Abstract
Inappropriate activation of PI3K signaling has been implicated strongly in human cancer. Although studies on the role of PI3K signaling in breast tumorigenesis and progression have focused most intensively on PI3Kα, a role for PI3Kβ has begun to emerge. The PI3Kβ isoform is unique among class IA PI3K enzymes in that it is activated by both receptor tyrosine kinases and G-protein–coupled receptors (GPCR). In previous work, we identified a mutation that specifically abolishes PI3Kβ binding to Gβγ (p110526KK-DD). Expression of this mutant in p110β-silenced breast cancer cells inhibits multiple steps of the metastatic cascade in vitro and in vivo and causes a cell autonomous defect in invadopodial matrix degradation. Our results identify a novel link between GPCRs and PI3Kβ in mediating metastasis, suggesting that disruption of this link might offer a novel therapeutic target to prevent the development of metastatic disease. Cancer Res; 76(10); 2944–53. ©2016 AACR.
Introduction
Metastasis is a highly complex process that requires coordinated, bidirectional interactions between tumor cells and their microenvironment (1). These interactions are often mediated by chemokines and their receptors, many of which belong to the G-protein–coupled receptor (GPCR) family of receptors (2). In tumors, local production of chemokines plays an important role in attracting leukocytes and other inflammatory cells (3). Similarly, tumor cells express receptors that respond to chemokines secreted by stromal cells, tumor-associated macrophages, tumor-infiltrating leukocytes, as well as chemokines secreted by adjacent tumor cells or in an autocrine manner (2). Numerous chemokines have been shown to promote cellular behaviors associated with breast tumor metastasis, including cell migration, increased expression and activity of matrix metalloproteinases (MMP), increased heparanase activity, and invasion (4–12). Expression of chemokine receptors such as CXCR4 and CCR7 in tumor cells has also been linked to the dissemination of breast cancer cells to secondary organ sites that express high levels of their respective ligands (12, 13). In addition to chemokine receptors, several other types of GPCRs, including protease-activated receptor-1 (PAR1), lysophosphatidic acid (LPA), and prostaglandin (EP) receptors, have been implicated in breast cancer metastasis (14–18).
Ligand-stimulated GPCRs signal through both activated Gα and Gβγ subunits (19). Among the downstream effectors of Gβγ are the class I PI 3-kinases, PI3Kβ and PI3Kγ (20–22). Gβγ activates these PI3K isoforms by directly binding to the p110β and p110γ catalytic subunits (23, 24). Given the high frequency of PIK3CA mutations in breast cancer patients (25), studies on breast cancer tumorigenesis and progression have focused most intensively on p110α (26), which is regulated primarily by receptor tyrosine kinases (RTK). However, a role for the GPCR-coupled PI3Ks has begun to emerge. For example, loss of p110γ activity reduces tumor growth and myeloid cell recruitment in the MMTV-PyMT breast cancer model (27). Similarly, p110β kinase-dead transgenic mice are protected from tumor progression in an ErbB2-driven mammary cancer model (28). Moreover, clinical data show that p110β expression correlates with poor prognosis and distant metastasis in breast cancer patients (29). A recent study also shows an important role for p110β in the resistance of tumors to inhibitors of p110α (30).
The p110β isoform is unique among class IA PI3Ks as it can be activated both by RTKs, through p85 SH2 domain binding to activated receptors, and by GPCRs (31). We previously defined the region of p110β that binds to Gβγ and identified a mutation that abolishes this interaction (p110β526KK-DD; ref. 23). This mutant p110β is activated normally by binding to tyrosine phosphorylated peptides, but is specifically defective for activation by GPCRs. Using p110β526KK-DD, we now demonstrate a previously unidentified role for GPCR signaling to p110β in tumor cell–macrophage paracrine interactions and in invadopodia function, which affects multiple steps of the metastatic process. Our data suggest that disruption of the Gβγ–p110β interaction might constitute a novel pharmacologic target for the prevention of breast cancer metastasis.
Materials and Methods
Cell culture and stable cell lines
The human breast cancer cell lines MDA-MB-231, BT-549, and T47D were obtained from the ATCC and maintained as recommended. Bone marrow–derived macrophages (BMM) were isolated from tibias and femurs of C57BL/6 mice as previously described (32) and grown in Complete BMM Medium (αMEM (Invitrogen) containing 15% FBS, 120 ng/mL recombinant human CSF-1, 0.02 mg/mL asparagine, 2 mmol/L glutamine, and 100 U/mL penicillin and 100 μg/mL streptomycin. Bac1.2F5 macrophages were grown in αMEM (Invitrogen) containing 10% FBS, 36 ng/mL recombinant human CSF-1, 0.02 mg/mL asparagine, and 2 mmol/L glutamine. 3B-11 endothelial cells were obtained from the ATCC and grown in DMEM containing 10% FBS. All media were supplemented with 100 U/mL penicillin and 100 μg/mL streptomycin
MDA-MB-231, BT-549, and T47D cells were infected with lentivirus-expressing shRNA against human p110β (target sequence: 5′GCGGGAGAGTAGAATATGTTT3′) and selected using 5 μg/mL puromycin (Sigma). Stable knockdown cell lines were confirmed by Western blot analysis for endogenous p110β and then transfected with cDNA expressing myc-tagged murine wild type p110β, p110βKK-DD or p110βK-R using Lipofectamine 3000 (Life Technologies). Stable transfectants were selected using 800 μg/mL Geneticin (Enzo Life Sciences) and expression of exogenous p110β was verified by Western blotting with antibodies against p110β and the myc epitope.
