Aquaporin-3-Dependent Cell Migration and Proliferation during Corneal Re-epithelialization

Abstract

purpose. To determine a role for the water- and glycerol-transporting protein aquaporin-3 (AQP3) in mammalian corneal epithelium, where it is expressed but has no known function.

methods. Corneal epithelial water and glycerol permeabilities were measured in living wild-type and AQP3-null mice using calcein fluorescence-quenching and ¹⁴C-glycerol-uptake assays, respectively. After removal of the corneal epithelium by scraping, re-epithelialization was followed by fluorescein staining. The contribution of AQP3-facilitated cell migration to corneal re-epithelialization was assessed using an organ culture model, in which initial resurfacing results from epithelial cell migration, as shown by BrdU analysis and 5-fluorouracil insensitivity, and by scratch wound assay using primary cultures of corneal epithelial cells from wild-type versus AQP3-null mice. Involvement of AQP3 in epithelial cell proliferation was investigated by morphometric and BrdU analysis of histologic sections, and by measurement of [³H]thymidine uptake in primary cultures of corneal epithelial cells.

results. AQP3 deficiency did not alter corneal epithelial thickness, morphology, or glycerol content, though both water and glycerol permeabilities were reduced. Time to corneal re-epithelialization in vivo was significantly delayed in AQP3-null mice compared to wild-type mice. Delays were also found in organ and primary cultures, demonstrating a distinct defect in cell migration arising from AQP3 deletion. Delayed restoration of full-thickness epithelia of AQP3-null mice over days after scraping suggested a separate defect in epithelial cell proliferation, which was confirmed by reduction in proliferating BrdU-positive cells in AQP3-deficient mice during healing, and by reduced proliferation in primary cultures of corneal epithelial cells from AQP3-null mice.

conclusions. The significant impairment in corneal re-epithelialization in AQP3-deficient mice results from distinct defects in corneal epithelial cell migration and proliferation. The results provide evidence for involvement of an aquaporin in cell proliferation and suggest AQP3 induction as a possible therapy to accelerate the resurfacing of corneal defects.

Aquaporins (AQPs) comprise a family of small transmembrane proteins that transport water, and in some cases, both water and small solutes such as glycerol. Functional studies on AQP knockout mice have revealed the importance of high transcellular water permeability in facilitating osmotically driven epithelial fluid transport, as with AQP1 and AQP3 in the kidney, ^{1 2} and in rapid near-isosmolar transport, as with AQP5 in salivary and submucosal glands, AQP1 in choroid plexus, and AQP1 and AQP4 in ciliary epithelium. ^{3 4 5 6} In addition to these classical functions of AQPs, recent phenotype analysis of knockout mice has revealed several unexpected cellular roles. For example, AQP1 facilitates tumor angiogenesis by enhancing cell motility by a mechanism that may involve facilitated water transport at the leading edge of migrating cells. ⁷ AQPs that transport both water and glycerol are involved in skin hydration and biosynthesis (AQP3) and fat metabolism (AQP7) by regulation of cellular glycerol content. ^{8 9} In skin, AQP3 deletion also slowed wound healing and recovery of barrier function after stratum corneum removal. ¹⁰  

The stratified corneal epithelia of mouse, rat, and human express the water-selective aquaporin AQP5, and the water- and glycerol-transporting aquaglyceroporin AQP3. ^{11 12 13} We previously found reduced transcorneal water permeability in mice lacking AQP5. ¹³ The function of AQP3 in the cornea is unknown. Maintenance of the corneal epithelial layer is crucial to providing a smooth and transparent refractive surface and a barrier to infection, requiring continued regeneration to replace normal epithelial cell loss from the surface. ¹⁴ Limbal stem cells located between the cornea and conjunctiva give rise to a single layer of centripetally migrating, mitotically active columnar basal cells that adhere to a basement membrane. These transient amplifying cells undergo several rounds of cell division before producing an intermediate layer of suprabasal wing cells one to three cells thick and finally a superficial layer of terminally differentiated squamous cells two to four layers thick. 

Corneal epithelial cell renewal is greatly increased during wound healing. Whereas suprabasal and basal epithelia of normal mouse cornea migrate centripetally at an average linear rate of 0.7 to 1.0 μm/h, ^{15 16} the marginal cells bordering a defect can migrate at 30 to 60 μm/h. ¹⁷ There is an extensive body of literature on the biology of corneal epithelial regeneration based on wound-closure models (reviewed in Ref. ¹⁸ ). Corneal epithelial replacement involves three distinct phases: the latent phase, when cells clear debris at the wound edge and alter their metabolic status (∼6 hours); the cell migration–adhesion phase, involving increased protein and macromolecule synthesis and glycogen utilization (up to 24 hours); and the cell proliferation phase (days). Recurrent or persistent corneal erosions in humans generally arise from trauma or various forms of epithelial basement membrane dystrophy, and may result in ulceration or perforation of the underlying stroma with associated pain and visual impairment. ^{19 20}  

We hypothesized that aquaporins might be involved in one or several aspects of corneal epithelial regeneration. In this study, we demonstrated the functional expression of AQP3 in corneal epithelium of mice. Significant impairment in corneal re-epithelialization was found in AQP3-null mice using an established mouse model of corneal epithelial removal, which was evaluated mechanistically by studies of corneal epithelial cell migration and proliferation in organ and primary cell cultures. Our results implicate AQP3 in the two processes fundamental to wound healing: cell migration and proliferation. 

