Eps15R and clathrin regulate EphB2‐mediated cell repulsion

Expression of Eph receptors and their ligands, the ephrins, have important functions in boundary formation and morphogenesis in both adult and embryonic tissue. The EphB receptors and ephrinB ligands are transmembrane proteins that are expressed in different cells and their interaction drives cell repulsion. For cell repulsion to occur, trans‐endocytosis of the inter‐cellular receptor‐ligand EphB‐ephrinB complex is required. The molecular mechanism underlying trans‐endocytosis is poorly defined. Here we show that the process is clathrin‐ and Eps15R‐mediated using Co115 colorectal cell lines stably expressing EphB2 and ephrinB1. Cell repulsion in co‐cultures of EphB2‐ and ephrinB1‐expressing cells is significantly reduced by knockdown of Eps15R but not Eps15. A novel interaction motif in Eps15R, DPFxxLDPF, is shown to bind directly to the clathrin terminal domain in vitro. Moreover, the interaction between Eps15R and clathrin is required for EphB2‐mediated cell repulsion as shown in a rescue experiment in the EphB2 co‐culture assay where wild type Eps15R but not the clathrin‐binding mutant rescues cell repulsion. These results provide the first evidence that Eps15R together with clathrin control EphB/ephrinB trans‐endocytosis and thereby cell repulsion.


| INTRODUCTION
EphB receptors are a large family of receptor tyrosine kinases that interact with the transmembrane ephrinB ligands. EphB/ephrinB signalling is important for contact-mediated cell repulsion and tissue patterning during development, but also in adult tissue to establish morphological borders in the intestine or to direct growing axons in the nervous system. 1-3 Furthermore, aberrant expression or mistargeted expression of Eph receptors is associated with cancer cell invasion in prostate, breast and colon cancer. [4][5][6][7][8][9] The adhesive interaction between EphB receptors and ephrinB ligands activates intracellular signalling pathways that regulate cell-cell repulsion, migratory behaviour, adhesion and cell polarity. The EphB receptor and the ephrinB ligands are not expressed in the same cell and their interaction can therefore lead to cell-cell repulsion. However, in order for the cells to repel the EphB/ephrinB complex needs to be physically removed or to dissociate. There are two mechanisms described to date: transendocytosis (trans-cellular internalisation event) of the receptorligand complex into one of the cells, 10,11 or cleavage of the extracellular domain by a protease. [12][13][14][15] The endocytosis of EphB/ephrinB complexes is an unusual type of endocytosis where two transmembrane proteins from neighbouring cells are internalised into one cell, thus forming vesicles from two plasma membranes. 11 The process, called trans-endocytosis, is dependent on dynamin scission and actin polymerisation but no association with either clathrin or caveolae has been found. 10,11,16,17 Thus, the molecular mechanism requires further investigation.
The EphB2 receptor interacts with the endocytic adaptor protein Numb. 18 Numb is a phosphotyrosine binding adaptor protein that regulates receptor trafficking. 19 Numb interacts directly with the endocytic adaptor protein AP2 and endocytic accessory proteins Eps15/R and intersectin1/2. [20][21][22] AP2 is the main adaptor protein that directs the clathrin coat formation. 23 The interaction of Numb with AP2 is mediated by a single DPF (aspartic acid, proline, phenylalanine) motif, an interaction that is too weak to functionally engage the AP2 complex and actively promote clathrinmediated endocytosis, [22][23][24] suggesting that the recruitment of endocytic accessory proteins is important for Numb-mediated EphB2 receptor endocytosis.
In this study, we identified novel components of the transendocytosis pathway for EphB/ephrinB internalisation. Here we cocultured colorectal cancer cell-lines, Co115, stably expressing the full-length transmembrane EphB2 receptors or ephrinB1 ligands.
