Research ArticleCancer

A combination therapy for KRAS-driven lung adenocarcinomas using lipophilic bisphosphonates and rapamycin

See allHide authors and affiliations

Science Translational Medicine  19 Nov 2014:
Vol. 6, Issue 263, pp. 263ra161
DOI: 10.1126/scitranslmed.3010382


Lung cancer is the most common human malignancy and leads to about one-third of all cancer-related deaths. Lung adenocarcinomas harboring KRAS mutations, in contrast to those with EGFR and EML4-ALK mutations, have not been successfully targeted. We describe a combination therapy for treating these malignancies with two agents: a lipophilic bisphosphonate and rapamycin. This drug combination is much more effective than either agent acting alone in the KRAS G12D–induced mouse lung model. Lipophilic bisphosphonates inhibit both farnesyl and geranylgeranyldiphosphate synthases, effectively blocking prenylation of KRAS and other small G proteins (heterotrimeric GTP-binding protein, heterotrimeric guanine nucleotide–binding proteins) critical for tumor growth and cell survival. Bisphosphonate treatment of cells initiated autophagy but was ultimately unsuccessful and led to p62 accumulation and concomitant nuclear factor κB (NF-κB) activation, resulting in dampened efficacy in vivo. However, we found that rapamycin, in addition to inhibiting the mammalian target of rapamycin (mTOR) pathway, facilitated autophagy and prevented p62 accumulation–induced NF-κB activation and tumor cell proliferation. Overall, these results suggest that using lipophilic bisphosphonates in combination with rapamycin may provide an effective strategy for targeting lung adenocarcinomas harboring KRAS mutations.


Lung adenocarcinomas account for about 50% of all non–small cell lung cancers (NSCLC), the most common type of human malignancy and a leading cause of cancer-related mortality worldwide. There has been rapid progress in developing targeted therapies for lung adenocarcinomas over the past decade, including gefitinib and erlotinib, which target EGF receptor mutations (1, 2), and crizotinib, which targets the transforming EML4-ALK fusion gene (3). However, KRAS mutations, which are commonly found in smokers and Caucasian patients, are not effectively targeted by currently available therapeutics and have low survival rates, as well as frequent drug resistance (4). KRAS mutations at amino acid position 12, 13, or 61 are widely found in human pancreatic, thyroid, lung, and colorectal cancers (5). They typically impair GTPase (guanosine triphosphatase) activity and lead to constitutive activation of downstream signaling pathways. It is difficult to develop potent KRAS-mutant–specific inhibitors that can directly restore intrinsic GTPase activity, although specific inhibitors of KRAS G12C have recently been reported (6), as have attempts to interfere with mutated KRAS function by altering its membrane localization, inhibiting its downstream effectors, as well as searching for synthetic lethality (7, 8).

Farnesylation and correct membrane localization are essential for the in vivo biological activity of RAS proteins (9, 10). CAAX peptidomimetics, farnesyltransferase and geranylgeranyltransferase inhibitors (FTI/GGTIs), farnesylthiosalicylic acid (salirasib), which mimics farnesylcysteine, as well as small-molecule inhibitors of KRAS-PDEδ interactions have all been developed to circumvent KRAS posttranslational modification and membrane anchoring (7, 1113). Knockout mouse models support the notion that disruption of protein prenylation severely impairs lung cancer development induced by KRAS mutations (14, 15). However, there has been little success in clinical trials with these small-molecule inhibitors, probably due to the existence of “cross-prenylation” (16), in which an FTI can fail due to alternative KRAS geranylgeranylation, suggesting the need for combination therapies. In addition to the protein prenyltransferase inhibitors, there is interest in the development of compounds, such as bisphosphonates (see fig. S1) that directly inhibit the biosynthesis of the two prenyldiphosphate substrates: farnesyldiphosphate (FPP) and geranylgeranyldiphosphate (GGPP), catalyzed by the respective synthases, FPP synthase (FPPS) and GGPP synthase (GGPPS). Bisphosphonates are used to treat a variety of bone resorption diseases and function by blocking FPPS activity in osteoclasts. In previous work, we developed “lipophilic” bisphosphonates in which hydrophobic side chains were added to a pyridinium bisphosphonate. These compounds do not bind to bone mineral, but maintain inhibitory activity against FPPS, as well as GGPPS (17), both of which can provide membrane anchoring 15- and 20-carbon isoprenoid chains for Kras posttranslational modification.

Impaired protein processing, folding, and trafficking usually induce endoplasmic reticulum (ER) stress and autophagy if the protective unfolded protein response (UPR) is not sufficient to clear the incorrectly processed proteins (18), and indeed, inhibitors of FPP and GGPP biosynthesis, such as bisphosphonates and statins, have been reported to initiate autophagy in cells (19, 20). The role of autophagy in tumorigenesis has been considered a “double-edged sword,” because it can inhibit tumor initiation at an early stage or get adopted by tumor cells as a survival mechanism at an advanced stage (21). Here, we sought a combination therapy that would ultimately stop KRAS prenylation and temporally modulate autophagy as an effective two-pronged approach against lung adenocarcinomas.


