Starvation-inactivated MTOR triggers cell migration via a ULK1-SH3PXD2A/TKS5-MMP14 pathway in ovarian carcinoma

ABSTRACT SH3PXD2A/TKS5 (SH3 and PX domains 2A) is a scaffold protein that promotes invadopodia formation and regulates cell migration; therefore, the overexpression of SH3PXD2A has been reported in various cancers. However, the molecular mechanisms of the SH3PXD2A-mediated cellular migration signaling pathway remain unknown. Here, we showed that the starvation-induced macroautophagy/autophagy or treatment with the MTOR inhibitor RAD001 elevated SH3PXD2A expression in ovarian cancer cell lines. SH3PXD2A formed a complex with ULK1 (unc-51 like autophagy activating kinase 1) and MTOR, as revealed by co-immunoprecipitation assay. Furthermore, ULK1 affected protein stability by phosphorylating SH3PXD2A at serine residues 112, 142, 146, 147, 175, and 348. Mutation of these six residues in SH3PXD2A reduced ULK1-mediated phosphorylation, blocked SH3PXD2A induction by treatment with RAD001, and reduced its binding to the cell membrane phospholipid phosphatidylinositol-3-phosphate (PtdIns3P) versus recruitment of MMP14 (matrix metallopeptidase 14) in ovarian cancer cells. Finally, the administration of RAD001 induced SH3PXD2A expression in tumor tissues, as revealed by the PDX mouse model. The clinical impact of SH3PXD2A was evaluated in ovarian and endometrial cancers using western blotting. The negative correlation between MTOR-mediated phospho-ULK1 and SH3PXD2A proteins was found in clinical specimen. Furthermore, the TCGA database revealed that the cumulative overall survival of ovarian cancer patients with higher SH3PXD2A RNA/protein expression in tumor lesions was reduced compared to those with lower SH3PXD2A expression. Our results suggest that the ULK1-SH3PXD2A-MMP14 axis might regulate the biological aggressiveness of ovarian cancer and serve as a therapeutic target in this malignancy. Abbreviations AMPK: AMP-activated protein kinase; CHX: cycloheximide; RAD001: everolimus; HBSS: Hanks’ balanced salt solution; LC-MS/MS: liquid chromatography-mass spectrometry/mass spectrometry; MMP14: matrix metallopeptidase 14; MTOR: mechanistic target of rapamycin kinase; MAPK: mitogen-activated protein kinase; RB1CC1/FIP200: RB1 inducible coiled-coil 1; PtdIns3P: phosphatidylinositol-3-phosphate; PX: phox homology; SH3: Src homology 3; SH3PXD2A/TKS5: SH3 and PX domains 2A; SH3PXD2A-[6A]: S112A S142A S146A S147A S175A S348A mutant; ULK1: unc-51 like autophagy activating kinase 1


Introduction
Ovarian cancer is the leading cause of death due to gynecological cancer because it is usually diagnosed at an advanced stage [1].The combination of surgery and chemotherapy remains a major treatment for ovarian cancer; however, treatment relapse, drug resistance, and recurrences are challenges for clinicians or researchers [2].Therefore, studying the molecular carcinogenesis mechanisms of ovarian cancer may help identify novel strategies for treatment.
Macroautophagy/autophagy maintains homeostasis by removing proteins or organelles by self-digestion in response to stress or starvation [3].Under starvation or stress conditions, the autophagy initiating kinase ULK1 (unc-51 like autophagy activating kinase 1) is activated by inhibiting the MTOR (mechanistic target of rapamycin kinase) complex [4], initiating the activation of downstream kinase complexes to form autophagosomes, and finally fusing with lysosomes to degrade engulfed proteins or organelles [5].However, autophagy is a double-edged mechanism in tumorigenesis.In the early stages of tumors, autophagy blocks tumor progression by removing damaged organelles, reducing reactive oxygen species/ROS levels, and causing DNA damage.Nevertheless, autophagy promotes tumor growth by blocking apoptosis and suppresses anti-tumor immunity in the advanced stage tumors [6].Elevated autophagy also involves an increase in cancer cell motility, drug responses, epithelial-mesenchymal transition/EMT, secretion of pro-migratory factors, and expression of MMP (matrix metallopeptidase) enzymes [7,8].Further investigation of the autophagy mechanism may help manipulate this process and improve cancer therapy [9].
ULK1 is a hub of the autophagy pathway.In nutrient-rich environments, the MTORC1 complex can phosphorylate ULK1 at serine 757, then inactivate ULK1 by separating ULK1 and AMP-activated protein kinase (AMPK).However, inhibiting the MTORC1 complex by depleting nutrients causes AMPK to phosphorylate ULK1 at serine 317 and 771 to activate ULK1 [10].Activated ULK1 forms a complex with ATG13, RB1CC1/FIP200 (RB1 inducible coiled-coil 1), and ATG101 and initiates the autophagy pathway [11].In addition to its role on autophagy pathway, ULK1 is involved in cell migration and cancer metastasis.In breast cancer, ULK1 phosphorylates EXO70 to block breast cancer cell metastasis [12].Degradation of ULK1 by MAPK1 (mitogen-activated protein kinase 1)-MAPK3 phosphorylation promotes breast cancer bone metastasis [13].In contrast, overexpression of ULK1 can promote cell migration in gastric cancer [14].Dominant-negative ULK1 reduces metastasis in neuroblastoma [15].Inhibition of ULK1 via its potential inhibitor MRT68921 reduced autophagy production and cell viability, suggesting that targeting ULK1 might be a potential therapeutic strategy for metastatic serous ovarian cancer [16].However, the underlying molecular mechanisms of ULK1-mediated signaling pathways remain unclear.
SH3PXD2A/TKS5 is a scaffold protein and interacts with other enzymes or proteins to regulate signaling pathways involved in cell migration [17].SH3PXD2A contains one phox homology (PX) domain in the N-terminal, five SRC homology 3 (SH3) domains that mediate protein interactions, and two SRC kinase phosphorylation sites at tyrosine 558 and 620 [18].Upon cytokine stimulation, SRC kinase is activated and phosphorylates SH3PXD2A protein to induce invadopodia formation [19].During this process, the PX domain of SH3PXD2A binds to phosphatidylinositol-3-phosphate (PtdIns3P) and PtdIns(3,4)P 2 in the cell membrane to dock the SH3PXD2A complex into the invadopodia region [20].Simultaneously, SH3 domains on SH3PXD2A interact with WASL/N-WASP, GRB2, and NCK2 to remodel ACTB in the invadopodia region to regulate invadopodium formation [21][22][23].In addition, SH3PXD2A also transports MMP2-, MMP9-and MMP14 containing vesicles to invadopodia and finally releases MMP2, MMP9 and MMP14 into the extracellular matrix at invadopodia to promote cell migration [24,25].SH3PXD2A is highly expressed in several solid tumors, including melanoma [24], gastric [26], breast [27], lung [28], oral [29], and colon cancers [30].Knockdown of endogenous SH3PXD2A expression results in decreased cell growth [27], cell invasion [31], invadopodia activity, and metastasis [28].Increased SH3PXD2A expression in clinical tumors is associated with poor prognosis of patients and correlates with tumor metastasis [26,27,30].Reviewing literatures, elevated expression of SH3PXD2A or ULK1 is associated with tumor progression; however, the regulation of ULK1 and SH3PXD2A and their interactions in response to autophagy in cancers has not been reported previously.
In this study, the relationship between autophagy and SH3PXD2A in ovarian cancer cell lines was disclosed.Furthermore, the regulation/interaction between MTOR, ULK1 and SH3PXD2A, and their impact on cell migration through activation of MMPs was evaluated in ovarian cancer cell lines.Finally, the clinical impact of SH3PXD2A expression in survival of patients with ovarian cancer was also discussed.The results of this study might shed light on the role of the MTOR-ULK1-SH3PXD2A axis in ovarian tumorigenesis, potentially paving the way for novel therapeutics.

