Molecular basis for MMP9 induction and disruption of epithelial cell–cell contacts by galectin-3 Free

Epithelial cells at the leading edge of wounds require significant fluidity to change their shape and rearrange their position to assume a migratory phenotype (Jacinto et al., 2001). Re-epithelialization, which begins immediately after tissue injury, is initiated through the release of a number of factors that cause disassembly of the epithelial junctions and contribute to the movement of the epithelial sheet to cover the wounded area (Gipson et al., 1993; Pal-Ghosh et al., 2011). Matrix metalloproteinases (MMPs), a family of Zn²⁺-dependent endopeptidases that have the combined capacity to degrade adhesion molecules on cell surfaces and extracellular matrices, are central to this process. Barely detected in unwounded tissue, their induction during wound healing is thought to play a crucial role in tissue repair (Inoue et al., 1995; Sivak and Fini, 2002). An imbalance in the proper regulation of MMPs contributes significantly to the pathogenesis of chronic wounds (Fini et al., 1996; Toriseva and Kähäri, 2009).

It has been established that the carbohydrate-binding protein galectin-3 is central to re-epithelialization, and high levels of galectin-3 can be observed at the leading edge of migrating keratinocytes, even before active cell migration (Cao et al., 2002; Liu et al., 2012). A mechanism by which galectin-3 promotes keratinocyte cell motility is through association with N-glycan ligands on laminin-332 and α3β1 integrin, resulting in enhanced cellular signaling and the formation of lamellopodia, respectively (Kariya et al., 2010; Saravanan et al., 2009). In cancer cells, galectin-3 also facilitates cell motility by inducing MMPs (Kim et al., 2011; Wang et al., 2012), but the molecular mechanism associated with this process is unknown.

The production and activation of MMPs are tightly regulated by complex mechanisms that include the extracellular MMP inducer CD147 (also known as EMMPRIN and basigin), a glycosylated type I transmembrane protein known to induce cell migration and invasion (Iacono et al., 2007; Toole, 2003). Both homo-oligomerization of CD147 and the addition of complex N-glycans to the protein are crucial for the stimulation of MMP production (Iacono et al., 2007; Sun and Hemler, 2001; Tang et al., 2004; Tang and Hemler, 2004; Yoshida et al., 2000). The fact that glycosylation is required for the induction of MMP expression suggests that this posttranslational modification might serve as a regulatory mechanism for CD147 function (Iacono et al., 2007). Galectin-3 has been recently found to induce clustering of CD147 in retinal pigment epithelial cells (Priglinger et al., 2013), but the biological significance of this interaction has remained elusive. In the current study, we reveal a hitherto unknown function of galectin-3 in destabilizing cell–cell interactions by interacting with and clustering CD147 on the epithelial cell surface. We further demonstrate that galectin-3 and its N-terminal self-associating domain initiates keratinocyte cell–cell disassembly by inducing MMP expression in a CD147-dependent manner. Finally, we show the requirement of galectin-3 for MMP9 induction in an in vivo model of wound healing. This new mechanism is likely to play a key role in destabilizing cell–cell contacts to promote the epithelial rearrangement and cellular plasticity that are associated with cell motility.

To determine whether galectin-3 mediates the biological functions of CD147, we first evaluated the interaction between CD147 and galectin-3. Cell extracts collected from monolayer and stratified keratinocyte cell cultures were chromatographed on a galectin-3 affinity column, and bound proteins were eluted by competitive inhibition with lactose. As shown by western blotting, CD147 bound specifically to galectin-3 (Fig. 1A). The interaction between cell surface galectin-3 and CD147 was verified independently using mass spectrometry (supplementary material Fig. S1). In line with these results, both proteins colocalized to cell membranes throughout the entire stratified epithelium in human corneal donor tissue (Fig. 1B).