Materials and Methods

Mouse Preparation

Transgenic mice deficient in AQP3 in a CD1 genetic background were generated by targeted gene disruption as described. ² Mice were bred and cared for at the University of California, San Francisco, Animal Facility. For each experiment, wild-type and knockout mice were matched by age and weight (6–9 weeks, 22–30 g). Investigators were blinded to mouse genotype in all functional studies until completion of data analysis. Protocols were approved by the University of California, San Francisco, Committee on Animal Research, and were in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Mice were anesthetized using 125 mg/kg 2,2,2-tribromoethanol (Avertin; Sigma-Aldrich, St. Louis, MO) intraperitoneally, and the dose was supplemented during experiments to maintain deep anesthesia. Core temperature was monitored with a rectal probe and maintained at 37 ± 1°C with a heating pad. For all maneuvers, mice were immobilized with the cornea under study oriented to face upward in a custom-built stereotaxic device with a rotating jaw clamp. After experiments, mice were killed by an overdose of the anesthetic and cervical dislocation, and whole eyes were enucleated with forceps. 

Models of Corneal Re-epithelialization

In vivo and organ culture models of mouse corneal re-epithelialization used the same wounding procedure. After anesthesia and topical proparacaine (0.5%; Akorn, Buffalo Grove, IL), the cornea was blotted dry with a surgical sponge (Medtronic, Chicago, IL). A 2.3-mm diameter region of central corneal epithelium was demarcated with a surgical trephine (Roboz, Gaithersburg, MD) under observation with a stereo epifluorescence microscope (SMZ1500, 1× objective, 2.8× zoom; Nikon, Tokyo, Japan) and the full-thickness corneal epithelium was mechanically removed with a number 69 Beaver blade (BD Biosciences, Franklin Lakes, NJ) and standard scraping procedures, without damage to the basement membrane. ²¹ Corneas were allowed to resurface for up to 48 hours, with epithelial defect size monitored with fluorescein staining (0.1% in PBS) just after scraping and 12, 18, and 24 hours later. Fluorescence was imaged using a cooled CCD camera (CoolSnap HQ; Photometrics, Tuscon, AZ) and two-dimensional projections of relative wound area were quantified using ImageJ software (available by ftp at zippy.nimh.nih.gov/ or at http://rsb.info.nih.gov/nih-image/ National Institutes of Health, Bethesda, MD). 

For studies of healing kinetics in vivo, tobramycin ointment (0.3%; Alcon, Fort Worth, TX) was applied to wounded eyes after scraping. Mice were then returned to their cages and allowed to awaken. Each subsequent wound area measurement was performed under light anesthesia followed by recovery. For organ culture studies, eyes were scraped, excised, and incubated as described. ^{22 23} After rinsing in PBS, each enucleated eye was placed in a well of a 24-well culture plate and immersed fully in 1 mL Dulbecco’s modified Eagle’s Medium (DMEM) with 25 mM glucose (Invitrogen-Gibco, Rockville, MD), and supplemented with 2% fetal bovine serum (FBS), penicillin G (100 U/mL), and streptomycin (100 μg/mL). Eyes were kept in a tissue culture incubator (37°C, 5% CO₂) for up to 30 hours. The left eyes served as the control, and the right eyes were incubated in some studies in culture medium supplemented with 10 μg/mL of 5-fluorouracil (5-FU) or paclitaxel (Sigma-Aldrich, St. Louis, MO), prepared at 2000× stocks in DMSO. 

AQP3 Immunodetection

Eyes were fixed for histology by immersion in 10% neutral buffered formalin (Accustain; Sigma-Aldrich) for 24 hours. The fixed tissue was processed with xylenes and graded ethanols and embedded in paraffin, and 5-μm sections were cut through the central cornea and optic nerve (Histoserv Inc., Germantown, MD). Sections were deparaffinized (Citrisolv; Fisher Scientific, Pittsburgh, PA), rehydrated in a series of graded ethanols, and then either stained with hematoxylin and eosin (H&E) or processed for immunohistochemistry. Slides were treated for epitope retrieval in citrate buffer (10 mM sodium citrate and 0.05% Tween 20) for 30 minutes at 95°C and while cooling, and then sections were hydrogen peroxide quenched (3% H₂O₂). After blocking with goat serum, sections were incubated with a rabbit anti-AQP3 polyclonal antibody (1:500; Chemicon, Temecula, CA) and washed in PBS. Bound antibody was detected using the rabbit avidin-biotin complex (ABC) kit (Vectastain; Vector Laboratories, Burlingame, CA) and developed using the substrate 3,3-diaminobenzidene. Photographs were taken on an upright microscope (model DM4000B; Leica, Solms, Germany) equipped with a cooled CCD camera (Spot; Diagnostic Instruments, Sterling Heights, MI). 

