Inhibiting G protein βγ signaling blocks prostate cancer progression and enhances the efficacy of paclitaxel

Aberrant activation of G protein-coupled receptors (GPCRs) is implicated in prostate cancer progression, but targeting them has been challenging because multiple GPCRs are involved in cancer progression. In this study, we tested the effect of blocking signaling via a hub through which multiple GPCRs converge — the G-protein Gβγ subunits. Inhibiting Gβγ signaling in several castration-resistant prostate cancer cell lines (i.e. PC3, DU145 and 22Rv1), impaired cell growth and migration in vitro, and halted tumor growth and metastasis in nude mice. The blockade of Gβγ signaling also diminished prostate cancer stem cell-like activities, by reducing tumorsphere formation in vitro and tumor formation in a limiting dilution assay in nude mice. Furthermore, Gβγ blockade enhanced the sensitivity of prostate cancer cells to paclitaxel treatment, both in vitro and in vivo. Together, our results identify a novel function of Gβγ in regulating prostate cancer stem-cell-like activities, and demonstrate that targeting Gβγ signaling is an effective approach in blocking prostate cancer progression and augmenting response to chemotherapy.


INTRODUCTION
Prostate cancer is the most common cancer affecting men in the Western World and the second leading cause of cancer death among American men [1]. The five-year survival rate for patients with localized prostate cancer is nearly 100%, but it decreases to 28% for patients in advanced stages with metastases [2]. Androgendeprivation therapy (ADT) is the standard of care for advanced cases; and although initially effective, most patients with metastatic tumors eventually relapse with castration-resistant prostate cancer (CRPC) [3].
Many mechanisms foster CRPCs. For example, androgen-receptor (AR) signaling can persist, despite low levels of circulating androgen; alternatively, any one of many signaling pathways controlled by receptor tyrosine kinases and G protein-coupled receptors (GPCRs) can become deregulated. Such changes can allow prostate cancer growth to become androgen independent [4][5][6][7][8]. Emerging evidence, however, suggests another source of androgen independent CRPCs may be a small subpopulation of cells that retain stem-like properties [9][10][11]. These stem-like cells can propagate tumors, have a survival advantage, and escape current chemotherapies, suggesting their stem-like qualities allow them to persist, become drug-resistant, proliferate, and metastasize. These cells have been termed cancer stem cells (CSCs).
CSCs are known to overexpress GPCRs (e.g., CXCR4), which when stimulated in prostate cancer cells trigger growth, migration, and invasion [12]. Activated GPCRs might exert their tumorigenic effects directly by stimulating heterotrimeric G-protein-dependent signaling,

Research Paper
or indirectly by trans-activating androgen and growthfactor receptors [13]. Thus, overexpressed, activated GPCRs could signal to CSCs to become tumorigenic and chemotherapy resistant [12].
Downstream of most GPCRs, signals are transmited through heterotrimeric G proteins, which consist of Gα, Gβ, and Gγ subunits [14]. Gα subunits in their inactive state heterodimerize with Gβ and Gγ subunits [14]. GPCRs activate G proteins by inducing GTP/GDP exchange on Gα subunits, leading to the dissociation of Gα from Gβγ subunits. Both GTP-bound Gα and liberated Gβγ subunits transmit signals to downstream effectors and both are implicated in tumor progression, but accumulating evidence indicates that Gβγ plays a particularly important role in tumor growth and metastasis, therefore representing an attractive therapeutic target [31].
Gβγ subunits, rather than Gαi/o subunits, transmit the primary proliferation signals for a large group of GPCRs overexpressed in prostate tumor cells (such as receptors for chemokines, bradykinin and LPA) [13,32]. Moreover, in prostate cancer cells, liberated Gβγ likely transactivates EGF receptors [33,34]. Inhibiting Gβγ has been shown to block prostate cancer PC3 cell growth in vitro and primary tumor growth in nude mice [35,36]. Gβγ also serves as a point of convergence for signals from multiple GPCRs in breast cancer cells, where Gβγ promotes tumor growth and metastasis [37,38], but whether it plays a similar role in increasing prostate cancer metastasis is unknown. Moreover, it has never been investigated whether Gβγ signaling mediates GPCR activity in increasing prostate cancer CSC tumorigenicity and sensitivity to chemotherapy.
In this study, we showed that inhibiting Gβγ signaling in several castration-resistant prostate cancer cell lines not only blocked progression of preexisting primary prostate tumors but also suppressed formation of tumor metastases in bone and soft tissues. Moreover, we provide evidence that, both in vitro and in vivo, Gβγ signaling may be involved in maintaining the population and tumorigenicity of prostate CSCs. Furthermore, we showed Gβγ blockade sensitized prostate cancer cells to the chemotherapeutic agent, paclitaxel. Our data thus reveal that prostate cancer CSC tumorigenicity is driven by a novel function of Gβγ signaling, and that targeting Gβγ signaling may be a new way to eliminate these cells to block tumor progression.

