ZO-1: Lamellipodial Localization in a Corneal Fibroblast Wound Model

Abstract

purpose. To explore the roles of ZO-1 in corneal fibroblasts and myofibroblasts in a model of wounding.

methods. Antibodies were used to identify ZO-1 in cultured rabbit corneal fibroblasts by immunocytochemistry, Western blot analysis, and immunoprecipitation. For colocalization studies, antibodies to β-catenin, cadherins, connexins, integrins, α-actinin, and cortactin were used. G- and F-actin were identified by DNase and rhodamine phalloidin, respectively. To study ZO-1 localization during cell migration, confluent corneal fibroblasts were subjected to scrape-wounding and evaluated by immunocytochemistry.

results. As predicted from previous studies, ZO-1 colocalized with cadherins and connexin 43 in intercellular junctions. The study revealed a new finding: ZO-1 was also detected at the leading edge of lamellipodia, especially in motile wounded fibroblasts and in freshly plated fibroblasts, before the formation of cell–cell contacts. In fibroblast lysates, ZO-1 largely partitioned to the detergent-soluble fraction compared with myofibroblast lysates, indicating that much of the fibroblast ZO-1 is not associated with insoluble structural components. Lamellipodial ZO-1 colocalized with G-actin, α-actinin, and cortactin, which are proteins involved with actin remodeling and cell migration. Integrins α5β1 and αvβ3 also localized to the leading edge of migrating fibroblasts, and the association of ZO-1 with integrin was confirmed by immunoprecipitation. Finally, alkaline phosphatase treatment of fibroblast lysate decreased the molecular mass of ZO-1 in lysates of cells grown in serum, demonstrating that, in activated fibroblasts, ZO-1 is phosphorylated.

conclusions. ZO-1’s appearance at the leading edge of migrating fibroblasts makes it a candidate for a role in the initiation and organization of integrin-dependent fibroblast adhesion complexes formed during migration and adhesion. Further, phosphorylation of ZO-1 may regulate its cellular localization.

In the stroma of the normal cornea, keratocytes communicate with one another by gap junctions. ^{1 2 3} Wounding disrupts these junctions, and, during repair, corneal fibroblasts and myofibroblasts reestablish this communication. ^{4 5 6} Using an in vitro corneal stromal cell wound-healing model, we have demonstrated that these connections are functional gap junctions. We have also identified cadherin-based intercellular junctions in fibroblasts and myofibroblasts. ^{7 8}  

We began the present study to examine whether ZO-1 was a candidate for organizing components of gap and adherens junctions of corneal fibroblasts during repair. In epithelia and endothelia, ZO-1 acts as a scaffolding molecule that localizes in adherens junctions, tight junctions, and gap junctions early in the assembly of actin-based junctional complexes. ^{9 10} Furthermore, ZO-1 has been shown to associate directly with connexins, facilitating their lateral aggregation before intercellular junction formation, and regulating their removal from the cell surface. ^{10 11} ZO-1 is a 220-kDa cytoplasmic protein that was first described as a component of tight junctions (zonula occludens) of epithelial and endothelial cells. ^{12 13 14 15 16} It is a member of the membrane-associated guanylate kinase (MAGUK) family of proteins characterized by protein–protein interaction domains (PDZ domains), Src homology domain (SH3), and the GUK domain. ^{14 15 17 18} Binding sites in the amino terminus of ZO-1 have been identified for claudins, connexins, and α-catenin and through the latter, to cadherins; and the carboxyl terminus of ZO-1 interacts with the actin cytoskeleton. ^{9 10 11 19} In addition, ZO-1 has tyrosines, serines, and threonines that are potential phosphorylation sites. ²⁰ The role of phosphorylated ZO-1 is uncertain, in that it has been implicated in very different functions for different cell types—for example, as a scaffolding protein for intercellular junctions and for promoting metastasis. ^{21 22}  

We evaluated the localization of ZO-1 in all three corneal stromal cell phenotypes: keratocytes (the quiescent fibroblasts freshly isolated from the cornea), corneal fibroblasts (the repair phenotype), and corneal myofibroblasts (the contractile phenotype). ^{23 24} We detected ZO-1 at intercellular junctions in each cell type. In migrating corneal fibroblasts, ZO-1 was also detected at the leading edge of the lamellipodia. Alkaline phosphatase treatment of fibroblast lysate indicated the presence of phosphorylated ZO-1, especially under conditions of extensive lamellipodial formation. By immunocytochemistry, lamellipodial ZO-1 colocalized with G-actin, α-actinin, and cortactin. The integrins α5β1 and αvβ3 were also detected at the leading edge. Furthermore, in fibroblasts that were newly plated, ZO-1 was detected at the cell’s periphery before the formation of intercellular contacts. Based on the temporal sequence of localization of ZO-1, cadherin, connexin (Cx)43, actin, cortactin, and integrins, we hypothesize that ZO-1 contributes to a scaffold for organizing molecules that assemble the actin cytoskeleton-adhesion complexes at the leading edge. By its presence at the leading edge, ZO-1 is in a position to participate in the initiation of new cell–cell contacts. 

