A viral fusogen hijacks the actin cytoskeleton to drive cell-cell fusion

Cell-cell fusion, which is essential for tissue development and used by some viruses to form pathological syncytia, is typically driven by fusogenic membrane proteins with tall (>10 nm) ectodomains that undergo conformational changes to bring apposing membranes in close contact prior to fusion. Here we report that a viral fusogen with a short (<2 nm) ectodomain, the reptilian orthoreovirus p14, accomplishes the same task by hijacking the actin cytoskeleton. We show that the cytoplasmic domain of p14 triggers N-WASP-mediated assembly of a branched actin network, directly coupling local force generation with a short membrane-disruptive ectodomain. This work reveals that overcoming energetic barriers to cell-cell fusion does not require conformational changes of tall fusogens but can instead be driven by harnessing the host cytoskeleton. Impact Statement A viral fusogen drives cell-cell fusion by hijacking the actin machinery to directly couple actin assembly with a short fusogenic ectodomain.

A key feature of viral and cell-cell fusogens is their tall ectodomains, which in their metastable pre-fusion state typically extend more than 10 nm from the membrane. Since the plasma membrane of cells is densely decorated with glycoproteins and glycolipids that could sterically block membranes from getting close enough to fuse, the tall ectodomains of viral and cell-cell fusogens may allow them to reach across the membrane gap and anchor to the apposing membrane, involving insertion of a fusion peptide for Class I viral fusogens or a fusion loop for Class II (Harrison, 2015;Podbilewicz, 2014). Once the fusogen links the two membranes, conformational changes cause the fusogen to fold back, pulling the two membranes into close contact and forming a stable post-fusion structure that promotes membrane fusion (Bullough et al., 1994;Chen et al., 1999;Harrison, 2015;Ivanovic and Harrison, 2015;Pancera et al., 2014;Podbilewicz, 2014;Sapir et al., 2008;Wilson et al., 1981). This conformational change, together with fusogen oligomerization and local cooperativity, is believed to be sufficient to provide the energy required to overcome the repulsive hydration barrier, which prevents membranes from coming closer than~2 nm (Chernomordik and Kozlov, 2003;Harrison, 2015;Ivanovic and Harrison, 2015;Pérez-Vargas et al., 2014;Rand and Parsegian, 1989).
However, in other instances of cell-cell fusion, transmembrane proteins required for fusion are short by comparison and do not appear to undergo conformational changes, raising the question of how they bring two plasma membranes into close contact (Figure 1a). One example is the reptilian orthoreovirus fusion protein p14, a non-structural, single-pass transmembrane protein, that is expressed after viral entry. One of seven members of the FAST family of reovirus fusion proteins discovered by Duncan and colleagues (Ciechonska and Duncan, 2014;Dawe and Duncan, 2002;Duncan et al., 1995;Duncan et al., 2004;Guo et al., 2013;Kim et al., 2015;Racine et al., 2009;Shmulevitz et al., 2004;Shmulevitz and Duncan, 2000;Wilcox and Compans, 1982), p14 has a membrane-disruptive ectodomain that is necessary to drive cell-cell fusion but extends only 0.7-1.5 nm from the plasma membrane Corcoran et al., 2006). This short ectodomain has minimal secondary structure but is sufficient to disrupt opposing membranes either when added as a soluble fragment or when bare liposomes containing p14 are brought into close apposition through divalent cations Top et al., 2005). However, cell surfaces are crowded with proteins that sterically block close apposition, and it is unclear how p14 overcomes the energetic barrier of cell surface protein crowding, as well as the~2 nm repulsive hydration barrier, to drive cell-cell fusion (Chernomordik and Kozlov, 2003;Harrison, 2015;Rand and Parsegian, 1989). Yet, expression of p14 alone in cultured cells is sufficient to drive fusion with neighboring naïve cells Duncan et al., 2004).
