The epidermis covering our skin and body maintains a vital and multidimensional barrier to water and the outside environment while renewing itself with rapid kinetics, even under normal physiological conditions (Kubo et al., 2012). The mechanisms through which new progenitor cells are produced at the base of this stratified epithelium, pace themselves through differentiation, and maintain tissue architecture and function in spite of a high rate of cell loss at the skin surface are only partially understood (Wells and Watt, 2018).

Keratin intermediate filaments are major protein constituents in epithelial cells and are encoded by a large family of 54 conserved genes that are individually regulated in a tissue- and differentiation-specific fashion (Schweizer et al., 2006). An outstanding question is the extent to which keratin, and other types of intermediate filaments (IFs), participate in basic processes such as cell differentiation and tissue homeostasis. The type I keratin 14 (K14) and type II K5 co-polymerize to form the prominent IF apparatus that occurs in the progenitor basal layer of epidermis and related complex epithelia (Nelson and Sun, 1983; Fuchs, 1995). Two main roles have so far been ascribed to K5-K14 IFs. First, to provide structural support and mechanical resilience to keratinocytes in the basal layer of epidermis and related epithelia (Coulombe et al., 1991a; Vassar et al., 1991; Fuchs and Coulombe, 1992). Second, to regulate the distribution of melanin with an impact on skin pigmentation and tone (Uttam et al., 1996; Betz et al., 2006; Gu and Coulombe, 2007). Dominantly-acting missense alleles in either KRT5 or KRT14 underlie the vast majority of cases of epidermolysis bullosa simplex (EBS), a rare genetic skin disorder in which trivial trauma results in skin blistering secondary to the lysis of fragile basal keratinocytes (Bonifas et al., 1991; Coulombe et al., 1991b; Fuchs and Coulombe, 1992; Lane et al., 1992). Such mutant alleles may also affect skin pigmentation (Gu and Coulombe, 2007), establishing the relevance of both roles of K5-K14 in both healthy and diseased skin.

Structural insight gained from solving the crystal structure of the interacting 2B regions of corresponding rod domain segments in human K5 and K14 highlighted the presence of a trans-dimer, homotypic disulfide bond involving cysteine (C) residue 367 (C367) in K14 (Coulombe and Lee, 2012). Conspicuously, residue C367 in K14 occurs within a four-residue interruption, or stutter, in the long-range heptad repeat of coil two in the central alpha-helical rod domain in virtually all IF proteins (Lee et al., 2012). We showed that K14 C367-dependent disulfides form in human and mouse skin keratinocytes (Lee et al., 2012), where they play a role in the assembly, organization and steady state dynamics of keratin IFs in live skin keratinocytes (Feng and Coulombe, 2015a; Feng and Coulombe, 2015b). We also showed that loss of the stutter cysteine alters K14’s ability to become part of the dense meshwork of keratin filaments that occurs in the perinuclear space of early differentiating keratinocytes (Lee et al., 2012; Feng and Coulombe, 2015a; Feng and Coulombe, 2015b). However, the physiological significance associated with the surprising properties conferred by a cysteine residue located in a mysterious conserved motif within the central rod domain of a keratin, namely K14, remained unclear.

Here, we report on studies involving a new mouse model that provides evidence that the stutter cysteine in K14 protein regulates entry into differentiation and thus the balance between proliferation and differentiation through regulated interactions with 14-3-3 adaptor proteins and YAP1, a terminal effector of Hippo signaling (Pocaterra et al., 2020). We also discuss evidence that this role likely applies to K10 and other type I keratins expressed in surface epithelia.