Antibodies and reagents
The rabbit p110β antibody was produced in-house. Rabbit anti-p110α, phospho-AKT (Thr308), total AKT, and phospho-ERK1/2 (Thr202/Tyr204) were purchased from Cell Signaling Technology. The pan ERK antibody was purchased from BD Biosciences. LPA and EGF were purchased from Sigma and Millipore, respectively. SDF-1 and IL8 were purchased from R&D Systems. TGX-221 was purchased from Echelon.
Western blotting
For Western blot experiments, MDA-MB-231, BT-549, or T47D cells expressing wild-type or mutant p110β grown in 6-well plates were starved for 18 hours in serum-free media supplemented with 0.1% BSA and then stimulated for 5 minutes at 37°C with 5 nmol/L EGF or 10 μmol/L LPA. Cells were lysed in 2X Laemmli sample buffer with phosphatase inhibitor cocktails (Sigma and Calbiochem) and boiled for 5 minutes. Cell lysates were separated by SDS-PAGE and blots were visualized using ECL (Thermo Scientific) and quantitated by densitometry.
Transwell migration and invasion assays
For migration assays, MDA-MB-231 cells stably expressing wild-type or mutant p110β were starved overnight in serum-free DMEM containing 0.1% BSA. A total of 3 × 104 cells were plated in starvation media in the upper wells of Transwell chambers (8 μm pore; BD Biosciences). For macrophage-stimulated migration, 1 × 106 BMMs were plated in the lower chamber in Complete BMM Medium and incubated overnight. The media were changed to serum-free starvation media supplemented with 120 ng/mL CSF-1 before plating the tumor cells in the transwells. For LPA-stimulated migration, starvation media without or with 10 μmol/L LPA were placed in the lower chamber. MDA-MB-231 cells were allowed to migrate for 8 hours.
For invasion assays, Transwell chambers (8-μm pore; BD Biosciences) were coated with 300 μg/mL growth factor-reduced Matrigel (BD Biosciences) for 2 hours at 37°C. MDA-MB-231, BT-549, or T47D cells were starved in serum-free media containing 0.1% BSA for 6 hours. For ligand-stimulated invasion, starvation media without or with 100 ng/mL SDF-1, 100 ng/mL IL8, or 5 nmol/L EGF was placed in the lower chamber. For macrophage-stimulated invasion, 1 × 106 BMMs were plated in the lower chamber in Complete BMM Medium and incubated overnight. The media were changed to serum-free starvation media supplemented with 120 ng/mL CSF-1 before plating the tumor cells in the Transwells. Hundred nmol/L TGX-221 was included as indicated. Tumor cells were allowed to invade for 24 hours.
After 8 hours for migration assays and 24 hours for invasion assays, cells on the upper surface of the membrane were removed with a cotton swab. Cells on the lower surface of the membrane were fixed with 4% paraformaldehyde in PBS for 10 minutes, washed twice with PBS, and stained with DAPI. Five random fields at ×10 magnification from at least two Transwells per condition were imaged, and the cells were counted using ImageJ.
3D collagen invasion assay
3D in vitro invasion assays were performed and quantified as described previously (33). Briefly, Bac1.2F5 macrophages and MDA-MB-231 cells stably expressing wild-type or mutant p110β were labeled with CellTracker Red CMPTX and CellTracker Green CMFDA, respectively. Eighty thousand tumor cells were plated on MatTek dishes without or with 200,000 Bac1.2F5 macrophages and grown in BAC1.2F5 medium for 16 hours. Cells were overlaid with 5.8 mg/mL type I collagen, incubated for 24 hours and fixed. Invasion into the collagen gel was quantified by laser scanning confocal microscopy detection of the fluorescent signals from the red and green CellTracker dyes.
Extravasation-transendothelial migration assay
Transwell chambers (8-μm pore; BD Biosciences) were coated with 300 μg/mL growth factor-reduced Matrigel (BD Biosciences) for 2 hours at 37°C. A total of 2 × 104 3B-11 endothelial cells were plated on the Matrigel layer and incubated for 48 hours at 37°C to allow the formation of a tight monolayer, as indicated by resistance measurement. A total of 104 BMMs were plated on the underside of the membrane and allowed to attach for 30 minutes. MDA-MB-231 or BT-549 cells stably expressing wild-type or mutant p110β were labeled with CellTracker Green CMFDA dye (Invitrogen) in serum-free medium for 30 minutes at 37°C. A total of 2 × 104 tumor cells were plated on top of the endothelial cell monolayer in the upper chamber of the Transwells and allowed to migrate for 36 hours at 37°C. After removing non-migrated cells with a cotton swab, cells on the lower surface of the membrane were fixed with 4% paraformaldehyde for 10 minutes and washed twice with PBS. Six random fields at ×20 magnification from duplicate or triplicate wells for each condition were imaged using a fluorescent microscope.
