A common C-terminal element stabilizes trimer formation by dUTPase and E4-ORF1

The crystal structure of human dUTPase predicts that NH₂-terminal β-strand 1 (β1) from one subunit and C-terminal β-strand 8 (β8) from an adjacent subunit mediate intermolecular interactions that stabilize trimer formation (Mol et al., 1996). Supporting this prediction and the hypothesis that E4-ORF1 and dUTPase adopt a related protein fold, results from SEC analyses with deletion mutants of dUTPase (ΔC16 and ΔC20) or E4-ORF1 (ΔC5 and ΔC7) demonstrated a requirement for β8-residues in trimer formation by both proteins (Figure 2a, Supplementary Figures S3 and S4).

Open in a separate window

Figure 2

A C-terminal element determines dUTPase and E4-ORF1 trimerization. (a) C-terminal mutants of human dUTPase and E4-ORF1. Gray shaded regions indicate residues needed for trimer formation. Substitution mutations are indicated in bold. E4-ORF1 PBM residues are underlined. (b) β8 mutations abolish E4-ORF1 trimer formation. 293 T cells were transfected with pGW1 encoding the indicated mutant E4-ORF1. Extracts in RIPA buffer were fractionated by SEC with elution buffer containing 0.1% DOC and blotted with E4-ORF1 antibody.

Identical experiments were conducted with four E4-ORF1 mutants having double (AAI, VAA and AKA) or triple (AAA) alanine substitutions in β8-residues VKI (Figure 2a). In SEC analyses, AAI and VAA eluted as trimers, whereas AKA eluted as a mixture of trimers and monomers, reflecting a partial defect in trimer formation (Figure 2b). By contrast, AAA eluted primarily as a monomer and, unlike VAA, failed to self-associate in a co-IP assay (Figure 2b, Supplementary Figure S4b). These results define β8-residues VKI as a crucial E4-ORF1 trimerization (TRI) element.

E4-ORF1 binds cooperatively to PDZ domains in Dlg1 but not MAGI-1 and MUPP1

The existence of E4-ORF1 monomers and trimers in cells might indicate that each form has distinct functions. In this regard, oligomeric ligands that target multi-PDZ domain proteins often bind cooperatively to two closely situated PDZ domains, termed a PDZ tandem (Grootjans et al., 2000; Long et al., 2003). As the E4-ORF1 trimer possesses oligomeric PBM ligands, this feature might uniquely permit a cooperative interaction with the Dlg1 PDZ1 + 2 tandem, which mediates binding to E4-ORF1 (Frese et al., 2006). Supporting this idea, E4-ORF1 co-IP’d wt Dlg1 but not Dlg1 mutants having PDZ1 and PDZ2 dually or individually inactivated (Supplementary Figure S5a).

Unlike the interaction with Dlg1, E4-ORF1 does not target a PDZ tandem in other PDZ proteins but rather binds either one PDZ domain in ZO-2 (PDZ1) and PATJ (PDZ8) or two non-tandem PDZ domains in MUPP1 (PDZ7 and PDZ10) and MAGI-1 (PDZ1 and PDZ3) (Glaunsinger et al., 2000, 2001; Lee et al., 2000; Latorre et al., 2005). While E4-ORF1 would not be expected to bind cooperatively to one PDZ domain, such an interaction is feasible with two non-tandem PDZ domains. Nonetheless, in co-IP assays, wt E4-ORF1 bound HA-tagged wt MAGI-1 but not MAGI-1 having non-tandem PDZ1 and PDZ3 dually deleted (PDZΔ1Δ3), yet retained binding to MAGI-1 having PDZ1 or PDZ3 individually deleted (PDZΔ1 and PDZΔ3), albeit at reduced levels compared to wt MAGI-1 (Supplementary Figure S5b) (Glaunsinger et al., 2000). The amount of E4-ORF1 binding to wt MAGI-1 equaled the sum of binding to PDZΔ1 and PDZΔ3 (Supplementary Figure S5b), revealing non-cooperative binding to these non-tandem PDZ domains. A similar conclusion was drawn for E4-ORF1 binding to non-tandem PDZ domains in MUPP1 (Lee et al., 2000).

These observations led to the hypothesis that E4-ORF1 monomers mediate non-cooperative interactions with individual PDZ domains in MUPP1, MAGI-1, ZO-2 and PATJ whereas E4-ORF1 trimers mediate cooperative interactions with the PDZ tandem in Dlg1.

