Although hDlg is known to localize to sites of cell-cell contacts, to interphase nuclei, and to the midbody of cells in cytokinesis [2, 18, 23], little is known about its binding partners at sites other than cell-cell contacts. Here we report that hDlg associates with activated MEK2, a protein specifically found at the midbody ring during cytokinesis . Using in vitro binding assays, we found that the PDZ domains of hDlg interact directly with the C-terminal peptide of MEK2, which contains a Class I consensus PDZ-binding motif. Interestingly, while others have very recently shown partial co-precipitation of hDlg and MEK2 from asynchronous HEK-293 and human vascular endothelial cell lysates , in the cellular contexts we tested, the association between hDlg and full-length MEK2 is only detectable when MEK2 is activated. Several lines of evidence, including these data, suggest that the C-terminal PDZ-binding motif of full-length MEK2 may only be accessible to hDlg in the activated protein. Structural studies have shown that upon activation, MEK2 likely undergoes a change of conformation common to other kinases [34, 35]. This change of conformation results in the relative rotation of the N- and C-terminal lobed-structures forming the catalytic domain. It is unknown how this rotation affects the C-terminal PDZ-binding sequence of MEK2 because the C-terminal seven amino acids, including this region, were not well ordered in the MEK2 structure . Nevertheless, the crystallographic structure of MEK2 bound to Mg-ATP and a small molecule non-competitive inhibitor, PD334581, clearly shows that in this active-conformation homodimer, the PDZ-binding sequences should be accessible to the PDZ repeats of hDlg. Although not structured, the C-terminal seven residues would extend at the periphery of the complex, near the ATP binding site, and not be buried in the dimer interface . Further supporting our hypothesis, we note that both the MEK2 clone isolated in our two-hybrid screen and the clone identified in a similar screen by Magai et al. (2011, ) coded for only the C-terminal portion of MEK2. Therefore in assays using a C-terminal fragment of MEK2 (Ref.  and herein), or a C-terminal MEK2 peptide (herein), the C-terminal PDZ-binding motif is likely more exposed than in full-length MEK2 clones, allowing for an interaction that is normally regulated by a phosphorylation-driven conformational change in the full-length protein. While our results and those of Maiga et al.  provide some initial information as to the binding specificities and the mode of interaction between hDlg and MEK2, a better understanding of how, and in which context, these two proteins can interact will require further structural studies. Other unresolved questions include how the intramolecular interaction between the GK and SH3 domains of hDlg, the self-association of hDlg proteins via their N-terminal ends, and the identity of hDlg splice variants contribute to the regulation of its binding to MEK2.
The role of MEK/ERK-dependent signaling in the control of cell-cycle has been extensively studied, particularly its role in G1/S phase transition. Recent studies have shown that activation of this signaling pathway is also required for mitotic progression. During late anaphase and cytokinesis, the two nascent daughter cells are separated by a cytoplasmic bridge characterized by a bundle of antiparallel microtubules forming the central spindle. At the half-point of the central spindle, a structure called the midbody determines the site of cell abscission and recruits proteins that are necessary for the completion of mitosis. The activated forms of MEK1/2, ERK and the ERK substrates RSK1-3 are found in the central spindle and in the midbody [29–31]. Inhibition of MEK activity prevents the completion of mitosis in cells released from nocodazole-mediated mitotic arrest, thus suggesting the role of MEK1/2 and downstream substrates in cytokinesis [31, 36]. The interaction observed between hDlg and activated MEK2 suggests that both proteins should co-distribute at specific stages of the cell cycle when MEK2 is activated. Indeed, we have found that certain variants of hDlg, like activated MEK2, localize specifically to the midbody ring structure during cytokinesis.
It is unclear how hDlg is targeted to the midbody, however, because the PDZ domains of hDlg are insufficient to target the protein to the midbody , it cannot rely on activated MEK2. Others have shown that a protein fragment spanning only the C-terminal SH3 and GK domains of hDlg localized to the midbody when overexpressed, competing with endogenous hDlg and disrupting cell growth . The GK domain of hDlg alone localizes to the midbody and causes a mild cytokinesis defect when overexpressed . This suggests that while the N-terminal region of hDlg, including its PDZ domains, may be dispensable for proper localization of hDlg at the midbody, it is essential to its function.