For immunoblot analysis, corneal epithelia of anesthetized mice were scraped using sterile Beaver blades and pooled (2–4 eyes/sample) in extraction buffer containing 250 mM sucrose, 10 mM EDTA, and 1% protease inhibitor mix (Sigma-Aldrich). Cells were mechanically disrupted using an insulin syringe, and samples were loaded on a 4% to 12% SDS polyacrylamide gel (3 μg/lane). Protein was transferred to a polyvinylidene difluoride (PVDF) membrane and incubated with rabbit anti-AQP3 antibody (1:1000) followed by anti-rabbit IgG horseradish peroxidase-linked antibody (1:10,000; GE Healthcare, Piscataway, NJ), and visualized using enhanced chemiluminescence (GE Healthcare). 

Transmission Electron Microscopy

Freshly enucleated globes were immersed in 1.2% paraformaldehyde and 0.8% glutaraldehyde in 0.1 M Sorensen buffer (pH 9.2) for 90 minutes. Corneas were dissected from globes, postfixed in 1% osmium tetroxide in sodium veronal buffer for 1 hour, dehydrated in graded ethanols, and embedded in Araldite epoxy resin. Ultrathin sections (70–100 nm) were stained with aqueous saturated uranyl acetate and Reynold’s lead citrate and screened at 2,000× to 30,000× magnification using an electron microscope (1200 EX; JEOL, Tokyo, Japan) operating at 80 kV. Also, 1-μm thick sections were cut and stained with Trump’s toluidine blue for orientation under the light microscope. Ultrastructure was compared by an observer blinded to genotype information. 

In Vivo Water and Glycerol Permeability Assays

Osmotic water permeability was measured using a calcein-quenching method. ¹³ Briefly, epithelial cells of anesthetized mice were loaded with calcein by exposure of the cornea for 30 minutes to 25 μL isosmolar PBS containing 10 μM calcein-AM (Invitrogen-Molecular Probes, Eugene, OR). A custom-built 33-μL microchamber with <50 ms solution exchange time was then positioned over the cornea for continuous measurement of cell calcein fluorescence under suction perfusion with solutions of specified osmolarities. The time course of fluorescence in response to solution osmolarity changes, F(t), was fitted to a single exponential time constant, τ: F(t) = A + Be ^{−t /τ}, where A and B are related to system sensitivity and background signal. 

For measurements of glycerol permeability in vivo, the corneal epithelium of anesthetized mice was exposed to 15 μL isosmolar PBS containing 1 mM glycerol (Fluka, Buchs, Switzerland) and 30 μCi/mL [¹⁴C]glycerol (specific activity 146 mCi/mmol; GE Healthcare) for 0 to 20 minutes. After anesthetic overdose, the ocular surface was washed with the same solution at 4°C (lacking radioactive glycerol), blotted with a surgical sponge, and scraped within the limbus to collect full-thickness corneal epithelium. Cell-associated ¹⁴C radioactivity was measured by scintillation counting (one eye per sample). Assay of total scraped protein per eye indicated <10% differences from eye to eye. 

Assays of Corneal Epithelial Glycerol and Glycogen Content

Cellular glycerol and glycogen contents were measured using coupled enzymatic assays. After anesthesia and application of topical proparacaine, the ocular surface was blotted with a surgical sponge and the corneal epithelium was scraped. Material from two to four eyes was dissolved in 30 μL of cold PBS (for glycerol) or 50 μL hot 5 N NaOH (for glycogen). Cellular glycerol was assayed by a glycerol kinase absorbance assay (Free Glycerol Reagent; Sigma-Aldrich) and normalized to tissue protein content. Glycogen was extracted in hot base, precipitated in ice-cold absolute ethanol, and hydrolyzed in 0.6 N HCl based on methods established for rabbit corneal epithelium. ^{24 25} Liberated glucose was measured using a glucose oxidase absorbance assay (Wako Chemicals, Neuss, Germany) and expressed relative to protein measured in the supernatant collected during ethanol precipitation. Protein concentration was measured using a DC protein assay kit (Bio-Rad, Hercules, CA). 