Methods

Cell Culture

Quiescent corneal fibroblasts were released from rabbit corneas (Pel Freeze, Rogers, AR) by collagenase according to Masur et al. ²⁵ To maintain the quiescent fibroblast phenotype we used culture conditions as defined by Jester et al. ²³ : freshly isolated cells (10⁵ cells/mL) were grown on type I collagen in serum-free medium consisting of DMEM-F12 supplemented with l-glutamine, glutathionine, RPMI vitamin mix, sodium pyruvate (Invitrogen-Gibco, Rockville, MD), antibiotic–anti-mycotic mix, gentamicin (Sigma-Aldrich, St. Louis, MO), and insulin/transferrin/selenium supplement (Invitrogen-Gibco). In situ, these cells are called keratocytes, but they are fibroblasts and should not be confused with epidermal fish keratocytes or epithelial keratinocytes of skin. The cultured keratocytes were studied within 24 hours of isolation. 

When cultures are grown in DMEM/F12 and 10% FBS, they are primarily fibroblasts but may have up to 25% myofibroblasts (Masur SK, et al. unpublished data, 1996). Therefore, to generate homogeneous fibroblast or myofibroblast cultures, we passaged corneal fibroblasts at 5 × 10⁴ cells/mL and added phenotype-specific growth factors as follows: To produce fibroblasts, cells were cultured in DMEM/F12 and 10% FBS plus FGF-2 (20 ng/mL; Invitrogen-Gibco) and heparin (5 μg/mL; Invitrogen-Gibco). To produce myofibroblasts, we cultured cells in DMEM/F12 and 1% FBS plus human TGFβ1 (0.25 ng/mL; R&D Systems, Minneapolis, MN). ^{26 27}  

In Vitro Scrape-Wound Model

Cells were cultured for 3 to 5 days on 13-mm glass coverslips or in 100-mm tissue culture dishes until confluent. A single scratch was made with a P200 pipette tip across the diameter of a confluent coverslip. For protein analysis by Western blot, we used confluent cultures in 100-mm dishes and wounded the culture at 3-mm intervals, both vertically and horizontally. After wounding, media were replaced with fresh volumes of the same medium for 4 hours, and cells were fixed for immunocytochemistry or lysed for immunoprecipitation and Western blot detection. 

Immunocytochemistry

Immunocytochemistry was performed on cells plated on 13-mm glass coverslips. Cells were plated at 1 × 10⁵ cells/mL in their respective phenotypic media. Freshly isolated corneal fibroblasts (keratocytes) were plated on collagen I (3.0 mg/mL Vitrogen; Cohesion, Palo Alto, CA), whereas fibroblasts, or myofibroblasts were trypsinized and plated on bare coverslips. After growth for 1 to 3 days, cells were fixed in 3% p-formaldehyde (Sigma-Aldrich) in PBS (pH 7.4), for 10 to 15 minutes at room temperature and immunodetected (see below). For evaluation of new intercellular junctions, cells were trypsinized and suspended in complete medium and allowed to attach for 15, 30, 60, and 120 minutes before fixation. 