Galectin-3 is upregulated during many epithelial processes involving rearrangement and migration, such as wound healing (Cao et al., 2002) and invasion in cancer (Newlaczyl and Yu, 2011). To dissect potential mechanisms underlying the interaction of galectin-3 with CD147 during these processes, we tested the effect of recombinant full-length galectin-3 (rGal-3) in regulating CD147 distribution in monolayer cultures of human keratinocytes. In these experiments, we also took advantage of a dominant-negative inhibitor of galectin-3 lacking the N-terminal domain (rGal-3C). The mutant rGal3C competes with full-length galectin-3; however, because it lacks the N-terminal domain, it is unable to oligomerize (Hirabayashi et al., 2002; John et al., 2003). The addition of exogenous full-length rGal-3 to cultures of human keratinocytes induced CD147 clustering in areas of cell–cell contact and, more remarkably, cell–cell detachment within 30 min following treatment (Fig. 1C,D; supplementary material Movie 1). Another outcome observed following treatment of keratinocytes with rGal-3 was a reduction of CD147 pericellular staining in areas lacking cell–cell contacts (Fig. 1C), suggesting destabilization of CD147 following disruption of epithelial cell–cell interactions. These effects were not observed with the dominant-negative rGal-3C, indicating that oligomerization of galectin-3 is essential to promoting CD147 redistribution and plays a central role in cell–cell detachment. Our results also revealed that cell viability is not affected by treatment with recombinant proteins (Fig. 1E).

To further ascertain the role of galectin-3 in epithelial rearrangement, we tested the effect of rGal-3 on occludin distribution. Occludin, an integral plasma membrane protein, is crucial for maintaining the stability of tight junctions and for regulating actin organization and epithelial migration (Cummins, 2012; Du et al., 2010; Fanning et al., 1998). As shown by immunofluorescence, rGal-3 induced loss of occludin localization within intercellular junctions, concomitant with reorganization of the actin cytoskeleton (Fig. 2A). This effect was dependent on the N-terminal polymerizing domain of galectin-3.

To investigate how galectin-3 regulates cell–cell detachment, the expression of MMP9 was analyzed. MMP9 is the primary metalloproteinase synthesized and secreted by basal cells at the leading edge of migrating epithelium (Sivak et al., 2004). Moreover, MMP9 has been associated with the cleavage of occludin through a CD147-dependent mechanism (Huet et al., 2011). In our experiments, addition of rGal-3 to keratinocytes resulted in increased secretion of MMP9 in a dose- and lactose-dependent manner, concomitant with upregulation of MMP9 mRNA levels (Fig. 2B,C). As shown by gel zymography, rGal-3 did not have a profound effect on the secretion of MMP2. Moreover, incubation of cells with the broad-spectrum MMP inhibitor GM6001 prevented the galectin-3-mediated increase in gelatinolytic activity, reduced the number of cell–cell detachment events, and prevented the loss of occludin staining within intercellular junctions (Fig. 2D,E). These results indicate that galectin-3 initiates cell–cell disassembly by triggering MMP9 expression.

The highly glycosylated species of CD147 are responsible for MMP production through a process that requires homophilic interactions (Iacono et al., 2007; Tang et al., 2004; Tang and Hemler, 2004; Toole, 2004; Yan et al., 2005). However, the underlying molecular mechanism that triggers MMP biosynthesis has not yet been elucidated. Here, we evaluated a potential role for galectin-3 and its N-terminal polymerizing domain in promoting CD147-mediated MMP9 expression. For these experiments, the expression of endogenous CD147 was knocked down by treatment with siRNA (supplementary material Fig. S2A). CD147-specific siRNA reduced the expression of CD147 protein by 63%±3 (mean±s.d.) compared to non-targeting siRNA. As shown by gel zymography and western blotting, knockdown of CD147 markedly reduced the rGal-3-induced MMP9 expression (Fig. 3A,B). Furthermore, dominant-negative rGal3C did not affect the levels of MMP9, indicating that the oligomeric form of galectin-3 is required to induce MMP9 biosynthesis. The fact that CD147 was glycosylated in our experiments was demonstrated, first, by treating cell extracts with PNGase F, an enzyme that cleaves N-linked oligosaccharides, and, second, by treating cell cultures with tunicamycin, an inhibitor of protein N-linked glycosylation (supplementary material Fig. S2B,C).

To further prove that galectin-3 mediates CD147-dependent MMP9 expression, we took advantage of a function-blocking anti-CD147 antibody. Blocking of CD147 function in keratinocytes with serial dilutions of anti-CD147 antibody prevented the galectin-3-mediated induction of MMP9 (Fig. 3C). This effect was not observed when an isotype-matched control antibody was used. Taken together, these data suggest that galectin-3 polymerizes glycosylated CD147 on the cell surface of keratinocytes to induce MMP9 expression and epithelial cell rearrangement.