Experimental metastasis
A total of 4 × 105 MDA-MB-231 cells stably expressing wild-type or mutant p110β were injected intravenously into the lateral tail vein of SCID mice, respectively. After 6 weeks, the mice were sacrificed. Lungs were collected, fixed in 10% neutral buffered formalin, and embedded in paraffin followed by serial sectioning. Lung sections were stained with hematoxylin and eosin and scanned. The tumor nodules were quantified by thresholding the images using ImageJ software to determine the number of nodules per lung section as well as the size of individual nodules, expressed in arbitrary units.
Xenografts
A total of 2 × 106 MDA-MB-231 cells stably expressing wild-type or mutant p110βwere injected into the right fourth mammary fat pad of 6- to 8-week-old SCID mice. Tumors were measured three times per week, and tumor mass was calculated using the formula tumor mass (g) = 0.1 × length in mm × (0.1 × width in mm)2. Mice were sacrificed when the tumor mass reached approximately 1 gram.
Gelatin degradation
MDA-MB-231 cells stably expressing wild-type or mutant p110β were plated on glass coverslips coated with Oregon Green 488–conjugated gelatin (Molecular Probes) as previously described (34). Briefly, coverslips were treated with 50 μg/mL poly-l-lysine for 10 minutes at room temperature followed by 0.5% glutaraldehyde for 10 minutes at room temperature. The treated coverslips were then coated with 200 μg/mL gelatin for 15 minutes at room temperature, treated with 0.1 mol/L glycine for 10 minutes, and extensively washed with PBS. A total of 4 × 104 tumor cells in DMEM containing 10% FBS were plated on the coverslips and incubated for 18 hours. Cells were then fixed, stained with rhodamine phalloidin, and immunostained for cortactin. At least 10 fields per condition were imaged at ×60 magnification as described above. To quantify matrix degradation, experiments were performed in triplicate with a minimum of 120 cells per condition examined. Cells with at least one degraded spot were counted as positive for gelatin degradation. The area of degradation per field was measured by thresholding the images using ImageJ software to determine the total area in the field that lacks fluorescence. The total area was then divided by the number of degrading cells in the field.
Statistical analysis
Quantitative data are expressed as the mean ± SEM from three independent experiments. Statistical analysis was performed using ANOVA followed by Holm-Sidak post hoc testing to correct for multiple comparisons. A P value less than 0.05 was considered statistically significant.
Results
Expression of GPCR-uncoupled or kinase-dead PI3Kβ inhibits GPCR signaling in breast cancer cells
We tested the role of PI3Kβ in two triple-negative breast cancer cell lines. MDA-MB-231 cells express mutant Ras and Raf, but PI3Ks and PTEN are wild-type. In contrast, BT-549 expresses a truncated catalytically inactive mutant of PTEN (35). We generated stable MDA-MB-231 and BT-549 human breast cancer cell lines in which endogenous p110β was knocked down by infecting cells with a lentivirus expressing an shRNA targeting p110β. The knockdown cells were then transfected with cDNA constructs encoding the murine wild type p110β or two previously described mutants that abolish its binding to Gβγ (mutation of residues 526KK to DD; p110βKK-DD; ref. 23) or its lipid kinase activity (mutation of residue 799K to R; p110βK-R; ref. 28). The exogenous murine p110β was expressed at levels that were comparable with that of endogenous p110β in control cells (Fig. 1A). Although a myc blot showed a slightly higher expression level for the p110β mutants as compared with wild-type p110β (Supplementary Fig. S1A), expression of p110α was not affected in either cell line (Supplementary Fig. S1B).
Stable replacement of endogenous p110β with mutant p110β inhibits PI3K but not MAPK signaling pathways in breast cancer cells. A, representative immunoblots showing expression of p110β and β-actin in MDA-MB-231 (top) and BT-549 (bottom) cells in which endogenous p110β was stably knocked down, followed by stable transfection with wild type p110β, p110βKK-DD, or p110βK-R. B, representative immunoblots showing pThr308-AKT and total AKT levels in stable MDA-MB-231 cell lines stimulated with LPA (top) or EGF (bottom). C, representative immunoblots showing pThr202/Tyr204-ERK1/2 and total ERK levels in stable MDA-MB-231 cell lines stimulated with LPA (top) or EGF (bottom). D, representative immunoblots showing pThr308-AKT/total AKT and pThr202/Tyr204-ERK1/2/total ERK levels in stable BT-549 cells stimulated with LPA (top) or EGF (bottom). Data are representative of at least two independent experiments for each condition.
