• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of embojLink to Publisher's site
EMBO J. Mar 22, 2006; 25(6): 1406–1417.
Published online Mar 2, 2006. doi:  10.1038/sj.emboj.7601030
PMCID: PMC1422156

Oncogenic function for the Dlg1 mammalian homolog of the Drosophila discs-large tumor suppressor

Abstract

The fact that several different human virus oncoproteins, including adenovirus type 9 E4-ORF1, evolved to target the Dlg1 mammalian homolog of the membrane-associated Drosophila discs-large tumor suppressor has implicated this cellular factor in human cancer. Despite a general belief that such interactions function solely to inactivate this suspected human tumor suppressor protein, we demonstrate here that E4-ORF1 specifically requires endogenous Dlg1 to provoke oncogenic activation of phosphatidylinositol 3-kinase (PI3K) in cells. Based on our results, we propose a model wherein E4-ORF1 binding to Dlg1 triggers the resulting complex to translocate to the plasma membrane and, at this site, to promote Ras-mediated PI3K activation. These findings establish the first known function for Dlg1 in virus-mediated cellular transformation and also surprisingly expose a previously unrecognized oncogenic activity encoded by this suspected cellular tumor suppressor gene.

Keywords: adenovirus, Dlg1, E4-ORF1, PI 3-kinase, tumor suppressor

Introduction

Dlg1, also known as hDlg or SAP97, is a mammalian homolog of the Drosophila discs-large (Dlg) tumor suppressor protein (Lue et al, 1994; Muller et al, 1995). These proteins localize to specialized cell–cell contact membrane sites (i.e., septate, adherens, and synaptic junctions) and are required for both junction formation and proper polarity establishment in cells (Humbert et al, 2003; Nguyen et al, 2003; Laprise et al, 2004). As MAGUK family proteins, they lack catalytic activity and instead consist of multiple protein interaction modules, including three PDZ domains, an SH3 domain, and a guanylate kinase-homology (GK) domain (Gonzalez-Mariscal et al, 2000). Accordingly, these polypeptides function as molecular scaffolds to organize their targets into supramolecular signaling complexes and to translocate them to the plasma membrane (Kim and Sheng, 2004).

Evidence supports the assertion that Drosophila Dlg and mammalian Dlg1 encode evolutionarily conserved tumor suppressor proteins. Drosophila Dlg mutations cause cellular overgrowth in the larval brain and imaginal discs, where cells exhibit a neoplastic morphology (Woods et al, 1996). Upon transplantation into wild-type (wt) flies, Dlg-mutant imaginal discs produce cancerous outgrowths resembling metastatic tumors (Woodhouse et al, 1998), perhaps reflecting the capacity of Dlg to prevent abnormal cellular invasion into adjacent tissues (Goode and Perrimon, 1997). Furthermore, transgenic expression of Dlg1 by Drosophila Dlg mutants rescues their cellular overgrowth phenotype (Thomas et al, 1997), demonstrating conservation of the tumor suppressor function. Dlg1 likewise prevents unscheduled cellular proliferation in ocular lenses of mice (Nguyen et al, 2003), sustains mutations accompanied by a loss of heterozygosity in human breast carcinomas (Fuja et al, 2004), and exhibits reduced expression in invasive human cervical carcinomas (Cavatorta et al, 2004; Lin et al, 2004).

Reports showing that three different human virus oncoproteins independently evolved to target Dlg1 in cells provide the most compelling evidence implicating this cellular factor in human cancer. Human adenovirus type 9 (Ad9) elicits mammary tumors in experimental animals (Javier et al, 1991), and high-risk human papillomavirus (HPV) and human T-cell leukemia virus type 1 (HTLV-1) are etiological agents for cervical carcinoma and adult T-cell leukemia, respectively (Barmak et al, 2003; Heise, 2003). Strikingly, the otherwise unrelated oncogenic determinants of these tumor viruses (Ad9 E4-ORF1, HPV E6, and HTLV-1 Tax) similarly possess a carboxyl-terminal PDZ domain-binding motif (PBM) that mediates binding to several different cellular PDZ proteins (Lee et al, 1997; Rousset et al, 1998; Mantovani and Banks, 2001), including Dlg1, which is the only PDZ-protein target common to all three viral proteins. Additionally, the PBM plays a key role in cellular transformation induced by these oncoproteins (Lee et al, 1997; Watson et al, 2003; Hirata et al, 2004). Evidence specifically implicating endogenous Dlg1 in this process is currently lacking, but it has been reasonably hypothesized that the tumorigenic potentials of these viral proteins stem partly from an ability to inactivate this suspected tumor suppressor protein. Consistent with this idea, HPV E6 promotes degradation of overexpressed Dlg1 (Gardiol et al, 1999) and HTLV-1 Tax overcomes cell-cycle arrest caused by Dlg1 overexpression (Suzuki et al, 1999), although one caveat is the uncertain physiological relevance of these findings.

The oncogenic potential of Ad9 E4-ORF1 was recently demonstrated to depend on its ability to promote constitutive, growth factor-independent stimulation of cellular phosphatidylinositol 3-kinase (PI3K) (Frese et al, 2003). This critical activity of E4-ORF1 requires the PBM, which mediates interactions with Dlg1 (Lee et al, 1997) and three other cellular PDZ proteins (MUPP1, MAGI-1, and ZO-2) (Glaunsinger et al, 2000, 2001; Lee et al, 2000). We report here that specific binding of E4-ORF1 to endogenous Dlg1 promotes Ras-mediated PI3K activation in cells, revealing the first established function for Dlg1 in viral oncoprotein-mediated cellular transformation. The additional finding that functional as opposed to inactivated Dlg1 mediates E4-ORF1-induced PI3K activation also exposed a surprising oncogenic activity for this suspected tumor suppressor protein.

Results

Mutation of MUPP1, MAGI-1, or Dlg1 fails to reproduce PI3K activation induced by E4-ORF1

Given that Dlg1 as well as ZO-2 (Chlenski et al, 1999a, 1999b, 2000) are suspected tumor suppressors and that E4-ORF1 sequesters MUPP1, MAGI-1, and ZO-2 in the cytoplasm of cells (Glaunsinger et al, 2000, 2001; Lee et al, 2000), we postulated that PI3K activation provoked by Ad9 E4-ORF1 may stem from inactivation of one of these cellular targets. If so, disruption of the corresponding cellular gene should recapitulate E4-ORF1-induced PI3K activation in cells.

