Dual Farnesyl and Geranylgeranyl Transferase Inhibitor Thwarts Mutant KRAS-Driven Patient-Derived Pancreatic Tumors

Members of the RAS family of GTPases are signal transducers that regulate many biological processes, including cell-cycle progression, cell survival, and differentiation (1–3). When mutated, they become persistently activated and promote several oncogenic events, including uncontrolled proliferation, resistance to apoptosis, sustained angiogenesis, invasion, and metastasis (1). Approximately 20%–30% of all solid tumors have mutations in the KRAS, NRAS, or HRAS isoforms (1–3). Among these three RAS isoforms, oncogenic mutant KRAS is associated with currently unsolved lethality of several solid tumor neoplasms, particularly where it is prevalent such as in 90% of pancreatic, 45% of colorectal, and 35% of lung carcinomas (4–7). For example, patients whose tumors harbor mutant KRAS respond poorly to chemotherapy and targeted therapies, have poor prognosis, and have more aggressive tumors (4–7). The U. S. NCI identified the targeting of KRAS as a high priority and has implemented several major initiatives with the ultimate goal of discovering therapies that specifically target patients whose tumors harbor mutant KRAS. However, targeting KRAS directly is challenging (8–12). Although recent efforts have led to encouraging results (10–12), no anticancer drugs targeting mutant KRAS are available in the clinic. Therefore, indirect and alternative approaches to targeting mutant KRAS human tumors are greatly needed.

One alternative approach is to inhibit KRAS prenylation, a lipid posttranslational modification that is required for the proper cellular localization and cancer-causing activity of KRAS (13). The two enzymes involved in KRAS prenylation are farnesyltransferase (FT) and geranylgeranyltransferase 1 (GGT-1), which transfer farnesyl and geranylgeranyl lipids to the cysteine sulfhydryl of proteins terminating at their carboxyl terminal with CAAX tetrapeptide sequences (where C, cysteine; A, aliphatic residues; and X, any amino acid). Farnesyltransferase prefers methionine or serine at the X position, whereas GGT-1 prefers a leucine or isoleucine at the X position (13). The fact that KRAS and other GTPases require prenylation for their ability to induce malignant transformation has led to the development of FT inhibitors (FTI) and GGT-1 inhibitors (GGTI) as potential anticancer agents (13). All RAS isoforms are exclusively farnesylated. However, when cells are treated with FTIs, KRAS and NRAS, but not HRAS, become geranylgeranylated by GGT-1, and the resulting geranylgeranylated KRAS and NRAS are fully functional (13). Blocking FT is sufficient to inhibit HRAS, but blocking both FT and GGT-1 is required to inhibit the prenylation and function of KRAS and NRAS (13). Therefore, a single molecule with dual inhibitory activities has the potential to abrogate mutant KRAS- and mutant NRAS–driven human cancers.

Pancreatic cancer is the fourth leading cause of cancer-related deaths in the United States, with less than 6% of patients surviving 5 years following diagnosis (14, 15). The persistently poor survival rate of patients with pancreatic cancer is due to resistance to current treatments, with mutant KRAS (>90% prevalence) as a major contributor (14). In this study, we describe the development of FGTI-2734, a potent CAAX tetrapeptide mimetic dual FT and GGT-1 inhibitor that prevented KRAS membrane localization and inhibited the in vivo growth of several mutant KRAS–driven human cancer cells and growth of patient-derived xenografts (PDX) from mutant KRAS tumors from four patients with pancreatic cancer. Using three-dimensional cocultures with chemoresistance-promoting pancreatic stellate cells, we found that FGTI-2734 inhibited the growth and viability of primary and metastatic mutant KRAS tumor cells derived from patients with pancreatic cancer. Importantly, FGTI-2734 suppressed major cancer-causing pathways (i.e., PI3K/AKT/mTOR and cMYC) while upregulating the tumor suppressor p53 and inducing apoptosis in PDXs in vivo.

Human lung cancer cell lines (A549, H460, and Calu6), colon cancer cell line (DLD1), and pancreatic cancer cell lines (MiaPaCa2 and L3.6pl) were obtained from the ATCC. Human lung cancer cell lines H522, H661, H322, and H2126 and normal lung fibroblast cell lines WI-38 and MRC-5 were kindly provided by Dr. Eric Haura (Moffitt Cancer Center, Tampa, FL). NIH3T3 cells stably transfected with constitutively active H-Ras61L (H-Ras/NIH3T3), K-RasGV12 (K-Ras/NIH3T3), N-Ras (N-Ras/NIH3T3), and empty vector pcDNA3 (pcDNA3/NIH3T3) were obtained from Dr. Channing Der (University of North Carolina, Chapel Hill, NC). Normal fibroblast cells WI-38 and MRC-5 were cultured in minimum essential medium, and all other cell lines were cultured in DMEM or RPMI1640 medium. All media were supplemented with 10% heat-inactivated FBS, 10 U/mL penicillin, and 10 μg/mL streptomycin. All cell lines were Mycoplasma-free, monitored regularly with HEK-blue2 cells and Mycoplasma Detection Kit from InvivoGen (catalog No. rep-pt1). All cell lines were authenticated by University of Arizona Genetics Core. FGTI-2734, FTI-2148, and GGTI-2418 were synthesized in-house as described previously (16–18). The FGTI-2734 and FTI-2148 were derivatized as mono-mesylate salt and bis-trifluoroacetate salts, respectively.