After fixation and blocking nonspecific binding with normal serum, coverslips were incubated with primary antibodies that were diluted in 1% bovine serum albumin (BSA) according to supplier directions. Myofibroblasts were identified by smooth muscle (SM) α-actin in stress fibers. ^{28 29} We used mouse anti-SM α-actin (Sigma-Aldrich), rabbit anti-Cx43 (Elliot Hertzberg, Albert Einstein College of Medicine, Bronx, NY), mouse anti-β-catenin (Transduction Laboratories, KY), rabbit anti-cadherin-11 (Karen Knudsen, Lankenau Medical Research Center, Philadelphia, PA), rat anti-ZO-1 (Chemicon, Temecula, CA), rabbit anti-ZO-1 (Zymed, South San Francisco, CA), mouse anti-ZO-1 (Zymed), mouse anti-cortactin (BD-Transduction Laboratories, Lexington, KY), mouse anti-α5β1 (VIF4; Ralph Isberg, Tufts, Boston, MA), and mouse anti-β3 (7E3, Barry Coller, Rockefeller University, New York, NY). The primary antibodies were visualized with anti-IgG raised against the appropriate animal species conjugated to Alexa Dyes 350, 488, and 546 (Molecular Probes, Eugene, OR). F-actin was detected with rhodamine or FITC-conjugated phalloidin (Sigma-Aldrich) and G-actin was detected using Texas red–conjugated DNase (Molecular Probes). The cells were viewed with a wide-field fluorescent microscope (Axioskop; Carl Zeiss Meditec, Jena, Germany) and images recorded with PhotoShop (Adobe Systems, Mountain View, CA). Each experimental condition was repeated at least three times. 

SDS-PAGE and Western Blot Analysis

Adherent control and 4-hour wounded cultures in 100-mm dishes were lysed in 1% NP-40 or Triton X-100 in lysis buffer plus protease and phosphatase inhibitors, as previously described, ²⁷ and separated into detergent-soluble (supernatant) and -insoluble fractions (pellet). The pellet was solubilized in 1.0% SDS. Protein concentrations were determined by protein assay (Bio-Rad, Hercules, CA) and 20 μg of protein per lane was separated by electrophoresis under reducing conditions in 6%, 7.5%, or 4% to 20% gradient gels (Bio-Rad), by SDS-PAGE, and transferred to nitrocellulose membranes (0.2 μm, Protran; Schleicher & Schuell, Keene, NH). After nonspecific binding was blocked (5% BSA in Tris-buffered saline plus 0.05% Tween-20), the blot was incubated with rabbit or mouse anti-ZO-1 (Zymed) or mouse anti-α5β1 primary antibodies (Sigma-Aldrich) overnight at 4°C with gentle rocking. This was followed by incubation with anti-IgG antibodies conjugated to horseradish peroxidase (HRP; Jackson ImmunoResearch Laboratories, West Grove, PA). A chemiluminescent substrate (SuperSignal, West Pico; Pierce, Rockford, IL) was used for antibody detection on imaging film (Biomax MS; Eastman Kodak, Rochester, NY). 

Alkaline Phosphatase

To determine whether ZO-1 was phosphorylated, NP-40 or Triton X-100 lysates with protease inhibitors, but not phosphatase inhibitors, were normalized for protein concentration, and then 30 μg of protein was incubated with 4.5 μL alkaline phosphatase (Roche, Mannheim, Germany) for 60 minutes at 37°C. ³⁰ Addition of phosphatase inhibitors to both the alkaline phosphatase and nonphosphatase lysate served as the control. The results were analyzed by SDS-PAGE and Western blot analysis with rabbit anti-ZO-1, as described earlier. 

Immunoprecipitation

Four hour scrape-wounded and unwounded cultures were washed twice with cold PBS and lysed in 0.5% Triton X-100 or 1% NP-40, 150 mM NaCl, 10 mM Tris, 1 mM MgCl₂, 1 mM CaCl₂, phosphatase inhibitors, 3 mM Na vanadate, 3 mM NaF, (Sigma-Aldrich), EDTA-free protease inhibitors (Complete; Roche), and phenylmethylsulfonyl fluoride (Sigma-Aldrich), as previously described. ^{31 32} Lysis and all subsequent steps were performed at 0°C to 4°C. The soluble fractions were precleared for 20 minutes with anti-mouse IgG–coupled agarose beads (Sigma-Aldrich). The precleared lysates were incubated overnight with mouse monoclonal anti-α5β1, 10 μL/500 μL lysate (MAB1999; Chemicon, Temecula, CA). The next day, 40 μL of anti-mouse IgG agarose beads were added to 500 μL lysate, for 60 minutes of shaking, followed by collection of bead-bound immune complexes by centrifugation at 4000 rpm for 2 minutes. Proteins were eluted from the beads with reducing sample buffer and boiled for 5 minutes. Samples were separated on 1 to 1.5 mm, 6% or 8% SDS-polyacrylamide gels and transferred to nitrocellulose (0.2 μm, Protran; Schleicher & Schuell) for Western blot analysis. 

