Autophagy mediates cell cycle response by regulating nucleocytoplasmic transport of PAX6 in limbal stem cells under ultraviolet-A stress

Limbal stem cells (LSC) account for homeostasis and regeneration of corneal epithelium. Solar ultraviolet A (UVA) is the major source causing oxidative damage in the ocular surface. Autophagy, a lysosomal degradation mechanism, is essential for physiologic function and stress defense of stem cells. PAX6, a master transcription factor governing corneal homeostasis by regulating cell cycle and cell fate of LSC, responds to oxidative stress by nucleocytoplasmic shuttling. Impaired autophagy and deregulated PAX6 have been reported in oxidative stress-related ocular surface disorders. We hypothesize a functional role for autophagy and PAX6 in LSC’s stress response to UVA. Therefore, human LSC colonies were irradiated with a sub-lethal dose of UVA and autophagic activity and intracellular reactive oxygen species (ROS) were measured by CYTO-ID assay and CM-H2DCFDA live staining, respectively. Following UVA irradiation, the percentage of autophagic cells significantly increased in LSC colonies while intracellular ROS levels remained unaffected. siRNA-mediated knockdown (KD) of ATG7 abolished UVA-induced autophagy and led to an excessive accumulation of ROS. Upon UVA exposure, LSCs displayed nuclear-to-cytoplasmic translocation of PAX6, while ATG7KD or antioxidant pretreatment largely attenuated the intracellular trafficking event. Immunofluorescence showing downregulation of proliferative marker PCNA and induction of cell cycle regulator p21 indicates cell cycle arrest in UVA-irradiated LSC. Abolishing autophagy, adenoviral-assisted restoration of nuclear PAX6 or antioxidant pretreatment abrogated the UVA-induced cell cycle arrest. Adenoviral expression of an ectopic PAX gene, PAX7, did not affect UVA cell cycle response. Furthermore, knocking down PAX6 attenuated the cell cycle progression of irradiated ATG7KD LSC by de-repressing p21 expression. Collectively, our data suggest a crosstalk between autophagy and PAX6 in regulating cell cycle response of ocular progenitors under UVA stress. Autophagy deficiency leads to impaired intracellular trafficking of PAX6, perturbed redox balance and uncurbed cell cycle progression in UVA-stressed LSCs. The coupling of autophagic machinery and PAX6 in cell cycle regulation represents an attractive therapeutic target for hyperproliferative ocular surface disorders associated with solar radiation.


Introduction
The corneal epithelium, an indispensable prerequisite for visual acuity, is postnatally maintained and regenerated by a pool of adult stem cells, termed limbal stem cells (LSC) [1][2][3][4]. Solar ultraviolet A (UVA) is a major environmental hazard causing acute photodamage in cornea and chronic exposure is often associated with hyperproliferative, yet degenerative ocular surface diseases, such as pterygium [5][6][7]. Cells respond to UVA stress by activation of antioxidant signaling pathways, dynamic regulation of cell cycle or apoptosis. To date, key cellular and molecular signaling events driving LSC's stress response remain unclear in UVA-related ocular pathology. Autophagy, a lysosomal degradation system, is essential for maintenance of stem cell characteristics, including self-renewal, differentiation and quiescence [8][9][10][11]. Accumulating evidence suggests that autophagy contributes to cellular defense mechanisms in somatic stem cells under various types of stress [12,13]. Whether autophagy plays a role in LSC's stress response to UVA remains elusive.
Paired-box protein 6 (PAX6) is a master transcription factor guiding corneal morphogenesis and homeostasis by regulating cell cycle and fate of tissue progenitor cells [14,15]. Recent reports suggest that PAX6 is implicated in corneal wound healing [16] and inflammatory response [17,18], both of which are cellular events with a transient surge of reactive oxygen species (ROS). Interestingly, Ou et al. demonstrated that PAX6 in corneal epithelial cells responded to oxidative stress by nuclear-to-cytoplasmic translocation [19]. However, the molecular mechanism and the physiological/pathological significance of the oxidative stresstriggered PAX6 relocation remains undetermined. In this study, we sought to characterize LSC's cellular response to UVA radiation and explored the potential role(s) of autophagy and PAX6 in the identified stress response.