Starvation induced endogenous SH3PXD2A expression through post-translational modifications
Starvation-induced autophagy can trigger migration and invasion of cancer cells [32].Autophagy was induced in ovarian cancer cells by treatment with Hanks' balanced salt solution (HBSS), which contained neither nutrients nor serum, to examine the relationship between autophagy and expression of endogenous SH3PXD2A.As shown in Figure 1A, the expression levels of SH3PXD2A and autophagy marker-LC3B-II increased in the HBSS-treated group compared to those in the untreated group in three ovarian cancer cell lines (ES2, SKOV3, and Kuramochi): suggesting starvation-induced autophagy and elevated endogenous SH3PXD2A (Figure 1A).The expression of phospho (p)-MTOR and p-ULK1 was repressed after treatment with HBSS in these ovarian cells (Figure 1A).Furthermore, the expression of SH3PXD2A also increased when the ovarian cells were treated with everolimus (RAD001), a rapamycin analog that inhibits MTOR and induces autophagy (Figure 1B).In contrast, HBSStreatment-induced SH3PXD2A protein expression in ovarian cancer cells was blocked in the presence of the MTOR activator-MHY1485 [33] (Figure 1C).Since HBSS rapidly induced SH3PXD2A protein expression in ovarian cancer cells (Figure 1A), we hypothesized that HBSS induced SH3PXD2A protein expression through post-translational modifications such as protein phosphorylation affected its stability [34].To address this issue, the phosphorylation of SH3PXD2A was analyzed using a phos-tag gel.As shown in Figure 1D, increased phosphorylation of SH3PXD2A (p-SH3PXD2A) was observed after treatment with HBSS in ovarian cancer cells (Figure 1D).Since inhibition of MTOR can activate the downstream kinase ULK1 to trigger autophagy [5].To determine whether ULK1 is an upstream effector of SH3PXD2A, endogenous ULK1 was knocked down with its specific siRNA in ES2, SKOV3 and Kuramochi cells with or without HBSS treatment.As shown in Figure 1E, the knockdown of endogenous ULK1 blocked HBSS-mediated SH3PXD2A expression in three ovarian cancer cell lines.Moreover, suppression of SH3PXD2A with a specific siRNA did not affect autophagy (Figure S1A).Nevertheless, in autophagy deficient cells achieved by transfection with ATG5 siRNA [35], the induction of SH3PXD2A by HBSS treatment was attenuated (Figure S1B).Taken together, endogenous expression of SH3PXD2A in ovarian cancer cells could be induced in an HBSS-dependent manner, treatment with RAD001 magnified autophagy and induced SH3PXD2A expression.Knockdown of endogenous ULK1 or treatment of MTOR activator-MHY1485 blocked HBSS-mediated SH3PXD2A activation, and such elevation on SH3PXD2A protein level was through its protein phosphorylation in ovarian cancer cells.These data suggested that SH3PXD2A was the downstream signal of autophagy, MTOR and ULK1 in ovarian cancer cells and its induction might be modulated by protein phosphorylation.

ULK1 interacted with SH3PXD2A in ovarian cancer cells
To determine whether ULK1 could interact with SH3PXD2A, in turn inducing phosphorylation of SH3PXD2A, the protein association between them was analyzed via coimmunoprecipitation assay.The endogenous ULK1 complex in ovarian cancer cells were pulled down using its specific antibody, which was further recognized by the antibody against SH3PXD2A, showing that SH3PXD2A was complexed with ULK1 in three ovarian cell lines (Figure 2A).Notably, this interaction was further confirmed by overexpression of both ULK1 and SH3PXD2A plasmids in SKOV3 following immunofluorescence staining then investigated using a confocal microscope, showing the co-localization of the two proteins in cytoplasm within the ovarian cells (Figure 2B; upper panel).The endogenous expression of ULK1 and SH3PXD2A obtained using immunofluorescence staining were further validated using a confocal microscopy after HBSS stimulation.The ULK1 and SH3PXD2A were colocalized in the cytoplasm of SKOV3 with puncta distribution after HBSS treatment [36] (Figure 2B; lower panel).
To further dissect the interaction domains of ULK1 and SH3PXD2A, serially truncated ULK1and SH3PXD2A were constructed and overexpressed in 293 cells due to its high transfection efficiency.The total proteins were extracted from transfected cells and pulled down by their specific tag to identify the specific interaction regions.As shown in Figure 2C, full-length SH3PXD2A was found to be critical for MTOR and SH3PXD2A interaction, deletion of any portions of SH3PXD2A decreased its interaction with MTOR, indicating that an intact SH3PXD2A protein structure may be important for MTOR binding and SH3PXD2A dimerization (Figure 2C).The N-terminal PX-SH3 domain of SH3PXD2A protein was dispensable for the interaction of SH3PXD2A and ULK1; whereas the other four SH3 domains of SH3PXD2A were essential for ULK1 interactions; especially the C terminus SH3 domain of SH3PXD2A (clone SH3 4-5) (Figure 2C).
To map the SH3PXD2A interaction domain in ULK1 protein, the deletion of the C-terminal CTD domain of ULK1 (1-828) dramatically reduced its interaction with SH3PXD2A (Figure 2D); nevertheless, the N terminus of ULK1 (1-279) still interacted with SH3PXD2A (Figure 2D).These data highlight that the secondary structure of ULK1 was crucial for SH3PXD2A binding; the CTD domain and kinase domain of ULK1 both might be important for SH3PXD2A interaction.Taken together, these data indicated that the interaction between endogenous SH3PXD2A and ULK1 was confirmed by coimmunoprecipitation and cellular colocalization in cytoplasm of ovarian cells.The C-terminal SH3 domain of SH3PXD2A (SH3 4-5) interacts with the C-terminal CTD and N-terminal kinase domain of ULK1.

