Phosphoinositide Conversion Inactivates R‐RAS and Drives Metastases in Breast Cancer

Abstract Breast cancer is the most prevalent cancer and a major cause of death in women worldwide. Although early diagnosis and therapeutic intervention significantly improve patient survival rate, metastasis still accounts for most deaths. Here it is reported that, in a cohort of more than 2000 patients with breast cancer, overexpression of PI3KC2α occurs in 52% of cases and correlates with high tumor grade as well as increased probability of distant metastatic events, irrespective of the subtype. Mechanistically, it is demonstrated that PI3KC2α synthetizes a pool of PI(3,4)P2 at focal adhesions that lowers their stability and directs breast cancer cell migration, invasion, and metastasis. PI(3,4)P2 locally produced by PI3KC2α at focal adhesions recruits the Ras GTPase activating protein 3 (RASA3), which inactivates R‐RAS, leading to increased focal adhesion turnover, migration, and invasion both in vitro and in vivo. Proof‐of‐concept is eventually provided that inhibiting PI3KC2α or lowering RASA3 activity at focal adhesions significantly reduces the metastatic burden in PI3KC2α‐overexpressing breast cancer, thereby suggesting a novel strategy for anti‐breast cancer therapy.


Introduction
Metastasis is a leading cause of mortality in cancer patients which results from a multistep process characterized by increased migration and invasion of cancer cells. [1] Such phenotype requires the rearrangement of the cytoskeleton and abnormal cell adhesion including disassembly of focal adhesions at the rear of the

PI3KC2 Overexpression in Breast Cancer Leads to Increased Cell Migration, Invasion and Metastasis
We previously reported that reduced expression of PI3KC2 in a cohort of more than 2000 breast cancer patients initially causes impairment of tumor growth but later leads to the convergent evolution of fast-growing clones with mitotic checkpoint defects. [12] Although we first focused on patients with low PI3KC2 expression levels, a large subset of cases from the same cohort showed increased PI3KC2 expression, correlating with an enhanced probability of distant-metastatic events. [12] In particular, we extracted from our previous analysis [12] the low-grade (1-2) and high-grade tumors (grade-3) of patients expressing either low or high levels of PI3KC2 and represented these data in a pie chart in Figure 1a. We observed that in patients with reduced PI3KC2 expression, only 26% of the tumors were scored as grade-3 ( Figure 1a, upper panel and Table 1; p<0.00001), in agreement with our findings showing that reduced PI3KC2 is initially protective in breast cancer. [12] In contrast, in the subset of patients expressing high levels of PI3KC2 , 49% of tumors were scored as grade 3 (Figure 1a, lower panels; Table 1; odds ratio, 3.03; 95% confidence interval (CI), 2.47-3.71; p<0.00001), suggesting that enhanced PI3KC2 expression was linked to increased aggressiveness and metastatic spreading. [12] We therefore analyzed PI3KC2 protein expression in different human and murine breast cancer cell lines. Increased PI3KC2 protein expression was observed in cells with higher invasive ability [13] (Figure 1b). Accordingly, analysis of PIK3C2A mRNA levels from a Cancer Cell Line Encyclopedia (CCLE) panel [14] in luminal (low aggressiveness) or basal-like/triplenegative (highly aggressive) human breast cancer cell lines showed a significant correlation between high PIK3C2A expression and basal-like/triple-negative subtype (Figure S1a, Supporting Information).
To study whether increased PI3KC2 expression was sufficient to induce tumor metastasis in vivo, we focus on two murine breast cancer cell lines, 168-FARN and 4T1, respectively with low and high metastatic ability (Figure 1b). [15] These two cell lines were engineered with a PiggyBac (PB) transposon vector to yield stable overexpression of PI3KC2 ( Figure S1b, Supporting Information) and orthotopically injected into the mammary fat pad in mice. [16] Although increased PI3KC2 levels did not induce significant changes in tumor growth ( Figure S1c-l, Supporting Information), overexpression of PI3KC2 resulted in a significantly increased number of lung metastases (Figure 1c, d). This effect was particularly evident in cells with low metastatic ability such as 168-FARN, where enhanced PI3KC2 expression led to a more than fivefold increase of lung metastases (Figure 1c, d).