Expression of Gαt selectively blocks GPCR signaling in prostate cancer cells
In our study of prostate cancer cells, Gβγ signaling was manipulated by expressing recombinant Gαt, a specific inhibitor of Gβγ that binds free Gβγ and selectively prevents the activation of Gβγ effectors without interfering with Gα signaling [37,39]. Here, recombinanat lentiviruses stably transduced castration-resistant prostate cancer cell lines (i.e., PC3, DU145, 22RV1) and an nontransformed prostate epithelial cell line (RWPE1) with tetracycline-inducible expression vectors encoding GFP (control) or Gαt ( Figure 1A). To assess how Gαt expression affected Gβγ signaling, we tracked AKT and ERK phosphorylation, in the presence of several GPCR agonists. These GPCR agonists included LPA, SDF1α and a protease-activated receptor 1 (PAR1) peptide agonist, all of which can activate AKT and ERK via Gβγ-dependent mechanisms [37]. As control, cells were stimulated in parallel with IGF or EGF, which activate receptor tyrosine kinases. As shown in Figure 1B-1F, in all cell lines tested, stimulation of GPCRs activated AKT and/or ERK phosphorylation, but the levels of response varied widely in different cell types and also dependent on the types of ligands used. Gαt expression inhibited GPCR-stimulated AKT and ERK phosphorylation in all cancer cell lines tested, but only suppressed GPCR-stimulated ERK but not AKT phosphorylation in RWPE1 cells. IGF and EGF signalling for activated AKT and ERK was unaffected by Gαt expression. Similar results were observed with a selective inhibitor of Gβγ, gallein, in RWPE1 (not shown) and PC3 cells ( Figure 1C), suggesting that GPCR stimulated AKT and ERK phosphorylation through a Gβγdependent pathway in prostate cancer cell lines, and both Gβγ-independent and dependent pathways in RWPE1 cells. In PC3 cells, LPA-, SDF1α and PAR1-stimulated AKT and ERK phosphorylation was also sensitive to pertussis toxin ( Figure 1C). Since pertussis toxin exclusively uncouples Gi/o proteins from their receptors by catalyzing the ADP-ribosylation of the Gai/o subunits [40], these findings suggest that these GPCRs activate AKT and ERK in cancer cell lines primarily through Gβγ freed from Gi/o proteins. www.impactjournals.com/oncotarget

Gβγ signaling promotes prostate cancer cell growth, proliferation and migration
We then asked how blocking Gβγ signaling affects cell growth. Compared to GFP expression, Gαt expression significantly decreased PC3, DU145 and 22Rv1 cell growth in monolayer culture in a XTT-based cell viability assay and a 5-bromo-2′-deoxyuridine (Brdu) incorporation assay (Figure 2A-2C). The slower proliferation of prostate cancer cells expressing Gαt was also evident by the 50 to 80% reduction in the size of colonies in a 3D-matrigel culture ( Figure 2D, 2E). Treatment of prostate cancer cells with PTx reduced proliferation of GFP-expressing cells to a level comparable to that of Gαt-expressing cells (Figure 2A), indicating the primary involvement of Gβγ subunits released from Gi/o proteins. Simuilar to Gαt expression, gallein treatment also reduced PC3 cell growth ( Figure 2A). Interestingly, the growth of RWPE1 cells was neither affected by PTx treatment nor Gαt expression ( Figure 2A). These findings suggest that Gβγ signaling is critical for the proliferation of prostate cancer cells but is dispensable for the growth of normal, prostate epithelial cells.

