Efficient ADAM22 surface expression is mediated by phosphorylation-dependent interaction with 14-3-3 protein family members

ADAM22 is a member of a large family of membrane proteins with a disintegrin and metalloproteinase domain (ADAM) (White, 2003; Wolfsberg et al., 1995). ADAM proteins share common structural components and are characterized as containing a pro-, metalloproteinase-like-, disintegrin-, cysteine-rich-, EGF-like, transmembrane- and cytoplasmic-domain (Fig. 1). ADAM proteins were originally found to be involved in sperm-egg membrane fusion (Evans, 2001; Primakoff et al., 1987). They have also been shown to play important roles in muscle development (Gilpin et al., 1998; Yagami-Hiromasa et al., 1995), ectodomain shedding (Blobel, 2005) and regulating neurogenesis and axonal outgrowth (Pan and Rubin, 1997; Sotillos et al., 1997; Yang et al., 2005). The pro-domain of ADAM proteins is removed in early processing shortly after transit through the medial Golgi by a furin-type pro-protein convertase (Lum et al., 1998). Removal of the pro-domain is necessary for ADAM proteins to become catalytically active (Loechel et al., 1999; Milla et al., 1999). Of the over thirty ADAM proteins described to date, only half are predicted to be proteolytically active based on the presence of the consensus sequence HEXXHXXGXXH, that is necessary for zinc-dependent metalloprotease catalytic activity (Jongeneel et al., 1989; Novak, 2004; Stocker et al., 1995) (updated tables are maintained by J. M. White, and T. G. Wolfsberg, and can be found at http://www.people.virginia.edu/∼jw7g/Table_of_the_ADAMs.html).

Catalytically inactive ADAM proteins are thought to function predominantly through their disintegrin and cysteine-rich domains. However, to date little research has been published regarding the molecular function of catalytically inactive ADAM family members. The disintegrin domains of catalytically active and inactive ADAM proteins are ligands for integrins and show high sequence homology to soluble snake venom disintegrins (Evans, 2001). The disintegrin domain of ADAM9 has been shown to competitively bind to integrins and decrease cell adhesion to plates coated with various ECM components (Mahimkar et al., 2005). ADAM proteins studied to date have been found to interact predominantly with integrins α9β1 and α6β1 (White, 2003). We have found recently that ADAM22 interacts with the integrin αvβ3 and also with integrin dimers containing α6 or α9 (D'Abaco et al., 2006).

Some ADAM proteins have sizeable cytoplasmic domains of up to 235 amino acids, often containing tyrosine residues and proline-rich Src homology 3 (SH3) docking sequences. It is thought that ADAM proteins may mediate signaling events via their cytoplasmic domains. Several studies have identified ADAM cytoplasmic domain-binding partners, the majority of which bind to proline-rich SH3 docking sequences. ADAM15 interacts with Src family protein tyrosine kinases and Grb2 (Poghosyan et al., 2002). ADAM12, ADAM15 and ADAM19 interact with Fish (Abram et al., 2003), an adapter protein and Src substrate (Lock et al., 1998), and co-localize at podosome structures in Src-transformed cells. ADAM17 has been shown to interact with MAD2, ADAM9 with MAD2β (Nelson et al., 1999), and ADAM19 with ArgBP1 (Arg binding protein) (Huang et al., 2002). The functional significance of these particular interactions is yet to be fully elucidated. Other proteins that interact with ADAM cytoplasmic proline-rich domains have been shown to regulate ADAM catalytic activity such as PACSIN2 (Tanaka et al., 2004), PACSIN3 (Mori et al., 2003), and Eve-1 (Cousin et al., 2000). In addition, over-expression of the cytoplasmic domain of ADAM9 has been shown to inhibit 12-O-tetradecanoylphorbol-13-acetate (TPA)-induced heparin-binding-EGF shedding by competing for cytoplasmic binding proteins (Izumi et al., 1998). In a similar study, ADAM12 was shown to interact via its cytoplasmic domain with α-actinin-2, an actin cross-linking protein, to exert its effects on cell fusion (Galliano et al., 2000). ADAM17 (TACE) was shown to interact with protein phosphatase H1 (PTPH1) via its carboxyl-terminal PDZ ligand domain. Over-expression of catalytically active forms of PTPH1 decreased ADAM17 levels and decreased PMA-induced shedding of TNF-α, compared to the over-expression of a catalytically inactive PTPH1 (Zheng et al., 2002). Finally, ADAM12 translocation to the cell surface was found to be mediated by its interaction with protein kinase Cϵ (Sundberg et al., 2004).