Using the co-culture system, we found that EphB2-mediated cell-cell repulsion is a clathrin-and dynamin-dependent mechanism. Moreover we identified a key component in the trans-endocytosis protein complex, Eps15R, that interacts with EphB2 via the adaptor protein Numb. Using shRNA knockdowns, morphological analysis, and motif mapping, we identified a novel non-canonical clathrin-binding motif in Eps15R that was functionally important for EphB2-mediated cellcell repulsion. These results suggest that Eps15R and clathrinmediated trans-endocytosis of the EphB2/ephrinB1 complex is an important mechanism for terminating this adhesive interaction and turning it into cell-cell repulsion.

| EphB2-mediated cell repulsion is regulated by clathrin-mediated endocytosis
To visualise EphB2-mediated cell repulsion we used the colorectal cancer cell line Co115 stably expressing EphB2 and EGFP, ephrinB1 and RFP, EGFP alone and RFP alone. 5 The cells were co-cultured for 48 h, fixed, imaged and the size of the clusters of EGFP-expressing cells was quantified as previously described. 5 During the co-culture the initially randomly mixed EphB2-and ephrinB1-expressing cells repelled over time to minimise contact and formed homogenous clusters (groups) of cells expressing either EphB2 receptor or ephrinB1 ligand, resulting in a pattern formation of the two cell lines, which is distinct from the co-culture of EphB2-expressing cells with RFP or two cell lines that do not express EphB receptor or ligand ( Figure 1A).
Quantification of the percentage of clusters of EGFP-positive cells showed that EphB2-expressing cells formed large clusters when cocultured with ephrinB1-expressing cells ( Figure 1B), but were more mixed (a higher proportion of small EGFP-positive cell clusters) when co-cultured with control RFP-expressing cells consistent with previous observations. 5 Cell clusters containing >31 cells were not present in control co-cultures ( Figure 1B). As the ephrins are involved in tissue patterning, we applied an established method for pattern analysis on our EphB-mediated cell repulsion assay that has been developed for analysing chimeric patterning in retina tissues. 25,26 This method has been shown to distinguish between random mixing and clustered patterns formed by two cell populations both in tissue and in computer models. 26 It has the advantage of correcting for unequal proportions of the two populations that may arise at the seeding stage of the repulsion assay in the 'pattern score' that is generated. 25 The images collected from our assay showed striking differences in patterning ( Figure 1A). The results showed that the method reliably discriminates between cell clustering in EphB/ephrinB co-culture and the random cell mixing in the two controls where RFP-expressing cells were co-cultured with either EphB2-or EGFP-expressing Co115 cells ( Figure 1C; P < 0.0001, Student's t test). Thus, we conclude that image analysis of patterning can be applied to EphB-ephrinB cell repulsion studies as we found it to produce reproducible data in agreement with previously published findings. 5 To analyse whether EphB2-mediated cell repulsion was dependent on the GTPase dynamin-1, a scission molecule that is involved in most endocytic pathways to sever membrane buds from the plasma membrane, we overexpressed the dominant negative GTPase mutant T65A in our co-cultures. 27 Inhibition of dynamin-mediated membrane scission strongly reduced the clustering of EphB receptor and ephrinB1 expressing cells ( Figure 1D-F; P < 0.0001). Because multiple endocytic pathways are dependent on dynamin-1 for membrane scission we next used a specific reagent to block clathrinmediated endocytosis, AP180 C-terminus. AP180 is a membranebinding protein that promotes clathrin polymerisation and formation of endocytic membrane buds. 28,29 The C-terminus of AP180 has multiple clathrin-binding motifs and overexpression of this domain efficiently blocks internalisation of receptors that are internalised via clathrin-mediated endocytosis. [29][30][31] Expression of AP180 C-terminus in the Co115 co-cultures with EphB2 (GFP) and ephrinB1 (RFP) inhibited EphB/ephrinB cell patterning and increased cell mixing (Figure 1D, F; P < 0.0001), thus establishing that EphB2-mediated cell repulsion is regulated by clathrin-mediated endocytosis.