Blocked Gβγ signaling impairs prostate tumor growth and metastasis in vivo
To examine whether blocking Gβγ signaling inhibits prostate cancer cell growth in vivo, 22Rv1 cells expressing inducible GFP or Gαt were injected, in equal numbers, into the prostate gland of nude mice. Without inducing transgene expression, bioluminescence imaging showed Gαt-expressing cells prolifereated at the same rate or slightly faster than GFP-expressing cells ( Figure 4A, 4B). After 21d, when transgene expression was induced with doxycycline, significant reduction in tumor growth was seen in mice injected with Gαt-expressing cells, as compared to GFP-expressing cells ( Figure 4A, 4B). These findings indicate that Gβγ signaling is critical for maintaining 22Rv1 orthotoptic xenograft growth.
To test if Gβγ signaling drives prostate cancer metastasis, we injected 22Rv1 cells expressing inducible GFP or Gαt into the left ventricle of nude mice, to disseminate tumor cells to multiple organs. Injected cells were allowed to form tumors in the absence of doxycycline induction for 21 days. Over this period, BLI revealed all injected cells grew at comprabe rates, throughout the animals' bodies ( Figure 5A-5C). Upon inducing GFP or Gαt expression, whole-body BLI analysis suggested Gαt-expressing cells proliferated more slowly, but the difference was not statistically significant ( Figure 5B). Ex vivo BLI, however, revealed that Gαt-expressing cells gave rise to fewer tumors, in multiple organs (i.e., brain, lung, kidney, leg and mandible; Table 1). Moreover, mice bearing Gαt-expressing cells were significantly improved in overall survival ( Figure 5C). Similar results were found for PC3 cells ( Figure 5D-5E and Table 2). These findings indicate that Gβγ signaling is also critical for the outgrowth of prostate cancer metastases in multiple organs.

Blocked Gβγ signaling targets aggressive, stemlike cells in prostate tumors
Prostate cancer cells harbor a small population of CSCs that may contribute to metastasis and recurrence [9]. Given that prostate cancer cell growth and metastasis was robustly inhibited by Gβγ blockade, we tested whether Gβγ signaling regulates the activities of their CSCs. Prostate cancer CSCs can be identified in vitro by their ability to grow primary and secondary tumorspheres upon serial passaging under non-attached conditions in ultralow-adhesive plates; and in vivo by their ability to generate tumors after serial transplantation into mice [41,42]. As reported [41,43], PC3 and DU145 cells generate increasing number of tumorspheres upon serial passaging ( Figure 6A, 6B). Inhibiting Gβγ signaling by inducing Gαt expression or gallein treatment decreased the number and size of tumorspheres generated from PC3 and DU145 cells ( Figure 6A, 6B). Similar results were found for 22Rv1 cells (data not shown).
To verify these findings in vivo, we cultured GFP-expressing and Gαt-expressing PC3 cells, without doxycycline induction, under tumorsphere-forming conditions for three passages to enrich for CSCs. Then GFP-expressing and Gαt-expressing tumorspheres were dispersed into a single-cell suspension and injected subcutaneously, (50,000, 20,000, or 10,000 per mouse) into nude mice. Mice were immediately fed doxycyclinecontaining diets to induce transgene expression. In parallel, we also treated another group of mice bearing GFP-expressing cells with the Gβγ inhibitor, gallein.
Two months post injection, all mice that had been inoculated with 50,000 GFP-expressing or Gαtexpressing cells formed palpable tumors (Table 3), but the size of tumors derived from Gαt-expressing cells or gallein-treated mice tended to be smaller (data not shown). Compared to mice inoculated with a comparable number of GFP-expressing cells, mice inoculated with 20,000 and 10,000 of Gαt-expressing cells developed fewer palpable tumors (60 and 80% vs 40 and 0%, respectively; Table 3). The frequency of tumor formation in mice inoculated with 10,000 GFP-expressing cells also decreased from 80% to 20% with gallein treatment www.impactjournals.com/oncotarget (Table 3). Immunohistochemical analysis of Ki67 expression and caspase 3 cleavage activation revealed that blocking Gβγ signaling by Gαt expression or gallein treatment significantly reduced prostate cancer cell proliferation and increased apoptosis in the xenograft tumors ( Figure 7A, 7B). These findings suggest Gβγ signaling may be required for maintaining and renewing the population of prostate CSCs.

