The data from our yeast two-hybrid screen, as well as the results from pull-down assays and co-immunoprecipitation experiments demonstrate binding of the NG2 proteoglycan to the multi-PDZ domain protein MUPP1. The ability of NG2 to utilize MUPP1 as a cytoplasmic ligand is consistent with the presence of a putative PDZ-binding motif at the C-terminus of the proteoglycan. This QYWV sequence fits the general requirements of a four amino acid PDZ-binding motif in which valine occupies the final position (position 0) and tyrosine is the hydroxyl group-containing residue at position-2 [Songyang et al., 1997
]. The importance of the C-terminal portion of the NG2 cytoplasmic domain for interaction with MUPP1 is illustrated by the failure of the truncated NG2/t3 mutant to yield positive results in pull-down assays using fusion proteins representing the first MUPP1 PDZ domain (). To define the PDZ binding motif of NG2 more precisely, we will need to perform additional experiments with NG2 mutants carrying smaller C-terminal deletions and with variants in which individual residues are mutated.
The fact that the yeast two-hybrid screen yielded two cDNA clones (NIP-2 and NIP-7) representing the first MUPP1 PDZ domain but no clones for other MUPP1 PDZ domains suggests that there could be considerable specificity for binding of NG2 to the amino-terminal portion of MUPP1. Although the NG2 binding capability of all 13 MUPP1 PDZ domains remains to be assessed in detail, this expectation is partially borne out by our comparison of the NG2 binding capability of MUPP1 PDZ domains 2/3, 10/11, and 12/13. Although the PDZ pairs 10/11 and 12/13 fail to bind NG2 in the pull-down assay, the 2/3 pair has activity comparable to that of PDZ-1. Additional tests with NG2 deletion mutants will be required to determine whether PDZ-1 and PDZ-2/3 interact with the same segment of the NG2 C-terminus, or whether the PDZ domains recognize distinct binding sites in the proteoglycan. Nevertheless, there appears to be a significant preference of NG2 for binding to PDZ domains in the N-terminal portion of MUPP1. Other known MUPP1 ligands such as the 5-HT2C
receptor [Ullmer et al., 1998
] and the viral 9ORF1 protein [Lee et al., 1997
] appear to bind to PDZ domains in the C-terminal half of MUPP1, giving this scaffolding protein a decided polarity in its ligand binding. This property might be important for the ability of MUPP1 to cluster ligands in specific spatial patterns. The structural details of MUPP1 PDZ domains that determine their respective specificities for NG2, 9ORF1, and the 5-HT2C
receptor remain to be defined. An additional indication of specificity in the interaction of NG2 with the first PDZ domain of MUPP1 comes from the yeast two-hybrid data showing that our NIP-2 and NIP-7 clones fail to interact with the cytoplasmic segments of syndecan-2 and Eph2B. Thus the first MUPP1 PDZ domain has the ability to discriminate between NG2 and other potential PDZ-binding transmembrane receptors.
The interaction of NG2 with a PDZ domain-containing cytoplasmic ligand presents an interesting parallel to the case of the syndecan family of proteoglycans, which also utilize PDZ-containing molecules as cytoplasmic binding partners. Syndecans have been shown to bind via their conserved C-terminal EFYA sequence to the PDZ-containing molecules syntenin [Grootjan et al., 1997
] and CASK/LIN-2 [Cohen et al., 1998
; Hsueh et al., 1998
]. The colocalization of CASK and syndecan-2 in neuronal synapses [Hsueh et al., 1998
] suggests a role for this interaction in the development and/or stabilization of these complex structures. In support of this idea, syndecan-2 has been shown to be specifically clustered on dendritic spines and to be involved in the morphologic maturation of the spines, which serve as primary sites for the formation of excitatory synapses [Ethell and Yamaguchi, 1999
]. The EFYA motif of the syndecan is required for its ability to mediate spine maturation, suggesting the involvement of the syndecan-PDZ interaction in this process.
