Physical disruption of cell–cell contact induces VEGF expression in RPE cells

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: To investigate the role of RPE cell–cell contact in vascular endothelial growth factor (VEGF) protein expression in cultures of primary human RPE (hRPE) cells and a human RPE cell line (ARPE-19).

Methods: Two in vitro methods, scratching and micropatterning, were used to control the physical dissociation of RPE cell–cell junctions. Scratching was performed by scoring monolayers of RPE cells with a cell scraper. Micropatterning was achieved by using a stencil patterning method. Extracellular VEGF expression was assessed by using an enzyme-linked immunosorbent assay (ELISA) kit. Immunocytochemistry (ICC) was performed to visualize the expression and localization of VEGF and intercellular proteins zonula occludens-1 (ZO-1), N-cadherin, β-catenin, and claudin-1 in RPE cultures.

Results: Higher expression of VEGF protein by cells on the edges of the scratched RPE layers was confirmed with ICC in short-term (1 day after confluency) and long-term (4 weeks after confluency) cultures. According to the ICC results, ZO-1, N-cadherin, β-catenin, and claudin-1 successfully localized to cell–cell junctions in long-term cultures of ARPE-19 and hRPE cells. However, unlike N-cadherin, β-catenin, and claudin-1, only ZO-1 localized junctionally in short-term cultures of both cell types. Moreover, removing cell–cell junctions by scratching resulted in the delocalization of ZO-1 from tight junctions to the cytoplasm. The loss of tight junction formation and the accumulation of ZO-1 in the cytoplasm correlated with increased VEGF expression. Micropatterning RPE cells on different sized circular patterns produced varying concentrations of cells with lost cell–cell junctions. When fewer cells formed intercellular junctions, increased extracellular VEGF secretion was observed from the ARPE-19 and hRPE cells.

Conclusions: VEGF expression increases after physical disruption of RPE cell–cell connections. This increase in VEGF expression correlates with the loss of intercellular junctions and the localization of ZO-1 in the cytoplasm of RPE cells.

The exudative (wet) form of age-related macular degeneration (AMD) is characterized by the abnormal growth of new leaky blood vessels in the choroid (choroidal neovascularization, CNV) and near the macula. CNV can cause RPE deformation and degeneration, leading to the irreversible loss of vision [1,2]. Although the exact causes of CNV are not completely understood, RPE-derived vascular endothelial growth factor (VEGF), a potent angiogenesis factor, is generally thought to be the major stimulator of CNV [3-10].

Appropriate levels of VEGF are crucial for the normal development of the choroid [11,12]. VEGF also functions as an important factor in maintaining RPE and endothelial cells [11]. However, abnormal levels of VEGF are also associated with retinal disease [3]. Moreover, overexpressing VEGF in rat RPE results in the development of CNV [4,5]. Accordingly, VEGF has been the foremost target in many experimental studies and clinical trials to inhibit CNV. The most successful treatment for CNV in wet AMD uses recombinant anti-VEGF to antagonize VEGF, slow vision loss, and improve visual acuity [6-10].

Even though anti-VEGF products slow the progression of CNV, there is no cure or prevention for CNV associated with wet AMD. The exact mechanisms resulting in the overexpression of angiogenic factors, including VEGF, in RPE cells remain unknown. A wide range of molecular and environmental factors has been implicated in elevated VEGF expression by RPE cells, including hypoxia [13-16] and inflammation due to increased levels of inflammatory cytokines or drusen components, such as C3a, C5a, and amyloid β [17-19].

Reduced RPE cell–cell adhesion, caused by RPE tears or RPE cell death in the latest stages of dry AMD, may also elevate VEGF gene expression. RPE tears occur during AMD from RPE detachment or CNV [20-24] and most commonly from intravitreal injection of anti-VEGF drugs during treatment [24-28]. RPE cell death, mediated by apoptosis and/or necrosis, in geographic atrophy (GA) is another in vivo phenomenon through which the physical contact between RPE cells is lost [29-31]. Two separate studies reported increased mRNA levels of VEGF after calcium-mediated dissociation of RPE cell–cell junctions [32,33].

