Ciliogenesis and autophagy are coordinately regulated by EphA2 in the cornea to maintain proper epithelial architecture

Abstract

Purpose:

To understand the relationship between ciliogenesis and autophagy in the corneal epithelium.

Methods:

siRNAs for EphA2 or PLD1 were used to inhibit protein expression in vitro. Morpholino-anti-EphA2 was used to knockdown EphA2 in Xenopus skin. An EphA2 knockout mouse was used to conduct loss of function studies. Autophagic vacuoles were visualized by contrast light microscopy. Autophagy flux, was measured by LC3 turnover and p62 protein levels. Immunostaining and confocal microscopy were conducted to visualize cilia in cultured cells and in vivo.

Results:

Loss of EphA2 (i) increased corneal epithelial thickness by elevating proliferative potential in wing cells, (ii) reduced the number of ciliated cells, (iii) increased large hollow vacuoles, that could be rescued by BafA1; (iv) inhibited autophagy flux and (v) increased GFP-LC3 puncta in the mouse corneal epithelium. This indicated a role for EphA2 in stratified epithelial assembly via regulation of proliferation as well as a positive role in both ciliogenesis and end-stage autophagy. Inhibition of PLD1, an EphA2 interacting protein that is a critical regulator of end-stage autophagy, reversed the accumulation of vacuoles, and the reduction in the number of ciliated cells due to EphA2 depletion, suggesting EphA2 regulation of both end-stage autophagy and ciliogenesis via PLD1. PLD1 mediated rescue of ciliogenesis by EphA2 depletion was blocked by BafA1, placing autophagy between EphA2 signaling and regulation of ciliogenesis.

Conclusion:

Our findings demonstrate a novel role for EphA2 in regulating both autophagy and ciliogenesis, processes that are essential for proper corneal epithelial homeostasis.

Introduction

Eph receptor tyrosine kinases (RTKs) are activated by ephrin ligands found on the surface of neighboring cells and represent the largest family of RTKs. This asymmetry between receptors and ligands creates a cell-cell communication-dependent activation, which leads to unique functions of Eph receptors that involve boundary formation, tissue morphogenesis, differentiation and migration of epithelial cells [1–5]. EphA2, a member of this RTK family, is abundantly expressed in epithelial tissues [6–9]. We have shown that the expression pattern of EphA2 is predominantly in the corneal epithelium and its ligand ephrin-A1 is principally in the limbal epithelium, which leads to segregated cell populations of limbal basal cells from the more differentiated corneal epithelium [10]. EphA2/ephrin-A1 signaling complexes have the potential to regulate epidermal and corneal epithelial cell migration and wound healing during tissue repair [2,5,10,11]. This signaling complex is also attributed to keratinocyte differentiation. For example, in the absence of EphA2 there is accelerated skin tumor growth in an experimental two-stage skin carcinogenesis model [12,13]. Interestingly, increased expression of unliganded EphA2 with low levels of its ephrin-A1 ligand was observed in psoriasis, a skin disease with a hyperproliferative phenotype. Unliganded EphA2 was antagonistic to proper M-phase progression while maintaining cortical rigidity and regulating mitotic spindle formation [14] as well as promoting cell migration and invasion [8].

In addition to the above biological functions attributed to Eph/ephrins, they are involved in signaling networks of structural proteins that affect epithelial cell shape and polarity. For example, Eph/ephrin signaling along with focal adhesion kinase can regulate apical constriction of ciliary band cells [15] and is important for actin cytoskeletal assembly [16]. Moreover, EphA2 signaling mediates epithelial cell polarization by suppressing Arf6 activity [17] and promoting E-cadherin junctions. There is also evidence for positive regulation of tight-junction assembly by EphA2 via direct interaction with afadin, a cell junction associated protein [9]. Afadin was shown to be important in the determination of cell polarity and also primary cilia distribution and organization [18] suggesting a role for Eph/ephrins in ciliogenesis.

Primary cilia are microtubule-based cell organelles that project from the surface of most epithelial cells including epidermal keratinocytes and limbal/corneal epithelial cells [19–28]. Cilia enable cells to receive and transmit extracellular signals, and have been implicated in Hedgehog, Wnt and Notch signaling pathways that impact polarity, proliferation, and differentiation during development and maintenance of tissue homeostasis [19–22]. In early epidermal development, cilia are required to maintain the proper balance between proliferation and differentiation via Notch signaling. Later in development, cilia are involved in correct hair follicle formation through Shh signaling [20,23,29,30]. In the corneal epithelium, cilia appear to play an important role in balancing proliferation and differentiation via Notch signaling [19]. Crucial to proper cilia formation and maintenance is the intraflagella transport (IFT), which is a bi-directional movement of non-membrane-bound particles along the doublet microtubules of the cilia. Disruption of the IFT leads to many disease phenotypes collectively known as ciliopathies and has contributed to our understanding of processes influenced by cilia [31,32].