SH3PXD2A was a substrate of ULK1
Since the SH3PXD2A interacted with ULK1, the data might suggest that SH3PXD2A can be phosphorylated by ULK1 in an autophagy-dependent manner in ovarian cells.To address this issue, the SH3PXD2A and ULK1 expression plasmids were transfected into 293 cells in the presence of RAD001; then the SH3PXD2A was affinity isolated and detected with an anti-phospho-serine/threonine antibody using western blotting (Figure 3A).The phosphorylated SH3PXD2A was detected in the ULK1 co-transfected cells, not in the vehicle alone in in vivo kinase assay (Figure 3A).To identify the amino acid residues phosphorylated by ULK1, wild-type SH3PXD2A products of the in vitro kinase assay were digested with trypsin and subjected to liquid chromatography-mass spectrometry (LC-MS/MS) to identify phosphosphopeptides.Four peptides and six phosphorylated sites at serine residues (S112, S142, S146, S147, S175, and S348) of the SH3PXD2A tryptic peptides were determined by LC-MS/MS and compared with the absence of ULK1 protein.(Figure 3B).The mono-phosphorylated cancer cells (ES2, SKOV3 and Kuramochi) were treated with 1 µM RAD001 for 24 h, the expression of endogenous SH3PXD2A and ACTB were examined using western blotting with their specific antibodies.(C) Cells (ES2, SKOV3 and Kuramochi) were pre-treated with 1 µM MTOR activator MHY1485 for 24 h, and then culture medium was replaced with HBSS in absence or presence of MHY1485 for 30 min.The expression levels of SH3PXD2A p-MTOR, MTOR, p-ULK1, ULK1and ACTB were examined using western blotting with their specific antibodies.(D) Cell lysates with or without HBSS treatment were separated in a Phos-tag gel then probed with anti-SH3PXD2A and anti-ACTB antibodies.Arrows indicate the position of phospho-SH3PXD2A (p-SH3PXD2A) and unphospho-SH3PXD2A (SH3PXD2A), expression of ACTB was as a loading control in this experiment.(E) the ovarian cells were transfected with control siRNA or ULK1 siRNA for 48 h, following treated with or without HBSS for 30 min.The endogenous protein levels of SH3PXD2A, p-MTOR, MTOR, p-ULK1, ULK1 and ACTB were detected with specific antibodies by western blotting.The intensity of western blotting of SH3PXD2A and LC3B-II was normalized for ACTB, whereas the p-MTOR and p-ULK1 were normalized by endogenous total MTOR and ULK1 expression levels; respectively, following normalized by expression level of ACTB.The protein levels in vehicle control or absence of HBSS treatment was considered as one-fold (1X); the expression level of SH3PXD2A, LC3BII, p-MTOR and p-ULK1 protein were further normalized by vehicle control as relative fold changes that showed in the bottom panel of each figure.All results are expressed as mean ± standard errors from three independent experiments.Statistical significance was calculated with Student's t-test *p < 0.05.**p < 0.01.

Figure 2.
SH3PXD2A interacted with ULK1 and MTOR complex.(A) the interaction of endogenous ULK1 and SH3PXD2A was analyzed using co-immunoprecipitation assay in ovarian cancer cells.The endogenous ULK1 complex in ovarian cancer cell lysates (ES2, SKOV3 and Kuramochi) were co-precipitated with either IgG or anti-ULK1 antibody, followed by western blot with anti-SH3PXD2A antibody (up panel).The 1/20 total input protein for co-immunoprecipitation assay was used to peptides identified by MS/MS spectra are shown in Figure S2.One of peptides of SH3PXD2A, 142 VWLSSWAESPK 153 , with multiple potential phosphorylation residues, has three potential mono-phosphorylation events (Figure 3B).The extracted ion chromatography spectra and predicted phosphorylation sites of these three potent monophosphorylation features are shown in Figure S3.To further confirm these phosphorylation sites of SH3PXD2A, the single mutants (S112A, S142A, S146A, S147A, S175A, and S348A) and all six serine residue mutants (SH3PXD2A-[6A]) proteins were purified, coincubated with ULK1 protein in the presence of ATP, and finally recognized by an anti-phospho-serine/threonine antibody using western blotting (Figure 3C).Mutations on each residues of SH3PXD2A attenuated phosphorylation by ULK1; however the phosphorylation levels were decreased 60% in SH3PXD2A mutant-SH3PXD2A-[6A] compared to its wild type in the presence of ULK1 using in vitro kinase assay (Figure 3C-E).Furthermore, the induction of SH3PXD2A protein levels by RAD001 treatment was only blocked in SH3PXD2A-[6A] in 293 cells (Figure 3F, G, H).Taken together, the six serine residues 112, 142, 146, 147, 175, and 348 of the SH3PXD2A protein was important for phosphorylation by ULK1 and induction by an MTOR inhibitor-RAD001.

ULK1 maintained SH3PXD2A protein stability
To further investigate the effect of ULK1-mediated phosphorylation of SH3PXD2A, the expression plasmids of ULK1 and wild type SH3PXD2A or SH3PXD2A-[6A] were exogenously transfected into 293 cells to examine the stability of SH3PXD2A protein or its mutants (SH3PXD2A-[6A]) with or without the protein synthesis inhibitor cycloheximide (CHX) (Figure 4A,B).Protein stability of SH3PXD2A increased when co-expressed with wild-type ULK1 after treatment with CHX for 3 h compared with the vehicle control in 293 cells (Figure 4A).However, the half-life of the SH3PXD2A-[6A] mutant was approximately one hour after CHX treatment, which was shorter than that of the wild-type SH3PXD2A in 293 cells (Figure 4B).Since ubiquitinated proteins are a marker for protein degradation through the proteasome pathway [37], the SH3PXD2A-or SH3PXD2A-[6A]-expressing plasmids were expressed in 293 cells and then co-precipitated to detect the ubiquitination of proteins in the presence of the proteasome inhibitor-MG132.As shown in Figure 4C, the level of protein ubiquitination in SH3PXD2A-[6A] was higher than in wild-type SH3PXD2A alone in 293 cells; suggesting these serine residues were crucial for stability of SH3PXD2A protein (Figure 4C).Taken together, these data suggest that ULK1 regulated SH3PXD2A stability via protein phosphorylation, mutation of the six serine residues of SH3PXD2A blocked ULK1 phosphorylation and promoted protein degradation.