To better understand the molecular mechanism by which PI3KC2 overexpression promotes metastasis, human breast cancer cell lines expressing low levels of PI3KC2 , like MCF7 and MDA-MB-468, were engineered with a PiggyBac (PB) transposomal vector to produce a stable overexpression of PI3KC2 ( Figure S1m, Supporting Information). After 7 days in culture, differential gene expression associated with increased cell migration and invasion was observed in MCF7 and MDA-MB-468 overexpressing PI3KC2 (PB-PI3KC2 ), compared to control cells (PB empty vector) (Figure 1e and Figure S1n, Supporting Information). Accordingly, increased extracellular matrix degradation was observed in PB-PI3KC2 compared to control PB cells in both gelatin degradation assay and transwell invasion assays (Figure 1f,g and Figure S2a,b, Supporting Information). No significant differences in cell proliferation were observed in PB-PI3KC2 compared to PB cells ( Figure S1o,p, Supporting Information).
Enhanced cell invasion is accompanied by changes in the actin cytoskeleton towards a pro-migratory phenotype. [17] To better understand if PI3KC2 overexpression triggers rearrangements in cell shape, that is, increased number of actin protrusions, such as filopodia and lamellipodia, PB and PB-PI3KC2 MCF7 were stained for F-actin. Breast cancer cells overexpressing PI3KC2 exhibited more stress fibers and intense ruffling at the cell edges, together with increased numbers and length of filopodia without significant alteration in the cell area (Figure 1h-j, Figure S2c,d, Supporting Information). Spreading of filopodia and lamellipodia in PB-PI3KC2 cells was even more obvious when cells were cultured on gelatin-coated plates ( Figure S2e, Supporting Information). In line with actin cytoskeleton remodeling and increased levels of pro-migratory genes, PI3KC2 -overexpression increased the migration velocity by threefold and doubled the distance traveled by breast cancer cells in a cell-tracking migration assay (Figure 1k,l). Wound healing and transwell migration assays further confirmed that PI3KC2 overexpression enhances migration in both human (MCF7 and MDA-MB-468) and murine (4T1) breast cancer cell lines (Figure 1m and Figure S2f-i, Supporting Information). Consistent with the previous report, [18] PI3KC2 knockdown impairs migration/invasion ability in MDA-MB-231 ( Figure S2j, Supporting Information).
Taken together, our findings demonstrate that PI3KC2 promotes breast cancer cell migration and invasion in vitro and its increased expression correlates with breast cancer metastasis in vivo.

Increased PI3KC2 Expression Reduced Focal Adhesion Number and Stability
Cell migration requires the integration and coordination of specific focal adhesion dynamics at the cell front, center, and rear. [2a,19] The regulation of adhesion turnover and disassembly involves a number of tyrosine kinases and phosphatases, most of which are engaged in FAK signaling pathways. [2a] To investigate the mechanisms underlying the PI3KC2 -dependent cell migration in breast cancer cells, the relationship between PI3KC2 and focal adhesion dynamics was further explored. First, we analyzed the phosphorylation status of FAK in cells overexpressing PI3KC2 in normal or pro-migratory conditions including treatment with fibronectin. We observed in PB-PI3KC2 cells that phosphorylation at Tyr397, the most important site for the regulation of FAK activity, [20] was significantly increased (Figure 2a and Figure S3a, Supporting Information), together with enhanced phosphorylation at Tyr925 (Figure 2a and Figure S3a, Supporting Information), which is known to regulate cross-talk between focal adhesion turnover and cell protrusion. [21] In addition, PB-PI3KC2 cells exhibited increased phosphorylation of the focal adhesion scaffold paxillin on Tyr118 (Figure 2a and Figure S3a, Supporting Information), a phosphorylation status commonly observed in metastatic breast cancer. [22] Phosphorylation of FAK at Tyr-397 promotes cell migration by inducing disassembly of focal adhesions at the cell tail. [23] In line with this, breast cancer cells with PI3KC2 overexpression and increased phosphorylation of Tyr-397 displayed a significantly reduced number of focal adhesions without changes in size (Figure 2b), suggesting rapid focal adhesion turnover. To test this directly, we performed a focal adhesion disassembly assay [24] and measured the focal adhesion lifetime. Nocodazole treatment was used to induce focal adhesion accumulation followed by nocodazole wash-out to measure the disassembly rate. At 30 min after release from the nocodazole block, PB-PI3KC2 MCF7 showed more than 60% of focal adhesion disassembly, compared with less than 20% in PB control cells (Figure 2c Collectively, our findings demonstrate that overexpression of PI3KC2 promotes activation of the FAK signaling pathway and shortens focal adhesion lifetime by facilitating their disassembly.