To explore this idea, we isolated mouse embryo fibroblasts (MEF) from wt mice and matched littermates carrying either a large deletion mutation encompassing the entire MUPP1 gene (Bell et al, 1995) or a targeted mutation interrupting either the MAGI-1 gene (in preparation) or the Dlg1 gene (Caruana and Bernstein, 2001). The lack of ZO-2 mutant mice precluded inclusion of the respective mutant MEF in this study; however, ZO-2 seems unlikely to play a role in E4-ORF1-induced PI3K activation given that E4-ORF1 proteins encoded by human Ad3, Ad5, and Ad12 activate PI3K (Frese et al, 2003), yet do not bind this single, unique PDZ-protein target of Ad9 E4-ORF1 (Glaunsinger et al, 2001).

MUPP1−/− or MAGI-1−/− MEF failed to express the respective wt gene product, as did Dlg1gt/gt MEF, which instead expressed the Dlg/β-geo fusion protein containing the amino-terminal 549 residues of the ~900-residue wt Dlg1 polypeptide (Supplementary Figure 1A and B). Following serum deprivation and subsequent stimulation by platelet-derived growth factor (PDGF), all mutant MEF and matched wt MEF displayed comparable basal and PDGF-induced levels of activated, phosphorylated protein kinase B (PKB) (Supplementary Figure 1A), a key downstream effector of PI3K. Throughout this study, Thr308- and Ser473-phosphospecific PKB antibodies yielded identical results, so each was used interchangeably. As Dlg/β-geo could feasibly retain some function, including the postulated capacity to suppress PI3K activation, we stably downregulated this protein in Dlg1gt/gt MEF using a Dlg1-specific short hairpin RNA (shRNA). Despite substantial, specific downregulation of Dlg/β-geo, Dlg1 shRNA-expressing Dlg1gt/gt MEF displayed basal and PDGF-induced levels of activated PKB identical to those of Dlg1gt/gt MEF (Supplementary Figure 1C). These observations did not support our hypothesis that E4-ORF1-induced PI3K activation simply results from functional inactivation of MUPP1, MAGI-1, or Dlg1 in cells.

PI3K activation and anchorage-independent growth induced by E4-ORF1 depend on Dlg1

Our negative results prompted consideration of the opposite hypothesis, wherein PI3K activation induced by E4-ORF1 requires one of its cellular PDZ-protein targets. This model predicted instead that MUPP1−/−, MAGI-1−/−, or Dlg1gt/gt MEF might fail to support E4-ORF1-induced PKB activation. Significantly, during transient E4-ORF1 expression, Dlg1gt/gt MEF, but not MUPP1−/− or MAGI-1−/− MEF, displayed a substantial defect in supporting this activity compared to matched wt or heterozygous mutant MEF (Figure 1A). Though prolonged exposures revealed that Dlg1gt/gt MEF retained a weak capacity to support E4-ORF1-induced PKB activation, this remnant activity was eliminated by Dlg1 shRNA expression (Figure 1B). Additionally, whereas Dlg1+/+ MEF stably expressing E4-ORF1 showed constitutive activation of endogenous PKB and grew in soft agar, Dlg1gt/gt MEF stably expressing E4-ORF1 lacked these phenotypes (Figure 2A and B), similar to the former MEF treated with the PI3K inhibitor drug LY294002 (LY) (Figure 2C and D). Also notable was that Dlggt/gt MEF and Dlggt/gt MEF expressing the Dlg1 shRNA failed to grow in soft agar (Figure 2B). These findings demonstrated a requirement for functional rather than inactivated Dlg1 in both PI3K activation and oncogenic transformation induced by E4-ORF1 in MEF.

Figure 1
E4-ORF1-induced PKB activation depends on Dlg1. (A) E4-ORF1-induced PKB activation is deficient in Dlg1gt/gt MEF. Cells on 6-cm dishes were lipofected with pGW1-HA-PKB (0.5, 1, or 1 μg for MUPP1, MAGI-1, or Dlg1 MEF, respectively) and either empty ...
Figure 2
Dlg1 is also required for E4-ORF1 to induce anchorage-independent growth in MEF. (A) Dlg1gt/gt MEF stably expressing E4-ORF1 fail to accumulate activated, endogenous PKB. Dlg1+/+ or Dlg1gt/gt MEF stably transfected with either empty pBABE ...

I3-containing Dlg1 isoforms support E4-ORF1-induced PI3K activation

Due to several short insertion elements (e.g., I1, I2, I3) that arise by alternative splicing (see Supplementary Figure 1B), multiple Dlg1 isoforms can be produced in cells (Lue et al, 1994; McLaughlin et al, 2002). Despite comparable binding of E4-ORF1 to the Dlg1-I2, -I3, -I1I2, and -I1I3 isoforms (Figure 3A), transient expression of green fluorescent protein (GFP)-tagged Dlg1-I3, but not GFP-Dlg1-I2, restored robust E4-ORF1-induced PKB activation in Dlg1 shRNA-expressing Dlg1gt/gt MEF (Figure 3B). A nucleotide mismatch between the mouse Dlg1 shRNA and the rat Dlg1 cDNA used to construct Dlg1 plasmids permitted GFP-Dlg1 expression in these cells.

Figure 3
I3-containing Dlg1 isoforms mediate E4-ORF1-induced PKB activation. (A) E4-ORF1 binds comparably to four Dlg1 isoforms. COS7 cells on 6-cm dishes were lipofected with pGW1-E4-ORF1 (1.5 μg) and either empty pGW1 (−) or the indicated pGW1-HA-Dlg1 ...

In Dlg1gt/gt MEF, E4-ORF1-induced PKB activation increased markedly even at low undetectable levels of HA epitope-tagged (HA)-Dlg1-I3 expression, peaked at moderate levels, and did not increase appreciably at higher levels (Figure 3C). By contrast, GFP-tagged Dlg1-I2, MUPP1, MAGI-1, ZO-2, or Dlg1-related SAP102 (Thomas et al, 1997) lacked any such activity (Figure 3D). GFP-Dlg1-I3 likewise enhanced PKB activation induced by related human Ad3 and Ad5 E4-ORF1 proteins (Frese et al, 2003) (Figure 3E), but not by cellular H-RasV12 or polyomavirus middle T (PyMT) (Figure 3F). Additionally, Dlg1-I1I2 and Dlg1-I1I3 mirrored Dlg1-I2 and Dlg1-I3, respectively, in their capacities to support E4-ORF1-induced PKB activation (e.g., see Figure 4B) (data not shown). Thus, wt Dlg1 mediates E4-ORF1-induced PI3K activation in cells, and I3-containing Dlg1 isoforms mainly provide this function. The lack of an I3 element in Dlg/β-geo explains, at least in part, its defect in supporting E4-ORF1-induced PI3K activation.

