A Presenilin-2–ARF4 trafficking axis modulates Notch signaling during epidermal differentiation

One fundamental question in developmental biology is how an individual cell may sense its environment to transmit extracellular signals that control cell signaling and proliferation during tissue morphogenesis. Once thought merely a vestigial structure, the primary cilium is now well established as a cell-sensory organelle that coordinates signal transduction pathways (Berbari et al., 2009). Although cilia have been most prominently linked to Sonic Hedgehog (SHH) signaling, their appreciation as cellular “antennae” that sense a wide variety of external signals likely explains why ciliary defects contribute to diverse human disorders and diseases, such as polydactyly, neural tube defects, Bardet–Biedl syndrome, retinal degeneration, polycystic kidney disease, and skin cancer (Badano et al., 2006; Satir and Christensen, 2007).

In response to external environmental cues during skin embryogenesis, ciliated epithelial progenitors within a single (basal) layer either stratify and differentiate to generate the epidermis or invaginate to make the buds that will develop into hair follicles (HFs; Fuchs, 2007; Ezratty et al., 2011). Hair bud formation requires Wnt and Shh signaling (Ouspenskaia et al., 2016), and given the cilium’s prominent role in Shh signaling, it is not surprising that primary cilia have a role in HF development (Ezratty et al., 2011). However, defects in ciliogenesis also cause a temporally and spatially distinct perturbations in epidermal differentiation (Croyle et al., 2011; Ezratty et al., 2011), a process thought to occur independently of Shh signaling (Mill et al., 2005), but require Notch-signaling (Rangarajan et al., 2001; Lefort and Dotto, 2004; Blanpain et al., 2006). The role of the cilium in this latter process remains poorly understood.

Notch signaling is activated when one of four (Notch 1–4) Notch receptors engages with Delta or Jagged ligands, typically presented on an adjacent neighboring cell. Upon ligand activation, Notch receptors are cleaved in a cascade of proteolytic events, culminating in Presenilin-mediated enzymatic cleavage and subsequent release of the Notch intracellular domain (NICD). NICD then translocates to the nucleus and associates with the DNA-binding protein RBPj to activate downstream target genes that are required for differentiation (Kopan, 2012; Hori et al., 2013). When Rbpj is conditionally ablated in the basal layer of embryonic epidermis, Notch signaling is abrogated and epidermal differentiation is impaired, but cilia are unaffected (Blanpain et al., 2006; Ezratty et al., 2011). This places ciliogenesis upstream of Notch signaling in embryonic skin.

When Ift88 is ablated postnatally in skin, the epidermis displays hyperproliferation and discontinuous keratin 1 (K1), suggestive of suppressed terminal differentiation (Croyle et al., 2011). Similarly, when Ift88 or several other Ift mRNAs are knocked down in embryonic skin by in utero, epidermal-specific delivery of lentiviruses harboring one of several different ciliary hairpin shRNAs, the epidermis displays hyperproliferation and diminished differentiation (Ezratty et al., 2011). Moreover, Ift mutant embryonic skin was accompanied by a reduction in canonical Notch reporter activity and nuclear pathway members NICD and HES1 (Ezratty et al., 2011). Given that ciliogenesis occurs before and independently of canonical Notch signaling, and epidermal ciliary mutants are defective in Notch-dependent epidermal differentiation, we became curious as to whether primary cilia may play a context-specific role in spatially and/or temporally regulating aspects of Notch signaling during embryogenesis.

In the present study, we sought to test the hypothesis that the intrinsic Notch signaling defects that we observe in ciliary mutants might be attributable to disrupted polarized trafficking resulting from ciliary loss. To dissect the molecular mechanisms by which cilia may influence the regulation of Notch signaling, we complemented in vitro culturing of epidermal keratinocytes with our powerful in utero lentiviral delivery method to selectively and efficiently transduce the skin epithelium of embryonic day (E) 9.5 embryos at a stage when it exists as a single-layer of progenitors (Beronja et al., 2010). Using an array of complex genetics, cell biology, and site-directed mutagenesis in developing embryonic epidermis, we demonstrate that Notch-processing enzyme Presenilin-2 targets to the basal body (centrosome), and this localization is required for proper differentiation and optimal activation of Notch signal transduction. We next examine a subset of putative cilia and ciliopathy-related proteins implicated in regulating either Notch signaling or trafficking to primary cilia. Our findings implicate the small GTPase ARF4, which functions in polarized exocytosis (Deretic and Wang, 2012). When ARF4 is absent, cilia are still present, but Presenilins no longer localize to basal bodies and Notch signaling and epidermal differentiation is perturbed.

We began by examining Notch activity in detail during embryonic skin development and epidermal differentiation. We used the GFP-transgenic Notch reporter (GFP-TNR) mouse, which has been reported to faithfully recapitulate Notch signaling in the developing nervous system (Gaiano et al., 2000). Confocal microscopy of whole-mount (WM) epidermis showed that GFP-TNR activity was only robust in a subset of epidermal cells within the planes of the first and second suprabasal (spinous) layers of developing E17.5 epidermal tissue (Fig. 1 A). By comparison, Notch reporter activity was considerably lower both within the basal epidermal plane and also in the third and more upper suprabasal spinous layer.

