SEPT9 negatively regulates ubiquitin-dependent downregulation of EGFR

Human septins constitute a family of guanine-nucleotide-binding proteins comprising 13 isoforms that are differentially expressed in a cell- and tissue-specific manner. Based on sequence similarity, these isoforms have been classified into the SEPT2, SEPT3, SEPT6 and SEPT7 subgroups (Kinoshita, 2003). All isoforms contain a central GTPase domain that allows them to assemble into hetero-oligomeric complexes (Sirajuddin et al., 2007). The G-domain is flanked by N- and C-terminal extensions, which provide interfaces for association with further binding partners. Accordingly, septin filaments have been described as molecular scaffolds that orchestrate the recruitment of downstream effectors. When purified from cells and tissues, septins form rod-shaped assemblies consisting of apolar hexamers and octamers. In the hexameric state, these complexes are usually composed of SEPT7 and members of the SEPT2 and SEPT6 subgroups. Individual members of the same subgroup can substitute for each other in the SEPT2–SEPT6–SEPT7 complex (Sellin et al., 2011). Octamers from dividing non-neuronal cells additionally contain SEPT9, which occupies the terminal positions. SEPT9 might thereby allow for filament elongation or abrogate the formation of septin superstructures, depending on the length of the individual splice variant incorporated (Kim et al., 2011).

Septin complexes bind to membrane surfaces enriched in negatively charged phospholipids, such as phosphoinositides (Tanaka-Takiguchi et al., 2009). They thereby form diffusion barriers that support the compartmentalization of membranes during a broad range of cellular processes, including cell division, membrane trafficking and signal transduction (Caudron and Barral, 2009). Given the fundamental importance of these processes, it is not surprising that the misregulation of certain isoforms has been correlated with the development of diseases, including tumorigenesis and neurodegeneration (Dolat et al., 2014; Mostowy and Cossart, 2012). Depletion of SEPT9, for example, has been shown to cause cytokinetic defects (Surka et al., 2002), owing to a misregulation of the abscission machinery at the midbody (Estey et al., 2010; Kim et al., 2011). Conversely, increased protein levels of SEPT9 have been detected in several tumors (Connolly et al., 2011a; Scott et al., 2005; Stanbery et al., 2010), suggesting that elevated levels of SEPT9 might induce or support the development of cancers. However, the molecular mechanisms underlying the cell biological functions of SEPT9 remain largely obscure.

The misregulation of growth factors and their receptors, in particular that of receptor tyrosine kinases (RTKs), has been correlated with several stages of cancer progression (Witsch et al., 2010). RTK signaling depends on the number of receptors exposed at the cell surface, which in turn is under dynamic regulation (Sorkin and Goh, 2009). Hence, endocytosis and intracellular trafficking of RTKs have profound effects on signaling strength and duration (Sigismund et al., 2008). In the case of the well-studied epidermal growth factor receptor (EGFR), internalization and degradative sorting have been shown to be regulated by the E3 ubiquitin ligase Cbl (Schmidt and Dikic, 2005). At high concentrations, ligand binding triggers the recruitment of Cbl to dimerized receptors at the plasma membrane and thereby induces receptor multi- and polyubiquitylation. Whether ubiquitylation is essential to drive endocytosis is still a matter of debate. However, at the level of endosomes, ubiquitin moieties are believed to support sorting of EGFR into the degradative pathway in an ESCRT-dependent manner (Haglund et al., 2003; Lu et al., 2003).

Cbl-interacting protein of 85 kDa (CIN85, also known as SH3KBP1) is an adaptor protein involved in the downregulation of RTKs such as EGFR. This function depends on its interaction with Cbl (Petrelli et al., 2002; Soubeyran et al., 2002; Szymkiewicz et al., 2002). CIN85 encodes three SH3-domains, which confer binding to atypical proline-arginine motifs (Kowanetz et al., 2003), a proline-rich middle domain and a C-terminal coiled-coil region implicated in membrane association (Zhang et al., 2009). Here, we show that SEPT9, by associating with CIN85 SH3 domains, negatively regulates ubiquitin-dependent EGFR degradation, thereby linking the molecular activity of SEPT9 to cellular growth control.

Given the fundamental role of SEPT9 during development and its involvement in cell division and differentiation (Estey et al., 2010; Füchtbauer et al., 2011), we set out to investigate potential changes in surface levels of growth factor receptors upon RNA interference (RNAi)-mediated depletion of SEPT9. We quantified surface EGFR by applying Alexa-Fluor-647–EGF (AF647–EGF) in the cold. Strikingly, SEPT9-, but not SEPT7-depleted HeLa cells displayed a near 50% reduction in EGFR surface levels when compared with control cells (Fig. 1A–C; supplementary material Fig. S1A). Knockdown of SEPT9 also slightly decreased the total levels of EGFR (supplementary material Fig. S1B). Other surface proteins, such as β1-integrin, remained unaffected (supplementary material Fig. S1C).