Figure 4
The I3, PDZ1+2, and U3 or SH3 elements of Dlg1 are required for E4-ORF1-induced PKB activation. (A) Illustration of Dlg1 isoforms and deletion mutants. Mutants below Dlg1-I3 are derived from this isoform. (B) The PDZ1+2 conformational ...

Multiple Dlg1 domains contribute to E4-ORF1-induced PI3K activation

We next examined the capacity of GFP-Dlg1-I3 deletion mutants (Figure 4A; group 1) to rescue E4-ORF1-induced PI3K activation in Dlg1gt/gt MEF. Like wt Dlg1-I3, Dlg1-I3 mutants missing the amino-terminal region (NT), PDZ3, SH3, or GK restored robust E4-ORF1-induced PKB activation (Figure 4B). The modest defect of Dlg1-I3ΔGK probably arises indirectly from its unique, partial nuclear accumulation (Kohu et al, 2002) (data not shown). By contrast, Dlg1-I3 mutants missing PDZ1 or PDZ2 exhibited modest or severe activity loss, respectively, whereas Dlg1-I3 mutants missing either all three PDZ domains or PDZ1+2, which forms a single conformational unit (Lue et al, 1996), lacked any activity (Figure 4B). Given that PDZ1+2 was both necessary and sufficient to mediate Dlg1 binding to E4-ORF1 (Figure 4C), the defects of the latter Dlg1 mutants likely reflect nearly complete or complete disruption of this interaction.

To assess possible roles for other Dlg1 domains in E4-ORF1-induced PI3K activation, we examined additional Dlg1 mutants that retain PDZ1+2 and I3, but lack certain other elements (Figure 4A; group 2). We excluded mutants with the ΔGK mutation due to their confounding partial or complete nuclear localization. Whereas Dlg1 mutants missing either the unique region between PDZ2 and PDZ3 (U3) or the SH3 domain restored E4-ORF1-induced PKB activation, Dlg1 mutants lacking both elements did not (Figure 4D), suggesting that U3 and SH3 have redundant functions for this activity. In summary, Dlg1 requires PDZ1+2, I3, and either U3 or SH3 to support E4-ORF1-induced PI3K activation.

E4-ORF1 causes endogenous Dlg1 to accumulate at the plasma membrane

E4-ORF1 sequesters endogenous MUPP1, MAGI-1, and ZO-2 aberrantly within detergent-insoluble complexes in the cytoplasm of cells, suggesting functional inactivation of these cellular targets. We postulated that E4-ORF1 differentially affects Dlg1 in cells because our results implicated functional rather than inactivated Dlg1 in E4-ORF1-induced PI3K activation. In rat CREF fibroblasts stably expressing E4-ORF1, endogenous ZO-2, but not endogenous Dlg1, was sequestered within detergent-insoluble complexes (Figure 5A) that appeared as cytoplasmic punctae in indirect immunofluorescence (IF) assays (Figure 5B). Close inspection of these IF assays revealed that, in ~50% of the E4-ORF1-expressing CREF cells, but in none of the control CREF cells, some Dlg1 accumulated at the plasma membrane (Figure 5B), the known location for E4-ORF1-induced PI3K activation (Frese et al, 2003). Moreover, treatment with chlorpromazine, an inhibitor of receptor-mediated endocytosis (Sieczkarski and Whittaker, 2002), increased detection of Dlg1 plasma membrane staining to 95% of E4-ORF1-expressing CREF cells, but to less than 10% of control CREF cells (Figure 5C), implying that E4-ORF1-induced Dlg1 membrane accumulation is a dynamic process continually counterbalanced by endocytosis.

Figure 5
Binding of E4-ORF1 to Dlg1-I3 promotes their translocation to the plasma membrane. (A) E4-ORF1 fails to sequester endogenous Dlg1 within detergent-insoluble complexes. Confluent CREF cells stably transfected with empty pBABE (vector) or pBABE-E4-ORF1 ...

Binding to E4-ORF1 triggers Dlg1-I3 to translocate to the plasma membrane

We also tested whether E4-ORF1 can cause GFP-Dlg1 to accumulate at the plasma membrane of 293T cells. When expressed alone, GFP-Dlg1-I2 and GFP-Dlg1-I3 similarly localized diffusely in the cytoplasm (Figure 5D), as did E4-ORF1, which additionally exhibited characteristic accumulation within cytoplasmic punctae (data not shown) (Weiss et al, 1996). Expression of GFP-Dlg1-I3 with E4-ORF1, however, caused both proteins to translocate to the plasma membrane in ~70% of transfected cells (Figure 5D). By contrast, expression of GFP-Dlg1-I2 with E4-ORF1 instead promoted both proteins to concentrate within cytoplasmic punctae in ~70% of transfected cells (Figure 5D). The common requirement for I3 in both PI3K activation and Dlg1 membrane translocation induced by E4-ORF1 served to link these two processes.

Unpublished data indicate that E4-ORF1 consists of two separate domains, the PBM and Domain 2 (D2), the latter of which mediates binding to an undetermined cellular phosphoprotein (pp70). Supporting this model, PBM mutations (e.g., V125A, T123D, and IIIA) eliminate E4-ORF1 binding to PDZ-protein targets, but not to pp70, and, conversely, D2 mutations (e.g., IA, IIA, and IIB) abolish E4-ORF1 binding to pp70, but not to PDZ-protein targets (in preparation) (Lee et al, 1997; Weiss and Javier, 1997). Disruption of either domain alone, however, suffices to inhibit E4-ORF1-induced PI3K activation (Frese et al, 2003).

