Rough polystyrene substrates were prepared for the HGK culture according to our previously described methods [10]. Hydrophobic polystyrene culture dishes were rubbed with silicon carbide papers (#4000 for the low-nanometer dimension, #1200 for the mid-nanometer dimension, and #200 for the high-nanometer dimension) under wet conditions for 3–4 minutes. In our previous reports, we showed that the roughness value rapidly increased during the first 2 minutes after the beginning of the rubbing and then remained stable [10]. Biases in the orientation of the scratches were reduced by moving the silicon papers in multiple directions. The substrate roughness was measured using an atomic force microscope (XE-100, Park Systems, Suwon, Korea). Three samples were obtained from each polystyrene dish. The roughness value of each sample was measured at 3 points. The roughness value of the smooth culture dish was measured at 3 points in each hydrophobic polystyrene dish. Substrates with the following average roughness (Ra) were obtained: 121.3±13.4 nm by rubbing with #4000 silicon carbide papers, 505.3±115.3 nm by rubbing with #1200 silicon carbide papers, 867.0±168.6 nm by rubbing with #200 silicon carbide papers, and 5.4±1.4 nm for the smooth dishes (Table 1, Figure 1). The Ra value of the enamel surface has been reported to be in the range of 37.0–127.9 nm [1112]. The Ra value of the root surface has been reported to be in the range of 0.41–1.12 µm [4]. Various methods of root planing have been reported to produce a root surface roughness in the range of 0.35–4.90 µm depending on the instruments used [451314]. The plates with an Ra value corresponding to a high-nanometer dimension (867.0±168.6 nm) and a mid-nanometer dimension (505.3±115.3 nm) were rougher than the enamel surface and within the range of the Ra reported on the untreated root surface or the root surface after root planing. Thus, the substrates with an Ra value corresponding to a low-nanometer dimension (121.3±13.4 nm) represented enamel surface that has been roughened physiologically to a greater or lesser extent. Acid-etching produces a roughened enamel surface in the range of 150–450 nm [1112]. The substrates with an Ra corresponding to a high-nanometer dimension (867.0±168.6 nm) and a mid-nanometer dimension (505.3±115.3 nm) represented untreated root surfaces or root surfaces after root planing. The 3 types of culture dishes with varying levels of roughness produced by this method showed a significant difference in the Ra (P<0.05). The substrates were coated with fibronectin (0.2–0.4 mg/mL) to facilitate cell adhesion because rubbing the surface exposed the hydrophobic polystyrene.

Antibodies for glyceraldehyde 3-phosphate dehydrogenase (GAPDH), c-Jun N-terminal kinase (JNK), phospho-c-Jun N-terminal kinase (p-JNK: Thr183/Tyr185), E-cadherin, ×10 cell lysis buffer, and horseradish peroxidase (HRP)-linked anti-rabbit immunoglobulin G (IgG) were purchased from Cell Signaling Technology (Waltham, MA, USA). Fluorescein isothiocyanate-labeled phalloidin (FITC-phalloidin), SP600125 (a JNK inhibitor), anisomycin (a JNK activator), puromycin, and 4′, 6-diamidino-2-phenylindole dihydrochloride (DAPI) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Cy3-conjugated anti-rabbit IgG antibody was obtained from Jackson ImmunoResearch (West Grove, PA, USA). The JNK small hairpin RNA (shRNA) (shJNK) plasmid was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Lipofectamine LTX and Plus reagents were obtained from Invitrogen (Carlsbad, CA, USA). psPAX2, a virus packaging vector, and pMD2.G, an envelope protein vector, were gifts from Dr. Zang-Hee Lee (Seoul National University, Seoul, Korea). Y-27632 (Tocris Cookson, Avonmouth, UK) was used to inhibit the activity of Rho-associated kinase (ROCK). Gibco 0.25% trypsin-EDTA was obtained from Fisher Scientific (Pittsburgh, PA, USA).

The HOK-16B cell line was a gift from Dr. N. H. Park (School of Dentistry, University of California, Los Angeles, CA, USA), and comprised a line of cells immortalized from periodontally healthy human retromolar gingival tissue [15]. The HOK-16B cells were cultured in keratinocyte growth medium (KGM) supplemented with bovine pituitary extract, hydrocortisone, recombinant human epidermal growth factor, gentamicin and amphotericin-B (GA-1000), recombinant human insulin (Lonza, Basel, Switzerland), and 1% penicillin. The transfection of cells was performed as described previously [9]. Briefly, HOK-16B cells were cultured in a culture medium containing lentiviral particles generated in HEK293T cells that had been transfected with the shJNK1/2 plasmid together with pMD2.G and psPAX2, using the Lipofectamine LTX and Plus reagents.