To address the question of how p14 promotes close contact between cells and drives membrane fusion, we studied cell-cell fusion in HEK293T cells transiently expressing p14. We found that p14 drives cell-cell fusion by hijacking the host cell actin cytoskeleton. Using a phosphorylation-dependent motif in its cytoplasmic domain, p14 triggers N-WASP-mediated assembly of branched actin, directly coupling actin assembly with a short membrane-disruptive ectodomain. This work suggests that energetic barriers to cell-cell fusion can be overcome by harnessing force generated from local actin assembly, and it points to an alternate means of promoting cell-cell fusion in processes where tall fusogens have not been identified.

Results
Expression of p14 in HEK293T cells caused the cells to fuse with neighboring wild-type and p14expressing cells, forming large multinucleated syncytia over the course of 24 hr (Figure 1b, Video 1), like previous reports for other cell types . Partial cleavage of the p14 cytoplasmic tail, which also occurs during reptilian orthoreovirus infection (Top et al., 2009), liberates the C-terminus mCherry fluorescent tag from the transmembrane protein and serves as a convenient cytoplasmic marker of p14-expressing cells (Figure 1b, Figure 1-figure supplement 1a,b, and Video 2). At 12 hr post transfection, 5% of p14-expressing cells had 3 or more nuclei. At 24 hr post transfection, 34% of p14-expressing cells had 3 or nuclei (Figure 1c and d). In comparison, at 24 hr post transfection, no multinucleated cells were seen in cells transfected with mCherry alone (Figure 1-figure supplement 1c).
To quantify cytoplasmic mixing during p14-mediated cell-cell fusion in a high-throughput manner, we expressed the two halves of splitYFP in two populations of HEK293T cells and mixed the cells together. When fusion between cells of the two populations occurred, the two halves of splitYFP self-associated in the mixed cytoplasm and fluoresced, allowing quantification by a plate reader ( Figure 1e, and Figure 1-figure supplement 1d and e). Repeating the cell-cell fusion experiments above, the increase in YFP intensity in p14-expressing cells between 12 hr and 24 hr post transfection compared well with the increase in the percent of multinucleated cells, as quantified by counting nuclei (Figure 1f). Since YFP intensity is dependent on the percent of HEK293T cells expressing the two halves of splitYFP, the resultant YFP intensity was normalized to that of p14 WT and referred to as fusion index. While the ectodomain of p14 is shorter than typical viral fusogens, its cytoplasmic domain is comparatively long (68 amino acids). To determine how the cytoplasmic domain of p14 might be involved in cell-cell fusion, we first truncated Q70-I125 (p14 Dcyto), retaining a polybasic motif needed for trafficking to the plasma membrane (Parmar et al., 2014). Although p14 Dcyto was properly trafficked to the plasma membrane ( . This is consistent with previous findings , suggesting that p14 may be interacting with cellular components through its cytoplasmic tail to enable cell-cell fusion. We next investigated whether post-translational modification of the cytoplasmic tail of p14 might play a role in cell-cell fusion. The p14 cytoplasmic tail is mostly disordered but has several tyrosines that could be phosphorylated ( To determine if these tyrosines are indeed phosphorylated, we immunoprecipitated p14 and probed with an antiphosphotyrosine antibody, which confirmed p14 phosphorylation ( Figure 2b). Next, we mutated each predicted tyrosine to phenylalanine (Y59F, Y77F, Y96F, Y100F, Y116F) and found that only one mutation (Y116F) decreased cell-cell fusion ( Figure 2c and Figure 2-figure supplement 1b). We then used NetPhos3.1 and Scansite 4.0 to analyze the cytoplasmic tail of p14, and they predicted that Y116 is phosphorylated by c-src kinase (Figure 2d; Blom et al., 2004;Obenauer et al., 2003). To test this prediction, we mutated all other predicted phosphotyrosines of p14 to phenylalanine (YEY; Y59F, Y77F, Y96F, Y100F) and co-expressed it with constitutively active c-src mutant (CA c-src; Y527F). Y116 phosphorylation subsequently increased, indicating c-src phosphorylates p14 ( Figure 2e). Consistent with this, Y116 phosphorylation was also increased when tyrosine phosphatases were broadly inhibited by addition of pervanadate ( Figure 2e). Finally, to confirm that c-src is sufficient to phosphorylate Y116, we carried out a modified in vitro kinase assay using a peptide including Y116 (Y116-N121), along with CA c-src and kinase dead c-src (KD c-src; Y527F/K295R) mutants immunoprecipitated from HEK239T cells (Figure 2-figure supplement 1c and d). We found that CA c-src was sufficient to phosphorylate p14 cytoplasmic tail peptide, but KD c-src was not (Figure 2f), showing that c-src kinase is sufficient to phosphorylate Y116 in vitro. To verify this in live-cells, we expressed p14 WT in mouse embryonic fibroblasts lacking src, yes, and fyn kinase (SYF), as well as SYF cells supplemented with c-src (SYF + c-src). Despite lower transfection efficiency and reduced cell-cell fusion compared to HEK239T cells, SYF + c-src cells fused more than SYF cells ( Figure 2g). Altogether, this demonstrates that c-src phosphorylates Y116 in the p14 cytoplasmic tail to drive cell-cell fusion.