Materials and methods
Cell culture of primary human LSC This research has been approved by the Ethical Committee of the Medical University of Vienna (IRB approval number MUW 1578/2013). Human corneal tissues procured in MUW cornea bank not suitable for corneal transplantation were used in the current study. Limbal epithelial sheets from corneoscleral rims were surgically isolated after overnight incubation in dispase II (10.7 U/mL, from Bacillus polymyxa, Sigma-Aldrich, St. Louis, MO) at 4˚C and subjected to trypsin digestion for 10 mins at 37˚C. Cells were seeded at densities ranging from 100 to 400 cells/cm 2 and cultured up to 12 days. Colonies were expanded in serum-and feeder cell-free conditions in Keratinocyte-Serum Free Medium (KSFM, Gibco, Thermo Fisher Scientific Inc., Waltham, MA) supplemented with 5 ng/mL human recombinant epidermal growth factor and 50 μg/mL bovine pituitary extract (both Gibco).

PAX6 and PAX7 overexpression
Pre-packaged human adenovirus (dE1/E3 serotype 5, Vector Biolabs, Malvern, PA) expressing the human PAX6 gene under control of the cytomegalovirus (CMV) promoter (Ad-CMV-PAX6) was used to ectopically express PAX6 in LSC colonies. To study whether the observed PAX6 gene functions were ocular tissue-specific, experiments of ectopic PAX7 gene expression in LSC colonies were performed in parallel. Adenoviral cassettes carrying viral backbones and an empty CMV promoter (Ad-CMV-null) served as controls. Colonies were infected on clonal day 10 with an MOI of 50 overnight. Cells were cultured for additional two days in adenovirus-free KSFM for protein expression before experiments were performed.

UVA irradiation
Sellamed 3000 UVA-1 irradiation device (Sellas, Ennepetal, Germany) was used to generate radiation filtered for the emission of UVA light in the spectral range of 340-440 nm. During irradiation, cells were transferred to Dulbecco's Phosphate Buffered Saline (DPBS, Gibco). A non-lethal dose of UVA titrated to the irradiance of 20 J/cm 2 was used throughout all experiments to study PAX6's regulation on cell cycle progression in UVA-stressed LSCs (S1 Fig).

Anti-oxidant treatment and ROS detection
Cells were treated with an antioxidant mixture (10 μM α-tocopherol and 15 mM N-acetyl-Lcysteine, both Sigma-Aldrich) for 16 hours prior to UVA irradiation. Cells treated with 20 μM H 2 O 2 for 6 hours served as positive controls for oxidative stress. ROS levels were visualized using CM-H 2 DCFDA (10 μM; Molecular Probes, Eugene, OR) according to the manufacturer's instructions. Live cell imaging was performed by laser scanning confocal microscopy 30 minutes after exposure to UVA. For quantification, mean fluorescence intensity of LSC colonies was determined.

Autophagosome and autophagic flux measurement
Autophagy activity was determined by autophagosome formation and autophagic flux. Autophagosomes were visualized 6 hours after UVA exposure using a cationic amphiphilic tracer dye to stain autophagosomal vesicles (CYTO-ID, Enzo Life Sciences, Farmingdale, NY). Cells treated with 10 μM rapamycin (from Streptomyces hygroscopicus, Sigma-Aldrich) for 6 hours served as positive controls. Autophagic cells were defined as cells displaying more than five perinuclear puncta [20,21]. A minimum of 250 clonal cells were counted in each sample and data are presented as percentage of autophagic cells per colony. Basal and UVA-stimulated autophagic flux were determined by western blot analysis of lipidated form of LC3B (LC3B-II). LC3B-II expression in the absence or presence of autophagic flux inhibitor Bafilomycin A1 (BafA1, 100 nM from Streptomyces griseus, Sigma-Aldrich) with or without UVA irradiation was assessed at different time points up to 24 hours [22]. Data were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) expression levels for densitometric analysis.