PI3KC2 is Recruited to the Focal Adhesion by the HBD Region and Produces PI(3,4)P2
To better understand how PI3KC2 overexpression affects focal adhesion dynamics, its intracellular localization was analyzed. Co-expression of GFP-PI3KC2 and FAK-mCherry in MCF7 showed that the two proteins are enriched and co-localized at plasma membrane regions corresponding to lamellipodia and filopodia (Figure 3a). Time-lapse imaging further revealed that GFP-PI3KC2 enrichment precedes the disassembly of the focal adhesion itself as FAK-mCherry decreased immediately following the full recruitment of GFP-PI3KC2 ( Figure 3b and Figure  S4a, Supporting Information). Next, we asked how PI3KC2 is recruited to focal adhesions. We took advantage of our recently  Table 1 [12] . b) Quantification and representative blots showing PI3KC2 protein expression normalized over -tubulin in human BC cell lines and normal mammary cell line HME1 (left panels). Quantification and representative blots showing PI3KC2 protein expression normalized over GAPDH in four isogenic murine BC cell lines (right panels). c,d) Lung metastasis analyzed in BALB/c mice orthotopically injected with PB or PB-PI3KC2 168-FARN (left panels). Lung metastasis analyzed in BALB/c mice orthotopically injected with PB or PB-PI3KC2 4T1 (right panels). Representative H&E staining is shown (c) (scale bar = 2 mm). Each dot in the graph (d) is representative of an injected mouse (n ≥ 5). e) Real-time analysis performed on PB or PB-PI3KC2 MCF7. mRNA levels of indicated genes are reported as ratio over control. f) PB or PB-PI3KC2 MCF7 were prepared and cultured on glass coverslips covered with cross-linked gelatin overlaid with Oregon GreenTM 488. Cells were stained with phalloidin to identify actin filaments (F-actin; red) and ToPRO3 (nuclei, blue). Degraded gelatin is identified as a dark area (white arrows) on the Oregon GreenTM 488 (green) background. g) Representative pictures and quantification of Transwell invasion assay in PB and PB-PI3KC2 MCF-7; scale bar = 100 μm. h) Immunofluorescence staining using Phalloidin to detect F-actin in PB or PB-PI3KC2 MCF-7. Extensive filopodia formation can be observed in PB-PI3KC2 group as indicated by red arrow. i) Frequency distribution showing filopodia number in PB or PB-PI3KC2 MCF7. n = 162 and n = 168 cells were imaged for PB and PB-PI3KC2 MCF7, respectively, in three independent experiments. j) Filopodia length measured in PB or PB-PI3KC2 MCF7 cells. n = 46 and n = 51 cells were imaged for PB and PB-PI3KC2 MCF7, respectively, in three independent experiments. k,l) Representative cell tracks over 10 h (k) and quantification of the migration distance and speed (l) are shown. Data were analyzed using Manual Tracking plugin (ImageJ). n = 34 and n = 35 cells for PB and PB-PI3KC2 respectively. m) Representative pictures and quantification of Transwell migration assay in PB and PB-PI3KC2 MCF-7; scale bar = 100 μm. All results are shown as mean of at least three independent experiments ± SEM (n.s., no significance, *P<0.05; **P<0.01; ***P<0.001).