However, because the exact effect of extracellular calcium ions on VEGF expression is unclear, alternative in vitro methods may elucidate the role of physical cell–cell adhesion in VEGF expression. Moreover, none of these studies, to our knowledge, has demonstrated how junctional cell–cell detachment affects the expression of the VEGF protein. In this work, we used two in vitro methods, scratching and micropatterning, without introducing exogenous components to study the role of RPE cell–cell adhesion in VEGF protein expression.

Scratching assays, also known as wound healing assays, are generally used to mimic tissue damage and study the proliferation and migration capability of cells [34,35]. Scratching has also been used to study the molecular events associated with the loss of cell–cell contact in RPE cells [36]. Specifically, the expression of platelet-derived growth factor (PDGF) increased in cells on the leading edge of a scratched RPE layer [37]. Another analysis demonstrated that scratching a monolayer of ARPE-19 cells altered the expression of the genes encoding hepatoma-derived growth factor (HDGF), mitogen-activated protein kinase, CD44, and other proteins [38]. Here, we used scratching assays to mimic damage to the RPE monolayer resulting from RPE degeneration in AMD and to study the changes in VEGF expression in RPE cells surrounding the scratched edge.

Although the scratching assay provides a view of differential VEGF expression in RPE cells around the edges, the extent of cell–cell contact loss cannot be well controlled. Accordingly, we also developed a micropatterning method to control cell–cell contact and study VEGF expression. Micropatterning techniques provide a platform to control spatial distribution of cells on various substrates and reconstruct the architecture of in vivo tissue. Micropatterning is a powerful tool for engineering the differentiation and function of cells [39,40], making in vitro tissue models [41], and studying cell–cell interactions [42]. A variety of micropatterning methods with different complexities and applications is currently available [43]. Most of these methods use cell adhesion proteins to promote cell adhesion to specific areas of substrates [44]. However, it has been shown that cell adhesion proteins can affect VEGF expression [14]. Stencil patterning, a new method of patterning, is a straightforward and reproducible alternative technique, which eliminates the need for cell adhesion proteins [45,46].

In the present study, an in vitro scratching method was used to mechanically dissociate cell–cell junctions in ARPE-19 (a human RPE cell line derived from a 19-year-old man) and human primary RPE (hRPE) cells. Both cell types are capable of expressing RPE markers [47-50]. Localized VEGF expression in scratched samples was assessed using immunocytochemistry (ICC). We also localized tight and adherens junctions in short-term (1 day after confluency) and long-term (4 weeks after confluency) RPE samples by staining N-cadherin, β-catenin, claudin-1, and ZO-1 to examine the formation of intercellular junctions. N-cadherin is the most abundant cadherin in RPE cells that mediates cell–cell adhesion at adherens junctions and modulates cellular phenotype [51-55]. Cadherins contribute to the formation of adherens junctions by associating with β-catenin [56,57]. β-catenin plays a key role in the formation of cadherin–catenin complexes through linking cadherins to α-catenin and actin fibers [58]. Claudin-1 is a member of the family of claudins and a marker of tight junctions [59,60]. ZO-1, another tight junction protein, belongs to the membrane-associated guanylate kinases (MAGUK) family [61]. VEGF phosphorylates ZO-1 and increases RPE permeability by disrupting tight junctions. Increased levels of VEGF also reduce ZO-1 expression in mesothelial, endothelial, and epithelial cells [62-66]. A recent study showed that in endothelial cells, reduced ZO-1 expression is associated with alterations in VEGF expression and cell proliferation [67].

A polydimethylsiloxane (PDMS) stencil patterning method was used to make micron-sized circular patterns (as small as 100 μm in diameter) of the ARPE-19 and hRPE cells. Growing cells on different size patterns produced varying concentrations of cells with lost junctional adhesions, allowing quantitative analysis of changes in VEGF expression. Here, using scratching assays and micropatterning techniques, we show a correlation between the loss of intercellular junctions and increased VEGF expression in RPE cells. Our findings add new insights into the mechanisms of increased VEGF expression from RPE cells during retinal degenerative diseases, such as AMD.

