Function of the Tetraspanin CD151–α6β1 Integrin Complex during Cellular Morphogenesis

Cell Culture and CD151 Mutants and Transfectants

The NIH3T3 mouse fibroblast cell line was obtained from American Type Culture Collection (Bethesda, MD) and cultured in DMEM supplemented with 10% fetal bovine serum (FBS), penicillin, and streptomycin. CD151 mutants were generated by recombinant PCR. As a template we used wild-type human CD151 cDNA, with an HA tag linked to its C-terminus, ligated into the eukaryotic expression vector pZeoSV (Invitrogen, San Diego, CA). As shown in Figure 5, the CD151 mutants generated in this study are as follows: 1) CD151-n-A15 (N-terminal cytoplasmic domain MGEFNEKKTTCGTVCLKYLLFTY of CD151 replaced by the corresponding METKPVITCLKTLLIIYS from A15); 2) CD151-c-A15 (C-terminal cytoplasmic domain SLKLEHY of CD151 replaced with FITANQYEMV from A15); 3) CD151-nc-A15 (both N- and C-termini of CD151 replaced by corresponding domains from A15); 4) CD151-c2-NAG2 (TM4 and C-terminal tail HLRVIGAVGIGIACVQVFGMIFTCCLYRSLKLEHY of CD151 replaced by corresponding regions NLLAVGIFGLCTALVQILGLTFAMTMYCQVVKADTYCA from TM4SF protein NAG-2); and 5) CD151-Δc(GFP) (C-terminal cytoplasmic tail of CD151 LYRSLKLEHY replaced by green fluorescent protein [GFP] moiety).

For stable expression of CD151 mutants, plasmid DNAs were transfected into NIH3T3 cells using Lipofectamine (Life Technologies, Bethesda, MD). After 48 h, cells were then cultured in media containing Zeocin (200 μg/ml; Invitrogen) for selection. After 2 weeks of selection, colonies were pooled, and CD151-positive cells were sorted by flow cytometry. For double transfectants, human α3 cDNA in eukaryotic expression vector pRcCMV was cotransfected (into NIH3T3 cells) with CD151 mutant plasmid DNA and selected using both G418 (1 mg/ml; Life Technologies) and Zeocin. A15/TALLA1 plasmid DNA was kindly provided by Dr. Osamu Yoshie (Kinki University, Osaka, Japan), subcloned into pRcCMV vector, and selected in G418 after stable transfection into NIH3T3 cells. To assess cell surface expression, NIH3T3 transfectants were analyzed by flow cytometry as previously described (Zhang and Hemler, 1999). Cells were incubated with negative control monoclonal antibody (mAb) and specific mAbs and then with FITC-conjugated goat anti-mouse IgG and were analyzed using a FACScan flow cytometer (Becton Dickinson, Mountain View, CA). Fluorescence with negative control mAb was subtracted to give specific mean fluorescence intensity (MFI) units.

Antibodies and Other Proteins

mAbs used in this study were anti-human CD151 mAbs 5C11 (Yauch et al., 1998) and 1A5 (Testa et al., 1999; provided by Dr. J Testa); anti-human integrin α3 subunit IIF5 (Weitzman et al., 1993), anti-human CD147 mAb 8G6 (Berditchevski et al., 1997), anti-human integrin αV subunit mAb P3G8 (Wayner et al., 1991), anti-mouse CD9 mAb KMC8 (PharMingen, San Diego, CA), anti-mouse integrin α6 subunit mAb GoH3 (PharMingen), anti-mouse CD44 mAb KM114 (PharMingen), and negative control mAb P3 (Lemke et al., 1978). A15 mAbs B2D and A2 M 30.3 were kindly provided by Dr. Osamu Yoshie (Kinki University, Osaka, Japan) and Dr. F. Lanza (Strasbourg, France), respectively. The rabbit polyclonal antibody 6843, against α6A integrin cytoplasmic domain, was a gift from Dr. V. Quaranta (The Scripps Research Institute, La Jolla, CA). Horseradish peroxidase (HRP)-conjugated goat anti-mouse and goat anti-rabbit secondary antibodies (Sigma, St. Louis, MO) were also used. Matrigel, a solubilized basement membrane matrix extracted from the Engelbreth-Holm-Swarm (EHS) mouse sarcoma, was purchased from BD Bioscience (Bedford, MA). Other ECM proteins used in this study were human plasma fibronectin (Life Technologies), and mouse laminin 1 (Life Technologies). PDGF-BB was from BD Bioscience, and bFGF was from Roche Molecular Biochemicals, Indianapolis, IN.