Blocking Gβγ signaling enhances the therapeutic efficacy of paclitaxel
CSCs contribute to prostate cancer resistance to chemotherapy [10,43]. Since targeting Gβγ signaling inhibits the self-renewal activity and tumorigenicity of prostate cancer CSCs, we tested whether blocking Gβγ signaling sensitizes prostate cancer cells to the chemotherapeutic agent, paclitaxel. We first evaluated the effect of mono-or combinational therapy on the tumorsphere-forming capacity of PC3 and DU145 cells in vitro. (To quantify tumorsphere formation, tumorspheres were cultured on Matrigel rather than ultralow-adhesive plates, because dead cells aggregate under the latter condition, obscuring tumorsphere size and number.) Paclitaxel alone inhibited PC3 and DU145 tumorsphere formation in a dose-dependent manner ( Figure 9A-9B); and the inhibitory effect was significantly enhanced by simultaneously blocking Gβγ signaling ( Figure 9A-9B).
The effect of combining Gβγ blockade with paclitaxel was greater than either treatment alone ( Figure 9A-9B), suggesting synergistic inhibition.
To translate our findings to an in vivo model, PC3 xenograft tumors were established from GFP-or Gαtexpressing cells isolated from tumorspheres grown on ultralow-adhesive plates, in the absence of doxycycline. When the tumors reached a size of ~300 mm 3 , the mice were fed doxycycline-containing diets, to induce GFP or Gαt expression. Silmutaneously, mice were given a sub-maximal dose of vehicle control or paclitaxel. As expected, paclitaxel treatment or Gαt expression alone inhibited tumor growth by ~1.6 fold. When combined with Gαt expression, paclitaxel decreased the rate of tumor growth by ~4-fold ( Figure 9C), suggesting a possible synergistic effect.

DISCUSSION
A growing body of evidence shows GPCRs play a critical role in prostate cancer cell proliferation and dissemination [46][47][48]. Our studies demonstrate that GPCR signals that drive prostate cancer growth and metastasis converge at one signaling hub-the Gβγ subunits. These findings are consistent with previous reports that in PC3 cells, inhibiting Gβγ signaling by overexpressing the Gβγ scavenger, the C-terminal tail of G protein-coupled receptor kinase 2, alleviated proliferation in vitro and subcutaneous tumor formation, in nude mice [35]. Moreover, we provide the first evidence that Gβγ signaling is a central regulator of the self-renewal activity and tumorigenicity of prostate CSCs. Importantly, in our studies, targeting Gβγ signaling inhibited the tumorigenicity of prostate CSCs and sensitized tumors to chemotherapy.
Our studies showed that blocking Gβγ signaling does not affect the growth of the non-transformed prostate epithelial cell line, RPWE1. Unlike in cancer cell lines, in RWPE1 cells, GPCRs stimulate AKT activation through a Gβγ-independent pathway. This suggests that Gβγ signaling may be dispensable for normal cell growth. Thus, GPCRs may drive normal and cancer cell growth through different mechanisms, and targeting Gβγ signaling may be an effective strategy to selectively target tumor cells, while sparing normal cell function. Interestingly, the effect of Gβγ inhibition on cell growth can be mimicked by a specific Gβγ inhibitor, gallein, or pertussis toxin treatment, indicating that the Gβγ is predominantly liberated from activated Gi/o proteins that couple to many GPCRs overexpressed in prostate cancer cells, such as CXCR4 and LPA receptors [13,32,36].
Tumor metastasis is the major cause of patient mortality [2], and bone is the primary site of prostate cancer metastasis [49]. Clinically, the bones that are most commonly involved are vertebrae, sternum, pelvic bones, ribs, and femurs [50]. In the intracardiac injection mouse model of prostate cancer metastasis, mandible, tibia and femur are most commonly affected [51]. Interestingly, inhibiting Gβγ signaling largely blocks tumor formation in tibial and femurs but has little effet in the mandible, suggesting GPCR or Gβγ signaling affects bone metastasis differently at different metastasis sites.
Increasing evidence indicates that prostate cancer progresses from a local to a metastatic disease, in part because of a reservoir of CSCs [9][10][11]. These cells display the stem-cell-like ability to self-renew and then generate diverse cells that constitute a tumor. Gβγ signaling appears to be critical for maintaining prostate CSCs. Inhibiting Gβγ signaling reduces the number and size of tumorspheres grown from PC3, DU145 and 22Rv1 on serial passages in vitro. And in a limiting dilution, tumorformation assay, PC3 cells generated far fewer tumors under Gβγ blockade through Gαt expression in tumor cells or systemic administration of the specific inhibitor of Gβγ, gallein. One caveat to these results is that it is difficult to separate general effects that Gβγ blockade may have on prostate cancer cell growth from more specific effects on  The number and percentage of mice detected with tumors at the indicated tissues by ex vivo BLI are indicated. The number and percentage of mice detected with tumors at the indicated tissues by ex vivo BLI are indicated.  Immediately post inoculation of PC3 cells, mice were fed with doxycycline-containing diet to induce GFP or Gαt expression and treated with or without gallein (10 mg/kg, twice daily). Mice were monitored weekly for palpable tumors. Data are expressed as percentage of mice with palpable tumors two month-post innoculation.