| EphB2 interacts with Eps15R and Eps15 via Numb
To further our understanding of EphB2 trans-endocytosis we next sought to identify novel components of the endocytic complex. Studies using immuno-precipitation and mass spectrometry methods to map the interactome of EphB2 have not yielded data that identify endocytic proteins. 32,33 We therefore decided to target our search to the previously identified endocytic adaptor protein of EphB2, Numb. 18 The interaction between Numb and EphB2 is mediated by the phosphotyrosine-binding (PTB) domain of Numb, and is functionally important during neural development. 18 Numb has previously been implicated in receptor-mediated endocytosis, particularly of receptor tyrosine kinases. 19 It is comprised of a PTB domain that binds to activated receptor tyrosine kinases, a motif domain containing proline-rich motifs binding SH3 domains, a single DPF motif that binds the clathrin adaptor AP2, and two NPF motifs that bind EHdomain containing proteins. 18,22,34,35 AP2 is the major adaptor protein for bringing in cargo to clathrin-coated pits. However the weak affinity interaction of a single DPF motif binding to AP2 (K d ≈ 1 μM) is not sufficient for formation of a stable protein complex that can nucleate clathrin coat formation. 24,36 We therefore performed a small interaction screen of EH domain-containing endocytic proteins to identify interaction partners of the NPF motifs in Numb. The NPF motif binds EH domains and there are a number of endocytic scaffold proteins that contain multiple EH domains; Eps15, Eps15R, intersectin-1, intersectin-2. EH domain scaffold proteins serve important functions in clathrin-mediated endocytosis, in particular by facilitating the formation of multi-protein complexes. [37][38][39] To investigate which endocytic EH-domain containing proteins interacted with Numb, a screen was performed with EH domains of Eps15R, Eps15, intersectin-1, and intersectin-2 expressed as recombinant GST-tagged protein and immobilised on beads. We found that Numb specifically bound the second EH domains of Eps15R and Eps15, but not the EH domains of intersectin-1 or -2 ( Figure 2A). Thus, Numb has a binding specificity for certain EH domains. We found that Eps15R and Eps15 both form a complex with EphB2 in vivo together with Numb, as shown in a co-immunoprecipitation assay from HEK293T cells ( Figure 2B). The interaction was more prominent upon stimulation of the EphB2-expressing cells with pre-clustered extracellular domain of ephrinB1. A kinase dead mutant of EphB2 was used as a control for  Error bars represent standard deviation (n = 5 random fields per condition). The frequency of cells in small clusters (<10 cells) is a representation of cell mixing and there is a significant difference between EphB2 co-cultured with ephrinB1 and RFP, and control EGFP plus RFP co-cultures (****P < 0.0001, Student's two-tailed unpaired t test). C, The EphB-mediated patterning was analysed across random field of co-cultured EphB2-and ephrinB1-expressing cells. The pattern score takes into account random variation of the proportion of the EGFP-and RFP-expressing cells. Data represent mean AE standard error of the mean. ****P < 0.0001, Student's two-tailed unpaired t test, n = 20 per condition in four independent experiments. D, Endocytosis was inhibited by expression of two reagents that have established dominant negative impact on endocytosis, dynamin1-T65A and AP180 Cterminus and the effect on EphB2-mediated cell repulsion in the Co115 co-cultures was evaluated. Scale bar, 500 μm. E, Cell mixing was quantified by counting the percentage of EGFP-positive cells forming clusters of various sizes in co-cultures of EphB2-and ephrinB1-expressing cells. In the control experiment a large proportion of cells were found in large (>30 cells) clusters, but this was dramatically reduced when clathrin-mediated endocytosis was inhibited by either expression of dynamin1-T65A or AP180 C-terminus. Error bars represent standard deviation (n = 5 random fields per condition, ****P < 0.0001, Student's twotailed unpaired t test). F, Bar graph showing the quantification of EphB2-mediated patterning of EphB2/ephrinB1 or control EphB2/RFP cocultures expressing control BFP, dynamin1-T65A, or AP180 C-terminus. Data represent mean AE standard error of the mean. ****P < 0.0001, Student's two-tailed unpaired t test, at least 20 images were analysed in three independent experiments ligand activation. We suggest that this protein complex consisting of EphB2-Numb-Eps15/R could function as the adaptor complex for EphB2 endocytosis mediated by clathrin, which has not been described previously. We therefore pursued further study of the role of Eps15R and Eps15 in the endocytosis of EphB2 receptor.