GFP-TNR epifluorescence appeared to faithfully recapitulate epidermal Notch signaling, as judged by immunolabeling for NICD, which displayed strong nuclear localization primarily in the first suprabasal layers (Fig. 1 B, sagittal section). Additionally, expression of the Notch ligand Jagged 1 and the downstream transcriptional activator HES1 was restricted to the first several suprabasal layers of the developing skin epidermis (Fig. 1, C and D). Collectively, these data indicated a spatial activation of Notch signaling within the early embryonic epidermis.

While examining Notch activity in the early differentiating layers of developing skin epidermis, we became intrigued that neither the Notch reporter nor immunolabeling for processed Notch (NICD) and the NICD-RBPj target HES1 was uniform within the spinous layer planes of the epithelium (Fig. 1; Blanpain et al., 2006). To explore this further, we examined the status of primary cilia, which can be visualized using antibodies against either acetylated tubulin or ARL13b. As shown by WM imaging in Fig. 2 A, cilia were readily detected in both basal and spinous layers of E17.5 epidermis. Notably, however, some suprabasal cells did not display a primary cilium (see boxed area in Fig. 2 A; magnified in Fig. 2 B). When examined in GFP-TNR embryos, it was readily apparent that those suprabasal cells lacking a visible primary cilium often showed low levels of GFP-TNR activity (Fig. 2 B, cells marked by an asterisk). This became even more apparent in the third and higher suprabasal layers, where cilia were rare. Quantifications indicated that in total, ∼80% of GFP (+) suprabasal cells were ciliated in the first two spinous layers, whereas only 30–40% of GFP (−) cells displayed primary cilia (Fig. 2 C). This differed from the basal layer, where cells were uniformly ciliated with little or no Notch signaling, a feature likely attributed to the largely suprabasal localization of Notch receptors 1–3 in E17.5 epidermis (Blanpain et al., 2006; Williams et al., 2011).

The statistically significant correlation of ciliogenesis with Notch reporter activity suggested that either directly or indirectly, cilia might function in spatially or temporally modulating Notch signaling during the rapid epidermal stratification that occurs at E15.5 to E17.5. In contrast to Shh, which signals long-range to activate cilia-localized Shh receptors and their downstream effectors (GLI family members; Rohatgi et al., 2007; Kim et al., 2009), Notch ligands are on the surface of juxtaposed cells to signal locally to their Notch-receptor–positive neighbors. Notch ligands are expressed basally (Delta and Jagged 2) as well as suprabasally (Jagged 1; Fig. 1 C; Watt, 2002), offering multiple routes for activating Notch signaling in the spinous cells. To this end, we wondered whether primary cilia might be spatially organized within tissue in such a way that cell-cell signaling, as typified by Notch, might be modulated. A priori, given the stratified nature of the developing epithelia, primary cilia could either poke up in between cells or protrude into a neighboring cell. Electron microscopy of sagittal sections of the ciliary axoneme revealed that the primary cilium of one epidermal cell protrudes into another neighboring cell, thereby juxtaposing the ciliary membrane of one cell with the plasma membrane of its neighbor (Fig. 2 D). By increasing the surface area of cell–cell contact, this spatial positioning of primary cilia within stratified differentiating epidermis could enhance cell–cell signaling for any transmembrane receptors that might exist within the primary cilium in the skin epithelium and whose activating ligands exist within the adjacent cell.

Our studies indicate that Notch signaling is enhanced preferentially in individual suprabasal epidermal cells that are ciliated. Next, we examined the more superficial differentiating cells where TNR-GFP signal, NICD, and HES1 expression waned. Interestingly, although Arl13b beautifully labeled the primary cilia in the basal progenitors and cells within the first two to three layers (Fig. 2, A and E), Arl13b immunolabeling became junctional in the most apical region of cells within more superficial layers (Fig. 2 E and Video 1). Intriguingly, this was reminiscent of the centrosomal protein NINEIN, which was reported to relocalize to the intercellular junctions during epidermal differentiation (Lechler and Fuchs, 2007). Presenilin-2 also localized to both centrosomes/basal bodies and cellular junctions in the epidermis, with enhanced junctional immunostaining in more superficial layers (Fig. 2, F and G; and Video 2). Quantification of the data from Figs. 1 and 2 suggests an inverse correlation between the suprabasal cells in which Presenilin-2 and ARL13b localized to intercellular junctions and HES1, indicative of active Notch signaling (Fig. 2 H). Conversely, HES1 was most readily detected in those epidermal layers where ARL13b and Presenilin-2 localized to cilia and basal bodies, respectively. Collectively, these observations suggest that relocalization of proteins from the cilia or basal bodies (in differentiating cells with high Notch signaling) to cellular junctions (in differentiating cells with low Notch signaling) could contribute to the spatial and temporal activation of Notch signaling observed during epidermal development.