All of these studies investigated interactions with the cytoplasmic domains of catalytically active ADAM proteins. We set out to identify binding partners of the cytoplasmic domain of ADAM22, one of three catalytically inactive proteins expressed predominantly in the brain (Poindexter et al., 1999; Sagane et al., 1998). Human ADAM22 has a conserved extracellular domain and at least eleven isoforms with distinct cytoplasmic domains resulting from alternate splicing (Sagane et al., 2005) (U.N., unpublished results). The cytoplasmic domains of ADAM22 variants range from 148 to 61 amino acids. ADAM22 variant 1 (ADAM22v1) has the largest cytoplasmic domain. It contains a potential carboxyl-terminal PDZ binding domain, a proline-rich sequence and has six tyrosine residues. The ADAM22 variant 4 (ADAM22v4) sequence is identical to that of ADAM22v1; however, due to differential splicing, an early stop codon results in a cytoplasmic tail that lacks 47 amino acids (Fig. 1) (Harada et al., 2000; Poindexter et al., 1999; Sagane et al., 1998). ADAM22-deficient mice were found to exhibit severe ataxia, suffered marked peripheral hypomyelination and died within 2 weeks of birth. This suggests that ADAM22 is important for the correct functioning of the nervous system, however, the mechanisms mediating these effects are not known (Sagane et al., 2005).

We sought to identify and characterize protein ligands of the cytoplasmic domain of ADAM22 to gain a better understanding of ADAM22 function in the brain. During the course of our studies, reports of ADAM22 interacting with some 14-3-3 proteins were published (Zhu et al., 2005; Zhu et al., 2003). Here, we report the interaction of ADAM22 with all 14-3-3 protein family members expressed in the brain and suggest a novel function for these interactions, where 14-3-3 binding may play a role in the forward transport of ADAM22 to the cell membrane.

We used the yeast two-hybrid system to identify ADAM22 binding proteins. Sequential transformations of pDBLeu-ADAM22v1 cytoplasmic domain and a human brain cDNA library into yeast cells resulted in an estimated 6.2×10⁶ transformants, which were subsequently screened for His3 reporter gene activation. The 23 clones deemed positive for His3 selection were then subjected to second round screening for lacZ reporter activation by X-gal assay, resulting in 13 positive clones. Sequencing of yeast plasmid DNA recovered in E. coli revealed the identity of 10 clones. All but one of these belonged to the 14-3-3 protein family, including isoforms beta (β), zeta (ζ) and tau (τ). The remaining clone consisted of unidentifiable genomic sequence.

Recovered 14-3-3 plasmid DNA or empty library vector pPC86 were re-introduced into yeast with either pDBLeu, pDBLeu-ADAM22v1 cytoplasmic domain or pDBLeu-ADAM22v4 cytoplasmic domain bait constructs. These yeast transformants were then assayed against the three reporter genes His3, lacZ and URA3 to confirm interactions in the yeast two-hybrid system (Fig. 2).

Previously 14-3-3β and 14-3-3ζ had been shown to interact with ADAM22 in the yeast two-hybrid system, whereas 14-3-3τ was shown not to interact (Zhu et al., 2005; Zhu et al., 2003).

We decided to investigate further the range of 14-3-3 isoforms that interact with ADAM22 and to characterize better these interactions in mammalian cells. ADAM22v4 was used for all subsequent studies.