Apoptosis detection
Apoptotic cells were quantified 6 hours after UVA irradiation using terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling assay (TUNEL; Click-iT TUNEL Imaging Assay, Thermo Fisher Scientific Inc.) according to the manufacturer's protocol.
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) Total RNA was isolated using RNeasy Plus Mini Kit (Qiagen, Hilden, Germany) and reverse transcribed using iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Inc., Hercules, CA). qPCR was performed using TaqMan Gene Expression Assays and Gene Expression Master Mix (both Applied Biosystems, Foster City, CA). Amplification reactions were carried out according to a standardized qPCR program of the ABI Prism 7500 Fast Real-Time PCR System (Perkin Elmer, Applied Biosystems). Data were analyzed using the "delta-delta method" as suggested by Applied Biosystems and Pfaffl [23]. Results display averaged relative quantification (RQ) target gene expression normalized to GAPDH values of at least three biological repeats.

Immunofluorescence (IF) confocal microscopy
For immunofluorescent staining, cells were fixed in 4% (v/v) formaldehyde and subsequently immersed in DPBS containing 0.1% (v/v) Triton X-100 for permeabilization. Samples were blocked using serum-free Protein Block (Dako, Glostrup, Denmark). Antibody diluent (Dako) was used to dilute primary and secondary antibodies. Mouse and rabbit IgG served as isotype controls (Vector Laboratories, Burlingame, CA). 4,6-diamidino-2-phenylindole dilactate (DAPI) and Hoechst 33342 (both Molecular Probes) were used as nuclear counterstains for fixed and viable cells, respectively. Images were acquired using an inverted laser scanning confocal microscope (LSM780, Carl Zeiss GmbH, Oberkochen, Germany) equipped with a 405-nm laser diode and an argon ion laser capable of providing an illumination of 488 nm. A helium-neon laser was used to generate an excitation wavelength of 594 nm. Fluorescent signals were detected by a 32-channel spectral detection unit (gallium-arsenide-phosphide detector 416-728 in 8.3 nm steps, Carl Zeiss). Imaging was routinely performed with a 20x air objective (numeric aperture 0.8) and a 63x oil lens (numeric aperture 1.4) was used to detect autophagosomal vesicles using ZEN software (2009, Carl Zeiss). Image processing software ImageJ and Fiji (Laboratory for Optical and Computational Instrumentation, Madison, WI) were used for image analysis [24,25].
Chemiluminescence was detected using ChemiDoc XRS system with Image Lab software (version 5.2.1, Bio-Rad Laboratories, Inc.).

Statistical analysis
Data represent average of at least 3 independent experiments. For IF data analysis, at least 10 colonies per condition were imaged, unless specified otherwise. Statistical analysis was performed using non-paired Student's t test, one-way ANOVA, Sidak's, Dunnett's and Tukey's multiple comparisons (Graphpad Prism 6.05, GraphPad Software, Inc., La Jolla, CA). Data are shown as means ± s.e.m. with p < .05 considered statistically significant.

Validation of ATG7KD as an in vitro model of autophagy-deficient human LSC
In order to generate autophagy-deficient human LSC in vitro, colonies during exponential growth phase (clonal days 8-10) were transfected with two siRNAs targeting human homolog of autophagy-related gene 7 (ATG7) while controls received scrambled siRNAs not complementary to any human gene sequence (SCR). The efficiency of siRNA-mediated ATG7KD was measured at both mRNA and protein levels. Quantitative PCR demonstrated a 65.1 ± 1.6% reduction of ATG7 mRNA expression in ATG7KD compared to SCR control (p < .05) ( Fig  1A). Western blot analysis and immunofluorescence verified KD efficiency of ATG7 in ATG7KD LSC at protein levels compared to SCR transfected controls (Fig 1B & 1C). To validate the functional abolishment of autophagy in ATG7KD LSCs, cellular activity of autophagosome formation in response to rapamycin, an autophagy inducer, was studied. Rapamycin treatment significantly increased the percentage of autophagic cells from basal autophagic levels of 35.7 ± 4.9% to the inductive level of 84.1 ± 4.1% in SCR (p < .01), while no inductive effect of rapamycin was observed in ATG7KD LSCs (30.2 ± 6.4% in non-treated ATG7KD versus 52.9 ± 8% in rapamycin-treated ATG7KD, p>.05) (Fig 1D & 1E). Taken together, these data validate a successful blockage of inductive autophagy by siRNA-mediated KD of ATG7 in our LSC clonal culture.