We examined whether the two E4-ORF1 domains also function as a unit to promote GFP-Dlg1-I3 translocation to the plasma membrane. In these assays with 293 T cells, D2 mutant IIB behaved like wt E4-ORF1, whereas PBM mutant IIIA lacked any appreciable activity (Figure 5E). Other E4-ORF1 D2 and PBM mutants exhibited phenotypes indistinguishable from those of IIB and IIIA, respectively (data not shown). Results with PBM mutants reinforced the link between E4-ORF1-induced PI3K activation and Dlg1 membrane translocation, and further suggested that the interaction per se between E4-ORF1 and Dlg1-I3 suffices to trigger the latter process. Thus, E4-ORF1 binding to Dlg1 PDZ1+2 may cause cytoplasmic Dlg1 in the inactive, closed conformation to assume an active, open conformation capable of binding membrane- and cytoskeleton-associated proteins (Wu et al, 2000). The dispensability of E4-ORF1 D2 for this process signified specific participation in the subsequent PI3K activation event.

Identical Dlg1 domains determine both PI3K activation and Dlg1 membrane translocation induced by E4-ORF1

We next tested the capacity of E4-ORF1 to promote membrane translocation of all Dlg1 proteins shown in Figure 4A. Identical to wt Dlg1, Dlg1 mutants expressed alone in 293 T cells localized diffusely in the cytoplasm (e.g., see Figure 5D) (data not shown). Again, the single exception was Dlg1-I3ΔGK, which partially accumulated in the nucleus, a property that likely explained the modest defect of this mutant in E4-ORF1-induced membrane translocation (Table I).

Table 1
Capacities of wt and mutant Dlg1 proteins to support E4-ORF1 functions

More important, the majority of Dlg1 proteins that could support wt levels of E4-ORF1-induced PKB activation were fully competent for E4-ORF1-induced membrane translocation (I1I3, ΔNT, ΔPDZ3, ΔSH3, ΔU3/PDZ3, ΔPDZ3/SH3, ΔNT/SH3, ΔNT/U3/PDZ3), whereas all Dlg1 proteins having little or no capacity to support E4-ORF1-induced PKB activation displayed very weak or no E4-ORF1-induced membrane translocation (I2, ΔPDZ2, ΔPDZ1+2, Δ3PDZ, ΔU3/SH3, ΔU3/PDZ3/SH3, ΔNT/U3/PDZ3/SH3) (Table I). These findings revealed an intimate connection between E4-ORF1-induced PI3K activation and Dlg1 membrane translocation, though some slight deviations from this trend were noted. For example, ΔPDZ1 displayed modestly impaired E4-ORF1-induced PKB activation, yet wt levels of E4-ORF1-induced membrane translocation, and, conversely, ΔSH3, ΔPDZ3/SH3, and ΔNT/SH3 exhibited wt levels of E4-ORF1-induced PKB activation, yet modestly impaired E4-ORF1-induced membrane translocation (Table I). These observations may indicate that PDZ1 and SH3 have some specific role in PI3K activation or Dlg1 membrane translocation, respectively, induced by E4-ORF1.

Ras mediates Dlg1-dependent PI3K activation induced by E4-ORF1

The fact that we consistently failed to detect PI3K associated with either Dlg1 or E4-ORF1 in cells (data not shown) implied that E4-ORF1 indirectly activates PI3K through a cellular intermediate, such as receptor or nonreceptor tyrosine kinases, heterotrimeric G-protein-coupled receptors, or monomeric G-proteins. Due to additional failures at detecting associations between PI3K and tyrosine-phosphorylated proteins in E4-ORF1-expressing cells (Supplementary Figure 2A) and at diminishing E4-ORF1-induced PKB activation with specific inhibitors of receptor tyrosine kinases (genistein) or src kinases (PP2) (data not shown), as well as heterotrimeric G-proteins (Supplementary Figure 2B), our subsequent efforts focused on Ras monomeric G-proteins. We found that expression of dominant-negative mutant H-Ras-N17 substantially diminished E4-ORF1-induced PKB activation in mouse 3T3 cells (Figure 6A), as well as in Dlg1gt/gt MEF reconstituted with GFP-Dlg1-I3 (Figure 6B). These results were specific because H-Ras-N17 did not inhibit PKB activation induced by constitutively activated PI3K (Auger et al, 2000) (Figure 6A). Moreover, expression of wt H-Ras, as well as wt N-Ras and K-Ras (data not shown), but not wt R-Ras, reproducibly augmented E4-ORF1-induced PKB activation (Figure 6A).

Figure 6
PKB activation by E4-ORF1 requires Ras. (A) Dominant-negative (DN) Ras-N17 inhibits E4-ORF1-induced PI3K activation. NIH 3T3 cells on 6-cm dishes were lipofected with pGW1-HA-PKB (0.5 μg) and the indicated combinations of pGW1-E4-ORF1 (25 ng), ...

Considering that Ras normally stimulates not only the PI3K pathway but also the Raf/MEK/ERK pathway, the notion that Ras mediates E4-ORF1-induced PI3K activation initially seemed contradictory to the reported absence of activated ERK1/2 in E4-ORF1-expressing CREF cells (Frese et al, 2003). In fact, we consistently detected lower levels of activated, phosphorylated ERK1/2 in E4-ORF1-expressing CREF cells than in control CREF cells (Figure 6C). Reports showing that activated PKB can inhibit Raf activation (Zimmermann and Moelling, 1999), however, provided a plausible explanation for these observations. For example, E4-ORF1 could conceivably increase Ras signaling selectively through the PI3K/PKB pathway and thereby promote PKB-mediated suppression of Raf. If so, we predicted that treatment with a PI3K inhibitor to block PKB activation would cause activated ERK1/2 to accumulate above basal levels in E4-ORF1-expressing cells, but not in control cells. The fact that experiments with CREF cells (Figure 6C) and MEF (unpublished data) yielded this exact outcome supported the contention that Ras mediates Dlg1-dependent PI3K activation induced by E4-ORF1.