To investigate the range of human 14-3-3 isoforms that interact with ADAM22 we co-expressed the 14-3-3 isoforms as C-terminally Myc- or HA-tagged proteins with full-length FLAG-tagged ADAM22v4 in HEK 293T cells and examined their ability to co-immunoprecipitate. All six human 14-3-3 protein family members investigated were found to co-precipitate with ADAM22 (Fig. 3). Expression of 14-3-3σ is confined mainly to epithelial cells, and therefore was not investigated (Prasad et al., 1992). The precursor form of ADAM22 was the predominant species co-precipitated by all 14-3-3 proteins. The mature form of ADAM22 was also co-precipitated by 14-3-3β, 14-3-3ϵ and 14-3-3η, although to a much lesser extent. High levels of both the precursor and mature forms of FLAG-tagged ADAM22 were detected in all HEK 293T cell lysates used in this study (Fig. 5B).

There are two 14-3-3 binding motifs, RSXpSXP and RXY/FXpSXP, present in almost all known 14-3-3 interacting proteins (Yaffe et al., 1997). Analysis of the cytoplasmic amino acid sequence of ADAM22 revealed two potential binding consensus motifs RSNSWQ [amino acids (aa)831-836] and RSNSTE (aa854-859; Fig. 1). Although both sites lack a proline at +2, recognition of the consensus sites is still likely to occur because an alanine substitution at +2 in synthetic peptide binding studies shows only a moderate effect on 14-3-3 binding (Muslin et al., 1996). 14-3-3 proteins often function as homo- or heterodimers once phosphorylated (Jones et al., 1995). The dimeric structure of 14-3-3 proteins allows them to bind at two sites simultaneously. To investigate the importance of each 14-3-3 protein binding motif we mutated both binding sites separately or together or deleted the C-terminal domain containing both binding sites by the introduction of an early stop codon after aa830. Co-immunoprecipitation studies were carried out using lysates of HEK 293T cells, expressing Myc-tagged 14-3-3β and FLAG-tagged ADAM22v4 or FLAG-tagged ADAM22v4 mutants (Fig. 4A). Co-immunoprecipitation of ADAM22 by 14-3-3β was drastically reduced when the first 14-3-3 binding site was mutated. Mutation of the second 14-3-3 binding site also impeded co-immunoprecipitation, although to a much lesser extent. Finally, mutation or deletion of both sites abolished co-immunoprecipitation of ADAM22 and 14-3-3β.

Next, we performed GST-14-3-3ζ fusion protein pull-down assays using HEK 293T cell lysates expressing the same ADAM22 wild-type and mutant proteins (Fig. 4B). Mutations of the first binding site similarly reduced the ability of GST-14-3-3ζ to pull down the ADAM22 precursor form. Mutation of the second site had very little effect, whereas double mutation or deletion of both binding sites abolished binding to the GST-14-3-3ζ protein.

As previously shown by co-immunoprecipitation, the precursor form of ADAM22 was also found to interact preferentially in a GST-14-3-3ζ fusion protein pull-down assay.

This finding suggests that the interaction between ADAM22 and 14-3-3 proteins may occur at the earlier stages of ADAM22 processing. In addition, in vitro binding assays using GST fusion proteins containing the cytoplasmic domain of ADAM22 protein failed to precipitate Myc-tagged 14-3-3 proteins (data not shown). These results indicate that mammalian post-translational processing or conformational changes may play a crucial part in the binding of 14-3-3 proteins to ADAM22. We hypothesized, on this basis, that the interaction between ADAM22 and 14-3-3 proteins may require phosphorylation of ADAM22.

Finally, we assessed the double 14-3-3 binding site mutant of ADAM22 in the yeast two-hybrid system and clearly demonstrated by three reporter assays that mutations at both 14-3-3 binding sites disrupt any interaction between ADAM22 and 14-3-3 proteins (Fig. 4C).

14-3-3 protein interactions are often dependent on serine phosphorylation of the target protein. However, 14-3-3 protein binding of nonphosphorylated targets has been reported previously (Masters et al., 1999), and can occur on nonphosphorylated RSXS consensus binding motifs (Campbell et al., 1997).

To determine whether phosphorylation of either ADAM22 or 14-3-3 proteins was important for an interaction to occur, separate HEK 293T cell lysates containing either FLAG-tagged ADAM22, HA-tagged 14-3-3τ or other Myc-tagged 14-3-3 proteins were treated with alkaline phosphatase to dephosphorylate proteins prior to mixing and co-immunoprecipitation (Fig. 5A). Dephosphorylation of ADAM22 significantly reduced binding of all 14-3-3 isoforms. By contrast, dephosphorylation of 14-3-3 proteins alone did not impact on their ability to co-immunoprecipitate ADAM22.