UVA activates autophagy by increasing autophagic flux
To test whether autophagy mediates LSC's ultraviolet stress response, formation of autophagosomal structures was assessed 6 hours after exposure to 20 J/cm 2 UVA by CYTO-ID assay. UVA radiation induced autophagy in SCR LSCs evidenced by an increase in autophagic cells from 35.7 ± 4.9% to 59.4 ± 3.3% compared to non-irradiated controls (p < .01) (Fig 2A & 2B). In contrast, the number of autophagic cells remained at baseline levels after UVA irradiation in ATG7KD LSC in comparison to untreated counterparts (30.2 ± 6.4% in ATG7KD and 30.8 ± 6.8% after UVA, p>.05). Data indicate that UVA increases autophagosomes in an ATG7-dependent mechanism.
To determine whether the observed autophagosome increase in irradiated SCR colonies results from UVA-impaired autophagosomal clearance or UVA-stimulated autophagic flux, we then employed BafA1, an autophagolysosomal inhibitor, to measure autophagic flux at homeostatic state and under UVA stress. In non-irradiated LSCs, LC3B-II was found elevated after 24 hours of BafA1 treatment (p < .05 compared to drug vehicle-treated group), indicative of a low homeostatic autophagic flux in LSC colonies (Fig 2C & 2D). In contrast, the elevation of LC3B-II started as early as 6 hours after UVA irradiation and remained elevated at 24 hours in BafA1-treated LSCs (p < .05 compared to BafA1-treated, non-irradiated group), suggesting that UVA activates autophagy in LSCs by increasing autophagic flux.

Autophagy regulates intracellular ROS levels post-UVA irradiation
Next, we sought to determine whether autophagy contributes to restore redox balance after UVA stress. To this end, intracellular ROS were evaluated by CM-H 2 DCFDA live imaging. After UVA irradiation, ROS remained at physiological level in SCR as well as antioxidant pretreated LSCs compared to non-irradiated controls (10.5 ± 1.3 a.u. after UVA and 15.1 ± 3.5 a.u. with antioxidant pre-treatment compared to 29.5 ± 2.5 a.u. in controls, p>.05) (Fig 3). In contrast, ATG7KD LSCs showed a significant increase of intracellular ROS upon UVA exposure compared to non-irradiated counterparts (70.1 ± 9 a.u. in ATG7KD and 130.9 ± 25.4 a.u. after UVA p < .05) as well as irradiated, autophagy-competent LSCs (12.5-fold increase, p < .01). Conversely, treating ATG7KD LSCs with antioxidant mixture 16 hours prior to irradiation largely alleviated the accumulation of intracellular ROS in ATG7KD LSCs (19.8 ± 8.7 a.u. versus 130.9 ± 25.4 a.u. in UVA-treated ATG7KD LSC, p < .05). These data suggest a crucial role of autophagy in regulating redox balance in LSC's stress response to UVA.
Autophagy facilitates nucleocytoplasmic transport of PAX6 in response to UVA-induced ROS To determine whether our model is suitable to study PAX6 response to oxidative stress, colonies were challenged with 20 μM H 2 O 2 and subcellular expression of PAX6 was examined by immunofluorescence. Notably, LSC colonies displayed high levels of nuclear PAX6 expression on clonal day 12 (70.9 ± 1.5%) (Fig 4A & 4B). H 2 O 2 -induced oxidative stress resulted in nuclear depletion of PAX6 (10.8 ± 2.9% PAX6 + nuclei, p < .01 compared to controls), validating our in vitro ROS-responsive PAX6 model. Since we found that autophagy balances UVAinduced oxidative stress, we next asked whether PAX6 was differentially regulated in autophagy-competent and-deficient LSCs confronted with UVA-elicited ROS. To this end, the subcellular localization of PAX6 was studied in UVA-stressed SCR and ATG7KD LSCs by immunofluorescence. Interestingly, the percentage of LSCs expressing nuclear PAX6 + was significantly decreased in SCR LSC colonies after UVA exposure (8.0 ± 1.6%) compared to nonirradiated SCR LSCs (22.1 ± 1.8%, p < .01) (Fig 4C & 4D). In contrast, ATG7KD LSCs retained baseline levels of PAX6 + nuclei after irradiation (35.3 ± 7.5% in ATG7KD versus 33.2 ± 3.1% after UVA, p>.05). Of note, the reduction of nuclear PAX6 + cells in irradiated SCR LSCs correlated with the increase of cells with cytoplasmic PAX6 expression. In addition, antioxidant pre-treatment abolished the UVA-triggered PAX6 nuclear depletion observed in SCR LSCs evidenced by 26.6 ± 1.3% of nuclear PAX6 + cells in irradiated SCR LSCs (p < .01 compared to 8.0 ± 1.6% in irradiated LSCs). Collectively, data presented here suggest: (1) Nuclear-to-cytoplasmic protein transport, rather than transcriptional repression or acceleration of cytoplasmic PAX6 protein degradation, is the dominant mechanism by which UVA induces nuclear reduction of PAX6. (2) The UVA-induced PAX6 nucleocytoplasmic transportation is both ROSand autophagy-dependent.