described crystal structure of PI3KC2 in which we identified a sequence insertion of about 100 amino acids between the RBD and the N-C2 domain that was absent in class I and class III enzymes. This domain, named HBD, displayed structural similarity with the focal-adhesion targeting (FAT) domain of Crk-associated substrate (Cas), and the F-actin binding domains of vinculin and -catenin. We posited that this protein-binding domain may be responsible for the recruitment of PI3KC2 to focal adhesions. To test this hypothesis, 293T cells were transfected with GFP-PI3KC2 (full length), GFP-HBD, or GFP-PI3KC2 ΔHBD together with HA-FAK. Co-immunoprecipitation (IP) experiments showed that full-length PI3KC2 or HBD alone could both interact with HA-FAK ( Figure 3c and Figure S4b, Supporting Information). Conversely, HBD deletion from full-length PI3KC2 completely abolished the binding to FAK (Figure 3c). Consistently, live-cell imaging showed that GFP-HBD co-localizes with Vinculin-RFP, while deletion of the HBD region from full-length PI3KC2 was sufficient to displace its localization from focal adhesions ( Figure S4c, Supporting Information). Taken together, our data demonstrate that the localization of PI3KC2 to focal adhesions is mediated by a previously unidentified association of PI3KC2 with FAK via its unique HBD domain.
Once recruited to the plasma membrane, PI3KC2 undergoes a conformational change towards an active state able to synthetize PI(3,4)P2. [25] We hypothesized that the FAK-mediated recruitment of PI3KC2 close to the plasma membrane could be responsible for the production of a selected PI(3,4)P2 pool leading to focal adhesion reduced stability. We generated breast cancer cells that co-express a GFP-TAPP1-PHx3 binding probe able to detect PI(3,4)P2 [26] and FAK-mCherry. PI(3,4)P2 showed significant co-localization with FAK and was particularly enriched at focal adhesions where PI3KC2 was overexpressed ( Figure 3d). Time-lapse analysis further showed that PI(3,4)P2 is enriched at focal adhesions immediately before focal adhesion dismantling (Figure 3e), suggesting a causal role in controlling focal adhesion lifetime. To investigate whether PI(3,4)P2 directly stimulates focal adhesion turnover, we analyzed the phosphorylation status of FAK and paxillin in cells overexpressing wild type (WT), kinaseinactive (KD), and KPLP-mutant PI3KC2 . Increased phosphorylation of FAK (p-Tyr397) and paxillin (p-Tyr118) was observed in cells overexpressing WT PI3KC2 but not in KD or KPLP mutants, demonstrating that focal adhesion turnover specifically requires PI(3,4)P2 ( Figure 3f). Finally, the migratory ability of cells overexpressing WT, KD, and KPLP-mutant was assessed in a wound-healing assay. In agreement with reduced FAK and paxillin activation, wound closure in KD-and KPLP-expressing cells was slower than in WT controls ( Figure 3g).
These results demonstrate that recruitment of PI3KC2 by its HBD region to the focal adhesion is required for PI(3,4)P2dependent FAK/paxillin activation and cell migration (Figure 3h).

PI3KC2 -Dependent PI(3,4)P2 Controls R-RAS Inactivation at Focal Adhesion
PI(3,4)P2 has recently emerged as a crucial regulator of cell migration in Dictyostelium, where it recruits two PH-domain containing RasGAP proteins, [27] suggesting that a RAS isoform and RasGAP pair may regulate mammalian cell migration in a similar manner. In particular, alteration in R-RAS activity leads to changes in epithelial cell motility and morphology. [28] Thus, we determined the intracellular localization of PI3KC2 and R-RAS in breast cancer cells. Significant co-localization between GFP-PI3KC2 and R-RAS at focal adhesion was observed (Figure 4a and Figure S5a, Supporting Information), suggesting a functional relationship between the two proteins. To further probe this, the R-RAS activity was evaluated in cells overexpressing PI3KC2 using pull-down assays. [29] R-RAS activity was significantly reduced in PB-PI3KC2 compared with PB controls (Figure 4b), indicating that increased PI3KC2 expression leads to R-RAS inactivation. Other RAS family members were unaffected (K-, N-, H-RAS and RAP1) ( Figure S5b, Supporting Information). To challenge our findings by an independent approach, a FRETbased sensor was used to directly visualize the activation status of R-RAS at focal adhesions. [29b,30] Raichu-R-Ras FRET probe comprised a modified YFP designated as Venus, R-Ras, the RA domain of the R-Ras effector RalGDS, a modified CFP designated as SECFP. Under R-Ras activation, the intramolecular binding of R-Ras to the RA domain brings CFP into proximity with YFP, triggering a FRET increase. We transfected cells with RFP-vinculin to visualize focal adhesions together with a FRET probe and FRET activity was measured specifically at vinculin-positive regions of the cell. In line with the data from the R-RAS pulldown assay, FRET activity was strongly reduced at focal adhesions in PB-PI3KC2 cells compared with PB controls (Figure 4c). Time-lapse imaging further revealed that focal adhesion disassembly is accompanied by inactivation of R-RAS and that in PB-PI3KC2 cells, the R-RAS inactivation is faster compared to PB controls ( Figure 4d, Figure S5c, Supporting Information). These findings demonstrate that increased PI3KC2 expression leads to selective and spatiotemporal inactivation of R-RAS at focal adhesions.