As these data suggest a possible role for SEPT9 in stabilizing EGFRs, we next investigated the effects of SEPT9 depletion on endosomal sorting of EGFR. Loss of SEPT9 substantially accelerated receptor degradation, reflected by an approximately twofold decrease in half-lives from 2 h to 1.1 h and 1.2 h (two independent siRNAs; Fig. 1D; supplementary material Fig. S1D). Accordingly, surface levels of EGFR recovered in knockdown cells when degradative processes were blocked by the addition of both proteasomal and lysosomal inhibitors (supplementary material Fig. S1E). Next, we performed pulse-chase experiments to monitor ligand-induced sorting of EGFR into degradative or recycling pathways. In knockdown cells, a larger fraction of ¹²⁵I-labeled EGF was degraded after 40 min of incubation as compared with that of control cells, and fewer receptors were recycled (Fig. 1E; supplementary material Fig. S1F).

Faster degradation might be the consequence of enhanced endocytic uptake of EGFR. As the mode of EGFR endocytosis is known to depend on ligand concentration (Sigismund et al., 2005), we determined the internalization rates of EGFR when stimulated with low (Fig. 1F) and high (supplementary material Fig. S1G) doses of ligand. SEPT9 depletion affected neither clathrin-dependent nor clathrin-independent endocytosis of EGFR. Taken together, these results thus demonstrate that SEPT9 modulates degradative sorting of EGFRs at the level of endosomes without affecting receptor internalization.

To gain insight into the molecular mechanisms that might underlie the SEPT9-dependent stabilization of EGFRs, we carried out knockdown-rescue experiments. Expression of full-length SEPT9, or of a truncation mutant comprising the G-domain alone (SEPT9ΔN), did not significantly alter the levels of surface EGFR (Fig. 1G; Fig. 2A). Importantly, upon re-expression of full-length SEPT9 in a SEPT9-knockdown background, surface EGFR levels almost completely recovered, whereas SEPT9ΔN, the mutant lacking the N-terminal domain, was unable to restore EGFR expression. This suggests a crucial role of the SEPT9 N-terminal domain during endolysosomal sorting of EGFR.

The SEPT9 N-terminal domain is enriched in proline residues. We therefore speculated that this region might link SEPT9 to SH3-domain-containing proteins implicated in endosomal sorting and/or cytoskeletal rearrangements (supplementary material Fig. S2A). Among the various SH3 domains tested, only the three SH3 domains of CIN85 were able to retain SEPT9 from cell extracts on a glutathione-S-transferase (GST) affinity matrix. CIN85 is an adaptor protein implicated in various steps of the vesicular transport of EGFRs, including internalization, degradation and recycling (Havrylov et al., 2010), and we thus considered CIN85 as a candidate factor linking SEPT9 to EGFR sorting. We probed for a potential interaction and found a substantial fraction of endogenous SEPT9 to be co-immunoprecipitated with native CIN85 from cell extracts (Fig. 2B).

Biochemical mapping experiments revealed that the SEPT9 N-terminal domain (GST–SEPT9_aa1–220) was sufficient to promote the interaction with CIN85 (Fig. 2C; supplementary material Fig. S2C). By contrast, the isolated G-domain (GST–SEPT9ΔN) was unable to associate with CIN85. Closer inspection of the SEPT9 primary sequence revealed the presence of three atypical proline-arginine (PR) motifs in its N-terminus, which would fit the consensus sequence known to mediate interactions with CIN85 SH3 domains. (Fig. 2A) (Kowanetz et al., 2003). One particular PR motif (SEPT9 PR129/130) is highly conserved between species (supplementary material Fig. S2B). Mutational inactivation of this motif (SEPT9 PR129/130AA) almost completely abolished complex formation with CIN85, whereas the two remaining motifs (PR26/27 and PR219/220) were largely dispensable for binding. A similar binding pattern was observed when the isolated SH3 domains of CIN85 (His₆–CIN85 SH3A-C) were probed (supplementary material Fig. S2C), indicating that the CIN85–SEPT9 interaction is direct.

As expected, the SH3 domains derived from CIN85 were sufficient to associate with SEPT9 from cell extracts, similar to other SH3 ligands, including dynamin-2 (Schroeder et al., 2010) (Fig. 2D; supplementary material Fig. S2A). A deletion variant of CIN85 lacking the SH3 domains (CIN85 aa334-end) was unable to interact with SEPT9. Interestingly, SEPT2/6/7-containing filaments known to assemble with SEPT9 (Kim et al., 2011) were co-purified on GST-fused CIN85 SH3A-C, but only when SEPT9 was present (Fig. 2E). Collectively, our results demonstrate that an atypical PR motif links SEPT9 to the SH3 domains of CIN85. Furthermore, our data suggest that this interaction serves to recruit SEPT9-containing filaments, rather than SEPT9 monomers, to activated receptors.

CIN85 associates with EGFR at the level of the plasma membrane, where – together with the ubiquitin ligase Cbl – it forms part of an EGF-induced signaling complex that remains assembled during internalization and endolysosomal sorting (Haglund et al., 2002; Kowanetz et al., 2004). To investigate at which stage SEPT9 might enter this complex, we applied indirect immunofluorescence microscopy using antibodies raised against SEPT9 and CIN85 (supplementary material Fig. S3A,B; Fig. 1A). Consistent with previous reports (Surka et al., 2002), endogenous SEPT9 localized to prominent fiber-like structures positioned in close proximity to the nucleus, and to thinner filaments in the cell periphery. Overexpression of mRFP-tagged CIN85 induced the formation of enlarged, aberrant endosomes, as described previously (Zhang et al., 2009), which accumulated late-endosomal markers such as LAMP1 (supplementary material Fig. S3C) and were clearly distinct from early EEA1-containing endosomes. Importantly, overexpression of CIN85 was sufficient to trigger the relocation of endogenous SEPT9 from filamentous structures onto CIN85-positive organelles.

