Non-catalytic role of phosphoinositide 3-kinase in mesenchymal cell migration through non-canonical induction of p85β/AP-2-mediated endocytosis

Class IA phosphoinositide 3-kinase (PI3K) galvanizes fundamental cellular processes such as migration, proliferation, and differentiation. To enable multifaceted roles, the catalytic subunit p110 utilizes a multi-domain, regulatory subunit p85 through its inter SH2 domain (iSH2). In cell migration, their product PI(3,4,5)P3 generates locomotive activity. While non-catalytic roles are also implicated, underlying mechanisms and its relationship to PI(3,4,5)P3 signaling remain elusive. Here, we report that a disordered region of iSH2 contains previously uncharacterized AP-2 binding motifs which can trigger clathrin and dynamin-mediated endocytosis independent of PI3K catalytic activity. The AP-2 binding motif mutants of p85 aberrantly accumulate at focal adhesions and upregulate both velocity and persistency in fibroblast migration. We thus propose the dual functionality of PI3K in the control of cell motility, catalytic and non-catalytic, arising distinctly from juxtaposed regions within iSH2.

The catalytic activity of PI3K is one of the major positive regulators in cell migration. In amoeboid cells such as Dictyostelium discoideum and mammalian neutrophils, chemoattractant induces PI(3,4,5)P 3 accumulation at the front of cells 17-19 , leading to the activation of the Rho family of small GTPases including Rac1 [19][20][21] and cell protrusions driven by the actin cytoskeleton. Mesenchymal cells such as broblasts also establish similar PI(3,4,5)P 3 polarity 22 . However, a recent study found that PI3K in broblasts acts as an ampli er of nascent lamellipodia instead of an initiator of protrusion 23 . Further research found that this PI3K-actin feedback loop originates from nascent adhesions, another unique feature of mesenchymal cell migration 24 . Therefore, amoeboid and mesenchymal cells utilize distinct mechanisms, at least at the level of PI3K, with yet elusive mechanisms.
In the face of the catalytic-role-centric studies, non-catalytic roles of p85 have also been reported. In ER stress response, p85 brings XBP-1s to the nucleus to upregulate unfolding protein response genes 25,26 . p85 also involves in receptor internalization through the interaction with an adaptor molecule insulin receptor substrate 1 (IRS-1), Rab GTPases activation, or ubiquitination on p85 itself [27][28][29] . In addition, p85 regulates cytoskeletal reorganization in concert with the small GTPase Cdc42 30,31 . It therefore is important to consider PI3K as a multifaceted molecule to fully understand its functions and regulations.
In this study, we combine bioinformatics and chemical biology approaches with live-cell uorescence imaging to reveal a previously uncharacterized non-catalytic function of PI3K in which a part of the p85β iSH2 domain induces endocytosis mediated by clathrin and dynamin. Using p85 knockout cells with genetic rescues, we show that this non-catalytic induction of endocytosis regulates cell migration properties through local regulation of p85 at focal adhesions.

Results
iSH2 domain of regulatory subunit p85 has AP-2 binding motifs To explore possible non-catalytic roles of PI3K, we analyzed the primary sequence of the regulatory subunits of class IA PI3K (p85α, p85β, and p55γ). Using Eukaryotic Linear Motif (ELM) prediction 32 , we found that iSH2 domain of the C-terminal region of p85β accommodates three consensus binding motifs for AP-2 33 , an adaptor protein for clathrin-mediated endocytosis, namely YxxΦ, di-leucine, and acidic clusters (Fig. 1a, Extended Data Fig. 1). Consistent with the crystal structure of p110 complexed with iSH2-cSH2 16 , the C-terminal region of iSH2 was predicted to be intrinsically disordered and unlikely a part of secondary structures based on primary sequence analysis of IUPred2A 34 , PrDOS 35 , and PONDR 36 (Extended Data Fig. 1). These results suggested possible interaction between p85 and AP-2, which could lead to endocytosis upon their membrane targeting.