In Vitro Morphogenesis Assay

The spontaneous formation of interconnecting web-like structures by NIH3T3 or HUVEC cells on Matrigel was used to assess morphogenic potential. Matrigel was plated into 24-well plates (0.4 ml/well) and allowed to solidify for 2 h at 37°C. NIH3T3 or HUVEC cells were then seeded into each well at a concentration of 1 × 10⁵ cells/well in a 0.6 ml volume of DMEM. The final concentration of fetal calf serum was 5%. Cells on Matrigel were cultured at 37°C in 10% CO₂ and photographed after ∼8, 10, or 24 h or 7 d, as indicated. Images were captured using Scion Image 1.60 (Scion Corp.. Frederick, MD) software through a video camera (TM-7AS; PULNiX America, Inc., Sunnyvale, CA) attached to an inverted phase contrast microscope. All results were obtained in at least two independent experiments.

Immunoprecipitation and Western Blot

Immuoprecipitations were carried out as described (Zhang and Hemler, 1999). Briefly, NIH3T3 transfectants were lysed in 1% Brij 99 lysis buffer (containing 150 mM NaCl, 25 mM HEPES, 2 mM phenylmethylsulfonylfluoride, 20 μg/ml leupeptin, 20 μg/ml aprotinin, 2 mM sodium vanadate, and 2 mM sodium fluoride), at 4°C for 1 h. Lysates were preincubated (2 times) with a combination of protein A– and protein G–Sepharose beads (Pharmacia Amersham Biotech, Uppsala, Sweden) at 4°C and each time were clarified by centrifugation at 10,000 rpm centrifugation. Next, mAb preabsorbed protein A– and protein G–Sepharose beads were incubated with cell lysate at 4°C overnight. Beads were washed with 1% Brij 99 lysis buffer three times, dissolved in Laemmli sample buffer and heated at 95°C for 5 min, and then proteins were resolved by 10% SDS-PAGE. After electrophoretic transfer, nitrocellulose membranes (Schleicher & Schuell, Keene, NH) were sequentially blotted with primary antibody and HRP-conjugated anti-mouse IgG (Sigma) and then visualized with chemiluminescence reagent (New England Nuclear Life Science, Boston, MA).

DIC Time-lapse Video Microscopy

As described elsewhere (Zhang et al., 2001a, 2001b), acid-washed glass coverslips were affixed to a 60-mm Petri dish, covering a 12-mm hole. Coverslips were coated 2 h at 37°C with 100 μl Matrigel. Immediately before image acquisition, NIH3T3 transfectants were detached with 2 mM EDTA in PBS, washed once with PBS, and plated onto coverslips in complete DMEM medium. Image acquisition was achieved using a Zeiss Axiovert 135 microscope (Thornwood, NY) with a VS25 shutter controlled by a Uniblitz D122 driver (Vincent Associates, Rochester, NY) and a video camera (TM-7AS; PULNiX America, Inc.) connected to a Power Macintosh 6500 equipped with a VG-5 frame grabber (Scion Corp., Frederick, MD) through a focusing monitor (PVM-137; Sony Corp., Parkridge, NJ). A macro written for Scion Image 1.60 (Scion Corp.) controlled the shutter driver and image acquisition. Images were captured every 5 min for 13 h, as cells were maintained in a humidified, 37°C, 10% CO₂ environment in a custom-built stage incubator. DIC images were obtained using a Hoffman Modulation Contrast system (Modulation Optics Inc., Greenvale, NY) consisting of a contrast objective, a condenser, and a contrast control polarizer.

CD151–α6β1-dependent Network of Cord-like NIH3T3 Cell Structures on Matrigel

Upon plating on basement membrane Matrigel, NIH3T3 cells assembled into a network of cord-like structures visible after 8 h (Figure 1A) and 30 h (Figure 1B). The process began at ∼5 h after cell plating, and the cord-like network pattern sometimes lasted more than 7 days (Figure 1C), depending on the batch of Matrigel. Typically, 1–1.5 × 10⁵ cells (in area = 2 cm²) were sufficient to yield cord-like structures, whereas 5 × 10⁴ cells was often insufficient. Cell network formation was diminished in the absence of serum but was strongly promoted in the presence of either PDGF-BB (40 ng/ml) or bFGF (40 ng/ml). For time-lapse images of cellular network formation and accompanying video, see Figure 5A, below.