We then analyzed the localization of endogenous SEPT9 upon stimulation with EGF. As expected, application of EGF facilitated the recruitment of CIN85 to activated EGFRs at the surface of cells, as shown by its colocalization with AF647-labeled EGF (Fig. 3A). At these sites, CIN85 also colocalized with SEPT9 and with haemagglutinin epitope (HA)-tagged c-Cbl, a ubiquitin ligase mediating ligand-induced ubiquitylation and downregulation of EGFRs (supplementary material Fig. S3D) (Schmidt and Dikic, 2005). Of note, the degree of colocalization between SEPT9 and EGF-positive structures significantly decreased upon knockdown of CIN85 (Fig. 3B,C). A similar effect could be observed for other septin isoforms co-assembling with SEPT9, for instance SEPT2, when CIN85 was depleted (Fig. 2E; Fig. 3D). Moreover, SEPT9 depletion decreased the colocalization of SEPT2 with surface-bound EGF, supporting the idea that septin filaments are recruited to the plasma membrane in a CIN85/SEPT9-dependent manner, rather than SEPT9 mono- or multimers.

To investigate a potential role of SEPT9 during degradative sorting of EGFRs, we monitored its localization during the transport of AF647–EGF to early (5 min) and late (15 min) endosomes. SEPT9 remained associated with filaments in the cell periphery, and displayed no colocalization with internalized EGFRs, irrespective of the presence or duration of the stimulus (Fig. 4A). Furthermore, SEPT9 did not translocate to early or to recycling endosomes upon overexpression of GTP-locked eGFP–Rab5 or of wild-type eGFP–Rab11, respectively (supplementary material Fig. S4A,B). Given the functional implication of CIN85 during the assembly of the ESCRT (endosomal sorting complex required for transport) machinery (Sadoul, 2006) and the role of septins in the recruitment of ESCRT components to the site of abscission (Renshaw et al., 2014), we also considered a potential function of the SEPT9–CIN85 complex during the formation of multivesicular bodies. However, degradative endosomes stalled by overexpression of dominant negative Vps4A(E228Q) (Baumgärtel et al., 2011) (supplementary material Fig. S4C) accumulated endogenous CIN85, as expected, but not SEPT9. Collectively, these data largely exclude the possibility that a native complex of SEPT9 and CIN85 is assembled on endosomes.

Septins can associate with membrane surfaces enriched in phosphoinositides (Tanaka-Takiguchi et al., 2009). We therefore performed biochemical fractionation experiments to gain insight into the subcellular distribution of SEPT9 upon stimulation with EGF. Under these conditions (e.g. at low temperatures, when SEPT9 is not localized to filaments), we found only a small pool of SEPT9 to be associated with membranes during starvation. As anticipated, stimulation with EGF triggered the translocation of the signaling adaptor Grb-2 from the cytosol to membranes in a dose-dependent manner (Fig. 4B,C). Interestingly, the same held true for SEPT9 and its cognate binding partners SEPT2 and SEPT7. In conclusion, these results indicate that SEPT9 – though regulating degradative sorting at the level of endosomes – is transiently recruited to ligand-engaged EGFRs at the plasma membrane in a CIN85-dependent manner, but is excluded from this complex during or soon after endocytosis.

Concomitant with the decrease in surface EGFR in knockdown cells, depletion of SEPT9 also significantly reduced phospho-AKT levels in cells upon stimulation with EGF (Fig. 4D; supplementary material Fig. S4D). Total AKT levels remained unaffected. By contrast, no changes in ERK1/2 signaling were detectable (supplementary material Fig S4D,E).

How can SEPT9 modulate degradative sorting of EGFR if it acts exclusively at the plasma membrane and not on endosomes? To answer this question, we followed a structure-based approach and applied nuclear magnetic resonance (NMR) spectroscopy to gain insights into the molecular mechanisms underlying the association of SEPT9 with CIN85. We designed a SEPT9-derived peptide containing the PR-signature motif needed for its association with CIN85, and incubated it with the ¹⁵N-labeled SH3A domain of CIN85 (Fig. 5A). Our experiments confirmed the direct association of the SEPT9 peptide with the SH3A domain of CIN85. Moreover, titration of the peptide induced changes in chemical shifts of amino acids implicated in binding to other PR-signature motifs (Fig. 5B) (Ceregido et al., 2013), indicating that both ligands compete for the same binding site. Accordingly, SEPT9- and Cbl-derived peptides occupied the same binding surface on SH3A domain of CIN85 (Fig. 5C). Affinity chromatography experiments performed in the presence of a SEPT9-derived peptide or of a mutant control peptide lacking the conserved PR-signature motif further confirmed that the binding of SEPT9 and Cbl to CIN85-SH3 domains is mutually exclusive (Fig. 5D). These structural and biochemical data suggest a molecular mechanism by which SEPT9 might regulate Cbl-mediated EGFR ubiquitylation and degradation. To test this directly, we assessed the extent of ubiquitylation of EGFRs in the presence or absence of SEPT9 upon stimulation of HeLa cells with ligand. Quantitative analyses revealed that EGFR affinity-purified from lysates of SEPT9-knockdown cells had incorporated significantly more ubiquitin than EGFR from control cells (Fig. 5E,F). These results provide a molecular explanation for the accelerated degradation observed in SEPT9-depleted cells.

