Connexon-mediated cell adhesion drives microtissue self-assembly

Abstract

Microtissue self-assembly is thought to be driven primarily by cadherins, while connexons have been examined mainly in intercellular coupling. We investigated whether connexon 43 (Cx43)-mediated cell adhesion modulates self-assembly of human KGN granulosa cells, normal human fibroblasts (NHFs), and MCF-7 breast cancer cells seeded into nonadhesive agarose gels. We found that treatment with anti-Cx43 E2 (112 μg/ml), which suppresses Cx43 docking, significantly inhibited the kinetics of KGN and NHF self-assembly compared to the preimmune sera control (41.1±4.5 and 24.5±10.4% at 8 h, respectively). Likewise, gap junction inhibitor carbenoxolone also inhibited self-assembly of KGN, NHF, and MCF-7 cells in a dose-dependent manner that was specific to cell type. In contrast, Gap26 connexin mimetic peptide, which inhibits channel permeability but not docking, accelerated self-assembly of KGN and NHF microtissues. Experiments using selective enzymatic digestion of cell adhesion molecules and neutralizing N-cadherin antibodies further showed that self-assembly was comparably disrupted by inhibiting connexin- and cadherin-mediated adhesion. These findings demonstrate that connexon-mediated cell adhesion and intercellular communication differentially influence microtissue self-assembly, and that their contributions are comparable to those of cadherins.—Bao, B., Jiang, J., Yanase, T., Nishi, Y., Morgan, J. R. Connexon-mediated cell adhesion drives microtissue self-assembly.

Successful tissue formation requires the coordinated aggregation of individual cells such that these cell-cell interactions become, at least temporarily, more energetically favorable than cell-matrix interactions. This is achieved through expression of surface adhesion molecules, such as cadherins, and their connections to the internal cytoskeletal network. A useful model to closely examine this cellular coalescence is that of self-assembly, whereby monodispersed cells aggregate into 3-dimensional (3-D) microtissues within a scaffold-free environment. Active self-assembly is thought to mimic embryogenesis and morphogenesis (1, 2), while fully assembled microtissues provide an opportunity to understand the mechanisms involved in steady-state tissue architecture and function (3).

Connexins are the constituent proteins of gap junction channels. Six connexins form a hexamer with a central pore termed a connexon. Adjacent cells become electrically and chemically coupled when apposed connexons dock, facilitating the passive diffusion of ions and small molecules up to 1 kDa through the adjoined pores (4). Docking requires rotation of one connexon 30° to facilitate interdigitation (5) of the apposed extracellular domains, forming a dual concentric β-barrel configuration (6). Since gap junction intercellular communication is involved in many important cellular processes, such as homeostasis (7) and resistance to tumorigenesis (8), docking is seen primarily as a prerequisite to channel formation. However, recent research has begun to examine connexins as cell adhesion molecules (CAMs) independent of their coupling properties. Transfection with Cx43 has been shown to enhance cell migration (9), a process that involves constant initiation of cell-cell contact and adhesion. Similarly, short-term rotary shaker assays show that Cx43 facilitates aggregation of cells into small clusters (10). In this study, we investigate the role of connexins in cellular self-assembly, another cell adhesion-driven process.

Although cadherins are perceived as the principal driving force of self-assembly (11), we recently demonstrated that other cellular components are actively involved. We showed that cytoskeletal-mediated tensile forces can affect the self-assembly kinetics of H35-hepatoma cells and normal human fibroblasts (NHFs) (12). In this study, we show that Cx43-mediated cell adhesion modulates the self-assembly of KGN granulosa cells, NHFs, and MCF-7 breast cancer cells. In all three cell types, Cx43 is the predominant connexin subtype (13–15). We found that carbenoxolone, which disassembles plaques (16), and an antibody to Cx43, which sterically inhibits connexon docking (17), slowed the kinetics of microtissue compaction. Interestingly, treatment with the peptide Gap26, which inhibits channel communication but not docking (18), accelerated self-assembly. Selective enzymatic digestion of CAMs and treatment with an N-cadherin neutralizing antibody showed that self-assembly was comparably constrained by inhibition of connexin and cadherin-mediated cell adhesion. These observations indicate that both connexon-mediated cell adhesion and coupling influence microtissue self-assembly, that their contributions are nonequivalent, and comparable to the well-established role of cadherins.