ULK1-SH3PXD2A pathway regulated cell migration
Since SH3PXD2A participates in invadopodia formation and regulates cell migration [38], cell mobility was analyzed using either the transwell migration assay or the wound healing assay to examine the role of the ULK1-SH3PXD2A axis in cell migration in ovarian cancer cell lines.Cell migration ability increased after treatment with RAD001 for 24 h compared to DMSO treated SKOV3 cells (Figure 4D,E).Knockdown of endogenous ULK1 or SH3PXD2A using its specific siRNA showed a large cellfree gap area or migrated cells indicating lower migration ability than control siRNA in the absence or presence of RAD001 (Figure 4D,E).Moreover, cell mobility was abolished when exogenously expressing the Halo-SH3PXD2A -[6A] plasmids into SKOV3 cells compared to its wildtype expression plasmid, as revealed by wound-healing or transwell migration assays (Figure 4F,G).Taken together, these findings indicated that cell migration increases in the ULK1-SH3PXD2A regulatory axis upon treatment with RAD001; mutation of phosphorylation sites of SH3PXD2A analyze the expression levels of SH3PXD2A, ULK1 and ACTB are shown in the lower panel.(B) SKOV3 cells were co-transfected with Halo-SH3PXD2A and NTAP-ULK1 vectors for 48 h, the ULK1-SH3PXD2A complex was detected with a confocal microscopy.SKOV3 cells were double stained with anti-Halo antibody (SH3PXD2A, red signal, upper-left panel), anti-CBP tag antibody (ULK1, green signal, upper-middle panel).The SKOV3 was treated with HBSS for 30 min, then detected endogenous expression of SH3PXD2A and ULK1 using their specific antibodies; anti-SH3PXD2A (red signal, lower-left panel), anti-ULK1 (green signal, lower-middle panel).The colocalization of two proteins was revealed as yellow signals by overlapping red and green signals (yellow signal, upper-right and lower-right panel).Negative control of SKOV3 cells were double stained with IgG control and anti-ULK1 antibody.Blue signals indicate the DAPI-stained cell nucleus.Yellow scale bar: 10 µm.(C) 293 cells were transfected with full-length RPTOR/Raptor (HA tag), full-length SH3PXD2A (CBP tag), the flag-tag SH3PXD2A constructs with full-length or truncated fragments (NC: control vector, FL: full-length, PX-SH3-1, PX-SH3-3, SH3 1-5 and SH3 4-5) for 48 h, then harvested for lysis.The amino acid sequence of SH3PXD2A was based on reference number NP_003556.2, the relative position of PX and SH3 domain versus deletion constructs are shown in the right panel.The protein lysate was pulled down using M2 beads (tag of SH3PXD2A), the SH3PXD2A-associated proteins complex was examined with anti-MTOR, anti-HA (RPTOR), anti-CBP (SH3PXD2A) and anti-ULK1 antibodies (upper panel: IP).The 1/20 total input protein for coimmunoprecipitation assay was used to analyze the expression levels of SH3PXD2A (flag-tag or CBP tag), ULK1, MTOR, RPTOR (HA) and TUBA are shown in the lower panel (Input).The intensity of western blotting results in pulled-down MTOR, HA-RPTOR, ULK1 and SH3PXD2A were normalized with the total protein input of MTOR, HA-RPTOR, ULK1 and SH3PXD2A expression levels.The intensity of western blotting with full-length flag-SH3PXD2A was considered as onefold; the relative fold changes for each pulled-down proteins were normalized by the level full length of flag-SH3PXD2A as shown in the right bottom panel.(D) Full-length SH3PXD2A (Flag tag) or control vector was co-transfected with the full-length or truncated ULK1 constructs (CBP tag, NC: control vector, FL: full length, 1-828, 1-279, 279-1050 and 828-1050) into 293 cells for 48 h then harvested for lysis.The protein lysate was pulled down using streptavidin beads (CBP-ULK1 vector also containing CBP and streptavidin binding peptide), and then detected SH3PXD2A protein using anti-Flag antibody by western blotting (IP, upper panel).The 1/20 total input protein for co-immunoprecipitation assay was used to analyze the expression levels of ULK1 (CBP tag), SH3PXD2A (flag) and GAPDH are shown in the lower panel (Input).The amino acid sequence of ULK1 was based on reference number NP_003556.2, the relative position of kinase and CTD domain versus deletion constructs are shown in right panel.The intensity of western blotting results in pulled-down SH3PXD2A was normalized for input SH3PXD2A expression levels, the expression level in full-length ULK1 transfectant was considered as one-fold.The relative fold changes of pulled-down levels in ULK1 deletion constructs were further normalized with full-length ULK1 that was showed in the bottom panel.All results are expressed as mean ± standard errors from three independent experiments.Statistical significance was calculated with Student's t-test *p < 0.05.**p < 0.01.

Site-direct mutagenesis SH3PXD2A reduced binding to cell membrane lipid and active-MMP14
SH3PXD2A is one of the substrates of SRC tyrosine kinase, which enhances and stabilizes SH3PXD2A binding to cell membrane phospholipids, further increases invadopodia formation [23]and accumulates matrix metalloproteinase MMP14 in the invapodia region to promote extracellular matrix degradation [39].To evaluate whether the phosphorylation sites of SH3PXD2A were crucial for binding to cell membrane phospholipid, the binding ability of purified SH3PXD2A and SH3PXD2A-[6A] proteins to phospholipids was examined using a phospholipid array.As shown in Figure 5A, mutation of the phosphorylation site of SH3PXD2A (SH3PXD2A-[6A]) decreased binding to PtdIns3P as compared to the wild-type SH3PXD2A protein.The activity of MMP14 was examined in ES2 cells overexpressing wild-type and mutant SH3PXD2A-[6A] expression plasmids, followed by pull-down with Halo resin following recognition by MMP14 antibody.The interaction with active MMP14 [40] was lower in SH3PXD2A-[6A] expressed cells than wild-type SH3PXD2A (Figure 5B).The pull-down of SH3PXD2A or SH3PXD2A-[6A] protein complex in ES2 cells was further examined by its MMPs activity using zymography assay, the SH3PXD2A-[6A] mutant showed lower MMP14 activity than the SH3PXD2A complex (Figure 5C).Taken together, the SH3PXD2A-[6A] mutant showed a lower binding ability to cell membrane phospholipids-PtdIns3Pand decreased activity of MMP14 in ovarian cells.