The effect of losing RPE–RPE adhesion on VEGF expression was first investigated using scratching assays and ICC. ARPE-19 and hRPE cells were grown for 4 weeks (long-term) after they had reached confluency and then scratched using a cell scraper. One day after scratching, the cells were stained with anti-human VEGF antibody, and the nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI; Figure 2). In addition, ICC was performed on the scratched cultures of the ARPE-19 and hRPE cells that had been grown for only 1 day after they reached confluency (short-term cultures). Confocal imaging of the scratched samples showed that disrupting RPE–RPE contact amplified VEGF expression in the cells on the edges of the scratched areas in the short- and long-term RPE cultures (Figure 2). In the confocal images, a marked decrease in the net signal intensity for VEGF was observed from the scratch edge toward the middle of the monolayer in all scratched samples (Figures 2D,H,L,P). VEGF intensity was also averaged over two intensity bins at 0–100 μm and 100–200 μm from the scratch edge. For the ARPE-19 and hRPE scratch samples, the average signal intensity of the VEGF channel was higher in the 0–100 μm distance. To visualize the formation of the intercellular junctions, we used ICC to stain for the tight junction components (ZO-1 and claudin-1) and the adherens junctions (N-cadherin and β-catenin) in the short- and long-term cultures of the ARPE-19 (Figure 3) and hRPE (Figure 4) cells. According to the ICC results, ZO-1, N-cadherin, β-catenin, and claudin-1 were localized to the intercellular space in the long-term cultures (n = 3). ZO-1 was present in 88 ± 5.0% of the cell–cell junctions in the ARPE-19 cells and 98 ± 2.0% of those in the hRPE cells (Table 1). High levels of junctional localization were also seen for N-cadherin (93 ± 6.0% in the ARPE-19 cultures and 69 ± 17% in the hRPE cultures), β-catenin (82 ± 8.0% in the ARPE-19 cultures and 49 ± 13 in the hRPE cultures), and claudin-1 (only 10 ± 5.0% in the ARPE-19 cultures and 53 ± 6.0% in the hRPE cultures, Table 1). In the short-term cultures (n = 3), however, only ZO-1 was junctionally localized, with 64 ± 14% and 84 ± 7.0% of the junctions covered with ZO-1 in the ARPE-19 and hRPE cultures, respectively (Table 1). The percentage of junctional coverage with claudin-1 and β-catenin in both cell types and N-cadherin in the hRPE cultures was less than 10% in the short-term cultures. In the short-term cultures of the ARPE-19 cells, only slightly more than 25% of cell junctions were covered with N-cadherin (Table 1). We also performed a confocal z-stack scan on the long-term ARPE-19 and hRPE cultures grown on the porous cell culture inserts to confirm the polarization of the RPE cells. The results showed that in the long-term ARPE-19 cultures, although ZO-1 successfully localized to the apical regions, VEGF failed to polarize properly because its expression was not localized to either side of the culture (Figure 3M). In the long-term hRPE cultures, ZO-1 was localized to apical and intercellular areas, and the expression of VEGF was mainly basolateral, confirming the polarization of cells after 4 weeks (Figure 4M).

Results of the scratching experiments followed by ZO-1 and VEGF staining showed that in the ARPE-19 and hRPE cultures, VEGF expression was higher on the edges of the cell patches, where the cell–cell contact was missing, and lower in the confluent areas of the monolayer with higher ZO-1 junctional localization. VEGF expression was inversely proportional to the localization of ZO-1 to the intercellular junctions. In other words, with less junctional localization of ZO-1, VEGF expression was higher and vice versa (Figure 5A–C).

Based on the scratching results, we hypothesized losing intercellular junctions, and therefore, reduced cell–cell interactions, may contribute to increased VEGF expression. Inducing VEGF expression from cells with a free edge in the short-term scratched cultures suggests that intercellular junctions that occur early after cell–cell contact, such as ZO-1, may be involved in VEGF regulation. To verify these results, we used a micropatterning method to quantitatively study and control the loss of cell–cell contact on VEGF expression in short-term cultures of ARPE-19 and hRPE cells.