Stable replacement of endogenous p110β with mutant p110β inhibits PI3K but not MAPK signaling pathways in breast cancer cells. A, representative immunoblots showing expression of p110β and β-actin in MDA-MB-231 (top) and BT-549 (bottom) cells in which endogenous p110β was stably knocked down, followed by stable transfection with wild type p110β, p110βKK-DD, or p110βK-R. B, representative immunoblots showing pThr308-AKT and total AKT levels in stable MDA-MB-231 cell lines stimulated with LPA (top) or EGF (bottom). C, representative immunoblots showing pThr202/Tyr204-ERK1/2 and total ERK levels in stable MDA-MB-231 cell lines stimulated with LPA (top) or EGF (bottom). D, representative immunoblots showing pThr308-AKT/total AKT and pThr202/Tyr204-ERK1/2/total ERK levels in stable BT-549 cells stimulated with LPA (top) or EGF (bottom). Data are representative of at least two independent experiments for each condition.
We previously showed that the p110βKK-DD mutant is defective for Gβγ binding, but exhibits comparable interactions with tyrosine phosphopeptides, Rab5 (23) and Rac1 (data not shown) as wild type p110β. To validate that the mutations in the p110β Gβγ-binding site disrupt p110β signaling in breast cancer cells, we measured AKT phosphorylation following stimulation with LPA, a GPCR ligand, or EGF, an RTK ligand (Fig. 1B–D, with quantitation in Supplementary Fig. S2). LPA-stimulated MDA-MB-231 and BT-549 cell lines expressing p110βKK-DD or p110βK-R all showed markedly reduced phosphorylation of AKT as compared with cells expressing wild-type p110β (Fig. 1B and D; top). However, when stimulated with EGF, cells expressing mutant or wild-type p110β showed similar levels of AKT phosphorylation (Fig. 1 B and D; bottom).
To analyze effects on the MAPK pathway, we measured ERK1/2 phosphorylation following stimulation with LPA or EGF. MDA-MB-231 cells expressing wild-type p110β showed high levels of basal ERK1/2 phosphorylation, which was not further stimulated by LPA (Fig. 1C; top) or EGF (Fig. 1C; bottom). This is consistent with the expression of activated forms of KRAS and BRAF in this cell line (35). Interestingly, quiescent MDA-MB-231 cells expressing p110βKK-DD or p110βK-R showed lower levels of ERK1/2 phosphorylation; the reason for this decrease is not clear. However, LPA and EGF-stimulated ERK1/2 phosphorylation in MDA-MB-231 cells expressing p110βKK-DD or p110βK-R was similar to that in cells expressing wild type p110β (Fig. 1B). BT-549 cells expressing wild-type or mutant p110β showed similar levels of basal as well as LPA-stimulated and EGF-stimulated phosphorylation of ERK1/2 (Fig. 1D; top and bottom, respectively), indicating that the MAPK signaling pathway is intact in these cell lines.
To investigate the role of p110β in breast cancer cell chemotaxis, we assayed the migration of the MDA-MB-231 cell lines towards LPA in a Transwell assay (Fig. 2). p110β knockdown cells showed decreased cell migration in response to LPA. This was rescued by expression of wild-type p110β, but not p110βKK-DD or p110βK-R; migration of cells expressing either mutant was similar to that seen with p110β knockdown cells. We also measured invasion in response to SDF-1 and IL8, which have been implicated in paracrine signaling between breast tumor cells and macrophages (36, 37). Migration to SDF-1 was almost completely suppressed in cells expressing mutant p110β (Fig. 2B). IL8 caused a trend toward increased migration in cells expressing wild-type p110β and p110βK-R, but not in cells expressing p110βKK-DD (Fig. 2C). As expected, migration toward the RTK ligand EGF was unaffected by expression of mutant p110β (Fig. 2D). Taken together, these data demonstrate that expression of p110βKK-DD or p110βK-R in breast cancer cells specifically inhibits GPCR regulation of AKT signaling and chemotaxis.
