perdiguero
,
P.S.
Yunes-Leites
,
R.
Acín-Pérez
,
M.
Zeini
,
A.
Garcia-Gomez
,
K.
Sreenivasan
,
M.
Jiménez-Alcázar
,
J.
Segalés
,
D.
López-Maderuelo
, et al
.
2016
.
The chromatin remodeling complex Chd4/NuRD controls striated muscle identity and metabolic homeostasis
.
Cell Metab.
23
:
881
892
.
Guinea-Viniegra
,
J.
,
R.
Zenz
,
H.
Scheuch
,
D.
Hnisz
,
M.
Holcmann
,
L.
Bakiri
,
H.B.
Schonthaler
,
M.
Sibilia
, and
E.F.
Wagner
.
2009
.
TNFalpha shedding and epidermal inflammation are controlled by Jun proteins
.
Genes Dev.
23
:
2663
2674
.
Heinz
,
S.
,
C.
Benner
,
N.
Spann
,
E.
Bertolino
,
Y.C.
Lin
,
P.
Laslo
,
J.X.
Cheng
,
C.
Murre
,
H.
Singh
, and
C.K.
Glass
.
2010
.
Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities
.
Mol. Cell.
38
:
576
589
.
Hu
,
Y.
,
Z.
Zhang
,
M.
Kashiwagi
,
T.
Yoshida
,
I.
Joshi
,
N.
Jena
,
R.
Somasundaram
,
A.O.
Emmanuel
,
M.
Sigvardsson
,
J.
Fitamant
, et al
.
2016
.
Superenhancer reprogramming drives a B-cell-epithelial transition and high-risk leukemia
.
Genes Dev.
30
:
1971
1990
.
Janich
,
P.
,
K.
Toufighi
,
G.
Solanas
,
N.M.
Luis
,
S.
Minkwitz
,
L.
Serrano
,
B.
Lehner
, and
S.A.
Benitah
.
2013
.
Human epidermal stem cell function is regulated by circadian oscillations
.
Cell Stem Cell.
13
:
745
753
.
Kashiwagi
,
M.
,
B.A.
Morgan
, and
K.
Georgopoulos
.
2007
.
The chromatin remodeler Mi-2β is required for establishment of the basal epidermis and normal differentiation of its progeny
.
Development.
134
:
1571
1582
.
Kashiwagi
,
M.
,
J.
Hosoi
,
J.F.
Lai
,
J.
Brissette
,
S.F.
Ziegler
,
B.A.
Morgan
, and
K.
Georgopoulos
.
2017
.
Direct control of regulatory T cells by keratinocytes
.
Nat. Immunol.
18
:
334
343
.
Keyes
,
B.E.
,
S.
Liu
,
A.
Asare
,
S.
Naik
,
J.
Levorse
,
L.
Polak
,
C.P.
Lu
,
M.
Nikolova
,
H.A.
Pasolli
, and
E.
Fuchs
.
2016
.
Impaired epidermal to dendritic T cell signaling slows wound repair in aged skin
.
Cell.
167
:
1323
1338.e14
.
Kim
,
J.
,
S.
Sif
,
B.
Jones
,
A.
Jackson
,
J.
Koipally
,
E.
Heller
,
S.
Winandy
,
A.
Viel
,
A.
Sawyer
,
T.
Ikeda
, et al
.
1999
.
Ikaros DNA-binding proteins direct formation of chromatin remodeling complexes in lymphocytes
.
Immunity.
10
:
345
355
.
Kolev
,
V.
,
A.
Mandinova
,
J.
Guinea-Viniegra
,
B.
Hu
,
K.
Lefort
,
C.
Lambertini
,
V.
Neel
,
R.
Dummer
,
E.F.
Wagner
, and
G.P.
Dotto
.
2008
.
EGFR signalling as a negative regulator of Notch1 gene transcription and function in proliferating keratinocytes and cancer
.
Nat. Cell Biol.
10
:
902
911
.
Lander
,
J.M.
,
D.M.
Supp
,
H.
He
,
L.J.
Martin
,
X.
Chen
,
M.T.
Weirauch
,
S.T.
Boyce
, and
R.
Kopan
.
2017
.
Analysis of chromatin accessibility in human epidermis identifies putative barrier dysfunction-sensing enhancers
.
PLoS One.
12
:e0184500.
Langmead
,
B.
,
C.
Trapnell
,
M.
Pop
, and
S.L.
Salzberg
.
2009
.
Ultrafast and memory-efficient alignment of short DNA sequences to the human genome
.
Genome Biol.
10
:
R25
.
LeBoeuf
,
M.
,
A.
Terrell
,
S.
Trivedi
,
S.
Sinha
,
J.A.
Epstein
,
E.N.
Olson
,
E.E.
Morrisey
, and
S.E.
Millar
.
2010
.
Hdac1 and Hdac2 act redundantly to control p63 and p53 functions in epidermal progenitor cells
.
Dev. Cell.