Results

ZO-1 Expression in Quiescent Fibroblasts

We found that ZO-1 was expressed in all phenotypic variants of the corneal stromal fibroblasts. To evaluate expression of ZO-1 in quiescent fibroblasts (keratocytes), we isolated corneal stromal cells and cultured them under conditions that maintained the keratocyte phenotype. ²³ At 24 hours, the cells had established the characteristic keratocyte stellate morphology (Fig. 1A) . By immunodetection of vimentin, a fibroblast intermediate filament protein, we confirmed that these cells were derived from the stroma and not the epithelium or endothelium (Fig. 1B) . ZO-1 was detected in keratocyte cell–cell contacts (Fig. 1A)similar to the identification of Cx43 in keratocyte cell–cell contacts. ²³ In contrast, cadherins have not been detected in either cultured keratocytes or in situ (Masur SK, unpublished data, 1999). In addition, when keratocytes were lysed after 24 hours in culture, antibody to ZO-1 detected a 220-kDa protein by Western blot (Fig. 1G) . It is important to note that ZO-1 was not detected in the first hours after isolation of the keratocytes (data not shown). This may be for two reasons; it was not present in situ or cell junction proteins were degraded and had to be resynthesized after the cells were isolated from the stroma. 

ZO-1 Expression in Cell-to-Cell Adherens Sites in Fibroblasts and Myofibroblasts

By immunocytochemistry, ZO-1 was localized at cell–cell contacts in punctate or elongated structures in corneal fibroblasts (Fig. 1C)and myofibroblasts (Fig. 1E)where it colocalized with the gap junction protein Cx43 and the adherens junction protein cadherin-11 (Fig. 2 , arrows). ^{2 7} To determine whether ZO-1 at cell junctions of fibroblasts is associated with actin as it is in epithelial cells, we used phalloidin detection of F-actin in combination with immunodetection of ZO-1. We found that ZO-1 was adjacent to the terminals of F-actin stress fibers in corneal fibroblasts, but without overlapping them (Figs. 1C 1D) . Similarly, in myofibroblasts, ZO-1 was adjacent to SM α-actin microfilaments without overlapping (Figs. 1E 1F) . In Western blot analysis of myofibroblast lysates, both SM α-actin and ZO-1 partitioned largely into the detergent-insoluble fraction (Fig. 1I) . In contrast, in fibroblast lysates, ZO-1 was mainly in the detergent-soluble fraction (Fig. 1H)and thus was not associated with highly polymerized cytoskeletal elements. Furthermore, we often detected a second, higher-molecular-mass band in the fibroblast lysate (described later). 

Newly Developing Fibroblast Cell Junctions: ZO-1, Cadherin, and Cx43

In established junctions, ZO-1 was colocalized with cadherin, and Cx43 (Fig. 2) . To address the question of ZO-1’s association with formation of new contacts, we used freshly plated cells and documented the sequence of the appearance of ZO-1 and the junctional proteins that localize and form the new cell–cell contacts. Using antibodies to β-catenin that identify both classic and nonclassic cadherins junctions, ³³ we found that β-catenin and ZO-1 colocalized in the earliest cell–cell contacts, 1 hour after plating (Figs. 3A 3B , overlay). In contrast, Cx43 was primarily detected in intracellular vesicles at this time (Fig. 3D) . Thus, it is likely that ZO-1 can participate in cadherin-based junction formation, and that formation of this junction precedes that of Cx43 gap junction formation. Also, ZO-1 was immunodetected along with β-catenin at the periphery of the cells (Figs. 1C 2B 3B , arrowheads) in lamellipodial-like, noncontact areas, as previously reported. ²¹ When the cells were plated on defined matrices in the absence of serum, a similar peripheral pattern of ZO-1 localization was detected (data not shown). 

ZO-1 Expression in Fibroblast Lamellipodia in a Scrape-Wound Model

Scrape-wounding confluent corneal fibroblasts provided a model for directed cell migration into a wound gap. In the scrape wound, rabbit corneal fibroblasts showed distinct lamellipodia (Fig. 4B , arrowhead). ZO-1 was immunodetected at the leading edge of these corneal fibroblast lamellipodia, an observation that has not been reported previously (Figs. 4E 5 6) . Lamellipodial ZO-1 outlined the periphery of the leading edge (Figs. 4E 5A 5C 5E , arrowhead), in contrast to the punctate or dash-like ZO-1 configurations that were seen in cell–cell contacts at the ends of stress fibers (Figs. 1 2 4E 5E , arrows). We also immunodetected ZO-1 in lamellipodia of cultured human corneal fibroblasts, human neonatal foreskin fibroblasts, and mouse embryonic fibroblasts (data not shown). 