Figure 7: Blocking Gβγ signaling slows proliferation and triggers apoptosis of prostate cancer cells. Tumor sections were
prepared from tumors derived from mice inoculated with 20,000 PC3 cells, as indicated in Table 3. Sections were stained with Ki67 (A) or cleaved caspase 3 (B). Representative images of the indicated tumor sections and quantitative data are shown. ***p < 0.001 vs GFP, n = 5.

Figure 8: Inhibition of Gβγ signaling decreases the CD133 + /CD44 + populations in PC3 cells and PC3 xenograft tumors.
Single cells were prepared from PC3 cells grown in 2D culture (A) or PC3 xenograft tumors derived from mice inoculated with 20, 000 cells/mouse, as indicated in Table 3  CSCs. However, Gβγ blockage decreases the frequency of CD133 + /CD44 + populations in PC3 cells in vitro and in tumor xenografts, supporting the idea that Gβγ blockade does affect CSCs. Taxanes, such as docetaxel and cabazitaxel, are first-line chemotherapeutic agents for metastatic CRPC patients [52,53]. After initial responses, patients almost always acquire resistance to taxane drugs, likely in part because of CSCs [10,43]. As compared to bulk tumor cells, CSCs are resistant to most chemotherapeutic agents so treatment with taxanes likely enriches for CSCs. This has been considered as a potential source of tumor relapse and resistance. Thus, targeting CSCs should inclued enhancing sensitivity to taxanes. Indeed, our data support this hypothesis. Inhibiting Gβγ signaling sensitizes PC3-derived tumorspheres to paclitaxel, both in vitro and in vivo, although further experiments would be necessary to determine if combining these agents have additive versus synergistic effects. These findings imply that aberrant GPCR signaling, through Gβγ, in CSCs may contribute to resistance to paclitaxel. How Gβγ activates the downstream signaling pathways to enhance CSC paclitaxel sensitivity remains to be determined. However, signaling molecules activated by Gβγ (e.g., PI3K/ AKT, MAPK, and STAT3) have been shown to regulate sensitivity to chemotherapeutic reagents, including taxanes [9,31,32,41,52,54,55]. Thus, Gβγ likely increases the sensitivity of prostate CSCs to taxanes and tumor progression through diverse signaling pathways.
In summary, we presented clear evidence that Gβγ is a convergence point for multiple GPCRs that promote prostate cancer growth and metastasis. Moreover, we showed that Gβγ signaling is critical for maintaining the tumorigenicity of reseident CSCs. Thus, targeting Gβγ may be a useful approach for eliminating prostate cancer CSCs, to halt tumor progression and augment sensitivity to chemotherapy. The tumorsphere size was quantified as in Figure 2. Data are expressed as percentage of the size of tumorspheres from GFP-expressing cells in the absence of paclitaxel treatment. *, **p < 0.05 and 0.01 vs GFP; ≠ p < 0.05 vs Gαt in the absence of paclitaxel. N = 3. (C) Mice (n = 6-7) were subcutaneously implanted with PC3 cells expressing inducible GFP or Gαt. Tumor growth was measured by caliper. When tumor size reached ~300 mm 3 , mice were fed doxycycline-containing diets and treated with vehicle or paclitaxel (10 mg/kg, i.p., twice per week). *, **p < 0.05 and 0.01 vs GFP; ≠ p < 0.05 vs Gαt or GFP+paclitaxel. www.impactjournals.com/oncotarget

Lentiviral production
The pSLIK lentiviral vectors for tetracycline inducible EGFP and Gαt expression were described previously [37,38]. Lentiviruses were generated by transfecting 293FT cells with the pSLIK vectors together with packaging vectors pMDL and pSPAX2 (Addgene).