Primary Culture of Mouse Corneal Epithelial Cells

Primary cultures of corneal epithelial cells from wild-type and AQP3 knockout mice were grown on either tissue culture plastic or on plastic coated with fibronectin (5 μg/mL; Roche Diagnostics, Indianapolis, IN) in supplementary hormonal epithelial medium (SHEM) according to the method of Kawakita et al. ²⁶ SHEM consisted of equivolume HEPES-buffered DMEM and F12 medium, containing 10 ng/mL mouse-derived EGF, 5 μg/mL insulin, 5 μg/mL transferrin, 5 ng/mL sodium selenite, 0.5 μg/mL hydrocortisone, 10⁻¹⁰ M cholera toxin A subunit (all from Sigma-Aldrich), 5% FBS, 50 μg/mL gentamicin, and 1.25 μg/mL amphotericin B. Enucleated eyes of 4- to 8-week-old mice were washed in SHEM and then enzymatically digested for 18 hours at 4°C in SHEM containing 15 mg/mL Dispase II (Roche Diagnostics) and 100 mM d-sorbitol. Each eye was then held under suction at its posterior pole by a Pasteur pipette, and the corneal–limbal epithelial cell sheet was removed intact by gentle shaking in SHEM. Sheets were broken up into single-cell suspensions in Hanks’ balanced salt solution containing 0.05% trypsin and 0.53 mM EDTA (Invitrogen-Gibco) by pipetting for 8 to 10 minutes at room temperature. Cells from eyes of each genotype were pooled, centrifuged (5 minutes at 800g), resuspended in SHEM, counted with a hemocytometer, and seeded in 12-well plates at a density of 6 × 10⁴ cells/cm² (for proliferation studies) or 1 × 10⁵ cells/cm² (for migration studies). Medium was replaced after 24 hours, to remove unattached suprabasal cells. 

Attached basal cells were grown for up to 5 days. To detect the corneal epithelial–specific marker, cytokeratin 12 (K12), cells were fixed with 4% paraformaldehyde, blocked with 1% BSA for 30 minutes, incubated with a rabbit anti-mouse K12 polyclonal primary antibody (1:50; Santa Cruz Biotechnology, Santa Cruz, CA) for 2 hours, and washed with PBS. Antibody binding was detected with a Cy3-conjugated anti-rabbit secondary antibody (1:200, Sigma-Aldrich). Epithelial cell morphology was observed by phase-contrast light microscopy. Samples for AQP3 immunoblotting were collected by cell scraping. 

Cell Proliferation Assay

For in vivo studies, 5-bromo-2′-deoxyuridine (BrdU, 12 mg/mL; Sigma-Aldrich) was injected intraperitoneally (100 mg/kg) 2 hours before euthanasia, after which eyes were processed as just described. For organ culture studies, BrdU was added to culture wells (100 μg/mL final concentration) 2 hours before fixation and staining. For BrdU immunohistochemistry, tissue sections were processed as described earlier, with the following differences: in place of citrate buffer epitope retrieval, sections were treated with 2 N HCl for 1 hour at 37°C and then with 0.1 M sodium borate solution (pH 9) twice for 15 minutes at room temperature. Sections were blocked with rabbit serum and incubated with a rat anti-BrdU monoclonal primary antibody (1:40; Abcam, Cambridge, MA). Bound antibody was detected using the rat ABC kit (Vectastain; Vector Laboratories). BrdU-positive cells were counted from limbus to limbus in a central corneal section for each eye. 

Cell proliferation was measured in primary cultures of mouse corneal epithelium at 5 days after seeding onto uncoated plastic by addition of [methyl-³H]thymidine (2 μCi/mL, specific activity 86 Ci/mmol; GE Healthcare) to cultures for 6 hours. Cells were washed three times with PBS and three times in 1 mL cold 10% trichloroacetic acid, and then solubilized by addition of 750 μL/well of 0.5 N NaOH for 30 minutes at room temperature. The solution was then neutralized by addition of 75 μL of 5 N HCl to each well. ³H radioactivity incorporated in 250 μL sample aliquots was measured by scintillation counting. Total DNA in 20 μL of each sample was assayed by Hoechst 33258 fluorescence (Sigma-Aldrich) in 2 M NaCl, 50 mM Na₂HPO₄, and 2 mM EDTA (pH 7.4) after an 8-hour incubation at room temperature. 

Cell Migration Assay

A scratch-wound–closure assay ^{7 27 28} was used to compare migration in primary cultures of wild-type and AQP3-deficient mouse corneal epithelia plated onto uncoated or fibronectin-coated (5 μg/mL; Roche Diagnostics) wells of 12-well plates. Five days after seeding, when cultures formed confluent monolayers, cells were synchronized by replacing the medium with SHEM containing low serum (1% FBS) and lacking EGF for 10 hours. Monolayers were wounded by two perpendicular linear scratches across each well with a 10-μL pipette tip, to produce 300-μm-wide strips. After unattached cells were washed away, the medium was switched to SHEM containing EGF and 5% FBS. Wounds were photographed immediately by phase-contrast imaging at 10× magnification, and at 18 hours after wounding at four marked regions per well near the crossing point. Wound healing for each culture was reported as the average linear speed of the wound edges over 18 hours. 

Statistics

Data are expressed as the mean ± SE. Significance between experimental groups was determined by a two-tailed Student’s t-test assuming a 95% confidence interval. 

Results

Corneal Epithelial AQP3 Expression

Immunohistochemistry showed AQP3 expression in corneal epithelium of wild-type mice, with no specific staining found in AQP3-null mice (Fig. 1A , top). AQP3 was most abundant in plasma membranes of basal epithelial cells. Immunoblot analysis detected specific bands corresponding to glycosylated and nonglycosylated AQP3 monomers in freshly scraped corneal epithelia of wild-type but not knockout mice (Fig. 1A , bottom). 