Cells were fixed with 5% paraformaldehyde and coated with palladium after freeze-drying or drying with a graded alcohol series. Surface images of the cells cultured on various substrates were obtained by field emission scanning electron microscopy (FE-SEM) (S4700, Hitachi, Tokyo, Japan).

Immunoblotting was performed according to the standard protocol. Briefly, the cells were lysed with a lysis buffer (150 mM NaCl, 1% deoxycholate, 20 mM Tris-HCl [pH 7.5], 1 mM EDTA, 1% Triton X-100, 1 mM EGTA, 2.5 mM sodium pyrophosphate, 1 mM glycerophosphate) containing a protease inhibitor mixture comprising 1 mM Na₃VO₄, 10 mM NaF, and 1 mM PMSF protease inhibitor (Boehringer Mannheim, Indianapolis, IL, USA), 1 µg/mL of leupeptin, and 1 µg/mL of aprotinin phosphatase inhibitors (Calbiochem, La Jolla, CA, USA). Cell lysates boiled in sample buffer were size-separated through sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. The membranes were blocked with 5% skim milk and incubated with primary antibodies overnight at 4°C. Then, the membranes were incubated with secondary antibodies in 5% skim milk for 1 hour at room temperature. The blots were developed using an enhanced chemiluminescent HRP substrate (Invitrogen). Images of the immunoblotting were digitized to compare the strength of the expression levels of targeted molecules using Image J (National Institutes of Health [NIH], Bethesda, MD, USA). The quantified relative strengths of the expression levels of the target molecules were obtained from the immunoblotting images using the Gels menu in Image J.

Immunocytochemistry was performed according to the standard protocol. Briefly, cells were fixed with 4% v/v paraformaldehyde and permeabilized in 0.5% v/v Triton X-100. The cells were incubated with a primary antibody against E-cadherin for 2 hours. Then, cells were treated with Cy3-conjugated anti-rabbit IgG antibody. F-actin was stained with FITC-phalloidin. The nucleus was stained with DAPI. Images of the stained cells were acquired using an LSM 700 confocal laser-scanning microscope (CLSM) (Zeiss, Oberkochen, Germany). In addition, images of the staining for E-cadherin and F-actin were merged to examine the co-localization of E-cadherin and F-actin.

Statistical analyses were performed in MedCalc version 17.2 for Windows (MedCalc Software, Ostend, Belgium). Since the quantitative data collected in the present study did not show a normal distribution by the Kolmogorov-Smirnov test for normality, the nonparametric Kruskal-Wallis test, followed by post hoc multiple comparison with the Mann-Whitney test at a Bonferroni-adjusted alpha value, was used for the comparison of multiple groups according to the guidelines presented in the MedCalc software (MedCalc Software).

HOK-16B cells, which are HGKs, were examined by FE-SEM to compare the spreading of the cells seeded on substrates with varying levels of roughness. Intriguingly, the surfaces of the cells initially attached to the substrates were rich in membrane folding and furrows of various dimensions, regardless of the type of substrate to which the cells were attached (Figure 2A). These characteristic folds and furrows became gradually unfolded at the bottom of the cell, where the cell membrane connected with the substrate (Figure 2B). The hidden membrane located deep between the furrows appeared to unfold in order to supply the cells with additional cell membrane, which was required for the cells to spread on the substrate.

The initial spreading of the cells differed between the cells cultured on rough substrates and those cultured on smooth substrates. A few cytoplasmic processes extended from the bodies of the cells on the rough substrates, whereas circumferential lamellipodia extended from the onset of cell spreading on the smooth substrates (Figure 2C and D). Fewer cytoplasmic processes extended as the substrate roughness increased (Figure 3). The cell processes of the HGKs on the rough substrates gradually developed into lamellipodia of varying widths depending on the roughness of the substrate (Figures 3 and 4). Finally, the cells developed circumferential lamellipodia after the provisional stage of developing cell processes on the rough substrate with a low-nanometer dimension (Ra=121.3±13.4 nm) (Figure 3). In contrast, the cells maintained narrow lamellipodia throughout the whole stage on substrates with a greater roughness (Ra=505.3±115.3, 867.0±168.6 nm) (Figures 3 and 4).