Human pancreatic cancer cell lines were derived from eight patient pancreatic tumors using institutional review board (IRB)-approved (MDA Cancer Center protocol LAB07-0854) and previously described methods (19, 20). The investigators obtained written consent from the subjects. The research was conducted according to Declaration of Helsinki. The cells were plated at 3,000/well in triplicate in a 96-well flat-bottom plate in RPMI medium with 10% FBS. Cells were subsequently treated for 72 hours with DMSO or FGTI-2734 at 3, 10, 30, and 100 μmol/L. For three-dimensional cultures, 15 mL of cold Matrigel growth factor–reduced, phenol red–free solution was added to wells, spread evenly, and allowed to incubate at 37°C for 30 minutes to solidify. Pancreatic cancer cells (3,000/well) were resuspended in RPMI-2% Matrigel medium-10% FBS overlaid on the solidified Matrigel and cultured for 24 hours before treatment. For three-dimensional coculture with pancreatic stellate cells, human pancreatic stellate cells were grown until confluent in RPMI with 10% FBS. Matrigel was added as in three-dimensional culture and pancreatic cancer cells and human pancreatic stellate cells were resuspended at 3,000 cells/well in RPMI-2% Matrigel medium supplemented with 10% FBS, overlaid at 1:1 ratio, and cultured for 24 hours before treatment.

GGT-1 and FT activities from 60,000 × g supernatants of human Burkitt lymphoma (Daudi) cells (obtained from ATCC) were assayed as described previously (17). We conducted tests to determine the ability of peptidomimetics to inhibit the transfer of [³H]geranylgeranyl and [³H]farnesyl from [³H]geranylgeranylpyrophosphate (PerkinElmer) and from [³H]farnesylpyrophosphate (Amersham Biosciences) to recombinant HRAS-CVLL and HRAS-CVLS, respectively.

To prepare whole-cell lysates, cells were trypsinized, washed twice with cold PBS, and lysed in mammalian protein extraction reagent (product No. 78501, Thermo Fisher Scientific) supplemented with protease inhibitor cocktail (product No. A32953, Thermo Fisher Scientific), consisting of 2 mmol/L phenylmethylsulfonyl fluoride, 2 mmol/L Na₃VO₄, and 6.4 mg/mL p-nitrophenylphosphate. Tumor tissue samples were lysed in tissue protein extraction reagent (product No. 78510, Thermo Fisher Scientific) with above supplements. The automatic hand-operated OMNI-TIP Homogenizer (Omni International, Inc.) was used to homogenize the tumor tissues. Lysates from whole cells and tumor homogenates were cleared by centrifugation at 12,000 × g for 15 minutes, and the supernatants were collected as whole-cell extracts. Protein concentrations were determined using the BCA protein assay kit. Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes (17), which were then blotted with antibodies specific for unprenylated RAP1A (catalog No. sc-1482), HRAS (catalog No. sc-520), NRAS (catalog No. sc-519), IGF-1Rβ (3G5C1; catalog No. sc-81167), AKT1/2 (N-19; catalog No. sc-1619), p53 (DO-1; catalog No. sc-126), and cMYC (pE10; catalog No. sc-40; all from Santa Cruz Biotechnology). Other antibodies included phospho-AKT (S473; catalog No. 9271S), phospho-ERK (catalog No. 9101L), ERK (catalog No. 9102L), cleaved-CASP-3 (catalog No. 9664L), cleaved-PARP (catalog No. 5625S), phospho-S6 ribosomal protein (catalog No. 2215S), and S6 ribosomal protein (catalog No. 2217S) from Cell Signaling Technology; RHOGDI/D4-GDI (catalog No. 66586E) from BD Pharmingen; anti-c-KRAS (Ab-1; catalog No. OP24) from Calbiochem; HDJ-2 from Lab Vision Corporation; and anti-β-ACTIN (catalog no A5441-.2ML) and vinculin (catalog No. V9131-.2ML) from Sigma-Aldrich.

Membrane and cytosolic fractions were prepared according to van der Hoeven and colleagues (21). Briefly, cells were grown in 10-cm dishes, treated for 72 hours with drugs (FGTI, FTI, and GGTI) or vehicle control (DMSO), and washed twice with cold PBS. Cells were then harvested by scraping into fractionation buffer [10 mmol/L Tris (pH 7.5), 25 mmol/L NaF, 5 mmol/L MgCl₂, 1 mmol/L EGTA, 1 mmol/L dithiothreitol, 100 μmol/L Na₃VO₄ plus protease inhibitors], incubated on ice for 30 minutes, and lysed by passing through a 23-gauge needle 26 times. Lysed cells were then centrifuged at 1,000 × g for 5 minutes at 4°C to remove nuclear fractions and unbroken cells, with supernatants then centrifuged at 100,000 × g at 4°C for 30 minutes in a refrigerated ultracentrifuge (Optima Max Ultracentrifuge, Beckman Coulter) to sediment plasma membranes. The cytosolic supernatant (S100) was separated from the membrane pellet. The crude membrane pellet was washed once (without disturbing the pellet) with the above fractionation buffer and then resuspended in the above buffer by sonication (3 × for 30 seconds each) and used as membrane fraction (P100). Protein concentrations of both fractions were determined using BCA protein assay kit, and 15–20 μg of both fractionation proteins underwent SDS-PAGE and Western blotting with the specific antibodies.