Here, we identify a novel key regulatory role of SEPT9 in modulating the surface expression levels of EGFR. This function involves a highly conserved PR motif encoded by a subset of SEPT9 splice variants, which promotes the association with the SH3 domains of CIN85. SEPT9 stabilizes EGFRs by protecting them from degradation and increases the fraction of receptors recycling back to the plasma membrane, without affecting receptor internalization. However, despite its involvement in degradative sorting, SEPT9 does not act at the level of endosomes. Instead, EGF stimulation triggers the recruitment of SEPT9 exclusively to the plasma membrane, dependent on the presence of CIN85. Finally, we demonstrate that SEPT9 competes with Cbl for binding to CIN85 and inhibits ubiquitylation of EGFRs in cells. Taken together, our data thus provide a molecular explanation for the observed effects of SEPT9 depletion on EGFR sorting.

CIN85 is known to remain associated with activated EGFRs during endosomal sorting (Haglund et al., 2002). Our observations suggest that SEPT9 joins this complex transiently, and is released during or soon after internalization. Based on our fractionation assays, we hypothesize that phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] and phosphatidylinositol (3,4,5)-trisphosphate [PI(3,4,5)P₃] accumulating in the vicinity of stimulated EGFRs might aid the initial recruitment of SEPT9-containing filaments to the plasma membrane and support their retention at the cell surface (Tanaka-Takiguchi et al., 2009). According to our rescue experiments, CIN85-associated septin filaments need to contain long isoforms of SEPT9 to protect EGFRs from degradation (Fig. 1G). Of note, long SEPT9 isoforms have been shown to be upregulated in various tumors and cancer cell lines (Amir et al., 2010; Connolly et al., 2014; Connolly et al., 2011b; Stanbery et al., 2010). It remains to be determined whether and to what extent their interaction with CIN85 might contribute to cancer progression.

SEPT9 can bind to SEPT2–SEPT6–SEPT7 complexes in a non-stoichiometric manner, whereby long isoforms of SEPT9 stabilize the formation of higher-order oligomers (Kim et al., 2011). Interestingly, in comparison to members of other septin classes, SEPT9 shows the fastest hydrolysis rate (Zent and Wittinghofer, 2014). As filament formation and stability additionally depend on the nucleotide-binding state of individual septin monomers, it is possible that interactions with the N-terminal region could affect the GTPase activity of SEPT9 in filaments and thereby induce a remodeling of local septin structures.

Septin filaments assembled in close proximity to active EGFRs could initiate the formation of a diffusion barrier, which would aid the retention of signaling complexes in subdomains of the plasma membrane (Caudron and Barral, 2009). Moreover, these filaments might serve as scaffolds that recruit further binding partners of the SEPT9 N-terminal domain or of other septin isoforms co-assembled in the same complex (Hall and Russell, 2012). For example, the SEPT9 N-terminal domain binds to a RhoA-specific guanine nucleotide exchange factor (Nagata and Inagaki, 2005), potentially allowing for crosstalk between septin filaments and the actin cytoskeleton. Several studies have further highlighted the functional interplay between septins and other elements of the cytoskeleton (Mostowy and Cossart, 2012). However, whether such local crosstalk between different components of the cytoskeleton might contribute to the modulation of EGFR trafficking remains an open question.

Our structural and biochemical data further suggest that SEPT9 competes with c-Cbl for complex formation with CIN85, thereby attenuating receptor ubiquitylation. SEPT9-sensitive ubiquitylation does not significantly affect clathrin-dependent or clathrin-independent pathways of EGF internalization, in line with the observation that several mechanisms act in a cooperative manner during EGFR endocytosis (Goh et al., 2010). Ubiquitylation, however, has an essential role during EGFR degradation (Huang et al., 2006). For instance, it has been shown that ubiquitylated EGFR associates with HRS, a component of the ESCRT-0 complex (Sigismund et al., 2005). Furthermore, perturbing the interaction between HRS and the ESCRT-I subunit Tsg101 (Lu et al., 2003), or depleting Tsg101 by RNAi (Hislop et al., 2004), attenuates EGFR degradation. Taking these observations into account, we hypothesize that SEPT9 limits the extent of EGFR ubiquitylation at the level of the plasma membrane, and thereby negatively regulates the ubiquitin-dependent association of EGFR with the ESCRT machinery at the level of endosomes. We note that CIN85 has recently been implicated in recycling of EGFRs (Schroeder et al., 2010). However, a direct role of SEPT9 in this process is unlikely, given that we found no evidence for the localization of SEPT9 on recycling endosomes (supplementary material Fig. S4B).