Because a major component of Matrigel is laminin-1, we considered that integrin α6β1 (α6β4 is not present in NIH3T3 cells) and associated CD151 may be involved in network formation. Indeed, anti-murine α6 function–blocking mAb GoH3 dramatically inhibited the 8 h (Figure 1A) and 30 h (Figure 1B) formation of cord-like structures by NIH3T3-pZeo mock transfectants. Network formation by NIH3T3-CD151 wt cells (transfected with wild-type human CD151) was inhibited not only by anti-α6 mAb GoH3 but also by anti-CD151 mAb 1A5 (Figure 1, A and B). Semiquantitative RT-PCR revealed that transfected human CD151 was present at a level two- to threefold greater than endogenous murine CD151. In the absence of human CD151, mAb 1A5 was not inhibitory (Figure 1, A and B, bottom left panels). In a separate experiment, after 7 d in culture, NIH3T3-CD151-α3β1 cotransfectants showed again a dramatic inhibition of cord-like networks by anti-α6 and anti-CD151 antibodies (Figure 1C). In contrast, antibodies to CD44, to another tetraspanin protein (CD9), or to α3 integrin had minimal effect after 7 d (Figure 1C) or after 30 h (our unpublished results), even though each of those molecules was present in these NIH3T3 cells at a level higher than either endogenous α6β1 integrin or transfected CD151.

As seen elsewhere, CD151 may physically associate with α6β1 integrin (Serru et al., 1999). To test for CD151–α6 integrin complex formation in NIH3T3 cells, cells were lysed, and then human CD151 immunoprecipitates were blotted for integrin α6 subunit. Levels of α6 associated with human CD151 (Figure 2, lane d) were comparable to the levels of total detectable α6 (lanes a and c), thus indicating a high stoichiometry interaction. These results are consistent with human CD151 being more prevalent than endogenous murine CD151 (which would also be expected to associate with α6 integrin). In a control experiment, anti-CD151 antibody failed to coimmunoprecipitate α6 from cells that did not contain human CD151 (lane b). Immunoprecipitation of the A15 tetraspanin protein yielded minimal associated α6 integrin (lane h), even though both the α6 integrin (lane g) and the A15 molecule (see Figure 4, below) were well expressed in NIH3T3-A15 transfectants. Exchange of the CD151 C-terminal cytoplasmic tails with the tail of A15 (see Figure 4 below) did not result in loss of α6 association (lane f). Thus, specificity for α6 association does not reside in the cytoplasmic tails of CD151. Importantly, expression of endogenous α6 was not perturbed upon transfection of mutant or wild-type CD151 (Figure 2, lanes a, c, and e) and as seen by flow cytometry.

CD151 C-terminal Tail Involvement

Next we discovered that the C-terminal tail of CD151 is clearly involved during CD151–α6β1-dependent network formation. A tail exchange mutation (CD151-c-A15) caused an almost complete loss of cellular network formation (Figure 3) without altering integrin association (see Figure 2 above). In contrast, the N-terminal tail exchange mutation (CD151-n-A15) had no discernible effect, whereas exchange of both tails (CD151-nc-A15) did again abolish the cord-like structures. Expression of wt A15 itself had no effect. To confirm the role of CD151 C-terminal tail of CD151, additional mutants were generated, including another exchange mutant (CD151-c-NAG2) and a deletion mutant (CD151-Δc-GFP). Again, CD151-dependent network formation among NIH3T3 cells was essentially abolished (Figure 3, bottom panels). Wild-type CD151, A15, and the various mutants were all stably expressed at comparable levels on the surface of NIH-3T3 cells (Figure 4). Surface expression of the CD151-Δc-GFP mutant was not analyzed by flow cytometry using FITC-conjugated second antibody (due to excessive GFP fluorescence) but instead was confirmed by immunoprecipitation (our unpublished results). Mutant human CD151 molecules were each present at levels two- to threefold greater than endogenous murine CD151, as indicated by semiquantitative RT-PCR.

In a time-lapse video microscopy study, wild-type CD151 transfectants initially showed a directional migration and alignment of cells. Next, there was cell–cell contact among aligned cells, and finally the cells merged into elongated rod-like structures, before condensing into thicker cellular cables (Figure 5A and attached video). In sharp contrast, CD151-c-A15 cells were relatively motile but showed no directional cell migration and no cell–cell alignment (Figure 5B and attached video).