To determine which cellular components could be interacting with p14 upon phosphorylation, we used the Eukaryotic Linear Motif (ELM) prediction tool to identify potential binding motifs (Dinkel et al., 2016). ELM predicted that phosphorylated Y116 binds to the SH2 domain https://elifesciences.org/articles/51358#video1 of Grb2 as part of a Grb2 consensus-binding motif, YVNI ( Figure 3a). To test this prediction, we carried out a co-immunoprecipitation assay and confirmed that p14 binds to Grb2 ( Figure 3b). To determine if Grb2 binding is necessary for p14-mediated cell-cell fusion, we introduced two point mutations that disrupt the predicted Grb2 binding site both individually (Y116F, N118A) and together (FVAI; Y116F/ N118A) (Figure 3b). While trafficking to the plasma membrane was intact, all three mutants severely attenuated cell-cell fusion (Figure 3c, Figure 3-figure supplement 1a,b and c), suggesting that Grb2 is important for p14-mediated fusion. To confirm that p14 is sufficient to recruit Grb2, we conjugated biotinylated p14 cytoplasmic tail peptide to streptavidin beads in vitro and incubated them with purified Grb2 fluorescently labeled with AF647 ( Figure  Having shown that Grb2 binds to the p14 cytoplasmic tail in a phosphorylation-dependent manner, we next sought to determine mechanistically how Grb2, an adaptor protein with two SH3 domains, plays a role in p14-mediated cell-cell fusion. The N-terminal SH3 domain of Grb2 binds to SOS, activating Ras, which in turn activates Raf kinase and the MAPK-ERK1/2 pathway, while the C-terminal SH3 domain of Grb2 binds to the actin nucleation promoting factor N-WASP, which binds to Arp2/3 and nucleates branched actin assembly ( Figure 4a). To determine if one or both pathways are important for fusion, we first treated cells expressing p14 with sorafenib tosylate, an inhibitor of Raf kinase, but found no effect on cell-cell fusion at up to 100 times the IC 50 ( Figure 4b). We next considered whether branched actin networks could be directly involved in p14-mediated cell-cell fusion. Building on previous work showing that cytochalasin D disrupts fusion of p14-expressing cells (Salsman et al., 2008), we treated p14-expressing cells with wiskostatin, an inhibitor of N-WASP. We found that p14 trafficking to the plasma membrane fusion was unperturbed, but cell-cell fusion was significantly reduced ( Interestingly, treating p14-expressing cells with smifH2, an inhibitor of formins, enhanced cell-cell fusion (Figure 4b), perhaps due to increased branched actin assembly that has been observed when formins are broadly inhibited (Burke et al., 2014).