It is well established that crosstalk exists between primary cilia signaling and autophagy [33–41]. Interestingly, Eph receptors induce autophagy of breast cancer cells [42]. Moreover, ephrin-A1 ligand-dependent activation of EphA2 in prostate cancer cells inhibits mTORC1 [43], which has been well-demonstrated to antagonize autophagy [44,45]. Our proteomic data identified novel potential binding partners of EphA2 associated with autophagy in primary human epidermal keratinocytes [9]. Such involvement of EphA2 in autophagy, as well as in aspects of ciliogenesis, lead us to hypothesize that EphA2 signaling is a common link between primary cilia and autophagy. We now report that EphA2 regulates, in part, end-stage autophagy, corneal epithelial basal cell commitment to post mitotic wing cells, as well as ciliogenesis. The recognition that EphA2 is a receptor coordinately regulating both processes highlights the importance of this RTK in epithelial biology.

Results

Impact of EphA2 on corneal epithelial thickness

EphA2 is specifically expressed in the cell-cell junctions of corneal epithelial cells but absent from the junctions of limbal epithelial cells [10]. Interestingly, corneas of EphA2^−/− mice showed an increase in corneal epithelial thickness due to an increase in the number of cells in the stratified layers (Fig. 1A–D). We did not see any changes in the thickness of the stroma between WT and EphA2^−/− mice. To determine the proliferative rate and potential of the corneal epithelium, we used a 1 h pulse labeling with bromodeoxyuridine (BrdU) and Ki67 immunostaining, respectively. Interestingly, we observed an increase in BrdU + cells (Fig. 1E–G) in the suprabasal layers of EphA2^−/− corneal epithelia and detected an increase in the Ki67+ cells in EphA2^−/− corneal epithelium (Fig. 1H) as well as an increase in BrdU + corneal epithelial cells in cultures depleted of EphA2 (Fig. 1I). When a basal cell commits to a wing cell, it becomes post-mitotic. Thus, this increased proliferation in the wing cell layer increases the number of cells per area and suggests a defect in corneal epithelial basal cell commitment to wing cells and could result in the thickening of the corneal epithelium.

Fig. 1.

EphA2 impacts cell fate decisions: H&E staining of corneas from EphA2 KO (B) and WT littermate control (A) mice (6 weeks old). The epithelial thickness in μm was measured using ImageJ (C). Number of cells above the basal layer was determined by counting the number of nuclei between basal cell layer and the most superficial layers (D). Scattered dots: values for an individual mouse. * = p < 0.05. BrdU staining after 1 h pulse in corneas from EphA2 KO (F) and WT littermate control (E) mice (6 weeks old). White arrows: BrdU + cells in the suprabasal layer. (G) the BrdU + cells in the basal layer, and suprabasal epithelium were counted using ImageJ. Scattered dots: values for an individual mouse. * = p < 0.05. (H) Ki67 staining in corneas from EphA2 KO and WT littermate control mice (6 weeks old) and their corresponding graphs show quantifications of the Ki67+ cells. Scattered dots: values for an individual mouse. Graph (I) shows the BrdU + cells in hTCEpi cells after knocked down of EphA2 (siEphA2) compared to control (siControl). Scattered dots: values for replica* = p < 0.05.

EphA2 is a novel regulator of ciliogenesis

It is well established that ciliogenesis plays a critical role in the regulation of polarized cell division [46–48]. Depletion of one of the proteins involved in ciliogenesis, IFT88, in corneal epithelium, leads to an increase in corneal epithelial thickness corresponding to an increase in proliferation in wing cell layers [19]. Moreover, our RNAseq analysis of EphA2^−/− corneal epithelia (Supplemental Table 1) showed that the genes regulated by EphA2 compared with the genes involved in ciliogenesis had a statistically significant overlap which was validated by EphA2 depletion (siEphA2) in cell culture (Fig. 2A, Supplemental Fig. 1). In order to determine the involvement of EphA2 signaling in the regulation of ciliogenesis, human corneal and limbal epithelial cells (HCECs and HLECs, respectively) were treated with siEphA2 or siControl and stained with acetylated tubulin to detect primary cilia and CEP152 to detect the basal body. Confocal analysis revealed that both cell types contained prominent primary cilia and basal bodies, whereas the EphA2 deficient cells had a dramatic loss of the primary cilium (Fig. 2B). Similarly, to determine whether EphA2 affected cilia in the corneal epithelium, we surveyed whole mounts of mouse corneal epithelium stained with acetylated tubulin to detect primary cilia and γ-tubulin to detect the basal body. In wild-type mice, ~20% of basal corneal epithelial cells have cilia (Fig. 2C), whereas whole mounts of the basal corneal epithelial cells from EphA2^−/− mice revealed significantly fewer cells with detectible ciliary staining (Fig. 2D and E). To determine the role of EphA2 in ciliogenesis, we used Xenopus embryonic skin, a widely accepted model system to study ciliogenesis. In this model, multiciliated cells contain approximately 150 motile cilia. EphA2 is expressed in the epithelial cells of Xenopus skin shown by Xenbase [49] with somewhat weak expression in the ciliated cell precursors [50]. To modify the levels of EphA2 in this model system, we took advantage of morpholino oligos specifically targeting endogenous expression of EphA2. Confocal microscopy of Xenopus skin stained with acetylated tubulin to detect the cilia showed numerous cilia within the epithelial cells (Fig. 2F). When EphA2 levels were reduced in the Xenopus embryonic skin, we observed a marked decrease in acetylated tubulin staining (Fig. 2G). Quantification of these images revealed that the number of the cilia in the EphA2 deficient cells were reduced (Fig. 2I). Overexpression of a morpholino resistant full-length Xenopus EphA2 protein into morphant skin dramatically rescued the number of cilia (Fig. 2H and I).