Bisphosphonates inhibit FPPS and GGPPS activity

We tested a library of 30 synthetic analogs of zoledronate (Fig. 1) for growth inhibition of two KRAS-mutant cell lines (6#, L2) and of control mouse embryonic fibroblasts (MEFs). We found the most anti-growth activity with BPH-1222, a zoledronate analog that has a C8 side chain and a 1-hydroxy (1-OH) group [~1 μM IC50 (median inhibitory concentration); Fig. 2A and fig. S2, A and B]. Compounds with very short or very long chains inhibited growth the least, whereas BPH-1222 and other intermediate chain length compounds had the most activity. In vitro inhibitory activities against human FPPS (Ki as low as ~1 nM) correlated well with activities in inhibiting cell growth, suggesting FPPS as one possible in vivo target (Fig. 2B, table S1, and fig. S2C). To uncover how BPH-1222 binds and inhibits FPPS, we determined the structure of the BPH-1222–FPPS complex by single crystal x-ray crystallography. In Fig. 1D, the bisphosphonate, head group, and imidazolium ring bind to human FPPS in essentially the same manner as does zoledronate (shown superimposed), with a 0.45-Å root mean square deviation for the common atoms (see table S2 for a summary of structural and refinement statistics).

Fig. 1. BPH-1222 and zoledronate bind to the same site of FPPS.

(A) Zoledronate. (B) Library of zoledronate analogs, side-chain length n = 0 to 15, X = H or OH. (C) BPH-1222. (D) Structure of BPH-1222 binding to FPPS was determined by single crystal x-ray crystallography. Zoledronate binding is shown superimposed with BPH-1222 (zoledronate in yellow and BPH-1222 in cyan).

Fig. 2. Bisphosphonates suppress cell growth by inhibiting FPPS and GGPPS.

(A) IC50 of zoledronate and its hydroxy analogs (see fig. S2 for desoxy analogs) was determined in cell lines derived from KRAS-shp53 mouse lung cancer model (6# and L2) and control MEFs using MTT assay. (B) Ki of compounds in (A) was measured in vitro against human FPPS or GGPPS. (C) Mouse lung cancer cells (6#) were treated with single drugs (FTI-277, 15 μM; GGTI-298, 15 μM; BPH-1222, 10 μM; FOH, 10 μM; GGOH, 10 μM; ascorbic acid, 50 μM) or drug combinations as indicated for 48 hours, and images were taken under phase-contrast microscope. Scale bars, 100 μm. (D and E) Mouse lung cancer cells (L2) and MEFs (matched genetic background), or (E) MEFs transformed by KRAS-shp53 and MYCL1-shp53-shRB1 were treated with different concentrations of BPH-1222 for 3 days, and cell survival was measured using Cell Proliferation Reagent WST-1 from Roche. All data were collected from two independent experiments (each has triplicate wells) and raw ODs (optical densities) were normalized to a common scale with GraphPad Prism. Data are presented as means ± SE, n = 6.

Moreover, the lipophilic bisphosphonates also inhibit human GGPPS (Fig. 2B, fig. S2C, and table S1). Zoledronate, as well as other bisphosphonates such as risedronate and alendronate, inhibits FPPS, but does not inhibit GGPPS, because the latter lacks the third Asp required for [Mg2+] coordination to the bisphosphonate (22). With the lipophilic bisphosphonates binding to GGPPS, the loss of the additional binding interaction due to the absence of the third Asp residue is likely made up for by increased van der Waals interactions in the hydrophobic tunnel that normally houses the allylic side chains of the growing isoprenoid diphosphate products. The intermediate chain length species have the best inhibitory activity for GGPPS, a Ki of ~300 nM for BPH-1222. There is little difference in GGPPS enzyme inhibition activity between the 1-OH and 1-desoxy species (table S1), so in this work, we focused on BPH-1222, the N-octyl analog of zoledronate, over its desoxy analog to eliminate the possibility of a retro-Michael reaction (in which the hydrophobic side chain would be lost) in vivo.

As can be seen in table S1, the Kis for FPPS inhibition by zoledronate (ID 1) and BPH-1222 (ID 9) are quite similar (1 to 2 nM), but the IC50 values for cell growth inhibition are ~5× to 10× lower with BPH-1222. This likely correlates with the lower GGPPS Ki values (~3 μM for zoledronate; 300 nM for BPH-1222) and with the higher clogP (the computed logarithm of the oil/water partition coefficient) value for BPH-1222 (0.25) versus zoledronate (–3.9).

Inhibition of protein farnesylation (with FTI-277), geranylgeranylation (with GGTI-298), or both (fig. S2D) had little effect on cell proliferation (Fig. 2C and fig. S2E). However, there was massive cell death within 3 days of starting treatment with BPH-1222. Supplementation with geranylgeraniol (GGOH), but not farnesol (FOH) or an antioxidant, substantially rescued cells from bisphosphonate-induced cell death (Fig. 2C). Similar rescue effects were seen with zoledronate (23), as well as with the lipophilic pyridinium bisphosphonate BPH-714 (17) and with simvastatin, a potent HMG-CoA (3-hydroxy-3-methylglutaryl coenzyme A) reductase inhibitor that shuts down all isoprenoid biosynthesis (fig. S2F). As noted by Goffinet et al. (23), the chemical target for zoledronate is FPPS (not GGPPS), but the main “biological effect” of zoledronate involves protein geranylgeranylation. Thus, dual FPPS/GGPPS-targeting bisphosphonates (such as BPH-1222) are expected to be particularly potent because formation of the FPP substrate for GGPP biosynthesis is blocked by FPPS inhibition, and GGPP production is particularly important for cell survival. Although inhibition of FPPS is expected to have effects on diverse metabolic pathways (such as sterol and steroid biosynthesis), these effects cannot be major ones responsible for cell death because the effects of BPH-1222 are reversible upon addition of GGOH.