To test whether the reduction in cell-cell fusion with wiskostatin and CK-666 was the result of a direct link between N-WASP and p14 or a more general inhibition of actin activity, we overexpressed Grb2 mutants that can only bind to either SOS or N-WASP by truncating either the N-or C-terminal SH3 domains, respectively. We found that both of these Grb2 mutants bound to p14 WT and co-localized with phosphorylated p14 in pervanadate-treated live cell images (Figure 4-figure  supplement 1c), confirming that the mutations did not disrupt interactions with p14. However, when the Grb2 N-terminal mutant that can bind only to SOS (Nterm) was over-expressed, the extent of p14-mediated cell-cell fusion was reduced, similar to when endogenous Grb2 binding was reduced by overexpression of the Grb2 SH2 domain (Figure 4c, Figure 4-figure supplement 1d,e and f). When a Grb2 C-terminal mutant that can bind only to N-WASP (Cterm) was over-expressed, p14-Video 2. Confocal timelapse of a HEK293T expressing p14-mcherry (magenta) fusing with a WT HEK293T cell that appears dark. Plasma membrane is marked with gpi-anchored pHluorin (green). Scale bar is 5 mm.
https://elifesciences.org/articles/51358#video2 mediated cell-cell fusion was restored to a level comparable to that of endogenous Grb2 in WT cells To determine whether N-WASP binding to Grb2 and the p14 cytoplasmic tail is sufficient to nucleate actin assembly, we used an in vitro actin-based motility assay. We bound biotinylated p14 cytoplasmic tail peptides to streptavidin beads in a purified actin motility mixture containing N-WASP (lacking the EVH1 domain), Arp2/3, profilin, cofilin, capping protein and actin,  supplemented with Grb2 (Figure 4-figure supplement 1h). When Y116 of the p14 cytoplasmic tail peptide was phosphorylated, actin tails were nucleated from the bead (Figure 4e), but when Y116 was not phosphorylated, actin tails were not observed (Figure 4e). This confirms that Grb2 is necessary and sufficient to recruit N-WASP to the p14 cytoplasmic tail and can nucleate localized branched actin networks when p14 is present.      We next investigated whether branched actin network assembly must be directly coupled to the fusogenic ectodomain, or whether the fusogenic ectodomain can simply be present in the same membrane as actin assembly by the cytoplasmic tail of p14. To test the necessity for direct coupling, we co-expressed the ectodomain deletion mutant (Decto; DM1-T35), which traffics to the plasma membrane and binds Grb2 ( Figure 5-figure supplement 1a and b) and the cytoplasmic tail deletion mutant (Dcyto) in the same cell. Interestingly, we found that cell-cell fusion was abolished ( Figure 5a, and Figure 5-figure supplement 1c), despite the presence of both halves of p14. This indicates that actin assembly localized to the p14 ectodomain is necessary to drive fusion. This also suggests why native Arp2/3-generated branched actin networks at cell-cell contacts (Efimova and Svitkina, 2018) are not sufficient to cause spontaneous cell-cell fusion. We note that the mechanism described here is in addition to the role of the actin cytoskeleton at E-cadherin-mediated cell-cell adhesion sites (Salsman et al., 2008).      If actin assembly directly coupled to p14 is necessary and sufficient for the fusogenic activity of p14, then it should be possible to drive cell-cell fusion by creating an alternate link between p14 and branched actin network assembly. To test this idea, we engineered a fusion protein that binds to p14 consisting of Grb2 SH2 domain and a 47-residue peptide from EspF U of enterohemorrhagic E. coli (EHEC) (Figure 5b). This 47-residue peptide binds to and relieves the auto-inhibition of endogenous N-WASP, nucleating branched actin network (Cheng et al., 2008;Sallee et al., 2008). Fusion of this peptide with the Grb2 SH2 domain, which we confirmed binds to phosphorylated p14 (Figure 5-figure supplement 1d), enables binding to WT p14. When we expressed this fusion protein (SH2-R47) together with p14 in HEK293T, cell-cell fusion was significantly higher than when the SH2 domain lacking R47 was expressed to a similar level (Figure 5c, Figure 5-figure supplement 1e,f  and g). This result demonstrates that direct coupling of branched actin assembly to p14 is necessary and sufficient for fusion.