The lentiCRISPRv2 plasmid was a gift from Feng Zhang (Addgene plasmid no 52961). Guide RNAs were designed using the CRISPR design tool (http://crispr.mit.edu). The nontarget scramble control (SC) and sgKRAS vectors were generated using Feng Zhang's protocol, which is available on Addgene's website (www.addgene.org). The guide RNA sequences of SC and sgKRAS are 5′-GCACTACCAGAGCTAACTCA-3′ and 5′-GCAATGAGGGACCAGTACATG-3′, respectively. For lentivirus production, the lentiviral vector (10 μg), pMD.2G (5 μg), and pspPax2 (5 μg) vectors were cotransfected into 293T cells using polyethylenimine (25 kDa, 1 μg/μL). At 48 hours, the virus supernatants were collected and concentrated at 1:100 ratio using the Lenti-X concentrator (Clontech, 631231). Cells were transduced overnight using 10 μL of concentrated virus with 8 μg/mL of polybrene and then cultured for an additional 72 hours before harvest and Western blot analyses.

The immunofluorescence vectors pLEX-GFP, pLEX-GFP-KRASG12V-CVIM, and pLEX-GFP-KRASG12V-SVIM were constructed in a modified pLEX-FLAG vector. A GFP sequence was inserted between NotI and BamHI sites, and KRAS sequences were inserted after GFP and between BamHI and AgeI sites. Lentivirus production was carried out as described above. Cells (0.8 × 10⁵) were plated onto glass coverslips in 12-well plates and left overnight, after which the cells were infected with 200 × 10⁸ copies of GFP-KRASG12V with wild-type CAAX box (GFP-KRASG12V-CVIM), GFP-tagged mutated prenylation cysteine 185 to the non-prenylable serine 185 (GFP-KRASG12V-SVIM), and corresponding empty vector (GFP-EV). After 24 hours of infection, cells were treated with vehicle control and different concentrations of FGTI-2734, FTI-2148, and GGTI-2418 for 48 hours. Cells were then washed once with warm Dulbecco's PBS (catalog No. 14190-144, GIBCO Life Technologies, Invitrogen) and fixed with prewarmed 4% paraformaldehyde solution in PBS (catalog No. sc-281692, Santa Cruz Biotechnology) for 5 minutes at room temperature. After fixation, cells were washed 3 times with Dulbecco's PBS, with coverslips mounted using Vectashield with 4′,6-diamidino-2-phenylindole (H-1200; Vector Labs). Localization of GFP was observed using a confocal microscope (40× magnification).

Cell viability was determined by the CellTiter-Glo Luminescent Cell Viability Assay (Promega) according to the manufacturer's protocol. Briefly, cells (10³ cells/well) were seeded in 384-well plates, allowed to adhere overnight, and treated with vehicle (DMSO) or drugs for 72 hours, after which they were processed for viability using CellTiter-Glo reagent. Data were normalized to percentage of control, and IC₅₀ values were determined. Each condition was performed in replicates of six wells. For the patient-derived cells, IC₅₀ was calculated using GraphPad Prism 7.02 software. Live cell imaging was carried out with the IncuCyte ZOOM. Wells were scanned at day 0 and at 72 hours of treatment. Cell number was determined using IncuCyte software, and day 0 scan was used as control.

Male SCID-bg mice (Charles River Laboratories) were maintained and treated in accordance with our Institutional Animal Care and Use Committee procedures and guidelines. Exponentially growing MiaPaCa2, L3.6pl, Calu6, A549, DLD1, and H460 cells were harvested via trypsinization, pelleted at 300 × g for 5 minutes, resuspended in sterile Dulbecco's PBS (Invitrogen) at 10 × 10⁶ cells (L3.6pl and A549), 5 × 10⁶ cells (MiaPaCa2, Calu6, and DLD1), and 3 × 10⁶ cells (H460) per 100 μL, and injected into the flank of each mouse. The tumor xenografts were monitored 3 days/week with an electronic caliper. Tumor volume was calculated using the formula V = (L²W)/2, where L is length and W is width, with width defined as the largest measurement and length as the smallest measurement. When the tumors reached approximately 200 mm³, the animals were randomized to the following treatment groups: (i) vehicle control group [40% 2-hydroxypropyl-β-cyclodextrin (Sigma) w/v in sterile water], administered once daily via 100-μL i.p. injections, and (ii) FGTI-2734 group, with FGTI-2734 administered at 100 mg/kg/daily i.p. in 40% 2-hydroxypropyl-β-cyclodextrin. Treatments were continued for 18–25 days. For both groups, animals showed no evidence of gross toxicity as measured by weight loss.

To determine whether pharmacologic inhibition of KRAS can inhibit the growth of PDXs, we obtained fresh tumor biopsies from patients with pancreatic cancer (University of Florida, IRB protocol 201600873) with KRAS mutation. The investigators obtained written consent from the subjects. The research was conducted according to International Ethical Guidelines for Biomedical Research Involving Human Subjects.

Specifically, fresh 2-mm tumor pieces were obtained from pancreatic resection and transported on ice to the animal surgery suite for subcutaneous implantation into NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. The mice were housed, maintained, and treated in accordance with our Institutional Animal Care and Use Committee procedures and guidelines. After 1-cm incisions were made on right flanks of anesthetized NSG mice, the subcutaneous layer underwent blunt dissection. A viable tumor piece was then placed, and the skin was closed with surgical clips (generation 1). Once engrafted and tumors reached 1.5 cm in diameter, tumors were divided evenly into 2-mm pieces and reimplanted into NSG mice as above (generation 2). Generation 3 was generated similarly (details as described in Delitto and colleagues; ref. 22). When the tumors of generation two or three mice reached approximately 100–200 mm³, the animals were randomized into vehicle and FGTI-2734 treatment groups as described above for the xenograft models.

