Time at Confluence for Human RPE Cells: Effects on the Adherens Junction and In Vitro Wound Closure

Abstract

purpose. To determine how time at confluence affects the properties of cultured human retinal pigment epithelium (RPE) cells, with emphasis on the adherens junction.

methods. Cultures were maintained at confluence without passage for intervals to several months. Adherens junction proteins (N-cadherin, E-cadherin,α -catenin, β-catenin, plakoglobin, and actin) and the proliferation marker Ki-67 were localized in the cultures by fluorescence microscopy, and in vitro wound healing was compared. Adherens junctions were analyzed for protein solubility in detergent buffers and sensitivity to disruption by treatment with anti-cadherin antibodies and low calcium conditions.

results. Compared with cultures in early-confluence (2–3 days), postconfluent cultures (weeks) had more mature adherens junctions characterized by a circumferential (rather than linear) actin organization, and a zonular (rather than punctate) distribution of more detergent resistant cadherin and catenins. Postconfluent cultures also had fewer Ki-67-positive cells and a higher cell packing density. Early-confluence cells migrated into in vitro wounds as dissociated single cells, whereas postconfluent cells moved as contiguous sheets, retaining an intact junction during wound-induced cell migration and proliferation. Mature junctions were not disrupted by treatment of living cells with N-cadherin antibodies, which bound to and remained detectable at junctions for several days. Calcium withdrawal displaced N-cadherin from mature junctions and rendered it more soluble, but the dominant circumferential pattern of actin was stable. Restoration of medium calcium resulted in a rapid (hours) recovery of a nearly complete zonular pattern of insoluble N-cadherin.

conclusions. Over long postconfluent periods, cultured RPE cells became more growth quiescent, and intercellular cadherin adhesions became more stable, exhibiting increased resistance to calcium removal and greater retention of junctional integrity during in vitro wound closure. Consideration should be given to whether the behavior of RPE cells in postconfluent cultures, where intercellular adhesions are more mature, more closely simulates RPE cells in situ than cells in early-confluence cultures, which are more commonly used for analysis.

Development of the zonular adherens junction in epithelial cells is a time-dependent process that has been studied in detail in cell lines that express E-cadherin, such as kidney-derived Madin–Darby canine kidney (MDCK) cells. ^{1 2 3 4} The initiating event for forming the calcium-dependent adherens junction is the intercellular contact that occurs at confluence. During junction formation, the transmembrane cadherin protein localizes to apicolateral cell borders where its extracellular domain binds with cadherins in adjacent cells. The cytoplasmic domain complexes with catenins, proteins that link the cadherin to a circumferential bundle of actin microfilaments. ^{5 6 7} The linkage stabilizes the junction and renders the cadherin and catenins resistant to detergent extraction. ^{1 2 3} After adherens junction formation, polarized plasma membrane domains develop, resulting in the phenotype characteristic of epithelial cells. ^{8 9 10}  

Retinal pigment epithelium (RPE) cells differ from most monolayer epithelial cells in that the dominant cadherin is N-cadherin, ^{11 12 13 14 15} and adherens junction development is markedly slower, at least in RPE cell subpopulations selected for their highly epithelioid phenotype. ^{15 16} Rather than forming a mature junction within hours after cell–cell contact, fully dissociated epithelioid RPE cells require weeks at confluence to develop an N-cadherin junction that is completely zonular and stable to detergent extraction. ¹⁵ E-cadherin is expressed by some epithelioid variants of RPE, but E-cadherin may not contribute to early stages of junction formation because it does not localize to junctions until after the N-cadherin–containing junction is already established. ¹⁷ Unselected cultures of RPE cells, of the type typically used for experimentation, are also likely to undergo a slow postconfluence junctional maturation. Time at confluence therefore might produce RPE cells with different levels of junction development and therefore different baseline properties in functional assays. 

A functional assay that is widely used to analyze the induction of cell growth and migration in epithelial cells, including the RPE, ^{18 19 20 21 22} is in vitro wounding. In wound experiments, cultures are typically grown to confluence to generate contiguous monolayers, then defects are produced by chemical or mechanical treatments. Confluence is used to establish a quiescent baseline state resulting from the long-established contact inhibition of cell growth ²³ and motility. ²⁴ Confluence also triggers the morphogenetic process that produces an epithelial phenotype. Confluent epithelial monolayers are therefore presumed to resemble mature, quiescent cells within the intact tissue and to respond to wounding in a manner that simulates the response of cells in situ. This presumption might be questioned for RPE cells in which adherens junction development is delayed, so that RPE cultures shortly after confluence may not manifest the epithelial properties that have been attributed to them. 