Previous reports have demonstrated that both galectin-3 and MMP9 are central to processes associated with wound healing and cancer invasion (Cao et al., 2002; Itoh et al., 1999; Sivak et al., 2004; Takenaka et al., 2002). During wound repair, galectin-3 levels increase in migrating epithelia, and re-epithelialization is delayed in mice lacking galectin-3 expression (Cao et al., 2002; Liu et al., 2012). Similarly, expression of MMP9 increases at early stages during human wound healing, and mice lacking MMP9 display delayed healing and re-epithelialization (Kyriakides et al., 2009; Salo et al., 1994). Interestingly, using a debridement wound model in which the corneal epithelium is removed with an Algerbrush®, we have observed sporadic colocalization of galectin-3 and CD147 at sites of cell–cell contact at the leading edge of migrating epithelia (Fig. 4A). Thus, we tested the requirement of galectin-3 for MMP9 induction using Gal3^−/− mice. As expected, corneas of control mice expressing galectin-3 had a substantial amount of MMP9 in the migrating epithelia of healing corneas (Fig. 4B) (Gordon et al., 2011; Sivak and Fini, 2002). By contrast, corneas of Gal3^−/− mice displayed a striking impairment in MMP9 expression, demonstrating that MMP9 biosynthesis in the leading edge of migrating epithelium is regulated by galectin-3.

Although considerable progress has been made in demonstrating the crucial role of proteolysis in allowing movement of epithelial cells through tissue barriers, the molecular mechanisms associated with the initiation of cell rearrangement and protease production are still not fully understood (Murphy and Gavrilovic, 1999). In this study, we identify the carbohydrate-binding protein galectin-3 as a new MMP-dependent initiator of epithelial cell–cell disassembly. Moreover, we provide evidence and a unique mechanism whereby galectin-3 induces MMP9 expression in a manner that is dependent on the interaction with and clustering of the MMP-inducer CD147 on the cell surface.

Galectin-3 is overexpressed in migrating cells during wound healing, and during tumor invasion and metastasis (Cao et al., 2002; Takenaka et al., 2002). Mechanisms of action by which galectin-3 can affect the intrinsic motility of cells during these processes include the remodeling of cytoskeletal elements associated with cell spreading and the binding or upregulation of integrins, a group of heterodimeric cell surface proteins that mediate cell adhesion to the extracellular matrix (Liu and Rabinovich, 2005; Saravanan et al., 2009). Interestingly, published findings support a role for galectin-3 in facilitating epithelial rearrangement before migration. These include rapid translocation of galectin-3 to the leading edge of keratinocytes before active cell migration (Liu et al., 2012) and the ability of this lectin to promote colony dispersion and cell scattering (Saravanan et al., 2009). Our data are consistent with this hypothesis and strongly indicate that enhanced galectin-3 expression destabilizes epithelial cell–cell interactions. Results from this study demonstrate that both the C-terminal carbohydrate-recognition domain and the N-terminal self-associating domain of galectin-3 are involved in the induction of cell–cell detachment. In addition, we show that full-length galectin-3 promotes loss of occludin localization at tight junctions and disorganizes the actin cytoskeleton, suggesting a role for this lectin in regulating epithelial cell plasticity by disrupting cell–cell junctions.

Affinity chromatography and confocal microscopy studies revealed that galectin-3 binds CD147 and promotes its cell surface redistribution in human keratinocytes, in a process that is dependent on the N-terminal polymerizing domain of galectin-3. The fact that galectin-3 interacts with CD147 is consistent with a recent report using retinal pigment epithelial cells (Priglinger et al., 2013), but the precise significance of this association has thus far remained elusive. Our findings reveal a hitherto unknown function of galectin-3 in regulating MMP9 expression through CD147 clustering on epithelial cell surfaces. This conclusion is based on data showing that full-length galectin-3, (1) binds and clusters CD147, and (2) increases both MMP9 secretion, in a dose- and lactose-dependent manner, and MMP9 expression. Moreover, (3) either blocking CD147 expression by using siRNA or CD147 binding by using a function-blocking antibody impairs galectin-3-mediated secretion of MMP9. Importantly, our data demonstrate that abrogation of gelatinolytic activity by use of the broad-spectrum MMP inhibitor GM6001 reduces the number of cell–cell detachment events induced by galectin-3, reinforcing the concept that galectin-3 orchestrates cell–cell disassembly by inducing secretion of MMP9.