We next investigated serine phosphorylation levels of the precursor and mature forms of ADAM22 using a phosphoserine antibody that recognizes phosphoserine consensus motifs similar to the 14-3-3 binding motifs of ADAM22 (Fig. 5C). This antibody has been used to detect the phosphorylation state of other 14-3-3 protein substrates in previous studies (Kovacina et al., 2003; Liu et al., 2002). The precursor form of ADAM22 showed significant levels of serine phosphorylation, whereas the mature form had only minimal levels, suggesting serine phosphorylation at 14-3-3 consensus binding motifs occurs in the early processing of ADAM22 and is subsequently reversed.

In several cases 14-3-3 binding has been shown to increase stability of target proteins (Jeanclos et al., 2001; Wang et al., 2004; Zheng et al., 2005). HEK 293T cells transiently expressing either FLAG-tagged ADAM22 wild-type or the double 14-3-3 binding site mutant were treated with puromycin for up to 8 hours to inhibit protein synthesis. Lysates of these cells were probed with α-FLAG antibody and indicate some difference in ADAM22 protein stability (Fig. 6A). Densitometric analysis of ADAM22 precursor bands, as normalized to β-tubulin, confirms that the ADAM22 double substitution mutant that can no longer bind 14-3-3 protein is more rapidly cleared than wild-type ADAM22 (Fig. 6B).

To assess the expression of endogenous 14-3-3 isoforms present in the cell lines used in this study we performed RT-PCR using gene-specific primers to amplify full-length transcripts of all 14-3-3 protein isoforms (Fig. 6C). Transcripts of 14-3-3β, 14-3-3τ, 14-3-3ϵ and 14-3-3η were detected in D645 cells, whereas transcripts of 14-3-3β, 14-3-3τ, 14-3-3γ, 14-3-3ϵ, 14-3-3η and 14-3-3ζ were detected in HEK 293T cells.

14-3-3 protein binding can modulate target protein localization (Muslin and Xing, 2000). To assess the effect of 14-3-3 protein binding on ADAM22 subcellular localization, we expressed FLAG-tagged ADAM22 proteins in D645 glioma cells and performed immunofluorescence using α-FLAG antibodies. D645 cells expressing ADAM22 with intact 14-3-3 protein binding sites exhibited distinct membrane staining (Fig. 7A,G, Fig. 8A,G). This membrane staining was strongly reduced when ADAM22 mutants were expressed, which had substitutions at both 14-3-3 binding sites, and as such, could no longer interact with 14-3-3 proteins (Fig. 7B,H, Fig. 8B,H). Immunostaining of cells with both Golgi and ER markers demonstrated co-localization with the ADAM22 14-3-3 binding site mutant (Figs 7, 8). GM130 is a Golgi matrix protein, which is associated with the cis-compartment, whereas p230 is associated with the trans-Golgi network. Calnexin and calreticulin are molecular chaperones and play important roles in glycoprotein folding within the ER. These proteins have been used extensively as Golgi and ER markers in numerous studies.

These results suggest that the interaction between ADAM22 and 14-3-3 proteins is necessary for efficient ADAM22 translocation to the cell surface. Recent papers have demonstrated that interactions with 14-3-3 proteins can override ER localization signals (O'Kelly et al., 2002; Shikano et al., 2005; Yuan et al., 2003). Di-arginine motifs cause retention of the protein in the ER and can inhibit transport to the cell surface until the localization signal is overridden by various cellular processes. Inspection of the ADAM22 sequence revealed the presence of three di-arginine RXR motifs overlapping both 14-3-3 protein binding sites (Fig. 1). Expression of an ADAM22 C-terminal mutant resulting in deletion of both 14-3-3 protein binding sites and ER localization motifs restored ADAM22 protein staining in the membrane (Fig. 7C,I, Fig. 8C,I).