The distribution of cysteine residues in mouse K14 protein is schematized in Figure 1A. Codon C367 in KRT14 (human) occurs at position 373 in Krt14 (mouse), and is conserved in the orthologous keratin of several other species (Figure 1B). Moreover, this codon is also conserved in many other type I keratin genes expressed in skin (Strnad et al., 2011; Lee et al., 2012; Figure 1B). To address the physiological significance of the conserved stutter cysteine in K14, we generated Krt14 C373A mutant mice using CRISPR-Cas9 technology (Figure 1C) and verified its presence through allele specific DNA-sequencing (Figure 1D). Krt14 C373A mice are born in the expected mendelian ratio, are viable and fertile, and show a normal body weight when reaching adulthood (Figure 1E). Analysis of total skin proteins from several body sites showed that steady state levels of K14 protein are unaffected in Krt14 C373A relative to WT skin. By contrast, the pattern of K14-dependent, high molecular weight disulfide-bonded species is markedly altered, given fewer species that occur at lower levels (Figure 1F,G). This is so especially in ear and tail skin (Figure 1F,G), prompting us to focus on these two body sites in subsequent analyses. The residual K14-dependent disulfide bonding occurring in Krt14 C373A mutant skin (Figure 1F,G) likely reflects the participation of cysteines located in the N-terminal domain of K14 (see Figure 1A and Feng and Coulombe, 2015a). These findings indicate that mice homozygous for Krt14 C373A allele are viable and appear macroscopically normal, although biochemically they exhibit a strikingly altered pattern of K14-dependent disulfide bonding, particularly in ear and tail skin.

The histology and barrier status of young adult Krt14 C373A mouse skin were analyzed next. By histology, the epidermis of Krt14 C373A mice is modestly but significantly thickened relative to WT in ear and tail skin (Figure 2A,B). Measurement of trans-epidermal water loss (TEWL) at the skin surface revealed an increase in Krt14 C373A mice relative to WT control. This is so both at baseline (7.02 ± 0.72 g/m2/h vs. 2.95 ± 0.49 g/m²/h) and after topical acetone application (19.20 ± 1.78 g/m2/h vs. 7.02 ± 0.72 g/m²/h) (Figure 2C), a standard challenge that puts the skin barrier under a mild and reversible stress (Denda et al., 1996). Skin barrier defects often trigger elevated expression of Danger-Associated Molecular Patterns (Lessard et al., 2013) (DAMPs, also known as alarmins). At baseline, DAMPs such as S100a8, S100a9, Mmp9, and Ptgs2 are upregulated by 4-fold or more at the mRNA level in Krt14 C373A skin compared to WT (Figure 2D). This aberrant state is markedly enhanced after a topical acetone challenge to the barrier (Figure 2E). Next, we analyzed cornified envelopes (CEs) isolated from epidermis, given that they are key contributors to skin barrier function (Eckhart et al., 2013). CEs harvested from WT mice appear relatively uniform in size and shape, are mostly oval-shaped, and feature clear and smooth outlines (Figure 2F,G). By contrast, CEs isolated from Krt14 C373A mice are smaller (85% of the area and 87% of the circumference of WT CEs), jagged, and less oval-shaped (aspect ratio of 1.3 compared to 1.1 in WT) (Figure 2F,G and Figure 2—figure supplement 1A,B). Thus, the morphological and molecular anomalies occurring in the epidermis are accompanied by significant defects in barrier function in Krt14 C373A skin.

We next assessed keratinocyte proliferation, transit time, and apoptosis in order to identify possible causes for the increased thickness and barrier defect in Krt14 C373A epidermis. At 2 hr after a single pulse of the nucleotide analog Edu (Chehrehasa et al., 2009), a significantly greater fraction of keratinocytes are labeled in the basal layer of Krt14 C373A epidermis compared to WT (by ~1.6 fold; p=0.024) (Figure 3A,B), indicating that keratinocyte proliferation is enhanced at baseline in mutant mice. Following a 1 day chase after the Edu pulse, this difference is accentuated (>2 fold; p=0.01) and Edu-labeled nuclei now occur in the suprabasal layers of epidermis in both genotypes, reflecting keratinocyte exit from the basal layer (Figure 3A,B). Following a 3-day chase after the Edu pulse, nuclear labeling remains high and stable in the basal layer of epidermis in both genotypes, but a clear additional difference emerges as there are significantly more Edu-labeled nuclei in the suprabasal layers of mutant epidermis. At the 7-day mark, the fraction of labeled cells in the basal layer has subsided in both genotypes but, again, the suprabasal epidermis of Krt14 C373A skin shows far more labeled nuclei (Figure 3A,B). This pulse-chase experiment shows that keratinocytes in Krt14 C373A epidermis show enhanced proliferation confined to the basal layer at baseline, and that keratinocytes exhibit a faster pace of movement across the suprabasal layers as they progress through differentiation. A similar phenotype has been previously described for Krt10 null mice (Reichelt and Magin, 2002). We also observed a greater frequency (~3 fold) of TUNEL-positive nuclei in Krt14 C373A epidermis compared to WT, with apoptotic cell death confined to the suprabasal compartment (Figure 3C,D). We also examined expression of p63, given its role as master regulator of epidermal stratification and differentiation (Soares and Zhou, 2018). A distinct immunostaining pattern was observed in Krt14 C373A epidermis relative to WT (Figure 3E). Upon quantitation, significant differences prevailed in terms of frequency of keratinocytes labeled and their distance from the basal lamina (Figure 3F), with p63-positive staining showing a conspicuous elongated shape and extending higher up in the mutant epidermis. When combined, these findings suggest that the modest increase observed in epidermal thickness (Figure 1) masks a more pronounced defect in epidermal homeostasis under baseline conditions in Krt14 C373A mice.