To check whether R-RAS inactivation was linked to the reduced focal adhesion lifetime and increased migration observed in cells  Figure S6a, Supporting Information). Furthermore, R-RAS activity was reduced only in cells expressing WT PI3KC2 but not cells expressing either the KD or the KPLP PI3KC2 mutant, indicating that R-RAS inactivation specifically requires a PI3KC2 -dependent production of PI(3,4)P2. (Figure 4g).
Finally, to monitor motility and tissue invasiveness of PI3KC2 -overexpressing breast cancer cells in vivo, we took advantage of zebrafish model, to trace metastasis by live imaging in embryos. [31] Hence, we injected PB (green), PB-PI3KC2 (red) in 48-hpf zebrafish ( Figure S6b, Supporting Information). As expected, we observed an increased incidence of metastases and of disseminated foci in fish injected with PB-PI3KC2 breast cancer cells (Figure 4h, i and Figure S6b, Supporting Information). Conversely, overexpression of R-RAS 38V in PB-PI3KC2 cells or treating PB-PI3KC2 cells with PI3KC2 selective inhibitor, PITCOIN was sufficient to block cell spreading (Figure 4h, i).
Collectively our data point to a specific requirement of PI(3,4)P2 at focal adhesions to coordinate R-RAS activity, focal adhesion stability, and cell migration (Figure 4j).

R-RAS Inactivation is Mediated by PI(3,4)P2-Dependent RASA3 Accumulation at Focal Adhesions
The ability of R-RAS to switch between the active and inactive state is regulated by the balance between GEFs and GAPs activity. [32] Our observation of reduced R-RAS activity points to a selective recruitment of a RasGAP able to interact with PI(3,4)P2 at focal adhesions. A potential candidate is the RasGAP RASA3, which constitutes a component of the PI(3,4)P2 interactome. [33] Localization of RASA3 was then analyzed by immunofluorescence in breast cancer cells. Significant co-localization between RASA3 and PI(3,4)P2 (stained with GFP-TAPP1-PHx3 probe) was observed at vinculin-positive focal adhesions (Figure 5a and Figure S6c, Supporting Information). Accordingly, cells overexpressing PI3KC2 with increased PI(3,4)P2 levels showed a twofold increase in RASA3 enrichment at focal adhesions (Figure 5b and Figure S6d, Supporting Information). Coherently, RASA3 knockdown in cells overexpressing PI3KC2 restored R-RAS activation, demonstrating a functional relationship (Figure 5c, d).
To link the recruitment of RASA3 at focal adhesions with its GAP activity towards R-RAS, we performed a FRET analysis to measure R-RAS activation. While in PB-PI3KC2 cells, R-RAS activity was significantly reduced at focal adhesions, knock-down of RASA3 was sufficient to restore R-RAS activity to levels comparable to control PB cells (Figure 5e). Therefore, RASA3 controls R-RAS activation at focal adhesion in a PI(3,4)P2-dependent manner. Then we checked whether RASA3-mediated inactivation of R-RAS directly affects cell migration in wound healing assays. While PB-PI3KC2 cells showed a fast closure of the wound, concomitant down-modulation of RASA3 was sufficient to slow the wound closure similar to the rate observed in control cells (Figure 5f).