Gβγ binding to p110β is required for MDA-MB-231 migration in response to GPCR ligands but not EGF. A, LPA-stimulated Transwell migration of MDA-MB-231 p110β knockdown cells without or with stable rescue by wild-type or mutant p110β. Results are expressed as the fold change in the number of migrated cells, relative to values obtained with LPA-stimulated parental cells. B–D, Transwell Matrigel invasion assay of stable MDA-MB-231 cell lines toward SDF-1 (B), IL8 (C), or EGF (D). Results are expressed as the fold change in the number of invaded cells, relative to values obtained with ligand-stimulated cells expressing wild-type p110β. Data, the mean ± SEM from three independent experiments. In C, the P value for WT versus WT with IL8 is 0.135 and that for WT IL8 versus p110βKK-DD IL8 is 0.156; *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Gβγ binding to p110β is required for MDA-MB-231 migration in response to GPCR ligands but not EGF. A, LPA-stimulated Transwell migration of MDA-MB-231 p110β knockdown cells without or with stable rescue by wild-type or mutant p110β. Results are expressed as the fold change in the number of migrated cells, relative to values obtained with LPA-stimulated parental cells. B–D, Transwell Matrigel invasion assay of stable MDA-MB-231 cell lines toward SDF-1 (B), IL8 (C), or EGF (D). Results are expressed as the fold change in the number of invaded cells, relative to values obtained with ligand-stimulated cells expressing wild-type p110β. Data, the mean ± SEM from three independent experiments. In C, the P value for WT versus WT with IL8 is 0.135 and that for WT IL8 versus p110βKK-DD IL8 is 0.156; *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Gβγ signaling to p110β is required for macrophage-stimulated breast cancer cell invasion
Macrophages are obligate partners of breast cancer cells during invasion (38). We therefore assayed cell invasion using a modified Transwell invasion assay, in which tumor cell invasion through Matrigel is measured in the absence or presence of macrophages plated in the lower well, in the absence of any added chemoattractant. As compared with wells where macrophages were absent, the presence of macrophages in the lower well increased the invasion of MDA-MB-231 and BT-549 cells expressing wild-type p110β by 7.5- and 13-fold, respectively (Fig. 3A and B). However, macrophage-stimulated invasion was reduced by approximately 50% in cells expressing p110βKK-DD or p110βK-R (Fig. 3A and B). Macrophage-stimulated invasion was also inhibited by the p110β-specific kinase inhibitor TGX-221 (Fig. 3C and D). We next evaluated the MDA-MB-231 cell lines in an assay that directly measures macrophage-dependent invasion into a 3D collagen gel (33). For tumor cells expressing wild-type p110β, the addition of macrophages caused a 4-fold increase in the number of tumor cells that invaded the collagen gel (from 14% to 58%; Fig. 3E). However, macrophage-stimulated invasion was completely abolished in MDA-MB-231 cells expressing either p110βKK-DD or p110βK-R (Fig. 3E). These data demonstrate that GPCR signaling to PI3Kβ in tumor cells is critical for macrophage-dependent tumor cell invasion.
Gβγ binding to p110β is required for macrophage-stimulated breast cancer cell invasion. Macrophage-stimulated invasion of MDA-MB-231 (A) or BT-549 (B) cells is expressed as the fold change in the number of invaded cells, relative to values obtained with unstimulated cells expressing wild-type p110β. Macrophage-stimulated invasion of parental MDA-MB-231 (C) or BT-549 (D) cells without or with treatment with TGX-221 is expressed as the fold change in the number of invaded cells, relative to values in the absence of macrophages and TGX-221. E, macrophage-stimulated invasion of MDA-MB-231 cells in a 3D in vitro collagen invasion assay is expressed as the fraction of cells that invaded ≥20 μm. F, invasion of macrophages, coplated with MDA-MB-231 cells in the 3D invasion assay, is expressed as the number of cells that invaded ≥20 μm. Data, the mean ± SEM from three (A, B, C, E, and F) or two (D) independent experiments; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Gβγ binding to p110β is required for macrophage-stimulated breast cancer cell invasion. Macrophage-stimulated invasion of MDA-MB-231 (A) or BT-549 (B) cells is expressed as the fold change in the number of invaded cells, relative to values obtained with unstimulated cells expressing wild-type p110β. Macrophage-stimulated invasion of parental MDA-MB-231 (C) or BT-549 (D) cells without or with treatment with TGX-221 is expressed as the fold change in the number of invaded cells, relative to values in the absence of macrophages and TGX-221. E, macrophage-stimulated invasion of MDA-MB-231 cells in a 3D in vitro collagen invasion assay is expressed as the fraction of cells that invaded ≥20 μm. F, invasion of macrophages, coplated with MDA-MB-231 cells in the 3D invasion assay, is expressed as the number of cells that invaded ≥20 μm. Data, the mean ± SEM from three (A, B, C, E, and F) or two (D) independent experiments; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Interestingly, macrophage invasion into the collagen gel was also reduced when they were co-cultured with MDA-MB-231 cells expressing p110βKK-DD or p110βK-R as compared with wild-type p110β (Fig. 3F). To test whether this was due to a reduction in secreted factors by the tumor cells, we repeated the Boyden invasion experiment, but instead measured the ability of macrophages in the upper chamber to invade through a Matrigel layer in response to tumor cells in the lower chamber. We saw no differences in migration toward tumor cells expressing wild type versus mutant p110β (data not shown), suggesting that tumor cell signaling to macrophages was unaffected by mutation of p110β.
Gβγ signaling to p110β is required for breast cancer cell extravasation
To examine the contribution of GPCR signaling to p110β during tumor growth, we injected MDA-MB-231 cells expressing wild-type p110β, p110βKK-DD, or p110βK-R into the mammary fat pad of SCID mice. Xenograft tumors derived from all three cell lines showed similar growth rates (Fig. 4), which is consistent with in vitro assays in which the cells exhibit similar proliferation rates under complete and low serum conditions (Supplementary Fig. S3A and S3B).
Expression of GPCR-uncoupled or kinase-dead p110β in MDA-MB-231 cells has no effect on tumor growth. Growth of MDA-MB-231 tumors expressing wild-type or mutant p110β in SCID mice. Data, the mean ± SEM from 10, 13, and 12 mice for wild-type p110β, p110βKK-DD, or p110βK-R tumors, respectively.