Next, we carried out cell surface biotinylation assays on D645 cells expressing ADAM22 wild-type and mutant proteins to further investigate the observation by immunolocalization studies that ADAM22 mutants lacking the ability to bind 14-3-3 proteins localize poorly to the cell surface. Accordingly, the ADAM22 mutant of the first and second 14-3-3 consensus binding motifs was cell surface biotinylated at significantly lower levels than both ADAM22 wild-type and C-terminal-deletion mutant proteins, indicating less efficient transport to the cell surface (Fig. 9).

14-3-3 proteins are well conserved and although widely expressed, are notably abundant in the brain where they represent 1% of total soluble protein (Boston et al., 1982). 14-3-3 proteins have been found to play diverse roles in almost every cellular process, most notably in regulating cell cycle (Davezac et al., 2000; Peng et al., 1997), cell death (Zha et al., 1996) and mitogenic signaling (Fu et al., 2000). Interestingly, they also play important roles in brain development and disease (Chen et al., 2003; Layfield et al., 1996; Toyo-oka et al., 2003; Ubl et al., 2002), adhesion (Garcia-Guzman et al., 1999; Han et al., 2001; Santoro et al., 2003), proliferation (Rosner and Hengstschlager, 2006) and cancer (Hermeking, 2003; Sugiyama et al., 2003).

Although 14-3-3 proteins function as homo- and heterodimers and can bind two separate ligands simultaneously, only very few of the >250 interactions described to date are found to act as intermolecular bridges (Mackintosh, 2004). Of particular exception are the complexes of Raf-1 with Bcr (Braselmann and McCormick, 1995), A20 (Vincenz and Dixit, 1996) and PKCζ (Van Der Hoeven et al., 2000).

14-3-3 proteins commonly bind to RSXpSXP or RXY/FXpSXP motifs, present in almost all known 14-3-3 interacting proteins (Yaffe et al., 1997). The ADAM22 cytoplasmic domain has two consensus motifs RSNSWQ (aa831-836) and RSNSTE (aa854-859). These sites are found in all known ADAM22 splice variants. Many 14-3-3 ligands have been shown to have multiple binding motifs, such as Cdc25C (Zeng et al., 1998), 3BP2 (Foucault et al., 2003), CBL (Zha et al., 1996), BAD (Zha et al., 1996), Raf-1 (Dumaz and Marais, 2003; Muslin et al., 1996) and FKHRL family members (Obsil et al., 2003). Often 14-3-3 binding is dependent on one or two high affinity sites and additional low affinity sites that may act to stabilize and enhance the interaction. This agrees with the `gatekeeper' model for 14-3-3 ligand binding (Yaffe, 2002). Indeed, 14-3-3 dimers can bind tandem sites on the same peptide with >30-fold higher affinity than a single 14-3-3 binding motif (Yaffe et al., 1997). Such high affinity binding by 14-3-3 to multiple binding sites on target sequences could cause ligand conformational change, thus exposing/masking regions that are otherwise inaccessible in the free ligand form. Here, we show the first 14-3-3 binding site of ADAM22 to be the most crucial for 14-3-3 binding by both co-immunoprecipitation and GST pull-down assays. However, the presence of both sites drastically improved 14-3-3 protein binding.

The majority of 14-3-3 interactions are dependent on phosphorylation of their target sequences (Pozuelo Rubio et al., 2004). Nevertheless, some phosphorylation-independent interactions with 14-3-3 proteins also occur. These include 14-3-3 binding to exoenzyme S, p190RhoGEF and the R18 peptide inhibitor (Masters et al., 1999; Wang et al., 1999; Zhai et al., 2001). Furthermore, it has been demonstrated that dephosphorylation can create a 14-3-3 binding motif (Waterman et al., 1998). We find the interaction between ADAM22 and 14-3-3 proteins to be dependent on the phosphorylation of ADAM22, but not 14-3-3 proteins. In addition, although we found some 14-3-3 proteins can bind to both the precursor and mature forms of ADAM22, the interaction preferentially occurs with the precursor form of ADAM22. This led us to believe that the interaction between ADAM22 and 14-3-3 proteins may take place in the early processing of ADAM22 protein, not at the cell surface. We found that the proteolytically processed, mature form of ADAM22 is less serine-phosphorylated, which suggests that serine phosphorylation during early processing of ADAM22 is the likely trigger for 14-3-3 binding prior to ADAM22 protein maturation.