Fig. 2.

EphA2 regulates the generation of cilia: (A) Venn diagram showing the overlap of potential EphA2 regulated genes determined by RNAseq analysis in corneal epithelium and ciliogenesis related genes (cilia related genes that obtained from ciliaCarta and GO term cilium, cilium organization). DEG: differentially expressed genes. Representation factor indicates overlapping: a higher number means more overlapping. (B) Graph showing the percent of cells with cilia in HCEC and HLEC treated with siControl or siEphA2. Scattered dots: values for replica. Immunostaining of acetylated tubulin (red) and α-tubulin (green) in whole mount corneas of WT (C) or EphA2 KO (D) mice (14 day old). Below each image is an enlargement of the highlighted inset. White arrows in inset: cilia. DAPI was used to stain nuclei. (E) Graph showing number of cilia from (C) and (D). Scattered dots: values for an individual mouse. (F–I) Confocal microscopy of Xenopus skin stained with acetylated tubulin to detect the cilium within the epithelial cells after (F) control morpholino (MO), (G) a morpholino targeting endogenous EphA2, or (H) overexpressing frog EphA2 in Xenopus skin treated with a morpholino targeting endogenous EphA2. Graphs showing percentage of cells with cilia (I) from images (F–H).

Loss of EphA2 results in a defect in end-stage autophagy

Having linked EphA2 as a regulator of ciliogenesis, we considered whether this RTK might also be involved in regulating autophagy, since cross-talk between autophagy and ciliogenesis has been established [33,36,38,40,41] and an unbiased proteomic approach [9] found that about 7.4% (20 out of 270) of the potential EphA2 interacting proteins were involved in autophagy (Fig. 3A). LC3 is a microtubule-associated protein recruited to autophagosomes [51]. Using mice that transgenically express a green fluorescent protein (GFP)-LC3 to assess autophagy, LC3 puncta were detected in all layers of the corneal epithelium. Interestingly, corneal epithelial wing and superficial cell layers have more LC3 puncta compared with the basal layer (Supplemental Fig. 2A). These wing and superficial cells are in relatively close apposition to the surface and are the first areas to experience environmental stress. Thus, it is not surprising that autophagy would be active in these cells. p62 is an autophagy substrate that is used as a reporter of autophagy flux, with high levels of p62 associated with an impairment of autophagy. p62 expression was specifically localized to the basal cells of the human corneal epithelium and was markedly absent in the wing and superficial cells (Supplemental Fig. 2b), suggestive that autophagy is more active in differentiated corneal epithelial cells. We have a sound understanding of how end-stage of autophagy is regulated in the limbal epithelium [52]; however, little is known about the regulation of autophagy in the corneal epithelium. We have reported that perturbations to HCECs (e.g., elimination of miRNAs) results in the formation of large vacuoles, and these vacuoles are retained due to a failure in end stage autophagy [52]. Knockdown of EphA2 (siEphA2) in HCECs resulted in the retention of large vacuoles (Fig. 3B and C), similar to what was observed by blocking end-stage autophagy [52]. It has been shown that excess autophagy can lead to apoptosis [53]. Interestingly, TUNEL staining revealed an increase in the apoptotic response in cells lacking EphA2, even though EphA2 KO mouse corneas were devoid of any positive TUNEL staining (Supplemental Fig. 3). Reduced apoptotic signal in mouse corneas could complement the proliferative response leading to the thickening observed in EphA2 KO corneas (Fig. 1A). To determine the nature of these vacuoles, we used immunostaining to localize markers associated with lysosomes and autophagosomes. Confocal microscopy showed that Lamp1, a lysosomal marker, and LC3, an autophagosome marker, colocalized to the large vacuoles (Fig. 3D), confirming their autolysosomal nature. To examine further the autophagic origin of these large vacuoles, the autophagy initiation regulatory proteins ATG5, Beclin1 or both were knocked down in cells with siEphA2. Consistent with the autophagic vesicle feature of large vacuoles, knockdown of ATG5 and/or Beclin 1 resulted in markedly decreased LC3II levels, indicative of blocking autophagic vesicle formation, which was accompanied with a significant reduction in large vacuoles in cells with siEphA2 (Supplemental Figs. 4A and B).