The effects of AQP3 on corneal epithelial ultrastructure were assessed by transmission electron microscopy. As shown in Figure 1B(top), full-thickness central corneal epithelia from wild-type and AQP3 knockout mice had the same total thickness (27.0 ± 0.6 vs. 26.7 ± 0.9 μm, ±SE, four eyes per genotype), basal cell height (10.3 ± 0.3 vs. 10.2 ± 0.1 μm); number of cell layers (6–7); and basal, wing, and superficial cell morphology. At higher magnification, comparable densities of desmosomal (Fig. 1B , bottom) and hemidesmosomal (not shown) extracellular communications were seen in basal cell layers of the two genotypes, and no pools of intercellular fluid were seen between AQP3-deficient epithelial cells (Fig. 1B , bottom, insets). These findings suggest grossly normal corneal epithelial cell morphology and attachments in AQP3-null corneas. 

Effect of AQP3 Deficiency on Epithelial Water and Glycerol Transport

Having confirmed expression of AQP3 protein in mouse corneal epithelium, water and glycerol permeabilities were measured in intact corneal epithelia of living mice, to demonstrate functionally significant AQP3 expression in wild-type mice. Osmotic water permeability of corneal epithelial cells was measured by a calcein fluorescence method in which cytoplasmic calcein fluorescence provided an instantaneous readout of cell volume. ¹³ The reversible kinetics of cell swelling is shown in Figure 2A(top) in response to serial perfusion of 470, 310, and 470 mOsM solutions. Osmotically induced cell volume changes were slowed by 2.4 ± 0.3-fold in AQP3-null mice (Fig. 2A , bottom). 

A [¹⁴C]glycerol uptake assay was developed to quantify AQP3-dependent corneal epithelial glycerol permeability in vivo. After anesthesia and immobilization, small aliquots of PBS containing [¹⁴C]glycerol were placed on proptosed eyes for specified times. Full-thickness corneal epithelium was harvested by scraping for measurement of cell-associated [¹⁴C]glycerol radioactivity. [¹⁴C]glycerol uptake was linear for at least 30 minutes in wild-type mice (Fig. 2B , top). [¹⁴C]glycerol uptake measured at 12 minutes was remarkably reduced in AQP3-null mice (Fig. 2B , bottom). However, despite the reduced glycerol permeability, steady state corneal epithelial glycerol content was not significantly affected by AQP3 deletion (Fig. 2C , top). There was a small but significant reduction in glycogen content (Fig. 2C , bottom). 

Effect of AQP3 Deficiency on Re-epithelialization In Vivo

To determine the role of AQP3 in corneal surface re-epithelialization, we used an established in vivo model of corneal epithelial wound healing in which healing was quantified based on fluorescein pooling on the bare stroma, indicating the advancement of marginal basal epithelia at the wound edge. The 2.3-mm-diameter wound size, chosen to avoid damage to the limbus, remained relatively constant for up to 10 hours after wounding (the latent phase) and then decreased from 10 to 24 hours (the migration phase). Re-epithelialization was quantified from the defect area during the time of rapid healing at 12, 18, and 24 hours after scraping (Fig. 3A) . At the 12- and 18-hour time points, a significant delay in resurfacing was found in AQP3-deficient corneas compared with wild-type corneas (Fig. 3B ; 12 hours: 10% ± 4% vs. 28% ± 3% area healed; 18 hours: 36% ± 4% vs. 51% ± 5% healed). AQP3 expression remained approximately constant in wild-type mice during the healing process, as demonstrated by immunoblot analysis and immunohistochemistry at 24 or 48 hours after corneal scraping (data not shown). 

Starting at approximately 24 hours and lasting for several days, the healing response is dominated by cell proliferation and the re-establishment of a multilayered epithelium (the proliferation phase). The re-stratification process was assessed by histology of central corneal sections from eyes fixed at various times after wounding (representative sections shown in Fig. 3C ). In both genotypes, large wounds with well-demarcated leading edges were visible at 12 hours. At 24 and 36 hours, AQP3-deficient corneas showed delayed monolayer surfacing and cell stratification–differentiation compared with wild-type corneas. Periodic acid-Schiff staining of adjacent sections during healing demonstrated comparable glycogen store depletion in basal cells of both genotypes (not shown). Figure 3Dshows incomplete recovery of epithelial thickness in AQP3-null corneas at 24, 36, and 48 hours. In contrast, restoration of full epithelial thickness was observed by 48 hours after wounding in wild-type mice. The central regions of AQP3-null corneas had 0 to 2 cell layers at 24 hours after wounding, compared to three to four layers in wild-type corneas. At 36 and 48 hours, AQP3-null corneas were three to four cell layers thick, less than the six to seven layers (full-thickness) found for wild-type corneas, suggesting impaired proliferation in AQP3 deficiency. 