Discussion
In current study, we identified autophagy as an UV stress sensor in LSCs. Dual functional roles for autophagy in cellular stress response were further suggested by alteration of redox status and cell cycle progression in UVA-irradiated LSC. Activation of autophagy in LSC contributes to restoration of intracellular redox balance and adaptive cell cycle response. We propose a novel PAX6-p21 molecular mechanism underpinning the autophagy-mediated stress response (Fig 10).
Autophagosome accumulation has been reported in the UV stress response of epidermal keratinocytes [21], whether autophagic flux is increased or decreased remains undetermined. Increased number of autophagosomes in stressed cells may result from elevated autophagic flux or impaired lysosomal degradation of autophagosomes. Our autophagic flux experiment suggests the autophagosome accumulation observed in UVA-irradiated LSCs, corneal keratinocyte progenitors, resulted from increased autophagic flux. Zhao et al. show that autophagy exerts a cytoprotective effect by selective degradation of UV-oxidized macromolecules [21]. Here, we identified autophagy as an alternative antioxidant mechanism in LSCs to reduce ROS elicited by UVA (Fig 3). Whether the autophagy-mediated redox balance contributes to reduce UVA-induced oxidative damage in LSCs requires further investigation. While it is known that high dose irradiation induces cellular apoptosis, physiological effects of low dose UVA radiation on LSCs remains unclear. Using a non-lethal dose of UVA irradiation, we found that stressed LSCs employ autophagy as a molecular machinery to curb cell cycle (Figs 5 & 8). Interestingly, low level of ROS was identified as a key signaling event in the autophagy-mediated cell cycle response, since antioxidant pretreatment was found to abrogate the UVA-induced cell cycle arrest. Impairing autophagic machinery by ATG7KD resulted in excessive ROS accumulation and failure of cell cycle response in UVAstressed LSCs. Paradoxically, high level of ROS in autophagy-deficient LSCs failed to induce cell cycle arrest (Figs 3, 5 & 8). Therefore, ATG7-mediated autophagy is intrinsic to redox balance and cell cycle response in LSC's stress response to mitigate UVA-elicited oxidative stress.
PAX6 deregulation and oxidative stress have been identified as early pathogenic events in various ocular surface diseases [17,[26][27][28][29]. A direct molecular interaction between PAX6 and ROS is suggested by rapid nucleocytoplasmic translocation of PAX6 in hydrogen peroxidetreated corneal cells [19]. Data presented in current study further suggest a functional role for the ROS-driven PAX6 mobilization in stressed corneal cells, i.e. mediation of cell cycle response. In our experiment, PAX6 and PCNA were found mostly co-localized in the nuclear compartment of proliferative LSCs in unstressed conditions (Fig 5). In contrast to the cytostatic growth effect in differentiated corneal cell [30], nuclear PAX6 at physiological dose in non-irradiated clonal cells might have a stimulatory effect in cell cycling. Following cytoplasmic relocation of PAX6, we observed a cell cycle arrest via p21-mediated PCNA downregulation (Figs 8 & 9). PAX6 is known to differentially regulate cell cycle via various mechanisms. For instance, PAX6 expression has been associated with proliferation and cell cycle progression of colorectal cancer cells, non-small cell lung carcinoma cells and breast cancer cells [31][32][33]. A mechanistic for the PAX6-driven cell cycle progression is recently suggested by a recent work of Li et al. [34]. In their work, Li and colleagues reported that lentiviral overexpression of PAX6 in retinoblastoma cells results in downregulation of p21 and impairs the p53-mediated cell cycle arrest response through reducing p53-p21 molecular interaction. Although the effect of PAX6 on cell cycle progression observed in their study is in line with ours, the underlying mechanism for PAX6-driven cellular proliferation might differ. As a mutually exclusive expression of PAX6 and p21 in the