Fig. 3.

EphA2 regulates autophagy: (A) Venn diagram showing the overlap of potential EphA2 interacting proteins and autophagy genes (GO consortium, GO:0006914). (B) Phase-contrast microscopy of hTCEpi cells 48hr after siControl or siEphA2 transfection. (C) Number of cells with vacuoles were counted in (B). Three fields (299 μm × 237 μm) per well in each experiment were counted. Each experimental replicates included 3 technical replicates. Scattered dots: values for replica. (D) Confocal imaging of LAMP1 (red) and LC3 (green) immunofluorescent staining in hTCEpi cells after siControl or siEphA2 transfection. White arrows: LC3 puncta on the membranes of LAMP + vesicles. (E–F) Western blotting showing EphA2, p62, LC3I and LC3II levels in hTCEpi cells after siControl or siEphA2 transfection (E) as well as in mouse corneal epithelium isolated from EphA2 KO and littermate control mice (F; 6 weeks old). Graphs indicate the densitometry analysis of blots for LC3II and p62 levels. * = p < 0.05. GAPDH was used as a loading control. (G–H) Fluorescence images showing GFP-LC3 puncta in EphA2 KO mice and littermate controls (WT) corneal epithelium. DAPI was used to stain nuclei. White dotted lines demarcate the basement membrane region. The GFP-LC3 signal in a region of wing cells in EphA2 KO cornea was enhanced to show the puncta. Number of LC3 puncta was counted using ImageJ (H). Scattered dots: values for an individual mouse.

LC3II is a well-accepted marker of autophagosomes and p62 is a negative marker of autophagy flux [45]. Accumulation of both markers suggests a defect in autophagy [45]. Using a genetic approach, when we knocked down EphA2 in HCECs and/or in corneal epithelium isolated from EphA2 null mice, a significant accumulation of both these markers was observed (Fig. 3E and F). Similarly, there was an enhancement in GFP-LC3 puncta in EphA2 null mouse corneas when compared to WT littermates (Fig. 3G and H). These observations indicate an inhibition of autophagy.

In end-stage autophagy, autophagic lysosome reformation (ALR), is an important process for autolysosome recycling [54,55]. We have demonstrated that inhibition of such a process in corneal and epidermal keratinocytes results in accumulation of large Lamp1+; LC3+ vacuoles [52,56] that resemble those in cells with EphA2 knockdown. Thus, we investigated whether the large vacuoles in siEphA2 cells are a result of fusion of autolysosomes due to the failure of autolysosome recycling at end-stage autophagy. Bafilomycin A1 (BafA1) prevents the formation of autolysosomes [45]. Thus, we used this drug to determine whether vacuole accumulation occurs during the early or late stages of autophagy. In hTCEpi cells, treatment with BafA1 significantly reduced the formation of large vacuoles after knockdown of EphA2 (Fig. 4A) without any significant effect on cell viability (Supplemental Fig. 5). This finding was confirmed using confocal microscopy in conjunction with LAMP1 staining (Fig. 4B). It should be noted that BafA1 markedly reduced the siEphA2-induced vacuoles; however, BafA1 did not prevent the accumulation of autophagosomes and lysosomes but rather prevented the fusion of lysosomes with autophagosomes (Fig. 4B). Furthermore, in hTCEpi, HLECs and HCECs, knockdown of EphA2 resulted in increased LC3II compared with siControl, while after BafA1 treatment, there were no significant differences in LC3II levels between siControl and siEphA2 (Fig. 4C–E). Moreover, treatment with an activator of autophagy via inhibition of mTORC1, rapamycin, did not reverse the vacuole formation or increase in LC3BII and p62 levels by EphA2 depletion (Supplemental Fig. 6). These observations suggest that accumulation of LC3II in cells with siEphA2 is not due to enhanced induction of autophagy but rather a reduced LC3II turnover, confirming an inhibition of autophagy flux at the late stage.

Fig. 4.