Defective Migration of AQP3-Null Epithelial Cells in an Organ-Culture Wound-Healing Assay

Delayed re-epithelialization in AQP3-deficient corneas may arise from impairment in cell migration, proliferation, or both. To distinguish between these processes, an organ culture wound-healing model was used that has been shown to represent cell migration. ^{22 23} In situ wounds were made as in the in vivo model, and then eyes were enucleated and cultured in serum-containing medium for up to 30 hours (Fig. 4A , inset). In preliminary experiments, no healing occurred when eyes were incubated in serum-free medium. To characterize the model, globes from wild-type mice were cultured in medium containing either 5-FU (10 μg/mL), paclitaxel (10 μg/mL), or dimethyl sulfoxide (DMSO) vehicle (Fig. 4A) . 5-FU impairs cell proliferation but not migration in cornea and other tissues, whereas paclitaxel inhibits both processes. ^{29 30 31 32} In this model, epithelial resurfacing occurred in the control group at a slightly slower rate than in vivo. Paclitaxel, but not 5-FU, significantly slowed corneal resurfacing at 18, 24, and 30 hours, supporting the utility of organ culture for in vitro assessment of corneal cell migration. 

There was a marked delay in re-epithelialization of AQP3-null versus wild-type corneas (without 5-FU or paclitaxel; Fig. 4B ), in agreement with the in vivo data. To confirm the absence of significant corneal epithelial cell proliferation in this in vitro assay, BrdU staining was performed on corneas at 24 hours (four corneas per genotype). Less than one BrdU-positive cell on average was found in an entire section of cornea from wild-type or AQP3 knockout mice (data not shown). In addition, whereas the full-thickness, multilayered morphology observed in uninjured epithelia (Fig. 4B , inset a) was restored after 24 hours of healing in vivo (Fig. 4B , inset b), organ-cultured eyes were always covered by only one to two layers of epithelial cells at 24 hours, including at and near the unwounded limbus (Fig. 4B , inset c). This observation supports the conclusion that basal cell proliferative capacity is arrested under these organ culture conditions. These experiments thus indicate impaired corneal epithelial cell centripetal migration in AQP3 deficiency. 

Defective Proliferation of AQP3-Null Epithelia after Wounding In Vivo

The potential contribution of AQP3 function to corneal epithelial cell proliferation in vivo was studied by BrdU incorporation. Experiments were performed in nonwounded mice and at different times after wounding. Figure 5Ashows representative images of midperipheral cornea (well outside of the advancing wound margin) stained for BrdU. No significant difference in BrdU staining was observed between the genotypes in the absence of wounding, with a greatly reduced number of BrdU-positive cells in AQP3-null and wild-type peripheral corneal epithelium at 12 hours after wounding (data not shown). As summarized in Figure 5B , at 24 hours AQP3-null corneas showed greatly diminished basal cell proliferation compared with wild-type corneas. A similar conclusion was reached when central, midperipheral, and peripheral areas of the corneal epithelium were counted separately (not shown). 

Defective Proliferation and Migration of AQP3-Null Epithelial Cells in Primary Culture

The role of AQP3 in corneal epithelial cell proliferation was further examined in a recently developed mouse primary cell culture model. ²⁶ Intact corneal–limbal epithelial sheets from wild-type and AQP3 knockout mouse eyes were isolated by Dispase II digestion, and nonpassaged cells were cultured on plastic in SHEM medium. Substrate attachment of wild-type and AQP3-deficient cells was comparable, as measured by cell counting at 24 hours after seeding both from trypsinized cultures using a hemocytometer (7.7 ± 0.3 × 10⁴ wild-type vs. 7.7 ± 1.4 × 10⁴ AQP3-null cells per well, ±SE; three wells per genotype) and from imaging randomly selected 10× fields in undisturbed cultures (92 ± 3 wild-type vs. 89 ± 5 AQP3-null cells/field, ± SE; nine wells per genotype). Approximately 30% of plated cells adhered to tissue culture plastic, as was reported previously. ²⁶ Cells from both genotypes grew similarly as monolayers of mostly large basal epithelial cells (Fig. 6A , top), which began to stratify at areas of confluence, as appreciated by altering the microscope’s focus. Consistent with the earlier report, larger (basal) cells showed cytoplasmic immunostaining for the corneal epithelial-specific marker, K12 (Fig. 6A , middle, left). Immunoblot analysis showed AQP3 expression in cell cultures from wild-type but not AQP3-null corneas (Fig. 6B , middle, right). 

Cell proliferation was measured after 5 days of culture, when cells were 50% to 70% confluent. Under maximum stimulation in the presence of serum and growth factors, AQP3-null cell cultures incorporated 3.8 ± 0.7 times less [³H]thymidine than cultures from wild-type corneas (Fig. 6A , bottom). Cultures from wild-type corneas were found by Hoechst fluorescence to contain approximately two times more DNA (per well) than AQP3-null cultures, and qualitatively approached confluence 20% to 30% faster in all six separate sets of primary cell cultures prepared for these studies. These results support defective corneal epithelial cell proliferation in AQP3 deficiency, which probably accounts for the delayed restoration of full-thickness cornea in vivo over the days after scraping. 