Plasma membrane recruitment of iSH2 domain induces endocytosis
Whether a given molecule is capable of inducing endocytosis can be tested by recruiting such molecules to plasma membranes 37,38 . With the help of a chemically inducible dimerization (CID) system 39 , we aimed to recruit iSH2 including the putative AP-2 binding motifs to the plasma membrane and see if this results in endocytosis. To achieve this, we used rapamycin-dependent heterodimerization of FK506binding protein (FKBP) and FK506-rapamycin-binding domain (FRB) to trap YFP-FKBP-iSH2 (YF-iSH2) at plasma membrane-anchored Lyn-CFP-FRB (Lyn-CR). Within several minutes after accumulation of YF-iSH2 at the plasma membrane, numerous mobile puncta became visible in the cytosol (Fig. 1b, Supplementary movie 1-3). The puncta were seen only with YF-iSH2 but not with a negative control YFP-FKBP (YF), suggesting that iSH2 is responsible for induction of puncta derived from the plasma membrane.
We then tested colocalization between the observed puncta and markers for endocytosis. When we used a membrane staining dye mCLING 40 , which gets internalized to endomembranes upon endocytosis, the puncta colocalized well with the dye (Extended Data Fig. 2).
Endocytic activity is highly sensitive to ambient temperature, likely due to critical involvement of dynamin GTPase which has an unusually high Q 10 temperature coe cient value 41,42 . When conducting iSH2 recruitment to the plasma membrane at a reduced temperature (37°C to 23°C), we observed much fewer puncta (Extended Data Fig. 3, Supplementary movies 1-3). This is consistent with the lack of documentation of such puncta upon iSH2 recruitment by our group and others in the past [43][44][45][46] . Collectively, these results strongly support the idea that membrane-recruited iSH2 induces endocytosis.

iSH2-mediated endocytosis is context independent
To test how well the iSH2-mediated endocytosis can be generalized, we repeated the CID recruitment assay with two modi cations. First, we used FRB anchored to the plasma membrane through six different targeting sequences (Supplementary Table 2). In all cases except KRas4B-CAAX, we observed puncta formation (Extended Data Fig. 4a, b). Furthermore, the endocytosis can be also triggered by a light inducible dimerization system (iLID-SspB) 47 (Extended Data Fig. 4c). Thus, iSH2-mediated endocytosis is not speci c to a certain type of plasma membrane targeting or dimerization scheme.

iSH2-mediated endocytosis depends on the AP-2 binding motifs
To determine if the predicted AP-2 binding motifs are necessary for iSH2-mediated endocytosis, we deleted 12 amino acids (aa) within the motif clusters (∆motif) or replaced the same region with a 3×SAGG exible linker (motifGS). When the recruitment assay was conducted with each of these iSH2 mutants, we saw little to no puncta, indicating the necessity of the 12 aa for inducing endocytosis (Fig. 1d, Extended Data Fig. 5). Then, we individually mutated the YxxΦ motif, di-leucine motif, and acidic cluster. Whereas point mutations in the di-leucine motifs drastically decreased endocytic activity, Y to A mutation in the YxxΦ motif did not show signi cant effect (Fig. 1d, Extended Data Fig. 5). Replacement of the acidic cluster EDEDA with GSAGG partially reduced the endocytic activity (Fig. 1d, Extended Data Fig. 5). These results suggest that the di-leucine motif and acidic clusters contribute to iSH2-mediated endocytosis.