CD147 is a widely distributed cell surface glycoprotein highly enriched on the surface of keratinocytes during wound healing and malignant tumor cells (Gabison et al., 2009; Yan et al., 2005). A major function of CD147 is the stimulation of MMP synthesis through self-association involving both heterotypic and homotypic cell–cell interactions, but the mechanism remains largely unclear. It is now established that glycosylation is crucial to CD147 biological function, as it is the highly glycosylated species of CD147 that are responsible for the induction of MMPs and are more susceptible to clustering (Iacono et al., 2007; Tang et al., 2004; Tang and Hemler, 2004). Interestingly, CD147 glycosylation is attributable to N-acetylglucosaminyltransferase V (MGAT5)-generated, β1,6-branched polylactosamine (Huang et al., 2013; Tang et al., 2004), a glycan structure known to act as a binding partner for galectins (Newlaczyl and Yu, 2011). Expression of MGAT5 has been shown to promote continued migration of skin keratinocytes (Terao et al., 2011), and it is now clear that Mgat5^−/− tumor cells are less metastatic in vivo (Lau and Dennis, 2008). Although not within the scope of our current study, it will be exciting to determine whether the effects observed for MGAT5 during wound healing and metastasis are due to its role in mediating CD147 glycosylation and, therefore, facilitate interaction of CD147 with galectin-3 and promote MMP9 secretion.

Based on our findings, we propose a mechanism by which galectin-3 promotes epithelial rearrangements at points of cell–cell contact to provide fluidity of cell movement and facilitate migration by inducing MMP9 (Fig. 4C). Successful tissue repair during wound healing requires coordinated spatiotemporal regulation of MMP9 proteolytic activity. MMP9 is rapidly expressed as an early response to injury, persisting during healing and declining upon re-epithelialization (Inoue et al., 1995). The increased activity of MMP9 observed in chronic diseases is known to correlate with the clinical severity of the disease and its recurrence (Pal-Ghosh et al., 2011; Rayment et al., 2008). Because the N-terminal domain of galectin-3 is sensitive to MMP9 cleavage (Ochieng et al., 1994; Ortega et al., 2005), it is tempting to speculate that MMP9 can act downstream to prevent the biological activities of full-length galectin-3 and provide protection from the recurrence of epithelial disease. A better understanding of these events at the cellular and molecular level will provide opportunities for therapeutic modulation of tissue repair.

Anti-human CD147 antibodies for use in western blotting (clone A-12) and function-blocking assays (clone UM8D6) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-human CD147 antibody for immunostaining (clone HIM6) was purchased from BioLegend (San Diego, CA, USA). Anti-mouse CD147 antibody was purchased from AbDSerotech (Oxford, UK). Anti-MMP9 and anti-occludin antibodies were purchased from Abcam (Cambridge, MA, USA) and Invitrogen (Carlsbad, CA, USA), respectively. Rhodamine–phalloidin and the in situ gel zymography kit were purchased from Molecular Probes (Eugene, OR, USA). rGal-3 and rGal-3C were cloned and expressed as previously reported (Mauris et al., 2013). Briefly, Rosetta E. coli clones carrying the rGal-3 or rGal-3C vectors were selected and grown using LB medium containing ampicillin (100 µg/ml) and chloramphenicol (34 µg/ml). Heterologous protein expression was induced by the addition of 0.3 mM of IPTG and the induced cultures were incubated overnight at 15°C with shaking. rGal-3 and rGal-3C were purified using lactosyl sepharose as previously described (Hsu et al., 2006; Levi and Teichberg, 1981). To eliminate contaminating bacterial endotoxins, galectin-3 and galectin-3C were further purified by polymyxinB affinity chromatography (Sigma-Aldrich, Saint Louis, MO, USA). The absence of lipopolysaccharide was confirmed using the ToxinSensor™ Chromogenic LAL Endotoxin Assay Kit (GenSript, Piscataway, NJ, USA). Protein solutions were concentrated by filtration (VIVASPIN, Littleton, MA, USA), dialyzed against PBS buffer containing 10% glycerol and stored at −20°C.