Model

Based on our cumulative results and published findings on Dlg1, we propose the following model for Dlg1-dependent PI3K activation by the viral E4-ORF1 oncoprotein (Figure 7). (1) In the cytoplasm, the PBM of E4-ORF1 mediates an interaction with Dlg1 PDZ1+2, which functions as a molecular switch. With PDZ1+2 unoccupied, Dlg1 favors a closed conformation and localizes in the cytoplasm because sterically hindered SH3, I3, and possibly U3 cannot bind membrane- and cytoskeleton-associated complexes at the plasma membrane. (2) With PDZ1+2 occupied by E4-ORF1, cytoplasmic Dlg1 in a closed conformation undergoes a dramatic structural change to an open conformation, where SH3, I3, and U3 are exposed and active to engage in protein interactions. (3) By binding to membrane- and cytoskeleton-associated proteins at the plasma membrane, (4) SH3, I3, and U3 recruit cytoplasmic E4-ORF1/Dlg1 complexes to the plasma membrane. (5) E4-ORF1 D2 recruits pp70 into the heterocomplex. (6) The resulting heterocomplex activates Ras, which constitutively localizes at the plasma membrane, by an undetermined E4-ORF1 D2-dependent mechanism that channels most signaling to the PI3K pathway.

Figure 7
Model for Dlg1-dependent PI3K activation by the adenovirus E4-ORF1 oncoprotein. See text for details.

Discussion

Tumor viruses represent proven powerful tools to reveal mechanisms commonly underlying the development of human cancers. Given that Dlg1 is targeted not only by human adenovirus E4-ORF1 but also by HTLV-1 Tax and HPV E6, deciphering functional consequences for the interaction between Dlg1 and E4-ORF1 may yield valuable insights into molecular events promoting adult T-cell leukemia, cervical carcinoma, and other cancers in people. Significantly, we showed here that binding of E4-ORF1 to Dlg1 specifically promotes oncogenic PI3K activation through a Ras-dependent mechanism. These findings identified the first known function for endogenous Dlg1 in virus-mediated cellular transformation. The absolute requirement for Dlg1 in E4-ORF1-induced PI3K activation further revealed a surprising proto-oncogene function for this suspected tumor suppressor gene.

It is important to point out that, though unexpected, our finding that Dlg1 possesses an oncogenic activity does not refute a substantial body of evidence supporting its designation as a candidate tumor suppressor gene. Viral oncoproteins typically disregulate normal functions of their cellular targets, so it may be pertinent that Dlg1 was recently reported to recruit PI3K into a membrane-associated complex with E-cadherin to promote both adherens junction formation and terminal differentiation in enterocytes (Laprise et al, 2004). As this activity is predicted to induce an antiproliferative state in cells, it may represent a tumor suppressor function for Dlg1. Therefore, an intriguing possibility is that, whereas Dlg1 normally functions to promote temporally and spatially restricted E-cadherin-dependent PI3K activation during adherens junction formation in cells, Dlg1 modified by E4-ORF1 instead supports constitutive and spatially unrestricted, Ras-dependent PI3K activation at the plasma membrane. This model interestingly hints that E4-ORF1 may simultaneously inactivate a tumor suppressor function and stimulate an oncogenic function of Dlg1.

We discovered nearly identical requirements for PDZ1+2, U3, SH3, and I3 in both Dlg1 membrane translocation and PI3K activation induced by E4-ORF1. The fact that Dlg1 translocation to the plasma membrane by definition must precede PI3K activation would seem to dictate that Dlg1 solely contributes to the latter process by trafficking E4-ORF1 to the plasma membrane. Our results cannot, however, rule out the possibility that one or more crucial Dlg1 domains function directly in both Dlg1 membrane translocation and PI3K activation induced by E4-ORF1. Consistent with this idea, despite exhibiting prominent localization at the plasma membrane, an E4-ORF1 PBM mutant tagged with a farnesylation signal failed to activate PI3K in cells (unpublished data).

Our findings showed that PDZ1+2 of Dlg1 mediates binding to E4-ORF1 and that this interaction functions as a molecular switch to trigger the resulting heterocomplex to translocate to the plasma membrane, possibly by converting cytoplasmic Dlg1 in an inactive, closed conformation to an active, open conformation, where I3, U3, and SH3 can bind membrane-associated proteins (Figure 7). Supporting this model, localization of Dlg1 to the plasma membrane in both epithelial cells and neurons depends in part on I3 (Hanada et al, 2003; Rumbaugh et al, 2003), which binds ERM family proteins that connect membrane-associated protein complexes to the cytoskeleton. Furthermore, U3 is reported to recruit Dlg1 into caveolae through an interaction with Caveolin-3 (Folco et al, 2004), and SH3 binds CASK (Nix et al, 2000) to form a membrane-associated complex. In addition to their roles in organizing cell–cell membrane junctions, domains equivalent to Dlg1 PDZ2, SH3, and I3 in Drosophila Dlg are also essential for its tumor suppressor function (Hough et al, 1997). This observation suggests that these domains of Dlg1 may dually regulate its oncogenic and tumor suppressor activities.

The fact that viral oncoproteins often mimic the cytoplasmic domain of activated growth factor receptors may provide insights into the mechanism whereby E4-ORF1 stimulates Dlg1-dependent Ras activation. Such activated cytoplasmic domains bind the Grb2 and Shc adaptor proteins, which, in turn, recruit the Sos guanine nucleotide exchange factor to the plasma membrane to stimulate Ras activity (Hancock, 2003). Future studies will investigate whether these factors also participate in Ras activation induced by E4-ORF1.

It is unclear whether, like E4-ORF1, high-risk HPV E6 and HTLV-1 Tax also activate the oncogenic function of Dlg1. Nonetheless, these viral oncoproteins similarly bind to Dlg1 PDZ1+2 (Kiyono et al, 1997; Suzuki et al, 1999), and our preliminary data suggest that they also promote GFP-Dlg1-I3 translocation to the plasma membrane. Also relevant is a report of PI3K pathway activation in HTLV-1 Tax-transformed cells (Liu et al, 2001). In overexpression assays, high-risk HPV E6 promotes degradation of Dlg1 (Gardiol et al, 1999), and substantially reduced Dlg1 levels are detected in late-stage invasive HPV-positive cervical carcinomas (Watson et al, 2002), consistent with the notion that HPV E6 inactivates Dlg1 in these contexts. Nevertheless, reduced levels of Dlg1 are found neither in HPV-positive premalignant cervical lesions (Watson et al, 2002) nor in HPV-infected keratinocytes (Lee and Laimins, 2004), revealing that HPV E6 has differential effects on Dlg1 at early versus late stages of cancer progression. Thus, an intriguing possibility is that HTLV-1 Tax and HPV E6 likewise activate the Dlg1 oncogenic function to provoke disregulated plasma membrane signaling and thereby drive early stages of cancer development.