EphA2 regulates end-stage autophagy: (A) Phase-contrast microscopy of hTCEpi cells after siControl or siEphA2 transfection with or without BafA1 treatment for 24hr. Graph shows number of cells with vacuoles. Three fields (299 μm × 237 μm) per well in each experiment were counted. Each experiment replicates included 3 technical replicates Scattered dots: values for replica. (B) Confocal imaging of LAMP1 (red) immunofluorescent staining in hTCEpi cells after siControl or siEphA2 transfection with or without BafA1 treatment (100 nM) for 24hr. Inset shows enlarged area with small vesicles. (C–E) Western blotting showing EphA2, LC3I and LC3II levels in hTCEpi cells (C), HLECs (D) or HCEC (E) after siControl or siEphA2 transfection with or without BafA1 treatment for 24hr. Graphs show the densitometry analysis of blots for LC3II levels. GAPDH was used as a loading control. * = p < 0.05. N.S. = non-significant. n = 3.

EphA2 and its novel binding partner PLD1 ensure end-stage autophagy via PKC signaling

With respect to the mechanism underlying large vacuole formation, we have demonstrated that the PLD1/PKCα pathway inactivates dynamin1 and consequently blocks end-stage autophagy in limbal epithelial cells, resulting in the accumulation of large vacuoles [52]. Interestingly, one of the potential EphA2 interacting proteins in our unbiased proteomics data was phospholipase D1 (PLD1), an upstream regulator of PKC signaling. Using benchtop immunoprecipitation of biotinylated proteins in close proximity to EphA2-BirA* [9], we were able to validate the association of PLD1 with EphA2 (Fig. 5A). PLD1 also co-immunoprecipitated with EphA2 in hTCEpi cells (Fig. 5B). To confirm EphA2/PLD1 association via an alternative approach, we took advantage of super-resolution structured illumination microscopy (SIM), and showed that PLD1 was colocalized with EphA2 (Fig. 5C and D). Significantly, knock down of PLD1 (siPLD1) partially rescued vacuole formation (Fig. 5E and F) and LC3II accumulation in cells deficient in EphA2 (Fig. 5G). This was further confirmed by elimination of these vacuoles in siEphA2 cells treated with two PKC inhibitors, RO-32–0432 and Rottlerin (Supplemental Fig. 7). More importantly, treatment of mouse eyes with an inhibitor of PKC, Sotrastaurin, eliminated the LC3II accumulation in corneal epithelium in EphA2 KO mice (Fig. 5H). Collectively, this suggests a role for PLD1/PKC in the regulation of end stage autophagy by EphA2.

Fig. 5.

EphA2 interaction with PLD1 regulates end stage autophagy: (A) Benchtop BioID pulldown from hTCEpi overexpressing LZRS-Bir* or LZRS-EphA2-Bir*. Immunoblotting was performed for PLD1 and EphA2. Strepravidin-HRP was used in whole cell samples (WC) to show the level of overall biotinoylation in samples overexpressing EphA2-Bir*. B* = BioID-LZRS, A2* = EphA2-BioID-LZRS, L = LZRS, A2 = EphA2-LZRS, NC = negative control (pulldown without protein sample). (B) Interaction of PLD1 with endogenous and ectopically overexpressed EphA2 in hTCEpi was detected by immunoprecipitation with anti-PLD1 antibody, followed by immunoblot analysis with anti-EphA2, or anti-PLD1 antibodies. GAPDH was used as a loading control. (C) SIM image in hTCEpi cells shows colocalization of PLD1 and EphA2. Inset: white arrows indicate the colocalization. (D) line scan analysis was conducted (white lines in C) and showed colocalization of PLD1 and EphA2 proteins. N = 3. (E) Phase-contrast microscopy of hTCEpi cells after siControl, siEphA2 or siEphA2+siPLD1 transfection. (F) Number of cells with vacuoles were counted in (E). Three fields (299 μm × 237 μm) per well in each experiment were counted. Each experiment replicates included 3 technical replicates. Scattered dots: values for replica. * = p < 0.05. n = 4. (G–H) Western blotting showing EphA2, LC3I and LC3II levels in hTCEpi cells after siControl or siEphA2 transfection with or without siPLD1 cotransfection (G), or in mouse corneal epithelium isolated from EphA2 KO mice and littermate control after treatment with PKC inhibitor, Sotrastaurin (0.5 g/kg) (H). Graphs show the densitometry analysis of blots for LC3II levels. LC3II bands normalized to GAPDH using densitometric analysis in ImageJ. GAPDH was used as a loading control. * = p < 0.05.