Expression of GPCR-uncoupled or kinase-dead p110β in MDA-MB-231 cells has no effect on tumor growth. Growth of MDA-MB-231 tumors expressing wild-type or mutant p110β in SCID mice. Data, the mean ± SEM from 10, 13, and 12 mice for wild-type p110β, p110βKK-DD, or p110βK-R tumors, respectively.
To investigate extravasation in vivo, the MDA-MB-231 cell lines were injected into the lateral tail vein of SCID mice. There was a 50% reduction in the number of lung metastases in mice injected with p110βKK-DD-expressing cells as compared with mice injected with cells expressing wild-type p110β. In contrast, the number of lung metastases in mice injected with p110βK-R-expressing cells was actually greater than observed for mice injected with wild-type p110β cells (Fig. 5A). The difference in the number of lung metastases observed for p110βKK-DD-expressing cells was not due to alterations in cell proliferation, because the size of the metastatic lesions was similar for all three cell lines (Fig. 5B).
Gβγ binding to p110β is important for breast cancer cell extravasation in vitro and in vivo. A, the number of metastases per lung section from SCID mice injected with MDA-MB-231 cells expressing wild-type or mutant p110β into the lateral tail vein. B, average size of lung metastases. Data, mean ± SEM from 4, 6, and 5 mice for wild-type p110β, p110βKK-DD, or p110βK-R, respectively. Macrophage-stimulated transmigration of MDA-MB-231 (C) or BT-549 (D) cells expressing wild-type or mutant p110β in the extravasation-transendothelial migration assay is expressed as the fold change in the number of transmigrated cells relative to values obtained from cells expressing wild-type p110β. Data, the mean ± SEM from three independent experiments; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns: not significant.
Gβγ binding to p110β is important for breast cancer cell extravasation in vitro and in vivo. A, the number of metastases per lung section from SCID mice injected with MDA-MB-231 cells expressing wild-type or mutant p110β into the lateral tail vein. B, average size of lung metastases. Data, mean ± SEM from 4, 6, and 5 mice for wild-type p110β, p110βKK-DD, or p110βK-R, respectively. Macrophage-stimulated transmigration of MDA-MB-231 (C) or BT-549 (D) cells expressing wild-type or mutant p110β in the extravasation-transendothelial migration assay is expressed as the fold change in the number of transmigrated cells relative to values obtained from cells expressing wild-type p110β. Data, the mean ± SEM from three independent experiments; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns: not significant.
At metastatic sites, a discrete population of macrophages has been shown to promote tumor cell extravasation and seeding (39). To directly measure tumor cell extravasation, we used an in vitro transendothelial migration assay in which macrophages plated on the lower face of the Transwell filter stimulate tumor cell invasion through an endothelial cell layer plated on the upper face of the filter (40). Consistent with the in vivo metastasis assay, MDA-MB-231 and BT-549 cells expressing p110βKK-DD showed 65% and 75% reductions, respectively, in the number of transmigrated cells as compared with cells expressing wild-type p110β (Fig. 5C and D). BT-549 cells expressing p110βK-R showed a similar reduction in transmigration (Fig. 5D), whereas MDA-MB-231 cells expressing p110βK-R showed only a 35% reduction (Fig. 5C). Taken together, these data demonstrate a requirement for GPCR signaling to PI3Kβ in breast tumor extravasation.
Gβγ signaling to p110β is required for invadopodia-mediated gelatin degradation in MDA-MB-231 cells
The ability of tumor cells to degrade the extracellular matrix (ECM) is crucial for invasion and metastasis. To degrade the ECM, tumor cells form specialized structures called invadopodia, which are actin-rich protrusions that allow for the local secretion of matrix metalloproteinases and matrix degradation (41). To investigate the role of GPCR signaling to PI3Kβ in the formation and activity of invadopodia, MDA-MB-231 cells expressing wild-type or mutant p110β were plated on fluorescent gelatin and assayed for matrix degradation. Degradative invadopodia were identified by the appearance of black foci in the fluorescent gelatin that co-localized with the invadopodial markers F-actin and cortactin (Fig. 6A; ref. 41). Whereas approximately 90% of cells expressing wild-type p110β degraded the gelatin, the expression of p110βKK-DD or p110βK-R significantly reduced matrix degradation by 40% and 60%, respectively (Fig. 6B). Moreover, the total degradative area was markedly suppressed in cells expressing p110βKK-DD and p110βK-R as compared with cells expressing wild-type p110β (Fig. 6C).
Gβγ binding to p110β is required for gelatin degradation. A, representative micrographs of MDA-MB-231 cells expressing wild-type or mutant p110β, plated on Oregon Green 488-conjugated gelatin. Cells were stained with rhodamine phalloidin and cortactin to identify invadopodia. B, the percentage of cells with degradative invadopodia. C, the average area of gelatin degradation per cell. Values were normalized to those obtained from cells expressing wild type p110β. Data, mean ± SEM from three independent experiments, with a total of 146, 138, and 120 cells for wild-type p110β, p110βKK-DD, or p110βK-R, respectively; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.