Expression levels of phospho-MTOR, phospho-ULK1, and SH3PXD2A in human ovarian and endometrial cancers
The previously established PDX models showed similar histologic and genetic characteristics in human gynecological cancers [41], which could be used as a model to analyze the ULK5-SH3PXD2A regulatory axis in vivo.Two PDX mice were treated with the vehicle control or RAD001 for four weeks.The expression of SH3PXD2A protein increased in tumor tissues of two PDX mice in the presence of RAD001, as revealed by western blotting or immunohistochemical staining, as compared with vehicle control (Figure 6A,B).To verify the ULK1-SH3PXD2A axis in clinical samples, the expression of p-MTOR, MTOR, p-ULK1, ULK1 and SH3PXD2A in ovarian and endometrial cancer tissues was examined by western blotting.The expression of SH3PXD2A in tumor tissues was characterized by lower p-MTOR and p-ULK1 expression (Figure 6C,D).To correlate the expression of SH3PXD2A with survival of patients with ovarian cancers, the RNA expression data obtaining from The Cancer Genome Atlas (TCGA) also showed a positive correlation between ULK1 and SH3PXD2A using ovarian cancer database (r = 0.52, p < 0.001, Figure 6E).Furthermore, patients with lower SH3PXD2A mRNA or protein expression were characterized by a more favorable overall survival (OS) in all cancer types or ovarian cancers (all cancers in TCGA RNA database, n = 4750, p < 0.001; ovarian cancer in TCGA RNA database, n = 212, p < 0.01; ovarian cancer in TCGA protein database, n = 114, p < 0.05) (Figure 6F-H).These findings suggest that SH3PXD2A may serve as a prognostic marker for patients with ovarian tumors.