19
:
807
818
.
Love
,
M.I.
,
W.
Huber
, and
S.
Anders
.
2014
.
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol.
15
:
550
.
Luo
,
J.
,
F.
Su
,
D.
Chen
,
A.
Shiloh
, and
W.
Gu
.
2000
.
Deacetylation of p53 modulates its effect on cell growth and apoptosis
.
Nature.
408
:
377
381
.
Mehic
,
D.
,
L.
Bakiri
,
M.
Ghannadan
,
E.F.
Wagner
, and
E.
Tschachler
.
2005
.
Fos and jun proteins are specifically expressed during differentiation of human keratinocytes
.
J. Invest. Dermatol.
124
:
212
220
.
Mi
,
H.
, and
P.
Thomas
.
2009
.
PANTHER pathway: an ontology-based pathway database coupled with data analysis tools
.
Methods Mol. Biol.
563
:
123
140
.
Nestle
,
F.O.
,
P.
Di Meglio
,
J.Z.
Qin
, and
B.J.
Nickoloff
.
2009
.
Skin immune sentinels in health and disease
.
Nat. Rev. Immunol.
9
:
679
691
.
Pittelkow
,
M.R.
2005
.
Psoriasis: more than skin deep
.
Nat. Med.
11
:
17
18
.
Racila
,
D.
,
M.
Winter
,
M.
Said
,
A.
Tomanek-Chalkley
,
S.
Wiechert
,
R.L.
Eckert
, and
J.R.
Bickenbach
.
2011
.
Transient expression of OCT4 is sufficient to allow human keratinocytes to change their differentiation pathway
.
Gene Ther.
18
:
294
303
.
Reynolds
,
N.
,
P.
Latos
,
A.
Hynes-Allen
,
R.
Loos
,
D.
Leaford
,
A.
O’Shaughnessy
,
O.
Mosaku
,
J.
Signolet
,
P.
Brennecke
,
T.
Kalkan
, et al
.
2012
.
NuRD suppresses pluripotency gene expression to promote transcriptional heterogeneity and lineage commitment
.
Cell Stem Cell.
10
:
583
594
.
Risse
,
G.
,
K.
Jooss
,
M.
Neuberg
,
H.J.
Brüller
, and
R.
Müller
.
1989
.
Asymmetrical recognition of the palindromic AP1 binding site (TRE) by Fos protein complexes
.
EMBO J.
8
:
3825
3832
.
Rubin
,
A.J.
,
B.C.
Barajas
,
M.
Furlan-Magaril
,
V.
Lopez-Pajares
,
M.R.
Mumbach
,
I.
Howard
,
D.S.
Kim
,
L.D.
Boxer
,
J.
Cairns
,
M.
Spivakov
, et al
.
2017
.
Lineage-specific dynamic and pre-established enhancer-promoter contacts cooperate in terminal differentiation
.
Nat. Genet.
49
:
1522
1528
.
Sano
,
S.
,
K.S.
Chan
,
S.
Carbajal
,
J.
Clifford
,
M.
Peavey
,
K.
Kiguchi
,
S.
Itami
,
B.J.
Nickoloff
, and
J.
DiGiovanni
.
2005
.
Stat3 links activated keratinocytes and immunocytes required for development of psoriasis in a novel transgenic mouse model
.
Nat. Med.
11
:
43
49
.
Segre
,
J.A.
2006
.
Epidermal barrier formation and recovery in skin disorders
.
J. Clin. Invest.
116
:
1150
1158
.
Sextius
,
P.
,
C.
Marionnet
,
F.X.
Bon
,
A.L.
de La Chapelle
,
C.
Tacheau
,
M.
Lahfa
,
A.
Mauviel
,
B.A.
Bernard
,
J.
Leclaire
,
F.
Bernerd
, and
L.
Dubertret
.
2010
.
Large scale study of epidermal recovery after stratum corneum removal: dynamics of genomic response
.
Exp. Dermatol.
19
:
259
268
.
Shaulian
,
E.
, and
M.
Karin
.
2002
.
AP-1 as a regulator of cell life and death
.
Nat. Cell Biol.
4
:
E131
E136
.
Shen
,
L.
,
N.
Shao
,
X.
Liu
, and
E.
Nestler
.
2014
.
ngs.plot: Quick mining and visualization of next-generation sequencing data by integrating genomic databases
.
BMC Genomics.
15
:
284
.
Takahashi
,
H.
,
M.
Ibe
,
S.
Nakamura
,
A.
Ishida-Yamamoto
,
Y.
Hashimoto
, and
H.
Iizuka
.
2002
.
Extracellular regulated kinase and c-Jun N-terminal kinase are activated in psoriatic involved epidermis
.
J. Dermatol. Sci.
30
:
94
99
.
Tong
,
J.K.
,
C.A.
Hassig
,
G.R.
Schnitzler
,
R.E.
Kingston
, and
S.L.
Schreiber
.
1998
.