We found that BPH-1222 was more toxic to cells harboring KRAS mutations, such as cell lines derived from a mouse model of KRAS-induced lung adenocarcinoma, as well as to MEFs transformed by KRAS in vitro (Fig. 2, D and E). Together, these results suggest that blocking protein prenylation by lipophilic bisphosphonates can be used as a targeted therapy for cancer cells that carry KRAS mutations, because potentially both FPPS and GGPPS can be targeted, and the lipophilic bisphosphonates have much better clogP values than do more conventional bisphosphonates. Furthermore, the lipophilic bisphosphonates do not bind to bone mineral, which rapidly removes conventional bisphosphonates from the circulatory system (24).

Bisphosphonates block KRAS prenylation and induce its degradation

“KRAS addiction” has been shown in a mouse lung cancer model using inducible KRAS G12D (25). Given the observation that cells bearing KRAS mutations were more sensitive to bisphosphonate treatment, we were interested in determining if the lipophilic bisphosphonates blocked KRAS prenylation. We used a cell fractionation assay to check KRAS protein prenylation status, because unprenylated KRAS proteins lose their ability to avidly associate with cell membranes and, consequently, appear in the cytosolic fraction (26). BPH-1222 treatment robustly inhibited protein farnesylation, as well as protein geranylgeranylation, as indicated by HRAS degradation and RAP1A dislodgement from the membrane fraction, respectively (27). Other bisphosphonates, including BPH-714 and zoledronate, showed similar effects, but at higher concentrations (Fig. 3A). As expected, a substantial amount of KRAS was unprenylated and was also partially degraded. There was also a reduction of downstream AKT activity, which had been induced by oncogenic KRAS expression (Fig. 3A). This result was further confirmed with HDJ2, RAP1A, and KRAS electrophoretic mobility shift assay (Fig. 3B and table S3) (28). To further verify whether cytosolic KRAS represents the unprenylated form of the protein, we determined the molecular weight of KRAS expressed in U2OS cells after BPH-1222 treatment. As expected, the molecular weights of KRAS proteins from the cytosolic fraction (supernatant) and the membrane-bound fraction (pellet) fit well with unprenylated and farnesylated forms, respectively (see fig. S3A and table S4 for a summary). The effects of a lipophilic bisphosphonate on KRAS function were further tested in a mouse lung cancer (M3L2) and two human pancreatic cancer (Panc-1 and MiaPaCa2) cell lines carrying endogenous KRAS G12D or G12C mutations (Fig. 3, C and D). BPH-1222 treatment greatly reduced the amount of GTP (guanosine 5′-triphosphate)–bound KRAS (representing the active form of the KRAS protein) and led to down-regulation of the AKT pathway, as well as the activation of apoptosis, as shown by increase in caspase-3 cleavage (Fig. 3C).

Fig. 3. Bisphosphonates inhibit KRAS prenylation and induce ER stress and autophagy.

(A) U2OS cells expressing Flag-KRAS G12D were treated with bisphosphonates (BPH-1222, 10 μM; BPH-714, 10 μM; zoledronate, 20 μM) or chloroquine (CQ; 30 μM) for 48 hours. Cellular distribution of proteins (HRAS, KRAS, and RAP1A) that require prenylation was examined by immunoblotting. p, pellet contains correctly prenylated proteins that bind avidly to membrane; s, supernatant contains unmodified proteins in the cytoplasm. *, HRAS signal left on the membrane. **, autophagy activation determined by the presence of LC3II (bottom band, PE-conjugated form). (B) Same cells were treated with FTI-277 (15 μM), GGTI-298 (15 μM), or BPH-1222 (5, 10, and 15 μM) for 48 hours. Cell lysates were separated with 15-cm SDS-PAGE (SDS–polyacrylamide gel electrophoresis) and blotted with indicated antibodies. *, KRAS mobility shift was observed. (C) Mouse lung cancer cells (M3L2) were treated with BPH-1222 for 48 hours and analyzed for KRAS, AKT, caspase-3 activation, and LC3 conversion. KRAS-GTP was pulled down from whole-cell lysate with RAF-1 RBD beads and immunoblotted with total KRAS antibody. (D) Human pancreatic cancer cells harboring KRAS mutations (Panc-1 and MiaPaCa2) were treated with BPH-1222 (10 μM) for 48 hours and analyzed the same way as in (C). (E) BPH-1222 treatment (10 μM) for 1, 2, or 3 days induced ER stress (CHOP, BiP), autophagy (PE-conjugated LC3II), and apoptosis (caspase-3) in mouse lung cancer cells (6#) in a time-dependent manner.