CD151 C-terminal Tail-spreading Functions

We hypothesized that within a functional CD151–α6β1 complex, effects of CD151 tail mutation should be best seen when α6β1 is engaged with ligand. To address this, we carried out cell-spreading assays on laminin-1 (to engage α6β1) and on fibronectin (to engage α5β1). As shown in Figure 6A, a high percentage of all NIH3T3 transfectants showed abundant spreading after a 30-min incubation on fibronectin. On laminin 1, the majority of mock, CD151 wild-type, and A15 transfectants were well spread after 30 min, but the CD151-c-A15 mutant showed severely impaired spreading. Photographs of representative spread cells are presented in Figure 6B. Cell spreading on laminin-1 and a coating of Matrigel yielded comparable results (Figure 6B, bottom row), consistent with laminin-1 being a major component of Matrigel. These results emphasize that functional effects of CD151 C-terminal domain mutations are obvious only when the α6β1 integrin is engaged. Previous results illustrate that CD151 has little effect on integrin-dependent cell adhesion (Yauch et al., 1998). Consistent with this, static cell adhesion to polymerized Matrigel (at levels identical to that used in morphogenesis experiments) was not markedly different among NIH3T3 transfectants (mock, wild-type CD151, CD151-c-A15, wild-type A15). Each transfectant showed ∼950–1100 adherent cells/mm², corresponding to ∼70–81% of 50,000 input cells, in a standard cell adhesion assay (Pujades et al., 1997). Consistent with the presence of laminin 1 in the Matrigel, adhesion of NIH3T3 transfectants was strongly inhibited (∼80%) by anti-α6 integrin antibody GoH3 (our unpublished results).

Perturbation of Endothelial Cell Structures

Early passage HUVEC cells also form a network of cord-like structures when cultured on Matrigel (Sincock et al., 1999). The appearance of these cord-like structures (Figure 7) was partially disrupted by antibodies to either CD151 (mAb 5C11), or anti-α6 integrin (mAb GoH3). However, anti-CD147 mAb 8G6 and anti-αV integrin mAb P3G8 had no obvious effect. These results indicate that CD151–α6β1 complexes in multiple cell types can play a critical role during morphogenesis into cellular networks on Matrigel.

The CD151–α6β1 Complex as a Functional Unit

CD151 formed a complex with integrin α6β1 as shown here in NIH3T3 cells and as shown elsewhere in HUVECs and other cell types (Fitter et al., 1999; Serru et al., 1999; Sincock et al., 1999). The specificity of the CD151–α6β1 interaction is underscored by our failure to observe A15/Talla1 tetraspanin association with α6β1 integrin. Elsewhere, the CD151–α6β1 interaction was highly specific and was similarly retained under conditions in which other integrin–tetraspanin interactions were disrupted (Serru et al., 1999).

Although CD151-α6β1 complexes have been demonstrated previously, their functional relevance had not been demonstrated. Here we show that the CD151–α6β1 complex is acting as a functional unit because CD151 C-terminal tail mutation perturbed α6β1-dependent functions (network formation, cell spreading) only when α6β1 was engaged (on purified laminin-1 or on Matrigel) and not when a different integrin (α5β1) was engaged (on fibronectin). Furthermore, antibodies to both CD151 and α6β1 strongly inhibited network formation by HUVECs (Figure 7), NIH3T3 cells, NIH3T3-α3 cells (Figure 1) and by an immortalized α3-deficient murine kidney epithelial cell line (Wang et al., 1999; our unpublished results).

In contrast to CD151–α6β1 complexes, CD151–α3β1 complexes did not influence NIH3T3 cell morphogenesis on Matrigel. Presumably, although the α3β1 integrin may interact with laminin-1, it may not interact in the same manner or to the same extent as α6β1. Alternatively, CD151–α3β1 complexes may transmit different cellular signals, not conducive to network formation. Whereas anti-CD151 antibodies inhibited α6β1 integrin–dependent function when cells were plated on laminin-1 (as shown here); elsewhere, anti-CD151 antibodies inhibited α3β1 functions on laminin-5 but not on laminin-1 (Stipp and Hemler, 2000). Thus, CD151–α6β1 and CD151–α3β1 complexes are functionally distinct.