Discussion
The results of this study indicated that the ULK1-SH3PXD2A pathway regulates ovarian cancer cell migration under nutrient-deficient conditions.SH3PXD2A formed a complex with MTOR and ULK1 under normal conditions (Figures 2 and  7A).Under starvation conditions, inhibition of MTOR triggered ULK1 to phosphorylate SH3PXD2A, promoted its binding to PtdIns3P in the cell membrane and recruiting MMP14 in invadopodia to facilitate cell migration of ovarian cancers.Furthermore, a reverse correlation between activated p-ULK1 and SH3PXD2A expression was also found in clinical specimens of ovarian or endometrial cancers.Other studies have also shown that inactivation of MTOR by starvation induces extracellular matrix degradation through SH3PXD2A-MMP14 in breast and pancreatic cancers [42], however, the association between MTOR and SH3PXD2A-MMP14 was unclear in the previous report.Our data revealed the last piece of the puzzle in this pathway and suggested a novel MTOR-ULK1-SH3PXD2A-MMP14 pathway in ovarian malignancies (Figure 7B).
During tumorigenesis, rapid tumor growth leads to nutrient deprivation owing to insufficient blood supply [43,44].Cell stress, such as hypoxia and ROS, accumulate during nutrient shortage to regulate cell behaviors, including pulled-down protein was then detected with an anti-serine/threonine antibody.The intensity of phosphoserine/threonine in vehicle control was considered one-fold; the relative fold change of phospho-serine/threonine in ULK1 transfectant was further normalized with vehicle control, which was shown in the lower panel.(B) the purified SH3PXD2A protein was subjected to in vitro kinase assay and subsequently characterized and potential phosphorylation sites were identified using LC-MS/MS; six phospho-serine residues were identified in four peptides.The amino acid position in SH3PXD2A protein is shown in each peptide; the red color (S: serine) in each peptide indicates detected phospho-serine residue using LC-MS/MS.(C, D) In vitro kinase assay.Wild-type, six single-point-mutants (S112A, S142A, S146A, S147A, S175A, and S348A) and all six serine residues mutant (SH3PXD2A-[6A]) incubated with or without ULK1 in presence of ATP.Phosphorylated SH3PXD2A was detected with an anti-threonine/serine antibody.(E) the intensity of phospho-serine/threonine (C, D) were normalized by input SH3PXD2A expression levels, the expression level of phospho-serine/threonine in wild type SH3PXD2A was considered as one-fold; the relative fold changes in mutants SH3PXD2A were further normalized with the level of wild type SH3PXD2A.(F, G) the Halo tag SH3PXD2A wild-type, six single-pointmutant (S112A, S142A, S146A, S147A, S175A, and S348A) and all six serine residues mutant (SH3PXD2A-[6A], G) were transfected into 293 cells and treated 1 µM RAD001 for 24 h in the completed medium before harvesting.Halo-tag SH3PXD2A proteins were analyzed using an anti-Halo antibody and endogenous TUBA was probed with the specific antibody.(H) the fold changes in SH3PXD2A protein levels were normalized with TUBA.The expression level of SH3PXD2A in DMSO treated cells was considered as one-fold and the relative level in SH3PXD2A in RAD001 treated cells was showed as fold change after normalized with the results of DMSO treatment.Results are shown as mean ± standard errors from three independent experiments.*p < 0.05, ***p < 0.001.autophagy and cell migration [43,45,46].Autophagy can recycle damaged proteins and organelles to generate metabolites and energy, in response to nutrient deprivation [47].Starvation-induced autophagy promotes cell migration in hepatocellular carcinoma and bladder cancer [32,48], nevertheless the detailed mechansims of the regulation of cell migration by autophagy are still unkown.Knockdown of autophagy pathway proteins inhibits cell migration and invasion in breast cancer and hepatocellular carcinoma [8,49].ULK1 phosphorylates EXOC7/EXO70 to inhibit metastasis in the breast [12].Starvation-induced autophagy activates ULK1 and RB1CC1 to suppress FAK activity and block cell migration [50].In contrast, autophagy enhances cell migration by degrading focal adhesion proteins such as NBR1, CBL/ c-CBL, and PXN (paxillin) [51,52].The genes involved in the epithelial-mesenchymal transition are induced by autophagy in response to stress [7,48].Repression of endogenous ATG9A (an autophagy core protein) impairs cell protrusion and migration [53].Suppression of ULK1 expression with specific siRNA blocks vascular smooth muscle cell migration induced by cytokine [54].These data indicated that autophagy plays diverse roles in regulating cell migration under different conditions and cell types.In our study, HBSS treatment increased autophagy and SH3PXD2A expression in a short time, furthermore such activation was also induced by RAD001.These findings suggest that in the early stages of starvation, cancer cells depend on the starvation-induced autophagy-promoted ULK1-SH3PXD2A pathway to promote migration and seek nutritional sufficiency to avoid starvationinduced apoptosis.
RAD001 is a rapamycin derivative, inhibits MTOR activity and promotes autophagy [55].Accumulated results support the hypothesis that RAD001 blocks cell proliferation, migration and angiogenesis in various cancers [56].However, unexpected results have been reported after treatment with RAD001 in various cancer cells.An in vivo study on pancreatic cancer showed the occurrence of distant metastasis when administrated with RAD001 [57].Mesenchymal marker genes, MMP9, and cell motility are upregulated by RAD001 in human-derived renal cells and pneumocytes [58,59].Renal transplant patients who received RAD001 as a clinical protocol had a high risk of developing lung fibrosis [59].Long-term treatment with RAD001 changed the expression of integrins and enhanced cell migration in bladder cancer and prostate cancer [60,61].In this study, we also showed that RAD001 promoted SH3PXD2A expression both in vitro (Figure 1B) and in vivo (Figures 6A,B).Furthermore, increased cell migration by RAD001 was observed in ovarian cancer cells (Figures 4D,F).Knockdown of ULK1 and SH3PXD2A by specific siRNA blocked cell migration ability induced by RAD001.These results suggest that this novel signaling pathway is involved in regulating cell migration by treating RAD001 in ovarian cancers, which might highlight that patients with SH3PXD2A positive tumors should avoid the RAD001 treatment.However, this study has some limitations.The enrolled cases were restricted and the SH3PXD2A positive rate was low (7/29 in ovarian cancer specimen; 6/19 endometrial cancer specimen) in our enrolled clinical specimen.However, the ovarian cancer patients (n = 114) with higher expression of SH3PXD2A protein in tumor lesion tended to have a worse prognosis than those with lower SH3PXD2A expression using TCGA protein database (Figure 6H).These findings suggest that SH3PXD2A may serve as a prognostic marker for patients with ovarian tumors.Nevertheless, the detailed mechanism of RAD001-induced cell migration requires further investigation.
As a core protein of autophagy, ULK1 is overexpressed in hepatocellular carcinoma [62], nasopharyngeal carcinoma [63], colorectal cancer [64], esophageal squamous cell carcinoma [65] and clear cell renal carcinoma [66], and is considered a poor prognostic marker in these cancers.Specific siRNAs or inhibitors to block ULK1 promote apoptosis and inhibit tumor cell growth [63,66].Our study identified a new ULK1 interaction protein, SH3PXD2A, associated with the MTOR complex (Figure 2C).ULK1 phosphorylated SH3PXD2A S112, S142, S146, S147, S175, and S348 residues and regulated SH3PXD2A protein stability (Figure 3).In comparison with the wild-type, the site-directed SH3PXD2A mutant had lower binding ability to PtdIns3P (Figure 5A), MMP14 (Figure 5B), and migration ability (Figures 4F,G).Ovarian and endometrial tumor tissues were further confirmed the expression of ULK1-SH3PXD2A pathway.Overexpression of SH3PXD2A in tumor tissues was associated with lower phospho-ULK1 at serine 757, which represents inactive ULK1 (12 out of 13 samples, Figures 6C,D).Therefore, ULK1 May play a role in ovarian tumorigenesis and serve as a potential drug target in ovarian cancer.
In conclusion, we identified a novel ULK1-SH3PXD2A-MMP14 axis that regulated cellular migration during starvationinduced autophagy in ovarian cancer.Specifically, we demonstrated that ULK1 phosphorylates SH3PXD2A to regulate its protein stability and migration ability.The expression levels of SH3PXD2A was negatively correlated with inactive phospho-ULK1 in tumor tissues and might be used as a therapeutic target to prevent metastasis in this malignancy.
SH3PXD2A and Halo-SH3PXD2A-[6A] were overexpressed in 293 cells and treated with MG132 (10 µM) for 5 h before harvesting.Halo-SH3PXD2A and Halo-SH3PXD2A -[6A] were pulled down with anti-Halo resin and subsequently detected ubiquitin-labeled proteins with an anti-ubiquitin antibody.Specific antibodies were used to detect Halo-SH3PXD2A, Halo-SH3PXD2A-[6A], and GAPDH in 1/20 input protein.Results for ubiquitined SH3PXD2A were normalized for input SH3PXD2A expression levels, the intensity of ubiquitined wild-type SH3PXD2A was considered one-fold; the relative fold changes in ubiquitined SH3PXD2A-[6A] were further normalized using wild-type SH3PXD2A level (bottom panel).(D, E) SKOV3 cells were transfected with specific siRNAs for 48 h, then (D) wound closure assay or (E) transwell migration for 24 h in the absence or presence of 1 µM RAD001 was performed.The cell-free gap area was quantified using ImageJ and plotted in the right panel.Cells with higher migration ability produced smaller cell-free gap areas.The migrated cells in each transfected cells (E) was calculated using an inverted microscope in 5 high-power fields.The migrated cells in wild type SH3PXD2A transfectant was considered as one-fold; whereas the migrated cells in SH3PXD2A-[6A] transfectant was showed as relative fold change after normalized with results of wild type SH3PXD2A.(F, G) SKOV3 cells were overexpressed in wild-type SH3PXD2A and SH3PXD2A-[6A], then (F) wound closure assay, and (G) transwell migration assay was performed.The migration ability was quantified by ImageJ for gap area and plotted in the bottom panel (F); whereas the migrated cells was counted and plotted in right panel (G).Results in (D, E, F, G) are presented as mean ± standard errors of the mean from three independent experiments.Asterisks denote statistical significance (*p < 0.05, **p < 0.01).(A) Purified wild-type or mutant SH3PXD2A were hybridized with the lipidprotein interaction array membrane at 4°C overnight.After washing, the membrane-bound protein was detected with a specific anti-SH3PXD2A antibody.The binding intensity of wild-type SH3PXD2A was considered one-fold; the relative fold changes for SH3PXD2A-[6A] were further normalized using wild type SH3PXD2A signal (bottom panel).(B) Wild-type or mutant SH3PXD2A complex was pulled down using Halo tag resin.The associated protein in the complex was detected using western blot with an anti-MMP14 antibody.Results for pulled-down MMP14 was normalized using input MMP14 expression levels.The intensity of pulled-down protein from wild-type SH3PXD2A was considered one-fold; the relative fold changes of SH3PXD2A-[6A] were further normalized using wild type SH3PXD2A signal (bottom panel).(C) Halo tag wild-type or mutant SH3PXD2A were overexpressed in ES2 cells, then purified with halo tag resin.The MMPs activity in the SH3PXD2A complex was measured with zymography.The expression of active MMP14 (as arrowhead indicated) of Halo-SH3PXD2A expressing cells considered one-fold; the relative fold changes of Halo-SH3PXD2A-[6A] were further normalized with level of Halo-SH3PXD2A (bottom panel).All results are expressed as mean ± standard errors from three independent experiments.Statistical significance was calculated with Student's t-test *p < 0.05.**p < 0.01.