Bisphosphonates enhance ER stress and initiate autophagy

Cancer cells usually exhibit a “stress phenotype” that consists of replicative stress, mitotic stress, metabolic stress, oxidative stress, and proteotoxic or ER stress. They are, therefore, vulnerable to further enhancement of these stresses by chemotherapy (29). Bisphosphonates, as demonstrated above, potently block protein prenylation by eliminating the source of isoprenoid chains and lead to the accumulation of incorrectly folded proteins, inducing the so-called UPR, or ER stress, as is also observed when cells are treated with HMG-CoA reductase inhibitors (30, 31). In Fig. 3E, treatment of cells with the lipophilic bisphosphonate BPH-1222 potently induced C/EBP homologous protein (CHOP) and binding immunoglobulin protein (BiP), the two main markers for the ER stress response (18). Notably, the ER stress response induced by this bisphosphonate is specifically due to blockade of protein prenylation, because supplementation with GGOH, but not antioxidants, completely abolished the up-regulation of CHOP, BiP, as well as phos-PERK induced by the bisphosphonate (fig. S3B). When cells are unable to handle excessive ER stress, they activate autophagy to eliminate incorrectly processed proteins as a defense mechanism for cell survival. In tumor cells treated with BPH-1222, autophagy was induced within 2 to 3 days, as indicated by the accumulation of the phosphatidylethanolamine (PE)–conjugated form of LC3 (Fig. 3, A and E). Similar results have been observed in cells treated with other bisphosphonates, and it has been proposed that the combination of inhibitors of autophagy with GGPPS inhibitors might be a therapeutic strategy, because autophagy is a defense mechanism in this context (19).

Rapamycin, but not chloroquine, sensitizes tumor cells to bisphosphonates in vivo

Recent reports have indicated that KRAS-driven tumor cells depend on autophagy to help reduce reactive oxygen species (ROS), as well as provide substrates to fuel cell metabolism (32, 33). We therefore tested the hypothesis that blocking autophagy might sensitize tumor cells to bisphosphonate treatment. We first determined the effects on cell survival of treatment with bisphosphonate and chloroquine, an autophagy inhibitor (34). As expected, essentially additive effects of the bisphosphonate and chloroquine were observed in all the lung cancer cell lines derived from the mouse KRAS lung cancer model (with KRAS G12D and p53 knockdown), as well as in human NSCLC cell lines with KRAS mutations (A549 and A427) (Fig. 4A and fig. S4A).

Fig. 4. Rapamycin but not chloroquine sensitizes tumor cells to BPH-1222 treatment in vivo.

(A) Mouse lung cancer cells (6#) were treated with BPH-1222 in the presence of different concentrations of chloroquine (CQ) for 3 days, and cell survival was examined with Cell Proliferation Reagent WST-1. All data were collected from triplicate wells and are presented as means ± SD, n = 3. (B) Mouse lung cancer (L2, infected with 5xκB-luci reporter) syngeneic grafts were treated with BPH-1222 (2 mg/kg) plus either chloroquine (60 mg/kg) or rapamycin (2.5 mg/kg) on alternating days for 3 weeks. Scale bar, 10 mm. (C) NF-κB activity in tumor grafts was examined by in vivo luciferase imaging system (IVIS) immediately after the 3-week treatment. (D) Mouse lung cancer cells (6#) were treated with BPH-1222 (10 μM) or combinations with CQ (30 μM) or rapamycin (Rapa) (0.1 μM) for 48 hours, and NF-κB target genes were examined by quantitative RT-PCR. All data were collected from triplicate wells and are presented as means ± SD, n = 3. (E to G) L2 tumor samples treated with different drug combinations were sectioned and immunostained with Ki-67 and cleaved caspase-3 antibodies. Caspase-3–positive cells were marked with arrowheads. Scale bars, 100 μm. Percentages of Ki-67– and caspase-3–positive cells were quantified and plotted in (F) and (G). ns, not significant. (H) Mouse lung cancer (L2) syngeneic grafts were treated with BPH-1222 (2 mg/kg), Rapa (2.5 mg/kg), or the combination for 3 weeks. Scale bar, 10 mm. (I) Mouse lung cancer cells (6#) were treated with BPH-1222 (10 μM), Rapa (0.1 μM), or the combination for 48 hours and examined by immunoblotting.

We next tested this combination in a syngeneic graft model using L2 cells derived from a mouse lung adenocarcinoma (with KRAS G12D and p53 knockdown) developed in C57B mice. Treatment was started 3 weeks after subcutaneous transplantation, when the tumor mass was palpable. Surprisingly, 24 days of BPH-1222 plus chloroquine therapy only minimally suppressed tumor growth in animals. In contrast, BPH-1222 plus rapamycin (a potent autophagy inducer) showed substantially better efficacy (Fig. 4B). Our results are more in line with those recently reported for pancreatic cancer (KRAS G12D and p53–/–) development, where it was shown that another autophagy inhibitor, hydroxychloroquine, tended to accelerate tumor formation (35). We noticed that tumors treated with BPH-1222 plus chloroquine had elevated nuclear factor κB (NF-κB) activity, as shown by in vivo imaging (Fig. 4C). This result was further confirmed in cultured cells treated with this combination of drugs by quantitative reverse transcription polymerase chain reaction (RT-PCR) of NF-κB target genes (Fig. 4D). The NF-κB activity (Fig. 4, C and D) was correlated with p62 protein accumulation in the cells, probably due to the blockage of autophagy flux by chloroquine (fig. S4B). Indeed, high p62 protein level in cells is known to activate NF-κB through the TRAF6-TBK1 pathway (36). We have reported earlier that high NF-κB activity in tumors may stimulate cell proliferation (37), and this appeared to be the case in tumor samples treated with BPH-1222 plus chloroquine (Fig. 4, E and F), although these tumors also showed an increased number of apoptotic cells (Fig. 4, E and G). We suggest that the elevated NF-κB activity and cell proliferation impaired the ability of bisphosphonates to reduce the size of tumors (Fig. 4B).