Gβγ binding to p110β is required for gelatin degradation. A, representative micrographs of MDA-MB-231 cells expressing wild-type or mutant p110β, plated on Oregon Green 488-conjugated gelatin. Cells were stained with rhodamine phalloidin and cortactin to identify invadopodia. B, the percentage of cells with degradative invadopodia. C, the average area of gelatin degradation per cell. Values were normalized to those obtained from cells expressing wild type p110β. Data, mean ± SEM from three independent experiments, with a total of 146, 138, and 120 cells for wild-type p110β, p110βKK-DD, or p110βK-R, respectively; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.
ECM degradation is a component of all the macrophage-stimulated metastasis assays used in this study. We therefore considered whether the reduction in matrix degradation in cells expressing mutant p110β could account for the observed decreases in invasion and extravasation, independent of the effects on GPCR-mediated paracrine signaling from macrophages to tumor cells. We used a simplified chemotaxis assay, in which macrophages promote the migration of tumor cells through a Transwell filter in the absence of Matrigel or other ECM. We observed that macrophages stimulated the migration of tumor cells expressing wild-type PI3Kβ. However, macrophage-mediated tumor cell migration was markedly reduced in cells expressing mutant PI3Kβ (Supplementary Fig. S4). These data suggest that the loss of GPCR-mediated paracrine signaling from macrophages to tumor cells contributes to the defects in metastasis observed cells expressing p110βKK-DD.
To test the role of GPCR signaling to p110β in the context of mutant PIK3CA, we knocked down p110β in T47D cells (Supplementary Fig. S5A), which express the oncogenic H1047R mutant of p110α (35). Although we were not able to restore p110β expression to the levels observed in parental cells (Supplementary Fig. S5A and S5B), cells expressing wild-type p110β showed enhanced macrophage-stimulated invasion as compared with cells expressing p110βKK-DD and p110βK-R (Supplementary Fig. S5C).
Discussion
A large number of reports implicate GPCR signaling in cancer initiation, progression, and metastasis (42). Specifically, signaling through the Gβγ subunit has been shown to play a critical role in breast cancer tumorigenesis and metastasis (43, 44). The PI3K signaling pathway, which is one the major pathways activated downstream of GPCRs, is also one the most commonly disrupted pathways in cancer (45, 46). Two Class I PI3K isoforms are activated downstream of GPCRs, p110β, and p110γ. However, defining the role of GPCR signaling to p110β has been challenging because it is also activated by RTKs. Our recent identification a p110β mutant that is deficient for Gβγ binding and specifically abrogates its activation by GPCRs, but not RTKs (23) has allowed us to specifically test the role of Gβγ binding to the p110β PI3K in breast cancer invasion and metastasis.
The ability of cancer cells to disseminate depends on signals from cells in the tumor microenvironment, including tumor-associated macrophages, tumor-associated neutrophils, myeloid-derived suppressor cells, regulatory T cells, and cancer-associated fibroblasts (47). A paracrine signaling loop involving secretion of EGF and CSF-1 from breast cancer cells and tumor-associated macrophages, respectively, has been shown to be crucial for breast cancer cell invasion and intravasation (33, 48). However, tumor infiltrating cells secrete a variety of chemokines and other GPCR ligands that signal to tumor cells and promote tumor progression and metastasis (49). Consequently, we reasoned that Gβγ-mediated activation of p110β in breast cancer cells might play a role in breast cancer metastasis by transducing prometastatic signals from GPCR ligands produced by the tumor microenvironment.
In this study, we established MDA-MB-231 (wild-type PTEN and PIK3CA), and BT-549 (mutant PTEN) stable cell lines in which we knocked down endogenous p110β and replaced it with exogenous wild-type or mutant p110β at levels comparable with the parental cell lines. The knockdown-replace approach is critical for studying the roles of specific PI3K isoforms in cellular processes, because all p110α, p110β, and p110δ compete for a common pool of p85 regulatory subunits (50). As a consequence deletion of a single p110 isoform can lead to alterations in signaling by the remaining isoforms. For example, deletion of p110β leads to an increase in RTK-stimulated phosphorylation of AKT through p110α (51). Using these cells lines, we showed that the Gβγ–p110β interaction is required for LPA, but not EGF-stimulated AKT phosphorylation. We also showed that Gβγ-mediated activation of p110β is required for migration of MDA-MB-231 cells toward LPA (52) in a Transwell migration assay. Because the expression of mutant p110β only reduced cell migration by approximately 50%, it is likely p110γ, which is also expressed in these cells (53), contributes to GPCR-mediated migration.
In accordance with previous reports (33), we found that macrophages greatly increased the 3D invasion of breast cancer cells into Matrigel and collagen matrices. Remarkably, cells expressing Gβγ-uncoupled or kinase-dead p110β showed a significant reduction in macrophage-stimulated invasion in the Transwell assay and were completely unresponsive to the presence of macrophages in the 3D invasion assay. Interestingly, the invasion of macrophages themselves was also significantly reduced when they were co-cultured with breast tumor cells expressing mutant p110β in the 3D collagen invasion assay. This suggests that the ability of tumor cells to stimulate macrophage invasion also depends on the Gβγ-p110β interaction within tumor cells. Notably, tumor cells expressing wild-type or mutant p110β stimulated macrophage chemotaxis to similar extents, suggesting that signaling from tumor cells to macrophages is not affected by disruption of the Gβγ–p110β interaction in tumor cells. However, recent studies have suggested that macrophage-stimulated tumor cell invasion is mediated, at least in part, by direct contact between the two cell types (54). Therefore, it is possible that contact-dependent signaling between tumor cells and macrophages requires GPCR signaling through PI3Kβ.