Genetic ablation studies have previously established a critical role for the two mouse Presenilins in Notch signaling within the skin epidermis (Pan et al., 2004). Presenilin-2 localization can be recapitulated in in vitro studies, where a shift to high (2 mM) calcium prompts intercellular junction formation and terminal differentiation of cultured primary mouse epidermal keratinocytes (1°MKs). As shown in Fig. 3 A, Presenilin-2 colocalized with γ-tubulin at the centrosome or basal body and, in addition, showed strong labeling at intercellular junctions between keratinocytes. In contrast, lentiviral-mediated shRNA delivery and knockdown (KD) of Psen2, but not control (Scramble), resulted in a complete loss of Presenilin-2 immunostaining, without loss of cilia per se (Fig. 3 B). Additionally, as judged by Western blot analysis, NICD levels were appreciably reduced in Presenilin-2–deficient (Psen2-shRNA KD) keratinocytes (Fig. 3 C). Together, these findings underscored the efficacy of the immunostaining pattern and were in agreement with the prior demonstration of the importance of Presenilins for Notch-signaling in differentiating keratinocytes.

To assess whether basal body localization of Presenilin-2 matters for Notch signaling in the context of epidermal differentiation, we sought to determine how Presenilin-2 localizes to the basal body. In this regard, we were intrigued by the presence of a conserved sequence VRPF within the C-terminal domain of Presenilin-2. Mutations in a similar VxPx motif in Rhodopsin disrupt its localization to the cilia of photoreceptor cells and have been linked to retinitis pigmentosa in humans (Mazelova et al., 2009), and similar mutations in polycystin-2, found in patients with polycystic kidney disease, also result in loss of ciliary localization and function (Ward et al., 2011). Thus, we posited that this sequence in Presenilin-2 might be important for localizing it to basal bodies or cilia during epidermal differentiation.

To test this hypothesis, we designed the Presenilin KD and Presenilin-2 rescue strategy illustrated in Fig. 3 D. We generated two lentiviral (LV) expression vectors. One harbored an shRNA that targets both Psen1 and Psen2 and a PGK-driven H2B-RFP to control for transduction efficiency; Western blot analysis demonstrated that both Presenilin-1 and Presenilin-2 are effectively depleted using this strategy (Fig. 3 E). The second LV harbored a cDNA expression vector encoding either GFP-Presenilin-2 or GFP-P486A-Presenilin-2, in which the putative VRPF trafficking motif was mutated to VRAF (Figs. 3 D and S1).

Cotransductions into 1°MK were achieved by high titer LV infections confirmed by the presence of both H2B-RFP and GFP-Presenilin. As shown in Fig. 3 F, GFP-Presenilin-2 harboring the wild-type (WT) VRPF motif was detected at both cell junctions and at the cytoplasmic basal body, identified by its colocalization with γ-tubulin. In contrast, P486A-GFP-Presenilin-2, mutated in the putative VxPx trafficking motif, localized to cell junctions, but not to basal bodies, as determined by lack of GFP colocalization with γ-tubulin (representative example shown in Fig. 3 G).

Having confirmed the ability to sever the basal body connection of Presenilins, but not their junctional localization, we then turned to testing the physiological consequences of doing so. For this purpose, we coinjected high-titer (>10⁹ colony-forming unit) Psen1,2 KD and rescue LVs into the amniotic sacs of E9.5 mouse embryos. This noninvasive strategy selectively and efficiently infects and transduces the single-layered embryonic skin epithelium, which stably integrates the LV within 24 h and thereafter propagates it to its progeny (Beronja et al., 2010). Embryos were allowed to develop in utero and then analyzed between E15.5 and birth. Embryos typically showed transduction of >85% head skin and ∼60–70% back skin. Because viral DNA stably integrated and embryos grew rapidly, large clonal patches of skin were transduced.

Analysis of E15.5 WM epidermis revealed WT GFP-Presenilin-2 at basal bodies of ARL13b-labeled cilia (Fig. 3 H). In suprabasal cells, WT GFP-Presenilin-2 also showed prominent labeling at cell–cell contacts (Fig. 3, I and J), consistent with the labeling patterns of endogenous protein (Figs. 2 and 3). In contrast, but consistent with our in vitro studies, P486A GFP-Presenilin-2 was only detected at intercellular junctions and not at basal bodies (Fig. 3 J). Quantifications revealed that 50–60% of cells expressing WT GFP-Presenilin-2 showed basal body localization, whereas <20% of cells expressing P486A GFP-Presenilin-2 displayed a basal body-localized GFP signal (Fig. 3 K). These data indicate that the VRPF motif is necessary for localization of PRESENILIN-2 to basal bodies or cilia in both differentiating keratinocytes and developing embryonic epidermis.

To determine whether Presenilin-2 basal body localization is required for efficient activation of Notch signaling in developing epidermis, we evaluated Notch signaling and differentiation markers in Psen1,2-depleted epidermis or epidermis cotransduced with WT or P486A GFP-Presenilin-2. shRNA-mediated depletion of both Presenilins 1 and 2 (Psen1,2) in vivo resulted in a loss of Presenilin immunofluorescence (IF; Fig. 4 A). Additionally, Psen1,2 KD resulted in a marked reduction in expression of terminal differentiation markers K10 and filaggrin, as well as a loss of NICD nuclear immunolabeling (Fig. 4, B–D).