Finally, to challenge our findings in an in vivo model, we used an in vivo zebrafish model. [31] To this aim, we injected MCF7 expressing either PB, PB-PI3KC2 , or PB-PI3KC2 together with siRNA targeting RASA3 in 48-hpf zebrafish. As expected, we observed an increased number of metastases and disseminated foci in fish injected with PB-PI3KC2 breast cancer cells. Conversely, knocking-down of RASA3 in PB-PI3KC2 cells was sufficient to block the migratory process (Figure 5g-i).
Collectively, our findings demonstrate that PI(3,4)P2-RASA3 mediated inactivation of R-RAS allows PI3KC2 -overexpressing breast cancer cells to acquire a pro-migratory and pro-invasive phenotype, leading to increased metastatic potential (Figure 5j). Moreover, we provide a proof of concept that inhibiting PI3KC2 or targeting RASA3 activity can block the metastatic spreading of breast cancer cells overexpressing PI3KC2 in vivo.

Discussion
By producing PI(3)P and PI(3,4)P2, PI3KC2 plays an essential role during development and its loss or inactivation leads to multiple pathological processes. [34] Besides its enzymatic activity, we recently described a scaffold function of PI3KC2 that is necessary for keeping genome integrity by preventing mitotic defects in breast cancer cells. [12] Accordingly, loss of PI3KC2 in breast cancer initially delays tumor growth but finally leads to the selection of more aggressive clones. Here we describe that, unlike patients with PI3KC2 loss, high levels of PI3KC2 correlate with tumor grade and probability of distant metastatic events in breast cancer patients. In agreement with this phenotype, we found that a PI(3,4)P2 pool synthetized by PI3KC2 at focal adhesions disturbs their stability, leading to enhanced breast cancer cell migration and invasion. Moreover, differential expression of with HA-FAK and GFP-PI3KC2a (WT, HBD or HBD deleted mutants). HA-FAK was immunoprecipitated with anti HA antibody (IP). Bound proteins were blotted with anti GFP or anti HA. d) Representative picture (left) and PI(3,4)P2 quantification at focal adhesion (right) in live cell imaging performed on MCF7 expressing GFP-TAPP1-PHx3 (green) and FAK-mCherry (red). n = 44 and n = 63 FAs for PB or PB-PI3KC2 respectively. Scale bar, 1 μm. e) Turnover dynamics of FAK-mCherry labeled focal adhesions (red) and GFP-TAPP1-PHx3 (green). Fluorescence intensity profiles measured as a function of time were normalized to the maximum FAK-mCherry fluorescence intensity for each focal adhesion and aligned relative to one another (n = 10 focal adhesions). genes involved in migration and invasion can be likely explained as a consequence of increased focal adhesion dynamics and cytoskeletal remodeling, [35] induced by PI3KC2 overexpression. This is particularly evident in light of the fact that Class-II PI3K has never been shown to function in gene transcriptional regulation. [34c] Although our model of increased tumor cell migration and invasion relies on the PI3KC2 catalytic activity, the recently described PI3KC2 HBD domain emerged as a critical region for protein-protein interaction. In particular, this domain appears pivotal to the localization of PI3KC2 to cytoskeletal structures, as initially we found it to localize this kinase to the mitotic spindle by binding with TACC3 [12] and now to target PI3KC2 to focal adhesions, by promoting its association with FAK. In line with these findings, the HBD shares structural similarities with the FAT domain of Cas, and the F-actin binding domains of vinculin and -catenin, both necessary and sufficient for protein localization to the focal adhesions. Altogether these findings suggest that the HBD domain of PI3KC2 has a structural role as a proteininteracting domain in driving fundamental association with different proteins, critically required to localize PI3KC2 activity in specific subcellular compartments.
While PI(3,4)P2 has been previously associated with increased cell migration [10a, 11,36] and an inhibitory RAS-PI(3,4)P2 feedback loop was recently proposed, [27] its relevance for tumor metastases in vivo remained largely elusive. Our findings point to a direct role for PI(3,4)P2 in controlling focal adhesion stability, and identified the R-Ras-GAP RASA3 as the main player in driving R-RAS inactivation during migration and invasion, both in vitro and in vivo. Based on our findings, PI(3,4)P2 produced by PI3KC2 recruits RASA3 at focal adhesions, leading to R-RAS inactivation and focal adhesion dismantling. The localization of endogenous RASA3 at focal adhesion complexes is required for cytoskeleton rearrangement and cell migration, and loss of RASA3 impairs focal adhesion turnover. [37] Importantly, our data shows that increased PI3KC2 -mediated PI(3,4)P2 production significantly reinforces the localization of RASA3 at focal adhesions.