The E4-ORF1 monomer binds and sequesters MUPP1, MAGI-1 and ZO-2

Our hypothesis predicted that monomeric AAA, as well as partially TRI-defective AKA and TRI-competent AAI and VAA, all of which retain an intact PBM element (see Figure 2a), should display wt binding to MUPP1, PATJ, MAGI-1 and ZO-2. In co-IP assays with epitope-tagged PDZ proteins, these predictions were affirmed for MUPP1 and MAGI-1 (Figure 3a, Supplementary Figure S6), as well as PATJ (unpublished data). All four mutants also exhibited wt proficiency for sequestering endogenous MUPP1 and MAGI-1 within insoluble complexes in CREF cells (Figure 3b), which do not express PATJ (Latorre et al., 2005). By contrast, neither wt nor mutant E4-ORF1 proteins sequestered endogenous Dlg1 (Figure 3b), consistent with previous findings (Frese et al., 2006). Based on results with AAA, we conclude that the E4-ORF1 monomer is sufficient to bind and sequester MUPP1 and MAGI-1 in cells. As will be discussed in a subsequent section, a similar conclusion was drawn for ZO-2, despite its reduced binding to TRI element mutants.

Open in a separate window

Figure 3

E4-ORF1 monomers bind and sequester MUPP1, MAGI-1 and ZO-2. (a) Monomeric AAA binds MAGI-1. 293 T cells were co-transfected with pGW1 encoding HA-MAGI-1 and either pGW1 (vector) or pGW1 encoding wt or the indicated mutant E4-ORF1. Extracts in RIPA buffer were IP’d with HA antibody. Recovered proteins and total cell extracts (Input) were blotted with HA or E4-ORF1 antibody. E4-ORF1 PBM mutant IIIA (see Figure 2a) is a negative control. (b) Monomeric AAA sequesters endogenous MUPP1 and MAGI-1. CREF cells were transduced with pBABE (vector) or pBABE encoding wt or the indicated mutant E4-ORF1. Extracts in RIPA buffer were separated into soluble supernatant (S) and insoluble pellet (I) fractions. Equivalent amounts of each fraction were blotted with MUPP1, MAGI-1, ZO-2, SAP97 (Dlg1) or E4-ORF1 antibody. (c–e) E4-ORF1 monomers bind PDZ domains of MUPP1, MAGI-1 and ZO-2. 293 T cells were transfected with pGW1 encoding wt E4-ORF1 and/or pGW1 encoding (c) HA-MUPP1 PDZ10, (d) HA-MAGI-1 PDZ1, or (e) HA-ZO-2 PDZ1. Extracts in RIPA buffer were fractionated by SEC with elution buffer lacking DOC and blotted with E4-ORF1 or HA antibody.

We also directly tested whether an E4-ORF1 monomer binds the PDZ domains of MUPP1, MAGI-1 and ZO-2. Extracts of 293 T cells expressing HA-tagged MUPP1 PDZ10 (12-kDa), MAGI-1 PDZ1 (14-kDa), or ZO-2 PDZ1 (13-kDa) alone or in combination with 14 kDa wt E4-ORF1 were prepared in RIPA buffer and analysed by SEC with elution buffer lacking DOC, conditions that convert E4-ORF1 into monomers. When expressed alone, E4-ORF1 and MAGI-1 PDZ1 eluted as 12-kDa monomers, whereas MUPP1 PDZ10 and ZO-2 PDZ1 eluted as 26-kDa dimers (Figures 3c–e). When co-expressed, E4-ORF1 and each PDZ domain co-eluted at approximately 27-kDa (Figures 3c–e), similar to the predicted 26–28-kDa masses for heterocomplexes composed of one E4-ORF1 molecule bound to one PDZ domain molecule. Elution sizes of MUPP1 PDZ10 and ZO-2 PDZ1 dimers seemed unchanged by heterocomplex formation as the incorporated E4-ORF1 monomer displaced a PDZ domain having a similar mass. By contrast, identical experiments conducted with SEC elution buffer containing 0.1% DOC, which converts E4-ORF1 into its trimeric form, showed that the E4-ORF1 trimer neither co-elutes with nor changes the elution size of the same PDZ domains (Supplementary Figure S7). Thus, the E4-ORF1 monomer specifically binds the PDZ domains of MUPP1, MAGI-1 and ZO-2.