Protein purification and in vitro kinase assay
Wild-type and mutant SH3PXD2A Halo tag expressing plasmids were overexpressed in 293 cells.Transfected cells were harvested 24 h after transfection, lysed and purified using the Halo tag protein purification system (Promega) according to the manufacturer's instructions.Briefly, cell pellets were lysed in Halo Tag purification buffer (1× PBS, 1 mM DTT [ThermoFisher Scientific, Y00147], 0.005% IGEPAL-CA630 [Sigma-Aldrich, 18896], and 1X protease inhibitor [Bionovas Biotechnology, Toronto, Canada, FC00700001]).After three freeze-thaw cycles, the supernatants were added to HaloLink resin (Promega, G1912), and the halo tag fusion protein was released from the resin using HaloTEV (Promega, G6601) enzyme to cleave the TEV cutting site between the Halo tag and SH3PXD2A protein.Purified SH3PXD2A proteins were determined with western blot using specific antibodies.GAPDH was used as the loading control.The red number indicates SH3PXD2A expression in clinical samples.The expression for SH3PXD2A were normalized using GAPDH.The expression of p-MTOR and p-ULK1 was normalized with MTOR and ULK1; respectively, following normalization with GAPDH.The fold changes for SH3PXD2A (red), p-MTOR (green) and p-ULK1 (blue) protein levels are reported in the bottom panel.(E) Correlation of the mRNA expression between ULK1 and SH3PXD2A in ovarian cancer using TCGA ovarian cancer mRNA sequence database (http://oncodb.Org/).(F-H) Kaplan-Meier plots of overall survival of all cancers (F), ovarian cancer (G) SH3PXD2A mRNA and (H) SH3PXD2A protein expression in the TCGA database.incubated with ULK1 (SignalChem, Canada, U01-11 G) in kinase assay buffer 1 (SignalChem, Canada, K01-09) in the presence of ATP at 30°C for 30 min.Finally, the phosphorylation levels of purified SH3PXD2A were determined using western blotting with an anti-phospho-serine/threonine antibody (Millipore, AB1603).

Lipid-protein interaction array
The interaction between the SH3PXD2A protein and phospholipids was evaluated using a PIP array (Echelon Biosciences, P-6100).In brief, purified SH3PXD2A proteins with Halo Taq purification system from 293 cells were incubated with PIP array membrane overnight at 4°C.After washing with PIP array membrane, membrane-bound SH3PXD2A was detected with anti-SH3PXD2A antibody (Cell Signaling Technology, 16619) and corresponding horseradish peroxidase-conjugated antibodies (Santa Cruz Biotechnology).

Transwell migration assay
SKOV3 cells (1 × 10 4 in 300 μL of serum-free DMEM/F12) were loaded onto the upper part of the chemotaxis chamber (8-μm pore; Corning, CLS3422), followed by insertion of the upper chamber into a 24-well plate with 10% DMEM/F12 (800 μL).After 24 h incubation at 37°C, the membrane was fixed in methanol for 10 min and stained in 0.5% crystal violet (percent weight/volume in 20% methanol) solution for 1 h.The cells that migrated to the membrane were counted under a microscope.

Wound closure assay
Wound closure assay was performed using a two-well migration insert with a defined cell-free gap (ibidi, 81176).Cells (density: 1 × 10 5 cells/well) were seeded overnight into a twowell migration insert, which was removed to create a cell-free gap.Upon addition of serum-free medium, images of the cellfree gap were acquired at baseline and at the indicated incubation time.Cell-free gap areas were quantified using ImageJ software (https://imagej.nih.gov/ij/) after normalizing the corresponding gap area at baseline.

Collection of human ovarian and endometrial cancer specimens
Serous type of ovarian cancer and endometrioid endometrial cancer specimens were collected from women who had undergone surgery after obtaining written informed consent.The study was approved by the appropriate Institutional Review Board (Linkou Chang Gung Memorial Hospital approval number: 202101991B0C601).

Patient-derived xenograft (PDX) model
The experimental and animal care protocols were approved by the Animal Care and Use Committee (IACUC) of Chang-Gung Memorial Hospital (IACUC No. 2018072601).All tumor tissues were obtained from our previous study [41].Briefly, fresh ovarian or endometrial cancer tissues were minced and inoculated into the subcutaneous regions of severe combined immunodeficiency (SCID) mice.Once the tumor volume reached 100 mm 3 , SCID mice were orally administered vehicle control or RAD001 (2.5 mg/kg) twice a week for four weeks.Finally, the xenografted tumor tissues were lysed in RIPA buffer for protein extraction or fixed with formalin.

Confocal microscopy
After overnight transfection on a coverslip, SKOV3 cells were fixed with 4% paraformaldehyde, immersed in ice-cold acetone for 20 min at − 20°C, and incubated with blocking buffer (Thermo Fisher Scientific, TA-060-PBQ) for 1 h at room temperature (RT).Halo-SH3PXD2A and NTAP-ULK1 or anti-SH3PXD2A and anti-ULK1 (Santa Cruz Biotechnology, sc -390,904) antibodies were detected by incubating cells with an antibody raised against the Halo tag and CBP tag or endogenous SH3PXD2A and ULK1, followed by exposure to the corresponding fluorescent antibodies (Thermo Fisher Scientific, Alexa Fluor 488, A11001; Alexa Fluor 549, A11010).Finally, the slides were examined under a Leica TCS SP2 confocal laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany).

Zymography
MMP activity in the SH3PXD2A complex was detected using zymography.Overexpressed wild or mutant SH3PXD2A was purified from ES2 cells and separated using SDS-PAGE with 1% gelatin under non-reducing conditions.The gels were washed with 2.5% Triton X-100 and incubated in the development buffer for 48 h.After staining with 0.25% Coomassie Brilliant Blue R-250 (Bio-Rad, 1610436) and de-staining in staining solution without dye, MMP activity was observed as clear bands in the blue gelatin gel.

Autophagy assay
Autophagic flux was detected using an autophagy assay kit (Abcam, Cambridge, UK, ab139484).In brief, trypsinized cells were incubated with a fluorescence indicator at 37°C for 30 min and subsequently fixed with 4% formaldehyde for 20 min at RT.After washing, the cells were examined using a BD LSRFortessa (BD Biosciences).

Protein digestion
Proteins from the ULK1 in vitro kinase assay were reduced with dithiothreitol, alkylated with iodoacetamide, and digested with sequencing-grade trypsin (Promega, V5111).Tryptic digests were acidified with trifluoroacetic acid and desalted using Sep-Pak C18 cartridges (Waters Associates, Milford, MA, USA).