We next assessed markers relevant to keratinocyte proliferation to identify possible causes for the defective barrier of Krt14 C373A skin relative to WT. We examined the distribution of K14 (basal cell layer), K10 (early differentiation), filaggrin and loricrin (late differentiation) in tail skin sections from young adult mice. The staining for filaggrin and loricrin were markedly decreased (~62% and 48% reductions, respectively) while the staining for K10 was modestly decreased (~37% reduction) in Krt14 C373A epidermis relative to WT (Figure 3G,H). We also examined markers of adherens junction (E-cadherin), desmosomes (desmoplakin), tight junctions (claudin 3) since epidermal differentiation entails a tightly coordinated rearrangement of intercellular junctions. Claudin 3 staining was decreased by ~33% in Krt14 C373A epidermis, consistent with the barrier defect. The signals for E-cadherin and desmoplakin appeared slightly increased (Figure 3I), an occurrence that may reflect the modest epidermal thicknening (Figure 3J). In contrast to tail and ear epidermis, several markers including K14, K10, loricrin and filaggrin appear normal in the thin epidermis of back skin (Figure 3—figure supplement 1A–C), consistent with the markedly lower yield of K14-dependent disulfide bonding in this body site (Figure 1). Together these observations link the anomalies observed in epidermal homeostasis and skin barrier to defects in terminal keratinocyte differentiation in Krt14 C373A mouse skin.

We previously showed that replacing Cys with Ala at position 367 in human K14 does not abrogate 10 nm filament formation but leads to a reduction in the perinuclear clustering of keratin filaments in cultured keratinocytes (Feng and Coulombe, 2015b). Transmission electron microscopy of epoxy-embedded skin tissue sections was next used to assess whether similar changes occur in vivo. In basal keratinocytes of WT epidermis, keratin IFs typically occur as bundles near the nucleus. In Krt14 C373A basal keratinocytes, however, keratin IFs are absent from the perinuclear region and appear redistributed towards the cell periphery (Figure 3—figure supplement 2A,B). Consistent with the macroscopic appearance of skin tissue there is no ultrastructural evidence of cell fragility in Krt14 C373A epidermis (Figure 3—figure supplement 2A,B and data not shown). We find that nuclei feature a more ellipsoid shape along with a greater frequency of cytoplasmic invaginations (by ~1.4 fold in basal keratinocytes and by ~1.7 fold in suprabasal keratinocytes, respectively) compared to WT controls; Figure 3—figure supplement 2C). The occurrence of ultrastructural anomalies in the perinuclear keratin IF network in Krt14 C373A basal keratinocytes extend previous live imaging observations (Feng and Coulombe, 2015b) and point to the possibility that the mechanical properties of the nuclear envelope or nucleus are altered in these cells.

To identify potential pathways regulated by K14-dependent disulfides, we performed K14 co-immunoprecipitation (co-IP) assays followed by mass spectrometry (MS) analysis in protein extracts prepared from newborn WT keratinocytes in primary culture in the presence of 1 mM Ca²⁺, a condition that induces keratinocyte differentiation (Hennings et al., 1980). This screen identified 14-3-3σ and other 14-4-3 isoforms as major interacting partners for K14 in WT cell cultures (Figure 4A and Supplementary file 1). There is a strong precedent for interactions between 14-3-3 proteins and keratins, including K18 (Liao and Omary, 1996; Ku et al., 1998), K17 (Kim et al., 2006), which occur in a phosphorylation-dependent fashion. Co-immunoprecipitation assays confirmed that transfected, HA tagged-14-3-3σ physically interact with endogenous WT K14 and Krt14373A mutant protein in mouse keratinocytes in primary culture (Figure 4B and data not shown). Further inspection of the top 100 MS-identified proteins in this targeted proteomics screen (Supplementary file 1) reveals, as expected, the presence of desmosomal proteins, known keratin-interacting proteins (e.g., annexins; (Chung et al., 2012), and several proteins with known roles in organelle transport and organization (e.g., rab family members; Ohbayashi and Fukuda, 2012), which is consistent with K5-K14’s established role in skin pigmentation (Gu and Coulombe, 2007).

14-3-3σ was deemed of interest because it regulates the proliferation and differentiation of keratinocytes in epidermis (Herron et al., 2005; Li et al., 2005). The latter is achieved in part by modulating the cellular localization of YAP (Li et al., 2005; Sun et al., 2015), a terminal effector of Hippo signaling (Schlegelmilch et al., 2011; Silvis et al., 2011; Sambandam et al., 2015). Hippo is an evolutionary conserved pathway with a primary role in regulating growth and homeostasis in organs and tissues (Pocaterra et al., 2020). We next assessed the distribution of 14-3-3σ and YAP using indirect immunofluorescence of tissue sections prepared from WT and Krt14 C373A tail skin. 14-3-3σ occurs mostly as aggregates in suprabasal keratinocytes of Krt14 C373A epidermis, which is in striking contrast to the diffuse distribution observed in WT controls (Figure 4C). Consistent with previous reports (Schlegelmilch et al., 2011; Sambandam et al., 2015), a strong signal for YAP occurs in both the nucleus and cytoplasm in basal keratinocytes, and otherwise YAP occurs as a weaker and diffuse signal in the cytoplasm (but is not seen in the nucleus) of suprabasal keratinocytes in WT epidermis (Figure 4D). In Krt14 C373A epidermis, strikingly, YAP localizes preferentially to nuclei in both basal and suprabasal keratinocytes, in a consistent fashion (Figure 4D). The latter finding suggests that YAP-dependent gene expression may be altered in mutant mouse skin. Follow-up RT-qPCR assays show that the steady state levels for several known YAP target gene mRNAs, including Cyr61, Zeb1, Ctgf and Snail2, are markedly elevated in Krt14 C373A relative to WT skin (Figure 4E). By western immunoblotting, the levels of endogenous YAP1 and Ser^127-phosphorylated YAP1 are similar in WT and Krt14 C373A skin (Figure 4F,G). Together these findings point to a misregulation of 14-3-3σ and YAP as likely contributors to the epidermal phenotype exhibited in the ear and tail skin of Krt14 C373A mice.

Next we asked whether the misregulation of YAP subcellular partitioning also occurs in primary culture. Keratinocytes were isolated from WT and Krt14 C373A newborn pups, cultured in the absence or presence of calcium (Hennings et al., 1980), and analyzed using microscopy-based readouts. K14-dependent disulfide bonding is low in the absence of calcium and rises of the course of days after adding calcium to primary cultures of WT mouse keratinocytes (Figure 4—figure supplement 1A,B). In the absence of calcium, the staining for YAP is concentrated in the nucleus in both WT and Krt14 C373A keratinocytes (Figure 5A,B). After addition of calcium (1 mM), which triggers differentiation and mimics a suprabasal state (Hennings et al., 1980), 73% of WT keratinocytes lose their nuclear YAP signal whereas 95% of Krt14 C373A keratinocytes exhibit predominantly nuclear YAP (Figure 5A,B). Western immunoblotting confirmed that, as expected, both K10 and filaggrin proteins occur at lower levels in calcium-treated primary cultures of Krt14 C373A relative to WT keratinocytes (Figure 4—figure supplement 1C,D), suggesting a delay or a defect in terminal differentiation. Moreover, PLA assays yielded evidence for decreased interactions between 14-3-3σ and YAP in Krt14 C373A keratinocytes in primary culture, relative to WT controls (Figure 4—figure supplement 1E,F), thereby extending the immunoprecipitation findings shown in Figure 4B. We conclude that the abnormal retention of YAP to the nucleus and abnormal differentiation of Krt14 C373A keratinocytes are both preserved outside of the skin tissue setting, further suggesting that these properties are linked and inherent to keratinocytes.

The five cysteine residues present in human K14 are conserved in the mouse ortholog (Lee et al., 2012), and cysteines at positions 4, 40, and 367 in human K14 participate in disulfide bonding (Feng and Coulombe, 2015a). We next asked whether the aberrant YAP localization in Krt14 C373A mutant epidermis is Cys373-specific. To this end we devised a rescue assay using Krt14 null mouse keratinocytes in primary culture, transfection of GFP-K14 or mutants thereof, and analysis of YAP subcellular partitioning. Consistent with previous findings (Sambandam et al., 2015), only 19.5% of GFP-K14WT-expressing keratinocytes exhibit nuclear YAP in the presence of calcium (Figure 5—figure supplement 1A,B). By contrast, cells expressing either a GFP-K14 cysteine free (CF) mutant, or a GFP-C367A single mutant, feature an abnormally high nuclear retention of YAP (83.8% and 81.9%, respectively) (Figure 5—figure supplement 1A,B). Restoring Cys367 in the K14-CF backbone (GFP-K14CF-C367 construct) rescued the abnormal nuclear retention of YAP, given that only 27.3% of transfected cells show YAP in the nucleus (Figure 5—figure supplement 1A,B). These findings directly implicate the stutter cysteine (C367 in human K14) as a calcium-dependent regulator of the subcellular partitioning of YAP in skin keratinocytes.

In an effort to directly relate K14 to the activity of YAP1 as a transcriptional regulator, we next conducted luciferase reporter assays in a heterologous cell culture setting. HeLa cells were selected because they are epithelia-derived, and grow and transfect well. HeLa cells normally show low levels of K14 expression (Moll et al., 1982) but respond well to K5-K14 co-transfection (Figure 5—figure supplement 1C). Transfection of a Ccn1 (Cyr61) gene promoter-driven Firefly luciferase plasmid, previously shown to effectively report on YAP transcriptional activity (Ma et al., 2017), led to a strong (~8 fold) induction of luciferase activity normalized to cells transfected with a reference Renilla luciferase plasmid (Figure 5C). Consistent with K14’s ability to retain YAP in the cytoplasm, Ccn1 promoter activity was significantly attenuated by co-expression of WT K14 along with assembly partner WT K5 (Figure 5C). In striking contrast, activity of the Ccn1 promoter construct was much less affected when cysteine-free K14 (K14 CF) was co-expressed with WT K5 (Figure 5—figure supplement 1C). Of note, the organization of filaments in HeLa cells co-expressing either K5 and K14WT or K5 and K14CF is comparable (Figure 5—figure supplement 1C). The findings from this heterologous assay substantiate the notion that K14-containing filaments impact YAP activity as predicted, and that this property depends on the presence of Cys residues in K14.

YAP is a key effector of mechanosensing and mechanotransduction (Dupont et al., 2011; Benham-Pyle et al., 2015; Panciera et al., 2017). Cells experiencing tension often respond by enhancing determinants such as F-actin stress fibers, acto-myosin contraction, the recruitment of α-catenin and vinculin to adherens junctions (Leckband and de Rooij, 2014; Yap et al., 2018), expression of lamin A/C in the nucleus (Swift et al., 2013), and frequency of binucleation (Cao et al., 2017). Monitoring the status of such elements provides a test for altered mechanosensing or mechanotransduction. In tail skin tissue sections, the immunostaining for α-catenin and for lamin A/C but not that for desmoglein 1 are markedly increased in Krt14 C373A mice relative to WT controls (Figure 6A,B). Use of the a-18 antibody that recognizes a mechanosensitive epitope on α-catenin (Shimoyama et al., 1992) confirms that epidermal keratinocytes are under altered tensile stress in Krt14 C373A epidermis (Figure 6C). In the setting of primary culture, we observed a greater incidence of multi-nucleated keratinocytes in Ca²⁺-treated Krt14 C373A newborn skin keratinocytes compared to WT control (14.3% versus 2.8%; Figure 6D). We also find that the immunofluorescence signal for α-catenin, lamin A/C, F-actin stress fibers and phosphorylated myosin light chain II (pMLC Ser19) are each increased in Krt14 C373A relative to WT (Figure 6E,F). Collectively these findings strongly suggest that, consistent with YAP misregulation (see Figure 5), Krt14 C373A mutant keratinocytes show alterations in mechanosensing and/or mechanotransduction relative to WT.

Our study establishes that residue cysteine 373 in mouse K14 partakes in regulating the balance between keratinocyte proliferation and differentiation in epidermis in vivo, and ultimately barrier function, in skin. The loss of cysteine 373 results in profound alterations in the i) pattern of K14-dependent disulfide bonding in epidermis; ii) regulation of 14-3-3sigma and YAP in early-stage differentiating keratinocytes; and iii) several mechanosensitive readouts in the young adult epidermis in situ and newborn skin keratinocytes in primary culture. Our findings support the conclusion that K14-dependent disulfide bonding involving the conserved ‘stutter cysteine’ residue impacts cell architecture, mechanosensing and Hippo signaling at an early stage of epithelial differentiation in the epidermis.

A model that conveys the significance our findings is given in Figure 7A. Consistent with the literature, the model posits that Hippo signaling is inactive in most keratinocytes in the basal layer of epidermis (Schlegelmilch et al., 2011; Beverdam et al., 2013; Sambandam et al., 2015; Totaro et al., 2017), with YAP localizing to the nucleus given a specific level of cellular crowding and/or integrin-mediated adhesion to the extracellular matrix (Panciera et al., 2017; Elbediwy and Thompson, 2018). In basal keratinocytes, K5-K14 filaments are organized in loose bundles that run alongside the nucleus (Coulombe et al., 1989; Lee et al., 2012) while 14-3-3σ occurs at low levels (Dellambra et al., 2000; Reichelt and Magin, 2002; Kim et al., 2006). The first prediction of our model is that reception of differentiation-promoting cues triggers K14-dependent disulfide bonding and creates binding sites for 14-3-3 on K5-K14 filaments. The latter likely occurs via site-specific phosphorylation, on K14 (e.g., Inaba et al., 2018). These events foster the known reorganization of keratin filaments into a prominent perinuclear network (Lee et al., 2012), along with the binding and sequestration of YAP1 to the cytoplasm, thus activating Hippo signaling as keratinocytes initiate terminal differentiation (Figure 7A). Several reports converged in establishing a role for mitochondria-derived reactive oxygen species as a key trigger towards the initiation of terminal differentiation in keratinocytes of epidermis (Tamiji et al., 2005; Bause et al., 2013; Hamanaka et al., 2013; Sun et al., 2015). Such species could non-enzymatically mediate K14-dependent disulfide bonding in the perinuclear cytoplasm of keratinocytes (Suzuki et al., 2019).

Many type I keratins expressed in epidermis (e.g., K10, K16, K17) feature a cysteine residue at the location corresponding to codon 373 in mouse K14 and codon 367 in human K14 (Lee et al., 2012). Four lines of evidence support the contention that K10, in particular, is a strong candidate for K14-like regulation of YAP1 subcellular partitioning and Hippo signaling in the epidermis. First, KRT10 (human) and Krt10 (mouse) expression is turned on at the earliest stage of terminal differentiation (Woodcock-Mitchell et al., 1982; Schweizer et al., 1984), and the K10 protein features a structurally and biochemically equivalent Cys residue at position 401 and is capable of interacting with 14-3-3 (Wilker et al., 2007; Huang et al., 2010). Second, the crystal structure of the interacting 2B segments of K1-K10 (Bunick and Milstone, 2017) shows an overall fold identical to our original K5(2B)-K14(2B) structure (Lee et al., 2012), including the presence of a trans-dimer, homotypic disulfide bond mediated by the stutter cysteine (C401) in K10. Third, we previously showed that K10 partakes in the formation of disulfide-dependent, dimer-sized species in skin keratinocytes (Lee et al., 2012). Fourth the Krt10 null mouse, described >15 years ago (Reichelt and Magin, 2002; Reichelt et al., 2004), exhibits a strong phenotype consisting of hyperproliferation, faster keratinocyte transit time, and impaired differentiation in the epidermis which, molecularly, correlates with a marked upregulation of 14-3-3sigma and c-Myc. MYC has since then been shown to be a bona fide YAP1 target gene (Schütte et al., 2014; Kim et al., 2017; Cai et al., 2018). Accordingly, our model (Figure 7A) predicts that the newly-defined role for K14 in regulating the subcellular partitioning of YAP and onset of keratinocyte differentiation in epidermis is picked up and maintained by K10 as differentiation proceeds.

Our findings extend previous reports linking 14-3-3σ to the regulation of YAP during terminal differentiation in epidermis (Sambandam et al., 2015; Sun et al., 2015). They suggest that the network of keratin filaments proximal to the nucleus acts as a key docking site for 14-3-3/YAP complexes during keratinocyte delamination and differentiation, taking on a role that has been held to this point by integrin-based adhesion sites (Elbediwy et al., 2016) and adherens junctions (via alpha-catenin; (Schlegelmilch et al., 2011; Silvis et al., 2011). In addition, a disulfide bonding-rich perinuclear network of keratin filaments could also afford protection to the nucleus and the genome during delamination (Lee et al., 2012), thus extending the known role of K5-K14 IFs in providing mechanical support in basal keratinocytes (Coulombe et al., 1991a; Ramms et al., 2013). Besides, keratinocyte delamination and differentiation in epidermis together entail a profound reorganization of cell-cell and cell-matrix adhesion, and of F-actin and microtubule organization (Sumigray and Lechler, 2015; Muroyama and Lechler, 2017; Miroshnikova et al., 2018; Rübsam et al., 2018; Wickström and Niessen, 2018). How all of these effectors and their inputs are integrated to result in a redistribution of intracellular tension and/or compressive forces (Miroshnikova et al., 2018; Rübsam et al., 2018), along with proper regulation of YAP function and Hippo signaling, awaits further investigation.

YAP’s role as a transcription factor requires its binding to TEAD protein, which is itself stably bound to the promoter of its target genes in a sequence-specific fashion in the nucleus (Vassilev et al., 2001). Data available through ENCODE for human foreskin keratinocytes in primary culture in the absence or presence of calcium (see Materials and methods) provides insight into the rationale for keratin protein involvement towards the regulation of YAP-driven transcription and Hippo signaling. The combination of DNAse one hypersensitivity mapping, ATAC-seq mapping, and presence of TEAD binding sites (in either orientation) on DNA (5’- CATTCC-3’; Heinz et al., 2010) suggests that keratin genes expressed in dividing basal cells of epidermis, for example the type I Krt14 and Krt15 and type II Krt5, are likely YAP target genes given the presence of consensus-matching TEAD binding sites occurring in open chromatin regions in the absence of calcium (Figure 7—figure supplement 1; Supplementary file 2). ENCODE data convey that the chromatin surrounding these genes becomes non-accessible in the presence of calcium (Figure 7—figure supplement 1 and 2), which promotes differentiation (Hennings et al., 1980). Keratin genes expressed in differentiating keratinocytes of epidermis, including the type I Krt10 and type Krt1 and Krt2, do not feature proximal TEAD binding sites (Figure 7—figure supplement 1; Supplementary file 2) and thus are not likely to be transcribed in a TEAD/YAP-dependent fashion in epidermis. The YAP1 and TEAD genes (human) are themselves poised to be positively regulated by YAP in progenitor keratinocytes (Figure 7—figure supplement 1; Supplementary file 2). SFN, ITGB1, CCN1 and TP63 (human) are among the additional genes of interest that we included in this analysis (Supplementary file 2). Accordingly, the YAP/TEAD axis may be set up to promote unabated cell proliferation in a fast-renewing epithelium such as epidermis until two post-translational events converging on K14 protein, phosphorylation and disulfide bonding, would lead to YAP protein sequestration in the cytoplasm and inactivation of the YAP/TEAD axis as part of a regulated negative feedback loop, enacted at the time of entry into differentiation in epidermis (Figure 7B).

Several mechanisms could account for the functional interplay between K14-dependent disulfide bonding, 14-3-3σ, and the regulation of YAP’s subcellular partitioning in keratinocytes. First, the perinuclear enrichment of keratin IFs, which is promoted by K14-dependent disulfide bonding, is poised to increase the local concentration of binding sites for 14-3-3σ and YAP near the nucleus (simple mass action law). Second, the occurrence of K14-dependent disulfides may create an optimal binding interface for 14-3-3σ and YAP in the cytoplasm residing proximal to the nucleus. Third, the nucleus is known to function as a mechanosensor (Cao et al., 2017), and local forces impacted by the perinuclear network of keratin filaments could alter the mechanical gating of YAP across nuclear pores (Elosegui-Artola et al., 2017). These three mechanisms, and possibly others, could act in combination. How K14 (and possibly K10), 14-3-3σ, YAP, and other crucial effectors bind each other as part of this newly defined signaling axis, its regulation, and its significance have now emerged as open issues of high significance for future studies.