Telomerase-immortalized human corneal keratinocytes were grown as previously reported. Briefly, cells were grown in keratinocyte serum-free medium (K-SFM) (Life Technologies, Carlsbad, CA, USA) to confluence (monolayer). For galectin-3 pulldown assays, cells were also stratified by incubation in Dulbecco's modified Eagle's medium (DMEM)/F-12 (Sigma-Aldrich, Saint Louis, MO, USA) supplemented with 10% calf serum and 10 ng/ml epidermal growth factor for 6 to 7 days. Monolayer cultures were treated with 0 to 100 µg/ml of rGal-3 or rGal-3C for 4 or 24 h in the presence of lactose as indicated. For phase-contrast microscopy, cells were imaged with a 20× objective using an inverted microscope (Nikon Eclipse TS100). For quantification of intercellular space, cells were fixed in 100% methanol and then stained with 1% Crystal Violet in 60% methanol. The area of intercellular space was quantified using ImageJ® software (NIH, Bethesda, MD, USA). Healthy human corneal epithelial tissue was collected from donors who underwent LASIK surgery under an Institutional Review Board-approved protocol for discarded tissue and frozen in optimal cutting temperature compound for sectioning using a cryostat.

Six- to eight-week-old, galectin-3 null (Gal3^−/−) mice generated by homologous recombination on a C57BL/6 background (Hsu et al., 1999), as well as age-matched wild-type mice, were used. Two-millimeter corneal wounds were produced on the right eye of each animal by Algerbrush® (Algercompany, Lago Vista, TX, USA) abrasion as described previously (Cao et al., 2002). The corneas were allowed to partially heal in vivo for 16–18 h. The eyes were removed and processed for whole mounts and confocal microscopy as described previously (Pal-Ghosh et al., 2011). All animal procedures in this study were performed in accordance with the Association for Research in Vision and Ophthalmology Resolution on the Use of Animals in Vision Research, the recommendations of the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and approved by Tufts University Division of Laboratory Animal Medicine in Boston, MA, USA; Protocol # B2011-15.

Cells were surface labeled with the Pierce® Cell Surface Protein Isolation kit (Thermo Fisher Scientific, Rockford, IL, USA) following the manufacturer's instructions. Briefly, cells were incubated with 1 mg/ml cell-impermeable sulfo-NHS-SS-Biotin. After washing with TBS, cells were lysed in the presence of Complete™ Protease Inhibitor Cocktail (Roche Diagnostics, Indianapolis, IN, USA). Lysates were incubated with 500 µl NeutrAvidin™ for 1 h at room temperature. Biotinylated proteins were finally eluted with DTT.

A galectin-3 affinity column was prepared by coupling 5 mg of rGal-3 to cyanogen bromide-activated Sepharose 4B (GE Healthcare, Milwaukee, WI, USA) following the manufacturer's instructions. Binding activity of rGal-3 conjugated to beads was assessed by incubation of 200 µg asialofetuin (Sigma-Aldrich, Saint Louis, MO, USA) with 50 µl rGal-3 beads in the presence or absence of 0.1 M lactose for 1 h at room temperature. Asialofetutin was detected on a 10% SDS-PAGE gel by GelCode® Blue Stain (Thermo Fisher Scientific). For galectin-3 pulldown, aliquots of biotinylated protein (50 µl) or cell extracts (50 µg) were incubated with 100 µl rGal-3-conjugated agarose beads in PBS, pH 7.5 for 1 h at 37°C with gentle mixing every 10 min. After washing with PBS, beads were eluted in sequence with 0.5 M sucrose and 0.5 M of lactose. Eluates were then resolved on 10% SDS-PAGE and stained with Coomassie Blue for analysis by MALDI TOF MS (Taplin Biological Mass Spectrometry Center, Boston, MA, USA). Aliquots of the eluate (20 µl) were also analyzed by western blotting using an anti-CD147 antibody.

Protein from cell cultures was extracted using RIPA buffer (150 µM NaCl, 50 µM Tris-HCl, pH 8.0, 1% NP 40, 0.5% deoxycholate, 0.1% SDS) supplemented with Complete™ Protease Inhibitor Cocktail (Roche Diagnostics, Indianapolis, IN, USA). After homogenization with a pellet pestle, the protein cell extracts were centrifuged at 12,000 g for 45 min, and the protein concentration of the supernatant determined using the Pierce BCA™ Protein Assay Kit (Thermo Fisher Scientific). Proteins in cell lysates (20 µg) or cell culture medium (20 µl) were resolved in 10% SDS-PAGE, and electroblotted onto nitrocellulose membranes (Bio-Rad, Hercules, CA, USA). Membranes were then incubated with primary antibodies in TTBS supplemented with 1% BSA overnight at 4°C, followed by the appropriate secondary antibodies coupled to horseradish peroxidase (Santa Cruz Biotechnology, Inc.). Peroxidase activity was detected on HyBlot CL autoradiography film (Denville Scientific, Inc. Plainfield, NJ, USA). Immunoblots were quantified using ImageJ® software.