The results obtained in Fig. 4 (B and E) suggested that a combination therapy with a lipophilic bisphosphonate together with the autophagy inducer rapamycin might be more effective in treating KRAS-induced lung adenocarcinomas. We found that in the L2 cell syngeneic graft model, the combination of BPH-1222 with rapamycin was indeed much more effective than either single agent acting alone (Fig. 4H). Additionally, rapamycin also potently blocked phosphorylation of the two mammalian target of rapamycin (mTOR) substrates (p70 S6 kinase and 4E-BP1), both of which are important for boosting metabolism in cancer cells (Fig. 4I).

Bisphosphonate and rapamycin combination suppresses tumor growth in lung cancer models

We next tested the BPH-1222 plus rapamycin combination in treating lung adenocarcinoma in both the syngeneic orthotopic graft model and the lentiviral vector–mediated model, which represent the lung microenvironment well. In the orthotopic model, tail veil injection of 2 × 105 M3L2 cells reproducibly generated massive lung tumors in FVB mice, resulting in a median survival of 33 days. For the treatment group, all mice (n = 13) were given BPH-1222 and rapamycin alternately over the course of 16 days after the transplantation, for a total of nine doses of each compound. BPH-1222 has good pharmacokinetic properties (fig. S4C), and this treatment potently suppressed tumor growth, although mouse survival was not significantly prolonged (37 days, n = 13, compared to 33.5 days, n = 8 in the control group, P = 0.06) due to internal bleeding in the thoracic cavity in some of the mice under treatment, even those with small tumor burdens (Fig. 5, A to C). We suspect that internal bleeding in the lung is related to tumor shrinkage, because we did not observe this side effect in the lung with the same drug combination when tumors were transplanted subcutaneously.

Fig. 5. Combination therapy inhibits tumor growth in orthotopic graft model and KRAS-shp53 lentiviral model.

(A to C) Mouse orthotopic grafts were induced by tail vein injection of M3L2 cells. Mice were left untreated or given treatment (BPH-1222 plus rapamycin) 16 days after the inoculation. (A) Lungs were collected when mice reached the Institutional Animal Care and Use Committee (IACUC) endpoint (having severe breathing difficulty). Tumor lesions larger than 1 mm in diameter were counted. (B) Ctrl, 29.8 average lesions per mouse, n = 8; Treat, 13.5 average lesions per mouse, n = 13. Kaplan-Meier curves of mice from control group and treatment group. All surviving mice were collected on day 50. Median survival time: Ctrl, 33.5 days, n = 8; Treat, 37 days, n = 13 (C). (D to H) Mouse lung adenocarcinomas were induced by intratracheal infection with KRAS-shp53 lentiviral vectors. Mice were left untreated or given treatment when luciferase signals from tumors were detectable (Low-Ctrl and Low-Treat groups, luciferase signal: 103 to 105; High-Ctrl and High-Treat groups, luciferase signal: >105). (D) Luciferase imaging results of one mouse from Low-Ctrl group and one mouse from Low-Treat group, to show the shrinkage of tumor after the combination therapy. See fig. S5 for results from all mice. (E) Fold changes of luciferase signal after 2 weeks of treatment. Negative value means shrinkage in tumor size. (F) Kaplan-Meier curves of mice from all groups. Median survival time: Low-Ctrl, 55 days, n = 8; Low-Treat, 75.5 days, n = 6; High-Ctrl, 28 days, n = 8; High-Treat, 54 days, n = 6. (G) Tumors from Low-Ctrl and Low-Treat groups were sectioned and immunostained with different antibodies and quantified in (H). Caspase-3–positive cells were marked with arrowheads. Scale bars, 100 μm (insets in 4E-BP1-p staining: scale bars, 20 μm).

In the lentiviral vector–mediated model, as we reported previously, lung tumors appeared “visible” by luciferase imaging at about 2 months after lentiviral vector infection. In the absence of any treatment, the tumors grew rapidly, and all mice (n = 8) died from full-blown lung cancer, with a median survival of 55 days from when the tumors were first detected by imaging. A combination therapy of BPH-1222 and rapamycin was given to 12 mice about 3 months after lentiviral infection (the first day of drug treatment is counted as day 1 in Fig. 5, D and E), for a total of 16 doses of each compound. We monitored the tumor load of each mouse every 9 to 10 days throughout the course of treatment. The 12 mice were divided into two groups according to their tumor burden at the beginning of treatment—a high-burden group (n = 6, initial luciferase signal >105) and a low-burden group (n = 6, initial luciferase signal 103 to 105). The combination treatment substantially delayed tumor development and prolonged mouse survival in both the low-burden group (median survival of 75.5 days, compared to 55 days without treatment, P = 0.0009) and the high-burden group (median survival of 54 days, compared to 28 days without treatment, P = 0.02; Fig. 5E). Tumor regression was observed in most of the mice during treatment, although tumor size did increase after treatment ceased (Fig. 5, D and F, and fig. S5). Tumors from treated mice showed no change in pathological pattern [hematoxylin and eosin (H&E), SPC staining], but a large reduction in cell proliferation (Ki-67 staining) and mTOR pathway activity (phos–4E-BP1), although apoptotic cells were rarely found (cleaved caspase-3) (Fig. 5, G and H). Unlike the results we reported previously with an IKK2 inhibitor that suppressed tumor progression by reducing ERK (extracellular signal–regulated kinase) signaling (37), there was no significant change of ERK phosphorylation after the bisphosphonate plus rapamycin treatment in vivo (Fig. 5, G and H). BPH-1222 (either alone or combined with other agents) slightly increased ERK phosphorylation in cell culture conditions (fig. S6A), but its mechanism of action here is unknown. Nevertheless, we tested if a combination with an ERK inhibitor might further improve treatment efficacy. However, ERK inhibitor U0126 increased KRAS protein and c-RAF, MEK (mitogen-activated protein kinase kinase), and AKT phosphorylation under all conditions tested (fig. S6B), perhaps due to the interruption of the ERK-negative feedback loop (38).