iSH2-mediated endocytosis depends on clathrin and dynamin
To understand molecular mechanisms of iSH2-mediated endocytosis, we examined possible association between iSH2 and AP-2 by applying an inducible co-recruitment assay 48,49 (Extended Data Fig. 6a). In this assay, we can semi-quantitatively assess a protein-protein interaction in living cells. Here, we recruit an iSH2 domain to the plasma membrane using the chemically inducible dimerization scheme, and measure how much a bait protein, AP-2, gets co-recruited under TIRF microscopy. After recruitment of YFP-FKBP-labelled iSH2 to the plasma membrane, we observed an increase in the uorescence intensity of AP-2-mCherry (co-recruitment index, CI: 1.23), but not mCherry control construct (CI: 1.03) (Extended Data Fig. 6b, c), implying that iSH2 and AP-2 could interact with each other. This AP-2 co-recruitment was reduced when we used iSH2 motif mutants, ∆motif (CI: 1.07) and motifGS (CI: 1.20) (Extended Data Fig. 6b,c). Similarly, we measured an extent of colocalization between AP-2 and iSH2 after recruitment of iSH2 to the plasma membrane. As a result, AP-2 uorescence signals on the plasma membrane colocalized with the membrane-recruited iSH2, but not with the motif mutant (Fig. 1e, Extended Data Fig. 6d, e). These results suggested that the AP-2 binding motif of p85 binds to and colocalizes with AP-2 on the plasma membrane.
Interestingly, colocalization of iSH2 and AP-2 was also observed when FRB-CFP-CAAX(KRas4B) was used as a plasma membrane anchor (Extended Data Fig. 6d, e), despite the poor endocytosis induction of CAAX(KRas4B) (Extended Data Fig. 4a, b). This result suggested that while iSH2 interacts with AP-2 regardless of the type of plasma membrane anchor, endocytic development including vesicle maturation and membrane remodeling were somehow stalled in the case of KRas4B-CAAX.
LY294002 binds to the ATP binding pocket of p110 and inhibit its catalytic function 58 , whereas iSH2-DN mutation abolishes iSH2-p110 interaction 59 . When we performed the iSH2 recruitment assay in the presence of either of these reagents, puncta formation occurred normally despite the production of PI(3,4,5)P 3 being suppressed in the same cells (Fig. 2a, Extended Data Fig. 7a). This indicates that iSH2-mediated endocytosis is independent of the p110 kinase activity and can be classi ed as a non-catalytic function of PI3K.
iSH2-mediated endocytosis is β isoform speci c The iSH2 domain is de ned in all three regulatory subunits of class IA PI3K (p85α, p85β, and p50γ) 1 . We then took iSH2 domains from different isoforms of human and mouse and asked if iSH2-mediated endocytosis is conserved among them by using the CID recruitment assay. iSH2 from p85β (both human and mouse) induced endocytosis, but α or γ isoforms did not (Fig. 2b, Extended Data Fig. 7b), indicating that endocytic activity is β isoform speci c. The mechanism of this isoform speci city is unknown, but slight sequential or structural differences may be involved as in the case of the reported isoform-speci c binding to In uenza A virus NS1 protein 60-62 .

aa disordered region is necessary and su cient for iSH2-mediated endocytosis
The iSH2 domain has been considered as a single domain whose main role is to bind to p110 and bring the catalytic subunit to the plasma membrane upon receptor stimulation. To locate exactly which part of iSH2 contributes to p110 binding, and which part contributes to the endocytosis induction, we performed a sequential truncation to the iSH2 domain. As a result, the C-terminal 46 aa was found to be both necessary and su cient to induce the endocytosis (