Having determined that EphA2 regulated autophagy via a PLD1 interaction and since autophagy regulates ciliogenesis [33,36,38,40,41], we reasoned that PLD1 may also be the mechanism by which EphA2 controls ciliogenesis. When HCECs treated with siEphA2 were transfected with siRNA targeting PLD1 (siEphA2+siPLD1), primary cilia were partially rescued as evidenced by the reappearance of positive staining for acetylated tubulin (Fig. 6A–C, E). Furthermore, when siEphA2+siPLD1 cells were treated with BafA1 to inhibit autophagy, siPLD1-induced rescue of cilia formation in EphA2 deficient cells was reduced (Fig. 6D and E). As expected, rapamycin treatment was unable to rescue cilia formation in cells depleted of EphA2 (Supplemental Fig. 8). Taken together, these findings suggest that the EphA2/PLD1 signaling pathway, which is critical for end-stage autophagy, also regulates ciliogenesis (Fig. 6F) in the corneal epithelium.

Fig. 6.

EphA2 regulates generation of cilia via PLD1 and autophagy. Immunostaining of Acetylated tubulin (Ac-tubulin; red) and CEP152 (green) in hTCEpi treated with siControl (A), siEPhA2 (B), siEphA2+siPLD1 (C) or siEphA2+siPLD1 with BafA1 treatment (D). DAPI was used to stain nuclei. (E) Graph shows the percent of cells with cilia. * = p < 0.05. Scattered dots: values for replica. (F) Model showing EphA2-PLD1 interaction in regulating both autophagy and ciliogenesis, processes that are essential for proper corneal epithelial homeostasis. EphA2 interacts and inhibits PLD1 to activate PKC signaling that leads to progression of end-stage autophagy. PLD1 is also involved in the regulation of ciliogenesis by EphA2, either directly or by regulating end-stage autophagy.

Discussion

Maintaining a proper stratified epithelium requires a well-organized cell fate decision machinery [57]. Recently, we have shown that EphA2 is a critical regulator of corneal epithelial homeostasis, specifically in cellular migration and limbal/corneal epithelial barrier function [2,10]. We now demonstrate a novel role for EphA2 in regulating both autophagy and ciliogenesis, processes that are essential for proper corneal epithelial organization. Specifically, our data demonstrate a link between EphA2 and end stage autophagy in the regulation of (i) ciliogenesis and (ii) epithelial cell passage from transit amplifying (TA) to post-mitotic cells.

Autophagy is a means by which cells adapt to differing intrinsic and extrinsic cellular stress-related situations. Autophagy regulates numerous physiological processes in which cellular components need to be degraded and recycled. Examples of such processes are the cell’s response to starvation and other stresses, removal of bacteria and viruses, elimination of damaged proteins and organelles that occur during aging, and maintenance of stem cell homeostasis [45,51,54,55,58,59]. We have shown that mice deficient in Beclin 1, an integral protein involved in the formation of the autophagosome, show a significant delay in proliferation following wounding [52], most probably due to a reduction in the stem cell/early TA cell compartment in the limbus [60]. Our proteomic data identified 270 novel potential binding partners of EphA2 in primary human epidermal keratinocytes [9]. Interestingly, Gene Ontology (GO) analysis showed that a significant number of these binding partners were associated with autophagy. We have shown that one of the binding partners of EphA2, PLD1, is involved in the regulation of end stage autophagy downstream from EphA2 signaling (Fig. 5) and this association was shown to regulate primary cilia formation in corneal epithelial cells in an autophagy dependent manner (Fig. 5, Fig. 6). Interestingly, EphA2 depleted corneas had increased basal cell proliferative rate and potential and cells lacking EphA2 had higher proliferation (Fig. 1) similar to what was seen in ciliogenesis-deficient mouse corneas [19]. The regulation of ciliogenesis by EphA2 in the basal cell layer of corneal epithelium might determine the transition of TA cells to the post-mitotic wing cells.

Our findings support the idea that EphA2 signaling is a common link between autophagy and ciliogenesis (Fig. 6F). The molecular network of the autophagy pathway (ATGs and the upstream signaling) has been studied over the last two decades [35,45,58,59,61–64]; however, much less is known about how autophagy affects cell biological changes within a cell and/or tissue, such as guiding a specific cell pathway. Specifically, little attention has been directed to autophagy as a regulator of the transition from a proliferating cell to a more differentiated state (e.g., corneal epithelial basal to wing cell). Of particular importance is the observation that depletion of one of the well-studied IFT proteins, IFT88 in corneal epithelium, led to an increase in corneal epithelial thickness primarily due to an accumulation of superficial or wing cells [19]. Interestingly, the phenotype of diminished EphA2 signaling (Fig. 1), bears a striking resemblance to altered ciliogenesis with an increased corneal epithelial thickness. In support of this resemblance, there were several cilia associated genes changed in corneas of EphA2 KO mice and cells depleted of EphA2 (Fig. 1A, Supplemental Fig. 1). We observed multiple basal cell layers, an increase in LC3 puncta in wing cell as well as in squamous cell layers in EphA2 null mouse corneas. The reduced apoptotic signal along with increased proliferative rate in corneal basal and suprabasal cells [65] could explain the increased thickness of the corneal epithelium in EphA2 null mice (Fig. 1).