nuclear compartment was observed during LSC's stress response in our work (Fig 8), we postulate de-repression of cell cycle inhibitor p21 by nuclear export of PAX6 as the mechanism for UVA-induced cell cycle arrest. In contrast, PAX6 downregulation was associated with cell cycle progression and proliferation of corneal epithelial cells in PAX6 +/corneal epithelia [35]. Genetic depletion of Pax6 in murine neuroepithelium led to clonal expansion of progenitor pool by downregulating p21 [36]. Li et al. showed that EGF induced cellular proliferation via PAX6 downregulation [37]. Furthermore, Ouyang et al. demonstrated that PAX6 overexpression induced cell cycle exit in proliferative corneal epithelial cells [30]. The differential cell cycle regulation between our study and the others suggest a stress context-dependent PAX6 function.
Recent epidemiological studies suggest associations of solar UV radiation and PAX6 deregulation in ocular surface diseases [26,[38][39][40][41][42]. Our data suggest that PAX6 mediates LSC's UVA stress response by regulating cell cycle. Although the functional significance of cell cycle arrest via nucleocytoplasmic PAX6 relocation in LSC's stress response remains unclear, we surmise that pausing cell cycle might contribute to stem cell's self-repair and autophagic clearance of UV-induced cellular damage. Interestingly, PAX7, a PAX family member expressed in muscle precursor cells, directs in ovo myogenesis by regulating proliferation of UV-stimulated myoblasts [43,44]. To attest whether ocular-resident PAX genes, i.e. PAX6, regulate UVA cell cycle response of LSCs, we expressed PAX7 in nuclear PAX6-depleted LSCs in UVA stress context. In contrast to adenoviral PAX6 overexpression experiments, forced PAX7 expression did not affect autophagy-mediated UVA cell cycle response (Fig 6). Based on these findings, we conclude that the "UVA-autophagy-PAX-cell cycle" axis is specific to PAX6 in ocular stem cells. While it is surmised that PAX7 cannot replace PAX6 to orchestrate complex ocular transcriptional network, the detailed mechanism for PAX6-specific cell cycle regulation in UVAstressed LSCs warrants further studies.
It is known that cells employ master transcription factors, such as p53 and Nrf2, to orchestrate stress responses, including apoptosis, cell cycle regulation and antioxidant defense to repair oxidative damage. The transcriptional activity of stress response transcription factors is controlled by dynamic subcellular localization with autophagy lately identified as a new regulatory mechanism [45][46][47]. Two mechanisms by which ROS drive PAX6 nuclear-to-cytoplasmic trafficking are oxidative modification of its nuclear export signal and oxidation of its adaptor proteins [48,49]. In the current study, we showed that low dose of ROS in autophagy-competent LSC induced cytoplasmic relocation of PAX6, while excessive levels of ROS in autophagydeficient LSCs failed to drive the relocation (Figs 3 & 4). These findings clearly indicate an autophagic regulation of PAX6' ROS response. Although physiological protein turnover of PAX6 mainly depends on proteosomal degradation [50], we do not rule out the potential of synergistic (or alternative) autophagic clearance of PAX6 in UVA stress context. Autophagy was reported to navigate intracellular trafficking of PAX6 by indirectly degrading PAX6 adaptor proteins, such as TRIM44 and SPARC [51][52][53][54]. Whether or not autophagy responds to ROS stress by selectively targeting PAX6 adaptor proteins to regulate cell cycle response requires further investigations.
In conclusion, our data indicate that autophagy mediates cell cycle response of LSCs following UVA exposure via a novel PAX6-p21 molecular cascade. The spatial dynamics of PAX6 recruitment to nucleocytoplasmic compartments represent a novel post-translational regulation for autophagy-mediated oxidative stress response. Functional enhancement of autophagy might thus represent a novel therapeutic strategy in treating UVA-associated, PAX6-deregulated ocular surface degenerative diseases.