The binding of 14-3-3 proteins to their target sequences has often been shown to increase target stability (Jeanclos et al., 2001; Wang et al., 2004; Zheng et al., 2005). We find that wild-type ADAM22 is more stable than a mutant form unable to bind 14-3-3 proteins.

We next investigated the effects of 14-3-3 binding on the localization of ADAM22. ADAM22 substitution mutants that could no longer bind to 14-3-3 proteins were no longer transported efficiently to the cell surface but were retained in the ER and Golgi. The relationship between 14-3-3 binding and forward transport of proteins is becoming more evident (Nufer and Hauri, 2003). 14-3-3 protein binding can override RXR motifs and contribute to post-Golgi sorting to the cell membrane. This has been clearly demonstrated with the HIV co-receptor GPR15 (Shikano et al., 2005), and also the potassium channels KCNK3 (O'Kelly et al., 2002; Shikano et al., 2005), and K_atp α subunit (Yuan et al., 2003). More recently, identification and characterization of a mode III 14-3-3 protein binding motif (SWTY) has further highlighted the important role 14-3-3 proteins may play in overriding ER-Golgi localization signals in membrane receptors (Coblitz et al., 2005). ADAM22 has three possible RXR di-arginine motifs located on or between both 14-3-3 binding sites (aa829-831, aa850-852, aa852-854). Removal of both the 14-3-3 binding sites and the di-arginine retention motifs of ADAM22 restored the ability of ADAM22 protein to accumulate efficiently at the membrane.

Our results suggest ADAM22 and 14-3-3 proteins interact to regulate ADAM22 transport to the membrane and that this interaction is initiated by serine phosphorylation of the precursor form of ADAM22. Previous reports show ADAM22 and 14-3-3 proteins co-operating to regulate cell adhesion and spreading on fibronectin-coated plates (Zhu et al., 2005; Zhu et al., 2003). In our attempts to replicate these results we used stably transfected cells that had been dislodged with EDTA rather than trypsinized prior to adhesion assays and could not confirm that ADAM22 over-expression increases cell adhesion to fibronectin-coated plates (data not shown). We found all HEK 293T cells adhere to fibronectin-coated plates within 10 minutes, irrespective of ADAM22 expression. Significant levels of α5 integrins, receptors for fibronectin (Argraves et al., 1986; Pytela et al., 1985), are endogenously expressed in HEK 293T cells (Li et al., 2001). In addition, we have demonstrated previously that ADAM22 does not interact with fibronectin receptors (D'Abaco et al., 2006). Our current data does not support previous reports that ADAM22 and 14-3-3 protein interactions can regulate fibronectin-mediated cell adhesion. Instead, we propose that 14-3-3 protein binding is an important event for ADAM22 processing and forward transport to the membrane.

The full length C-terminally HA-tagged ADAM22v1 and FLAG-tagged ADAM22v4 pcDNA3.1 mammalian expression constructs were generously provided by Dr Sagane (Tsukuba Research Laboratories, Eisai Co, Ibaraki, Japan), and have been described previously (Sagane et al., 1998). Point mutations in the cytoplasmic domain of the pcDNA3.1 ADAM22v4 mammalian expression vector were generated using the QuickChange II XL site-directed mutagenesis kit (Stratagene) as instructed. These included substitution mutants at the first 14-3-3 binding site (R831Q/S832G), the second 14-3-3 binding site (R854Q/S855A) and double mutations at both 14-3-3 consensus sites. Similarly, a C-terminal deletion mutant (ΔC-term), was made by in vitro mutagenesis, linking an 830 to the FLAG-tag.

To clone bait constructs for use in the yeast two-hybrid system, cDNA fragments of ADAM22v1 cytoplasmic domain (aa760-906), ADAM22v4 cytoplasmic domain (aa760-860) or the ADAM22v4 cytoplasmic domain with substitution mutants of the first and second 14-3-3 consensus binding motifs were cloned into the SalI/NcoI sites of pDBLeu (Invitrogen).

