Binding to Syntenin-1 Protein Defines a New Mode of Ubiquitin-based Interactions Regulated by Phosphorylation^*

Cell Lines and Reagents

HEK293T and HeLa cells were transfected with plasmids using polyethyleneimine (Sigma) and cultured in DMEM (Invitrogen) supplemented with 10% FCS (Invitrogen). The cells were then harvested and the lysates used for Western blotting or immunoprecipitation. Plasmids used to express syntenin-1 included pcDNA3-syntenin-1(WT)-HA, pcDNA3-syntenin-1ΔN, and syntenin-1ΔC as described previously (3). The FLAG^TM-tagged syntenin-1WT and point mutants were generated using PCR or QuikChange site-directed mutagenesis (Stratagene). The Myc-tagged Ulk1WT and kinase-dead Ulk1-KD (K46R) mutant DNA constructs were provided by Dr. S. Tooze (London Research Institute, London, UK). The plasmid encoding the glutathione S-transferase (GST)-tagged form of human syntenin-1 was used as reported previously (3), and point mutants were generated. Expression plasmids pET30a containing chemically synthesized His₆-tagged SmpA (control protein), wild-type Ub, or Ub mutant (L71A/L73A/R74Q and I44A) cDNAs inserted between the Nde1-Xho1 site were purchased from GeneScript (Piscataway, NJ). Pure His-Ub mutant proteins (G76A, ΔGG, G75A/G76A, K33R) and His-tagged poly-Ub (Lys⁶³ and Lys⁴⁸) were purchased from Boston Biochem and Enzo Life Sciences, respectively.

Antibodies used included anti-Ub mouse mAb, clone FK2 (Enzo Life Sciences); mouse mAb against CASK (Transduction Laboratories); anti-FLAG epitope mouse mAb (clone M2), and rabbit pAb (Sigma). Rabbit anti-Hrs pAbs were generously provided by Dr. S. Urbe (Liverpool). Rabbit anti-HA Ab (Santa Cruz Biotechnology); rabbit anti-Myc pAb (Cell Signaling Technology); mouse anti-actin mAb (Sigma); and mouse anti-syntenin-1 mAb (Abnova) were also used. Electroporation of HeLa cells was carried out using siRNA 5′-GCUAUAGCAUAGCUGCUUAtt-3′ and Silencer^TM negative control #1 siRNA (Ambion catalogue no. 4611) and assessed for knockdown by Western blot analysis. Mouse mAb against CD63, 6H1, has been previously described (3).

Recombinant Protein Purification

GST-tagged human syntenin-1 and His-SmpA/His-Ub were expressed in Escherichia coli BL21 (DE3), lysed, and affinity-purified using GSTrap or HisTrap FF column (GE Healthcare), followed by size exclusion chromatography using HiLoad Superdex 75 26/60 (GE Healthcare) in 50 mm NaPO₄, pH 7.5, 0.15 m NaCl, 0.5 mm tris(2-carboxyethyl)phosphine buffer. Tobacco etch virus protease was used to remove the GST tag from proteins, prior to size exclusion chromatography.

Immunoprecipitation, Pulldown, and Western Blotting

Immunoprecipitation and pulldown experiments from cellular lysates were performed as described previously (3). In brief, cells were lysed in 1% Triton X-100/PBS containing PMSF (2 mm), leupeptine (10 μg/ml), aprotinin (10 μg/ml), and the lysates were incubated with 15 μl of mono-ubiquitin (mUb) covalently bound to agarose beads (Sigma). Proteins retained by mUb-agarose were eluted from the beads in Laemmli buffer, and the eluate was divided into the appropriate number of equal aliquots, which were resolved by SDS-PAGE. The proteins were transferred to the nitrocellulose membrane and visualized with specific antibodies. In some of the experiments involving cellular lysates the functionality of mAb-agarose beads was additionally tested by adding 3 μg of hHR23A tandem UBA domain, TUBE1 (Boston Biochem), as a “Ub-binding tracer,” to all aliquots of cellular lysates used (supplemental Figs. S4 and S5). For pulldown experiments with recombinant proteins, syntenin-1 was incubated with His-tagged proteins immobilized on nickel-Sepharose beads, and bound proteins were eluted with 50 mm EDTA. The protein samples were resolved using SDS-PAGE and visualized by Coomassie staining or used for Western blotting.