In the current study we analyzed RPE cultures in early confluence (within 2–3 days of cell contact), and in late confluence (approximately 8 weeks after confluence) to evaluate the developmental state of the adherens junction and the functional response of cells to in vitro wounding. Wound closure was also examined in a few cultures after several months at confluence. We conclude that maintaining RPE cultures for extended postconfluent periods produces more growth-quiescent cells with stable intercellular adhesions whose behavior during wound closure differs from cells in early-confluence cultures, when most epithelial cells are typically analyzed. 

Methods

Cell Culture and Wounding

Human RPE cell cultures were produced as previously described ²⁵ and maintained with biweekly feedings of minimum essential medium (MEM) containing 10% fetal bovine serum (FBS). Cultures in passages 2 through 13 from 22 human donors aged 9 to 78 years were used for the experiments. All experiments were performed on a minimum of three cell populations from different donors. Cells were plated in flasks, multiwell plates or eight-chamber slides (Laboratory-Tek; Nalge Nunc, Naperville, IL) at a density that produced visual confluence in 2 to 4 days. Cultures were analyzed within 2 to 3 days of confluence (early confluence), or maintained undisturbed at confluence for approximately 8 weeks (late confluence or postconfluence), or 4 to 6 months (very late confluence). Linear wounds were produced in cultures with 1-mm pipette tips, and repair was observed by phase contrast and fluorescence microscopy. 

Fluorescence Microscopy

RPE cultures in chamber slides were fixed for 15 minutes in cold 3% phosphate-buffered paraformaldehyde (pH 7.4) then permeabilized by incubation for 5 minutes at room temperature in 0.5% Triton X-100 in phosphate-buffered saline (PBS). To visualize detergent-insoluble, cytoskeletally linked protein, some cultures were extracted before fixation with a detergent buffer containing protease inhibitors using a previously described protocol. ²⁶  

For immunostaining, fixed monolayers were incubated with primary antibodies overnight at room temperature, followed by three rinses in blocking buffer (10 mg/ml bovine serum albumin in PBS) and a 1-hour incubation in secondary antibodies. The following mouse monoclonal primary antibodies were used: anti-N-cadherin (clone GC-4; Sigma, St. Louis, MO), anti-E-cadherin (clone HECD-1; Zymed Laboratories, San Francisco, CA), anti-α/catenin (Becton Dickinson, San Jose, CA), anti-β-catenin and anti-γ-catenin/plakoglobin (Transduction Laboratories, Lexington, KY), and anti-nuclear antigen Ki-67 (a proliferation marker; MIB-1; Immunotech, Westbrook, ME). Monoclonal and polyclonal pancadherin antibodies (Sigma) were also used. The pancadherin antibodies, which are directed against an epitope in the cytoplasmic domain of the cadherins, react primarily with N-cadherin. ^{17 27} Appropriate fluorescein- or rhodamine-conjugated secondary antibodies were from Jackson ImmunoResearch (West Grove, PA). F-actin was visualized by treatment of fixed cells for 30 minutes at room temperature with fluor-conjugated phalloidin (Sigma). After staining, coverslips were mounted with antifade reagent (FluoroGuard; BioRad, Hercules, CA), and specimens were examined and photographed with a fluorescence microscope. 

Western Blot Analysis

The sensitivity to detergent extraction of adherens junction proteins was examined by preparing detergent-soluble and -insoluble protein extracts from early- and late-confluence RPE cultures from three donors. Soluble protein was extracted by the protocol described, using 50 μl buffer/cm² of culture surface, then mixed with an equal volume of Laemmli ²⁸ electrophoresis buffer. The detergent-insoluble protein that remained in the culture dish was rinsed three times with fresh detergent buffer and lysed with a 1:1 mixture of detergent buffer and electrophoresis buffer. 

For electrophoresis, reducing agent (5 mM β-mercaptoethanol) was added to the soluble and insoluble fractions. Samples were boiled for 10 minutes, loaded in equivalent volumes on gels, and electrophoresed using 10% sodium dodecyl sulfate (SDS) separating gels. Proteins were transferred to polyvinylidene (PVDF) membranes by electroblotting. ²⁹ Membranes were blocked by gentle agitation for 1 hour in 15% nonfat dry milk and 0.1% Tween-20 in PBS (pH 7.4). Primary and appropriate secondary antibodies (listed in the previous section) were diluted in PBS containing 0.1% Tween-20, and each was incubated with the membrane for periods ranging from 1 hour to overnight. Bands were visualized by the enhanced chemiluminescence (ECL) detection system (Amersham Pharmacia Biotech, Piscataway, NJ). 

Antibody Treatment and Calcium Depletion

Several protocols were conducted in which RPE cells were treated with anti-cadherin antibodies that are directed against the extracellular domain of the protein and have been reported to block adhesive function, leading to cell dissociation. ^{30 31} N-cadherin antibodies were used at a concentration of 45 to 200 μg/ml in complete culture medium (MEM with 10% FBS). In a few experiments, function-blocking E-cadherin antibodies (4 μg/ml) were also used. 