| The role of Eps15R and Eps15 in clathrinmediated endocytosis
Having established that trans-endocytosis is clathrin-mediated and that the EphB2 receptor forms a complex with Eps15R and Eps15, we next wanted to explore the molecular mechanisms further. Eps15 Western blot analysis of BSC1 cell lysates showed an efficient knockdown of Eps15R or Eps15 protein levels, and did not show a compensatory increase in either Eps15 or Eps15R expression ( Figure 3C).
This suggested that Eps15R and Eps15 are functionally diverse and we therefore focused further experiments on Eps15R.
Knockdown of Eps15R showed a reduced uptake of fluorescent transferrin in both HeLa and BSC1 cells treated with shRNA targeting the EPS15R gene (shEPS15R; Figure 3D) demonstrating that clathrin-

| Eps15R formed a direct interaction with clathrin heavy chain terminal domain
The prominent effect on clathrin-coat morphology following Eps15R knockdown led us to next examine whether Eps15R interacted directly with clathrin. Clathrin showed a stronger association with EGFP-Eps15R compared to EGFP-Eps15 in a co-immunoprecipitation assay from HeLa cell extracts ( Figure 4A). No clathrin was coimmunoprecipitated by control EGFP. AP2 was used as a control and was co-immunoprecipitated by both EGFP-Eps15R and EGFP-Eps15.
Thus, Eps15R can bind both clathrin and AP2 in vivo. We next mapped the region responsible for clathrin-binding using C-terminally truncated constructs of EGFP-Eps15R. Deletion of the ubiquitin interacting motif (UIM) domains (leaving aa 1-861) or UIMs plus the FIGURE 3 Legend on next page.
proline-rich region (leaving aa 1-747) did not impact on clathrin coimmunoprecipitation, whereas a construct (aa 1-596) additionally lacking the motif domain lost both clathrin and AP2 binding ( Figure 4B). To further narrow down the clathrin-binding site we used GST-tagged constructs comprising the motif domain and truncations thereof in a pull-down assay. The shortest construct that bound clathrin from HeLa cell extract was aa 717-729 ( Figure 4C). However, it should be noted that a slightly longer construct (aa 699-729) bound clathrin more efficiently ( Figure 4C). To identify a clathrin-binding motif in Eps15R, we used the shortest construct identified in Figure 4C and mutated individual amino acids (aa 718-728) to alanine residues. These mutated peptides were used in a GST pull-down assay. Our affinity purification assay using multiple peptides with alanine substitutions identified a binding motif between Eps15R and clathrin, DPFxxLDPF ( Figure 4D). An alignment of Eps15R orthologues showed that the DPFxxLDPF motif was conserved across vertebrates, and highlighted an additional LDPF motif starting at amino acid 700 that is only present in placental mammals ( Figure 4E, Figure S1, Supporting Information). In comparison, Eps15 does not contain any LDPF motifs. Pair-wise sequence alignment of Eps15R aa 699-729 with the corresponding DPF-containing region of Eps15 (aa 687-723) only showed a 14.6% identity. We compared the binding of clathrin to this region of Eps15R and Eps15 in a GST pulldown, and found that Eps15R but not Eps15 associated with clathrin ( Figure 4F). Next we investigated the significance of having two LDPF motifs in tandem in the Eps15R sequence. The first LDPF sequence (aa 699-709) did not show significant binding to clathrin on its own ( Figure 4C). We investigated whether this additional LDPF motif could explain the increased binding efficiency of Eps15R aa 699-729 compared to aa 717-729 ( Figure 4C). Indeed, mutation of F703A, F722A, or F728A in a longer Eps15R peptide (aa 699-729) reduced clathrin binding ( Figure 4G). The LDPF motif is a noncanonical clathrin motif, similar to the classic clathrin boxes DLL, LLxLD and PWDxW that contain hydrophobic residues and interact with clathrin heavy chain terminal domain. [45][46][47][48] Finally, to investigate whether the DPF motif that is conserved between the two LDPF motifs would also contribute to clathrin binding we compared two mutants, D707A and F709A from this DPF motif, to wild type Eps15R and found that F709A but not D707A reduced clathrin binding ( Figure 4H). In summary, having two LDPF motifs in tandem in