Generation of MEF cell lines with p85β AP-2 binding motif mutants and their biochemical characterization
To investigate how the unexpected link between p85β and AP-2 in uences the cellular functions of PI3K, we took an advantage of p85α/β double knock out (DKO) in mouse embryonic broblasts (MEFs) 63 to which a series of p85 variants, with or without mutations in AP-2 binding motifs, were individually introduced via lentiviral infection (Extended Data Fig. 8a). Since both the di-leucine motif and the acidic cluster contribute to endocytic activity (Fig. 1d), we created two p85β mutants whose 12 aa motif region was either truncated or replaced with 3×SAGG, serving as AP-2 motif de cient forms of p85β. YFP was tagged on the rescued p85 to sort the virus-infected cells and validated the consistency in the expression level of rescued p85 variants (Extended Data Fig. 8b).
Using these genetic resources, we rst assessed a possible regulatory role of the AP-2 binding motif in a receptor tyrosine kinase pathway ( Fig. 3a). Consistent with a previous report 63 , expression of wild type p85β in DKO MEFs could rescue the elevated levels of Akt phosphorylation (pTyr-308) in response to PDGF addition (Fig. 3b). When we tested this with the mutant p85β cell lines, there was no signi cant difference from the wild type. In assessing cell proliferation, we then found similar proliferation rates for cells rescued with wild type and motifGS mutant (Fig. 3c). Thus, mutations in the AP-2 binding motif of p85β did not show an apparent effect on Akt response or cell growth. Considering the possibility that AP-2 binding of p85β regulates receptor internalization, we next measured the effect on ERK, the other major pathway regulated by endocytic tra c of receptor tyrosine kinase (RTK) 64 . However, wild type and mutant rescued cells showed a similar pattern in ERK response (Extended Data Fig. 8c). We also tested the effect on transferrin receptors, a typical cargo of clathrin-dynamin endocytosis, and found no signi cant change in transferrin internalization between wild type and mutant rescue cells (Extended Data Fig. 8d). Therefore, the binding between p85β and AP-2 did not seem to in uence on RTK signaling or general endocytic functions.
Mutations in AP-2 binding motif causes localization of p85β at focal adhesions Besides the RTK response, PI3K locally controls cellular morphodynamics in association with focal adhesions 24,30,65,66 . To determine if AP-2 binding motifs are involved in such subcellular regulation, we next investigated the intracellular localization of wild type and mutant p85β using confocal microscopy. Strikingly, the 3×SAGG and ∆motif p85 cell lines showed signi cantly enhanced accumulation at focal adhesions (Fig. 3d). Previous studies found that p85 localizes to focal adhesions where it binds to focal adhesion kinase (FAK) through the interaction between its SH3 domain and auto-phosphorylated tyrosine of FAK (pY397) 65, 67-70 . We thus tested the effect of the AP-2 motif mutation on FAK. Western blot analysis did not detect signi cant differences in the expression or phosphorylation level of FAK among the p85-rescued cell lines (Fig. 3e). Using TIRF microscopy, we further performed live-cell imaging of p85 fused to YFP which was co-expressed with a focal adhesion marker mCerulean3-Paxillin 71 in the presence or absence of an FAK inhibitor PF-573228 72 . The results showed that both wild type and mutant p85 dissociated from focal adhesions after FAK inhibition with identical kinetics (Fig. 3f, Extended Data 9). Together, the data suggest that AP-2 binding motifs are involved in sequestration of p85β from focal adhesions. Since the observed sequestration did not affect the interaction between the SH3 domain of p85β and pY397 of FAK, there is another mechanism underlying a trigger of the sequestration.
Fibroblasts with impaired AP-2 binding motifs migrate faster and more persistently Focal adhesions function as a molecular clutch for a cell to transmit mechanical force to the external environment 73 , while simultaneously serving as a biochemical hub for PI3K-Rho GTPase-actin to extend lamellipodial protrusion 24,66 . Since mutation in AP-2 binding motifs altered localization of p85β at focal adhesions, we hypothesized that AP-2 binding motifs regulate cell migration through focal adhesions. To test this, we characterized migratory properties in a series of DKO MEFs in the presence of 10% FBS to trigger random migration (Fig. 4a, Extended Data Fig. 10a). DKO MEFs exhibited slower migration speed than wild type counterpart MEFs (Fig. 4b), consistent with the reduced Rac activity and less lamellipodia formation in the knockout cells 63 . Interestingly, rescuing the DKO cell line with wild type p85β further decreased migration speed (Fig. 4b, c). In contrast, the cells rescued with AP-2 binding motif mutants of p85β or p85α did not show the decrement, suggesting that the AP-2 motif negatively regulates migration (Fig. 4b, c, Extended Data Fig. 10).
Dominant negative mutation of p85 (DN), which lacks 470 to 504 aa residues necessary for p110 binding and decouples catalytic activity of PI3K from receptor activation 59 , and pharmacological inhibition of PI3K and FAK completely suppressed the migration. This basal level of migration was signi cantly lower than the migration activity of wild type p85β-rescued cells (Fig. 4b, Extended Data Fig. 10a, b). These results suggest that p85β has two layers of regulations on cell migration: positive regulation through PI3K catalytic product, PI(3,4,5)P 3 and negative regulation through AP-2-mediated sequestration of p85β from focal adhesions.
We then calculated persistence ratio of cell motility de ned as the ratio between displacement (d) and the total path length (D), which decreased over the course of migration assays. The decrease in wild type p85β-rescued cells was more prominent over time than mutant p85-rescued cells, suggesting that the link between p85 and AP-2 is involved in a negative regulation of cell migration with a temporal delay from PI(3,4,5)P 3 -mediated positive regulation (Fig. 4d, Extended Data Fig. 10c). Difference in migration speed between wild-type p85 rescue cells and AP-2 motif mutant rescue cells was also seen with PDGF as a stimulant, instead of FBS (Extended Data Fig. 10d), suggesting that the AP-2-mediated motility control is at play under growth factor signaling.

Role of the AP-2 binding motif in chemotaxis
To test migration behavior in a physiologically relevant context, we performed chemotaxis assays where cells are guided to migrate in a directed manner according to a chemoattractant gradient (Fig. 4e). In line with the random migration results, p85β-rescued cells migrated more slowly than that of DKO, p85α-rescued, and p85β motif mutant-rescued cells (Fig. 4f, g). Although the persistent ratio drew slightly different curves from those of random migration, wild type p85β-rescued cells consistently showed the least persistency among the tested cells (Fig. 4h). These data support the negative regulation of chemotaxis by the AP-2-mediated endocytosis. To examine its role in gradient sensing during chemotaxis, we quanti ed the forward migration index (FMI) de ned as a ratio between forward displacement (y) and the total path length (D) (Fig. 4i). As a result, there was no signi cant difference in FMI among the conditions tested; wild type cells, DKO cells, and DKO cells rescued with p85α, p85β, or p85β-motifGS (Fig. 4j). These data suggest that the AP-2-mediated endocytosis downregulates migration properties such as speed and persistence, but not gradient sensing, during chemotaxis.