MATERIALS AND METHODS

Design and fabrication of micromolded agarose gels

Micromolded hydrogels were produced as described previously (19). Briefly, wax molds were designed using computer-assisted design (Solid Works Corporation, Concord, MA, USA) and produced with a Thermojet rapid prototyping machine (3D Systems Corp., Valencia, CA, USA). Reprorubber (Flexbar, Islandia, NY, USA) negatives were then produced by pouring a 1:1 ratio of catalyst to base into the wax molds in a 6-cm Petri dish (Corning, Corning, NY, USA) and allowed to cure at room temperature for 20 min. Polydimethylsiloxane (PDMS; Dow-Corning, Midland, MI, USA) replicates of Reprorubber negatives sprayed with epoxy Parafilm release agent (Flexbar) were then made using 90% PDMS and 10% accelerator, followed by brief degassing. PDMS replicates were allowed to cure at 95°C for ≥1 h before separation from the Reprorubber. PDMS replicates were rinsed with 70% ethanol and dH₂O and sterilized by autoclaving. A 3% molten agarose solution was prepared by heating autoclaved Ultrapure agarose (Invitrogen, Carlsbad, CA, USA) powder and sterile water until dissolution. Molten agarose solution (1.0 ml) was pipetted into each PDMS replicate. After setting for 10 min, the agarose micromolds were separated from the PDMS replicates using a spatula, transferred to a 24-well tissue culture plate, and equilibrated overnight in the appropriate culture medium.

To form spheroid or rod microtissues, agarose micromolds with a rectangular seeding chamber (7×5 mm) containing a staggered array of 95 cylindrical, round-bottomed recesses (400 μm diameter, 800 μm depth), or 14 trough-shaped, round-bottomed recesses (1200 μm length, 400 μm width, 800 μm depth) were used, respectively.

Cell culture and production of self-assembled microtissues

KGN granulosa cells (20) and NHFs derived from neonatal foreskins were expanded in DMEM (Invitrogen). MCF-7 human breast cancer cells (NCI-60, National Cancer Institute, Frederick, MD, USA) were expanded in Roswell Park Memorial Institute Medium (RPMI; Invitrogen). Both DMEM and RPMI media were supplemented with 10% FBS and 1% penicillin/streptomycin. KGN and MCF-7s were cultured in a 37°C incubator with 5% CO₂ atmosphere, while NHFs were cultured with 10% CO₂ atmosphere. Self-assembled microtissues were produced, as described previously (19). Unless otherwise stated, cells were trypsinized, counted, and resuspended to the desired cell density. Agarose gels were then seeded with monodispersed cells by pipetting 100 μl cell suspension into the seeding chamber. Samples were incubated for 15 min to allow cells to settle before 1 ml of the appropriate culture medium was added.

Microtissue viability and Cx43 immunostaining

Microtissue viability was determined using the Live/Dead Viability/Cytotoxicity Kit (Invitrogen), as previously reported (21). For Cx43 immunostaining, a primary antibody directed against the C terminus of Cx43 (Sigma-Aldrich, St. Louis, MO, USA) developed in rabbit was used. Corresponding goat anti-rabbit Alexa 488-conjugated secondary antibodies were used (Sigma). Cx43 immunostaining was performed with modifications to a previous protocol (22). Microtissues seeded 24 h earlier were washed in PBS, fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned (5 μm thickness). Sections were deparaffinized in xylene, rehydrated in graded alcohols, and washed in PBS. Heat-induced antigen retrieval was performed using a citrate buffer (Invitrogen), according to the manufacturer's protocol. Samples were blocked for 1 h in PBS containing 3% BSA and 0.1% Triton X-100 at room temperature. Microtissues were then incubated with primary anti-Cx43 antibody, diluted at 1:200 in PBS with 1% BSA, overnight at 4°C. After being washed with PBS, samples were incubated with Alexa-488 conjugated secondary antibody, diluted at 1:200 in PBS with 1% BSA for 1 h at room temperature. DAPI (Invitrogen), diluted to 300 nM in PBS, was used as a nuclear stain. Sections were mounted with the ProLong Antifade Kit (Invitrogen) and imaged with a Carl Zeiss LSM 510 Meta Confocal Laser Scanning Microscope (Carl Zeiss MicroImaging, Thornwood, NY, USA).