NMR Spectroscopy

Lyophilized untagged ¹⁵N-labeled Ub (ASLA Biotech, Riga, Latvia) was reconstituted to 1 mm in NMR buffer (50 mm NaPO₄ pH 7.5, 150 mm NaCl). NMR spectra were recorded on 100 μm Ub at 30 °C in a Varian Inova 800-MHz spectrometer equipped with a 5 mm ¹H/¹³C/¹⁵N z-PFG cold probe and analyzed using CcpNmr Analysis 2.2. Chemical shift assignments for ¹⁵N-Ub at pH 7.5 were obtained from BMRB entry 4769 and as published (18). The concentration of syntenin-1 was increased stepwise as ¹H,¹⁵N-resolved two-dimensional NMR spectra of ¹⁵N-labeled Ub were collected. Titration was discontinued beyond 1.4-fold excess of syntenin-1 due to extensive line broadening of Ub resonances. The intensities of individual cross-peaks in free Ub versus the Ub-syntenin-1 complex were normalized, and decreases in cross-peak intensities were mapped onto the Ub structure. Molecular graphics images were produced using the Chimera package (Computer Graphics Laboratory, University of California, San Francisco).

Immunofluorescence Staining

HeLa cells grown on glass coverslips were transfected with appropriate DNA plasmids using GeneJammer (Stratech). Staining was performed 40–48 h after transfection. Cells were fixed with 2% paraformaldehyde/PBS for 10–15 min. The staining with primary and fluorochrome-conjugated secondary Abs was carried out as described previously (19). The images were captured using Zeiss LSM510 META confocal system with 63× oil immersion objective (NA 1.4). The images were captured using Zeiss LSM510 META confocal system with 63× oil immersion objective (NA 1.4). Co-localization was quantified using the “Profile” function of Zeiss software.

Syntenin-1 Links Transmembrane Partners to Ubiquitylated Proteins

We investigated whether syntenin-1 may control trafficking of its transmembrane partners via ubiquitylation-dependent pathways. Pulldown assays with immobilized syndecan-2, TGFα, and CD63 C-terminal peptides (“C-tails”) on HeLa cell lysates specifically retained ubiquitylated protein species (Fig. 1A and supplemental Fig. S1). Importantly, the signals for all three peptides were diminished in lysates from syntenin-1 knockdown cells. In the control experiments we found that the peptide corresponding to the C-terminal cytoplasmic region of the tetraspanin Tspan-12, which does not bind syntenin-1, did not retain ubiquitylated proteins (Fig. 1A, lane 5). This indicates that syntenin-1 can link a diverse range of transmembrane partners with ubiquitylated proteins.

We then tested whether any of the C-tail-bound proteins correspond to ubiquitylated forms of syntenin. Pulldown and immunoprecipitation experiments revealed a characteristic ladder of ubiquitylated species of syntenin-1, suggesting its poly- or multiubiquitylation (Fig. 1B). Notably, the most abundant of the ubiquitylated syntenin-1 molecules were also detected in the CD63 pulldown lane. Moreover, the syntenin-1 immunoprecipitate contained ubiquitylated proteins that are clearly distinct from the ubiquitylated syntenin-1 species (Fig. 1B). Thus, syntenin-1 interacts with a set of ubiquitylated proteins which it links to transmembrane partners, forming Ub-based molecular hubs.

Syntenin Binds Directly to Ubiquitin

Binding assays showed that agarose-conjugated mUb specifically retained endogenous syntenin from the lysates of HeLa cells, whereas CASK, an unrelated PDZ domain-containing protein, did not bind Ub (Fig. 2A). Interestingly, the interaction of Ub with Hrs, a protein with a well recognized double-sided Ub-interacting motif (20), was somewhat weaker (Fig. 2A). These results indicate that syntenin-1 (or one of its partners) interacts with Ub. To analyze whether interaction between syntenin-1 and Ub is direct we tested the binding of Ub to purified recombinant syntenin-1 protein. As illustrated in Fig. 2B, syntenin-1 binds to immobilized mUb in pulldown assays (lanes 2 and 3). Importantly, syntenin-1 interacted equally well with Lys⁴⁸- or Lys⁶³-linked poly-Ub chains (Fig. 2B, lanes 6–9). Together, this shows that full-length syntenin-1 binds directly and tightly to Ub, including its polyubiquitylated forms.