Discussion
The iSH2 domain is characterized as a positive regulator of PI3K since it stabilizes and recruits the catalytic subunit p110 to the plasma membrane 74 . Our present study demonstrates that the iSH2 domain of p85β has concurrent negative regulation of cell migration through AP-2-mediated endocytosis which originates from the C-terminal disordered region. Disruption of this linkage between p85β and AP-2 led to abnormal accumulation of p85β at focal adhesions (Fig. 3) and also increased speed and persistency of cell migration (Fig. 4). Based on these ndings, we propose that the iSH2 domain, originally assigned as a single domain for a single function, consists of two parts with distinct, antagonistic functions: the p110 binding coiled-coil region to promote cell migration, and the AP-2 motif-encoding disordered region to induce endocytosis for negative regulation of cell migration. One may wonder why PI3K elicits two opposing signals for cell motility control. Such a seemingly meaningless regulation may be explained by the kinetic difference. Upon stimulation, PI(3,4,5)P 3 production can initiate within milli-seconds to seconds timescale 75 , while clathrin-mediated endocytosis occurs more gradually (tens of seconds to a few minutes) 76 . The temporal difference creates an autonomous delayed negative feedback loop, which is one of the signature characteristics necessary for self-organized signal transduction often proposed in directed cell migration 77 . Thus, for PI3K to send out counteracting signals of different kinetics may be of importance for this intricate cell function.
We also determined that AP-2 motif regulates p85β localization at focal adhesions. Since cell protrusion signaling consisting of PI3K and actin is closely coupled with cell adhesions 66,23,24 , sequestration of PI3K from focal adhesions could act as a negative regulator of chemotaxis. Considering that mutations to the AP-2 binding motif did not affect the expression level or FAK phosphorylation (Fig. 3), the p85-mediated endocytosis likely regulates the signals downstream of PI3K without drastically altering molecular composition of the focal adhesions. Interestingly, under PDGF stimulation, mutations in the AP-2 binding motif increased cell migration speed without affecting other major pathway effectors such as Akt and ERK (Fig. 3b, c, Extended Data Fig. 8c, 10d). How does AP-2-mediated regulation discriminate a speci c signaling molecule from others? Two interesting observations may be of help to answer this question -the AP-2 binding motif resides within the intrinsically disordered region (Extended Data Fig. 1), and many of the membrane anchors that led to the p85-mediated endocytosis (Extended Data Fig. 4) colocalize with ordered lipid domains. Both properties are known to form unique molecular organizations such as liquid droplets and lipid rafts. It is thus intriguing to speculate that it is this unique lipid-protein interaction that results in biomolecular organization prerequisite for the p85-mediated endocytosis.
PI3K activity at focal adhesion is a major driver of mesenchymal cell migration. Earlier works showed that mesenchymal cells initiate protrusion with lopodia extension from nascent adhesions and that a positive feedback loop consisting of PI3K and actin dilates these adhesion-associated protrusions to develop mature lamellipodia 23,24 . Given that p85β has greater a nity to focal adhesion than p85α 65 , p85β is assumed to play a dominant role in cell migration. We determined that AP-2 binding of p85β negatively regulates its focal adhesion residence. As extension/retraction of membrane protrusions and their lifetime are all proportional to the PI3K activity 78 , this AP-2-mediated sequestration of p85β could act as a brake for migrating cells. Indeed, our data indicated that speed and persistency of cell migration correlate with extent of p85β localization at focal adhesions. Furthermore, the AP-2-mediated sequestration could ful ll a condition for long-sought negative feedback regulation of the PI(3,4,5)P 3 excitability 24 . Further exploration of molecular mechanisms underlying the observed p85β dissociation from focal adhesion should help reveal the understudied negative feedback regulation.
Of great interest, iSH2-mediated endocytosis is speci c to the β isoform and not observed with α or γ isoforms. Their opposing effects are reported elsewhere. For instance, p85α and p85β act as a tumor-suppressor and an oncogene, respectively 65,79-82 . Such a difference may have something to do with the endosomal PI3K signaling driven by p85β, but not by p85α. Recent studies revealed a role of endosomal PI(3,4,5)P 3 in Akt signaling 10,83 . In addition, the AP-2 binding motif region coincides with the hinge region that determines the oncogenicity of p85β 82 . Thus, iSH2-mediated endocytosis possibly contributes to hyperactivate endosomal PI3K-Akt signal. T cell regulation may also be a target of p85β endocytosis. It was shown that T cell coreceptor CD28 preferentially binds to the p85β isoform 84 , and that a PI3Kdependent endocytic process determines the CD28 pathway activity 85 . It is therefore tempting to speculate that iSH2-mediated endocytosis associates with the enigmatic difference in immune phenotypes between p85α and p85β knockout mice 1,11,86,87 . Accordingly, the impact of p85β-mediated endocytosis on physiological functions, as well as the molecular mechanisms leading to the difference between α and β, are fundamental to comprehensive understanding of the multi-faceted PI3K molecule in both normal and cancer cells. All the authors contributed to the nal version of the manuscript.