Dye transfer functional coupling assay

To examine intercellular coupling via gap junction formation, we modified an existing assay (23). Cells were trypsinized, counted, and resuspended in serum-free DMEM at 1 × 10⁶ cells/ml. Half of the cells were double-labeled with 5 μl/ml of 10 μM DiIC₁₈ (dialkyl carbocyanine; Invitrogen) and 0.5 μl/ml of 0.5 μM calcein-AM (Invitrogen), covered, and allowed to incubate at room temperature for 30 min. After incubation, the cell suspension was washed 3 times in serum-free DMEM and resuspended again to 1 × 10⁶ cells/ml. To obtain the desired ratio of labeled:unlabeled cells, the appropriate volumes of each were mixed, centrifuged, and resuspended, and seeded into agarose gels at 0.3 × 10⁶ cells/gel. Samples were covered and incubated at 37°C for 30 min before 1 ml of DMEM with 1% penicillin/streptomycin was added. Samples were imaged at 1-h intervals over 24 h, while maintained at 37°C and 5% CO₂ atmosphere.

Microscopy and image analysis

Phase contrast and epifluorescent images were obtained using an Axio Observer Z1 equipped with an AxioCam MRm camera (Zeiss) and Xcite 120 XL mercury lamp (Exfo Life Sciences Division, Mississauga, ON, Canada). Time-lapse images were acquired at 10-min intervals over 8 to 24 h. A humidified chamber with temperature (37°C) and CO₂ control (5%) was used (Zeiss). Quantification of rod microtissue length was performed using ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA). Using the line tool, rod length was defined as the length from end to end of the microtissue (long-axis length). Percentage inhibition by drug/antibody treatment was defined as the percentage difference between drug-treated and control rod length, at the indicated postseeding time. Viability of microtissues stained with Live/Dead stain was quantified using MetaMorph (Molecular Devices, Downington, PA, USA). The average intensities of green and red pixels were measured in the projected microtissue area, corresponding to the live and dead signal, respectively. Live/dead ratios were calculated from the signal intensities and normalized to live/dead ratio of the serum-free control.

Selective protease digestion of CAMs

Different classes of CAMs were enzymatically removed using a previously published method (24, 25). Cell monolayers were incubated in Ca²⁺/Mg²⁺-free Hank's balanced salt solution (HBSS) containing 1 mM EDTA for 15 min at 37°C. Cells were then mechanically dissociated using a cell scraper (BD Biosciences, Bedford, MA, USA). After 2 washes with Ca²⁺-free DMEM to remove residual EDTA, dissociated cells were counted and incubated for 15 min in one of the following conditions: condition A, Ca²⁺-free HBSS containing 1 mM EDTA (positive control, all CAMs are intact); condition B, 0.0001% trypsin in Ca²⁺-free HBSS containing 1 mM EDTA [calcium-independent CAMs (CIDs) only]; condition C, 0.01% trypsin in HBSS containing 1 mM Ca²⁺ [calcium-dependent CAMs (CADs) only]; condition D, 0.01% trypsin in Ca²⁺-free HBSS containing 1 mM EDTA (negative control, no CAMs intact). Cells were washed twice in Ca²⁺-free DMEM containing 10% FBS to stop all enzymatic reactions, counted, resuspended in either Ca²⁺-containing medium (conditions A and C) or Ca²⁺-free medium (conditions B and D), and seeded into gels.

Gap junction-inhibiting drugs and neutralizing antibodies

Carbenoxolone (Sigma) working solutions were prepared by diluting appropriate volumes of a 10 mM stock solution into serum-free medium. 1-Heptanol (Sigma) working solutions were prepared by diluting appropriate volumes of a 2 mM stock solution into serum-free medium. Anti-Cx43 E2, a rabbit polyclonal Ab against the second extracellular loop of Cx43 (E2, amino acid residues 186–206) was used to inhibit Cx43 docking interactions and was produced as described previously (26). Mouse mAb against N-cadherin (A-CAM clone GC-4, Sigma) was used to neutralize N-cadherin bond formation. Appropriate controls for the neutralizing Cx43 and N-cadherin antibodies were purchased from Sigma (rabbit preimmune sera and mouse IgG1 isotype control, respectively). Gap26 connexin mimetic peptide, possessing sequence homology with the Gap26 domain (amino acid residues 204–214) and a scrambled peptide control were purchased from Alpha Diagnostic International (San Antonio, TX, USA). For the drug, neutralizing antibody, and peptide experiments, gels were preequilibrated overnight in medium containing the appropriate agent. To avoid digestion of CAMs by trypsinization, monolayer cells were incubated in HBSS containing 1 mM EDTA for 15 min before being mechanically dissociated with a cell scraper. Cells were then resuspended in medium containing one of the three agents and seeded into gels.