Besides c-Cbl/Cbl-b, numerous other factors have been found to associate with the SH3 domains of CIN85 (Buchse et al., 2011; Kowanetz et al., 2004), including proteins involved in EGFR signaling and subcellular trafficking. It is not unlikely that SEPT9 also regulates the formation of those complexes, at least as far as they are assembled at the plasma membrane, and thereby serves to coordinate their activities in time and space. Whether and how SEPT9 modulates the composition of the EGFR interactome will be subject of future studies. CIN85 and its paralog CD2AP have not only been implicated in the regulation of the surface levels of EGFR, but also of other receptor tyrosine kinases, such as c-Met (Petrelli et al., 2002). Interestingly, several septins, including SEPT9, are recruited to stimulated Met during invasion of Listeria and Shigella into HeLa cells (Mostowy et al., 2009). Moreover, the subsequent entry of bacteria appears to depend on CIN85, as well as Cbl-mediated ubiquitylation of Met (Veiga and Cossart, 2005). Whether SEPT9 regulates CIN85 function during this process as well remains to be determined.

In general, the deregulation of growth factors and their receptors is known to promote cancer progression at various stages (Witsch et al., 2010). It is interesting to note that the reduced levels of surface EGFR in SEPT9-knockdown cells correlate with reduced levels of phospho-AKT detectable upon stimulation with EGF. By contrast, phospho-ERK1/2 levels remained unchanged, indicating additional functions of SEPT9 in fine-tuning EGFR signaling. Given the role of PI(3,4,5)P₃ during cellular migration, this suggests that SEPT9 might modulate the migratory potential of cancer cells. In line with this hypothesis, cultured embryonic fibroblasts derived from conditional SEPT9-knockout mice have been shown to migrate more slowly than wild-type cells (Füchtbauer et al., 2011). Also in this regard, our study provides a novel molecular explanation for the genetic association of SEPT9 with cancer (Connolly et al., 2011a).

Full-length and deletion variants of human SEPT9_v3 (BC021102; isoform c) and human CIN85_v1 (BC015806) were cloned into pGEX (Amersham Biosciences) or pET28a+ (Novagen) for bacterial expression or a pcDNA3-based plasmid encoding an in-frame fusion with HA, EGFP or mRFP tags for expression in mammalian cells. Site-directed mutagenesis was performed by overlapping PCR. All constructs were verified by dsDNA sequencing (MWG, BioTez). The following truncations of CIN85 were used in this study: SH3 A (amino acids 1–69), SH3 B (amino acids 96–163), SH3 C (amino acids 256–333), SH3 A-C (amino acids 1–333), C-terminal region (amino acids 334–end). The following truncations of SEPT9 were used in this study: N-terminal region (amino acids 1–220), ΔN (amino acids 238–end). For rescue experiments, an siRNA-resistant, untagged version of SEPT9 was generated.

Human EGF was purchased from Peprotech. Mouse Alexa-Fluor-647-conjugated (AF647)–EGF was purchased from Life Technologies. ¹²⁵I-EGF for uptake experiments was obtained from Bachem (H-7112) and for recycling experiments was obtained from Perkin Elmer (NEX42800). MG132 was obtained from Calbiochem and dissolved in DMSO. NH₄Cl was obtained from Roth and dissolved in water.

For immunofluorescence experiments, antibodies were used against HA (clone HA.11, Babco/Convance), EEA1 (BD transduction, #610456), β1-integrin (EMD Millipore, #MAB1981), EGFR (clone R-1, Santa Cruz, #sc-101) and LAMP-1 (BD Pharmingen, #555801). For western blotting experiments, antibodies were used against β-actin (clone AC-15, Sigma), clathrin heavy chain (clone TD.1), dynamin 2 (BD Bioscience, #610245), c-Cbl (BD transduction, #610441), EGFR (Cell signaling, #4267), tubulin (Sigma, #T5168), ubiquitin (clone P4D1, Cell Signaling, #3936), SEPT6 (Sigma, #HPA005665), SEPT2 (Sigma, #HPA018481), Grb2 (BD Transduction, #610112), AKT1/2/3 (Cell Signaling, #9272), phospho-AKT-Ser473 (Cell Signaling, clone D9E, #4060), ERK1/2 (Abcam, #ab17942), phospho-ERK1/2 (Sigma, #M8159) and His₆-tag (Novagen, #70796-3). Quantification from western blots was performed using the western blot plugin from ImageJ.

Polyclonal antibodies against SEPT7, SEPT9 and CIN85 were raised in rabbits by injecting two different peptides. All antisera were affinity-purified and tested for specificity. For the SEPT7 antibody, rabbits were immunized against the peptide NH₂-CYEFPETDDEEENKLV-COOH. For the SEPT9 antibody, the peptides NH₂-CSTQKFQDLGVKNSEP-CONH₂ and NH₂-CTELSIDISSKQVEN-CONH₂ were co-injected. The epitope was designed such that the resulting antibody detects all long isoforms of SEPT9 (v1, v2, v3 and v8) that contain the PR motif located at position 129-130 of isoform v3. This antibody was used throughout the whole study for immunofluorescence and immunoblotting, but not for the coimmunoprecipitation experiment in Fig. 2A. Here, the SEPT9 antibody from Abnova (H00010801-M01) was used. To allow for parallel detection of endogenous CIN85 and SEPT9, our SEPT9-specific antibody was directly coupled to Atto565 using the Lightning Link™ Rapid Conjugation System (Innova Biosciences, #351-0010) according to the manufacturer's instructions. For the generation of CIN85-specific antibodies used during immunoprecipitation, the peptides NH₂-CLQMEVNDIKKALQSK-COOH and NH₂-CKQLLSELDEEKKIRL-CONH2 were co-injected. For CIN85 antibodies used in immunofluorescence, purified His₆-tagged CIN85 (amino acids 429–end) was injected as an antigen. A CIN85 antibody from Upstate (Clone 179.1.E1) was used for immunoblotting. Secondary antibodies were purchased from Molecular Probes (conjugated to fluorescent dyes) or Dianova (horseradish peroxidase conjugated).