Scattered individual HGKs aggregated and developed intercellular junctions through contacts between the ends of lamellipodia of various sizes. Cell spreading characteristics were reflected in the development of the intercellular junctions. The cells surrounded by circumferential lamellipodia made relatively long and tight cell-cell junctions (CCJs) through contacts between the wide surfaces of circumferential lamellipodia on the smooth substrate (Figures 3 and 4). The HGKs rapidly aggregated to form compact islands on the smooth substrates. The HGKs that developed circumferential lamellipodia on the rough substrate with low-nanometer dimensions also formed long intercellular junctions, as observed in the cells on the smooth substrates (Figures 3 and 4). However, the HGKs required more time to form compact islands on the rough substrate with low-nanometer dimensions than on the smooth substrate because these cells had an extra-provisional stage in which cell processes were extended before the development of the circumferential lamellipodia (Figure 3). In contrast, the cells extending narrow lamellipodia on the rough substrates with mid- or high-nanometer dimensions (Ra=505.3±115.3, 867.0±168.6 nm, respectively) developed defective intercellular junctions (Figures 3and 4). The development of the intercellular junctions did not appear to strongly promote the formation of compact aggregates by the cells. Therefore, the cells on the rough substrates with mid- or high-nanometer dimensions did not establish compact aggregates, and wide intercellular gaps between the cells remained (Figures 3 and 4).

In HGKs, the intercellular junctions were ECJs; E-cadherin could be strongly detected by CLSM (Figures 3 and 4). An examination of the cell clusters by CLSM confirmed the development of the defective ECJs in cells cultured on substrates with high levels of roughness. The cells cultured on the smooth substrate or the rough substrate with low-nanometer dimensions rapidly formed compact islands that were surrounded by tight ECJs (except for the free membranes of the cell cluster); in contrast, the clusters of cells cultured on the rougher substrates were slow to develop ECJs. A larger percentage of the cell membranes remained without ECJs in cells cultured on the rougher substrates when examined at the early stage of the cell culture (Table 2). The area that was not covered by cells increased with increasing substrate roughness. After a longer culture time of up to 4 days, the cells on the smooth substrate and the rough substrate with low-nanometer dimensions became more compact (Figure 4). However, the cells on the highly rough substrates remained loosely aggregated. These results indicate that cells cultured on moderately rough substrates overcame the roughness when cultured for longer periods of time, but the cells on highly rough substrates were not able to overcome this roughness, even after a long culture period.

Next, we examined whether the defective development of the ECJs in cells cultured on the rough substrates with mid- or high-nanometer dimensions resulted from a reduced chance of cell contacts forming due to the wider spaces between the cells, which were less spread out (Figure 5). Compared with ECJ development in the cells cultured on the smooth substrate, ECJ development in the cells cultured on the substrate with a high roughness was defective even though the cells were cultured at a high density, permitting unconstrained cell contacts. These results confirm that on the rough substrate, the cell membranes of HGKs were less adhesive than those of cells on the smooth substrate.

Next, we examined whether there was a difference in the stability of the ECJs in cells cultured on the smooth substrate and those cultured on rough substrates. Compact cell aggregates formed in cells cultured for 4 days on the smooth and moderately rough substrates and were treated with EDTA or trypsin-EDTA, which are known to activate RhoA-ROCK and to dissociate intercellular ECJs [16]. These treatments disrupted the intercellular junctions of the cells cultured on the rough substrates earlier than in the cells cultured on the smooth substrate (Figure 6). These results indicate that intercellular ECJs were more vulnerable in HGKs cultured on rough substrates than in those cultured on smooth substrates, although the cells assembled compact cell aggregates on both substrates.

In a previous report, we showed that JNK downregulated ECJ development in HGKs [9]. Thus, we examined whether defective ECJ development in the cells cultured on the rough substrates was associated with JNK activity. The expression level of p-JNK was assessed by immunoblotting for p-JNK in cells cultured on substrates with varying levels of roughness. The expression levels of p-JNK were higher in the cells cultured on the rough substrates than in the cells cultured on the smooth substrate, although the difference in JNK activity decreased with time after cell seeding, excluding those on the rough substrates with high-nanometer dimensions (Ra=867.0±168.6 nm) (Figure 7A). The finding of reduced JNK activity after the initial culture phase suggests that JNK primarily influenced the initial stage of ECJ development.