Western blot analysis

For Western blots, cells were lysed in Nonidet P-40 lysis buffer (Invitrogen) prior to protein quantification using a BCA protein assay (Thermo Scientific, Rockford, IL, USA). Equal amounts of protein were loaded (30 μg/lane) and separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. Membranes were immunoblotted with the indicated primary antibodies: Cx43 (Sigma), Pan-cadherin (Abcam, Cambridge, MA, USA), and N-cadherin (Abcam). Membranes were then immunoblotted with appropriate secondary antibodies: goat anti-rabbit (Cell Signaling Technology, Danvers, MA, USA), goat anti-mouse (Invitrogen), and goat anti-rabbit (Cell Signaling Technology), respectively. Specific antibody binding was determined using the Pierce ECL Western blotting detection system (Thermo Scientific). To assure equal gel loading, nitrocellulose membranes were stripped in Restore stripping buffer (Thermo Scientific) and reprobed for actin using a rabbit polyclonal anti-actin antibody(Sigma).

RESULTS

Micromolded agarose gels with either circular or trough-shaped recesses were used to investigate the self-assembly of cells into spheroid and rod microtissues in a nonadhesive scaffold-free environment, respectively (Fig. 1). Monodispersed KGN, NHF, and MCF-7 cells were seeded onto micromolded gels and imaged over 24 h. KGN cells formed stable spheroids (Fig. 1B) and rods (Fig. 1C) by 24 h. NHF and MCF-7 cells also formed stable spheroids and rods by 24 h (data not shown), although self-assembly kinetics and 24 h microtissue dimensions varied with cell type and seeding density.

Figure 1.

Production of self-assembled microtissues. A) Monodispersed KGN cells were seeded into micromolded gels with circular or trough-shaped recesses and allowed to self-assemble. B, C) Twenty-four-hour time-lapse phase-contrast images of KGN spheroid (B) and rod (C) microtissue formation. Scale bars = 100 μm (B); 400 μm (C).

To determine whether cells within 3-D microtissues formed Cx43 plaques at sites of cell-cell contact, spheroid microtissues were fixed, sectioned, and immunostained for Cx43 (Fig. 2). Confocal images show punctate areas of fluorescence representing Cx43 plaques on the plasma membranes of cells within KGN and NHF spheroids 24 h postseeding. Perinuclear staining was also observed, similar to that seen by others (27). Western blot densitometry analysis showed that Cx43 expression is up-regulated approximately two-fold in KGN and NHF spheroid microtissues compared to adherent cells (Supplemental Fig. 1).

Figure 2.

Microtissues express Cx43 at sites of cell-cell contact and perinuclear regions. Cx43 immunostaining and confocal microscopy of KGN and NHF spheroids show positive Cx43 expression at sites of cell-cell contact (plaques) and perinuclear regions. Cx43 plaque expression appears higher in the microtissue interior compared to the surface. H/E and DAPI staining are provided as reference for microtissue morphology and nuclei, respectively. Scale bars = 50 μm.

To determine whether cells within microtissues were coupled via functional gap junction channels, a modified dye transfer assay was used (Fig. 3). Briefly, NHF cells were double-labeled with calcein AM, a cytosolic fluorophore, and DiI, a lipophilic fluorophore, before they were mixed with unlabeled NHF cells (5% labeled, 95% unlabeled) and seeded into gels. Intercellular coupling facilitates the transfer of only calcein AM to surrounding cells. Epifluorescent images taken immediately after seeding (Fig. 3, row 1) show that both DiI and calcein AM signals were punctate and colocalized, confirming that coupling required the intercellular contact that drives cell aggregation. Time-lapse images show that by 3 h postseeding, the calcein AM signal became larger and radially diffuse, while the DiI signal remained punctate and localized to the original cells (Fig. 3, row 2), indicating that coupling occurs parallel to self-assembly. As a control, the double-labeled/unlabeled NHF mixture was seeded into gels preequilibrated with 200 μM carbenoxolone and showed no evidence of coupling at 3 h (Fig. 3, row 3).