Although the functional importance of NG2 binding to MUPP1 has not been established in this type of detail, the general properties of PDZ domain-containing molecules suggest some interesting possibilities. As indicated by the foregoing set of observations on syndecans, molecules containing PDZ domains are thought to serve as cytoplasmic scaffolds for anchoring proteins or assembling them into complexes that serve either structural or signaling functions. For example, PDZ-containing proteins are postulated to be involved in localizing ion channels, cell adhesion molecules, neurotransmitter receptors, and Eph receptors and their ephrin ligands at key sub-cellular sites [Kornau et al., 1997
; Ponting et al., 1997
; Craven and Bredt, 1998
; Hata et al., 1996
; Irie et al., 1997
; Torres et al., 1998
]. A scaffolding function of this nature would appear to be relevant to two types of phenomena involving NG2. First, the detailed pattern of NG2 localization on the cell surface suggests an association with the actin cytoskeleton [Lin et al., 1996a
]. The mechanism for this association is unknown, but binding to MUPP1 could provide a means for linking NG2 to other molecules that have a cytoskeletal localization. Second, engagement of NG2 by the substratum induces reorganization of the actin cytoskeleton leading to changes in cell morphology and cell motility [Fang et al., 1999
]. This is consistent with the ability of NG2 to activate cytoplasmic signaling cascades that control cytoskeletal dynamics [Eisenmann et al., 1999
]. Binding of NG2 to MUPP1 could provide a link to these signaling pathways. In this regard it will be important to determine the relative cellular localizations of NG2 and MUPP1. Unfortunately, the existing MUPP1 antibodies do not give a clear picture of MUPP1 localization, so investigation of this aspect of the NG2/MUPP1 interaction will require the development of better reagents.
Although MUPP1 itself contains no obvious catalytic motifs, the 13 PDZ domains in the molecule provide a wealth of potential sites for bringing NG2 together with key structural and/or signaling molecules. At present only a few such MUPP1 ligands have been identified. One such ligand is the 5-HT2C
receptor [Ullmer et al., 1998
], which is mainly expressed in the central nervous system [Molineaux et al., 1989
; Mengod et al., 1990
]. It is not immediately apparent how the interaction of MUPP1 with this serotonin receptor might be relevant to the function of NG2 in the brain, since the receptor and the proteoglycan are almost certainly found in different neural cell types (neurons and oligodendrocyte progenitors, respectively). However, the closely-related 5-HT2A
receptors also contain PDZ-binding motifs similar to the one found in the 5-HT2C
receptor [Hoyer et al., 1994
], and these isoforms are found in peripheral cell types such as vascular smooth muscle and cardiac muscle [Corson et al., 1992
; Loric et al., 1992
], which also express NG2 [Grako and Stallcup, 1995
; Grako et al., 1999
]. The 5-HT2
serotonin receptors activate phospholipase C-β/protein kinase C signaling pathways [Foguet et al., 1992
; Wainscott et al., 1993
], so that interaction of NG2 with the MUPP1/5-HT2
complex might have an impact on signaling cascades that rely on phosphatidylinositol turnover.
In addition, the PDZ-mediated interaction of MUPP1 with a second type of ligand, represented by 9ORF1 and other viral transforming proteins [Lee et al., 1997
], suggests that association of NG2 with this complex could play a role in growth control. Since 9ORF1 appears to compete with the APC tumor suppressor protein for binding to the PDZ-containing protein hDLg/SAP97 [Matsumine et al., 1996
; Lee et al., 1997
], it seems possible that APC might also be able to interact with the 9ORF1 binding site in MUPP1. If the APC/MUPP1 or 9ORF1/MUPP1 interactions were modulated by NG2 binding to MUPP1, this would provide a mechanism for involvement of NG2 in pathways that are relevant to cell proliferation.
We have previously suggested such a role for NG2, based on its pattern of expression during development. Although NG2 is expressed in a variety of developing tissues, a common theme in each case is the expression of the proteoglycan on immature, mitotic progenitor cells followed by its down-regulation once these progenitors undergo terminal differentiation [Stallcup and Beasley, 1987
; Nishiyama et al., 1991b
; Nishiyama et al., 1996
; Grako and Stallcup, 1995
; Levine and Nishiyama, 1996
]. Moreover, NG2 expression is up-regulated in a number of neoplastic cell types, including melanomas [Real et al., 1985
], glioblastomas [Schrappe et al., 1991
], chondrosarcomas [Leger et al., 1994
], and lymphomas [Behm et al., 1996
], and some evidence has been presented that NG2 is responsible for the increased rate of proliferation of these tumor cells [Harper and Reisfeld, 1987
; Burg et al., 1998
]. Although the relationship between the NG2/MUPP1 complex and APC or 9ORF1-mediated processes is speculative at this point, such a model nevertheless provides a framework for testing specific hypotheses involving these molecules. The results of these experiments, coupled with identification of additional MUPP1 ligands, may shed more light on the functional importance of the NG2/MUPP1 interaction.