All siRNAs were purchased from MWG and synthesized with a 3′ TT overhang. siRNA sequences were as follows: SEPT9 #1, 5′-GGAGGAGGUCAACAUCAAC-3′; SEPT9 #2, 5′-AGACCAUCGAGAUCAAGUC-3′; CIN85, 5′-GGCACAGAAUGAUGAUGAA-3′; SEPT7, 5′-CUUGCAGCUGUGACUUAUA-3′; Control, 5′-AUCGUUGACUUACAAGAGA-3′. The sequences of the peptides used are as follows: SEPT9 PxxxPR, NH₂-EVLGHKTPEPAPRRTE-COOH; SEPT9 PxxxAA, NH₂-EVLGHKTPEPAAARTE-COOH.

HeLa cells (HeLa M clone) and A431 cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum (FCS) and antibiotics (0.1 U/ml penicillin, 0.1 µg/ml streptomycin). Plasmids were transfected into cells using JetPrime (Polyplus transfection). siRNAs were transfected into cells using Oligofectamine (Invitrogen) and the samples were analyzed after 48 h.

For all biochemical experiments except the EGFR ubiquitylation assay, cells were lysed in extract buffer (20 mM HEPES pH 7.4, 100 mM KCl, 2 mM MgCl₂, 1% Triton X-100) supplemented with 1 mM PMSF and protease inhibitor cocktail (Sigma). For experiments using EGF-stimulated cells, phosphatase inhibitor cocktails were added (Sigma, #P0044 and #P5726). Lysates were precleared by centrifugation at 13,000 g for 1 min and ultracentrifuged for 15 min at 180,000 g. Extracts were incubated with antibodies immobilized on Protein-A/G-PLUS–agarose (Santa Cruz Biotechnology) for 2 h at 4°C on a rotating wheel. Beads were washed with extract buffer and bound proteins were eluted with SDS-PAGE sample buffer.

His₆-tagged (vector pET28a) or GST-tagged proteins (vector pGEX) were expressed in BL21 (DE3) and purified from Escherichia coli using His-Select Nickel Affinity Gel (Sigma) or GST Bind Resin (Novagen) according to the manufacturer's instructions.

Cell extracts were prepared as described above and incubated with purified GST-tagged proteins for 2 h at 4°C on a rotating wheel. Beads were washed with extract buffer and bound proteins were eluted with SDS-PAGE sample buffer.

Eluted His₆-tagged proteins were cleared from possible precipitates immediately prior to the experiment by ultracentrifugation at 180,000 g. A total of 115 pmol of GST-tagged proteins was incubated with equimolar amounts of His₆-tagged proteins for 30 min at 4°C on a rotating wheel. Beads were washed with extract buffer and bound proteins were eluted with SDS-PAGE sample buffer. During binding and washing, 10 mM imidazole was added.

For NMR studies, ¹⁵N-labeled His₆–CIN85 SH3A (amino acids 1–58, subcloned into pET28a) was purified in 20 mM Tris-HCl pH 8, 50 mM NaCl and subjected to thrombin cleavage overnight at 4°C. The His₆-tag was removed by gel filtration on a Superdex 75 (GE Healthcare). Purified SH3A was dialyzed overnight into a buffer containing 50 mM cacodylate pH 7.4. A reference assignment of the resonances of the ¹H, ¹⁵N-HSQC of the protein was kindly provided by others (Ceregido et al., 2013). Because the protein construct used here slightly differed from the one documented previously, we performed measurements to confirm the assignment of the protein resonances. A sample of 600 µl of ¹⁵N-labeled 120 µM SH3-A was used in a 5-mm NMR tube at a temperature of 298 K. Experiments were performed on a AV600 Bruker spectrometer (600 MHz 1H frequency) using a QXI probe equipped with a self-shielding z-gradient. Three 3D experiments were performed: a ¹⁵N-NOESY-HSQC using 16 scans, a data size of 512(¹H)×64(¹⁵N)×80(¹H) complex points, t_H,max = 51.2 ms, t_N,max = 21.2 ms and t_H,max(i) = 9.6 ms; a ¹⁵N-TOCSY-HSQC using eight scans, a data size of 512(¹H)×64(¹⁵N)×80(¹H) complex points, t_H,max = 51.2 ms, t_N,max = 21.2 ms and t_H,max(i) = 9.6 ms; and a HNHA using eight scans, a data size of 512(¹H)×48(¹⁵N)×46(¹H) complex points, t_H,max = 51.2 ms, t_N,max = 15.9 ms and t_H,max(i) = 10.6 ms.