A New Dimension in Integrin Signaling

It is well established that integrin cytoplasmic domains play critical roles in determining the functional consequences of ligand binding (Liu et al., 2000). For example, integrin α6 cytoplasmic domains can regulate MAP kinase activation and cell migration (Gimond et al., 1998; Wei et al., 1998). Now we demonstrate that another cytoplasmic tail, that of CD151, may be just as important for α6 integrin function as integrin tails themselves. A transmembrane linker role for tetraspanins has been proposed (Hemler, 1999) because extracellular domains of tetraspanin proteins such as CD151 provide specificity for integrin association, and intracellular domains may determine association with signaling molecules such as PtdIns 4-K (Yauch and Hemler, 2000) and PKC (Zhang et al., 2001a, 2001b). Results here support the concept of TM4SF proteins as transmembrane linkers. First, a distinct region of CD151, likely extracellular and not involving the cytoplasmic tails, is needed for α6β1 integrin association. Second, the CD151 C-terminal cytoplasmic tail makes an essential contribution to NIH3T3 morphogenesis, as evidenced by tail deletion and two different tail exchange mutations. Most likely, our CD151 C-terminal tail mutants are having a dominant negative effect on endogenous CD151. While retaining integrin association, they disrupt critical CD151 tail-dependent signaling pathways that complement α6β1 integrin–signaling pathways. Neither integrin α6β1 expression levels or α6β1-dependent cell adhesion were affected by transfection of either wild-type or mutant CD151 into NIH3T3 cells. These latter results are in agreement with previous studies showing that CD151 and other tetraspanin proteins have little or no effect on cell adhesion (Hemler et al., 1996; Yauch et al., 1998). Thus, our dominant negative CD151 is altering outside-in rather than inside-out integrin signaling.

Among tetraspanin proteins, there has been little precedent for cytoplasmic domains having clear functional relevance. Now we demonstrate that the CD151 C-terminal tail is clearly distinct from the A15 and NAG2 tails with respect to its influence on cell network formation. With the CD151 tail having only ∼9 residues, it should be readily feasible to identify the critical individual amino acids in future studies. It remains to be demonstrated whether the CD151 C-terminal tail may also play a critical role in the functioning of CD151–α6β4 complexes in hemidesmosomes (Sterk et al., 2000) or in CD151–α3β1 complexes during neurite outgrowth (Stipp et al., 2001) and in cell migration (Yauch et al., 1998).

Formation of Cord-like Networks on Matrigel

Several cell types may form cord-like networks on a variety of different extracellular matrices (Vernon and Sage, 1995). Although the role of integrins (including α6β1) during network formation has been well established (Bauer et al., 1992; Berdichevsky et al., 1994; Davis and Camarillo, 1995; Vernon and Sage, 1995; Stahl et al., 1997; Sun et al., 1998), a major role for a tetraspanin protein has not previously been observed. In a prior study of endothelial cells on Matrigel, anti-CD151 antibody inhibition effects were more subtle than shown here, perhaps because of the use of different antibodies (Sincock et al., 1999).

In several previous studies, formation of cord-like structures on Matrigel has been seen as an in vitro model of angiogenesis. Indeed, our antibody inhibition results seen in NIH3T3 cells were confirmed using HUVECs, thus suggesting that CD151 may contribute to angiogenesis in particular and to branching morphogenesis in general. For fibroblasts in particular, a network of cord formation on Matrigel may be a model also for wound healing and development (Vernon and Sage, 1995). On Matrigel, cells that exert mechanical traction forces on the matrix may then align along “matrix guidance pathways” to form cord-like structures (Davis and Camarillo, 1995; Vernon and Sage, 1995).

Our time-lapse video results suggest that CD151-dependent NIH3T3 cell morphogenesis has at least three phases: cellular alignment due to tractional forces, cell motility, and cell–cell contact. It remains to be determined which of these phases is specifically facilitated by CD151. CD151 could play a key role at cell–cell contact sites (Yánez-Móet al., 1998; Sincock et al., 1999). However, because CD151 effects on cell alignment and migration precede cell–cell contact, that seems unlikely. Instead, CD151 may promote cell movement, because tetraspanin proteins in general and CD151 in particular are well established as regulators of cell motility (Hemler et al., 1996; Maecker et al., 1997; Yauch et al., 1998). Also, CD151 may promote mechanical traction–related forces. Consistent with this, CD151 tail mutation abolished cell spreading, another event requiring tractional forces.

In conclusion, our results provide perhaps the best support to date for a tetraspanin protein (CD151) having a transmembrane linker function. The extracellular portion associates with the integrin, whereas the cytoplasmic tail determines the specific consequences of integrin outside-in signaling (without altering cell adhesion). We suggest that while α6β1 is interacting with laminin-1, α6β1-associated CD151 is optimally localized to promote cell alignment with mechanical traction forces, cell motility, cell spreading, and/or other events needed for NIH3T3 cell morphogenesis into a cellular reticulum. This points to a new functional role for CD151. Finally, CD151 and α6β1 integrin also play critical roles during endothelial cell morphogenesis, illustrating the generality of our findings.