Discussion
This study reveals how a viral pathogen hijacks branched actin network assembly to drive cell-cell fusion, reminiscent of how the pathogens Listeria monocytogenes and vaccinia virus hijack branched actin network assembly to move within and between cells (Welch and Way, 2013). Here we show that the reptilian orthoreovirus fusogen p14 accomplishes this by presenting a c-src kinase substrate in its cytoplasmic tail, binding the host cell adaptor protein Grb2, and initiating branched actin assembly through the host nucleation promoting factor N-WASP. Since p14 ectodomain extends only 0.7-1.5 nm from the plasma membrane (Corcoran et al., 2006), it cannot interact with apposing membrane through the~2 nm repulsive hydration barrier. If that barrier can be overcome, for example by using divalent cations to bring bare membranes containing negatively charged PS together in vitro, the p14 ectodomain is able to disrupt and mix membranes (Top et al., 2005). However, it is unclear how such close membrane-membrane interaction is achieved on the crowded plasma membrane where the lipid bilayer is sterically blocked by dense membrane proteins.
We propose that the primary role of branched actin network assembly by p14 is to physically push the fusogenic ectodomain into the opposing plasma membrane, a step normally carried out by conformational changes of tall fusogens. In this model, localized force generation by branched actin network assembly around p14's cytoplasmic domain drives close membrane apposition by coupling to and pushing against the cortical actin cytoskeleton, much like other actin-based membrane protrusions. This close apposition in the presence of p14's membrane-disruptive ectodomain, which contains hydrophobic residues and myristoylation that are known to be necessary for fusion Top et al., 2009), may be sufficient to drive cell-cell fusion. Several other roles of the actin cytoskeleton may also be important for p14-bsaed cell-cell fusion. The actin cytoskeleton could promote clustering of p14 ectodomains into sub-diffraction-limited domains (Kö ster and Mayor, 2016), although enrichment of p14 or phosphorylated p14 was not readily observed at sites of fusion ( Figure 5-figure supplement 2a,b,c and d). The actin cytoskeleton could also change local membrane curvature and tension (Kozlov and Chernomordik, 2015), block fusion pore closure, and/or actively expand the fusion pore (Kozlov and Chernomordik, 2015), all of which would promote cell-cell fusion by p14.
Membrane fusion driven by other fusogens has long implicated the actin cytoskeleton (Kondo et al., 2015;Sanderson et al., 1998). The proposed roles of the actin cytoskeleton include regulating the activity, localization and enrichment of fusogens at fusion sites (Yang et al., 2017), regulating protein-protein interactions (Wakimoto et al., 2013), and enlarging the fusion pore (Wurth et al., 2010;Zheng and Chang, 1990). In cases where the putative fusogen remains unidentified or the mechanism of fusion remains unclear, the actin cytoskeleton has been suggested as a key player. Examples include formation of signaling scaffolds and protrusive structures during osteoclast fusion (Oikawa et al., 2012), macrophage fusion (Faust et al., 2019), and myoblast fusion in Drosophila Sens et al., 2010). Protrusive membrane structures generated by branched actin assembly have also been observed at fusogenic synapses (Shilagardi et al., 2013) and can be formed in vitro (Liu et al., 2008a;Simon et al., 2019).
Our findings may be relevant to putative fusogens like myomixer, a short extracellular peptide that is required for myoblast fusion (Bi et al., 2018;Leikina et al., 2018;Millay et al., 2013;Quinn et al., 2017;Sampath et al., 2018;Zhang et al., 2017) and does not resemble conformational-change based fusogens. Like p14, myomixer, together with its partner myomaker, could harness the actin cytoskeleton to drive cell-cell fusion by physically forcing a short fusogenic ectodomain through dense cell surface proteins and into contact with an apposing membrane. In this model of cell-cell fusion, local forces generated by the cytoskeleton, rather than by conformational changes in a tall fusogen, provide a powerful tool for cells and pathogens to create multinucleated structures.