RASA3 is known to function as a dual GAP for R-RAS and RAP1 small GTPases. [38] Here we found that, at least in breast cancer epithelial cells, overexpression of PI3KC2 leads to increased PI(3,4)P2 enrichment at focal adhesion accompanied by an increased activity of RASA3 towards R-RAS. Considering that the switch of RASA3 from RAS-GAP to RAP-GAP involves its PH domain, [39] we hypothesize that PI(3,4)P2 might affect this conformational change by promoting RASA3 activity in an R-RAS selective manner. Further studies are needed to better elucidate this mechanism of action.
Previous works showed that the C-terminal region of R-RAS contains a focal adhesion targeting signal and that targeting and activation of R-RAS are linked processes in the formation of focal adhesion in epithelial cells. [28a,b] Interestingly, only R-RAS-GTP is recruited to focal adhesions, enhancing both cell adhesion and cell spreading. [28a,40] Conversely, deactivation of R-RAS leads to its exclusion from focal adhesions and reduced focal adhesion stability. [28a] In line with this, expression of dominant negative R-RAS (41A) enhances migration persistence and membrane protrusion. [41] Our data further expanded these findings, showing that in breast cancer cells, the R-RAS inactivation and the consequent reduction in focal adhesion number and stability rely on PI(3,4)P2-mediated recruitment of the Ras-GAP, RASA3 to focal adhesions. We speculate that the RASA3-driven inactivation of R-RAS can be responsible for R-RAS release from the focal adhesions and their consequent destabilization and enhanced turnover, leading to increased cell migration. Additionally, the zebrafish xenograft model conclusively demonstrated that the PI3KC2 /RASA3/R-RAS axis controls breast cancer distant metastasis in vivo; nonetheless, future studies in mouse models will better elucidate the importance of targeting PI3KC2 and RASA3 in metastatic breast cancer therapy.
Despite recent advances in anticancer therapies targeting the primary mammary gland tumor, treating metastatic breast cancer has remained challenging. Therefore, the identification of new druggable pathways promoting metastasis remains an unmet medical need. The finding of the crosstalk between PI3KC2 and RASA3 identifies two potential therapeutic targets. On the one hand, the development of more potent and selective PI3KC2 inhibitors might open new therapeutic strategies for metastatic breast cancer. On the other, GAPs like RASA3 are potentially druggable and our proof-of-concept study provides evidence that inhibition of RASA3 GAP activity can significantly reduce metastasis of breast cancer irrespective of the subtype and likely even in other cancer types.

Experimental Section
Human Subjects: To assess the clinical relevance of PI3KC2 expression to breast cancer, a series of 1779 operable breast cancer patients was analyzed, available on tissue microarray (TMA), who underwent surgery at the European Institute of Oncology (IEO) in Milan from years 1997 to 2000. Details were previously shown. [12] The study was approved by the Institutional Review Board of the European Institute of Oncology (Milan, Italy) and informed consent was obtained from all subjects.