Zoledronate itself has been shown to potentiate the killing of osteosarcoma cells by RAD001 (a rapamycin analog) (39), and, although less potent than BPH-1222, zoledronate still synergizes with rapamycin in vivo for killing KRAS-mutant tumors (fig. S7). This suggests the possibility of a more immediate translation into clinical trials, with zoledronate plus rapamycin (or an analog) for lung adenocarcinoma, given that both drugs have been prescribed for many years with good safety records.


RAS mutations (including HRAS, KRAS, and NRAS) are commonly found in a variety of human cancers including lung, colon, and pancreatic cancers (5). Here, we investigated the efficacy of treating lung adenocarcinomas carrying a KRAS mutation with a combination of a lipophilic bisphosphonate, an analog of zoledronate, and rapamycin. Bisphosphonates are a class of drugs widely used for treating osteoporosis and for preventing bone metastasis of certain cancers (40). Mechanistically, they tightly bind to bone mineral and inhibit FPPS in osteoclasts. This results in impaired protein prenylation and function, inducing cell death of osteoclasts. However, this strong binding to bone mineral (24)—a desirable feature for a drug to treat bone resorption diseases—makes them less suited to treat solid tumors. Lipophilic bisphosphonates, on the other hand, do not bind to bone mineral and, in addition to inhibiting FPPS, they also target GGPPS. Compared with the most potent commercial bisphosphonate drug, zoledronate, the lipophilic bisphosphonate BPH-1222 is more efficient in killing tumor cells with KRAS mutations both in vitro and in vivo. This enhanced efficacy is due to a combination of factors: good FPPS and GGPPS inhibition, as well as greatly enhanced lipophilicity (clogP of 0.25 for BPH-1222 versus –3.9 for zoledronate). In addition, BPH-1222 has good pharmacokinetic properties (41, 42).

Targeting protein prenylation (including farnesylation and geranylgeranylation) has been pursued for more than 20 years, ever since researchers first found that RAS requires posttranslational prenylation for its malignancy-transforming activity. FTIs and GGTIs were developed early on to kill tumor cells in vitro; however, little success has been achieved using these compounds in animals. More interestingly, responses to these inhibitors do not always correspond to RAS mutation status. This observation strongly suggests the existence of other targets (27). Here, we show that KRAS prenylation and activity are largely inhibited by lipophilic bisphosphonate treatment and suggest that this is one of the major mechanisms of action of this class of compounds. However, other small G proteins, such as RAL, RHO, RAC, and CDC42, require exclusively geranylgeranylation, and all of these proteins have been shown to be involved in RAS-induced transformation in a context-dependent manner (15, 4345). Indeed, results from this as well as other studies all support the idea that suppression of protein geranylgeranylation is critical for bisphosphonate-mediated cytotoxicity. On the basis of this observation, dual-target (FPPS + GGPPS) inhibitors should be particularly potent in inhibiting tumor cell growth. They might also be particularly toxic in vivo, so their use in combination therapies might be of more interest. For example, it has been reported that high-dose treatment with dual prenyltransferase inhibitor (DPI; 1000 to 2000 mg/kg) led to substantial lethality (11). However, in our experiments, we have not observed any severe toxicity in animals even up to 6 weeks after treatment.

KRAS-mutant tumor cells have been shown to rely on autophagy for providing metabolic intermediates and clearing excess ROS (32, 33). In vitro, the additive effects of blocking protein prenylation and autophagy with chloroquine resulted in very potent tumor cell killing. However, in animals, the pro-proliferative effect induced by p62 accumulation and NF-κB activation became dominant, and the drug combination was ineffective. In sharp contrast, the combination of a lipophilic bisphosphonate with rapamycin was far more effective, because rapamycin not only facilitated autophagy but also inhibited the mTOR pathway that is critical for tumor cell survival. Thus, the combination of a lipophilic bisphosphonate plus rapamycin offers a promising therapeutic lead for treating KRAS-related lung cancers.

Here, we showed substantial tumor suppression by BPH-1222 plus rapamycin treatment in the lentiviral model; however, some mice developed ascites after 6 weeks of repeated intraperitoneal drug delivery, and the treatment had to be discontinued. We suspect that this was due to BPH-1222–induced topical irritation, because when the drug was given via tail vein, the damage only occurred at the injection site. Future studies will need to focus on optimizing the dose and route of drug delivery to prolong the course of treatment and achieve better efficacy.