Constitutively active chick-src kinase was amplified from pLNCX chick src Y527F and inserted into pcDNA 3.1 with linker (GGGS) and C-terminus tags (FLAG and mTagBFP2). pLNCX chick src Y527F was a gift from Joan Brugge (Addgene plasmid # 13660). K295R was introduced to constitutive active chick-src kinase to render it kinase dead with primers. cDNA of Human Grb2 (GE Dharmacon, cloneID: 3345524) was amplified and inserted into pGEX4T2 with a N-terminus GST tag and TEV cleavage site for purification of Grb2. For overexpression of Grb2 mutants, IRES Puromycin was amplified from pLV-EF1a-IRES-Puro (a gift from Tobias Meyer, Addgene plasmid #85132) and inserted into lentiviral pHR backbone to create pHR-IRES-Puro. Grb2 N-terminus SH3 domain and SH2 domain (N-termSH3, 1-159), Grb2 C-terminus SH3 domain and SH2 domain , and Grb2 SH2 domain (58-159) were amplified from cDNA of Human Grb2 and inserted into pHR-IRES-Puro with C terminus FLAG tag. 47 residues from EspF U of enterohemorrhagic E. coli (268-314) was synthesized and inserted with GGGS linker downstream of Grb2 SH2 domain (58-159) and FLAG tag into pHR-IRES-Puro.
Cell culture, transfection and generation of mutant Grb2 overexpression cells HEK293T cells were obtained from UCSF Cell Culture Facility. SYF (CRL-2459) and SYF + c-src (CRL-2498) cells were obtained from ATCC. N-WASP -/-and +/+ mouse embryonic fibroblasts were a kind gift from Scott Snapper. All cells were grown in DMEM (Life Technologies) supplemented with 10% heat-inactivated FBS (Life Technologies) and 1% Pen-Strep (Life Technologies), at 37˚C, 5% CO 2 . Cells were negative for mycoplasma as verified with Mycoalert mycoplasma detection kit (Lonza).
Cells were transfected with TransIT-293 (Mirus Bio) according to manufacturer's instructions.
HEK293T cells were transduced with lentivirus, and 24 hr post transduction selected with 3 mg/ml puromycin (Clontech) to select for mutant Grb2 expression. Cultures were maintained in 3 mg/ml puromycin.
splitYFP cell-cell fusion assay pHR-splitYFPa and pHR-splitYFPb were co-transfected with second generation packaging plasmids, pMD2.G and p8.91 in HEK293T to generate lentivirus. WT HEK293T cells were transduced with splitYFPa and splitYFPb lentivirus. The cells were passaged for at least a week before use in cell-cell fusion assay.
To quantify cell-cell fusion, HEK293T cells stably expressing splitYFPa and splitYFPb were mixed at 50:50 ratios and 1.33 Â 10 5 of cells were plated into each well of 48 well plate. The next day, the cells were transfected with . 18 hr post transfection, cells were moved to 30C , 5% CO 2 incubator to mature the splitYFP fluorophore. 24 hr post transfection, cells were lifted with 150 ml of 2 mM EDTA and placed into 96 well black bottom plate. splitYFP was excited at 510 nm and emission at 530 nm was quantified using a plate reader (Tecan).
Fusion index was quantified as (I sample -I cell ) /(I p14WT -I cell ), where I cell is the YFP intensity of nontransfected HEK293T cells expressing splitYFPa and splitYFPb, I sample is the YFP intensity of HEK293T cells transfected with plasmid as specified, I p14WT is the YFP intensity of HEK293T cells transfected with p14 WT and treated with DMSO as vehicle control. Average and standard deviation of fusion index is calculated from 3 independent transfections of 3 wells each. Statistical significance was determined using two-tailed, two-sample Student's t-test.

Nuclei count
3.8 Â 10 5 cells were plated into 24 well plates and transfected the next day with designated plasmid with TransIT-293 (Mirus Bio) according to manufacturer's instructions. 2 hr post transfection, cells were lifted with 150 ml of 2 mM EDTA, re-suspended with 850 ml of media, and 300 ml of cell suspension was plated onto a fibronectin-coated glass bottom chamber (Cell-vis). At 6 hr and 18 hr post transfection, cells were transferred to 30˚C, 5% CO 2 incubator. After 6 hr incubation at 30˚C, nuclei were labeled with 0.6% Hoechst 33342 (Life Technologies), and plasma membrane were labeled with 0.05% CellMask Deep Red (Thermo Fisher Scientific) for 20 min at 37˚C. Cells were imaged using spinning disk confocal microscopy. About 80-100 random field of views are taken for each sample to image almost the entire imaging well, and the number of nuclei in p14 expressing cells are manually counted. Average and standard deviation of binned nuclei count is calculated from 3 independent transfections. Statistical significance was determined using two-tailed, two-sample Student's t-test. Statistical significance between distributions of nuclei in p14-expressing cells were determined using Kolmogorov-Smirnov test (ks-test).