Materials and methods

Plasmids

pGW1, pCMVBam3-Neo, pBABE encoding wt or mutant E4-ORF1 cDNAs, as well as pGW1-HA-PKB, were previously described (Weiss et al, 1997; Weiss and Javier, 1997; Frese et al, 2003). Untagged and HA epitope-tagged wt and mutant Dlg1 cDNAs, derived from SAP97 cDNAs in pRK174-SAP97-I3 (Muller et al, 1995) and pGW1-SAP97-I2 (Kim and Sheng, 1996), were inserted into pGW1, peGFP-C3, or peYFP-C1. The mouse Dlg1 shRNA oligonucleotide (AGGTCCTAAAGGTCTTGGG) was inserted into the pSUPER-retro plasmid (Oligoengine, Seattle). MUPP1, MAGI-1c, and ZO-2 cDNAs were inserted into peYFP-C1. H-Ras and R-Ras cDNAs, generated by PCR amplification from plasmids kindly supplied by Julian Downward and Andrew Chan, were introduced into pGW1. Plasmids were also kindly provided by Silvio Gutkind (pCEFL-m2, pCDNA3-CD8βARKCT), Michael Schneider (pSV-Sport-RasN17), Tom Roberts (pCDNA3-myr-p110α), Robert Kamen (pPyMT1), and Nathalie Sans (peGFP-SAP102).

Cells

Cells were maintained in culture medium (Dulbecco's modified medium supplemented with gentamicin and 6% or 10% fetal bovine serum), which was also supplemented with 20 or 5 μg/ml of puromycin (Sigma) for cells stably transfected with E4-ORF1, PyMT, or Dlg1 shRNA. MUPP1, MAGI-1, and Dlg1 mutant MEF were isolated from 13-day embryos of mice missing the Mpdz (MUPP1) gene (Bell et al, 1995) or having the MAGI-1 gene (in preparation) or the Dlg1 gene specifically disrupted (Caruana and Bernstein, 2001), respectively.

Transfections and treatments

Transfections were performed using TransIT-LT1 (Mirus) or Lipofectamine (Invitrogen Life). Transient experimental analyses were routinely conducted 48 h post-transfection. In some experiments, cells were serum starved for 16–24 h, during which time they either remained untreated or were treated with 20 ng/ml of PDGF-BB (Invitrogen Life), 10 or 50 μM LY294002 (LY) (Cell Signaling Technologies), 100 μM carbachol (Sigma), 50 ng/ml pertussis toxin (PTx) (Sigma), 10 μg/ml chlorpromazine (cpz), or DMSO vehicle as a control.

Antibodies

Rabbit antisera to MUPP1, MAGI-1, ZO-2, SAP97, and E4-ORF1 were described (Javier, 1994; Lee et al, 1997; Glaunsinger et al, 2000; Glaunsinger et al, 2001). Other antibodies were purchased from Cell Signaling Technologies (Akt, phospho-Akt(Ser473), phospho-Akt(Thr308), p42/44 MAPK, phospho-p42/44 MAPK (Thr202/Tyr204)), Santa Cruz Biotechnologies (phospho-tyrosine PY20), BD Biosciences (H-Ras, Dlg1, GFP), Upstate Biotech (p85 PI3K), Covance (HA 16B12), Southern Biotechnology Associates Inc. (horseradish peroxidase-conjugated goat anti-rabbit and anti-mouse IgG), and Molecular Probes (goat anti-rabbit IgG conjugated to either AlexaFluor 594 or AlexaFluor 488).

Extracts, immunoprecipitations, immunoblots, and pulldowns

Cell extracts were prepared in RIPA buffer supplemented with protease and phosphatase inhibitors as described (Lee et al, 1997). For certain experiments, cell extracts in RIPA buffer were fractionated by centrifugation (10 000 g, 5 min at 4°C) into detergent-soluble supernatants and detergent-insoluble pellets, and the latter fractions were resuspended in a volume of sample buffer equal to that of the detergent-soluble supernatant. Equal volumes of each fraction were analyzed. GST pulldown, immunoprecipitation, and immunoblot analyses were carried out as described previously (Weiss et al, 1996).

Soft agar assays

Soft agar assays were conducted as described previously (Javier, 1994) and were photographed with an Olympus DP11 microscope digital camera mounted on an Olympus CK40 inverted microscope.

Microscopy

Indirect IF assays were performed as described previously (Frese et al, 2003). Cells were visualized with a Zeiss Axioplan 2 epifluorescence microscope and photographed with a CoolSnap HQ CCD camera (Photometrics).

Supplementary Material

Supplementary Figure 1

Supplementary Figure 2

Acknowledgments

KKF and IJL were supported by predoctoral fellowships from the NCI (Viral Oncology Training Grant T32 CA09197) or the US Army (Breast Cancer Training Grant DAMD17-94-J4204), respectively. This research was funded by an NIH grant (RO1 CA58541) to RTJ.