Cell culture and establishment of stable cell lines
The human prostate cancer cell lines, PC3, DU145, and 22RV1, and the normal prostate epithelial cell line RWPE1, were obtained from ATCC. PC3 and DU145 cells were cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (FBS), while 22Rv1 cells were cultured in RPMI-1640 (Life Technologies) supplemented with 10% FBS. RWPE1 cells were cultured in keratinocyte serum-free medium (Life Technologies) supplemented with 0.05 mg/ml bovine pituitary extract and 5 ng/ml EGF. PC3, DU145 and 22Rv1 cells were transduced with lentivirus prepared from the pLenti PGK Neo vector to express firefly luciferase and selected with G418 (0.5 mg/ml).
To assess the ability of prostate cancer cells to form colonies in three-dimensional cultures, cells were suspended in complete growth medium, supplemented with 2% growth factor-reduced Matrigel (BD Biosciences) and cultured on top of a thin layer of Matrigel in 8-well chamber slides (1,000 cells/well) [37,38]. Doxycycline was added to cells the next day; and the medium was replaced every 3 days. On day eight of culture, cells were assessed via microscopy and multiple images were taken. Colony sizes were analyzed by ImageJ software.

BrdU incorporation assay
Three days after treatment with doxycycline to induce GFP or Gαt expression, prostate cancer cells, grown on coverslips in a 24-well plate, were serum starved overnight. A BrdU incorporation assay was then performed for 16 hours, as described [38].

Cell migration assay
PC3, DU145, and 22RV1 cells were induced to express GFP or Gαt for three days. After serum starvation overnight, cells were detached from plates using 2 mM EDTA/PBS. After washing with serum-free DMEM twice, cells were resuspended in serum-free DMEM and subjected to transwell migration assay for 4 h at 37°C, as described previously [37,38].

Xenograft mouse models
All animal studies were conducted in accordance with an Institutional Animal Care and Use Committeeapproved protocol at the University of Iowa. 6 to 8-weekold, male, nude mice were used for studies. PC3 or 22Rv1 cells expressing inducible GFP or Gαt were implanted into mice either by injection into the left ventricle (2 × 10 5 cells in 100 µl PBS) or the dorsolateral prostate lobe (2 × 10 5 cells in 20 µl PBS). 21 days post-injection, mice were fed a doxycycline-containing diet (TD.01306, Harlan laboratories, 625 mg/kg) continuously, to induce GFP and Gαt expression. Tumor progression was monitored weekly by BLI. Mice were euthanized when they lost more than 15% of their body weight. The formation of tumors in various tissues was determined by post-mortem ex-vivo BLI. The formation of bone tumors was confirmed by X-ray and histological analysis by H&E staining. www.impactjournals.com/oncotarget BLI Mice were anaesthetized and injected retroorbitally with 100 µl D-luciferin (15 µg/ml in PBS). BLI was performed using a Xenogen IVIS200 system. The bioluminescent data were analyzed by the software Living Image (Xenogen) and expressed as photon flux (photons/s/ cm 2 /sr) [37,38].

Tumorsphere culture
Single PC3, DU145, and 22RV1 cells, expressing inducible GFP or Gαt, were plated (1,000 cells/ml) on ultralow-attachment, 6-well or 10-cm plates, coated with poly(2-hydroxyethl methacrylate (5 mg/ml in 95% ethanol) in serum-free prostate epithelial cell basal medium (Lonza) supplemented with 4 μg/ml insulin, 20 ng/ ml EGF, 20 ng/ml FGF, and B27 (Life Technology) as described previously [41]. After 7 to 14 days of culture, the number of tumorspheres was quantified under an inverted microscope. To determine the size of tumorspheres, images were taken at 5-10 randomly chosen areas and analyzed by ImageJ software. Spheres were dissociated with Accutase (Invitrogen) to generate single cell suspension for subsequent expansion and flow cytometry analysis.
To quantify the sensitivity of GFP-or Gαt-expressing tumorspheres to paclitaxel, third passages were dispersed to single cells and cultured in Matrigel, as described above. Cells were then treated with different concentrations of paclitaxel together with doxycycline for 8 days.