Psen1,2 KD epidermis expressing WT GFP-tagged Presenilin-2 showed a partial rescue of K10 and filaggrin (Fig. 4, A–C, middle). In contrast, the GFP-tagged Presenilin-2 VRAF mutant P486A showed little or no improvement in rescuing the differentiation defects caused by Psen1,2 KD (Fig. 4, A–C, right). Similar results were obtained in cultured primary keratinocytes transduced with Psen1,2 ± WT or mutant GFP-Presenilin-2 (Fig. 4, E and F).

Finally, and in agreement with the results on epidermal differentiation, only the WT GFP-Presenilin-2, and not the P486A GFP-Presenilin-2 mutant, showed rescue of suprabasal nuclear NICD. Thus, as shown by the quantifications in Fig. 4 G, when Psen1,2 were depleted, <10% of H2B-RFP (+) nuclei colocalized with nuclear NICD signal, consistent with the known requirement of both Presenilins in regulating NICD processing in the developing epidermis (Pan et al., 2004). Upon coexpression with WT Presenilin-2 GFP, 25% of H2B-RFP (+) nuclei were NICD immunolabeled, whereas very little improvement in NICD was seen with the P486A presenelin-2 mutant (Fig. 4 G). These data suggest that the VxPx trafficking motif of Presenilin is required for optimal Notch signaling and epidermal differentiation.

Polarized exocytosis has emerged as a primary mechanism for delivering proteins to the primary cilia (Jin et al., 2010). Our observations suggest that a conserved VxPx trafficking motif targets Presenilins to basal bodies or cilia and that this subcellular localization is required for optimal Notch signaling and epidermal differentiation during development.

To determine if known ciliary trafficking mechanisms contribute to Presenilin localization and activation of Notch signaling, we examined a panel of candidate cilia/ciliopathy-related proteins and used shRNA-mediated KD to ascertain their contribution to differentiation and Notch signal transduction in 1°MKs. Upon a shift to high Ca⁺², cells were analyzed for keratinocyte differentiation (K10) or GFP (Notch reporter activity). The data were quantified as shown in Fig. 5 (A and B). As expected from our previous study (Ezratty et al., 2011), KD of either Ift74 or Ift88 significantly diminished not only K10, but also Notch reporter expression in most transduced cells. Pin1 KD also showed a marked inhibition of both K10 and Notch reporter expression. We included Pin1, encoding a prolyl isomerase, because it had surfaced in a proteomic analysis of primary cilia (Ishikawa et al., 2012) and also was previously shown to stabilize and/or enhance Notch activity (Rustighi et al., 2009; Baik et al., 2015). Our focus was on a subset of ARL/ARF family small GTPases (Zhang et al., 2013) that, when knocked down, showed a marked reduction in overall levels of K10 and Notch reporter expression (Fig. 5, A and B). These proteins had not been previously implicated in epidermal differentiation or Notch signaling.

We focused on ARF4, a small GTPase that recognizes VxPx trafficking motifs (Mazelova et al., 2009). We used BBS2 as a control, because the BBSome has also been implicated in trafficking to the basal body, yet Bbs2 shRNAs showed little or no effect on K10 or GFP-TNR expression (Fig. 5, A and B; IF examples in Fig. 5, C and D). As shown in Fig. S2, both Arf4 and Bbs2 shRNAs knocked down their respective proteins as judged by IF and/or immunoblot analyses. These data suggest that the defects in K10 and GFP-TNR expression could be attributable to a specific role for ARF4 trafficking in regulating aspects of Notch signaling rather than a general defect associated with disruption of trafficking via other basal body–associated trafficking modules (such as the BBSome).

To ascertain the effects of ARF4 on epidermal differentiation in vivo, we cloned a second, equally strong shRNA Arf4-2 and scrambled control shRNA into lentiviral H2B-RFP expression vectors and injected them into amniotic sacs of living E9.5 mouse embryos in utero. Although the data shown are for Arf4-2, similar results were obtained with Arf4-3. Consistent with defective Notch signaling, clones of ARF4-2 shRNA epidermis showed reduced K10 (Fig. 5 E). Importantly, K10 was still expressed in adjacent uninfected skin patches, suggesting that the defects in epidermal differentiation were cell-autonomous and not caused by general skin perturbations. Epidermis depleted of ARF4 also showed marked defects in the expression of key terminal differentiation markers filaggrin and loricrin (Fig. 5, F and G), when compared with either scrambled control epidermis (Fig. 5, F and G) or nontransduced littermate controls (not depicted). These data suggest that ARF4 is required for efficient epidermal differentiation during development.

ARF4 and BBS2 are both known to be required for trafficking of ciliary localized proteins and ciliary signaling, but not for ciliogenesis (Follit et al., 2014). Consistent with this notion, differentiating 1°MKs that had been depleted of either ARF4 or BBS2 still generated cilia (Fig. 6 A, representative cilia are shown at higher magnification on right). Cilia were also present when ARF4 was depleted in vivo, as illustrated best by WM IF confocal microscopy (Fig. 6 B, right panel is a magnified view of boxed region from the left panel). Collectively, these findings uncouple the epidermal differentiation defects from ciliogenesis.