MiaPaCa2 cells were treated with vehicle and FGTI-2734 (30 μmol/L) or infected with lentiviral plex-Flag-GFP (EV), plex-Flag-GFP-KRAS (C185), and plex-Flag-GFP-KRAS (S185). Forty-eight hours after treatment or infection, cells were harvested with Cell Lysis Buffer [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 0.1%Triton X-100, 2.5 mmol/L sodium pyrophosphate, 1 mmol/L β-glycerophosphate,1 mmol/L Na₃VO₄, 1 μg/mL Leupeptin, and 1 mmol/L PMSF]. The cell lysates (1.5∼2 mg, 1,000 μL each) were used for immunoprecipitation with RAF-1 antibody (Santa Cruz Biotechnology, No. sc-7267) or anti-Flag M2 affinity gel (Sigma, No. A2220) for 3 hours at 4°C. The protein bound beads were washed twice with cell lysis buffer and twice with kinase buffer [25 mmol/L Tris (pH 7.5), 5 mmol/L β-Glycerolphosphate, 10 mmol/L MgCl2, 1 mmol/L DTT, and 0.1 mmol/L Na3VO4]. The protein bound beads were incubated with 40-μL kinase buffer supplemented with 1 μg of inactive MEK1 (Invitrogen, No. P3093) and 200 μmol/L ATP (Invitrogen, No. AM8110) for 30 minutes at room temperature. The phosphorylated MEK1 was detected by immunoblotting using anti-phospho-MEK1 (Ser217/221) antibody (Cell Signaling Technology, No. 9121), and total MEK1 detected with anti-MEK1 antibody (Cell Signaling Technology, No. 9122). For the other immunoblotting, the blots were incubated with antibodies of RAF-1, Flag (Sigma, No. F3165), anti-c-K-RAS (Millipore, No. OP24), and KSR-1(Santa Cruz Biotechnology, No. sc-515924), respectively.

All data generated or analyzed during this study are included in this article (and its Supplementary Data).

As shown in Fig. 1A, FTI-2148 inhibited FT and GGT-1 activities in vitro with IC₅₀ values of 1.4 nmol/L and 1,700 nmol/L, respectively, showing a 1,214-fold selectivity for FT over GGT-1. GGTI-2418 inhibited GGT-1 and FT activities in vitro with IC₅₀ values of 9.4 nmol/L and 58,000 nmol/L respectively, demonstrating a 6,170-fold selectivity for GGT-1 over FT. In contrast, FGTI-2734 inhibited FT and GGT-1 activities with similar IC₅₀ values of 250 and 520 nmol/L, respectively. This twofold difference was not statistically significant (Fig. 1A).

We next evaluated the effects of these three inhibitors on RAS membrane association in HRAS-, NRAS-, and KRAS-transformed murine NIH3T3 cells and in human cancer cells harboring mutant KRAS. Western blots in Fig. 1B show that FTI-2148 inhibited the farnesylation of the exclusively farnesylated protein HDJ2 in all three RAS-transformed NIH3T3 cells (non-prenylated proteins migrate more slowly than prenylated proteins in SDS-PAGE gels; refs. 13, 16, 17). In contrast, FTI-2148 did not inhibit the geranylgeranylation of the exclusively geranylgeranylated RAP1A (as documented by an antibody that recognizes only non-geranylgeranylated RAP1A; ref. 13, 16, 17), confirming in intact cells the high selectivity in vitro of FTI-2148 for FT over GGT-1 (Fig. 1A). GGTI-2418 inhibited the geranylgeranylation of RAP1A but not the farnesylation of HDJ2 (Fig. 1B), confirming its in vitro selectivity for GGT-1 over FT (Fig. 1A). In contrast, FGTI-2734 inhibited the prenylation of both HDJ2 and RAP1A (Fig. 1B), confirming its in vitro dual FT and GGT-1 inhibitory activity (Fig. 1A). Importantly, FGTI-2734, but not FTI-2148 and GGTI-2418, inhibited the prenylation of KRAS and NRAS (Fig. 1B). As expected, the prenylation of the exclusively farnesylated HRAS was inhibited by FGTI-2734 and FTI-2148 but not by GGTI-2418 (Fig. 1A).

To determine whether FGTI-2734 can inhibit KRAS and NRAS membrane association in human cancer cells, human pancreatic (MiaPaCa2) and lung (H460) cancer cells were treated with the inhibitors and processed for membrane/cytosol cellular fractionations as described in Materials and Methods. As shown in Fig. 1C, FTI-2148 inhibited the membrane association of the farnesylated protein HRAS, leading to its accumulation in the cytosol, but did not affect membrane association and cytosolic accumulation of the geranylgeranylated protein RAP1A. Although GGTI-2418 induced the cytosolic accumulation of non-geranylgeranylated RAP1A, it did not affect membrane association and cytosolic accumulation of HRAS (Fig. 1C). In contrast, FGTI-2734 induced the cytosolic accumulation of both HRAS and RAP1A. Importantly, FGTI-2734, but not FTI-2148 and GGTI-2418, inhibited the membrane association of KRAS and NRAS, which led to the accumulation of these proteins in the cytosol (Fig. 1C, bottom).