Human corneal tissue sections (10 µm), mouse whole mount corneas, and keratinocyte cultures grown in 4-well chamber slides (Thermo Fisher Scientific) were fixed for 10 min with methanol at −20°C and washed with PBS. After blocking for 10 min in PBS with 3% BSA, slides were incubated overnight with primary antibody diluted in PBS with 1% BSA. Secondary antibodies were incubated for 1 h at room temperature. Nuclei were counterstained using Vectashield with DAPI (Vector laboratories, Burlingame, CA, USA). F-actin was stained with Rhodamine–phalloidin. Incubation with primary antibodies was routinely omitted in control experiments. The sections and cultures were viewed by confocal microscopy using a DM 6000 CS confocal laser scanning microscope (Leica Microsystems, Buffalo Grove, IL, USA).

Total RNA was isolated from keratinocyte cultures using the extraction reagent TRIzol (GibcoBRL, Carlsbad, CA, USA) following the manufacturer's instructions. Residual genomic DNA was eliminated by DNase I digestion of the RNA preparation. A total of 1 µg total RNA was used for cDNA synthesis (SuperScript II Reverse Transcriptase; GibcoBRL, Carlsbad, CA, USA). For qRT-PCR, MMP9 gene expression was measured using KAPA SYBR® FAST qPCR kit (Kapa Biosystems, Wilmington, MA, USA) and commercial MMP9 and GAPDH primers (Bio-Rad).

Depletion of CD147 in human corneal keratinocyte cultures was achieved using Silencer® Select siRNAs (Life Technologies). The sequences for siRNA targeting CD147 were: siRNA sequence 1 5′-GCCAAUGCUGUCUGGUUGCtt-3′ (ID147251) and siRNA sequence 2 5′-GCUACACAUUGAGAACCUGtt-3′ (ID215973). Silencer® Select Negative Control siRNA (ID4390843) was used as scrambled sequence. For knockdown, preconfluent keratinocytes were transfected by a 6-h incubation with 500 nM siRNA in Lipofectamine 2000 (1 µl/100 mm², Life Technologies) dissolved in Opti-MEM reduced-serum medium (Life Technologies). After transfection, the cells were incubated in K-SFM for an additional 3 days.

Cell culture medium was collected and centrifuged at high speed (21,130 g in a microfuge) for 5 min to remove cells and cellular debris. The supernatant (40 µl) was mixed with non-reducing loading buffer (50 mM Tris-HCl pH 6.8, 10% glycerol, 1% SDS, and 0.01% Bromophenol Blue) and resolved on 7.5% SDS-PAGE gels containing 1 mg/ml gelatin (bovine skin type B). Gels were then incubated in 50 mM Tris containing 5 mM CaCl₂ and 2.5% Triton X-100 overnight at room temperature. After washing with distilled water, gels were incubated in collagenase buffer (50 mM Tris-HCl pH 7.6, 5 mM CaCl₂) for 24 h followed by staining in Coomassie Brilliant Blue solution (40% methanol, 10% acetic acid, 0.025% Coomassie Brilliant Blue R-250). Gels were then washed in distilled water for 2 h and photographed. Gelatinase activity was quantified using ImageJ® software.

The in situ gelatinase activity was measured using the EnzCheck Gelatinase Assay kit (Molecular Probes, Carlsbad, CA, USA) following the manufacturer's instructions. Briefly, cells cultured on two-well chamber slides were incubated with 40 µg DQ-gelatin-fluorescein isothiocyanate, examined under a fluorescence microscope, and photographed.

Cell viability was assessed in human corneal keratinocyte cultures by means of the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay following the manufacturer's instructions (Molecular Probes). Briefly, cultures were exposed to increasing concentrations of rGal-3 and rGal-3C for 24 h and then incubated with a 1.2 mM MTT solution at 37°C for 4 h. The absorbance values of Blue Formazan were determined at 540 nm. Cell viability was expressed as the amount of MTT uptake in treated cells, normalized to untreated cells.

We thank Ilene Gipson and Zaki Shaikhibrahim at The Schepens Eye Research Institute for providing the human corneal keratinocyte cell line and assistance with the pull-down assays, respectively. We also thank Ula Jurkunas at the Massachusetts Eye and Ear Infirmary for collection of human corneal epithelial tissue. Gal3^−/− mice were a gift from Fu-Tong Liu at the University of California Davis.