Tandem mass spectrometry-based protein identification
Tandem mass spectrometry was performed using a Q Exactive TM HF mass spectrometer coupled with an UltiMate™ 3000 RSLCnano System (Thermo Fisher).Peptide mixtures were directly injected into an analytical column (ES903 C18 Column, 100 A, 2 µm, 75 µm × 500 mm, Thermo Scientific) and separated through a 120 min acetonitrile gradient with a flow rate of 200 nL/min.Peptide spectra were acquired in the positive ion mode (scan range: 400-1600 m/z).The 15 most abundant precursor ions were dynamically selected for further fragmentation in high-collision dissociation mode (with normalized collision energy set at 28 or 34).The resolution of the full MS scan was set at 60,000 with an m/z of 200 and automatic gain control (AGC) of 3e6.The following parameters were used for the MS/MS scan: resolution, 15000; AGC target,5e4; maximum injection time, 50 ms.Dynamic exclusion of the selected precursor ions was set to 20 s.

MS-based proteomic data analysis by proteome discoverer software and mascot search engine
The MS raw files were uploaded to Proteome Discoverer (version 2.4, Thermo Scientific) software.The peak list for protein identification was generated using the Mascot search algorithm (version 2.5.1,Matrix Science, London, UK) against the Swiss-Prot human protein database.The methylthio group of cysteine was set as a fixed modification.Oxidized methionine, phosphorylated serine or threonine, N-terminal glutamate of peptides converted to pyroglutamate, and N-terminal acetylation of proteins were considered as dynamic modifications.Searches were done by allowing a single missed cleavage and with a mass accuracy set to ±10 ppm for precursor ions and ±0.02 Da for peptide fragments.An overall false discovery rate of below 1% was used to obtain confidence in peptide and protein identification.Moreover, only proteins identified with at least two unique and highly confident peptides were accepted.

Data analysis
Detailed descriptions of the statistical tests are provided in the figure legends.Correlation analysis between ULK1 and SH3PXD2A mRNA expression was performed on the OncoDB website (http://oncodb.org/)[68] using the TCGA sequence dataset for ovarian cancer mRNA.Overall survival was examined using gene and expression profiling interactive analysis using ovarian cancer and all TCGA cancer RNA datasets (GEPIA2, http://gepia2.cancer-pku.cn/#survival),and protein expression dataset (NIH Proteomic Data Commons, https://pdc.cancer.gov/pdc/,study ID: PDC000114).Protein database analyses were performed using SPSS, version 18 (IBM, USA).

Figure 1 .
Figure 1.SH3PXD2A was activated by starvation and inhibition of MTOR activity in ovarian cancer cells.(A) the expression of endogenous SH3PXD2A protein was detected in ovarian cancer cells (ES2, SKOV3 and Kuramochi) after treatment with Hanks' balanced salt solution (HBSS) in various times using western blotting.The expression of phospho-MTOR (p-MTOR), total MTOR, phospho-ULK1 (p-ULK1), total ULK1, LC3B-I/II and ACTB were also detected using western blotting.(B) Ovarian

Figure 3 .
Figure 3. SH3PXD2A was a ULK1 kinase substrate.(A) In vivo kinase assay.The flag-tag SH3PXD2A expression plasmid was co-transfected with vehicle control or CBPtag ULK1 expression plasmid into 293 cells for 24 h then treated with 1 µM RAD001, followed by purifying the expressed SH3PXD2A protein using M2 beads.The

Figure 4 .
Figure 4. ULK1 regulated the stability and function of SH3PXD2A.(A) Halo-SH3PXD2A were co-transfected with control vector or NTAP-ULK1into 293 cells for 48 h.(B) Halo-SH3PXD2A and Halo-SH3PXD2A-[6A] were overexpressed in 293 cells for 48 h, then treated with 10 µg/mL cycloheximide (CHX) at indicated time before harvesting.The expression of Halo-SH3PXD2A and Halo-SH3PXD2A-[6A] were detected with an anti-Halo antibody, whereas NTAP-ULK1 was probed with an anti-CBP antibody.Endogenous GAPDH or ACTB was used as loading control, the expression of SH3PXD2A followed by normalization with expression level of GAPDH (A) or ACTB (B).The expression level of SH3PXD2A before CHX treatment (0 h) was considered as one-fold, the relative fold changes in Halo-SH3PXD2A or Halo-SH3PXD2A -[6A] protein levels were shown in the bottom panel.Results are shown as mean ± standard errors from three independent experiments.*p < 0.05.(C) Halo-

Figure 5 .
Figure 5. SH3PXD2A mutant decreases binding ability to PtdIns3P and active MMP14.(A)Purified wild-type or mutant SH3PXD2A were hybridized with the lipidprotein interaction array membrane at 4°C overnight.After washing, the membrane-bound protein was detected with a specific anti-SH3PXD2A antibody.The binding intensity of wild-type SH3PXD2A was considered one-fold; the relative fold changes for SH3PXD2A-[6A] were further normalized using wild type SH3PXD2A signal (bottom panel).(B) Wild-type or mutant SH3PXD2A complex was pulled down using Halo tag resin.The associated protein in the complex was detected using western blot with an anti-MMP14 antibody.Results for pulled-down MMP14 was normalized using input MMP14 expression levels.The intensity of pulled-down protein from wild-type SH3PXD2A was considered one-fold; the relative fold changes of SH3PXD2A-[6A] were further normalized using wild type SH3PXD2A signal (bottom panel).(C) Halo tag wild-type or mutant SH3PXD2A were overexpressed in ES2 cells, then purified with halo tag resin.The MMPs activity in the SH3PXD2A complex was measured with zymography.The expression of active MMP14 (as arrowhead indicated) of Halo-SH3PXD2A expressing cells considered one-fold; the relative fold changes of Halo-SH3PXD2A-[6A] were further normalized with level of Halo-SH3PXD2A (bottom panel).All results are expressed as mean ± standard errors from three independent experiments.Statistical significance was calculated with Student's t-test *p < 0.05.**p < 0.01.

Figure 6 .
Figure 6.The expression of SH3PXD2A in vivo and its clinical impact on survival of patients with ovarian cancers.(A, B) Ovarian cancer and endometrial cancer PDX mice were treated with vehicle or RAD001(2.5 mg/kg) twice in a week for four weeks.SH3PXD2A expression levels were measured using (A) western blot, and (B) immunohistochemistry. ACTB was used as the loading control in (A).Expression of SH3PXD2A was normalized by expression level of ACTB.(C, D) Endogenous expression levels of SH3PXD2A, p-MTOR (S2448), MTOR, p-ULK1(S757), ULK1, and GAPDH in ovarian cancer (C) and endometrial cancer (D) were

Figure 7 .
Figure 7. Schematic representation of the starvation-induced cell migration through ULK1-SH3PXD2A-MMP14 axis.(A) Summary of the interaction domains between SH3PXD2A and ULK1.(B) Starvation blocked MTOR activity and activated ULK1.Upon phosphorylation by ULK1, SH3PXD2A interacted with PtdIns3P and MMP14 in invadopodia, to promote cell migration.Inhibition of ULK1 activity resulted in SH3PXD2A degradation.