Cultures were incubated with the antibodies for intervals from 6 to 24 hours. Parallel control cultures were exposed to medium containing an amount of nonimmune IgG equal to the antibody. Control cultures were immunostained for N-cadherin, using anti-N-cadherin antibodies and/or pancadherin antibodies, by the protocol described. For treated cultures, antibody that had bound to the living cells was visualized with the fluor-conjugated secondary antibody alone, with or without costaining with pancadherin antibodies. In some experiments on postconfluent cells, paired cultures were treated with E-cadherin function blocking antibodies or with a combination of E-cadherin and N-cadherin antibodies. For these experiments, immunostaining for E-cadherin was also included. 

After we observed that N-cadherin antibodies bound to the mature junctions of living cells without disrupting the junctions, we performed experiments to examine retention of bound antibodies. Late-confluence RPE cultures were incubated for 6 hours with N-cadherin antibodies or nonimmune IgG (control), then the medium containing the antibodies was removed and replaced with antibody-free medium. Cultures were harvested at intervals to three days after removal of the antibody, fixed and stained with secondary antibodies alone (to detect antibody pre-bound to living cells), and costained with pancadherin antibodies (to determine total N-cadherin distribution). 

The effect of calcium removal on the calcium-dependent cadherin junctions was examined by treating early- and late-confluence RPE cultures for intervals from 1 to 45 minutes with the calcium chelator EGTA (2 mM in complete serum-containing culture medium). EGTA-treated cultures and control cultures incubated in complete medium lacking EGTA were fixed and stained to localize N-cadherin and F-actin as described. In postconfluent cultures, in which 45 minutes of EGTA treatment was necessary to displace N-cadherin from the junctions of most cells, recovery experiments were also performed. Cultures were EGTA-treated for 45 minutes, refed with EGTA-free medium, and harvested 2 hours later. To determine whether recovery of N-cadherin at junctions required protein synthesis, recovery experiments were also conducted on paired cultures in the presence of the inhibitor cycloheximide (CHX, 100 mM). CHX was present during the 45-min EGTA treatment period and the 2-hour recovery interval. Preliminary experiments confirmed that CHX did not affect the displacement of N-cadherin from junctions with EGTA treatment. In some experiments the effect of EGTA treatment on the solubility of N-cadherin and actin in postconfluent cells was examined. EGTA-treated cultures were extracted with detergent buffer followed by examination of detergent-resistant protein by fluorescence microscopy or by western blot analysis of detergent-soluble and -insoluble fractions by using the protocols described. 

To examine the effects of prolonged exposure to calcium-depleted medium, postconfluent RPE cultures were incubated in culture medium containing a concentration of calcium (5 μM), below that necessary to support cadherin adhesion. ^{32 33 34} Incubation in low-calcium medium lasted for intervals to 1 month, with biweekly changes of the culture medium. Paired control cultures were similarly maintained in medium containing normal calcium levels (1.8 mM). 

Results

Characteristics of Early- and Late-Confluence RPE Cultures

RPE cells within 2 to 3 days of visual confluence (early confluence; Fig. 1A ) were larger, less tightly packed, and more irregular in shape than cells in cultures that were maintained undisturbed at confluence for several weeks (late confluence; Fig. 1G ). In early confluence, actin microfilaments were organized in linear, stress fiber–type arrays or in incomplete circumferential bands, often displaced from the perimeter of cells (Fig. 1B) . With time at confluence, actin microfilaments showed an epithelioid pattern of circumferential bundles closely associated with cell borders (Fig. 1H) . In unselected RPE cultures of the type used in these experiments that have several epithelioid variants, ²⁵ circumferential actin bundles vary in size and compactness. 

In early confluence, N-cadherin was diffuse and nonjunctional, as well as localized to the borders of many cells, where staining was often punctate or segmented rather than completely zonular (Fig. 1C) . This immature pattern was more apparent after detergent extraction, which spared only the foci of detergent-resistant protein (Fig. 1D) . Proliferation continued in early confluence as indicated by Ki-67 staining in many RPE cell nuclei (Fig. 1C) . After several weeks at confluence, Ki-67-positive cells were infrequent (Fig. 1I) , and N-cadherin formed a nearly complete and detergent-resistant zonular band at cell borders (Fig. 1I 1J) . 

The distribution and detergent solubility of the cadherin-linked protein β-catenin was similar to that of N-cadherin. In the immature junctions of early-confluence RPE cultures, β-catenin distributed in a punctate pattern at cell borders (Fig. 1E) that was partially resistant to detergent extraction (Fig. 1F) . As junctions matured with time at confluence, β-catenin was more completely zonular (Fig. 1K) and largely detergent resistant (Fig. 1L) . Similar results were obtained for α-catenin and plakoglobin (not shown). 