Figure 3.

Microtissues demonstrate gap junction coupling by 3 h postseeding. Five percent NHF cells double-labeled with DiI and calcein AM were mixed with 95% unlabeled NHF cells and seeded into spheroid gels. Epifluorescent images at 0 h postseeding demonstrate colocalization of DiI and calcein AM, showing that coupling has not occurred yet. At 3 h, significant radial diffusion of calcein AM, but not DiI, indicates that coupling via gap junctions has occurred. Control microtissues composed of double-labeled/unlabeled NHF cells seeded into 200 μM carbenoxolone showed no evidence of coupling at 3 h. Scale bar = 100 μm

To evaluate the role of connexons in self-assembly, microtissues were treated with carbenoxolone and 1-heptanol, two well-known gap junction inhibitors. Cells were resuspended in medium containing carbenoxolone or 1-heptanol, and then seeded into gels preequilibrated with the appropriate drug. To quantify the kinetics of self-assembly, rod microtissue lengths were measured over 12 h and normalized to the initial rod length. Carbenoxolone exhibited dose-dependent inhibition of rod microtissue contraction for all three cell types (Fig. 4A, only KGN data are shown). Higher concentrations (200 μM) showed both an earlier onset and greater percentage inhibition of KGN self-assembly at 12 h (36.2±2.8%) compared to 50 μM (24.2±9.8%). Pretreatment with 50 μM carbenoxolone for 5 h prior to seeding increased the 12 h inhibition (30.5±3.0%), making it comparable to the nonpretreated 200-μM sample. Likewise, 1-heptanol inhibited KGN self-assembly in a similar dose-dependent manner (Fig. 4B). Microtissue viability was evaluated at 12 h to address potential toxicity due to the nonspecific effects observed with carbenoxolone and 1-heptanol at high concentrations (Fig. 4C). Live/dead analysis confirmed that the viability of drug-treated microtissues remained comparable to the untreated control over the range of concentrations used. To determine whether inhibition of self-assembly by carbenoxolone varied with cell type and time postseeding, the percentage inhibition was determined for KGN, NHF, and MCF-7 cells at 5, 10, and 15 h postseeding (Fig. 4D). Of the different cell types, MCF-7 cells were the most sensitive to carbenoxolone, whereas NHF cells were the least sensitive, requiring considerably higher concentrations (400 μM) to show comparable inhibition to the 50 μM-treated MCF-7 cells. KGN cells were mixed with HeLa cells and seeded into spheroid gels to determine whether connexin-expressing cells (KGN cells) segregate from connexin-negative cells (HeLa cells) (Fig. 4E). The cells are clearly segregated by 16 h postseeding, with KGN cells sorting to the center and HeLa cells sorting to the periphery. Treatment with carbenoxolone (50 μM) disrupted sorting and led to increased mixing. Control experiments confirmed that carbenoxolone does not affect HeLa self-assembly or compromise the mixed KGN/HeLa microtissue viability (Supplemental Fig. 2).

Figure 4.

Gap junction inhibitors carbenoxolone (CBX) and 1-heptanol inhibit microtissue self-assembly and self-sorting. A) Carbenoxolone inhibited KGN rod self-assembly in a dose-dependent manner, with increased inhibition on pretreatment. B) 1-Heptanol also inhibited self-assembly in a dose-dependent manner. C) Live/dead analysis confirms that microtissue viability was not compromised, n = 35. D) Percentage inhibition of rod contraction by carbenoxolone varied with postseeding time and cell type; n = 18 (KGN), 11 (MCF-7), and 19 (NHF). KGN cells labeled with CellTracker Green (CMFDA) were mixed with HeLa cells labeled with CellTracker Blue (CMAC) at a 1:1 ratio and seeded into spheroid gels. E) Cells were clearly segregated by 16 h postseeding, with the connexin-expressing cells (KGN) sorting to the center and the connexin-negative cells (HeLa) sorting to the periphery. Treatment with carbenoxolone (50 μM) disrupted sorting, with an increased degree of mixing. EF, epifluorescence; CF, confocal. Scale bar = 200 μm.