We also achieved partial rescue of p110β knockdown in T47D cells (PIK3CA mutant). Consistent with the data in MDA-MB-231 and BT549 cells, expression of mutant p110β inhibited macrophage-stimulated invasion.
p110β has been implicated in tumor growth in certain genetic backgrounds (e.g., PTEN loss or ERBB2 overexpression; refs. 23, 28, 55). However, we observed that expression of mutant p110β had no effect on orthotopic MDA-MB-231 tumor growth in SCID mice. In contrast, GPCR signaling to p110β was critical for extravasation in both in vivo and in vitro assays.
Mutation of p110β had pronounced effects on matrix degradation, a process that is critical for all of the metastatic steps measured in this study. The reduction in degradative invadopodia occurred independently of macrophages, and thus represents an intrinsic loss of function in tumor cells expressing mutant p110β. The molecular mechanisms by which Gβγ–p110β interactions promote the formation of these degradative structures will require further investigation. Although alterations in invadopodia assembly or maturation likely contribute to defects in invasion and extravasation observed for p110βKK-DD-expressing tumor cells, our data also demonstrated a loss of paracrine signaling between macrophages and p110βKK-DD-tumor cells. It is therefore likely that both the loss of ECM degrading capacity and the loss of responsiveness to GPCR signaling from macrophages contributed to the observed defects in metastatic behavior.
p110β has been shown to have kinase-independent functions (55, 56), and may act as a scaffold to integrate signaling by its binding partners Gβγ, Rac, Rab5, and other proteins (23, 57, 58). A similar scaffolding function has been demonstrated for the other Gβγ-regulated PI3K, p110γ (59). Disruption of Gβγ-mediated targeting of PI3Kβ to cell membranes may inhibit both kinase-dependent and kinase-independent functions of p110β. It is therefore plausible that loss of Gβγ binding might have more severe effects on some cell behaviors than loss of kinase activity alone, which could account for the less pronounced effect of kinase-dead p110β in the transendothelial migration assay. However, it is surprising that expression of kinase-dead p110β inhibited transendothelial migration in vitro but slightly enhanced extravasation in SCID mice. In vivo, tumor cells are influenced by multiple stromal cell types (1), and are subjected to a more complex range of GPCR-dependent stimuli than occurs in vitro. Tumor cell responses to some of these stimuli might require p110β scaffolding functions that are intact in cells expressing kinase-dead p110β. Thus, a p110β kinase-dependent pathway that is rate limiting for transendothelial migration in vitro may be supplanted by a different pathway in vivo, one that does not require p110β activity but still requires Gβγ-mediated targeting of p110β. Alternatively, it is possible that loss of p110β kinase activity disrupts a negative feedback loop in vivo that is independent of Gβγ binding.
The finding that disruption of p110β binding to Gβγ has a more deleterious effect on experimental tumor metastasis than disruption of kinase activity suggests that targeting the p110β–Gβγ-binding interface might provide an effective alternative to small molecule inhibitors that compete for the ATP-binding site of p110β. Disruption of p110β-Gβγ binding could constitute a novel therapeutic pharmacologic approach to the treatment of metastasis in breast cancer patients.
Disclosure of Potential Conflicts of Interest
J.S. Condeelis has ownership interest (including patents) in Metastat and is a consultant/advisory board member for Metastat and Deciphera. J.M. Backer is a consultant/advisory board member for Karus Therapeutics. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: B.D. Khalil, A.R. Bresnick, J.M. Backer
Development of methodology: C. Hsueh, Y. Wang, J.S. Condeelis
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.D. Khalil, C. Hsueh, Y. Cao, W.F. Abi Saab, J.S. Condeelis, A.R. Bresnick, J.M. Backer
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.D. Khalil, W.F. Abi Saab, J.S. Condeelis, A.R. Bresnick, J.M. Backer
Writing, review, and/or revision of the manuscript: B.D. Khalil, A.R. Bresnick, J.M. Backer
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Hsueh, Y. Cao
Study supervision: A.R. Bresnick, J.M. Backer
Other (set up the MDA-MB-231 xenografts): Y. Wang, J.S. Condeelis
Acknowledgments
The authors thank Drs. Jeffrey Segall and David Entenberg for helpful discussions.
Grant Support
This work was supported by NIH grants 1P01 CA100324 (J.M. Backer, A.R. Bresnick, J.S. Condeelis, and B.D. Khalil), GM112524 (J.M. Backer and A.R. Bresnick), the Intravital Imaging, Cell Isolation and Fate Mapping Core of the Program Project in Motility and Invasion (1 P01 CA100324), and the Histotechnology and Comparative Pathology Core of the Einstein Cancer Center (P30-CA013330).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.