Western blot analysis of detergent-soluble and -insoluble fractions of RPE cells illustrated the increasing resistance to extraction of N-cadherin and all catenins as the junction matured with time at confluence (Fig. 2) . The extent to which the proteins partitioned between soluble and insoluble fractions differed among proteins and varied among cell populations. For all proteins, however, there was a shift to increasing insolubility in late-confluence cultures. E-cadherin was undetectable in early-confluence cultures, but was found in late confluence, as previously reported, ¹⁷ where it partitioned to both fractions. 

Wound Closure in RPE Cultures at Different Confluence Intervals

The wound edge immediately after wounds were produced is illustrated in cultures immunostained for β-catenin (Figs. 1E 1F 1K 1L) . Within 1 day after wounding, RPE cells at the wound edge moved into the defect in both early- and late-confluence cultures, although the pattern of movement differed (Figs. 3A , 3E) . In early confluence, cells at the leading edge separated from one another and moved as single cells with linear actin filaments oriented in multiple directions (Fig. 3B) . Stable contacts among cells at the leading edge were infrequent. Only irregular accumulations of detergent-resistant N-cadherin and β-catenin were seen, and these are at the perimeter of some but not all cells (Fig. 3C 3D) . At 4 days, shortly after wounds were closed, the contiguous cells within wounds had variable patterns of linear and circumferential actin (Fig. 3J) and punctate distributions of N-cadherin (Fig. 3K) and β-catenin (Fig. 3L) . As shown in Figure 1C , the nuclei of most cells in early-confluence cultures were Ki-67 positive before wounding, and they remained positive throughout the 4 days of wound closure (not shown). 

In contrast to early-confluence cultures, cells in late confluence remained contiguous throughout wound closure and actin retained an epithelioid, circumferential organization (Fig. 3F 3N) . Cells spread in the direction of the defect and could become much larger, but cell–cell association was retained. Staining for detergent-resistant N-cadherin and β-catenin showed retention at cell–cell borders, although staining for the junctional proteins was separated into short segments (Fig. 3G 3H 3O 3P) rather than forming a complete zonular band, as before wounding (Fig. 1I 1J 1K 1L) . On day 1 after wounding, Ki-67-positive nuclei remained infrequent (not shown), similar to the intact postconfluent monolayer. However, by 4 days, many cells within wounds stained positively for this proliferation marker (Fig. 3O) . The segmented zonular pattern of junctional N-cadherin was similar in cells with and without Ki-67 staining (Fig. 3O) . 

Although the pattern of cell movement into wounds differed, the timing of closure of 1-mm linear wounds was similar, with both early- and late-confluence cultures showing a restored monolayer by 3 to 4 days. Six to 7 days after wound closure (day 10 after wounding), the site of the wound could not be identified in the early-confluence cultures because cells within the wound resembled those in the adjacent monolayer in density and gross phenotype (Fig. 4A ). However, the wounded region of late-confluence cultures remained identifiable on day 10 as a zone of less tightly packed cells that were irregular in shape, often still oriented in the direction of movement into the wound. Cultures that were maintained even longer at confluence before wounding showed delays in the timing of closure. In very-late-confluence cultures (approximately 6 months), wounds were not yet closed at 10 days after wounding (Fig. 4C) . Wounds in very-late-confluence cultures required 3 to 4 weeks for closure, and the site remained identifiable as a zone of less epithelioid cells for at least 2 months after wounding (not shown). 

Effects of Treatment with Cadherin Antibodies and Calcium Chelators

The retention of junctional N-cadherin during closure of late- but not early-confluence wounds suggested an increasing stability of the RPE adherens junction with time at confluence. Incubation of early-confluence cultures with function-blocking N-cadherin antibodies for 6 to 24 hours induced a partial loss of the immature junctional sites (Fig. 5) . However, the same antibody treatment had no detectable effect on junctional integrity in late-confluence cultures (Fig. 6) . (A 6-hour antibody treatment interval is shown for both [Figs. 5 6 ]). In postconfluent cultures, N-cadherin antibodies added to living cells stained the complete zonular junction (Fig. 6C) , codistributing with pancadherin staining (Fig. 6G) . The amount and distribution of pancadherin staining in N-cadherin antibody–treated cultures (Fig. 6G) did not differ from that in untreated control cultures (Fig. 6E) or in control cultures treated with mouse IgG (Fig. 6F) . Junctions in late-confluence RPE cultures were similarly unaffected by treatment with function-blocking E-cadherin antibodies, either alone or in combination with N-cadherin antibodies (not shown). N-cadherin antibodies that were bound to the junctions of postconfluent RPE cells during a6-hour incubation remained detectable for at least 3 days after antibody removal (Fig. 6D) . 