Study design

The objective of this study was to develop combination therapies with lipophilic bisphosphonates, analogs of zoledronate, in mouse lung cancer models. Thirty bisphosphonates were screened for activity against their putative targets, farnesyl diphosphate synthase (FPPS) and geranylgeranyl diphosphate synthase (GGPPS), and for cytotoxic activity in vitro against a panel of tumor cell lines. The most effective compounds were then tested in vivo in mouse lung cancer models. As monotherapy, the bisphosphonates had only moderate activity. We then sought to determine whether efficacy could be improved by using combination therapies with chloroquine or rapamycin in vivo, in both syngeneic transplantation (subcutaneous and orthotopic) and lentiviral models. Tumor size was monitored one to two times every week during the treatment, by either palpation or in vivo luciferase imaging. At least five mice were used for each group (either control or treatment). All the treatment experiments have been independently repeated at least twice. Mice used for control and treatment were randomly grouped from a pool of model mice.

Chemical reagents

Lipophilic bisphosphonates were synthesized as described before (17, 46). Protein FTI/GGTIs (FTI-277 and GGTI-298) were purchased from Calbiochem. MEK inhibitor U0126 was purchased from Cell Signaling Technology. FOH and GGOH, ascorbic acid, Trolox, lipoic acid, Morin hydrate, and chloroquine diphosphate salt were purchased from Sigma. Simvastatin was purchased from Tokyo Chemical Industry Co. Rapamycin was purchased from Alfa Aesar.

Stable cell lines and cell survival assay

Mouse lung cancer cell lines (6#, L2 and M3L2) were derived from primary tumors of LSL-KrasG12D/+ mice infected with CA2Cre-shp53 lentiviral vector (37). L2 and M3L2 cells were derived from mouse tumors with pure C57B and pure FVB background, respectively, so that they would form syngeneic grafts in matched recipient mice. To monitor NF-κB activation in tumor grafts, L2 cells were stably infected with 5xκB-luci lentiviral vector. Human cancer cell lines (A549, A427, Panc-1, and MiaPaCa2) were purchased from the American Type Culture Collection (ATCC). MEFs were prepared from a mouse embryo with matched genetic background and immortalized by short hairpin RNAs (shRNAs) against p53 and Rb1. Cell survival after drug treatment was measured by Cell Proliferation Reagent (WST-1) from Roche or MTT Cell Proliferation Assay Kit (30-1010K) from ATCC. Dose-response curves and corresponding IC50s were fitted with GraphPad Prism.

Lentiviral vector–mediated mouse lung cancer model and syngeneic graft model

LSL-KRASG12D/+ Rosa26luc/luc mice were used for the lentiviral vector (CA2Cre-shp53) –mediated lung cancer model as described before (37). Tumor size was monitored by in vivo luciferase imaging system (IVIS 100) from Caliper Life Sciences. Syngeneic graft experiments were done by either subcutaneously transplanting 106 L2 tumor cells in the flank region of C57B mice or tail vein injection of 2 × 105 M3L2 cells into FVB mice. Subcutaneous tumor size was measured every 7 days after the transplantation. For single-drug treatment, animals were given the drug every day. For combination therapies, animals were given the drugs on alternating days, with BPH-1222 on days 1, 3, and 5, and the other drug on days 2, 4, and 6. BPH-1222 (2 mg/kg) and chloroquine (60 mg/kg) were diluted in phosphate-buffered saline (PBS), and rapamycin (2.5 mg/kg) in 75% dimethyl sulfoxide (DMSO) and 25% PBS. All drugs were given intraperitoneally in a volume of 100 μl. All mouse studies were carried out according to the protocols that were approved by the IACUC of Salk Institute.

Cellular fractionation

Cells (106) treated with drugs for 48 hours were washed with PBS and resuspended in 0.5 ml of 0.1 M tris-HCl (pH 7.5) with protease inhibitors for 15 min. Supernatant from 10,000g, 30 min, centrifugation was collected as cytosolic fraction. The pellet was resuspended in 0.5 ml of lysis buffer containing 1% Triton X-100 for 20 min and centrifuged. The supernatant was collected as membrane-bound fraction.

Histology, immunofluorescence staining, and immunoblotting analysis

Mouse lung tumor and syngeneic graft samples were fixed with 10% formalin, paraffin-embedded, and sectioned for H&E and immunofluorescence staining. Elite ABC system (Vector Labs) was applied where staining signal was weak. Immunoblotting analysis was performed according to standard protocols. For detecting HDJ2, RAP1A, and KRAS prenylation with mobility shift assay, proteins were separated with 15-cm 8% (HDJ2) and 13% SDS-PAGE (RAP1A and KRAS). Antibodies were purchased from Millipore (SPC, 1:2000), Cell Signaling Technology (phos-ERK, total ERK, phos-AKT, total AKT, phos-MEK, phos–c-RAF, cleaved caspase-3, CHOP, BiP, phos-PERK, LC3I/II, phos-p70 S6K, phos-4E-BP1, all 1:1000), Abgent (p62, clone 2C11, 1:2000), Vector Labs (Ki-67, 1:500), and Santa Cruz Biotechnology (KRAS, HRAS, RAP1A, HDJ2, ACTIN, all 1:1000).

Quantitative RT-PCR

Total RNA isolated from the treated cells was reverse-transcribed with SuperScript III system (Invitrogen) and random primers. Quantitative PCR was performed in triplicate using 7900HT Fast Real-Time PCR System with SYBR Green method (Applied Biosystems). Results were analyzed for the relative expression of mRNAs normalized against GAPDH and cyclophilin. A list of primers used for PCR is in table S5.

KRAS matrix-assisted laser desorption/ionization mass spectrum analysis

Purified protein samples were spotted onto a matrix-assisted laser desorption/ionization (MALDI) target using sinapinic acid as the matrix. Spectra (averages of 200 laser shots) were obtained on an Applied Biosystems Voyager-DE STR instrument in linear mode with delayed extraction. The accelerating voltage was set to 25,000 V, and the laser repletion rate was 20 Hz. Masses were corrected using human Activin A (MH+ave = 25,934.8) as an external standard.

FPPS and GGPPS inhibition assay and crystallization of FPPS with BPH-1222

Human FPPS and GGPPS proteins were prepared and used for enzymatic inhibition assay as described before (17). Crystallization of FPPS with BPH-1222 was carried out as reported (46). Diffraction data were collected at the Life Sciences Collaborative Access Team (LS-CAT) beamline 21-ID-Gat Argonne National Laboratory. Data were collected at 100 K with wavelength of 0.97857 Å. Data were processed with HKL-3000 and refined by using CCP4 and Coot (4749). Crystallographic figures were drawn with PyMOL ( Data collection and refinement statistics are shown in table S2. After refinement, the Ramachandran statistics showed that the percentages of the most favorable and additional allowed regions are 97.6 and 2.4%, respectively.

Pharmacokinetics test

Pharmacokinetic studies were performed with three female Sprague-Dawley rats (230 to 240 g body weight). Plasma concentrations were measured at 1 min, 5 min, 15 min, 30 min, 1 hour, 2 hours, 4 hours, 8 hours, 24 hours, and 48 hours after a single intravenous injection of BPH-1222 at 5 mg/kg. Blood (0.3 ml) was taken via retro-orbital bleeding each time. Data were analyzed with DAS2.0 software.

Statistical analysis

Statistical analyses were performed with GraphPad Prism software. Statistical significance of the differences in tumor number, Ki-67, and cleaved caspase-3–positive staining was evaluated with Student’s unpaired two-tailed t test. The Kaplan-Meier curves were analyzed by log-rank test. P values less than 0.05 were considered statistically significant.

Supplementary Materials

Fig. S1. Structural diagram of bisphosphonates.

Fig. S2. FPPS/GGPPS activity and cell growth inhibited by bisphosphonates.

Fig. S3. ER stress induced by blocking protein prenylation.

Fig. S4. Comparison of drug combinations.

Fig. S5. Combination therapy in KRAS-shp53 lentiviral model.

Fig. S6. Effect of ERK inhibition.

Fig. S7. Combination therapy with zoledronate and rapamycin.

Table S1. FPPS/GGPPS activity and cell growth inhibited by bisphosphonates.

Table S2. Data collection and refinement statistics.

Table S3. Summary of protein prenylation inhibition by various compounds.

Table S4. Summary of KRAS molecular weight (M.W.) measured by mass spectrometry.

Table S5. PCR primer sets used in the study.

References and Notes

  1. Acknowledgments: We thank R. Alvarez Rodriguez, J. Moscat, M. T. Diaz-Meco, and J. P. Noel for helpful discussions, and L. Li for technical assistance. Funding: I.M.V. is an American Cancer Society Professor of Molecular Biology and holds the Irwin and Joan Jacobs Chair in Exemplary Life Science. This work was supported in part by grants from the NIH (R01-AI048034 from the National Institute of Allergy and Infectious Diseases to I.M.V. and R01CA158191 from the National Cancer Institute to E.O.), Ipsen Biomeasure, the H.N. and Frances C. Berger Foundation, and the Leona M. and Harry B. Helmsley Charitable Trust grant #2012-PG-MED002 (I.M.V.); a Harriet A. Harlin Professorship (E.O.); and the 1000 Young Talents Program and start-up funds from Tsinghua University (Y.Z.). Use of the Advanced Photon Source, an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science by Argonne National Laboratory, was supported by the U.S. DOE under contract no. DE-AC02-06CH11357. Use of the LS-CAT Sector 21 was supported by the Michigan Economic Development Corp. and the Michigan Technology Tri-Corridor (grant 085P1000817). Y. Xie was supported by the post-doctoral fellowship from Tsinghua-Peking Center for Life Sciences. Author contributions: Y. Xia and Y.Z. designed the experiments. Y. Xia performed the experiments with the help of Y.Z., Y.-L.L., Y. Xie, W.Z., F.G., X.Z. (compound synthesis, enzymatic assay, crystallization analysis, pharmacokinetics test, and cell survival assay), S.S. (animal works), N.Y. (quantitative RT-PCR), and W.F. and W.L. (mass spectrometry). I.M.V., E.O., and Y.Z. provided funding and supervised the project. Y. Xia, E.O., and I.M.V. wrote the paper. Competing interests: A provisional patent for “Lipophilic bisphosphonates and methods of use” has been filed. Data and materials availability: Coordinates and structure factors have been deposited in the Protein Data Bank with access code 4N1Z for human FPPS in complex with BPH-1222.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article