References

  • Auger KR, Wang J, Narsimhan RP, Holcombe T, Roberts TM (2000) Constitutive cellular expression of PI 3-kinase is distinct from transient expression. Biochem Biophys Res Commun 272: 822–829 [PubMed]
  • Barmak K, Harhaj E, Grant C, Alefantis T, Wigdahl B (2003) Human T cell leukemia virus type I-induced disease: pathways to cancer and neurodegeneration. Virology 308: 1–12 [PubMed]
  • Bell JA, Rinchik EM, Raymond S, Suffolk R, Jackson IJ (1995) A high-resolution map of the brown (b, Tyrp1) deletion complex of mouse chromosome 4. Mamm Genome 6: 389–395 [PubMed]
  • Caruana G, Bernstein A (2001) Craniofacial dysmorphogenesis including cleft palate in mice with an insertional mutation in the discs large gene. Mol Cell Biol 21: 1475–1483 [PMC free article] [PubMed]
  • Cavatorta AL, Fumero G, Chouhy D, Aguirre R, Nocito AL, Giri AA, Banks L, Gardiol D (2004) Differential expression of the human homologue of Drosophila discs large oncosuppressor in histologic samples from human papillomavirus-associated lesions as a marker for progression to malignancy. Int J Cancer 111: 373–380 [PubMed]
  • Chlenski A, Ketels KV, Engeriser JL, Talamonti MS, Tsao MS, Koutnikova H, Oyasu R, Scarpelli DG (1999a) zo-2 gene alternative promoters in normal and neoplastic human pancreatic duct cells. Int J Cancer 83: 349–358 [PubMed]
  • Chlenski A, Ketels KV, Korovaitseva GI, Talamonti MS, Oyasu R, Scarpelli DG (2000) Organization and expression of the human zo-2 gene (tjp-2) in normal and neoplastic tissues. Biochim Biophys Acta 1493: 319–324 [PubMed]
  • Chlenski A, Ketels KV, Tsao MS, Talamonti MS, Anderson MR, Oyasu R, Scarpelli DG (1999b) Tight junction protein ZO-2 is differentially expressed in normal pancreatic ducts compared to human pancreatic adenocarcinoma. Int J Cancer 82: 137–144 [PubMed]
  • Folco EJ, Liu GX, Koren G (2004) Caveolin-3 and SAP97 form a scaffolding protein complex that regulates the voltage-gated potassium channel Kv1.5. Am J Physiol Heart Circ Physiol 287: H681–H690 [PubMed]
  • Frese KK, Lee SS, Thomas DL, Latorre IJ, Weiss RS, Glaunsinger BA, Javier RT (2003) Selective PDZ protein-dependent stimulation of phosphatidylinositol 3-kinase by the adenovirus E4-ORF1 oncoprotein. Oncogene 22: 710–721 [PMC free article] [PubMed]
  • Fuja TJ, Lin F, Osann KE, Bryant PJ (2004) Somatic mutations and altered expression of the candidate tumor suppressors CSNK1 epsilon, DLG1, and EDD/hHYD in mammary ductal carcinoma. Cancer Res 64: 942–951 [PubMed]
  • Gardiol D, Kuhne C, Glaunsinger B, Lee SS, Javier R, Banks L (1999) Oncogenic human papillomavirus E6 proteins target the discs large tumour suppressor for proteasome-mediated degradation. Oncogene 18: 5487–5496 [PubMed]
  • Glaunsinger BA, Lee SS, Thomas M, Banks L, Javier R (2000) Interactions of the PDZ-protein MAGI-1 with adenovirus E4-ORF1 and high-risk papillomavirus E6 oncoproteins. Oncogene 19: 5270–5280 [PMC free article] [PubMed]
  • Glaunsinger BA, Weiss RS, Lee SS, Javier R (2001) Link of the unique oncogenic properties of adenovirus type 9 E4-ORF1 to a select interaction with the candidate tumor suppressor protein ZO-2. EMBO J 20: 5578–5586 [PMC free article] [PubMed]
  • Gonzalez-Mariscal L, Betanzos A, Avila-Flores A (2000) MAGUK proteins: structure and role in the tight junction. Semin Cell Dev Biol 11: 315–324 [PubMed]
  • Goode S, Perrimon N (1997) Inhibition of patterned cell shape change and cell invasion by Discs large during Drosophila oogenesis. Genes Dev 11: 2532–2544 [PMC free article] [PubMed]
  • Hanada T, Takeuchi A, Sondarva G, Chishti AH (2003) Protein 4.1-mediated membrane targeting of human discs large in epithelial cells. J Biol Chem 278: 34445–34450 [PubMed]
  • Hancock JF (2003) Ras proteins: different signals from different locations. Nat Rev Mol Cell Biol 4: 373–384 [PubMed]
  • Heise A (2003) The clinical significance of HPV. Nurse Pract 28: 8–19 [PubMed]
  • Hirata A, Higuchi M, Niinuma A, Ohashi M, Fukushi M, Oie M, Akiyama T, Tanaka Y, Gejyo F, Fujii M (2004) PDZ domain-binding motif of human T-cell leukemia virus type 1 Tax oncoprotein augments the transforming activity in a rat fibroblast cell line. Virology 318: 327–336 [PubMed]
  • Hough CD, Woods DF, Park S, Bryant PJ (1997) Organizing a functional junctional complex requires specific domains of the Drosophila MAGUK discs large. Genes Dev 11: 3242–3253 [PMC free article] [PubMed]
  • Humbert P, Russell S, Richardson H (2003) Dlg, Scribble and Lgl in cell polarity, cell proliferation and cancer. BioEssays 25: 542–553 [PubMed]
  • Javier R, Raska K Jr, Macdonald GJ, Shenk T (1991) Human adenovirus type 9-induced rat mammary tumors. J Virol 65: 3192–3202 [PMC free article] [PubMed]
  • Javier RT (1994) Adenovirus type 9 E4 open reading frame 1 encodes a transforming protein required for the production of mammary tumors in rats. J Virol 68: 3917–3924 [PMC free article] [PubMed]
  • Kim E, Sheng M (1996) Differential K+ channel clustering activity of PSD-95 and SAP97, two related membrane-associated putative guanylate kinases. Neuropharmacology 35: 993–1000 [PubMed]
  • Kim E, Sheng M (2004) PDZ domain proteins of synapses. Nat Rev Neurosci 5: 771–781 [PubMed]
  • Kiyono T, Hiraiwa A, Fujita M, Hayashi Y, Akiyama T, Ishibashi M (1997) Binding of high-risk human papillomavirus E6 oncoproteins to the human homologue of the Drosophila discs large tumor suppressor protein. Proc Natl Acad Sci USA 94: 11612–11616 [PMC free article] [PubMed]
  • Kohu K, Ogawa F, Akiyama T (2002) The SH3, HOOK and guanylate kinase-like domains of hDLG are important for its cytoplasmic localization. Genes Cells 7: 707–715 [PubMed]
  • Laprise P, Viel A, Rivard N (2004) Human homolog of disc-large is required for adherens junction assembly and differentiation of human intestinal epithelial cells. J Biol Chem 279: 10157–10166 [PubMed]