To determine the effects of the prenylation inhibitors on KRAS cellular localization by immunofluorescence, lentiviruses coding for a GFP-tagged mutant KRAS-G12V with a wild-type C-terminal CAAX box (GFP-KRASG12V-CVIM) and a corresponding empty vector construct (GFP-EV) were generated. We also mutated the cysteine 185 of the CAAX box to serine 185, hence generating a lentivirus coding for GFP-tagged mutant KRASG12V with a mutant CAAX that cannot be prenylated and therefore cannot associate with the membrane (GFP-KRASG12V-SVIM). MiaPaCa2 and H460 cells were infected with the lentiviruses, treated with inhibitors, and processed 48 hours after drug treatment for immunofluorescence microscopy as described in Materials and Methods. As shown in Fig. 2A, cells infected with GFP-EV displayed a diffuse pattern of GFP fluorescence. In contrast, cells infected with GFP-KRASG12V-CVIM showed a distinct punctate pattern of primarily plasma membrane localization but also endomembrane, consistent with previous reports (22). As expected, cells infected with GFP-KRASG12V-SVIM showed a diffuse pattern of GFP fluorescence, confirming the lack of KRAS prenylation and membrane localization (Fig. 2A). FGTI-2734 treatment prevented GFP-KRASG12V-CVIM localization to the membrane, resulting in a diffuse pattern similar to the one observed in the cells infected with the GFP-KRASG12V-SVIM. Importantly, cells treated with FTI-2148 and GGTI-2418 did not prevent GFP-KRASG12V-CVIM membrane localization (Fig. 2A). The ability of FGTI-2734 to prevent membrane localization was also demonstrated in other human cancer cell lines, including Calu6, A549, and DLD1 cells (Fig. 2A).

Figures 1C and 2A demonstrated that FGTI-2734 prevents KRAS membrane association and induces accumulation of KRAS in the cytoplasm. We next evaluated the effects of shifting mutant KRAS from the membrane to the cytoplasm by determining whether it is still able to bind one of its major effectors RAF-1, and whether RAF-1 kinase activity is affected. To this end, we treated MiaPaCa2 cells with FGTI-2734 as described above, and processed the cells for RAF-1 immunoprecipitation, followed by immunoblotting of KRAS and KSR, a scaffolding protein that binds RAF-1 and mediates RAS activation of RAF-1. Figure 2B shows that in the absence or presence of FGTI-2734, RAF-1 was able to bind KRAS, suggesting that both membrane and cytosolic KRAS are able to bind RAF-1. However, FGTI-2734 inhibited the binding of RAF-1 to KSR, suggesting that the binding of KSR to RAF-1 requires the membrane environment. To determine the effects of FGTI-2734 on RAF-1 kinase activity, RAF-1 was immunoprecipitated and its kinase activity was assayed in the immunoprecipitates using recombinant inactive MEK1 as a substrate and Western blotting with an antibody to phosphorylated Serines 217 and 221, two RAF-1 phosphorylation sites on MEK1. Figure 2B shows that FGTI-2734 inhibited RAF-1 kinase activity, suggesting that although cytosolic KRAS binds RAF-1, it requires KSR and the membrane environment to fully activate RAF-1. We next determined whether RAF-1 binds KRASG12V-SVIM, the CAAX mutant that cannot be prenylated and that does not bind to the membrane (see Fig. 2A). To this end, we infected MiaPaCa2 cells with lentiviral FLAG-tagged KRASG12V-CVIM and its nonmembrane bound mutant FLAG-tagged KRASG12V-SVIM, and processed the cells for coimmunoprecipitation and RAF-1 kinase assays. Figure 2C shows that the KRASG12V-SVIM mutant is still able to bind RAF-1 and that the RAF-1 kinase activity in this cytosolic complex is inhibited, confirming the FGTI-2734 results of Fig. 2B.

We next determined whether blocking KRAS prenylation with FGTI-2734 can induce apoptosis in human cancer cells that harbor mutant KRAS and whether human cancer cells that depend on mutant KRAS for survival are more sensitive to FGTI-2734. To this end, we first constructed KRAS and scrambled guide RNAs (gRNA) in LentiCRISPR vectors, and determined whether the viability of six human cancer cell lines depends on KRAS by CRISPR/Cas9-targeted knockout of KRAS. As shown in Fig. 3, KRAS gRNA led to KRAS knockout in all cell lines evaluated, including MiaPaCa2 and L3.6pl (pancreatic), Calu6, A549, and H460 (lung), and DLD1 (colon). Although all six cell lines harbor mutant KRAS, KRAS knockout only resulted in apoptosis (as determined by CASPASE-3 and PARP cleavage) in MiaPaCa2, L3.6pl, and Calu6 cells (Fig. 3A) but not in A549, H460, and DLD1 cells (Fig. 3B). Therefore, only MiaPaCa2, L3.6pl, and Calu6 cells depended on KRAS for survival, consistent with previously published studies by us (23) and others (24, 25). Treatment with FGTI-2734 inhibited the farnesylation of HDJ2 and the geranylgeranylation of RAP1A in all six cell lines (Fig. 3). However, FGTI-2734 treatment only induced CASPASE-3 and PARP cleavage in MiaPaCa2, L3.6pl, and Calu6 cells (Fig. 3A) but not in A549, H460, and DLD1 cells (Fig. 3B), indicating that FGTI-2734 induces apoptosis only in the three cancer cell lines that are dependent on mutant KRAS for viability. Furthermore, FGTI-2734 did not induce apoptosis in wild-type RAS human cancer cells, including H2126, H522, H661, and H322 lung cancer cells (Supplementary Fig. S1). Finally, FGTI-2734 did not induce apoptosis in normal lung fibroblasts, as shown in the WI-38 and MRC-5 cell lines (Supplementary Fig. S1).