Syntenin-1 Contacts the C Terminus of Ubiquitin

The Ub binding pocket for full-length syntenin-1 was mapped by NMR spectroscopy. Significant and progressive line broadening of a set of cross-peaks in ¹H,¹⁵N-resolved two-dimensional spectra of Ub was seen with the addition of increasing amounts of unlabeled syntenin-1 (Fig. 3A). This indicated the intermediate exchange of Ub residues between free and complexed states with syntenin-1. The interface was mapped by monitoring Ub resonances which exhibited decreased intensities upon syntenin-1 binding. As seen in Fig. 3B, syntenin-1 binds to a well defined region spanning the C terminus and centered on Arg⁷², Leu⁷³, and Arg⁷⁴, resulting in severe decrease in the cross-peak intensities. In addition, Leu⁸, Gly¹⁰, and Thr¹² also exhibited significant intensity reduction. Other regions exhibiting decreased peak intensities involved residues Lys³³, Ile³⁶, Gln⁴⁰, Gly⁴⁷, and Leu⁷¹, all nearby and circumscribing a novel C-terminal binding site. Apart from Leu⁸ there was very minimal perturbation of the conservative hydrophobic pocket spanning Ile⁴⁴ and Val⁷⁰. The extreme C terminus of Ub was not involved in syntenin-1 binding because no significant shift or broadening of the resolved Gly⁷⁶ resonance was induced during NMR titration. Although Gly⁷⁵ could not be observed by NMR under the pH conditions employed in this study, it was not required based on the mutational analysis as shown below. The entire HSQC spectral comparison of syntenin-bound and free Ub is shown in supplemental Fig. S2. The affinity (K_D) of the Ub-syntenin-1 interaction was 27.3 μm, as measured by isothermal calorimetry (supplemental Fig. S3). This value is consistent with the intermediate exchange kinetics apparent by NMR, and is relatively tight compared with most ubiqutin binding domains (21).

The Ub binding site for syntenin-1 was validated by mutagenesis. Although the wild-type syntenin-1 exhibited compromised binding to the Ub^LLR/AAQ Ub mutant (Leu⁷¹ to Ala, Leu⁷³ to Ala, and Arg⁷⁴ to Gln substitutions) (Fig. 3C, lane 8), binding of syntenin-1 to three different C-terminal glycine mutants including Ub^G76A, Ub^G75A,G76A (GG/AA mutant), and Ub^ΔG75,G76 (ΔGG deletion mutant) was not affected in the pulldown assays (Fig. 3C, lanes 4–7). Although the resonances of the Ub Ile⁴⁴ residue showed no perturbation, the Ile⁴⁴ to Ala mutant exhibited reduced binding to syntenin-1 (Fig. 3C, lane 9). We speculate that this could be due to an indirect effect caused by a local disruption of the hydrophobic pocket around Ile⁴⁴ in the mutant protein. Ub^K33R still bound syntenin-1 (Fig. 3C, lane 7), suggesting that any basic residue will suffice at position 33. Thus, we concluded that residues 71–74 represent the core of the syntenin recognition site, with residues 8–12 (but not the C-terminal glycine residues) providing additional stabilizing interactions.

Syntenin Dimerization Is Essential for Ubiquitin Interaction

To identify the region of syntenin-1 recognized by Ub we initially tested the N- and C-terminal deletion mutants of syntenin-1 (synt-ΔN and synt-ΔC). Pulldown assays using cellular lysates containing these truncated syntenin-1 proteins revealed that the presence of the first 101 and last 25 residues were critical for Ub interaction (Fig. 4A, lanes 6–8). The requirement for both N and C termini for the syntenin-1 Ub interaction suggests their spatial juxtaposition during Ub binding. Head-to-tail dimerization could yield such proximity of the termini. Indeed, the crystal structure of the PDZ tandem of syntenin-1 reveals head-to-tail dimers (22), and the native protein dimerizes in cells (23). To investigate whether syntenin-1 protein dimerization is important for Ub binding, we performed Ub pulldown experiments using a Synt^F195A/L233A mutant (Synt-mono), which is impaired in dimer formation (24). Fig. 4B illustrates that the Synt-mono was not retained by mUb-agarose (lane 6). Together, this indicates that dimerization of syntenin-1 is critical for Ub binding.