For micropatterning, a PDMS stencil micropatterning method was used to make micropatterns of the RPE cells. The SEM images showed that the diameter of the through holes in the PDMS membranes was 102.7 ± 4.60, 213.4 ± 3.800, and 307.3 ± 10.50 μm for the 100, 200, and 300 µm patterns, respectively (n = 9; Figure 6A–C; Table 2). These membranes made consistent circular patterns of ARPE-19 and hRPE cells on the glass substrates. Figure 6D–F shows ARPE-19 cells attaching to the exposed glass in the presence of PDMS stencils. Figure 6G–I shows single patches of hRPE cells patterned on glass coverslips after the PDMS stencil was removed. Figure 7A shows that varying concentrations of cells on the periphery of the patterns could successfully be produced using this micropatterning method. The average cell patch size was 101.5 ± 4.000, 213 ± 11.0, and 311.5 ± 10.00 μm in diameter, and the average cell count per patch was approximately 6.0 ± 2.0, 28 ± 6.0, and 96 ± 16, respectively (n = 21; Table 2). Patterning on the smallest pattern size (100 μm) produced the highest concentration of peripheral cells with approximately 93% of all cells losing junctional cell–cell adhesion at least on one side. This percentage decreased to 54 and 33 for the 200 and 300 μm pattern sizes, respectively (n = 21; Figure 7A; Table 2). The size of the cell patches increased for both cell types 48 h after the stencil was removed, with this increase greater for ARPE-19 patches (Figure 7B). The viability of micropatterned ARPE-19 and hRPE cells was greater than 93% for all pattern sizes.

ICC was used to localize the expression of ZO-1 and VEGF in the micropatterned samples. The results showed that the expression of VEGF was higher on the periphery of the ARPE-19 and hRPE micropatterns. Similar to the scratched samples, this increase in VEGF expression was associated with the translocation of ZO-1 from the junctional zones (Figure 5D–I). VEGF expression in the spent media of the patterned ARPE-19 and hRPE cells was analyzed using a VEGF ELISA kit (n = 9). To quantitatively compare VEGF expression in cells grown on different pattern sizes, specific VEGF expression was calculated by dividing the total VEGF measured in the spent media after 24 h by the total number of cells (VEGF/ml.cell). With smaller pattern sizes, increasing levels of VEGF expression were observed from the ARPE-19 and hRPE cells (Figure 7C,D). To further investigate the effect of the patch size, the cells were grown for another 24 h in fresh media before the spent media were analyzed for VEGF expression. Specific VEGF expression slightly decreased at day 2. This change in VEGF expression at day 2 was greater in the ARPE-19 patterns (Figure 7C,D).

Although there is a large body of evidence linking increased VEGF expression in the RPE to pathologic neovascularization of the retina [3-10], the exact mechanisms leading to the imbalance of VEGF in wet AMD are not fully understood. The failure of RPE–RPE adhesion is known as a consequence of neovascularization but is not usually considered a mechanism that contributes to the pathogenesis of CNV. The present study demonstrates that losing RPE–RPE attachment can induce VEGF overexpression and that early cell–cell contact can regulate VEGF expression.

Scratching and micropatterning methods were developed to study the role of junctional RPE adhesion on VEGF expression. In particular, immunostaining the scratched samples for VEGF showed a higher level of VEGF in cells with lost junctional adhesion mainly on the edges of the RPE cell monolayers (Figure 2), suggesting a regulatory role for RPE cell–cell adhesion in VEGF expression. Junctional localization of claudin-1, N-cadherin, β-catenin, and ZO-1 in long-term hRPE cultures and the polarization of VEGF and ZO-1 expression demonstrated the proper formation of the intercellular junctions and the maturity of the hRPE cells. Despite the lack of proper polarization of ARPE-19 cells in the long-term cultures, the effect of losing cell–cell adhesion on VEGF expression in the ARPE-19 cells was similar to that in the hRPE cells. Aberrant localization of VEGF in ARPE-19 cells has previously been observed [50,69]. Here, we found that overexpression of VEGF occurs in short-term cultures, implying that early contact between neighboring cells is a key step in VEGF regulation. Therefore, it is likely that early-forming junctional complexes have a role in regulating VEGF expression.