As detailed in Materials and Methods, MiaPaCa2, L3.6pl, Calu6, A549, H460, and DLD1 were subcutaneously implanted in mice; after tumors reached an average size of approximately 200 mm³, mice were randomized and treated daily with vehicle or 100 mg/kg body weight FGTI-2734 i.p. As shown in Fig. 4, growth of mutant KRAS–dependent MiaPaCa2 tumors from mice treated with FGTI-2734 was significantly inhibited as early as 5 days after initiation of treatment and continued to be significantly inhibited for the rest of the treatment period. Mice displayed no gross toxicity (no effects on mouse weight, food intake, or activity). Compared with MiaPaCa2 tumors from vehicle control–treated mice, which grew from an average of 224 ± 50 to 1,015 ± 235 mm³, tumor growth in FGTI-2734–treated mice was nearly suppressed, growing from 218 ± 22 to only 302 ± 46 mm³ (Fig. 4, top left). Similar results were obtained with the other two mutant KRAS–dependent tumors (L3.6pl and Calu6; Fig. 4, middle and bottom left). For the mutant KRAS–independent tumors (A549, H460, and DLD-1), FGTI-2734 had no effect on growth (Fig. 4, middle). Thus, although all six tumors harbor mutant KRAS, FGTI-2734 only inhibited tumor growth in mutant KRAS–dependent tumors but not in mutant KRAS–independent tumors.

For the rest of the article, we focus on evaluating FGTI-2734 in clinically relevant models using human tumors harvested from patients with pancreatic cancer, as these tumors are known to be driven by mutant KRAS. To this end, we examined the effects of FGTI-2734 on the growth of PDX in mice from four patients with pancreatic cancer. Patient 1: T3N1 ductal adenocarcinoma with G12V-mutant KRAS, moderately differentiated, 2/17 LN, +LV/PN invasion, 7.8 cm; was refractory to adjuvant gemcitabine and 5-FU. Patient 2: T3N1 ductal adenocarcinoma with G12D-mutant KRAS, well differentiated, 4/11 LN, +LV, PN invasion, 3.0 cm; had received treatment with adjuvant gemcitabine. Patient 3: T3N0 ductal adenocarcinoma with G12D mutant KRAS, moderately differentiated, 0/10 LN, 4.5 cm, +LV/PN invasion, +large vessel invasion; had been treated with adjuvant gemcitabine and radiotherapy. Patient 4: T3N1 ductal adenocarcinoma with G12V-mutant KRAS, well differentiated, 1/14 LN, +LV/PN invasion, unable to identify grossly; had been treated with adjuvant gemcitabine. Freshly resected tumors were subcutaneously implanted in NSG mice as described previously (26) and randomized into vehicle control (four to six mice) or FGTI-2734 (five to seven mice) groups for each of the PDXs. Mice implanted with PDXs from patients 1, 2, and 3 were treated with 100 mg/kg body weight/day FGTI-2734. PDX from patient 4 and a separate PDX from patient 3 were treated with 50 mg/kg body weight/day FGTI-2734. Throughout treatment, FGTI-2734 at 50 or 100 mg/kg body weight significantly inhibited tumor growth of all four PDXs (Fig. 4, right; Supplementary Fig. S2). By the last day of vehicle treatment, PDXs from patients 1, 2, and 3 showed average growth of 253%, 640%, and 411%, respectively. In contrast, FGTI-2734 treatment resulted in only 92%, 236%, and 136% growth, respectively (Fig. 4, right). Differences in tumor growth between vehicle- and FGTI-2734–treated mice were statistically significant (P < 0.05) starting at day 2 (patients 1 and 2) and day 3 (patient 3; Fig. 4, right). At 50 mg/kg body weight/day, FGTI-2734 also significantly inhibited patient 3 and 4 PDX growth, although less effectively than at 100 mg/kg body weight/day (Supplementary Fig. S2).

We next investigated the effects of FGTI-2734 in vivo on the canonical RAS signaling pathways and other tumor survival pathways in patient-derived tumors. To this end, freshly resected tumors from patient 2 were subcutaneously implanted in 13 NSG mice, which were then randomized for daily 21-day treatment with vehicle (six mice) or 100 mg/kg body weight FGTI-2734 (seven mice). Two hours after the last treatment, tumors were harvested and lysates were processed for Western blotting. As shown in Fig. 5A, FGTI-2734 inhibited the farnesylation of HDJ2 and the geranylgeranylation of RAP1A. Furthermore, FGTI-2734 suppressed P-AKT and P-S6 levels, suggesting that FGTI-2734 significantly affected the PI3K/AKT/mTOR signaling pathway. Data quantification showed that, on average, P-AKT/AKT and P-S6/S6 levels were inhibited by 75% ± 4% (P = 0.00005) and 82% ± 15% (P = 0.003), respectively (Fig. 5B). In contrast, FGTI-2734 minimally affected the other Ras canonical Raf/Mek/Erk pathway. Indeed, P-Erk1-2/Erk1-2 was only inhibited by 18% ± 6% (P = 0.04; Fig. 5B). FGTI-2734 suppressed the levels of the oncogene cMYC while upregulating the levels of the proapoptotic tumor suppressor p53 and in parallel induced CASPASE 3 activation (Fig. 5A). FGTI-2734 inhibited cMYC/β-ACTIN levels by 71% ± 4% (P = 0.0006; Fig. 5B) and induced p53/β-ACTIN and CASPASE 3/β-ACTIN levels by 2.9 ± 0.6-fold (P = 0.027) and 1.71 ± 0.04-fold (P = 0.001), respectively (Fig. 5C).