To further examine the stability of the calcium-dependent adherens junction, early- (Fig. 7) and late-confluence RPE cultures (Fig. 8) were treated with the calcium chelator EGTA. For early-confluence cultures, 5 minutes of EGTA treatment produced a loss of N-cadherin from cell–cell junctions (Fig. 7D) , and an increase in the prominence of the stress fiber organization of actin filaments (Fig. 7C) . Late-confluence cultures were more resistant to EGTA treatment, showing little effect after 5 minutes of treatment (not shown). After longer treatment intervals (30 minutes; Fig. 8D ), N-cadherin became displaced from sites of cell–cell contact, codistributing with actin in a contracted irregular ring in some cells and becoming largely diffuse in others (Fig. 8C) . By 45 minutes, N-cadherin was diffuse in most cells, although remnants of cadherin associated with the inner part of contracted actin rings could occasionally be found (Fig. 8F) . After 45 minutes in EGTA, the actin bundle was less compact and displaced from cell borders, but the overall circumferential actin pattern was retained (Fig. 8E) . When calcium levels in the medium were restored for cells treated for 45 minutes with EGTA, nearly complete zonular N-cadherin staining was recovered in 2 hours in both the absence (Fig. 8H) and presence of cycloheximide (Fig. 8J) . Immunostaining showed that the diffusely distributed N-cadherin in EGTA-treated cultures was partially susceptible to extraction with detergent buffers, but the cadherin associated with the contracted actin rings remained detergent resistant (Fig. 9) . Western blot analysis confirmed increased N-cadherin detergent solubility after EGTA treatment (Fig. 9E) . After 2-hour recovery in calcium-containing medium, the extent of partitioning of N-cadherin to the detergent insoluble fraction recovered and even became slightly greater than before EGTA treatment (Fig. 9E) . A greater protein insolubility after the 2-hour recovery was also observed for actin (Fig. 9E) . 

Similar to the retention of actin’s circumferential organization seen after short-term treatment of postconfluent RPE cells with EGTA, the zonular pattern of actin was also retained after prolonged treatment with low-calcium (5 μM) culture medium (Fig. 10) . N-cadherin became diffuse and nonjunctional after 1 to 2 days in low calcium medium and retained this distribution throughout the 1-month exposure period (Fig. 10D) . Circumferential actin filaments became less compact and less tightly associated with cell borders (Fig. 10C) , but actin did not revert to the stress fiber–type organization that was found in early-confluence cells (Fig. 1B) , despite the absence of junctional cadherin. 

Discussion

Confluence State, Junction Maturation, and In Vitro Wound Repair

Confluence is an operational term that refers to cultures in which the cells are contiguous. It is readily identified by visual inspection, but the biologic consequences of cell–cell contact at confluence are complex and not fully understood. Contact inhibition of cell growth and motility has been recognized for many years. ^{23 24} There seems to be a widespread impression that contact inhibition occurs relatively abruptly and completely at confluence, and cultures are therefore often described as subconfluent or confluent without consideration of the time the cells have been confluent. However, confluence-triggered cellular changes are time dependent, and some are gradual. Cessation of growth, for example, occurs over days or even weeks after confluence. ²⁶ As shown here, expression of the proliferation marker Ki-67 continues after confluence, and ongoing growth over weeks produces progressively denser RPE cultures. 

The development of an epithelial phenotype is another gradual process that is triggered by cell–cell contact at confluence. The process has been studied in some detail in the kidney epithelial cell line MDCK, which has served as a model of in vitro epithelial morphogenesis. In these cells, early stages of phenotype development occur during the hours after intercellular contact as junctions form and strengthen, ^{1 2 3 4} followed by a more protracted postconfluent maturation. ^{26 35} For RPE cells, however, even initial phenotype development is slow. RPE subpopulations selected for their highly epithelioid phenotype ^{15 16 25} and the less uniformly epithelioid RPE cultures grown by commonly used methods (shown here) require several weeks at confluence to develop epithelial features: an actin cytoskeleton that is largely circumferential and an adherens junction that is largely zonular and detergent stable. 

At least two features distinguish postconfluent from early-confluence RPE cultures: greater growth quiescence and a more highly developed adherens junction. Growth state is a well recognized variable affecting cell function, but differences in junctional development are likely to have many functional ramifications as well. Cell–cell junctions are signaling sites, in addition to sites of physical adhesion, and the signaling pathways initiated at junctions converge with pathways regulating growth, ^{36 37 38 39} motility, ⁴⁰ and cell–substrate adhesion. ⁴⁰ Because of the differing developmental states of their adherens junctions, early- and late-confluence RPE cultures may therefore be nearly as functionally divergent as subconfluent and confluent cultures of epithelial cell lines. Indeed, it is likely that cells maintained for extended postconfluent periods have many biologic properties that differ from cells at other stages of in vitro propagation. 