| Eps15R regulates AP2-clathrin complex formation in vitro and in vivo
To investigate whether the Eps15R F703A/F722A/F728A triple mutant (Eps15R mut) associated less readily with clathrin we performed a co-immmunoprecipitaiton assay. Immunoprecipitation of the mutant Eps15R (EGFP-Eps15Rmut) from cell lysates showed that less clathrin was co-precipitated compared to the wild-type protein ( Figure 5A). Quantification of the clathrin immunoblots from three independent experiments showed a halving of the amount of clathrin that was co-immunoprecipitated with Eps15Rmut compared to wild type protein ( Figure 5B). The amount of AP2 that coimmunoprecipitated with Eps15Rmut was also reduced, and we hypothesise that it was likely to be due to the reduced amount of clathrin that was accumulated since the high-affinity AP2-binding site (FxxF) in Eps15R was not compromised. To examine whether Eps15R may strengthen clathrin-AP2 interactions by providing an expanded binding platform for these proteins, we examined the effect on clathrin-binding to AP2-β2-appendage with its hinge (aa 616-937) in lysates from cells over-expressing wild type and mutant Eps15R ( Figure 5C). AP2 functions as an important adaptor for clathrin and it contains clathrin-binding motifs in its β2 hinge, and is known to be a key factor in regulating clathrin polymerisation. 23,36,49 To investigate the contribution of Eps15R to the clathrin-AP2 interaction and how efficiently the AP2-clathrin complex forms in the presence or absence of Eps15R we performed GST pull-downs with AP2 β2-hinge containing the clathrin-binding motif. Western blot analysis of the pull-down experiments indicated that a strong reduction of clathrin association with AP2-β2- FIGURE 4 Legend on next page.
appendage-hinge occurred in samples where mutant Eps15R was over-expressed ( Figure 5C-D). Similar amounts of wild type and mutant Eps15R were bound to AP2-β2-appendage-hinge and the expression of clathrin in the two cell extracts was similar ( Figure 5C, D). Thus, these experiments suggest that Eps15R regulates clathrin-AP2 interactions. The clathrin-AP2 interaction has been suggested to be the dominant mechanism that controls clathrin coat assembly, 23 and we hypothesise that Eps15R adds an additional layer of regulation to this mechanism.
Expression of the triple mutant EGFP-Eps15R (F703, F722, F728) in HeLa cells showed that it still targeted to the plasma membrane and co-localised with clathrin and AP2 ( Figure 5E-F). Thus, mutation of the LFPF motifs did not perturb the trafficking of Eps15R, and suggested that Eps15R is not recruited to clathrincoated pits by clathrin. This is not surprising based on the literature available describing the high avidity protein interactions for Eps15, AP2 and clathrin and furthermore the potential for heterodimerization between Eps15R and Eps15. 23,36 However, we did notice that our mutant Eps15R formed expanded punctae compared to wild type Eps15R ( Figure 5E-G), implying that it does have an impact on clathrin coat formation. Based on our biochemistry data we suggest that this is due to regulation of clathrin-AP2 interactions. Taken together these results suggest that Eps15R is not directly recruited by clathrin, but can regulate the maturation of clathrin-coated pits together with clathrin adaptor proteins such as AP2.