Figure 2 shows that VEGF is induced in the RPE cells at the edges of the scratched areas. According to the ICC results (Figure 3 and Figure 4), ZO-1 is one of the first junctional proteins to localize to intercellular zones. Figure 5 shows that the delocalization of ZO-1 correlates strongly to increased VEGF expression in the scratched zones and along the periphery of the micropatterns. ZO-1 is an important component of intercellular signal transduction in epithelial and endothelial cells [70,71]. Previous studies have implicated VEGF in the phosphorylation and delocalization of ZO-1 [66,72]. A recent study showed that changes in ZO-1 expression can affect VEGF expression in endothelial cells [67]. Moreover, aberrant angiogenesis has been reported in the yolk sac of ZO-1-deficient mice [73]. In RPE cells, inhibiting ZO-1 expression induces proliferation. More effort is needed to clarify whether ZO-1 or other junctional proteins that contribute to early stages of junction formation can regulate VEGF expression in RPE cells. Nevertheless, our results show a clear relationship between the loss of junctional RPE cell adhesion and VEGF expression.

To quantitatively evaluate the hypothesis that the observed increase in VEGF expression is due, at least in part, to reduced cell–cell contact, we produced varying concentrations of cells with reduced cell–cell contact by growing RPE cells on different pattern sizes (Table 2; Figure 7A) using micropatterning. One advantage of the physical detachment of cells by micropatterning compared with chemical methods, such as calcium-mediated dissociation, is that it eliminates any possible side effect on VEGF expression from the chemicals typically used for dissociating cellular junctions [74]. By using PDMS stencils, we also eliminated the need for cell adhesion proteins, which may affect VEGF expression [14]. The quantitative ELISA results from the micropatterning experiments confirmed our hypothesis that losing cell–cell adhesion increases VEGF expression (Figure 7C,D). These findings confirm previous observations from calcium-mediated RPE dissociation studies [32,33]. In the patterning experiments, the ARPE-19 cells behaved similarly to the hRPE cells in terms of VEGF expression. The VEGF expression from ARPE-19 cells was only slightly lower than that of the hRPE cells. The decreased VEGF expression at day 2 of patterning, observed for both cell types, is presumably due to cell growth leading to an increase in patch size and consequently the decreased concentration of the cells on the edges of the patterns. This change in VEGF expression at day 2 decreased to a greater extent in the ARPE-19 cells than in the hRPE cells, potentially due to the shorter doubling time of the ARPE-19 cells compared with that of the hRPE cells, leading to faster growth of the size of the cell patches (Figure 7B–D).

During the latest stages of dry AMD, the loss of physical contact between RPE cells occurs due to RPE cell death in GA [29-31]. In a study on patients with AMD with GA, CNV started at the peripheral borders of the atrophic sites in nearly 70% of the eyes that developed CNV, implying that cell–cell contact may have a role in the progression of dry AMD to wet AMD [75]. RPE cells may also physically detach from each other because of breaks (tears) in the RPE monolayer. RPE tears occur spontaneously secondary to neovascular AMD or after intravitreal injection of antiangiogenic drugs [20-28]. The present findings suggest that physical perturbations in the RPE, which may result from RPE degeneration, RPE tears, drusen formation, or apoptosis associated with aging, can induce the initiation and/or progression of CNV due to the overexpression of VEGF. Particularly, our results are consistent with a previous finding that the expression of VEGF isoforms was markedly increased in the RPE cells along the edges of a surgically removed subretinal vascular membrane of a patient with AMD [33]. The present results may also explain observations that CNV is associated with the surviving RPE cells in atrophic areas of retinas with AMD [76]. It is possible that this association of CNV with surviving RPE cells may be a consequence of tight junction loss and increased VEGF expression in and from RPE cells at atrophic sites. Given that CNV occurs during the latest stages of AMD, it may be possible to prevent CNV by implementing anti-VEGF therapies earlier, as soon as defects in RPE–RPE integrity are detected. Moreover, the present findings suggest that anti-VEGF therapies may be required even after the development of RPE tears to prevent excessive VEGF expression due to the loss of cell–cell contact in the RPE.

The ELISA kit used here can quantify the levels of VEGF₁₆₅. This isoform is the most potent and most abundant VEGF isoform in the RPE [77,78]. Nevertheless, further research is needed to study the effect of losing cell–cell contact on the expression of other isoforms of VEGF, such as VEGF₁₂₁, and to identify the association between RPE cell–cell signaling and VEGF overexpression in wet AMD. Work in this direction is ongoing in our laboratory.