Interestingly, most Arf4-2 (or Arf4-3), but not scrambled, transduced embryos displayed polydactyly of both hind and forelimbs (Fig. 6 C). This feature is a classical sign of overactive SHH signaling by loss of GLI3, a transcription factor downstream of SHH that typically functions as a repressor. Notably, however, polydactyly is also seen in double loss-of-function mutations for both Shh and Gli3 (Litingtung et al., 2002), i.e., analogous to what would be expected in ciliogenesis mutants. Based upon these data, ciliary function and epidermal differentiation appeared to be compromised by loss of ARF4 in the skin epithelium, even though the primary cilia were still intact. These data suggest that the differentiation defects we observe in the ARF4-depleted epidermis are not caused by ciliogenesis defects per se, although trafficking within cilia could still be perturbed as suggested by the polydactyly phenotype of Arf4 KD embryos.

ARF4 has been implicated in Golgi to basal body trafficking (Deretic and Wang, 2012), and our mutagenesis studies suggest that localization of Presenilin-2 to basal bodies is important for robust Notch signal transduction and epidermal differentiation. In contrast to a Scrambled control shRNA, where Presenilin-2 was localized to basal bodies, Arf4 KD led to intracellular accumulation of Presenilin-2 (Fig. 7 A). Quantification of Presenilin-2 localization to basal bodies revealed that, as previously shown, 50% of cells had Presenilin-2 localized at basal bodies, but only 10% of Arf4 KD cells showed this localization (Fig. 7 B).

Consistent with a defect in polarized trafficking, immunolocalization also revealed what appeared to be an accumulation of Notch receptor in a perinuclear compartment that colocalized with GM130, a marker for the Golgi apparatus (Fig. 7 C). Quantifications revealed significantly reduced levels of Notch 2 and Golgi IF in control as compared with Arf4 KD keratinocytes. These data suggest that ARF4 may regulate trafficking of Notch components, because aberrant accumulation of Presenilin-2 and Notch 2 were observed upon Arf4 depletion during keratinocyte differentiation. In vivo, Notch 2 also accumulated in what appeared to be a perinuclear localization in suprabasal, but not basal, cells of Arf4 KD epidermis (Fig. 7 E). Expression of the downstream NICD/Rbpj target gene Hes1 was also significantly reduced in the absence of ARF4 function (Fig. 7 F). Collectively, these observations suggest that ARF4-dependent trafficking of Notch components contributes to proper activation of Notch signaling and is required for robust epidermal differentiation during development.

In the present study, we extended the curious links among Notch signaling, cilia, and epidermal differentiation that we and others have observed in mice (Croyle et al., 2011; Ezratty et al., 2011). Although ciliogenesis is downstream of Notch signaling in most tissues, in the developing epidermis, cilia are present as early as E11.5, when skin is a single-layered epithelium, and before, when either Notch signaling or Shh signaling acts in epidermal or HF morphogenesis, respectively (Oro and Higgins, 2003; Ezratty et al., 2011; Williams et al., 2011). Our findings that Notch 2, Notch 3, and Presenilin-2 immunolocalize to primary cilia and/or basal bodies and that ciliogenesis is temporally and spatially associated with Notch reporter activity and differentiation during embryonic skin development have provided a framework with which to probe deeper into the context-specific role of primary cilia and its potential regulation of Notch signal transduction.

Our study revealed a role for the small GTPase ARF4 in epidermal differentiation and Notch signaling both in cultured keratinocytes and in developing embryonic skin. Consistent with other studies, we found that ARF4 is not required for ciliogenesis in keratinocytes (Follit et al., 2014). Rather, it appears to participate in trafficking Notch components from the Golgi to basal bodies or primary cilia. Arf4 knockout mice die at midgestation (E9.5–10.5), shortly before node formation, and display trafficking defects in nonciliated endoderm but normal left–right patterning, a feature dependent on ciliogenesis (Follit et al., 2014). Although the early embryonic lethality of these mice precluded an analysis of skin differentiation and development, it is interesting that ARF4 has been implicated in megalin and fibrocystin trafficking, as both of these proteins undergo intracellular proteolysis, regulated by either γ-secretase (Presenilin) or γ-secretase–like enzymes (Biemesderfer, 2006; Kaimori et al., 2007). Given our findings, we hypothesize that ARF4 may play a general role in regulating the trafficking of membrane receptors that are destined for regulated intracellular proteolysis after their activation at specific subcellular structures. Such a function would be independent of whether the trafficking occurs at cilia or at some other subcellular structure. According to this paradigm, in the context of skin differentiation, the traffic happens to occur at the basal body or primary cilia, where ARF4 would function in localizing γ-secretase and processing Notch receptors. Although more studies will be needed to corroborate these facets, the model serves to conceptualize the cell type–specific differences in ciliary roles that go beyond Shh signaling and may be relevant for understanding the etiology of ciliopathies that cannot be explained solely via loss of Shh signaling.