To begin to evaluate whether and how time at confluence affects RPE cell function, early- and late-confluence cultures were subjected to a commonly used experimental protocol: in vitro wounding. The response to injury differed, with early-confluence RPE cells migrating largely as single cells, and late-confluence cells moving as a contiguous sheet. The early-confluence movement pattern resembles the pattern shown by others who have examined wounds in RPE cultures shortly after they became confluent. ^{18 19 21} This migration pattern should not be considered characteristic of RPE, however, because simply allowing a longer period at confluence before wounding produced a different outcome. Occasionally, primary cultures of RPE are used for analysis. Cell number is a limitation of this type of culture, but it is presumed that cells recently harvested from eyes function more like cells in situ. In vitro wounds in primary RPE cultures close by movement of contiguous sheets of cells, ^{20 22} similar to the postconfluence cultures shown here. It appears that RPE cells that are propagated in culture by standard methods can be made to reacquire some of the desirable features of primary cultures if a long postconfluence interval is allowed for redevelopment of phenotype. 

Time at confluence for RPE cells affected other events in the wound repair process aside from migration pattern, including phenotype recovery in cells that filled the defect. As might be expected, in early-confluence RPE cultures, cells within newly closed wounds were almost immediately indistinguishable from those in the adjacent monolayer, where cells were only a few days ahead in phenotype development. Cells within late-confluence wounds, however, required several additional days to reacquire the appearance and density of cells outside the wounds. After very long postconfluence periods (months), phenotype recovery was even more protracted and possibly indefinite. In this regard, postconfluent cultures may resemble RPE cells within eyes, which exhibit an irregular phenotype at sites of a retinal detachment, ⁴¹ suggesting that cell shape may not be restored after damage to the monolayer in situ. 

Although postconfluent cultures cannot be considered identical with the RPE monolayer within eyes, their slower growth and better developed junctions may make late-confluence cultures a more relevant model for RPE functional analyses than cultures in early confluence. 

The Mature Cadherin Junction of Postconfluent Cells

As indicated previously, the development of the adherens junction has been studied in some detail, especially in epithelial cell lines that express E-cadherin. However, less has been done to probe the properties of the adhesion after it has formed. In cells with nascent E-cadherin or N-cadherin adhesions, anti-cadherin antibodies with function blocking properties have been used to induce cell dissociation. ^{30 31} The mechanism whereby this occurs has not been explored in detail, but the presumption is that antibodies directed against the extracellular cadherin domain compete with the cadherin on adjacent cells and inhibit binding of the newly delivered cadherin. Antibody treatment would therefore disrupt formed junctions as they turn over. Early-confluence cells showed partial junctional loss with function-blocking N-cadherin antibody treatment, but antibody treatment had no effect on junctional integrity in postconfluent cells. On the assumption that even low levels of E-cadherin in postconfluent cells ¹⁷ might stabilize the junction, function-blocking E-cadherin antibodies were also used, and still no junction disassembly was observed. It appears that the cadherin adhesion of RPE cells, although slow to form, may be become stable with a low turnover. This conclusion is supported by the observation that N-cadherin antibodies that were bound to the zonular junctions of postconfluent cells remained detectable for several days after antibody removal. 

Increasing junctional stability in RPE cells with time at confluence was also indicated by an increasing resistance of junctional proteins to detergent extraction. Junctional proteins are not alone, however, in showing a time-dependent increase in detergent resistance. The amount of total cellular protein that is detergent stable increases after confluence at a rate that is similar to junctional proteins, so that the relative amount of cadherin and catenins remains nearly constant in detergent resistant protein extracts. ²⁶ This observation was interpreted to suggest that junctions are among the protein compartments of epithelial cells that form increasingly stable cytoskeletal associations during postconfluence maturation. ²⁶  

Another indicator of the stability of the formed junction of RPE cells was its maintenance during wound closure, despite the upregulation of cell movement and growth. In postconfluent RPE cultures, cells moving into the wound often spread dramatically and, although N-cadherin junctions separated into segments, cell–cell contact was retained. For E-cadherin junctions in other cells, adhesions first form as puncta that later coalesce into linear segments, then complete zonular arrays. ^{1 2 3} Similar punctate and segmented accumulations of N-cadherin are seen during the postconfluence formation of the zonular RPE junction. ¹⁵ With migration into the wound from a postconfluent monolayer, the N-cadherin junction again becomes segmented, suggesting that the mature zonular junction consists of discrete segmental units that give the junction plasticity, allowing cells to spread without losing intercellular contact. 