  • Lee C, Laimins LA (2004) Role of the PDZ domain-binding motif of the oncoprotein E6 in the pathogenesis of human papillomavirus type 31. J Virol 78: 12366–12377 [PMC free article] [PubMed]
  • Lee SS, Glaunsinger B, Mantovani F, Banks L, Javier RT (2000) Multi-PDZ domain protein MUPP1 is a cellular target for both adenovirus E4-ORF1 and high-risk papillomavirus type 18 E6 oncoproteins. J Virol 74: 9680–9693 [PMC free article] [PubMed]
  • Lee SS, Weiss RS, Javier RT (1997) Binding of human virus oncoproteins to hDlg/SAP97, a mammalian homolog of the Drosophila discs large tumor suppressor protein. Proc Natl Acad Sci USA 94: 6670–6675 [PMC free article] [PubMed]
  • Lin HT, Steller MA, Aish L, Hanada T, Chishti AH (2004) Differential expression of human Dlg in cervical intraepithelial neoplasias. Gynecol Oncol 93: 422–428 [PubMed]
  • Liu Y, Wang Y, Yamakuchi M, Masuda S, Tokioka T, Yamaoka S, Maruyama I, Kitajima I (2001) Phosphoinositide-3 kinase-PKB/Akt pathway activation is involved in fibroblast Rat-1 transformation by human T-cell leukemia virus type I tax. Oncogene 20: 2514–2526 [PubMed]
  • Lue RA, Brandin E, Chan EP, Branton D (1996) Two independent domains of hDlg are sufficient for subcellular targeting: the PDZ1-2 conformational unit and an alternatively spliced domain. J Cell Biol 135: 1125–1137 [PMC free article] [PubMed]
  • Lue RA, Marfatia SM, Branton D, Chishti AH (1994) Cloning and characterization of hdlg: the human homologue of the Drosophila discs large tumor suppressor binds to protein 4.1. Proc Natl Acad Sci USA 91: 9818–9822 [PMC free article] [PubMed]
  • Mantovani F, Banks L (2001) The human papillomavirus E6 protein and its contribution to malignant progression. Oncogene 20: 7874–7887 [PubMed]
  • McLaughlin M, Hale R, Ellston D, Gaudet S, Lue RA, Viel A (2002) The distribution and function of alternatively spliced insertions in hDlg. J Biol Chem 277: 6406–6412 [PubMed]
  • Muller BM, Kistner U, Veh RW, Cases-Langhoff C, Becker B, Gundelfinger ED, Garner CC (1995) Molecular characterization and spatial distribution of SAP97, a novel presynaptic protein homologous to SAP90 and the Drosophila discs-large tumor suppressor protein. J Neurosci 15: 2354–2366 [PubMed]
  • Nguyen MM, Nguyen ML, Caruana G, Bernstein A, Lambert PF, Griep AE (2003) Requirement of PDZ-containing proteins for cell cycle regulation and differentiation in the mouse lens epithelium. Mol Cell Biol 23: 8970–8981 [PMC free article] [PubMed]
  • Nix SL, Chishti AH, Anderson JM, Walther Z (2000) hCASK and hDlg associate in epithelia, and their src homology 3 and guanylate kinase domains participate in both intramolecular and intermolecular interactions. J Biol Chem 275: 41192–41200 [PubMed]
  • Rousset R, Fabre S, Desbois C, Bantignies F, Jalinot P (1998) The C-terminus of the HTLV-1 Tax oncoprotein mediates interaction with the PDZ domain of cellular proteins. Oncogene 16: 643–654 [PubMed]
  • Rumbaugh G, Sia GM, Garner CC, Huganir RL (2003) Synapse-associated protein-97 isoform-specific regulation of surface AMPA receptors and synaptic function in cultured neurons. J Neurosci 23: 4567–4576 [PubMed]
  • Sieczkarski SB, Whittaker GR (2002) Dissecting virus entry via endocytosis. J Gen Virol 83: 1535–1545 [PubMed]
  • Suzuki T, Ohsugi Y, Uchida-Toita M, Akiyama T, Yoshida M (1999) Tax oncoprotein of HTLV-1 binds to the human homologue of Drosophila discs large tumor suppressor protein, hDLG, and perturbs its function in cell growth control. Oncogene 18: 5967–5972 [PubMed]
  • Thomas U, Phannavong B, Muller B, Garner CC, Gundelfinger ED (1997) Functional expression of rat synapse-associated proteins SAP97 and SAP102 in Drosophila dlg-1 mutants: effects on tumor suppression and synaptic bouton structure. Mech Dev 62: 161–174 [PubMed]
  • Watson RA, Rollason TP, Reynolds GM, Murray PG, Banks L, Roberts S (2002) Changes in expression of the human homologue of the Drosophila discs large tumour suppressor protein in high-grade premalignant cervical neoplasias. Carcinogenesis 23: 1791–1796 [PubMed]
  • Watson RA, Thomas M, Banks L, Roberts S (2003) Activity of the human papillomavirus E6 PDZ-binding motif correlates with an enhanced morphological transformation of immortalized human keratinocytes. J Cell Sci 116: 4925–4934 [PubMed]
  • Weiss RS, Gold MO, Vogel H, Javier RT (1997) Mutant adenovirus type 9 E4 ORF1 genes define three protein regions required for transformation of CREF cells. J Virol 71: 4385–4394 [PMC free article] [PubMed]
  • Weiss RS, Javier RT (1997) A carboxy-terminal region required by the adenovirus type 9 E4 ORF1 oncoprotein for transformation mediates direct binding to cellular polypeptides. J Virol 71: 7873–7880 [PMC free article] [PubMed]
  • Weiss RS, McArthur MJ, Javier RT (1996) Human adenovirus type 9 E4 open reading frame 1 encodes a cytoplasmic transforming protein capable of increasing the oncogenicity of CREF cells. J Virol 70: 862–872 [PMC free article] [PubMed]
  • Woodhouse E, Hersperger E, Shearn A (1998) Growth, metastasis, and invasiveness of Drosophila tumors caused by mutations in specific tumor suppressor genes. Dev Genes Evol 207: 542–550 [PubMed]
  • Woods DF, Hough C, Peel D, Callaini G, Bryant PJ (1996) Dlg protein is required for junction structure, cell polarity, and proliferation control in Drosophila epithelia. J Cell Biol 134: 1469–1482 [PMC free article] [PubMed]
  • Wu H, Reissner C, Kuhlendahl S, Coblentz B, Reuver S, Kindler S, Gundelfinger ED, Garner CC (2000) Intramolecular interactions regulate SAP97 binding to GKAP. EMBO J 19: 5740–5751 [PMC free article] [PubMed]
  • Zimmermann S, Moelling K (1999) Phosphorylation and regulation of Raf by Akt (protein kinase B). Science 286: 1741–1744 [PubMed]

Articles from The EMBO Journal are provided here courtesy of The European Molecular Biology Organization
PubReader format: click here to try

Formats: