High Phosphoantigen Levels in Bisphosphonate-Treated Human Breast Tumors Promote V γ 9V δ 2 T-Cell Chemotaxis and Cytotoxicity In Vivo

The nitrogen-containing bisphosphonate zoledronic acid (ZOL), a potent inhibitor of farnesyl pyrophosphate synthase, blocks the mevalonate pathway, leading to intracellular accumulation of IPP/ApppI mevalonate metabolites. IPP/ApppI accumulation in ZOL-treated cancer cells may be recognized by V γ 9V δ 2 T cells as tumor phosphoantigens in vitro . However, the significance of these findings in vivo remains largely unknown. In this study we investigated the correlation between the anticancer activities of V γ 9V δ 2 T cells and the intracellular IPP/ApppI levels in ZOL-treated breast cancer cells in vitro and in vivo . We found marked differences in IPP/ApppI production among different human breast cancer cell lines post-ZOL treatment. Coculture with purified human V γ 9V δ 2 T cells led to IPP/ApppI-dependent near-complete killing of ZOL-treated breast cancer cells. In ZOL-treated mice bearing subcutaneous breast cancer xenografts, V γ 9V δ 2 T cells infiltrated and inhibited growth of tumors that produced high IPP/ApppI levels, but not those expressing low IPP/ApppI levels. Moreover, IPP/ApppI not only accumulated in cancer cells but it was also secreted, promoting V γ 9V δ 2 T-cell chemotaxis to the tumor. Without V γ 9V δ 2 T-cell expansion, ZOL did not inhibit tumor growth. These findings suggest that cancers producing high IPP/ApppI levels after ZOL treatment are most likely to benefit from V γ 9V δ 2 T-cell– mediated immunotherapy. cancer cells and intracellular IPP/ApppI levels in ZOL-treated breast cancer cells was investigated in vitro and in vivo IPP ApppI. xenografts IPP and cytotoxicity of V γ 9V δ 2 T cells . These results provide some support for an adjuvant role of NBPs in breast cancer. There is now clinical evidence that adding ZOL to endocrine therapy improves disease-free survival in estrogen-responsive early breast cancer (4,5). ZOL has also been shown to durably activate γδ T cells in patients with breast cancer (25). The immunomodulating role of NBPs on human γδ T cells might explain, at least in part, the anticancer effects of ZOL observed in these adjuvant clinical trials. Overall, our findings suggest that cancers producing high IPP/ApppI levels after ZOL treatment are most likely to benefit from V γ 9V δ 2 T-cell–mediated immunotherapy.


INTRODUCTION
Nitrogen-containing bisphosphonates (NBPs) are potent inhibitors of osteoclastmediated bone resorption and can treat bone-loss disorders such as postmenopausal osteoporosis (1), aromatase inhibitor-associated bone loss, and cancer treatment-induced bone loss (2). In preclinical models, NBPs have direct and indirect antitumor properties (3).
NBPs inhibit farnesyl pyrophosphate synthase (FPPS), thus blocking the mevalonate pathway and preventing prenylation of G-proteins (e.g., Ras and Rho). In addition to effects on bone, this can inhibit tumor-cell invasion, proliferation, and adhesion (3). Additional studies suggest that NBPs may also inhibit angiogenesis (3). Importantly, there is now clinical evidence that adding the NBP zoledronic acid (ZOL) to endocrine therapy improves disease-free survival in estrogen-responsive early breast cancer (4,5) and that adding ZOL to neoadjuvant chemotherapy reduces residual invasive breast cancer tumor size (6). How ZOL mediates this activity is not understood.
In preclinical models, Vγ9Vδ2 T cells activated by NBPs have potent antitumor effects in vitro (7,(15)(16)(17)(18), and Vγ9Vδ2 T cells expanded in vitro maintain antitumor activity upon adoptive transfer into immunodeficient mice inoculated with melanoma or pancreatic adenocarcinoma cells (15). ZOL also enhances Vγ9Vδ2 T-cell cytotoxicity in experimental models of chronic myelogenous leukemia, lung, colon or bladder cancer (18)(19)(20)(21). Additionally, clinical evidence suggests anticancer effects of the Vγ9Vδ2 T cells in patients with lymphoid malignancies treated with the NBP pamidronate + interleukin-2 (IL-2) (7,22), and in castration-resistant prostate cancer patients and advanced metastatic breast cancer patients treated with ZOL + IL-2 (23,24). A single dose of ZOL has also been shown to durably activate γδ T cells in patients with breast cancer (25).
Cancer cells may be especially vulnerable to Vγ9Vδ2 T cells (10,11), even though the antitumor activity of Vγ9Vδ2 T cells against different tumor cells treated with NBPs varies (26). Interestingly, IPP/ApppI production after NBP treatment in different cell lines also varies (27) and these mevalonate metabolites could be recognized by Vγ9Vδ2 T cells as tumor phosphoantigens (10,14), suggesting that the anticancer potency of Vγ9Vδ2 T cells might depend on intracellular IPP/ApppI levels in NBP-treated tumors. However, there is no in vivo evidence that a NBP can induce IPP/ApppI accumulation in tumors. In addition, the reasons why IPP/ApppI-producing tumors may become vulnerable to Vγ9Vδ2 T cells in vivo remain unexplored. In the present study, the correlation between the potency of Vγ9Vδ2 T cells to kill cancer cells and intracellular IPP/ApppI levels in ZOL-treated breast cancer cells was investigated in vitro and in vivo. penicillin/streptomycin (Invitrogen). Other breast cancer cell lines were cultured in DMEM medium (Invitrogen) supplemented with 10% fetal bovine serum and penicillin/streptomycin.

Cellular Uptake Assay for ZOL
T47D, MCF-7, and B02 cells were seeded overnight to 10-cm Petri dishes at 4×10 6 cells/dish then treated with 14 C-labeled ZOL for 1h. Cells were then rinsed with PBS, scraped, and extracted with acetonitrile and water. Extracts were separated by centrifugation (14,000×g, 2 minutes). Precipitates were analyzed for total protein content by a modified Bradford procedure. The soluble acetonitrile/water extracts were evaporated and redissolved in 120-µL Milli-Q water, then mixed with OptiPhase HiSafe3 scintillation cocktail (PerkinElmer Wallac, Turku, Finland) and evaluated by liquid scintillation (TriLux Microbeta™, PerkinElmer Wallac). For each sample, 50-µg protein was used for SDS-polyacrylamide gel electrophoresis, and transferred onto a PVDF membrane (Invitrogen), which was then washed with PBS/0.1% Tween and incubated overnight at 4°C with a rabbit polyclonal anti-HMG-CoA Ab (1/500 in 5% nonfat milk; Upstate, Lake Placid, NY). After washing, membranes were incubated with HRP-conjugated anti-rabbit Ab (GE Healthcare, Little Chalfont, UK) and developed using Western Lightning™ Chemiluminescence (PerkinElmer LAS Inc., Boston, MA). As a control, washed membranes were incubated with HRP-conjugated anti-mouse Ab (GE Healthcare) then developed as before. Densitometric analysis of blots was carried out using Quantity One software (Bio-Rad Laboratories, Hercules, CA).

Cytotoxicity Assay
T47D, MCF-7, and B02 breast cancer cells had different growth rates in culture. For cytotoxicity assessments, it was therefore necessary to adjust the cell number at the beginning of the experiment in order to have cell monolayers at a similar confluence at the 18h later, breast cancer cells were cocultured with or without purified Vγ9Vδ2 T cells (cancer cell:Vγ9Vδ2 T-cell ratio was 1:12.5 or 1:25) for 4h or 24h. Viability was assessed by MTT assay.

Experimental Tumorigenesis
Five-week-old female SCID-NOD mice were injected subcutaneously in the flank with . Experiments were carried out as previously described (28). The mitotic index was expressed as percentage of Ki-67-positive nuclei.

Real-time PCR
Total RNA was extracted from T47D and B02 tumors and infiltrating Vγ9Vδ2 T cells.

Transwell Migration Assay
T47D, MCF-7, and B02 breast cancer cells were seeded overnight to 6-well plates at

ZOL Induces Human Vγ9Vδ2 T-Cell Expansion In Vitro and In Vivo
ZOL stimulated, in a time-and dose-dependent manner, expansion of Vγ9Vδ2 T cells from human PBMCs, peaking at day 14 with a 10-µM ZOL concentration ( Figure 1A). These results are in accordance with previous reports showing that NBPs stimulate proliferation of Vγ9Vδ2 T cells in vitro (29). In addition, mass spectrometry revealed rapid IPP accumulation in PBMCs after 1-2 days of ZOL treatment ( Figure 1B), whereas ApppI was detected after 2-4 days ( Figure 1C).
No reactivity of murine γδ T cells to phosphoantigens such as IPP or ApppI has yet been demonstrated and human γδ T cells do not recognize cells from animals other than humans (11,30). Thus, murine PBMCs treated with an amino-bisphosphonate do not activate murine or human γδ T cells. The effect of ZOL on Vγ9Vδ2 T-cell expansion was therefore studied in vivo after intraperitoneal injection of human PBMCs into SCID-NOD mice and ZOL+IL-2 treatment. ZOL stimulated expansion of Vγ9Vδ2 T cells such that they comprised up to 16% of T lymphocytes, versus 2.7% for IL-2 alone (Figures 1D and 1E). This is explained by the fact that ZOL is internalized by antigen-presenting cells from human PBMCs where it induces the intracellular accumulation of IPP/ApppI and subsequent activation of Vγ9Vδ2 T-cell expansion (12), making our animal model most suitable for studying the role of ZOL in cancer immunotherapy in vivo. Moreover, although ZOL was administered more frequently in our model than the current clinical dosing regimen in humans, the dose given (30 µg/kg) was lower than the dosing in humans (100 µg/kg), supporting the clinical relevance of our findings.

ZOL Induces IPP/ApppI Production in Human Breast Cancer Cells In Vitro
Prior studies showed the intracellular accumulation of IPP and ApppI in NBP-treated cancer cell lines in vitro (10,27). In our study, ZOL-induced IPP/ApppI production was  (31). Variations in endocytosis between the cancer cell lines might explain differences in IPP/ApppI production induced by ZOL. Uptake of 14 C-labeled ZOL was therefore measured in T47D, MCF-7, and B02 cells. Uptake was similar in T47D and MCF-7 cells, but 3-fold lower in B02 cells ( Figure 2C). Additionally, we found a strong correlation between HMG-CoA reductase expression, a mevalonate pathway enzyme upstream of FPP synthase, and ZOL-induced IPP/ApppI accumulation ( Figure 2D). As shown by Western blotting, lovastatin (an HMG-CoA reductase inhibitor) substantially increased HMG-CoA protein levels in ER-positive T47D and MCF-7 cells ( Figure 2D). By contrast, HMG-CoA reductase was undetectable, and only a faint protein band was observed in lovastatin-treated ER-negative B02 cells ( Figure 2D). These results agree with data reported by Borgquist et al. 13 demonstrate that IPP/ApppI production in breast cancer cells depends on cellular uptake of NBP and on the activity of the mevalonate pathway. Moreover, we suggest that luminal neoplastic mammary epithelial cells might be more sensitive to bisphosphonate therapy than basal neoplastic mammary epithelial cells.

Breast Cancer Cells In Vitro and In Vivo
The antitumor potency of human Vγ9Vδ2 T cells against ZOL-treated breast cancer cell lines was first examined in vitro (Figures 3A-3F). Incubation for 1h with ZOL (1-25 µM) did not affect survival of T47D, MCF-7, and B02 cells ( Figures 3A, 3C, and 3E, respectively), whereas these bisphosphonate concentrations did induce intracellular IPP/ApppI accumulation in T47D and MCF-7 cells (Figures 2A-B). Coculture of ZOL-treated T47D cells with purified Vγ9Vδ2 T cells led to dose-dependent cancer cell death, which was statistically significant with ZOL at concentrations as low as 1 µM for 1h ( Figure 3B). Additionally, Vγ9Vδ2 T cells caused ZOL-treated MCF-7 cell death; concentrations of ZOL ≥10 µM were required to "prime" MCF-7 cells for Vγ9Vδ2 T-cell-mediated cytotoxicity ( Figure 3D). In sharp contrast, Vγ9Vδ2 T cells were not cytotoxic against ZOL-treated B02 cells ( Figure 3F).
Although human cancer cell lines exhibit susceptibility to Vγ9Vδ2 T-cell-mediated cytotoxicity upon NBP treatment, the in vitro antitumor activity of Vγ9Vδ2 T cells against tumor cells treated with NBPs varies greatly (10,11,15). For example, Vγ9Vδ2 T cells fail to efficiently kill a variety of human renal (ACHN, Caki-2, A-704) and gastric (MKN45, MKN74) cancer cell lines after NBP pretreatment (11). Here, our results strongly suggest that ZOLinduced IPP/ApppI accumulation in breast cancer cells was responsible for Vγ9Vδ2 T cellmediated cytotoxicity. In this respect, co-treatment of T47D cells with ZOL+lovastatin almost completely eliminated IPP/ApppI phosphoantigen production by these cells and substantially reduced Vγ9Vδ2 T cell cytotoxicity (supplementary data; Figure S1). It is therefore most conceivable that the lack of cytotoxicity of Vγ9Vδ2 T cells against some cancer cell lines is We next conducted in vivo experiments to examine the contribution of IPP/ApppI accumulation in breast tumors to Vγ9Vδ2 T-cell cytotoxicity. Earlier experimental in vivo studies used human Vγ9Vδ2 T cells expanded in vitro and then purified before inoculation (15,18,19,20). Here, we showed that ZOL + IL-2 treatment induced Vγ9Vδ2 T-cell expansion from human PBMCs injected intraperitoneally into SCID-NOD mice ( Figure 1E). In the clinic, it has been recently shown that the treatment of advanced breast cancer patients with ZOL and low-dose IL-2 triggers Vγ9Vδ2 T cell amplification in vivo (24). We therefore chose to use this therapeutic strategy to treat tumor-bearing animals. SCID-NOD mice bearing subcutaneous breast cancer xenografts were inoculated with human PBMCs, with expansion of Vγ9Vδ2 T cells in vivo (Figure 4). Treatment with ZOL alone or PBMC+IL-2 did not inhibit subcutaneous tumor growth (Figures 4A and 4B). In contrast, PBMC + IL-2 + ZOL completely blocked growth progression of T47D tumors, compared to that observed in placebo-treated animals (P<.01; Figure 4A). In addition, in situ immunodetection of the proliferation marker Ki-67 nuclear antigen in T47D tumors from mice treated with PBMC + IL-2 + ZOL showed drastic reduction in proliferative index versus tumors from placebo-treated mice ( Figure 4C). However, this treatment did not inhibit growth or proliferative index of B02 subcutaneous tumors (Figures 4B and 4D). Of note, Vγ9Vδ2 T-cell infiltrates were detected in T47D but not B02 tumors from mice treated with PBMC+IL-2+ZOL ( Figures 4E, 4F, and 4G), as judged by both RT-PCR and immunohistochemistry. Moreover, serum levels of human IFN-γ were high in T47D-tumor-bearing mice treated with PBMCs + IL-2 + ZOL (942 ± 240 pg/mL, n = 5) and moderate in animals treated with PBMCs + IL-2 (563 ± 370 pg/mL, n = 5), whereas the cytokine was not detected in animals treated with placebo or ZOL alone, demonstrating that Vγ9Vδ2 T cells were activated in vivo.
Because Vγ9Vδ2 T cells infiltrated subcutaneous T47D tumor xenografts in mice treated with PBMC+IL-2+ZOL ( Figures 4E and 4G Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on June 6, 2011; DOI: 10.1158/0008-5472.CAN-10-3862 extracts by mass spectrometry. As expected, IPP and ApppI were barely detectable in T47D tumors from placebo-treated mice ( Figure 5A), whereas IPP levels were substantial 24h after the last ZOL dose in mice treated with PBMC + IL-2+ ZOL (3.7 pmol; 82 nM) ( Figure 5B). By contrast, ApppI remained undetectable at 24h. Conversely, lower IPP levels were detected at 48h after the last ZOL dose in PBMC + IL-2 + ZOL-treated mice (0.7 pmol; 16 nM), whereas ApppI levels were measured (0.11 pmol; 2.5 nM) ( Figure 5C). Of note, IPP and ApppI were also detected in T47D tumor extracts from mice treated with ZOL alone. In contrast, ZOLinduced IPP/ApppI was not detected in B02 tumor xenografts (not shown).
These data, to the best of our knowledge, show for the first time that subcutaneous tumors can produce IPP/ApppI after ZOL administration, demonstrating the uptake of the  P < .01) versus assay buffer. Migration of Vγ9Vδ2 T cells was much higher with conditioned medium from ZOL-treated T47D and MCF-7 breast cancer cells (Figures 6B and 6C).
Additionally, conditioned medium from T47D cells treated with ZOL+lovastatin resulted in less migration of Vγ9Vδ2 T cells ( Figure 6C). To clarify whether IPP/ApppI acted as migration factors for Vγ9Vδ2 T cells, we used mass spectrometry to measure IPP levels in conditioned medium from ZOL-treated breast cancer cell lines (T47D, MCF-7, B02). We found substantial IPP levels in conditioned medium from ZOL-treated T47D and MCF-7 cells, but not B02 cells (data not shown). Moreover, 0.5-µM IPP or 1-µM ApppI significantly stimulated γδ T-cell migration ( Figure 6C). Because ApppI can be converted into IPP on hydrolysis by nucleotide pyrophosphatases (14), it also may be processed into IPP for subsequent recognition by the Vγ9Vδ2 TCR. The cytokines IL-6, IL-8, and MCP-1 are also γδ T-cell chemoattractants (30).
However, we found no differences in these or any other cytokines in a human cytokine antibody array of conditioned medium from untreated or ZOL-treated T47D, MCF-7, and B02 cells ( Figure 6D). Although B02 cells produced a greater range of cytokines than T47D cells, cytokine profiles were similar for ZOL-treated or untreated cells ( Figure 6D). By ELISA, untreated B02 cells (control) produced more IL-8 and MCP-1 compared with T47D or MCF-7 control cells ( Figure 6E); IL-6 was detected in B02 cells only.
In conclusion, our study shows for the first time that ZOL induces the production of IPP/ApppI phosphoantigens by breast tumors in vivo which, in turn, promote chemotaxis and cytotoxicity of Vγ9Vδ2 T cells. These results provide some support for an adjuvant role of NBPs in breast cancer. There is now clinical evidence that adding ZOL to endocrine therapy improves disease-free survival in estrogen-responsive early breast cancer (4,5). ZOL has also been shown to durably activate γδ T cells in patients with breast cancer (25). The immunomodulating role of NBPs on human γδ T cells might explain, at least in part, the anticancer effects of ZOL observed in these adjuvant clinical trials. Overall, our findings suggest that cancers producing high IPP/ApppI levels after ZOL treatment are most likely to benefit from Vγ9Vδ2 T-cell-mediated immunotherapy.     Identification of IPP and ApppI in T47D tumors of placebo-and ZOL-treated mice was performed by high-performance liquid chromatography negative ion electrospray ionization mass spectrometry. Selected reaction monitoring chromatograms are shown. They correspond to tumor extracts obtained from placebo-treated mice (A) and animals that received PBMC + IL-2 + ZOL whose tumors were collected 24h (B) and 48h (C) after the last