Discovery of the Syntenin Motif Recognized by Ub

We observed that the syntenin-1 N terminus contains three conserved LYPXL sequences centered at positions 5, 47, and 51, respectively, where X is any residue. To investigate their role in Ub interaction, we examined Ub binding by syntenin-1 mutants in which each central proline was mutated to Ala. Mutation of Pro⁵ to Ala (Synt P5A) completely abolished binding of syntenin-1 to Ub in pulldown experiments with cellular lysates (Fig. 4C, lane 8) or with recombinant proteins (Fig. 4E, lane 3). By contrast, P47A and P51A substitutions had no detectable effects on the interaction (Fig. 4C, lanes 9 and 10). Accordingly, immunofluorescence experiments reveal that the co-localization of Ub puncta with the Synt P5A mutant was severely compromised (Fig. 5).

The first LYPXL motif of syntenin-1 is unique in containing a conserved Ser residue at the X position. To investigate the role of Ser⁶, we generated Synt S6A and Synt S6E mutants. Either mutation completely abolished the interaction of syntenin-1 with Ub (Fig. 4D, lanes 7 and 8). Results from the pulldown assays using recombinant syntenin-1 mutant proteins further corroborated this observation (Fig. 4E, lanes 4 and 5). Accordingly, Ser⁶ was important for the recruitment syntenin-1 to ubiquitylated proteins in cells (Fig. 5). Together, this establishes the L⁴YPSL⁸ sequence of syntenin-1 as a novel Ub binding motif and a determinant in the formation of Ub-based regulatory complexes.

Ulk1 Controls the Syntenin-Ubiquitin Interaction

The importance of Ser⁶ prompted us to define the role of phosphorylation in the regulation of Ub-syntenin-1 interaction. Co-expression of syntenin-1 and Ulk1, a partner for syntenin-1 (25), in HEK293T cells yielded a slower migrating additional protein band (Fig. 6A, lane 3). This band disappeared when cellular lysates were treated with alkaline phosphatase (Fig. 6B, lane 6), indicating that syntenin-1 phosphorylation was responsible for this slower migrating species. Accordingly, no extra band was detectable when syntenin-1 was co-expressed with the kinase-dead mutant of Ulk1 (Fig. 6A, lane 4). Next, we examined the extent and effect of syntenin-1 phosphorylation by Ulk1. Although the S6A mutant migrates faster than the phosphorylated variant of the wild-type syntenin-1 in SDS-PAGE, its position was still above the nonphosphorylated form (Fig. 6C, middle). Not only did these results show that Ulk1 phosphorylates syntenin-1 on Ser⁶, but they also indicate additional phosphorylation sites in syntenin-1. Pulldown experiments were carried out to examine whether Ulk1-dependent phosphorylation of syntenin-1 regulates its binding to Ub. mUb-agarose did not retain the phosphorylated form of syntenin-1 from Ulk1-overexpressing cellular lysates (Fig. 6D, middle, lanes 5 and 6). Hence, Ulk1-dependent phosphorylation of syntenin-1 on Ser⁶ can represent a physiological switch which controls the interaction of Ub with syntenin-1. Together, this establishes the substrate specificity of Ulk1 and implicates syntenin-1 as its critical downstream target.

Ubiquitylation Influences Recruitment of Synteinin-1 to Endosomes

Ubiquitylation of transmembrane proteins serves as an important signal for their endocytic sorting (26). Accordingly, ubiquitylated proteins could be found on both early and late endosomes (27). To examine whether binding to Ub has a role in distribution of syntenin-1 to various endocytic compartments, we carried out immunofluorescence experiments using HeLa cells expressing the wild-type syntenin-1 and its mutants which lost their Ub binding property (i.e. Synt S6A and Synt P5A). There was only limited co-localization of the wild-type syntenin-1 or the mutants with markers of early endosomes (i.e. EEA-1 and STAM) (Fig. 7A and data not shown). Similarly, we found no evidence for co-localization of syntenin-1 with transferrin receptors on recycling endosomes (data not shown). By contrast, a significant proportion of syntenin-1-positive vesicles also contained CD63, a well established marker for late endosomes and lysosomes (Fig. 7B). Importantly, recruitment of both syntenin-1 mutants to these compartments was markedly suppressed. These results demonstrated that binding to ubiquitylated proteins may be an important factor that controls the distribution of syntenin-1 to late endocytic compartments.