Cell migration in confluent primary cultures of mouse corneal epithelia was measured using a standard wound scratch assay. Cultures grown on fibronectin-coated wells showed accelerated wound closure rates compared to cultures grown on uncoated plastic, in agreement with previous studies. ^{28 33} In vitro wound healing was thus quantified in fibronectin cultures after a relatively short (18-hour) healing period to avoid confounding effects of cell proliferation on healing. Figure 6Bshows delayed wound closure in AQP3-deficient corneal epithelia at 18 hours after creation of linear wounds (4.5 ± 0.2 μm/h for wild-type cells vs. 2.8 ± 0.1 for AQP3-null cells, ± SE; four wells per genotype), supporting the conclusion from in vivo and organ culture experiments that AQP3 deletion impairs corneal epithelial cell migration after wounding. 

Discussion

This study demonstrates AQP3-facilitated water and glycerol transport in mouse corneal epithelium and provides evidence for distinct roles of AQP3 in corneal epithelial cell migration and proliferation during re-epithelialization. Migration and proliferation are fundamental to normal corneal epithelial cell turnover (taking place over weeks) and become greatly accelerated during wound healing. The expression of AQP3 in basal cells of the corneal epithelium suggested a role in one or both of these processes. Defective corneal resurfacing over the first 24 hours after wounding implicated a role for AQP3 in cell migration, which was proved using an organ culture model of wound healing in which cell proliferation was confirmed to be essentially absent by BrdU staining, epithelial morphology, and 5-FU insensitivity and by using an vitro wound-closure assay comparing corneal epithelial cultures from wild-type and AQP3-null mice. 

A separate defect in cell proliferation during re-epithelialization was discovered in AQP3-null corneal epithelium, providing evidence for a previously unrecognized role for an aquaporin in cell proliferation. Defective corneal epithelial cell proliferation was shown by a reduced number of BrdU-positive cells at 24 hours after wounding, and substantial slowing of restratification. Wild-type corneal epithelia achieved their normal thicknesses and cell counts by 48 hours, whereas AQP3-deficient epithelia were delayed by both measures. AQP3-deficient corneal epithelia cells thus manifest impaired cell movement and DNA synthesis more slowly than wild-type cells after wounding. This conclusion is supported by the slowed cell proliferation measured in AQP3-null primary cultures of corneal epithelial cells. 

The AQP3-dependent effects on cell migration and proliferation could each, in principle, be explained by deficiencies in water and/or glycerol transport, or by a mechanism independent of transporting functions. The migration phase of re-epithelialization involves lamellipodial and filopodial extension by marginal cells at the wound’s leading edge. ¹⁷ Such protrusions and the rate of cell migration are reduced in a variety of aquaporin-deficient cells, ^{7 27 34} suggesting a role for local water transport at the leading edge of migrating cells. Corneal epithelial cell migration also requires mobilization of energy stores, particularly glycogen. ³⁵ Defective glycerol transport in AQP3 deficiency may impair glycogen synthesis or utilization by direct or indirect effects on glycolysis. ⁹ However, basal epithelial glycerol content was similar in wild-type and AQP3-null mice, and although glycogen stores were slightly diminished in AQP3-null mice, periodic acid-Schiff staining of corneal sections during healing showed no evidence of differential basal cell glycogen-utilization kinetics compared with wild-type mice. 

The involvement of AQP3 in cell proliferation may be related to its glycerol-transporting function. AQP3-mediated glycerol transport was found previously to be important for lipid biosynthesis in skin, ⁸ as glycerol is the backbone of phosphoglyceride, a major phospholipid in plasma membranes. The role of glycerol in corneal metabolism remains unexplored. Glycerol is a common ingredient in eye drop formulations, primarily because its humectant properties promote ocular surface hydration. ³⁶ Alternatively, the effect of AQP3 on cell proliferation may not involve glycerol transport, but instead AQP3 protein–protein interactions or modulation of membrane physical properties. 

In conclusion, the involvement of AQPs in cell migration appears to represent a general phenomenon with an exact mechanism that remains to be elucidated, but is probably dependent on water movement. The involvement of aquaglyceroporins, such as AQP3, in cell proliferation may also be a general phenomenon, which may be related to AQP3-dependent glycerol transport or to some nontransporting role of AQP3. 

 Supported by National Eye Institute Grants EY13574; National Heart, Lung, and Blood Institute Grants HL73856 and HL59198; National Institute of Diabetes and Digestive and Kidney Diseases Grants DK35124 and DK72517; National Institute of Biomedical Imaging Grant EB00415; and Drug Discovery and Research Development Program grants from the Cystic Fibrosis Foundation. MHL was an MD-PhD student supported by a National Institute of Health Medical Scientist Training Grant.

 Submitted for publication March 28, 2006; revised May 12, 2006; accepted August 1, 2006.