In stratified epithelia such as corneal epithelium and epidermis, proliferating basal cells ultimately undergo a commitment to become suprabasal cells, which are designated as wing cells in corneal epithelium. During this transition, basal TA cells lose proliferative capability and migrate vertically into the suprabasal layer [66]. Asymmetric division among the proliferating basal cells plays a key role in activating this basal to suprabasal fate switch [67,68]. During asymmetrical division, one of the daughter cells remains a basal cell and the other commits to a suprabasal or wing cell [[68], [69], [70]]. It is thought that the new basal cell has an excess of stem/progenitor cell factors and thus remains relatively undifferentiated with additional proliferative capacity. The suprabasal positioned, committed wing cell, inherits an excess of lineage commitment factors (e.g., Notch) [71] and loses proliferative capability. A similar loss of Notch signaling was observed in ciliogenesis deficient, IFT88 knockout corneas and correlated with the thickening of the corneal epithelium via vertical migration of basal corneal epithelial cells [19]. Moreover, in many tissues, autophagy is highly active during cell fate commitment, and we now show that autophagy is integral for the proper corneal epithelial cell transition from TA to more mature, post mitotic epithelial cells. We now show that, central to this regulation is EphA2 signaling in corneal epithelium, upstream from autophagic vacuole reformation and primary cilia formation. Our findings support the idea that in healthy corneal epithelium, EphA2 signaling maintains proper primary cilia in order to transition TA cells from the basal layer to the more mature epithelial cells via regulation of end stage autophagy. This signaling could be important to drive cell fate commitment during homeostasis and diseases in all stratified epithelia.

Material and methods

Cell culture

Immortalized corneal epithelial cells, hTCEpi, were cultured in keratinocyte serum free media (Life Technologies) as described previously [72]. Limbal (HLEC) and corneal epithelial (HCEC) cells were isolated as explants from human corneas obtained from the Eversight eye banks (Ann Arbor, MI, USA) as described previously [52]. HLEC and HCEC were cultured in CnT-PR media (CELLnTEC, Bern, Switzerland), and first passage cells were used for experiments. In order to deplete EphA2 in cell culture, two different siRNA oligonucleotides (Dharmacon, CO; Cat# LQ-003116–00-0005) were combined (10 nM each) to specifically target EphA2. DharmaFECT1 (Dharmacon, CO) was used as carrier. Most of the experiments were performed in hTCEpi cells due to more efficient transfection of the siRNAs. However, these experiments were validated in HLEC and HCEC as presented.

Animal studies

Animal procedures were approved by the Northwestern University Institutional Animal Care and Use Committee and adherence to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. EphA2 null mice (EphA2^−/−) were obtained from Jackson laboratories and backcrossed on a C57Bl/6 background. Mouse whole eyes were fixed in 10% neutral buffered formalin solution for paraffin embedding for histological analysis. H&E staining was conducted as described previously [73] and imaged using an Zeiss Axioplan 2 microscope system (Carl Zeiss, Oberkochen, Germany). For the BrdU pulse labeling assay, mice were injected with BrdU (50 μg/kg) intraperitoneally 1 h prior to euthanasia, whole eyes were fixed in 10% neutral buffered formalin solution and processed for paraffin embedding. To control for the axis of the section of mouse corneas when embedding, the line crossing the medial canthus and lateral canthus was aligned with the long side of OCT embedding mold or perpendicular with the bottom of paraffin embedding mold. In some experiments, mice were injected intraperitonially and corneas topically treated twice with Sotrastaurin (0.5 μ g/kg) [74] to inhibit PKC signaling. Corneal epithelial sheets were isolated by incubating whole globes in PBS containing 20 mM EDTA for 1 h at 37 °C, as described previously [75] for western blotting analysis. For bulk RNAseq analysis, WT and EphA2^−/− corneal epithelial layers were isolated as before [76]. Total RNA from these epithelial sheets was isolated using miRNeasy Mini kit (RNeasy; Qiagen, Valencia, CA), RNAseq analysis was performed by the NUseq core using Illumina HiSeq 75bp Paired-End platform (deposited as GSE174649). Genes regulated by EphA2 based on changes seen with EphA2−/− was compared to ciliogenesis related genes (cilia related genes that obtained from ciliaCarta, GO term cilium and The SysCilia Gold Standard (SCGS)) [77–79].

Gene Ontology analysis

Functional Annotation Clustering was performed in DAVID Functional Annotation Bioinformatics Resources v6.7 and GeneGo. EphA2 interactors from our BioID assay [9] were compared to autophagy related genes (GO consortium, GO:0006914).