Chromatin deacetylation by an ATP-dependent nucleosome remodelling complex
.
Nature.
395
:
917
921
.
Welter
,
J.F.
, and
R.L.
Eckert
.
1995
.
Differential expression of the fos and jun family members c-fos, fosB, Fra-1, Fra-2, c-jun, junB and junD during human epidermal keratinocyte differentiation
.
Oncogene.
11
:
2681
2687
.
Williams
,
C.J.
,
T.
Naito
,
P.G.
Arco
,
J.R.
Seavitt
,
S.M.
Cashman
,
B.
De Souza
,
X.
Qi
,
P.
Keables
,
U.H.
Von Andrian
, and
K.
Georgopoulos
.
2004
.
The chromatin remodeler Mi-2β is required for CD4 expression and T cell development
.
Immunity.
20
:
719
733
.
Wurm
,
S.
,
J.
Zhang
,
J.
Guinea-Viniegra
,
F.
García
,
J.
Muñoz
,
L.
Bakiri
,
E.
Ezhkova
, and
E.F.
Wagner
.
2015
.
Terminal epidermal differentiation is regulated by the interaction of Fra-2/AP-1 with Ezh2 and ERK1/2
.
Genes Dev.
29
:
144
156
.
Xue
,
Y.
,
J.
Wong
,
G.T.
Moreno
,
M.K.
Young
,
J.
Côté
, and
W.
Wang
.
1998
.
NURD, a novel complex with both ATP-dependent chromatin-remodeling and histone deacetylase activities
.
Mol. Cell.
2
:
851
861
.
Yamada
,
T.
,
Y.
Yang
,
M.
Hemberg
,
T.
Yoshida
,
H.Y.
Cho
,
J.P.
Murphy
,
D.
Fioravante
,
W.G.
Regehr
,
S.P.
Gygi
,
K.
Georgopoulos
, and
A.
Bonni
.
2014
.
Promoter decommissioning by the NuRD chromatin remodeling complex triggers synaptic connectivity in the mammalian brain
.
Neuron.
83
:
122
134
.
Yoshida
,
T.
,
I.
Hazan
,
J.
Zhang
,
S.Y.
Ng
,
T.
Naito
,
H.J.
Snippert
,
E.J.
Heller
,
X.
Qi
,
L.N.
Lawton
,
C.J.
Williams
, and
K.
Georgopoulos
.
2008
.
The role of the chromatin remodeler Mi-2β in hematopoietic stem cell self-renewal and multilineage differentiation
.
Genes Dev.
22
:
1174
1189
.
Yoshida
,
T.
,
Y.
Hu
,
Z.
Zhang
,
A.O.
Emmanuel
,
K.
Galani
,
B.
Muhire
,
H.J.
Snippert
,
C.J.
Williams
,
M.Y.
Tolstorukov
,
F.
Gounari
, and
K.
Georgopoulos
.
2019
.
Chromatin restriction by the nucleosome remodeler Mi-2β and functional interplay with lineage-specific transcription regulators control B-cell differentiation
.
Genes Dev.
33
:
763
781
.
Zenz
,
R.
,
R.
Eferl
,
L.
Kenner
,
L.
Florin
,
L.
Hummerich
,
D.
Mehic
,
H.
Scheuch
,
P.
Angel
,
E.
Tschachler
, and
E.F.
Wagner
.
2005
.
Psoriasis-like skin disease and arthritis caused by inducible epidermal deletion of Jun proteins
.
Nature.
437
:
369
375
.
Zhang
,
Y.
,
G.
LeRoy
,
H.P.
Seelig
,
W.S.
Lane
, and
D.
Reinberg
.
1998
.
The dermatomyositis-specific autoantigen Mi2 is a component of a complex containing histone deacetylase and nucleosome remodeling activities
.
Cell.
95
:
279
289
.
Zhang
,
Y.
,
T.
Liu
,
C.A.
Meyer
,
J.
Eeckhoute
,
D.S.
Johnson
,
B.E.
Bernstein
,
C.
Nusbaum
,
R.M.
Myers
,
M.
Brown
,
W.
Li
, and
X.S.
Liu
.
2008
.
Model-based analysis of ChIP-Seq (MACS)
.
Genome Biol.
9
:
R137
.
Zhang
,
J.
,
A.F.
Jackson
,
T.
Naito
,
M.
Dose
,
J.
Seavitt
,
F.
Liu
,
E.J.
Heller
,
M.
Kashiwagi
,
T.
Yoshida
,
F.
Gounari
, et al
.
2011
.
Harnessing of the nucleosome-remodeling-deacetylase complex controls lymphocyte development and prevents leukemogenesis
.
Nat. Immunol.
13
:
86
94
.
This article is distributed under the terms of an Attribution–Noncommercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.rupress.org/terms/). After six months it is available under a Creative Commons License (Attribution–Noncommercial–Share Alike 4.0 International license, as described at https://creativecommons.org/licenses/by-nc-sa/4.0/).

Supplementary data