Protein Analysis: Cells and tissues were homogenized in lysis buffer (120 mm NaCl, 50 mm Tris-HCl pH = 8, 1% Triton X-100) supplemented with 25x protease inhibitor cocktail (Roche), 50 mm sodium fluoride, and 1 mm sodium orthovanadate. Lysates were cleared by centrifugation at 13 000 rpm for 15 min at 4°C. Protein concentration was determined by Bradford method and supernatants were analyzed for immunoblotting with the indicated antibodies. Membranes probed with the indicated antibodies were then incubated with HRP conjugated secondary antibodies (anti mouse used 1:10 000, anti rabbit 1:5000, Sigma) and developed with enhanced chemiluminescence (ECL, BD). GAPDH or -tubulin were used for loading control as indicated above. The phosphorylation status of FAK and paxillin were normalized by their total levels. For IP assays, cells were lysed in 50 mm Tris-HCl (pH ), 150 mm NaCl, 1% NP-40, 1 mm EDTA, 10% glycerol, and protease and phosphatase inhibitors. 1 mg of pre-cleared extracts were incubated with 1 μg of the indicated antibody at 4°C on a rotating rack. After 1.5 h, 15 μL of protein G-Sepharose (Amersham Biosciences, Buckinghamshire, UK) were added for 30 min. Samples were collected by centrifugation (13 000 rpm 1 min) and washed six-times with lysis buffer. Bound protein complexes were then eluted by adding 30 μL Laemmli sample buffer. For pull-down experiment, HEK293T cells homogenized in lysis buffer (120 mm NaCl, 50 mm Tris-HCl pH = 8, 1% Triton X-100) supplemented with 25x protease inhibitor cocktail (Roche), 50 mm sodium fluoride, and 1 mm sodium orthovanadate. Lysates were cleared by centrifugation at 13 000 rpm for 15 min at 4°C. 1mg of cell lysate was incubated with 15 μL of recombinant protein-coupled with glutathione Stransferase agarose (GE, Buckinghamshire, UK) for 1 h at 4°C. Beads were washed four times with 1 mL of reaction buffer and analyzed by Immunoblotting after the addition of 30 μL of Laemmli buffer. Gene Silencing and Inhibitors: Rasa3 (5'-GCGCTTTGGGATGAAGAAT-3' and 5'CCTGAAGTTTGGAGATGAA-3') siRNAs were purchased from Horizon Discovery/Dharmacon. PI3KC2 -selective PITCOIN1 inhibitor was provided by Prof. Volker Haucke and described in a seperate study in preparation (Lo WT, et al.).
Quantitative RT-PCR: Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA). cDNA was synthesized from 1000 ng of total RNA using cDNA reverse transcription kits (Applied Biosystems, Foster City, CA). Relative mRNA level was analyzed by real-time PCR (ABI 7900HT FAST Real-Time PCR system, Applied Biosystems, Foster City, CA) with Taqman assays, using the Universal Probe Library system (Roche Applied Science, Penzberg, Germany). 18S gene was used as housekeeping control. The primers are listed in Table S1, Supporting Information.
Immunofluorescence: Immunofluorescence was performed by icecold methanol or 4% PFA fixation of the cells followed by standard procedures. [12,34a] Next, cells were permeabilized with either 0.1% Saponin or 0.3% Triton X-100 for 5 min and then blocked in 2% BSA for 20 min, followed by the incubation with indicated primary antibodies for 60 min. AlexaFluors secondary antibody (Alexa 488, Alexa 568, or Alexa 633) were used 1:1000 for 45 min. Cells were stained with DAPI and examined with, Leica TCS-II SP5 or Leica TSC-II SP8 confocal microscope. Raw images were digitally processed only to normalize the background and enhance the contrast. Z-stacks were acquired and processed with the Maximum Projection tool.
Proliferation Assay: Proliferation assay was performed by using CellTiter-Glo Luminescent Cell Viability Assay (Promega, Mannheim, Germany). Cells were seeded into 96-well plates in octuplicate at 4×10 3 cells/well. Absorbance was measured at the indicated time points.
Single Cell Tracking: Transfected MCF7 cells were seeded on μ-Slide 8 Well (ibidi, Germany) and cultured overnight. Cells were maintained at 37°C and 5% CO 2 and cell migration was monitored by using Leica TCS-II SP5 confocal microscope (10x objective). Cells were imaged for 24 h every 10 min. To assess cell migration, speed, and distance, single cells were tracked by using Manual Tracking plug-in from ImageJ.
Zebrafish Strains and Metastasis Assay: All procedures using zebrafish (Danio Rerio) were authorized by the Ethical Committee of the University of Torino and the Italian Ministry of Health. The wild-type fish strains Tuebingen was used. Adult fish were routinely maintained under a 14h light and 10h dark photoperiod at approximately 28°C, bred and genotyped according to standard procedures. Eggs were generated by natural mating, and following fertilization were collected, treated, and maintained under a 12h light and 12h dark photoperiod at 28°C. Embryos were treated with 0.003% 1-phenyl-2-thiourea (PTU, #P7629, Sigma) at 24hpf to prevent the formation of melanin pigment, which could interfere with the visualization