Yet another indicator of the stability of the formed RPE junction was its resistance to disruption by calcium depletion. Longer EGTA treatments were required to displace N-cadherin from mature junctions in postconfluent cells than immature junctions in early-confluence cultures. An unexpected feature of the cadherin junction in postconfluent RPE cells was the stability of the circumferential actin bundle, even when cadherin adhesion was inhibited for weeks by low-calcium treatment. As junctions form, the circumferential actin bundle develops in parallel with the zonular cadherin adhesion ¹⁵ to which it is linked. It might be expected that actin would lose its circumferential distribution when cadherin adhesion is blocked for extended periods. However, although the actin bundle became less compact and partially displaced from lateral cell borders, the circumferential actin array was maintained with a prolonged low-calcium treatment that rendered N-cadherin nonjunctional. All cadherin-mediated adhesion would be blocked by this protocol, including the dominant RPE cadherin (N-cadherin), less abundant RPE cadherins (E- and P-cadherin ¹⁷ ), and other unidentified cadherins, should they exist. Perhaps other intercellular adhesion proteins in the RPE ¹⁵ help maintain the formed circumferential actin bundle in the absence of cadherin linkage. 

With EGTA treatment, the actin in some RPE cells contracted within minutes into a smaller ring, presumably because the disruption of cadherin adhesion by rapid calcium removal produced a relatively abrupt release of tension on the circumferential actin bundle. N-cadherin remained transitorily associated with the contracted actin ring, then developed a diffuse extrajunctional distribution. The diffuse N-cadherin may be less tightly associated with the cytoskeleton after EGTA treatment, because treatment renders N-cadherin more soluble in detergent buffer. Nonetheless, restoration of calcium levels to cells in which cadherin had become largely diffuse produced a rapid recovery of the zonular N-cadherin distribution and of detergent insolubility of the protein. Junctional recovery was accompanied by an increased stability of the actin cytoskeleton as well, which also became more detergent resistant. Because the junction reformed rapidly and in the presence of a protein synthesis inhibitor, it appears that the junction is reassembled from preexisting rather than newly synthesized cadherin. How this is accomplished is not yet determined. Little is currently known about the turnover of mature cadherin adhesions, although a recent investigation of the E-cadherin junction of MDCK cells indicated that E-cadherin is continuously removed from and shuttled back to the cell surface in endocytic vesicles and that calcium chelation can induce cadherin internalization. ⁴² It remains to be determined whether the N-cadherin of RPE cells is handled in the same way. 

The uncoupling of the zonular distribution of cadherin and actin by calcium removal may provide a clue to why the adherens junction in RPE cells is slow to form. When the junction forms de novo in RPE cells that were fully dissociated, actin often has a stress fiber–type organization and the process of developing a zonular organization of both N-cadherin and actin requires weeks at confluence. However, the reformation of a zonular junction in cells in which actin is already circumferential, as after calcium removal and restoration, is rapid, requiring only hours. This time differential suggests that reorganizing the dominant actin network from linear to circumferential arrays may be the step in the process that retards the de novo formation of a zonular junction. The importance of the organization of actin filaments to the development of an epithelial-type adherens junction was previously recognized. ⁴³ Indeed, most cells that express N-cadherin as their dominant cadherin are not epithelial cells, and determining how RPE cells organize actin into a circumferential epithelial pattern may therefore be critical for understanding RPE morphogenesis. 

²  Present address: Osaka City University Medical School, Japan.

 Supported by National Institutes of Health Grants R01-EY10832 (JMB) and P30-EY01931, and by unrestricted grants from the Posner Foundation (Milwaukee, Wisconsin) and Research to Prevent Blindness. JMB is a recipient of a Senior Scientific Investigator Award from Research to Prevent Blindness.

 Submitted for publication November 5, 1999; revised April 24, 2000; accepted May 22, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Comparison of early- (2–3 days, A through F) and late- (8 weeks; G through L) confluence RPE cultures. Phase-contrast appearance and distribution of actin, N-cadherin, and β-catenin are shown. Nuclear staining for the proliferation marker Ki-67 (costained with N-cadherin, C, I) ranges from weak (arrowheads) to intense (arrows). For N-cadherin and β-catenin, cultures without (−deterg) and with (+deterg) detergent extraction are shown. (E, F, K, L, arrows) Wound edge immediately after wounding. Scale bar applies to each pair of images, (A, G), 200μ m; all others, 10 μm.

Figure 2.

 

Western blot analysis of extracts from early- and late-confluence RPE cell populations from three donors (cell pop 1–3). Detergent-soluble (S) and -insoluble (I) fractions were extracted in equivalent volumes and loaded equivalently on gels. Blots were probed for N-cadherin,α -catenin, β-catenin, plakoglobin, and E-cadherin. For N-cadherin, the entire blot is shown to illustrate that there was a single major band. A single major band was also found for all other proteins; the region of the blot showing the band is provided. Molecular mass markers (right, molecular masses in kilodaltons) were run on all gels.

Figure 3.