A possibility raised by our data is that membrane-bound I3-variants of hDlg are redistributed to the midbody. I3-variants of hDlg were previously found to be localized to the membrane at sites of cell-cell junctions . Membrane-bound hDlg was demonstrated to be closely associated with E-cadherin-containing adhesion complexes in human epithelial cells and cells deficient in hDlg expression fail to organize their cortical actin cytoskeleton and are unable to stabilize their adherens junctions . E-cadherin protein was also previously found in the cleavage furrow of mitotic epithelial cells, consistent with the observation that the daughter cells form adherens junctions with each other immediately following cytokinesis and are not separated by neighboring cells . Herein, we demonstrate that E-cadherin expression is required for localization of hDlg to the midbody in the same epithelial cells. E-cadherin depletion also led to a marked narrowing of the midbody structure. While our data does not specifically address whether E-cadherin depletion itself or its effect on the central spindle are the direct cause of the mis-localization of hDlg, because E-cadherin and hDlg are known to physically interact in confluent epithelial cells , we surmise that E-cadherin may directly recruit hDlg to the midbody. Taken together, these data suggest that hDlg may also be important for the formation of adherens junction between the two daughter cells.
Membrane-bound I3-containing variants of hDlg could either relocalize to the midbody from existing adherens junction sites or be recruited to the midbody from intracellular membrane systems. In this report, we show that the localization of hDlg to the midbody ring precedes the constriction of the intracellular cytoplasmic bridge, while post-Golgi secretory vesicles are delivered to the site of abscission after membrane constriction, during the late stage of cytokinesis , arguing against recruitment from intracellular membranes. Although only I3-variants of hDlg are preferentially associated with membrane systems [18, 19], others have shown that the transport and/or anchoring of overexpressed hDlg to the midbody is independent of I2 and I3 [23, 24]. Taken together our findings and those from these previous reports are consistent with the idea that SH3- or GK-binding proteins are responsible for the transport and or anchoring of hDlg to the midbody, while alternative splicing of the I3- vs. I2-insertion may determine which pool of hDlg relocalizes to this site, perhaps by regulating the accessibility of the GK domain.
hDlg plays an important role in clustering signaling molecules at different intracellular sites. In this report, we describe the interaction of hDlg with MEK2 which was also reported recently by Maiga et al. . In contrast to hDlg, however, the recruitment of phosphorylated MEK2 to the midbody seems to occur normally in E-cadherin-depleted cells indicating that although other assays show that hDlg can bind directly to phosphorylated MEK2, hDlg is not required for MEK2 recruitment to the midbody. Nevertheless, we cannot exclude that hDlg may serve as a scaffold to form a signalosome that is responsive to MEK/ERK signaling during cytokinesis. Previous studies have identified several other kinases that bind to hDlg. First, PBK/TOPK, a kinase which is activated during mitosis by the cyclin B/CDK1 complex and then phosphorylates p38-MAPK, interacts with the PDZ repeats of hDlg [9, 38]. Interestingly, an activated form of PBK/TOPK was recently found to be associated with the central spindle and to promote cytokinesis ; this association with the central spindle may reflect its interaction with hDlg. A second kinase interacting with hDlg is p85/PI3K, which it recruits to E-cadherin-mediated sites of cell-cell contact in human intestinal epithelial cells where hDlg plays a key role in the organization and stabilization of adherens junctions . Importantly, this recruitment of p85/PI3K is dependent on a change in the phosphorylation pattern of hDlg triggered by a yet unidentified kinase . Finally, the SAPK3/p38γ MAPK is also known to interact with the PDZ repeats of hDlg and to phosphorylate hDlg. Phosphorylation of hDlg by SAPK3/p38γ MAPK results in its dissociation from the cytoskeleton, apparently by disrupting the interaction between the GK domain of hDlg and its binding partner GKAP . Therefore, not only is hDlg implicated in the recruitement of several kinases at specialized sites in cells but, in turn, the activity of hDlg, and potentially its localization, is modulated by kinases.