The cDNA sequences of 14-3-3 isoforms β, ϵ and η were amplified from a Proquest pPC86 human brain cDNA library (Invitrogen). The full length sequences were cloned into the AscI site of pEF-MYC-1 (a gift from Warren Alexander, Walter and Eliza Hall Institute, Melbourne, Australia) for C-terminal Myc-tagged expression. Mammalian expression pcDNA3 constructs encoding Myc-tagged 14-3-3γ, Myc-tagged 14-3-3ζ and HA-tagged 14-3-3τ were a gift from Alastair Aitken, University of Edinburgh, UK.

Full-length 14-3-3ζ was cloned into pGEX-4T1 for GST-fusion protein expression.

A ProQuest human brain cDNA library, which is maintained in the yeast two-hybrid vector pPC86, was screened essentially as described in the ProQuest manual (Invitrogen). Briefly, the yeast reporter strain MaV203 was transfected with a pDBLeu-ADAM22v1 cytoplasmic domain construct. Yeast expressing the pDBLeu-ADAM22v1 cytoplasmic domain construct was then also transformed with the cDNA library and plated on synthetic complete medium plates lacking leucine, tryptophan and histidine and containing 30 mM 3-amino-1,2,4-triazole (3AT). Clones deemed positive for histidine selection in the primary screen were further screened for URA3 and lacZ reporter expression. Plasmid DNA from positive interacting clones was extracted as recommended and sequenced. Plasmid DNA was then re-transformed into yeast for confirmation by repeat reporter assays.

The human D645 glioma cell line (a gift from David Ashley, Royal Children's Hospital, Melbourne, Australia) and the human embryonic kidney (HEK) 293T cell line were propagated in DMEM in the presence of 10% fetal bovine serum. Cells were transfected using Lipofectamine 2000 (Invitrogen).

For co-immunoprecipitation studies, HEK 293T cells, grown in six-well plates, were transfected with 1 μg of plasmid DNA for expression of FLAG-tagged ADAM22v4 and 1 μg of the different Myc- or HA-tagged 14-3-3 isoform plasmids. The following day, cells were lysed in cell lysis buffer (0.15 M NaCl, 10 mM Tris-HCl pH 7.4, 1% Triton X-100, 1 mM EDTA, 0.5 mM orthovanadate, 0.5 μg/ml leupeptin, 300 U/ml aprotinin) and the 14-3-3 proteins were immunoprecipitated using anti-Myc or anti-HA antibodies. The immunoprecipitates were electrophoresed on 7.5% acrylamide/SDS gels, transferred onto Hybond C-super membrane (Amersham Biosciences) and probed with M2 anti-FLAG antibody (Sigma).

To evaluate the effects of protein dephosphorylation on ADAM22 and 14-3-3 protein interactions, separate HEK 293T cells were prepared expressing FLAG-tagged ADAM22v4, HA-tagged 14-3-3τ or other Myc-tagged 14-3-3 proteins and were lysed in cell lysate buffer lacking sodium vanadate before treatment with alkaline phosphatase (Promega) as indicated. Sodium vanadate was then added to samples prior to incubation with the corresponding cell lysates. Pooled lysates were immunoprecipitated with anti-HA or anti-Myc antibodies and immunoblots probed with anti-FLAG antibody.

Serine phosporylation of ADAM22 was analysed by western blotting of anti-FLAG immunoprecipitates from lysates of HEK 293T cells transfected with FLAG-tagged ADAM22v4 and probed with anti-phospho-(Ser) Akt substrate antibody (Cell Signaling). This antibody recognizes motifs containing phosphoserine preceded by arginines at the -2 and -5 positions as is present at both 14-3-3 consensus binding sites of ADAM22.