To compare the relative contributions of CIDs and CADs in self-assembly, KGN, NHF, and MCF-7 cells were exposed to enzyme solutions designed to selectively remove CADs and/or CIDs prior to seeding (Fig. 5A–C). For all cell types, self-assembly was most rapid in the control (condition A) where all CAMs were left intact. KGN and MCF-7 rod self-assembly were comparably inhibited by digestion of CADs (condition B) and CIDs (condition C). For NHFs, inhibition by selective CID digestion was minimal compared to the effect of removing CADs. In all cell types, protease digestion of both CADs and CIDs with no calcium added (condition D) led to the complete cessation of self-assembly. Western blot analysis (Fig. 5D) confirmed high Cx43, N-cadherin, and total cadherin expression in KGN and NHFs, the two rapidly aggregating cell types. MCF-7s, which aggregate more slowly, showed low Cx43 and total cadherin expression, while N-cadherin levels were undetectable. All blots were stripped and relabeled for actin to ensure equal protein loading.

Figure 5.

Selective protease digestion of CAMs shows comparable inhibition of rod self-assembly by removal of CADs and CIDs. A–C) KGN (A) and MCF-7 (C) rod contraction were inhibited by enzymatic removal of both CADs and CIDs, while NHF rod contraction (B) was inhibited by removal of CADs only. Condition A: positive control, all CAMs intact. Condition B: CIDs only. Condition C: CADs only. Condition D: negative control, no CAMs intact. D) Western blot analysis confirmed high Cx43, N-cadherin, and total cadherin expression in KGN and NHF cells, the two rapidly aggregating cell types. Slower aggregating MCF-7 cells showed minimal Cx43 and total cadherin levels, with undetectable N-cadherin expression. Blots were immunolabeled for actin to ensure equal loading.

To determine whether connexon-mediated cell adhesion modulates self-assembly, microtissues were treated with anti-Cx43 E2, an antibody that sterically inhibits Cx43 docking. Anti-Cx43 E2 inhibited the self-assembly of KGNs and NHFs, the two cell types with high Cx43 expression, within 2 h postseeding compared to the preimmune sera control (P<0.05, n=16; Fig. 6A, B). Anti-Cx43 E2 exhibited dose-dependent inhibition of KGN rod contraction. Higher concentrations (112 vs. 56 μg/ml) resulted in greater percentage inhibition at 8 h (41.1±4.5 vs. 24.8±9.3%, respectively). NHF self-assembly was unaffected by increasing the concentration, showing comparable 8 h percentage inhibition for 56 and 112 μg/ml (19.8±9.0 and 24.5±10.4%, respectively). To evaluate the role of connexon-mediated intercellular communication in self-assembly, microtissues were treated with Gap26, a small Cx43 mimetic peptide that inhibits channel permeability but not Cx43 docking (18). In contrast to anti-Cx43 E2, Gap26 (300 μM) accelerated the rate of KGN and NHF rod contraction compared to the scrambled peptide control (Fig. 6C, D). To compare the roles of Cx43 and N-cadherin-mediated cell adhesion in self-assembly, microtissues were treated with anti-NCad, a neutralizing antibody that inhibits homotypic N-cadherin bond formation. Anti-NCad inhibited the self-assembly of KGNs and NHFs, cell types shown to express N-cadherin. Unlike anti-Cx43 E2, which showed greater inhibition of KGNs, anti-NCad demonstrated a greater percent inhibition of NHFs than KGNs at 8 h (32.6±18.9% and 10.4±5.7%, respectively). Interestingly, simultaneous treatment with anti-Cx43 E2 and anti-NCad resulted in significantly greater inhibition of KGN self-assembly at 8 h (38.1±3.4%, n=10, P<0.05) than with either antibody alone (Supplemental Fig. 3).

Figure 6.

Connexon-mediated adhesion and intercellular communication differentially modulate self-assembly of high-Cx43-expressing cell types. A, B) KGN (A) and NHF (B) rod self-assembly is significantly inhibited by anti-Cx43 (56 μg/ml) compared to preimmune serum control; n = 16, P < 0.05. C, D) Increased anti-Cx43 E2 concentration (112 μg/ml) further inhibited self-assembly of KGN cells (C) but not NHF cells (D). In contrast, Gap26 connexin mimetic peptide (300 μM) accelerated self-assembly of KGN (C) and NHF cells (D); n = 10. E, F). Inhibition by anti-Cx43 E2 is comparable to that seen with neutralizing anti-N-cadherin (40 μg/ml); n = 10.