Acknowledgement
We thank Brendan Manning for p85 double knockout cells; Andrew Ewald for HEK293FT cells; Sandra B. Gabelli for human p85α plasmid; Gerald R.V. Hammond for Rab5 and LAMP1 plasmids; Justin W. Taraska for AP180 plasmid; Yi Wu for Paxillin plasmid. We also thank Yuta Nihongaki for technical assistance on lentivirus and FACS experiments. We appreciate Yoshihiro Adachi, Hiroshi Senoo, Miho Iiijma, and Hiromi Sesaki for technical support on lentivirus and western blot experiments. Our appreciation extends to Shigeki Watanabe, Yuuta Imoto, Atsuo Sasaki, Sho W. Suzuki, and Chuan-Hsiang Huang for insightful comments on the research project, and to our lab members including Hideki Nakamura, Allister Suarez and Helen D. Wu for fruitful discussions. We also thank Robert DeRose for manuscript proofreading and experimental support. This study was supported by the National Institutes for Health (R01GM123130 and R01GM136858 to TI, T32GM007445 to AFP), the DoD DARPA (HR0011-16-C-0139 to TI), and the PRESTO program of the Japan Science and Technology Agency to HTM (JPMJPR20KA). HTM was supported by Postdoctoral Fellowships from the Japan Society for the Promotion of Science.

Reagents and antibodies
Rapamycin was purchased from LCLab (R-5000), prepared as 100 µM stock solution in DMSO, and stored at -20°C. Alexa Fluor 647 conjugated transferrin was purchased from Thermo Fisher Scienti c (T23366), reconstituted with Milli-Q water to obtain 5 mg/mL stock solution in PBS, and stored at 4°C. mCLING-ATTO 647N-labeled was purchased from Synaptic Systems (710 006AT1), reconstituted with Milli-Q water to obtain 50 µM stock solution in PBS, and stored at -80°C. LY294002 was purchased from Selleck Chemicals (S1105), prepared as 50 mM stock solution in DMSO, and stored at -20°C. Fibronectin was purchased from Sigma-Aldrich (F4759), reconstituted with Milli-Q water to obtain 1 mg/mL stock solution, and stored at -20°C. Once frozen bronectin was thawed, the remainder was kept at 4°C. PDGF-BB was purchased from Sigma-Aldrich (P3201), reconstituted with 4 mM HCl containing 0.1% BSA to obtain 50 µg/mL stock solution, and stored at -20°C. FAK inhibitor PF-573228 was purchased from Selleck Chemicals (S2013), prepared as 20 mM stock in DMSO, and stored at -20°C. Hoechst 33342 (10 mg/mL solution in water) was purchased from Thermo Fisher Scienti c (H3570) and