NMR experiments to study the interaction with a SEPT9-derived peptide were performed on a AV600 Bruker spectrometer (600 MHz ¹H frequency) using a TCI cryoprobe equipped with a self-shielding z-gradient. For the titration, the peptide was added from a stock solution of 5 mM or 50 mM for the higher concentrations to avoid dilution of the sample. Five ¹H,¹⁵N-HSQCs were recorded using four scans, a data size of 512(1H)*128(15N) complex points and a t_H,max = 51.2 ms and t_N,max = 42.5 ms. The protein∶peptide ratio was 1∶0, 1∶0.5, 1∶1, 1∶2 and 1∶10. No attempt was made to completely saturate the protein. Differences in chemical shift were calculated from the peak positions in the individual spectra using the formula shift = sqrt[(Δδ(¹H))²+(Δδ(¹⁵N)/10)²], where Δδ is the difference in chemical shift in the respective dimensions.

All spectra were recorded and processed using topspin 3.1. (Bruker Biospin, Karlsruhe, Germany). The processed data were first converted to UCSF format (Goddard and Kneller, SPARKY 3. University of California, San Francisco) and subsequently transferred to CCPN (Vranken et al., 2005) for assignment. A molecular model was created using the X-ray structure B2Z8 by replacing the amino acid side chains and minimizing the energy using Sybyl.

HeLa cells were starved for 4 h in serum-free medium and incubated for the indicated times with 500 ng/ml EGF in presence of 10 µg/ml cycloheximide. For harvesting, cells were placed on ice, washed once with ice-cold PBS and scraped in extract buffer. Lysates were centrifuged for 5 min at 13,000 g and supernatants were subjected to western blot analysis. EGFR levels were quantified, corrected to a loading control (actin or tubulin) and each timecourse was normalized to time-point zero.

Cells were seeded on Matrigel-coated coverslips and fixed with 4% paraformaldehyde (PFA), 4% sucrose in PBS for 10 min or with methanol for 5 min at −20°C. Standard protocols were used for immunostaining. For immunostaining of surface proteins, blocking and antibody decoration was performed in medium containing 10% FCS and PBS was used for washing. Images were acquired on a Zeiss Axiovert 200M-based spinning-disc confocal microscope (Perkin Elmer Inc.) under the control of Volocity software (Improvision Inc.) or by epifluorescence microscopy under the control of MicroManager. Data analysis of epifluorescence data was performed with ImageJ. Quantification of confocal images was performed with Volocity.

HeLa cells were seeded on Matrigel-coated glass coverslips and starved for 3 h in serum-free medium. Cells were stimulated for the indicated times with serum-free medium containing AF647–EGF at 37°C and washed twice with PBS containing 2 mM MgCl₂ at room temperature. Upon fixation with 4% PFA/4% sucrose in PBS for 10 min, cells were analyzed by confocal microscopy.

HeLa cells were seeded on Matrigel-coated glass coverslips and starved for 3 h in serum-free medium. Cells were incubated with 500 ng/ml AF647–EGF for 30 min at 9°C, transferred to ice and washed twice with ice-cold PBS containing 2 mM MgCl₂. Upon fixation with pre-cooled 4% PFA/4% sucrose in PBS for 10 min at room temperature, cells were analyzed by immunofluorescence microscopy. For quantification of surface EGF level, cells were analyzed by quantitative epifluorescence microscopy. A subtraction of the area of the DAPI staining from the surface EGF signal was needed to decrease background fluorescence. Sum fluorescence intensity per cell was then calculated and normalized to that of the control condition. For colocalization studies with surface-bound EGF, cells were analyzed by confocal microscopy. The amount of the respective protein colocalizing with surface EGF was calculated with an intersection mask. Sum fluorescence intensity of this mask was normalized to total fluorescence intensity of EGF.

HeLa cells were starved for 3 h in serum-free medium, cooled on ice and treated with 100 ng/ml EGF for 30 min at 13°C to prevent internalization. Cells were lysed in RIPA buffer [50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate, 1 mM PMSF, protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktails (Sigma, #P0044 and #P5726)]. A total of 1 mg of lysates was ultracentrifuged for 15 min at 180,000 g and incubated with EGFR-antibody (sc-101) immobilized on Protein-A/G-PLUS–agarose (both Santa Cruz Biotechnology) for 2 h at 4°C on a rotating wheel. Beads were washed with RIPA buffer and bound proteins were eluted with SDS-PAGE sample buffer and analyzed by immunoblotting. To compensate for the decrease in EGFR level in SEPT9-knockdown samples, a 1.6-fold volume of the immunoprecipitation sample was loaded onto the gel compared to the control condition. For analysis, the amounts of ubiquitin were corrected for the amount of immunoprecipitated EGFR, which was determined from parallel gels.

HeLa cells were starved for 3 h in serum-free medium, cooled on ice and treated with the indicated concentrations of EGF for 60 min at 13°C to prevent internalization. Cells were scraped in 20 mM HEPES pH 7.4, 100 mM KCl, 2 mM MgCl₂ and cracked using a cell cracker device (HGM, Heidelberg, Germany) following three freeze-thaw cycles in liquid nitrogen. Nuclei and cell debris were discarded following centrifugation at 1000 g for 5 min. Cytosol and membranes were separated by ultracentrifugation for 1 h at 440,000 g. The membrane pellet was dissolved in hot SDS-PAGE sample buffer and thoroughly homogenized. Equal amounts of both fractions were analyzed by immunoblotting.