| The Eps15R-clathrin interaction is necessary for EphB2-mediated cell-cell repulsion
Finally, we wanted to examine the significance of the Eps15R-clathrin interaction in trans-endocytosis of EphB/ephrinB complexes. We again used the Co115 co-culture assay. First, we stably knocked down Eps15R in EphB2-expressing cells and co-cultured them with ephrinB1-expressing cells or control RFP-expressing cells ( Figure 6A-B). We quantified the cell pattern score as before, and showed that treatment with Eps15R shRNA resulted in a significant reduction in EphB2-mediated patterning ( Figure 6A, B). However, it should be noted that we did not observe a complete inhibition of cell-cell repulsion as the cells are not mixed to the same extent as the controls ( Figure 6A, B). These observations are in agreement with the slowing rather than complete inhibition of clathrin-mediated endocytosis that we demonstrated earlier by live cell imaging of CCPs ( Figure 3A, B).
Knockdown of Eps15 did not significantly alter the patterning compared to the control ( Figure 6A, B).  trafficking. 32,33 In this study we therefore instead used a targeted approach and employed the EphB2 interaction partner Numb 18   ephrinB ligands is known to be actin dependent. 10,11,16 In general, the actin cytoskeleton is involved in internalisation of larger endocytic intermediates such as macropinocytosis, phagocytosis and bacterial internalisation but is generally not associated with clathrinmediated endocytosis. 53,54 Because the EphB/ephrinB endocytic membrane invaginations are large assemblies containing plasma membranes from two neighbouring cells the requirement of force generated by the actin cytoskeleton could be compared to cells under high membrane tension or with a polarised membrane where clathrinmediated receptor endocytosis is actin-dependent. 55,56 It was recently shown that the mechanism of internalisation for soluble ephrinB1 ligand is different to that of membrane bound ligand, 16 which may explain the discrepancy between our results and data from other groups suggesting that trans-endocytosis is actindependent but clathrin-independent. 10,11 In our experiments we used co-cultures of cells expressing full-length EphB2 and ephrinB1. Moreover previous studies only assessed co-localisation of clathrin with EphB and ephrinB rather than perturbing clathrin interactions. We suggest that EphB trans-endocytosis is both actin-and clathrindependent.

| Plasmids
Rat AP2-β appendage + hinge (aa 616-937) was cloned into pGEX4T2. Fixed samples were imaged on a Zeiss 780 confocal microscope equipped with a 63× (1.4NA) and a 10× objective. Images within an experiment were collected using fixed laser settings and exposure times. Images were analysed using ImageJ software (National Institutes of Health), and the ClonalTools macro was used for pattern analysis as described previously. 25,26 For pattern analysis each image was analysed by randomly applying perpendicular lines to the image and the corrected patch width along the lines was quantified. The analysis of percentage of EGFP-positive cells was performed as in Cortina et al. 5 In brief, the number of EGFP-expressing cells growing in defined groups without mixing with RFP-expressing cells were quantified in randomly selected areas of images and expressed as percentage of the total number of cells in that area. Clustered cells were defined as the number of cells of one population localised together without the disruption of cells from the second population, here RFP-expressing cells.

| Statistical analysis
For microscopy-based experiments where multiple experiments were analysed the number of images for each sample were chosen to provide statistically significant data for each sample (20-50 images). For pairwise comparisons where the samples had a normal distribution an unpaired two-sided t test was used, while samples where the data was not normally distributed was analysed using the Mann-Whitney test. All statistical analysis was done using GraphPad Prism software.