Cell culture
HeLa, Cos-7 and HEK293FT cells (a kind gift from Andrew Ewald lab) were cultured in a DMEM (Corning, 10-013-CV) medium supplemented with 10% fetal bovine serum (Sigma-Aldrich, F6178). Wild type and p85 double knock out (DKO) mouse embryonic broblast (MEF) cells were kind gifts from Brendan Manning lab and cultured in DMEM with 10% FBS.
Generation of YFP-p85 rescued MEF cells EYFP-p85 rescued cells were established by lentivirus transduction. Lentiviruses were produced by transfecting HEK293FT cells as follows. Five hundred micro litter of Opti-MEM was mixed with 10 µg FUGW-puro-EYFP-p85, 7.5 µg ∆8.9, and 3.5 µg VSV-G plasmids. Another 500 µL of Opti-MEM was mixed with 63 µL of 1 mg/mL polyethylenimine. Two solutions were mixed and kept at room temperature for 20 minutes, then added to HEK293FT cells seeded one day before at 6×10 6 cells/10 cm dish density. Two and three days after transfection, media were collected. The virus-containing media were mixed with 1/3 volume of 40% (w/v) PEG-8000, 1.2 M NaCl, 1×PBS (pH 7.0-7.2) and kept at 4°C for more than 45 min. The viruses were precipitated by centrifugation (1,500×g for 45 min at 4°C) and resuspended with PBS (200 µL for 10 cm dish cells). Aliquoted viruses were ash-frozen in liquid nitrogen and stored at -80°C. To infect p85 DKO cells with the viruses, p85 DKO cells were seeded one day before infection at 4×10 4 cells/well (6-well) density. On the day of infection, medium was replaced with fresh 500 µL of medium and virus suspension (10-100 µL depending on titer) and nal 10 µg/mL polybrene were added. YFP positive cells were sorted by SH800S (SONY).

Microscopes and imaging
Confocal imaging was performed on a spinning-disk confocal microscope. The microscope was based on an inverted Axiovert 200 microscope (Zeiss) and equipped with the spinning disk confocal unit (CSU10; Yokogawa) and triple-band dichroic mirror (Di01-T442/514/647, Semrock). Excitations of CFP, YFP, and mCherry were conducted with diode lasers and a semiconductor laser (COHERENT, OBIS 445 nm LX 75 mW, OBIS 514 nm LX 40 mW, OBIS 561 nm LS 50 mW), which were ber-coupled (OZ optics) to the spinning disk unit.
Images were taken with a Neo Fluor ×40 objective (Zeiss) and a CCD camera (Orca ER, Hamamatsu Photonics) driven by or MetaMorph or Micro-Manager 1.4 (Open Imaging). Images of live cell CID assay was typically taken every 1 min for 40 min. Epi imaging for mCLING assay sample and ERKKTR live cell Imaging was performed by an Eclipse Ti inverted uorescence microscope (Nikon) equipped with a ×60 oil-immersion objective lens and Zyla 4.2 plus sCMOS camera. TIRF imaging of focal adhesion was performed by an Eclipse Ti inverted uorescence microscope (Nikon) equipped with a ×100 oil-immersion TIRF objective lens and pco.edge sCMOS camera (PCO). Nikon microscopes were driven by NIS-Elements software (Nikon).
All the live cell imaging was performed in the imaging media containing DMEM (Corning, 17-205-CV) and 1×Glutamax (Thermo Fisher Scienti c, 35050061) with temperature (37°C), CO2 (5%), and humidity control by a stage top incubator and a lens heater (Tokai Hit). For xation, typically, cells were chilled on ice, washed 2 times with ice-cold PBS, xed by xation solution (4% paraformaldehyde and 0.15 % glutaraldehyde in PBS) for 10 min at room temperature, washed 2 times with ice-cold PBS, and stored at 4°C in PBS.
Image processing and analysis were performed by Fiji software 96 .
Chemically-inducible co-recruitment assay: EYFP-FKBP was fused to iSH2 or indicated mutants, while FRB-CFP is tethered to the inner lea et of plasma membrane using the CAAX-region of K-Ras. Upon rapamycin addition, FKBP binds to FRB which brings the bait (mVenus-FKBP-iSH2) and the prey capable of binding (AP-2-mCherry or mCherry) to the plasma membrane. Recruitment of the bait and the prey to the plasma membrane were detected by TIRF microscopy as an increased uorescence signal (Extended Data Fig. 6a-c). For quanti cation, after background subtraction, co-recruitment levels of prey were measured by increase in mCherry (prey) signal normalized to the intensity before rapamycin addition. Only cells showing at least 30% increase in mVenus (bait) intensity after Rapamycin addition were considered.