Using previously described methods (19, 27), we generated low-passage (<20) human pancreatic cancer cells to evaluate the lethality of FGTI-2734 in standard and three-dimensional culture systems. These lines were derived from primary and metastatic human tumors from eight patients with pancreatic cancer, in which previously performed sequencing (28) identified point mutations in codons 12 and 13 in all of the cell lines (Table 1). When the eight patient-derived tumor cell lines were treated with FGTI-2734 in two-dimensional standard conditions, we observed a dose-dependent inhibition of viability (Fig. 6A and B), with IC₅₀ values ranging between 6 and 28 μmol/L (Table 1). Importantly, FGTI-2734 was more potent, with IC₅₀ values between 5 and 9 μmol/L, when the patient-derived tumor cells were cultured in three-dimensional conditions alone or when cocultured with the chemoresistance-promoting human pancreatic stellate cells harvested from pancreatic adenocarcinoma cells (Fig. 6C and D).

Patients with mutant KRAS tumors suffer from poor prognosis, increased tumor aggressiveness, and metastasis, and are less likely to respond to chemotherapy and targeted therapies (4–7, 14, 15). Most affected are patients with pancreatic adenocarcinoma where the mutant KRAS prevalence is 90% (14). Although recent studies have reported on small molecules that are capable of targeting KRAS (3, 10–12), so far no approved KRAS therapies are available in the clinic (3, 8–12). Attempts to target these cancers by blocking KRAS farnesylation have been unsuccessful in clinic, at least in part due to geranylgeranylation of KRAS by GGT-1 under FTI pressure (13). Here, we designed a dual FT and GGT1 inhibitor, FGTI-2734, which can prevent, unlike the selective FTI-2148 and GGTI-2418 inhibitors, membrane localization of KRAS, hence solving this major KRAS resistance problem. Our pharmacologic and genetic approaches show that cytosolic KRAS is still able to bind one of its major effectors RAF-1 but is not able to activate it, possibly due to inhibition of binding to the scaffolding protein KSR (see Fig. 2B), which plays a major role in RAS activation of Raf (29). The observation that cytosolic KRAS is able to bind but not to activate RAF-1 is consistent with our earlier report with the exclusively farnesylated HRAS, where we showed that treatment of mutant HRAS expressing NIH3T3 cells with the selective FTI, FTI-277, resulted in the accumulation of inactive cytoplasmic HRAS–Raf complexes (30).

An important finding of our study is that, although FGTI-2734 inhibited KRAS membrane association in all six human cancer cell lines that harbor mutant KRAS, it only induced apoptosis and inhibited tumor growth in mice with tumors in which CRISPR/Cas9/gRNA-targeted knockout of KRAS induced apoptosis (MiaPaCa2, Calu6, and L3.6pl) but not in cells where KRAS knockout did not induce apoptosis (A549, H460, and DLD1). The demonstration that FGTI-2734 is efficacious at selectively inhibiting growth in mutant KRAS–dependent versus mutant KRAS–independent tumors and at doses that were not toxic to mice underscores the specificity of FGTI-2734. Consistent with its specificity and lack of toxicity in our in vivo investigations, we also found that FGTI-2734 did not induce apoptosis in normal cells and in human cancer cells with wild-type RAS.

The fact that not all human tumors that harbor mutant KRAS are dependent on KRAS, coupled with the demonstration that FGTI-2734 is efficacious in only those that depend on mutant KRAS, suggests that cancers such as pancreatic cancer, which is known to be driven by mutant KRAS, may be responsive to FGTI-2734 treatment. Consistent with this, FGTI-2734 inhibited the in vivo growth of PDXs from four patients with pancreatic cancer. The FGTI-2734 antitumor activity shown in PDXs from patients with aggressive tumors that were refractory to chemotherapy underscores its potential in overcoming tumor resistance. These antitumor effects were independent of the mutant KRAS isoforms, which varied from G12C, G12D, and Q61K (cell lines) to G12D, G12V, and G13D (patients). This is an important finding, as the position and the type of mutations in KRAS are thought to influence its transforming capacity, as well as drug response, in patients with cancer.

In addition to its in vivo antitumor activity against PDXs, FGTI-2734 was also efficacious at inhibiting the viability of early-passage mutant KRAS–driven cancer cells derived from patients with pancreatic cancer. This is an important finding, as we have previously demonstrated that early-passage mutant KRAS–driven cancer cells derived from patients with pancreatic cancer are generally resistant to chemotherapeutic agents, such as gemcitabine, with IC₅₀ values over 100 μmol/L (19). Importantly, as a single agent, FGTI-2734 was significantly more potent when patient-derived pancreatic tumor cells were cocultured in a three-dimensional environment with pancreatic stellate cells. This is an important finding as pancreatic stellate cells are known to promote chemotherapeutic resistance through cross-talk with pancreatic cancer cells (31).

We found that FGTI-2734 effectively inhibited KRAS membrane localization and growth of mutant KRAS–dependent tumors at nontoxic doses. This is in contrast to the work of Lobell and colleagues (32), in which KRAS prenylation could not be inhibited without toxicity using L-778123, an inhibitor of both enzymes but with 50-fold selectivity for FT over GGT-1 (unlike FGTI-2734, which is an equipotent dual FT and GGT-1 inhibitor). Furthermore, our results are consistent with knockout mouse studies where simultaneous suppression of FT and GGT-1 was highly effective at preventing mutant KRAS–induced lung tumors and at improving mouse survival without toxicity (20). Thus, these genetic studies are consistent with our pharmacologic findings and together warrant further development of dual FT and GGT1 inhibitors as safe therapeutics for cancer treatment.