¹²⁵I-EGF uptake was performed as described previously (Kornilova et al., 1996). HeLa cells were seeded in 24-well plates and starved for 4 h in serum-free medium containing 0.1% BSA and 20 mM HEPES pH 7.4. Cells were stimulated with ¹²⁵I-EGF in starvation medium at 37°C and washed twice on ice with PBS. Surface-bound ¹²⁵I-EGF was removed by an acid wash with 0.2 M acetic acid, 0.5 M NaCl for 5 min on ice. The acid wash (corresponding to surface-bound EGF) was collected and radioactivity was measured using a scintillation counter (HIDEX 300SL). Cells were dried at room temperature for 5 min and lysed with 1 M NaOH for 60 min. Radioactivity of the lysate was measured (corresponding to internalized EGF). Non-specific binding was measured for each time-point in the presence of a 300-fold excess of cold EGF and was subtracted from all values. The ratio of internalized to surface-bound EGF was plotted relative to time.

¹²⁵I-EGF recycling was performed as described previously (Kornilova et al., 1996). HeLa cells were seeded in 24-well plates and starved for 4 h in serum-free medium containing 0.1% BSA and 20 mM HEPES pH 7.4. Cells were stimulated with 20 ng/ml ¹²⁵I-EGF in starvation medium for 15 min and washed twice on ice with PBS. ¹²⁵I-EGF remaining on the cell surface was removed by a mild acid wash (0.2 M sodium acetate, 0.5 M NaCl, pH 4.5) on ice for 3 min. To allow for endosomal progression of EGF-bound receptors, cells were washed twice with pre-warmed starvation medium and further incubated in starvation medium containing 4 µg/ml unlabeled EGF for 40 min at 37°C. Cells were placed on ice, the medium was collected and subjected to a trichloroacetic acid (TCA) precipitation (5% TCA, 0.5% BSA; overnight at 4°C). This step serves to separate recycled, intact EGF from degraded EGF released into the medium. After washing cells twice with PBS, a minor fraction of recycled EGF that remained surface bound was removed by an acid wash (0.2 M acetic acid pH 2.5, 0.5 M NaCl for 2 min on ice). The acid wash was collected and radioactivity was measured. Cells were dried at room temperature for 30 min and lysed in 1 mM NaOH for 30 min. Cell lysates were subjected to TCA precipitation as described above. TCA-precipitated (left-over internal, intact EGF) and TCA-soluble (degraded EGF) ¹²⁵I were separated by a centrifugation at 13,000 g for 5 min. The pellet was dissolved in 1 M NaOH, and both fractions, supernatant and pellet, were analyzed in a γ-counter (Wallac 1470 Wizard). Non-specific counts were measured for each time-point in the presence of a 300-fold excess of unlabeled EGF and were subtracted from all values.

Total internalized EGF was calculated as the sum of measured radioactivities in all fractions derived from one well and set to 100%. Radioactivity found in TCA precipitates of the medium supernatant was added to the radioactivity in the acid wash and defined as recycled EGF. Degraded EGF was determined by the sum of the radioactivity in the TCA-soluble fractions of both medium supernatant and lysate. TCA-precipitable radioactivity in the lysate was defined as leftover internalized EGF that had remained in the cell. Recycled, degraded and leftover internalized EGF were all normalized to total internalized EGF. Receptor trafficking as depicted in Fig. 1E was plotted as the ratio of recycled to degraded EGF.

Phosphorylation of AKT at Ser473 was measured using the pSer473-Akt1/2 ELISA kit (Enzo Life Sciences, #ADI-900-162). HeLa cells were starved overnight in medium containing 0.1% FCS and stimulated for 5 min at 37°C with EGF. Cells were placed on ice, washed twice with PBS and lysed on ice in 20 mM Tris-HCl pH 7.4, 500 mM NaCl, 10 mM EDTA, 1% Triton X-100, 20 mM NaF, 20 mM β-glycerophosphate, 1 mM PMSF, protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktails (Sigma, #P0044 and #P5726). Lysates were centrifuged for 5 min at 13,000 g and supernatants were used for the ELISA assay at a final concentration of 40 µg/ml, according to the manufacturer's instructions. Total levels of pSer473-Akt1/2 were determined using a standard curve, which was processed in parallel for each experiment. Values were normalized to control cells under starved conditions. Additionally, lysates were subjected to western blot analysis to control total AKT1/2/3 levels.

Statistical analyses were performed using the unpaired t-test for all quantified experiments except for the data shown in Fig. 5F (EGFR ubiquitylation after knockdown of SEPT9) and supplementary material Fig. S1B (EGFR level after knockdown of SEPT9). Here, a one-sample t-test was used.

We thank Dr M.A. Ceregido and Dr N.A.J. van Nuland (Vlaams Instituut voor Biotechnologie, Brussels, Belgium) for the assignment of the resonances of the ¹H, ¹⁵N-HSQC of SH3A of CIN85, Dr G. Krause (Leibniz-Institut für Molekulare Pharmakologie, Berlin, Germany) for the creation of the structural model and Dr Hans-Georg Krausslich (Universitätsklinikum Heidelberg, Germany) for the plasmid encoding EGFP–Vps4 E228Q.