For wound closure, fibroblast migration requires the actin-dependent extension of the leading edge regulated by proteins that sequester, cap, sever, nucleate, and depolymerize actin. ¹⁸ To determine whether ZO-1 might be associated with actin assembly and disassembly, we performed immunodetection for G-actin, α-actinin, and cortactin, as well as ZO-1. In Figure 5 , the colocalization of ZO-1 with these proteins was detected in the forward-moving lamellipodia. It was particularly interesting that myofibroblasts, which were less migratory than fibroblasts, had less extensive lamellipodia and less ZO-1 at the leading edge than did fibroblasts in the scrape-wound model (Figs. 5G 5H) . 

Integrins and ZO-1 Localization in Lamellipodia

During migration, cells move on the extracellular matrix using integrins, which are transmembrane proteins. In the wound-closure assay we found that migration required the presence of serum and wondered whether the matrix molecules including fibronectin and vitronectin in serum might also be ligands for integrins at the leading edge. Because integrins in turn, interact with the actin cytoskeleton via scaffolding proteins such as FAK, vinculin, and α-actinin, ³⁴ we postulated that ZO-1 might also act as a scaffold at the leading edge and interact with integrins at these sites. Both the fibronectin receptor α5β1 and the vitronectin receptor αvβ3 were predominantly immunodetected at the leading edge of cells migrating into the wound (Fig. 6) . Some but not all the immunodetected integrin colocalized with ZO-1 at the leading edge (Figs. 6C 6F , arrowheads). In contrast no integrin was detected with ZO-1 at cell–cell junctions (arrow). Further, the antibody to α5β1 coimmunoprecipitated ZO-1 from lysates of scrape-wounded cultures (Fig. 6G) . These data suggest that ZO-1 may act as a scaffolding molecule for integrin-mediated cellular adhesion in the forward-moving lamellipodia. 

ZO-1 Phosphorylation

Serum-containing medium is necessary to generate lamellipodia and induce cell migration after wounding. To determine whether serum stimulates ZO-1 phosphorylation, cells were incubated in serum-free media (SFM) or with serum (FBS). In lysates from cells incubated overnight in SFM, the band of immunodetected ZO-1 is denoted by an asterisk (Fig. 7 , lane 1). In contrast, when cells were incubated in 10% FBS (with or without FGF-2), the immunodetected ZO-1 was quite broad and included a dense lower band (asterisk) and a wide, more diffuse band above it (lane 5). This broad band of immunodetected ZO-1 suggested that serum induces the phosphorylation of ZO-1 on multiple sites. To test this, we incubated each of the lysates with alkaline phosphatase, a broad-spectrum phosphatase. Whereas the ZO-1 band in lysates of SFM fibroblasts did not change (lane 2), the broad ZO-1 band in lysates from cells grown with FBS was significantly reduced (lane 6). Furthermore, the molecular mass of the dense, lower ZO-1 band appeared lower than before alkaline phosphatase treatment and was now at a mass that was comparable to the one detected in the absence of serum. These data demonstrate that under SFM conditions, ZO-1 was not phosphorylated, whereas in the presence of FBS, and therefore when the cells had many lamellipodia, ZO-1 was highly phosphorylated. 

We used the scrape-wound model to generate many lamellipodia compared with nonwounded cultures. To determine whether wounded cells had increased ZO-1 phosphorylation, cultures were scrape-wounded and incubated for 4 hours in the presence of FBS (lanes 3 and 4). The same banding pattern was seen as with nonwounded cells in serum, and alkaline phosphatase similarly reduced the ZO-1 band. Thus, growing the cells in FBS was enough of a stimulus to induce ZO-1 phosphorylation—scrape-wounding did not enhance it further. 

These data suggest that ZO-1 can be differentially phosphorylated and that ZO-1 localization may be affected by changes in the phosphorylation state. This finding does not exclude the possibility that ZO-1 splice variants, one containing an 80-amino-acid insertion (termed ZO-1α⁺), could also contribute to ZO-1 localization and the broad band in serum-stimulated cells. ³⁵  

Discussion

Since ZO-1 was first identified in the detergent-insoluble zonula occludens complex, studies have focused on its role in the construction of intercellular junctions of epithelial and endothelial cells. ¹⁵ We have confirmed the more recent identification of ZO-1 in intercellular junctions of fibroblasts and report that ZO-1 colocalized with endogenous cadherins and connexins in cell–cell adherens and gap junctions in corneal fibroblasts and myofibroblasts. In a novel finding, we found that ZO-1 localized to the leading edge of lamellipodia in corneal fibroblasts migrating into a scrape wound. These observations confirm and extend to primary cells an earlier report by Howarth et al., ³⁶ who described the immunodetection of ZO-1 in the periphery of sarcoma cells. That ZO-1 is situated at the leading edge of lamellipodia suggests a role for it in organizing new cell-contact complexes at the leading edge of migrating cells. ^{21 36 37} Our finding of ZO-1 in the earliest cell junctions formed by corneal fibroblasts supports the idea that the initial position of lamellipodial ZO-1 may contribute to organizing cytoskeletal and membrane proteins in effective new cell–cell junctions. 