Our findings may also have major implications for patients whose tumors harbor NRAS mutations. Indeed, NRAS is found mutated frequently in a number of cancers, including hematologic malignancies, sarcomas, and melanomas; like KRAS, NRAS is also alternatively geranylgeranylated when cancer cells are treated with FTIs (13). FGTI-2734, but not FTI-2148 or GGTI-2418, inhibited NRAS membrane localization, suggesting that FGTI-2734 may be efficacious in the treatment of patients with cancer whose tumors are driven by mutant NRAS.

Another important finding is that, FGTI-2734 abrogated major signal transduction pathways downstream of KRAS that are known to be pivotal to survival and therapy resistance in a large number of human cancers. Indeed, treatment of PDX mice with FGTI-2734 suppressed the phosphorylation of AKT and S6, indicating that the activation of the PI3K/AKT/mTOR oncogenic pathway was thwarted by FGTI-2734 treatment. In contrast, P-Erk1/2, which mediates another major canonical pathway downstream of KRAS, was minimally affected, suggesting a nonpromiscuous effect of FGTI-2734 on signaling pathways. FGTI-2734 also suppressed cMYC levels, increased p53 levels, and activated CASPASE 3, suggesting that the ability of FGTI-2734 to inhibit tumor growth could be due, at least in part, to the suppression of oncogenic pathways such as PI3K/AKT/mTOR and cMYC and the upregulation of tumor suppressor pathways such as those mediated by the upregulation of p53. The ability of FGTI-2734 to prevent mutant KRAS from localizing to the membrane and inhibiting it from signaling downstream could explain the observed abrogation of the PI3K/AKT/mTOR pathway downstream of KRAS. However, considering that FGTI-2734 inhibits FT and GGT-1, enzymes that have many substrates (13), the inhibition of other farnesylated and/or geranylgeranylated proteins could also contribute to the FGTI-2734 effects on signaling, induction of apoptosis, and inhibition of tumor growth.

Finally, FGTI-2734 may be efficacious in a wide spectrum of human tumors, as pathways mediated by PI3K/AKT/mTOR, cMYC, and p53 mediate oncogenesis in most human cancers. For example, in a recent study of 19,784 tumor samples representing over 40 cancer types, 38% of patients had an alteration in one or more PI3K/AKT/mTOR pathway components, most commonly PTEN loss and mutations in PIK3CA, PTEN, or AKT1 (33). Importantly, activation of the PI3K/AKT/mTOR pathway is associated with poor patient survival (34). Furthermore, p53 is either deleted or mutated in about 50% of human cancers or inactivated/downregulated in the other 50% (35). Finally, cMYC overexpression through gene amplification is also common in human cancers, including pancreatic (14%), breast (27%), ovarian (31%), colorectal (6%), lung (5%), and renal cell carcinoma (23%; ref. 36).

Taken together, we have designed FGTI-2734, a dual FT and GGT-1 inhibitor capable of suppressing major cancer-causing pathways, upregulating the tumor suppressor p53, and activating CASPASE 3, leading to thwarted growth in vivo and in vitro of patient-derived tumors from 12 patients with pancreatic cancer with different KRAS mutations. Further preclinical studies are warranted to support the clinical development of FGTI-2734 in tumors where mutant KRAS is a major contributor to their malignancy and resistance to therapy.

Said Sebti and Andrew Hamilton are co-inventors of FGTI-2734. No other potential conflicts of interest were disclosed.

Conception and design: S.M. Sebti, A. Kazi, S. Fletcher, C. Cummings, A.D. Hamilton

Development of methodology: A. Kazi, S. Xiang, H. Yang, L. Chen, P. Kennedy, M. Ayaz, C. Cummings, F. Beato, Y. Kang, J.B. Fleming, A.D. Hamilton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Kazi, S. Xiang, L. Chen, P. Kennedy, M. Ayaz, C. Cummings, H.R. Lawrence, P.W. Underwood, J.B. Fleming, J.G. Trevino

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.M. Sebti, A. Kazi, S. Xiang, P. Kennedy, M. Ayaz, C. Cummings, H.R. Lawrence, F. Beato, J.G. Trevino

Writing, review, and/or revision of the manuscript: S.M. Sebti, A. Kazi, S. Xiang, H. Yang, M. Ayaz, H.R. Lawrence, F. Beato, M.P. Kim, P.W. Underwood, J.B. Fleming, J.G. Trevino

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.M. Sebti, F. Beato, M.P. Kim, A. Delitto, J.G. Trevino, A.D. Hamilton

Study supervision: S.M. Sebti, H.R. Lawrence, A.D. Hamilton

This work was funded in part by NIH grant R35 CA197731-01 (to S.M. Sebti, PhD) and was supported in part by the Chemical Biology, Molecular Genomics and SAIL Core Facilities at the H. Lee Moffitt Cancer Center & Research Institute; an NCI-designated Comprehensive Cancer Center (P30-CA076292). We thank these cores for their outstanding assistance and expertise. Part of the work was also supported in part by MD Anderson Cancer Center P30-CA016672. We also thank Michelle Blaskovich for technical assistance and Rasa Hamilton for editorial assistance. Synthesis of some of the compounds reported in this article was supported by NIH grant R50CA211447 (to H.R. Lawrence).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.