It is of crucial importance to clarify the molecular linkage between l-afadin and the COOH-terminal half of α-catenin. We have previously shown that l-afadin does not directly interact with α-catenin as estimated by affinity chromatography using the purified samples (
Sakisaka et al. 1999). Therefore, we assumed that there might be a protein that interacts with both l-afadin and α-catenin and connects them. On the basis of this assumption, we have been attempting to isolate an l-afadin–binding protein in the presence of α-catenin or an α-catenin–binding protein in the presence of l-afadin by using various methods currently available, including yeast two-hybrid, affinity chromatography, immunoprecipitation, and blot overlay. We have not yet obtained any candidate proteins, but we have found, by the yeast two-hybrid method, that full-length l-afadin directly interacts with the COOH-terminal half of α-catenin. We have confirmed this interaction by affinity chromatography, although the stoichiometry of this interaction is small. We have attempted to increase this small stoichiometry by the use of the NH
2-terminal or COOH-terminal half of l-afadin. The COOH-terminal half directly interacted with the COOH-terminal half of α-catenin, but the stoichiometry did not increase (data not shown). We have added the NH
2-terminal half of ZO-1 because it has been shown to directly interact with the AF-6 protein/s-afadin (
Yamamoto et al. 1997) and with the COOH-terminal half of α-catenin (
Itoh et al. 1997), but the NH
2-terminal half of ZO-1 does not increase the stoichiometry. We have not yet obtained the conditions where the stoichiometry is increased. Moreover, l-afadin and the COOH-terminal half of α-catenin overexpressed in COS7 cells are not coimmunoprecipitated to a significant extent. Thus, our results concerning the interaction of l-afadin with the COOH-terminal half of α-catenin are apparently inconsistent depending on the methods used: one positive, one semi-positive, and one negative. However, the negative results obtained from the immunoprecipitation experiments do not necessarily reflect in vivo negative protein–protein interactions, because coimmunoprecipitation of two proteins is sometimes affected by an extraction buffer used in experiments, and sometimes is not observed when they form a very complicated multicomplex. Taken together, l-afadin may directly interact with α-catenin under appropriate conditions, but the molecular linkage between these two proteins may not be so simple and another factor and/or posttranslational modifications of l-afadin and α-catenin may be necessary for this linkage.
Several other possible mechanisms of the linkage between the nectin/afadin and cadherin/catenin systems are conceivable. One possibility is that l-afadin indirectly interacts with α-catenin through ZO-1, because it has been shown that ZO-1 directly interacts in vitro with the AF-6 protein/s-afadin (
Yamamoto et al. 1997) and with the COOH-terminal half of α-catenin (
Itoh et al. 1997). However, we have previously shown that the stoichiometry of the interaction of l-afadin with ZO-1 is negligible as estimated by affinity chromatography (
Sakisaka et al. 1999). We have shown here that ZO-1 does not affect the interaction of l-afadin with the COOH-terminal half of α-catenin. Therefore, this possibility is unlikely. The second possibility is that l-afadin indirectly interacts with α-catenin through vinculin, because it has been shown that the COOH-terminal half of α-catenin directly interacts with vinculin in vitro (
Weiss et al. 1998) and that vinculin interacts with ponsin (
Mandai et al. 1999). However, consistent with previous reports (
Watabe-Uchida et al. 1998;
Imamura et al. 1999), we have shown by the use of nEαC-L and nEαN-L cells that vinculin is colocalized with the NH
2-terminal half of α-catenin (nEαN), but not with the COOH-terminal half (nEαC) or l-afadin. Therefore, this possibility is unlikely either. The third possibility is that F-actin is involved in the linkage between l-afadin and α-catenin, because l-afadin (
Mandai et al. 1997) and the COOH-terminal half of α-catenin (
Rimm et al. 1995) directly interact with F-actin. This possibility cannot be excluded, but the in vivo interaction of α-catenin with F-actin remains unknown.
It may be noted that nectin has a potency to recruit α- and β-catenins and E-cadherin through l-afadin at nectin-based cell–cell adhesion sites in the absence of E-cadherin or without its trans-interaction. It is likely that α-catenin is recruited there by the direct or indirect interaction with l-afadin, and that β-catenin is recruited there by the direct interaction with α-catenin. E-cadherin may be recruited there by the direct interaction with the α- and β-catenin complex through l-afadin. However, it remains unknown whether E-cadherin, which is concentrated at adhesion sites between L and EL cells stably expressing nectin, forms a cis-dimer or not.
In contrast to the recruitment of the cadherin-catenin complex to nectin-based cell–cell adhesion sites without the trans-interaction of E-cadherin, we have shown that nectin-1α, of which trans-interaction is inhibited by gD, is not recruited to E-cadherin–based cell–cell adhesion sites between two nectin-1α-EL cells. Thus, nectin trans-interacts independently of the trans-interaction of E-cadherin, and nectin, of which trans-interaction is inhibited or which lacks the ability to interact with l-afadin, does not appear to be recruited to E-cadherin–based cell–cell adhesion sites. However, it remains unresolved whether the trans-interaction of E-cadherin is dependent on the trans-interaction of nectin, because nectin has three isoforms, and cell adhesion activities of all the isoforms cannot be inhibited simultaneously at this time (
Satoh-Horikawa et al. 2000).
On the basis of these present and previous observations, we propose here at least two models for the formation of cell–cell AJs. One model is that nectin and E-cadherin independently form the respective trans-interactions and the nectin/afadin and cadherin/catenin systems recruit each other to form compact cell–cell AJs. The other model is that nectin first forms a trans-interaction that recruits, through l-afadin, the cadherin-catenin system in which E-cadherin does not trans-interact, followed by the trans-interaction of E-cadherin at nectin-based cell–cell adhesion sites, finally leading to the formation of compact cell–cell AJs. It is currently unknown which is the case, but it is likely that the nectin/afadin system plays a key role in the organization of cell–cell AJs in cooperation with the cadherin-catenin system.
We have previously shown by the use of epithelial cells in afadin (−/−) mice and (−/−) embryoid bodies that not only cadherin-based AJs, but also claudin/occludin-based TJs are impaired in these mutant cells (
Ikeda et al. 1999). Claudin and occludin are Ca
2+-independent homophilic cell adhesion molecules at TJs, of which cytoplasmic domains interact with ZO-1, -2, and -3 (
Tsukita and Furuse 1999;
Tsukita et al. 1999). ZO-1 and -2 are F-actin–binding proteins that connect the cytoplasmic regions of claudin and occludin to the actin cytoskeleton (
Tsukita et al. 1999). Moreover, we have shown that behavior of nectin and l-afadin is different from that of E-cadherin and similar to that of ZO-1 during the formation of TJs in cultured MDCK cells (
Sakisaka et al. 1999;
Asakura et al. 1999). These results suggest that the nectin/afadin system plays a key role in proper organization of not only cadherin-based cell–cell AJs, but also claudin/occludin-based TJs. It remains unknown how the nectin/afadin system organizes TJs properly, but l-afadin may also indirectly connect nectin to the component of TJs through an unidentified factor. It is of crucial importance to clarify the molecular linkages among these three different cell–cell adhesion systems.