 Disclosure: M.H. Levin, None; A.S. Verkman, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Corneal epithelial AQP3 expression and morphology. (A) AQP3 protein expression in mouse corneal epithelia. Top: immunohistochemistry of sections through central cornea with AQP3 plasma membrane staining indicated (arrowheads). Bottom: immunoblot of corneal epithelia from mice of indicated genotypes. (B) Transmission electron microscopy representative of four corneas from four mice of each genotype. Top: full-thickness central corneal epithelium showing six to seven cell layers. Bottom: basal cells with adjacent cell nuclei (*) and desmosomes (arrowheads). Insets: magnified regions of intercellular desmosomal attachments. Scale bar: (A) 20 μm; (B, top) 5 μm; (B, bottom) 1 μm; (B, insets) 0.5 μm.

Figure 2.

 

Water and glycerol permeabilities of corneal epithelium. (A) Osmotic water permeability of corneal epithelial cells in mice in vivo, where the ocular surface was exposed to osmotic gradients. Top: representative time course of cellular calcein fluorescence, reflecting changes in cell volume, in response to changes of perfusate osmolality between 310 and 470 mOsM. Bottom: summary of averaged reciprocal single exponential time constants (1/τ) fitted to kinetics of calcein fluorescence (±SE, four mice per genotype, with data from each mouse representing averaged data for 6–10 curves; *P < 0.05). (B) [¹⁴C]glycerol uptake in corneal epithelium in vivo. Top: time course of [¹⁴C]glycerol accumulation in wild-type full-thickness corneal epithelium (±SE, four to six eyes at each time point). Bottom: [¹⁴C]glycerol uptake after 12 minutes of exposure to [¹⁴C]glycerol-containing buffer (±SE, six wild-type and four AQP3-null eyes, *P < 0.01). (C) Corneal epithelial glycerol (top) and glycogen (bottom) content (±SE, four samples per genotype pooled from two to three eyes each; *P < 0.05).

Figure 3.

 

Aquaporin-dependent corneal re-epithelialization in vivo after wounding. (A) An epithelial defect (diameter, 2.3 mm) created at the center of corneas (left column) was monitored (right columns). (B) Imaging of corneal resurfacing after wounding. Percent remaining epithelial defects determined by fluorescein staining at 12, 18, and 24 hours after wounding (±SE, 8–14 eyes per genotype at each time point; *P < 0.01, comparing to wild-type). (C) Histology showing unwounded corneas (top row) and wounded corneas after 12, 24, and 36 hours of healing (bottom rows). (D) Averaged central corneal epithelial thicknesses before and at indicated times after wounding (±SE, four corneas from each genotype analyzed in two central sections separated by 50 μm; *P < 0.01).

Figure 4.

 

Corneal epithelial cell migration during wound healing in organ culture. (A) Eyes were enucleated after corneal scraping and cultured (inset) in DMEM containing 2% FBS without or with 10 μg/mL 5-FU or paclitaxel. Average percentages of resurfaced wound quantified by fluorescein staining (±SE, six to seven eyes per group; *P < 0.05). (B) Averaged healing (±SE, eight eyes per group, *P < 0.05). Inset: H & E-stained sections from uninjured wild-type cornea (a) and after 24 hours of healing in vivo (b) versus organ culture (c).

Figure 5.

 

AQP3-dependent corneal epithelial cell proliferation during re-epithelialization. (A) Representative sections of midperipheral cornea showing BrdU-immunoreactive cells (dark brown nuclei) before (top row) and at 24 (middle row) or 48 (bottom row) hours after corneal scraping. (B) Number of BrdU-positive cells distributed from limbus to limbus in uninjured corneal epithelia and during healing (±SE, four corneas per genotype at each time point, *P < 0.01).

Figure 6.

 

AQP3 deficiency impairs cell proliferation and migration in primary cultures of corneal epithelial cells. (A) Primary cultures of mouse corneal epithelial cells. Top: phase-contrast light microscopy of confluent areas of wild-type and AQP3-null cells at 7 days after seeding. Middle, left: K12 immunofluorescence staining of 5-day wild-type culture. Middle, right: AQP3 immunoblot of 5-day wild-type and AQP3-null cultures. Bottom: averaged [methyl-³H]thymidine incorporation normalized to DNA content after 6 hours of incubation in [methyl-³H]thymidine containing medium (±SE, 4 wild-type and 5 AQP3-null 5-day cultures in 12-well plates, *P < 0.01). Representative of two sets of experiments. (B) In vitro wound closure model to study corneal epithelial cell migration. Top: light micrographs of wounded cell monolayers grown on fibronectin-coated wells, showing delayed wound closure at 18 hours in AQP3-deficient corneal epithelia. Dashed lines: wound margin immediately after scraping. Bottom: speed of wound edge in wild-type and AQP3-null cells (±SE, four cultures per genotype; *P < 0.01). Scale bar: (A, top) 30 μm; (A, middle) 10 μm; (B) 50 μm.

The authors thank Liman Qian for mouse breeding and genotype analysis, Vibeke Pederson for processing and imaging tissue for electron microscopy, Mariko Hara-Chikuma for valuable input on AQP3 biology, and Jie Hu for advice on cell culture studies.