DISCUSSION

Cellular self-assembly, the aggregation of monodispersed cells into 3-D microtissues that is driven by cell-cell contact and intercellular adhesion, has been thought to be mediated primarily by cadherins. In this paper, we show that connexons also play a significant role. In conjunction with our previous findings and those of others, it is becoming increasingly clear that self-assembly is a complex integrative process involving multiple components, beyond cadherins.

Several lines of evidence have established the role of cadherins in cell aggregation. During embryogenesis, cadherins facilitate the cellular compaction of loosely adherent blastomeres into a blastocyst (28). Similarly, cells expressing higher levels of cadherins have been shown to form aggregates with higher surface tension (29). Recent work by our laboratory and others has begun to highlight the complexity of self-assembly. While homotypic cadherin bond specificity was initially implicated as an important modulator of cell aggregation (30), others have recently shown that cells with different cadherin subtypes are still able to aggregate effectively (31). Further, surface force measurements have shown that homotypic and heterotypic cadherin bonds have similar strengths and dissociation rates (32). Work by our lab and others has identified the actin cytoskeleton as an important determinant of self-assembly (12, 33, 34, 35), suggesting that cell aggregation may involve additional components beyond classic CAMs.

In this work, several complementary experiments implicate the involvement of connexons in self-assembly. We observed that carbenoxolone, a putative gap junction inhibitor, disrupts microtissue self-assembly. Carbenoxolone dephosphorylates Cx43, in turn, promoting plaque disassembly in a time- and dose-dependent manner (16). Likewise, inhibition of rod microtissue contraction by carbenoxolone was dose dependent. Higher concentrations resulted in an earlier onset and increased magnitude of inhibition without compromising viability. Given that plaques are composed of docked connexons (36), plaque internalization by carbenoxolone could reduce the number of connexon-connexon adhesive interactions that facilitate cell aggregation. Sensitivity to carbenoxolone also varied with cell type. MCF-7 cells were the most sensitive, while NHFs were the least. This difference could be due to variation in plaque turnover, or alternatively, suggest that the role of connexons to self-assembly is specific to cell type. Another explanation may be that sensitivity to carbenoxolone is related to connexin expression. Higher Cx43-expressing cells like NHFs may have larger reservoirs of unpaired membrane connexons that can be recruited to form new plaques (37) to maintain cell adhesion on plaque disassembly. Low Cx43-expressing cells, like MCF-7 cells, may not have such a reservoir and so are more susceptible to carbenoxolone. An alternative explanation may be that the cell types vary in their metabolism and/or accumulation of carbenoxolone, possibly due to differences in the expression of drug resistance transporter proteins.

Experiments using selective CAM digestion further highlight the involvement of noncadherin cell adhesion in self-assembly. KGN, NHF, and MCF-7 cells pretreated with protease solutions that digest CADs, primarily cadherins (24), and maintained in calcium-free medium were still able to effectively self-assemble, although less so than the control. While reinforcing the role of cadherins, this observation also confirms the participation of calcium-independent adhesive processes in self-assembly. Short-term rotary assays have previously shown that connexon-mediated cell adhesion is calcium-independent (38). Interestingly, self-assembly of KGN and MCF-7 cells was comparably inhibited by digestion of CADs and CIDs, while NHFs were inhibited more so by the loss of CADs. This difference could be due to variable and/or incomplete CAM digestion among cell types. Alternatively, it confirms the notion that the relative importance of cadherin and connexon-mediated adhesion in self-assembly is specific to cell type. The nonadhesive agarose micromolds provide a scaffold-free environment that minimizes involvement of CIDs that facilitate cell-matrix adhesion, such as integrins, suggesting that other CIDs may be involved. To definitively identify connexins, high Cx43-expressing cells (KGN and NHF cells) were treated with anti-Cx43 E2. This antibody, directed against the second extracellular loop of Cx43, sterically inhibits connexon docking and subsequent channel formation. It has been shown to inhibit plaque formation and increase the ultrastructural distance between cells (17). In our study, anti-Cx43 E2 inhibited rod contraction almost immediately, suggesting that the establishment of Cx43 adhesive contacts helps drive the self-assembly of microtissues.