The 14-3-3ζ cDNA was cloned into the pGEX4T-1 vector (Amersham Pharmacia Biotech) and GST fusion protein was expressed in Escherichia coli BL21 (Stratagene) by the addition of 1 mM isopropyl β-D-thiogalactopyranoside to the culture for 3 hours. Bacteria were harvested and lysed by sonication. Insoluble material was removed by centrifugation at 16,000 g. Soluble GST fusion protein was then bound to glutathione-Sepharose beads (Amersham Pharmacia Biotech), washed five times for 3 minutes in PBS containing 0.25% Triton X-100 and mixed with lysates of ADAM22v4 expressing HEK 293T cells for 1 hour at room temperature on a rocker. The beads were then pelleted, washed three times with lysis buffer, boiled in SDS-sample buffer and loaded onto a 7.5% acrylamide/SDS gel, subjected to electrophoresis, blotted onto Hybond C-super membrane (Amersham Biosciences) and probed with anti-FLAG antibodies.

To assess if 14-3-3 binding to ADAM22 had any effect on the stability of the ADAM22 protein, we subjected HEK 293 cells, transiently transfected with either wild-type FLAG-tagged ADAM22v4 or the double 14-3-3 binding site substitution mutant to incubation for 0-8 hours with puromycin (Sigma), which blocks protein synthesis. Cells were then lysed and analysed by gel electrophoresis and probed using anti-FLAG antibodies. Blots were re-probed with anti-β-tubulin (Sigma) antibody as loading control. Densitometry readings for ADAM22 precursor and β-tubulin bands were made using SynGene Gene tools software (SynGene Laboratories, Cambridge, UK). ADAM22 precursor bands were normalized to β-tubulin and expressed as percentage protein of the wild-type control at the corresponding time points.

Total RNA from D645 and HEK 293T cells was extracted using an SV total RNA isolation kit (Promega). 2 μg of total RNA were reverse transcribed in a 20 μl reaction using M-MLV reverse transcriptase (Promega) and an oligo(dT₂₀) primer as instructed. Standard PCR reactions were carried out using 1 μl cDNA template and sets of 20-25mer primers located at the initiation and terminal coding sequences of all 14-3-3 protein isoforms. The complete coding sequences were amplified under the following PCR cycling conditions: 94°C for 30 seconds, 50°C for 30 seconds and 72°C for 45 seconds for 30 cycles and analysed on an agarose gel.

D645 cells were plated on round 12 mm glass coverslips (Menzel-Glaser, Lomb Scientific, Taren Point, Australia) and transfected with 1 μg pcDNA3.1 ADAM22 constructs to express FLAG-tagged ADAM22v4, ADAM22v4Δc-term or ADAM22v4 double 14-3-3 binding site mutants. Twenty-four hours after transfection the cells were stained and examined for immunofluorescence as described previously (D'Abaco et al., 2006). FLAG-tagged proteins were detected with either rabbit or M2 mouse anti-FLAG antibodies (Sigma) at a 1:1000 dilution. In addition, co-staining with either the Golgi markers p230 and GM130 (BD Biosciences), or the endoplasmic reticulum (ER) markers anti-calnexin and anti-calreticulin (Stressgen Bioreagents, Ann Arbor, MI), were performed at dilutions of 1:500.

D645 cells were transfected with ADAM22 constructs to express FLAG-tagged ADAM22v4, ADAM22v4Δc-term, or ADAM22v4 double 14-3-3 binding site mutants as described previously. 24 hours after transfection cells were washed three times in PBS and treated with 2.5 μg/ml sulfo-NHS-LC-biotin (Pierce) for 5 minutes at room temperature. The reaction was quenched at 4°C in PBS plus 25 mM lysine hydrochloride for 15 minutes, washed three times in PBS and lysed in cell lysis buffer. Biotinylated proteins were isolated by pull down with streptavidin agarose (Sigma) for 1 hour, washed three times in cell lysis buffer and analysed by gel electrophoresis and immunoblotting with anti-FLAG antibodies.

The authors are grateful to Steve Cody and Eva Tomaskovic-Cook for discussion and advice with regard to immunofluorescence studies and Melissa Inglese for protein biotinylation protocols. We also wish to thank Lyndal Kerr-Bayles and Greg Collier for generous advice for setting up yeast two-hybrid assays. This work was supported by funds from the John T. Reid Charitable Trusts and The Friends of The Royal Melbourne Hospital Neuroscience Foundation.