Next, the effect of JNK inhibition on ECJ development was examined. The treatment of the cells with SP600125, a pharmacological inhibitor of JNK, clearly promoted ECJ development in the HGKs cultured on the rough substrate with low-nanometer dimensions and on the smooth substrate (Figure 7B). The knockdown of JNK1 or JNK2 via lentiviral transfection of shRNA JNK1 or shRNA JNK2 into the cells also promoted ECJ development in HGKs cultured on the rough substrate with low-nanometer dimensions (Figure 7C). In addition, involvement of JNK in the regulation of ECJ development was further confirmed by showing that anisomycin, a pharmacological activator of JNK, disrupted the ECJs of the cell aggregates (Figure 7D). These results suggest that the delay in ECJ development may have been associated with JNK activity in the cells cultured on the rough substrate. However, the responsiveness of ECJ development to JNK regulation was observed in HGKs cultured on the rough substrate with a low-nanometer dimension, but not in HGKs cultured on substrates with higher levels of roughness. There was no acceleration of ECJ development in response to JNK inhibition in the cells grown on substrates with higher levels of roughness (Figure 7B). These results indicate that the physical barrier of the very rough substrates could not be overcome by the biological regulation of JNK activity when the cells were scattered at a low density. However, ECJ development in response to JNK inhibition in the cells grown on the substrates with greater levels of roughness was clearly upregulated when the cells were cultured at a high density to reduce the distance between cells (Figure 5). These results suggest that JNK inhibition can be used to promote ECJ development when oral keratinocytes are packed into the JE on rough surfaces with high-nanometer dimensions.

Next, the association between the activity of JNK and regulatory F-actins, such as cortical actin, in ECJ development was examined to investigate the downstream effects of JNK in the regulation of ECJ development, which are largely unclear. Due to its stiffness, plasma membranes lined with cortical actin have been hypothesized to suppress the development of intercellular junctions [171819]. The F-actin of HGKs was stained with phalloidin to determine whether cortical actin was associated with the development of defective ECJs in the cells grown on the rough substrates. On the smooth substrates, as soon as the cells began to spread, they rapidly extended circumferential lamellipodia devoid of cortical actin (Figure 4). These circumferential lamellipodia were rich in filopodia instead of cortical actin. In contrast, in the HGKs grown on rough substrates, cortical actin without filopodia was prominent in the smooth plasma membrane, except at the ends of the cell processes at the onset of cell spreading (Figure 4). In HGKs on the rough substrate with low-nanometer dimensions, cortical actin disappeared from the circumferential lamellipodia, which were slow to develop (Figure 3). However, most of the cortical actin shell remained at the plasma membranes of the HGKs cultured on the substrates with higher levels of roughness, except at the ends of the cell processes, where small lamellipodia-like cytoplasmic extensions developed. ECJs developed only at the margins of the cells in which cortical actin was absent (Figure 3). The thickness and proportion of the plasma membrane that was lined with cortical actin increased as substrate roughness increased. Thus, the cells grown on rougher substrates appeared to require longer periods of time to contact the adhesive membranes of the other cells. These observations confirm that the primary requirement for ECJ development in HGKs is the disassembly of the stiff cortical actin shell beneath the plasma membrane to increase plasma membrane pliability [171819].

The cells in which JNK activity was inhibited by treatment with SP600125 or the transfection of shJNK1/2 did not develop cortical actin during ECJ development, regardless of the roughness of the substrates (Figure 8). In contrast, the cells in which JNK activity was upregulated by anisomycin developed thick cortical actin (Figure 8). The cells in which JNK was constitutively activated by anisomycin did not spread even on the smooth substrates. These cells maintained the spherical morphology of the cells suspended on the smooth substrates, which suggested that their cortical tension was extremely high. Interestingly, the effect of anisomycin, which induced the formation of thick cortical actin, was reversed by inhibiting actomyosin contraction via ROCK inhibition (Figure 8). HGKs co-treated with anisomycin and Y-27632 developed ECJs. Together, these results indicate that the cortical actin shell regulated ECJ development in HGKs, and that JNK activity was associated with the development of this stiff F-actin structure.