Another interesting finding was that connexon-mediated intercellular communication also appears to modulate self-assembly. While connexon docking is a prerequisite for channel formation and is seen primarily as such, this study suggests that their roles in self-assembly are not equivalent. In contrast to anti-Cx43 E2, Gap26 connexin mimetic peptide, which inhibits channel permeability but not docking (18, 39), accelerated KGN and NHF rod contraction. While connexon-mediated adhesion drives self-assembly in a manner similar to other CAMs, it appears that gap junction intercellular communication plays a regulatory role in microtissue formation. Both anti-Cx43 E2 and Gap26 inhibit unpaired connexon hemichannels (18, 26), which facilitates the release of ATP, glutamate, and other factors during mechanical and ischemic stress (40). It is certainly possible that unopposed hemichannels are also involved in self-assembly. However, given the opposing effects of anti-Cx43 and Gap26, their exact role cannot be deciphered from the data presented here. More experimentation using an inhibitor specific for hemichannels is required. Interestingly, the uncoupling agent 1-heptanol inhibited self-assembly, in contrast to Gap26. Unlike Gap26, which is specific for connexins and nontoxic (18), the nonspecific and potentially cytotoxic effects of 1-heptanol are well documented. Because 1-heptanol does not affect gap junctions themselves, but rather disrupts the fluidity of the membrane lipids on which they reside (41), resultant nonspecific channel blockade is frequently reported. In turn, disruption of self-assembly by 1-heptanol could be due to blockage of non-gap junction channels.

These findings complement a growing body of evidence dissociating the adhesive and communicative properties of connexins. In cell migration, a process that like self-assembly is driven by dynamic CAM and cytoskeletal interactions, connexon-mediated adhesion and coupling have been shown to be functionally distinct. In cardiac neural crest cells, Cx43-mediated cell contact and migration are maintained despite loss of coupling (9). In Cx43-transfected glioma cells, a different pattern is seen. Reduced cell migration is coincident with diminished channel communication, while Cx43 adhesive contacts are actually up-regulated (42). Most striking, cells expressing chimeric connexins capable of docking, but not channel formation, show no change in their ability to migrate (43) or adhere (10). Recent evidence indicates that channel formation following adhesion is rarer than previously imagined. While plaques composed of as few as a dozen channels have been reported (44), detectable coupling occurs only in plaques containing >400 channels (45). Further, <2% of these channels are open, suggesting that the function of connexon adhesive contacts may extend beyond coupling. This study reveals that in active microtissue self-assembly, the contributions of connexon docking and coupling are related yet functionally distinct.

The role of connexons in self-assembly appears comparable to that of cadherins. Treating KGN and NHF cells with anti-Cx43 E2 and anti-N-Cad to inhibit Cx43 and N-cadherin bond formation, respectively, reveals that the temporal profile and magnitude of inhibition by the neutralizing antibodies are comparable, at least in order of magnitude. While their contributions seem analogous, cadherin- and connexin-mediated cell adhesion are structurally and mechanistically distinct. Cadherins are connected to the actin cytoskeleton by α/β catenins, acting as foci for the transmission of cytoskeletal mediated tension (12) during cell motility and aggregation. In contrast, connexons in plaques are highly mobile, not directly attached to catenins, and can directly connect the cytoplasms of adjacent cells through channel formation (36). While homotypic cadherin bond specificity no longer appears to dictate cell aggregation, this is not the case with connexons. The compatibility, or lack thereof, of certain heterotypic connexon pairs is well known (40) and may modulate the self-assembly of cells expressing different connexin subtypes. However, the numerous lines of evidence detailing the close interrelationships between connexins, cadherins, and the actin cytoskeleton prevent the complete segregation of their functions. We found that inhibition of self-assembly was further augmented when both Cx43 and N-cadherin were blocked compared to either alone, supporting findings that antibodies against N-cadherin can alter gap junction plaque assembly, while antibodies against Cx43 can inhibit adherens junction assembly (46). Cx43 and N-Cadherin colocalize to the same focal cell contact regions (47) and can even form multimeric protein complexes (48). The C-terminus of Cx43 directly associates with the actin cytoskeleton via zonula occludens-1 (49), and actin microfilaments participate in the degradation of gap junctions (50).

Through complementary experiments using pharmacologic gap junction inhibitors, selective enzymatic CAM digestion, and neutralizing antibodies/peptides, we have highlighted the role of connexons in microtissue self-assembly. We have shown that connexon-mediated adhesion and intercellular communication differentially modulate self-assembly and that the contribution of the former is comparable to the well-established role of cadherins.