Within the multidomain structure of ZO-1, two regions have been identified that underlie ZO-1’s potential to coordinate extracellular signals with actin organization. A member of the MAGUK protein family, ZO-1 is characterized by repeated PDZ domains in its N terminus. In tight junctions and synapses, PDZ-containing proteins act as molecular scaffolds in the assembly of transmembrane proteins and signaling molecules. ^{14 15 17 18} In the adherens junctions, whereas the PDZ domain interacts with locale-specific transmembrane adhesion proteins during establishment of junctional complexes, the proline-rich C terminus of ZO-1 can associate with actin. ^{9 38 39 40} Cortactin, G-actin, and α-actinin are involved with actin-assembly and colocalize with ZO-1 at the leading edge. Future experiments will explore the potential interaction of ZO-1 with these and other molecules that are essential to actin-assembly–dependent polarized migration and typically localize to the leading edge, including Mena/Vasp, Ezrin/Moesin/Radixin, WASP/Scars, and the Arp2/3 complex. ^{40 41}  

In light of our localization of ZO-1 and integrins at the leading edge, it is of interest that a GFP-tagged integrin subunit, α5-GFP, transfected into CHO cells, also localized to the leading edge of lamellipodia (see Fig 2bin Laukaitis et al. ⁴² ). This α5-GFP lamellipodial immunolocalization is identical with what we found for ZO-1 and α5 and αv integrins. It should be noted that, in the corneal fibroblasts, both ZO-1 and the α5 were also detected separately in their characteristic discrete dash-like configurations, either in cell–cell contacts or focal adhesions. 

The nature of the link between integrins and ZO-1 is not clear at this time. Two pieces of data suggest that signaling from integrin–matrix binding may participate in ZO-1’s localization at the leading edge. First, anti-α5-antibody coimmunoprecipitated ZO-1, suggesting that they interact in situ, consistent with the appearance of both at the leading edge. Second, the scrape-wound assay in which we immunodetected ZO-1 at the leading edge was performed in the presence of serum-containing media, which exposed the cells’ surface to vitronectin and fibronectin in the serum. Arguing against this interpretation was our inability to prevent ZO-1 localization using function-blocking anti-α5β1 and -αvβ3 (data not shown). Thus, the relocalization of these molecules to the leading edge may be elicited by lysophosphatidic acid and other components found in serum that signal cytoskeletal reorganization rather than by serum ECM components. 

The identification of ZO-1’s subcellular locations was based on immunodetection by commercial antibodies to ZO-1. However, since commercial antibodies do not distinguish between ZO-1’s phosphorylation states or splice variants, ZO-1 in each site may differ from one another. ^{35 43 44} Studies relating ZO-1 phosphorylation and the tightness, or integrity, of cell junctions have yielded various results. ^{22 36} Our data support the likelihood that lamellipodial ZO-1 is a phosphorylated form of ZO-1. Increased ZO-1 phosphorylation was detected in lysates from cells grown in serum, a condition that promoted lamellipodial formation; and the pattern of decreased immunodetectable ZO-1 produced by alkaline phosphatase treatment of lysates suggests that ZO-1 has multiple phosphorylation states in activated corneal fibroblasts. Determination of the sites of phosphorylation may provide clues to molecular interactions and to the kinases responsible for ZO-1’s lamellipodial and intercellular locations. 

We have not yet evaluated whether fibroblasts also express both the alternatively ZO-1 splice variants that differ in an internal 80 amino acids. Previous studies on multiple cell types using RT-PCR and splice-specific antibodies have demonstrated cell-type–specific expression of one or both variants. ³⁵ An isoform-specific antibody will determine whether the dual localizations of ZO-1 in corneal fibroblasts correlate with the expression of ZO-1 as splice variants in these cells. 