Two residues in the E4-ORF1 TRI element also influence PBM-mediated binding to ZO-2 and Dlg1

As only the E4-ORF1 monomer binds ZO-2 PDZ1 (see Figure 3e), we predicted that all four TRI element mutants would bind proficiently to ZO-2. Compared to wt E4-ORF1, however, AAI and AKA or VAA and AAA displayed either modest or moderate binding defects, respectively, to HA-tagged ZO-2 (Figure 4a). Our hypothesis additionally predicted that TRI-competent AAI and VAA would exhibit wt binding to Dlg1 whereas partially TRI-defective AKA or monomeric AAA would show a reduced ability or fail to bind Dlg1, respectively. While AKA and AAA yielded anticipated results, AAI or VAA unexpectedly displayed a reduced capacity or failed to interact with HA-tagged Dlg1, respectively (Figure 4b). In CREF cells, TRI element mutants showed corresponding defects in sequestering endogenous ZO-2 within insoluble complexes (Figure 3b) and in binding endogenous Dlg1 (Supplementary Figure S8). In agreement with the fact that the oncogenic potential of E4-ORF1 depends on binding to ZO-2 and Dlg1 (Glaunsinger et al., 2001; Frese et al., 2006), AAI and AKA or VAA and AAA also showed a reduced ability or failed, respectively, to promote phosphorylation of the PI3K-effector protein kinase B (PKB) and focus formation (Supplementary Figure S9).

Open in a separate window

Figure 4

Residues KI in the TRI element additionally control E4-ORFl binding to ZO-2 and Dlg1. (a) E4-ORF1 TRI element mutants have trimerization (TRI)-independent defects in binding ZO-2. 293 cells were co-transfected with pGW1 encoding HA-ZO-2 and either pGW1 (vector) or pGW1 encoding wt or the indicated mutant E4-ORF1. Extracts in RIPA buffer were IP’d with HA antibody. Recovered proteins and total cell extracts (Input) were blotted with HA or E4-ORF1 antibody. (b) E4-ORF1 TRI element mutants have TRI-independent defects in binding Dlg1. Identical experiments as those described in (a) were conducted, except pGW1 encoding HA-Dlg1 was used.

We drew three conclusions from these unexpected results. (1) The observation that TRI-competent VAA exhibited reduced binding to ZO-2 and failed to bind Dlg1 indicates that the mutated KI residues function not only in E4-ORF1 trimer formation but also in enhancing or determining the ability of the adjacent PBM (residues ATLV) (see Figure 2a) to mediate binding to ZO-2 or Dlg1, respectively. Thus, the KI residues dually control both TRI and PBM element activity, redefining the E4-ORF1 PBM element as the expanded six-residue sequence KIATLV. (2) Because KI residues enhance E4-ORF1 binding to ZO-2, this interaction was reduced for TRI-competent VAA and monomeric AAA; however, the fact that these mutants showed comparable weak binding to ZO-2 supports the conclusion that an E4-ORF1 monomer is sufficient to bind this protein. (3) Because KI residues are required for E4-ORF1 to interact with Dlg1, this protein failed to bind both TRI-competent VAA and monomeric AAA, confounding our initial attempt to show that an E4-ORF1 trimer specifically mediates this interaction.

Antibodies

E4-ORF1, MUPP1, MAGI-1, ZO-2 and SAP97/Dlg1 antisera were described (Lee et al., 1997; Weiss et al., 1997; Glaunsinger et al., 2000, 2001; Lee et al., 2000). Antibodies to Dlg1 (BD Transduction Laboratories, San Jose, CA, USA), HA (Covance Research Products, Emeryville, CA, USA), and His (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were purchased, as were HRP-conjugated goat anti-rabbit IgG and goat anti-mouse IgG (Southern Biotechnology Associates, Birmingham, AL, USA).

Cells and extracts

Transfections were performed with TransIT LT1 (Mirus Bio Corporation, Madison, WI, USA). MLV vectors (Soneoka et al., 1995) were produced by co-transfecting 293T cells with pHIT60, pME-VSVG (gift from K Maruyama, DNAX Research Institute), and pBABE-puro encoding E4-ORF1 genes. Cell extracts were prepared in RIPA buffer (50 mm Tris–HCl, pH 8.0, 0.15 m NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) or NETN buffer (20 mm Tris–HCl, pH 8.0, 0.1 m NaCl, 1 mm EDTA, 0.5% NP-40) containing protease and phosphatase inhibitors. Separation of extracts into soluble supernatant and insoluble pellet fractions was done as described (Latorre et al., 2005).

Immunoprecipitations and immunoblots

These assays were conducted as described (Frese et al., 2006).

We thank Andy Rice and Chris Herrmann for use of their FPLC system and Richard Sutton for retroviral vector plasmids. RSW and KKF were supported by predoctoral fellowships from the US. Army (Breast Cancer Training Grant DAMD17-94 J4204) and the National Cancer Institute (Viral Oncology Training Grant T32 CA09197). This research was funded by Public Health Service Grants CA58541 (RTJ) and AI36040 (BVVP).