It is intriguing to speculate that basal bodies/primary cilia might be compartmentalized subcellular domains where regulated intracellular proteolysis occurs in a temporally regulated manner during development. Our observations of Notch reporter activity in developing epidermis suggest that such regulation might occur, and primary cilia are poised to spatially and temporally influence these aspects of Notch signaling.

The proteasome has been localized near centrosomes (Fabunmi et al., 2000), and disruption of basal body proteins, such Bardet–Biedl complex proteins BBS1 and BBS4, compromise proteasomal function that impinges on Wnt signaling (Gerdes et al., 2007). Moreover, such proteins were recently shown to regulate Notch signaling via endosomal trafficking of the Notch receptor in zebrafish (Leitch et al., 2014). This latter study found that loss of BBS1 or BBS4 resulted in increased Notch reporter activity, accompanied by decreased localization of Notch receptor at both primary cilia and the plasma membrane. Consistent with this view, BBS2 loss consistently showed sustained or elevated Notch reporter activity and K10 expression in our studies, whereas depletion of other ciliary or cilia-related proteins consistently suppressed Notch signaling and K10-mediated differentiation in cultured keratinocytes. We did not carefully analyze endosomal trafficking of Notch receptor upon depletion of BBS2, but our results are consistent with the zebrafish studies showing increased Notch activity. Collectively, these data suggest that basal body proteins may have different regulatory functions than proteins found within, or that have access to, the ciliary compartment.

Presenilins have been localized to multiple subcellular compartments, mainly the Golgi and endoplasmic reticulum, but to our knowledge, this is the first study to implicate that localization of a Presenilin to basal bodies may have functional consequences in regulating Notch signaling and epidermal differentiation. During mouse preimplantation development, Presenilins redistribute to centrosomes when the proteasome is inhibited (Jeong et al., 2000), and Förster resonance energy transfer studies of APP and Presenilin-1 suggest close interaction at the centrosome in human neuroglioma cells (Nizzari et al., 2007). Additionally, two different proteomic studies of isolated primary cilia detected Presenilins, Notch receptor, and the Notch regulator PIN1 (Ishikawa et al., 2012; Narita et al., 2012), suggesting that Notch components may be generally present in the cilia/ciliary membrane in at least two different cell types besides keratinocytes. The data further strengthen the notion that Notch components may, in some contexts, be bona fide components of the ciliary proteome.

Our studies manipulating the VxPx trafficking motif of Presenilin-2 lend further support to this notion. The ability of WT, but not P486A mutant, Presenilin-2 to rescue Notch-signaling defects in cells and embryos depleted of both Presenilin-1 and Presenilin-2 underscores the importance of this sequence in targeting to the basal body in the context of epidermal differentiation. One caveat of our studies is that although Presenilin-1 has been detected at cilia in the aforementioned proteomic studies, we have only been able to detect Presenilin-2 at centrosomes or primary cilia using antibodies or GFP-tagged constructs (unpublished data). This VxPx trafficking motif may alternatively regulate other aspects of the structure or function of Presenilin-2 in addition to its subcellular localization, and it will be important to determine whether this motif is a bona fide ciliary targeting sequence or whether this mutation regulates other aspect of Presenilin biology and Notch signal transduction.

Our data support a model where ARF4-dependent exocytosis is required for localization of Presenilin-2 to basal bodies or primary cilia, where γ-secretase has roles in modulating Notch signal transduction, either directly or indirectly. There are likely other levels of regulation of Notch signaling that involve either basal body or primary cilia-dependent mechanisms. One intriguing possibility is that endocytosis of components at the cilia may regulate various aspects of Notch signaling during skin development. The ciliary pocket has been implicated as a “hot spot” for endocytosis (Molla-Herman et al., 2010), and either Notch receptors or perhaps Notch ligands, both of which require endocytosis for proper signaling function, could be selectively or preferentially internalized in this specialized compartment of the plasma membrane.

Another intriguing potential mechanism for Notch regulation at primary cilia involves the newly described release of exosomes from the primary cilia. In Chlamydomonas reinhardtii, the cilium has been shown to additionally function as a secretory organelle via release of protease containing ciliary ectosomes (Wood et al., 2013). These lytic-enzyme containing extracellular membrane vesicles function to liberate daughter cells after mitosis. Because Notch receptors undergo a multiplicity of cleavage events, both extra- and intracellularly, it is intriguing to speculate that ciliary-mediated release of a bioactive factor could regulate signal transduction and cell–cell communication. It is currently unknown whether primary cilia in tissue cells, such as those of the stratified skin epithelia, can signal to adjacent cells. Our studies documenting perturbed Notch signal transduction in the absence of cilia and our observations of cilia poking into adjacent cells suggest that primary cilia may regulate cell–cell communication during development and unveil yet another fascinating potential function of the primary cilia that merits additional study.

CD-1 mice obtained from The Jackson Laboratory were used for in utero lentiviral injections. TNR mice (Mizutani et al., 2007) were obtained from The Jackson Laboratory and outbred to a CD1 background, where they were maintained as homozygotes.

Primary keratinocytes were isolated from WT CD-1 or TNR mice and cultured using previously described methods (Nowak and Fuchs, 2009). HEK293T cells were obtained from Takara Bio Inc. and used for lentiviral production as previously described. For lentiviral infections, keratinocytes were plated at 100,000 cells per six wells and incubated with lentivirus in the presence of 100 μl polybrene. 24–48 h after viral infection, cells were selected with 1–2 mg/ml puromycin. 96 h postinfection, keratinocytes were grown to confluency on fibronectin-coated coverslips and shifted to 2 mM Ca²⁺ growth media as previously described (Ezratty et al., 2011). Coverslips were fixed in either −20 Methanol (to visualize the microtubule cytoskeleton) or 4% PFA before processing for IF. Paraformaldehyde-fixed samples were permeabilized for 5 min in 0.5% Triton X-100. All antibodies were diluted in 10% normal goat serum in PBS.

In utero injection of E9.5 embryos and production of high-titer lentivirus were performed as previously described and are extensively detailed elsewhere (Beronja et al., 2010). Lentiviral shRNAs were provided by the Broad Institute’s Mission TRC-1 Library (Sigma-Aldrich) and for some experiments were cloned from the library vector into a modified pLKO backbone containing H2B-RFP. For the KD panel shown in Fig. 5, five different shRNAs to the proteins indicated from the TRC-1 library were pooled and used to generate lentiviral supernatant for KD. For experiments where individual shRNA were used, the following target sequences were used: Arf4-2, 5′-GCTGTCAAATGAACTTTCAAA-3′; Arf4-3, 5′-CCACTTGTGCTACACAAGGAA-3′.

For Psen1,2 shRNA-mediated targeting both Psen1 and Psen2, the following conserved target sequence was used: 5′-CCTCCCCATCTCCATCACCTT-3′. The following forward and reverse oligos were synthesized (Operon): 5′-CCGGCCTCCCCATCTCCATCACCTTCTCGAGAAGGTGATGGAGATGGGGAGGTTTTTG-3′; 5′-AATTCAAAAACCTCCCCATCTCCATCACCTTCTCGAG AAGGTGATGGAGATGGGGAGG-3′.

Oligos were annealed and cloned into pLKO.1-H2B-RFP using previously described methodology (Williams et al., 2011). Full-length mouse cDNA for Presenilin-2 was obtained from OriGene and directionally cloned into pEGFP-N1 (Takara Bio Inc.) via EcoRI and Xba sites. To generate the WT Presenilin-2 GFP lentiviral construct, H2B-RFP was excised from and Presenilin-2 GFP was ligated into pLKO.1 using previously described methodology (Beronja et al., 2010). This plasmid was used as the backbone to generate the P486 Presenilin-2 GFP construct, and the gene mutation was generated using Quick Change II Site-directed Mutagenesis kit (Agilent Technologies) according to the manufacturers protocol. Positive clones were identified and verified via restriction digest of a region predicted to be unique in the mutated clone. For Psen1,2 KD and presenlin-2 rescue experiments in vitro or in vivo, Psen1,2 H2B-RFP pLKO.1 and/or WT Presenilin-2 GFP or P486 Presenilin-2 GFP high-titer LV particles were generated and coinjected in CD-1 WT E9.5 embryos (see schematic in Fig. 3 C).

For WM immunofluorescence, embryos were fixed in 4% formaldehyde in PBS for 1 h at room temperature and washed extensively in PBS. Backskins were dissected and then permeabilized for 10 min in 0.3% Triton X-100/PBS and blocked in Gelatin Block (2.5% fish gelatin, 2.5% normal donkey serum, 2.5% normal goat serum, 0.5% BSA, 0.1% Triton X-100, and 1× PBS) containing MOM reagent (Vector Laboratories) for 1 h at room temperature. All primary antibodies were used at 1:200 dilution and incubated overnight at 4°C. Primary antibodies were removed by extensive washing in 0.3% Triton X-100/PBS, followed by 1-h room temperature incubation with secondary antibodies.

For IF on sagittal sections, embryos were embedded and frozen in OCT, cryosectioned (10–12 µm), and fixed 10 min in 4% formaldehyde in PBS. Sections were blocked 1 h in 2.5% fish gelatin, 2.5% normal donkey serum, 2.5% normal goat serum, 0.5% BSA, 0.1% Triton X-100, and 1× PBS. The following primary antibodies were incubated for either 2 h at room temperature or overnight at 4°C at the following dilutions: mα acetylated tubulin (1:200; Sigma-Aldrich), E-cadherin (1:200; Fuchs laboratory), α-tubulin (1:200; EMD Millipore), rbαK14 (1:400; Fuchs laboratory), rtα Nidogen (1:2,000; Invitrogen), rbα K10 (1:500; Covance), Rb α involucrin (1:1,000; Covance), Rb αfilaggrin (1:1,000; Covance), Rb αHes1 (1:200; Fuchs laboratory), chicken αGFP (1:5,000; Abcam), Rb αGFP (1:5,000; Invitrogen), Rb monoclonal antibody α Notch 1 (4147, 1:200; Cell Signaling Technology) Rbα hamster monoclonal antibody α Notch 2 (1:400; BioLegend), Rb α Notch 3/NICD3 (ab23426, 1:400; Abcam), rbα Presenilin-2 (1:200; Abcam), rbαJagged1 (ab7771, 1:200), rbαPin1 (ab76309, 1:200; Abcam), rbαBBS2 (ab86158, 1:200; Abcam), Rbα GM130 (D1E11, 1:400; Cell Signaling Technology), and Rbα Arf4 (1:200; Abcam). Secondary antibodies conjugated to Alexa Fluor 488 (Molecular Probes), Cy3, RRX, or Cy5 (The Jackson Laboratory) were diluted 1:200–1:500 in gelatin block and incubated 1 h at room temperature. Tissue was extensively washed in PBS before mounting with Pro-Long Gold containing DAPI (Invitrogen).

Cultured keratinocytes lysed in RIPA buffer containing a cocktail of protease inhibitors (Roche). Gel electrophoresis, Western blotting, and infrared imaging were performed as previously described (Ezratty et al., 2011). Primary antibodies used were rat anti–α-tubulin (1:5,000; EMD Millipore), hamster mAb α Notch 3 (1:500; BioLegend), Rb α Notch 3/NICD3 (1:500; ab23426; Abcam), rbα Presenilin-2 (1:10,000; Abcam) mα Presenilin-1. (1:5,000; Abcam) Rbα Arf4 (1:200; Abcam), and rbαBBS2 (1:200; ab86158; Abcam). Secondary antibodies were conjugated to IRDye680 or IRDye800CW (LI-COR Biosciences and Rockland) and were used at 1:15,000.

Confocal images were acquired with a Zeiss LSM510 laser-scanning microscope (ZEISS) through a 63× oil objective. (N.A. 1.4) For WM imaging, Z-stacks of 20–40 planes (0.25 µm) were captured, but only representative single Z-planes are presented. Images were recorded at either 512 × 512 or 1,024 × 1,024 square pixels. For wide-field epifluorescence, images were acquired using an Axioplan 2 (ZEISS) 20×/0.8 air or 63×/1.4 oil Plan-Apochromat objectives equipped with the following Chroma filter sets: 49003 ET YFP (YFP), 49008 ET TR C94094 (mRFP1), 49004 ET dsR C94093 (Cy3, DyLight549), 41008 Cy5 (Cy5), and 41001 FITC (Alexa Fluor 488/GFP).

For quantification of HES1 (nuclear) versus ARL13b/Presenilin-2 (junctional) staining, the number of nuclei in either sagittal or single confocal planes of E17.5 epidermis was determined by thresholding the DAPI or HES1 image in ImageJ and using the measure function to count the number of nuclei, and then cell junction staining was manually determined on the basis of its appearance in various epidermal layers. Suprabasal layers 1–3 were manually determined and ImageJ was used to select layers for analysis as described. For quantification of ciliogenesis, K10 or Notch reporter expression during epidermal development or in cultured keratinocytes, the number of nuclei were determined in single confocal planes by thresholding the image (either DAPI or H2B-RFP [+] nuclei) in ImageJ and using the Measure function to count the number of objects. The number of cilia/K10⁺/GFP⁺ cells were then manually counted in either basal or suprabasal single confocal planes. For quantification of Presenilin-2 basal body localization, the number of nuclei was determined using the measure function in ImageJ, and then the number of basal bodies with a GFP-Presenilin-2 signal was manually determined in single confocal planes of suprabasal cells. For quantification of H2B-RFP/NICD colocalization in Presenilin rescue experiments, the colocalization function in Metamorph was applied to thresholded images and the total pixel area of colocalization was determined and expressed as a percentage of the total nuclear area. Unless otherwise noted, all error bars represent SEM, and an unpaired two-tailed Student’s t test was performed on datasets collected from multiple independent experiments to determine statistical significance where indicated.

Fig. S1 shows relative levels of expression of WT GFP-Presenilin-2 versus P486-GFP-Presenilin-2 in cultured keratinocytes. Fig. S2 shows immunofluorescence and a Western blot of BBS2 or ARF4 in keratinocytes transduced with indicated shRNAs. Video 1 shows a confocal stack of E16.5 epidermis stained with an antibody against Arl13b. Video 2 shows a confocal stack of E16.5 epidermis stained with antibodies against acetylated tubulin and Presenilin-2. DAPI labels nuclei.

We thank D. Oristian for his expert assistance in carrying out all the in utero lentiviral injections reported in this manuscript. We also thank N. Stokes and L. Polak for their expertise and assistance in the mouse facility and Alison North and the Rockefeller University Bioimaging Resource Center for assistance with image acquisition. We are grateful to Meellis Kadaja, Anita Kulukian, Scott Williams, and other members of the Fuchs laboratory for helpful discussions.

Research reported in this publication was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under award numbers K99AR063161 (E. Ezratty) and R37-AR27883 (E. Fuchs).

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. E. Fuchs is an investigator in the Howard Hughes Medical Institute.

The authors declare no competing financial interests.