We hypothesize that ZO-1 in the leading edge contributes to cell migration. This is supported by other examples of migratory cells where ZO-1 is no longer limited to cell–cell junctions. For instance, a previous report noted that ZO-1 was “delocalized” from tight junctions in transition from confluent endothelium or epithelium to a more migratory condition. Of particular interest is the response of the corneal endothelium after wounding. As the cells became migratory, the ZO-1 adjacent to the tight junctions exhibited fragmentation. ⁴⁵ Similarly, in studies exploring the epithelial-to-mesenchymal transition, ZO-1 was lost from tight junctions and became cytoplasmic when the cells became migratory. ^{17 46}  

After corneal injury, healing without long-term scarring and corneal opacity requires precise localization and remodeling by the cells of the corneal stroma. This study has uncovered a potential role for ZO-1 at the initiation of the wound–healing process in the dynamic reorganization of adhesions that occurs at the leading edge. Of particular importance is the finding that in fibroblasts, the migratory corneal stromal cell phenotype, ZO-1 was detected in both cell–cell junctions and the highly developed lamellipodia and was detected in both the NP-40-soluble and -insoluble fractions. This is in contrast to the nonmigratory myofibroblasts in which ZO-1 was detected primarily in cell–cell junctions and in the NP-40-insoluble fractions but not in lamellipodia. We propose that the leading edge localization of ZO-1 could contribute to fibroblast migration to replace lost keratocytes and repair corneal stroma. The fact that we found ZO-1 in the lamellipodia only in fibroblasts is particularly interesting in light of recent studies on the urokinase (uPA) pathway, where we demonstrated that uPA and its receptor, uPAR, are expressed in the corneal fibroblast, with little in the myofibroblast and none in keratocytes (Bernstein et al., manuscript submitted). Furthermore, in corneal fibroblasts, migration was induced by uPA binding to its receptor, thus anchoring the ligand-receptor complex in the actin cytoskeleton. ⁴⁷ Together, these uPA/uPAR and ZO-1 studies better define the corneal fibroblast phenotype and elucidate pathways specific to wound healing in the corneal stroma. 

 Supported by National Eye Institute (NEI) Grant R01 EY09414 (SKM), an NEI Core Grant, and Research to Prevent Blindness grants. AMB was supported by NEI Grant F32 EY07049. Microscopy was performed at the Mount Sinai School of Medicine Microscopy Shared Research Facility, supported in part, with funding from NIH-NCI shared resources Grant 1-R24-CA095823 and National Science Foundation Major Research Instrumentation Grant DBI-9724504. LT received a postdoctoral research fellowship from Fight for Sight, Prevent Blindness America.

 Submitted for publication February 11, 2004; revised September 10, 2004; accepted September 17, 2004.

 Disclosure: L. Taliana, None; M. Benezra, None; R.S. Greenberg, None; S.K. Masur, None; A.M. Bernstein, 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.

 

Immunodetection of ZO-1 in fixed, cultured keratocytes, fibroblasts, and myofibroblasts (A, C, E). ZO-1 was immunodetected at the cell–cell contacts (arrows) of keratocytes (A), fibroblasts (C), and myofibroblasts (E), as well as in lamellipodia of fibroblasts (C, arrowhead). Areas with ZO-1 in cell contacts, indicated by rectangles in adjacent panels, have been enlarged and printed with ZO-1 (green) and cytoskeleton (red) superimposed as overlays. In keratocytes, ZO-1 was detected at intercellular contacts (A, arrows), adjacent to vimentin intermediate filaments (B). In fibroblasts, ZO-1 was detected at intercellular contacts (C, arrows), adjacent to F-actin stress fiber terminals (D). In myofibroblasts, ZO-1 was immunodetected at intercellular contacts (E, arrows) adjacent to SM α-actin stress fiber terminals (F). ZO-1 was detected in Western blots of keratocytes (G), FGF-2 treated-fibroblasts (H), and TGFβ-induced myofibroblasts (I). In the lysates of keratocytes and fibroblasts, ZO-1 was immunodetected in the NP-40-soluble fraction (S), whereas in lysates of myofibroblasts, it was immunodetected in the NP-40-insoluble pellet (P). The myofibroblast phenotype was confirmed by reblotting for SM α-actin (I, SMA). The same immunocytochemical detection patterns were seen with each of the three commercially available antibodies to ZO-1. (J, K, L) Controls for specificity of the anti-ZO-1 antibodies. Mouse (J), rabbit (K), and rat (L) IgG was followed by the appropriate secondary antibody conjugated to Alexa 488 (Molecular Probes, Eugene, OR).

Figure 2.