 

Early- and late-confluence RPE cultures on day 1 (1d) and day 4 (4d) after wounding. In all images, the wound is to the right. (A, B, arrows) Position of the initial wound edge. For the day 1 wounds (A through E), the distribution of actin, N-cadherin (N-cad), and β-catenin (β-cat) is shown in cells at the leading edge of migration into the wound. At day 4 wounds were closed (I through P). Distribution of the same proteins is shown in cells within the wound. Cells stained for β-catenin within day 4 late-confluence wounds are contiguous (P), but the entire perimeter of all cells is not shown within the high-magnification field because of extreme spreading in some cells. For N-cadherin and β-catenin, cultures were extracted before fixation to show detergent-resistant protein. Ki-67 costaining with N-cadherin is shown to illustrate increased proliferation in cells within day 4 wounds in late-confluence cultures (O). Scale bar applies to each group of four images showing the same feature, (A, E, I, M), 200 μm; all others, 10 μm.

Figure 4.

 

Phase-contrast micrographs of early- (A), late- (B) and very late- (C) confluence cells on day 10 after wounding. Arrows: Position of the original wound edge; wounds are above the arrows. Wounds in early- (2–3 days) and late- (8 weeks) confluence cultures had been closed for 6 to 7 days, and cells within wounds in the early- but not the late-confluence cultures resembled cells within the adjacent monolayer. For very-late-confluence cultures (6 months), cells were not yet contiguous within wounds. Scale bar, 200 μm.

Figure 5.

 

Early-confluence RPE cultures in an N-cadherin function-blocking antibody experiment costained to show N-cadherin and pancadherin. (A, C) Control culture treated for 6 hours with mouse IgG and costained for N-cadherin and pancadherin. (B, D) Culture treated for 6 hours with monoclonal N-cadherin antibodies, then fixed and stained as usual for pancadherin to show partial disruption of junctions. For N-cadherin, the secondary antibody alone was used after fixation to show that N-cadherin antibodies did not bind to and remain detectable at junctions (compare with postconfluent cells in Fig. 6C ). Scale bar, 5 μm.

Figure 6.

 

Postconfluence RPE cultures in an N-cadherin function-blocking antibody experiment costained to show N-cadherin and pancadherin. (A, E) Control, untreated culture costained for N-cadherin and pancadherin. (B, F) Control culture treated for 6 hours with mouse IgG, fixed, and stained as usual for pancadherin. For N-cadherin, the secondary antibody alone was used after fixation to show no nonspecific binding of the IgG to junctions. (C, G) Culture treated for 6 hours with monoclonal N-cadherin antibodies, fixed, and stained as usual for pancadherin. For N-cadherin, the secondary antibody alone was used after fixation to show N-cadherin at junctions. (D, H) culture treated for 6 hours with N-cadherin antibodies, followed by removal of the antibody and fixation of the cells 3 days later. Staining is as described for (C, G). Scale bar, 5 μm.

Figure 7.

 

Early-confluence RPE cultures costained for actin and N-cadherin, before treatment (A, B) and after a 5-minute treatment with EGTA (C, D). After 5 minutes in EGTA, actin stress fibers became more prominent in many cells (C) and N-cadherin staining became nonjunctional and diffuse (D). Scale bar, 10 μm.

Figure 8.

 

Late-confluence RPE cultures costained for actin and N-cadherin before treatment (A, B) or after treatment with EGTA. After 30 minutes in EGTA (C, D), the actin circumferential bundle was displaced from cell borders and appeared out of focus because the filaments were less compact and the bundle was thicker. N-cadherin codistributes in many cells with the center of the contracted actin ring. Arrows: Same positions in (C) and (D). By 45 minutes N-cadherin was diffuse in most cells, although faint codistribution with actin rings could be seen in some cells (E, F, arrows). (G through J) Cultures treated for 45 minutes with EGTA followed by 2-hour recovery in medium containing normal calcium in the absence (−CHX) or presence (+CHX) of cycloheximide (CHX). Scale bars apply to each actin and N-cadherin pair, 5 μm.

Figure 9.

 

Late-confluence RPE cultures treated with EGTA. Fluorescence micrographs showing actin and N-cadherin costaining after 30 minutes of EGTA treatment without (A, B) or with (C, D) extraction with a detergent buffer. Extraction slightly diminished the diffuse N-cadherin staining, but the ringlike staining was unaffected. (E) Western blot analysis showing detergent-soluble (S) and -insoluble (I) extracts of cultures without EGTA treatment (−EGTA) after 45 minutes of EGTA treatment (+EGTA), when N-cadherin was largely diffuse (see Fig. 8F ), and after 45 minutes of EGTA treatment followed by a 2-hour recovery in medium containing normal calcium (recov). Blots were probed for N-cadherin and actin. Right: Molecular mass (in kilodaltons). Scale bar, 5μ m.

Figure 10.

 

Late-confluence RPE cultures costained for actin and N-cadherin. After 8 weeks at confluence, cultures were incubated for an additional 1 month in medium containing normal calcium (1.8 mM: A, B) or low calcium (5 μM: C, D). After low-calcium incubation, the actin bundle was less compact and displaced from cell borders but still circumferential (C), whereas N-cadherin was nonjunctional (D). Scale bar, 10μ m.