Streptavidin pulldown and co-immunoprecipitation assay

Cells were transduced with either LZRS-EphA2-BirA* or LZRS-BirA* (control) for 48 h. Then, cells were incubated with biotin for 24 h prior to harvest. Biotinylated proteins were pulled down using streptavidin beads (Sigma-Aldrich, St. Louis, MO) as described [73]. Co-immunoprecipitation (Co-IP) assay was performed as previously described [73]. Cells were transduced with either LZRS or LZRS-EphA2 for 48 h. Protein lysates were incubated with protein A/G PLUSAgarose beads (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) plus PLD1 antibody (Cell Signaling Technologies). Sample mixtures were rotated for 2 h at 4 °C. The beads were washed three times with ice-cold phosphate-buffered saline and subjected to immunoblot analysis for EphA2 and PLD1.

Western blotting

Western blots were performed as described [73]. The following antibodies were used: GAPDH (Santa Cruz Biotechnology, Texas, USA), LC3B, PLD1, p62 (Cell Signaling Technologies, Massachusetts, USA), EphA2 (Millipore).

Immunofluorescence and whole mount staining of mouse corneas

Immunofluorescent (IF) studies were conducted as described [80]. Sections or methanol fixed cells were incubated with primary antibodies (1:50) overnight at 4 °C. The following antibodies were used: BrdU (G3G4, Developmental Studies Hybridoma Bank), Acetylated tubulin (Sigma-Aldrich, St. Louis, MO), CEP152 (Bethyl Laboratories), LAMP1, PLD1, LC3 (Cell Signaling Technologies, Massachusetts, USA) and EphA2 (Millipore). Sections were counterstained with DAPI. Images were taken using an epifluorescence Zeiss Axiovision Z1 microscope with an AxioCam MRm digital camera (Carl Zeiss) or A1R; Nikon using a 60× oil Plan-Apochromat objective lens with a 1.4 NA. Axiovision or Nikon Elements Software was used for all acquisition and image processing. For whole mount staining of corneas, eyes were processed as described [10,81,82] and fixed in 4% PFA for 20 min. Corneal rims were further fixed in ice-cold acetone in −20 °C for 5 min, before processing for immune staining for Acetylated Tubulin (Sigma 6–11B-1) and gamma-Tubulin (Sigma T3195). Imaging was performed using A1 Nikon confocal microscopy. For BrdU staining, antigen retrieval of the paraffin sections was in 2 N HCL for 20 min at room temperature. After blocking in PBS containing 5% goat serum and 0.01% Tween-20, sections were incubated overnight with BrdU monoclonal antibody (1:10; Developmental Studies Hybridoma Bank, Iowa, USA). After washing, sections were incubated with Alexa555-linked secondary anti-mouse IgG (Thermo Fisher Scientific, Massachusetts, USA). Images were taken using an AxioVision Z1 fluorescence microscope system (Carl Zeiss, Oberkochen, Germany). Cell counting, thickness measurement and relative fluorescence analysis were conducted by ImageJ. When measuring cell count and thickness, the central portion of the cornea in 6 different locations, equally separated from each other, were analyzed. For super-resolution structured illumination microscopy (SIM), imaging was performed using an Nikon N-SIM microscope equipped with an EM-CCD camera iXon3 DU-897E (Andor Technology, Belfast United Kingdom) and a 100 × apo 1.49-NA objective lens as described previously [52,73]. Reconstruction was conducted by the Nikon NIS Elements software package.

Xenopus skin as a model system to study ciliogenesis

All Xenopus experiments were done as described [83]. Briefly, Xenopus embryos were obtained by in vitro fertilization using standard protocols [84] approved by the Northwestern University Institutional Animal Care and Use Committee. Morpholino antisense oligonucleotides (GeneTools) were injected into one or more blastomeres at the four-cell stage with membrane RFP as a tracer to inhibit the expression of Xenopus EphA2 (Entrez ID: 394370). A morpholino was used to target EphA2 at the initiation site: EphA2 MO: 5′- CAATCCGCTCCGCTCTCATGTTCAT-3′; Control MO: 5′-CCTCTTACCTCAGTTACAATTTATA-3′. For the rescue experiments, Xenopus EphA2 was cloned into the CS2+ XLT vector with a C-terminal GFP. This was co-injected together with the MO at the two- or four-cell stage with 10–20 pg of EphA2-GFP construct.

Statistical analysis

All values are expressed as mean ± SD. The significance of the differences between 2 groups was evaluated by an unpaired Student’s t-test. For the differences between 3 or more groups, a 1-way ANOVA with post hoc pair-wise t-test comparisons using Bonferroni’s correction for multiple comparisons was conducted. Parameters with values P < 0.05 were considered significant.