 

ZO-1 was colocalized with cadherin-11 and Cx43 in cell–cell contacts of fibroblasts (A, B) and myofibroblasts (C, D). Arrows: colocalization (yellow) of ZO-1 (green) and cadherin-11 (red) or Cx43 (red). In a fibroblast lamellipodium, ZO-1 was expressed without Cx43 (B, arrowhead, green). Note that in the Figure 3overlay inset, lamellipodial ZO-1 was coexpressed with β-catenin.

Figure 3.

 

ZO-1 and β-catenin were immunodetected in fibroblasts’ cell–cell contacts before Cx43. In fibroblasts fixed 1 hour after plating in serum, (A) ZO-1 and (B) β-catenin were immunodetected in newly formed cadherin-based junctions (arrow) and were also immunodetected in the periphery of cells in lamellipodia (arrowhead). In the overlay, colocalization (yellow) of ZO-1 (green) and β-catenin (red) was visible. In contrast at 1 hour, Cx43 (D, arrows) was detected in intracellular vesicles but not at cell–cell junctions where ZO-1 (C, arrow) was expressed.

Figure 4.

 

ZO-1 was immunodetected in lamellipodia of scrape wounds. Phase micrographs of scrape-wounded confluent cultures at (A) 0, (B) 4, (C) 24, and (D) 48 hours after wounding. ( ) Site of the scrape wound. In the presence of serum, fibroblasts migrated into the wound within 24 hours (C). At 4 hours after wounding, the lamellipodia had ZO-1 at the leading edge of scrape wounds (B, E, arrowhead) and at intercellular contacts (E, arrow).

Figure 5.

 

ZO-1 colocalized with G-actin, α-actinin, and cortactin in lamellipodia of 4-hour scrape wounds in fibroblasts. (A, C, E, G) ZO-1 immunodetection. (B) G-actin, was detected by DNase binding and (D) α-actinin, (F) cortactin, and (H) SM α-actin by immunodetection. ZO-1 was colocalized with G-actin, α-actinin, and cortactin in lamellipodia (arrowheads). ZO-1 was also detected in cell–cell contacts (E, arrow). (G, H) A myofibroblast (MF) identified by immunodetection of SM α-actin stress fibers had little lamellipodial ZO-1 (arrowhead) compared with the adjacent fibroblasts (F).

Figure 6.

 

Integrins and ZO-1 were immunodetected at the leading edge of lamellipodia in 4-hour scrape wounds. (A) ZO-1 and (B) αvβ3 were immunodetected at the leading edge in lamellipodia (arrowhead). In the overlay, yellow indicates colocalization of ZO-1 (green) and αvβ3 (red). Similarly (D) ZO-1 and (E) α5β1 were both detected in lamellipodia (arrowhead) and colocalized (yellow) in (F). In contrast, ZO-1 was expressed in cell–cell contacts without integrins (C, F, arrow, green). (G) Four-hour scrape-wounded fibroblasts were lysed and immunoprecipitated with monoclonal anti-α5β1 and then ZO-1 detected with polyclonal antibody to ZO-1 (arrow). ZO-1 immunoprecipitation by anti-α5β1 supports the hypothesis that some α5β1 integrin and ZO-1 interacted.

Figure 7.

 

ZO-1 was phosphorylated in fibroblasts under conditions promoting extensive lamellipodial formation. (A) ZO-1 was immunodetected in NP-40 lysates of fibroblasts grown in SFM (lanes 1, 2), in the presence of FBS and FGF/heparin, 4 hours after wounding (lanes 3, 4) or in the presence of serum and FGF/heparin, in unwounded cultures (lanes 5, 6). In SFM, a light band of ZO-1 was detected (*), whereas in the presence of FBS and FGF/heparin, a dense band of ZO-1 was detected (*) with an additional broad band above it. Lysate was incubated with alkaline phosphatase decreasing the size and extent of the broad band (lanes 4, 6). In addition, the dark, lower ZO-1 band (*) was decreased to a molecular mass comparable to that in lysates of SFM fibroblasts (compare lanes 1, 2, 4, 6). The same blots were detected for Hsp90 as an indicator of protein loaded per lane. (B) Lysates in the same sequence as in (A) were incubated in the presence of phosphatase inhibitors. In the presence of these inhibitors, there was no difference in the ZO-1 band before (lanes 1, 3, 5) or after alkaline phosphatase treatment for each condition (lanes 2, 4, 6).

The authors thank Scott Henderson for expert assistance at the MSSM Microscopy Shared Research Facility and Dania Zekaria and Anurina Mitra for technical support.