Quanti cation and statistical analysis
All the quanti ed data were obtained from 3 or more independent experiments except for Extended data Fig. 10d. To statistically compare a pair of data, wilcox.test was used in R as Wilcoxon rank sum test. To statistically compare multiple data, pSDCFlig (Asymptotic option) of NSM3 library was used in R as Steel-Dwass test.
Quanti cation of iSH2 puncta index Following the method described in Supplementary Figure 13 of a previous paper 97 , we created 5×5 median-ltered images of YF-iSH2 images and divided the raw image by the ltered images. iSH2 puncta index was measured by quantifying standard deviation of cytosolic region of the divided YF-iSH2 images. To avoid including intensity uctuation caused by plasma membrane, regions of interest were manually drawn. We used Cos7 cells for the analysis of iSH2 mutants and variants since the cell showed more homogenous background (e.g., in the case of negative control YF) than HeLa cells.
Western blot 3.6×10 5 cells/well (6-well) were seeded ~16 hours before experiment. The cells were serum-starved for 5-6 hours, stimulated as described in gure legends with 5% CO 2 at 37°C. The reaction was stopped by directly replacing the culture media with 100 µL ice-cold RIPA buffer (Cell Signaling, 9806S) supplemented with cOmplete protease inhibitor (1×, Roche, 11873580001), 1 mM PMSF, and phosphatase inhibitors (1× for each, Sigma P5726 and P0044). Since cooling on ice was not su cient to stop dephosphorylation, it was critical to immediately replace the media with RIPA buffer. Soluble fraction was collected as supernatant after centrifugation (14,000×g for 10 min at 4 °C) and the protein concentration was measured by Bradford assay. The samples were mixed with SDS-sample buffer, boiled at 95°C for 5 min, and separated on polyacrylamide gel. Proteins were transferred to methanol pre-treated PVDF membrane by using Criterion Blotter (BioRad, 1704070JA). The membrane was blocked by rocking in blocking buffer (3%BSA, 1×TBS) for 30-60 min at RT, stained with primary antibodies by rocking in antibody buffer (3%BSA, 1×TBS, 0.1% Tween 20, 0.1% NaN 3 ) overnight at 4°C, washed (5 min×3 times) with TBS-T, stained with secondary antibodies in antibody buffer for 1 hours at rt, and washed again (5 min×3 times) with TBST. Immuno uorescence Immuno uorescence against vinculin was performed as follows. 25×10 3 cells/well MEF cells were seeded on bronectin-coated 8-well chambers and incubated overnight in DMEM supplemented with 10% FBS. Cells were then washed with PBS twice, xed with 4% paraformaldehyde in PBS at room temperature for 15 minutes, washed again with PBS twice, permeabilized 0.1 % Triton X-100 in PBS at room temperature for 2.5 minutes, and blocked with blocking buffer (1% BSA in PBS) at room temperature for 30 minutes. Antibody against vinculin was used as ×500 dilution in the blocking buffer and the binding was performed at 4°C overnight. The secondary antibody Alexa Fluor 568-conjugated anti-Mouse IgG was used as ×1000 dilution in the blocking buffer and the binding was performed at room temperature for 1 hour. Each antibody binding steps were followed by 3 times of 5 minutes wash with TBST.

Proliferation assay
For proliferation assay, 2.5-5×10 4 cells were seeded on asks, cultured in DMEM supplemented with 10% FBS for 50-72 hours, and the nal number of cells were counted. Doubling time was calculated by Initial and nal number of cells assuming the cell growth is exponential.
Random migration assay 24-well plate were coated with 10 µg/mL bronectin (5 µg/cm 2 ) >30 min at 37°C. 1×10 4 MEF cells were seeded and incubated in DMEM supplemented with 1% FBS for roughly 20 hours. Cells were washed once with fresh DMEM supplemented with 1% FBS and the media were replaced with DMEM supplemented with 10% FBS and 0.25 µg/mL Hoechst 33342. Cells were left in a 37°C and 5% CO 2 incubator for 2 hours (Hoechst stain seemed to delay in the presence of bronectin or collagen coating). Random migration was performed at 37°C and with 5% CO 2 and humidity. Images were captured every 10 minutes for 16 hours through DAPI channel and phase contrast and analyzed by TrackMate 98 plugin in Fiji software 96 .

Chemotaxis
Chemotaxis assay was performed on µ-slide chemotaxis chambers (ibidi, 80326) by following manufacturer's protocol. Brie y, 2.4×10 6 /mL WT MEF or 3.0×10 6 /mL p85 DKO and rescued MEF were seeded. After incubation at 37°C with 5% CO 2 and 95% humidity for 2-3 hours, right reservoir was lled with imaging media supplemented with 1% FBS and 0.25 µg/mL Hoechst 33342 and left reservoir was lled with imaging media supplemented with 20% FBS and 0.25 µg/mL Hoechst 33342. The chamber was further incubated for 2 hours to allow the FBS gradient to be established. Chemotaxis was performed at 37°C with 5% CO 2 and humidity. Images were captured every 10 minutes for 16 hours through DAPI channel and bright eld and analyzed by TrackMate plugin 98 in Fiji software 96 .