Drug treatment
To broadly inhibit the protein-tyrosine phosphatases, pervanadate is prepared by incubating 10 mM sodium orthovanadate with 0.15% hydrogen peroxide in 20 mM HEPES for 5 min at room temperature. Pervanadate is neutralized with catalase and added to cells immediately. For western blot, cells were lysed 10 mins after pervanadate addition, for live-imaging, cells were imaged immediately after pervanadate addition.
To perturb the actin cytoskeleton, 4 hr post transfection, the media was replaced with complete media supplemented with cytoskeletal drugs CK-666 (Sigma Aldrich), Wiskostatin (Krackeler Scientific) and smifH2 (EMD Millipore) at specified concentrations. DMSO was used as vehicle control. 18 hr post transfection, splitYFP was matured at 30˚C. At 24 hr post transfection splitYFP fluorescence was quantified as described above.
To inhibit Raf kinase, 4 hr post transfection, the media was replaced with compete media supplemented with sorafenib tosylate (Selleckchem). DMSO was used as a vehicle control. 18 hr post transfection, splitYFP was matured at 30˚C. At 24 hr post transfection splitYFP fluorescence was quantified as described above.

Surface biotinylation
HEK293T cells were transfected with GFP tagged p14 WT, p14 FVAI, p14 Dcyto, p14 Decto. P14 WT expressing cells were treated with 20 mM Wiskostatin or 100 mM CK-666 9 hr post transfection. 18-24 hr post transfection, cells were washed with 1 mM CaCl2/PBS on ice, lifted off the dish with icecold 2 mM EDTA/PBS. Cells were pelleted at 200 rcf for 5 min and washed with PBS (pH = 8) three times on ice and incubated with 1 mg NHS-Biotin in 0.5 mL PBS (pH = 8) for 1 hr on ice. The biotinylation reaction was quenched with 100 mM glycine in PBS for 10 min at room temperature. Cells were washed three times with 100 mM glycine in PBS and lysed with 1 mL lysis buffer (1x RIPA buffer supplemented with 1x HALT protease inhibitor (Thermo Fisher Scientific)) for 30 min on ice and sonicated for 3 min. Cell debris was pelleted for 5 min at 17,900 rcf at 4˚C. 30 ml of streptavidin magnetic beads (ThermoFisher) were washed with RIPA buffer at 4˚C, the supernatant was added and incubated overnight. Beads were washed 5 times with RIPA buffer, suspended in Laemmli sample buffer (30 ul), denatured at 95˚C for 5 min and separated on 4-20% acrylamide gradient gels by SDS-PAGE. Proteins were transferred onto nitrocellulose membrane with iBlot (Thermo Fisher Scientific). The membrane was blocked with 5% milk powder in TBST for 1 hr and probed with primary antibodies a-GFP (1:5000, g1544, Sigma Aldrich), a-tubulin (1:5000, Clone YL1/2, Thermo) in 5% milk in TBST overnight at 4˚C. The membrane was then probed with secondary antibodies, a-rabbit HRP (1:5000, 65-6120, Thermo Fisher) and, a-rat AlexaFluor 647 (1:5000, Life Technologies). Western blots were imaged on ChemiDoc (Bio-Rad).

Protein purification
GST-TEV-Grb2 (human) was expressed and purified from E. coli as previously described (Su et al., 2016).
N-WASP (DEVH1) was a kind gift from D. Wong and J. Taunton (University of California, San Francisco). Actin was purified from rabbit skeletal muscle as previously described (Spudich and Watt, 1971). Capping protein was a kind gift from S. Hansen and D. Mullins (University of California, San Francisco). Arp2/3 was purchased from Cytoskeleton, Inc Profilin, cofilin, and utrophin actin binding domain (1-261) were purified as previously described (Bieling et al., 2016;Harris et al., 2019).