Tumorigenicity of prostate cancer stem cells in vivo
PC3 cells expressing inducible GFP or Gαt were maintained in tumorsphere culture, in the absence of doxycycline for three generations. Single cells were prepared from the third generation of tumorspheres and various numbers (50,000, 20,000, and 10,000 cells) were subcutaneously injected into the upper and lower flanks of nude mice. Following injection, mice were immediately fed doxycycline-containing diets to induce GFP and Gαt expression. To evaluate the effect of Gβγ inhibitor, gallein, on tumor formation, mice implanted with GFP expressing PC3 cells were treated with gallein (10 mg/kg in PBS, i.p., twice daily). Tumor formation was monitored for 8 weeks by palpation and caliper measurement.
To determine the sensitivity of tumorspherederived xenograft tumors to the combination of paclitaxel treatment and the blockage of Gβγ signaling, 1 × 10 6 of PC3 cells dissociated from the third passages of GFP-or Gαt-expressing tumorsheres formed in the absence of doxycycline treatment were injected subcutaneously to 6 to 8-week-old male mice. Primary tumor growth was initially monitored by bioluminescence imaging. Once tumors became palpable, their growth was recorded with caliper measurement of tumor length (L) and width (W) every 5 days. Tumor volume was calculated by the formula of length × width 2 × 0.5. When the tumors reached a size of ~300 mm 3 , mice were fed doxycyline-containing diet and treated with paclitaxel (10 mg/kg, i.p., twice per week) or vehicle.

Flow cytometry analysis
Cells were dissociated from two-dimensional cultures using Accutase. To dissociate cells from xenograft tumors, tumors were cut into small pieces and digested with collagenase 3 (Collagenase Type3, Worthington) at 37°C for 2-3 hours as reported [41]. Cells were resuspended in PBS containing 2% FBS and 1mM EDTA and stained with APC-conjugated anti-CD44 and PEconjugated anti-CD133 antibodies at 4°C for 30 min. To exclude stromal cells in tumor samples, dissociated cells were also stained first with biotin-labeled anti mouse CD31, CD45 and Ter119, followed by PE/Cy7-conjugated streptavidin. Samples were analyzed on a LSR II violet flow cytometer (Becton Dickinson). At least 1 × 10 6 viable cell events were collected per sample with appropriate negative and single-color positive staining controls. For tumor samples, CD31 -/CD45 -/Ter119cells were analyzed for CD44 and CD133 expression.

Immunohistochemistry
Resected xenograft tumors were fixed with 4% paraformaldehyde, dehydrated with 20% sucrose solution, and embedded in paraffin. Tissues were sectioned at 10-mm intervals, deparaffinized and stained with rabbit anti-Ki67 (1:100, GenTex) or anti-cleaved caspase 3 (1:100, Cell Signaling Technology) antibodies, followed by the HRP-conjugated goat anti rabbit IgG secondary antibody (1:200) using the ultra-sensitive ABC peroxidase staining kit (Pierce) [37]. At least 5 images per section at random fields were taken by a Leica ICC50HD microscope using 10x lens and analyzed by the Image J software.

Statistical analysis
Survival curves were analyzed according to the Kaplan-Meier method using the Graphpad Prism 6 Software, and the differences between curves were evaluated by the log-rank test. To compare xenograft tumor growth rates after doxycycline and/or paclitaxel treatment, the tumor size was plotted against time. A linear regression analysis was performed using the Graphpad Prism 6 Software, to obtain the slopes, and the difference in the slope values were evaluated by the analysis of covariance. Data are expressed as mean ± S.E.M. Means between two groups were compared with a two-tailed, unpaired Student's t test, while comparisons of means from multiple groups with each other or against one control group were analyzed with one-way ANOVA, followed by Tukey's range test. A p value of less than 0.05 was considered to be statistically significant.

ACKNOWLEDGMENTS AND FUNDING
This work was supported in part by grants from NIH grant GM094255 (S. Chen) and DOD PCRP breakthrough award PC111628 (S. Chen). This work utilized core resources supported by Cancer Center Support Grant (P30 CA086862).