Identification of a farnesol analog as a Ras function inhibitor using both an in vivo Ras activation sensor and a phenotypic screening approach

Abstract

Mutations in Ras isoforms such as K-Ras, N-Ras, and H-Ras contribute to roughly 85, 15, and 1 % of human cancers, respectively. Proper membrane targeting of these Ras isoforms, a prerequisite for Ras activity, requires farnesylation or geranylgeranylation at the C-terminal CAAX box. We devised an in vivo screening strategy based on monitoring Ras activation and phenotypic physiological outputs for assaying synthetic Ras function inhibitors (RFI). Ras activity was visualized by the trans-location of RBD_Raf1-GFP to activated Ras at the plasma membrane. By using this strategy, we screened one synthetic farnesyl substrate analog (AGOH) along with nine putative inhibitors and found that only m-CN-AGOH inhibited Ras activation. Phenotypic analysis of starving cells could be used to monitor polarization, motility, and the inability of these treated cells to aggregate properly during fruiting body formation. Incorporation of AGOH and m-CN-AGOH to cellular proteins was detected by western blot. These screening assays can be incorporated into a high throughput screening format using Dictyostelium discoideum and automated microscopy to determine effective RFIs. These RFI candidates can then be further tested in mammalian systems.

Introduction

Ras proteins are monomeric small guanosine triphosphatases (GTPases) which regulate normal cellular proliferation [1]. Aberrant signaling through Ras pathways occurs both as the result of mutations in Ras and from the mis-regulation of genes upstream and downstream of Ras [1-3]. 20 % of human tumors have activating point mutations in Ras, with most found in KRAS (about 85 % of total), then NRAS (about 15 %), and lastly HRAS (<1 %) [2]. These mutations all affect the GTPase activity of RAS, preventing GTPase-activating proteins from promoting hydrolysis of GTP on RAS and therefore causing RAS to accumulate in the GTP-bound active form [2, 4]. Ras GTPases activate four major effector pathways including Raf protein kinases, phosphatidyl inositol 3-kinase (PI3K), Ral guanine nucleotide dissociation stimulator (RAL GDS), and phospholipase C-epsilon. While Raf regulates cell cycle progression and transcription, PI3K plays a role in cell survival, transcription, translation, and cytoskeletal signaling [5]. Ral GDS regulates transcription, vesicle transport, and cell cycle progression [2].

Post-translational prenylation plays a critical role in the proper localization and activation of Ras [2, 6-8]. Post-translational farnesylation of Ras catalyzed by protein farnesyltransferase (FTase) is obligatory for protein function and sub-cellular localization. FTase catalyzes the transfer of a farnesyl group from farnesyl diphosphate (FPP) to proteins with a cysteine residue located in a C-terminal CAAX motif where C is the modified cysteine, A is often an aliphatic residue, and X is Ser, Met, Ala, or Gln [9-12]. When X is a Leu, Ilu, or Val, proteins are geranylgeranylated by geranylgeranyl transferase type 1 (GGTase I) [9]. After prenylation, the AAX peptide is cleaved by the endopeptidase Ras-converting enzyme1. This is followed by methylation of the carboxyl terminus of the terminal farnesylated cysteine residue by S-isoprenyl cysteine O-methyltransferase [13-15].

K-Ras is a major driver of the aggressive nature of many cancers, including tumor growth, invasiveness, metastasis, and therapeutic resistance [16-22]. As a result, methodologies for therapeutically targeting K-Ras and Ras signal transduction pathways have been highly sought. To date, there are ~20 new Ras-directed therapeutic agents in clinical trials for cancer. Targets include FTase, GGTase I, transforming protein p21-mRNA (H Ras mRNA), c-Raf1 mRNA, mitogen-activated protein kinase kinase, Raf, Epidermal growth factor receptor (EGFR), and ERBB2/HER2/neu [2]. This knowledge led to the development and clinical testing of FTase inhibitors (FTIs) in multiple cancer types [18, 23] including pancreatic cancer [24]. FTIs in clinical trials include CAAX peptidomimetics and FPP analogs [2, 25]. Although initial FTI clinical trials in breast cancer showed substantial responsiveness, more so than other solid tumors [19], the lack of clear survival benefits halted these trials. The lack of FTI clinical efficacy is attributed to alternative prenylation of FTase substrates such as K-Ras, which become geranylgeranylated by GGTase I when FTase is inhibited [26-28]. Therefore, there is a clear need for pharmacological agents that specifically target K-Ras and which can also avoid alternative prenylation [29, 30].

These trials suggest that the development of Ras function inhibitors (RFIs) which do not leave the target proteins as substrates for alternative prenylation by GGTase I is a promising area of research. FTIs have traditionally been screened by measuring the incorporation of radiolabeled precursors of FPP and GGPP into farnesylated proteins [31, 32]. The drawback of this method is low sensitivity, radioactive contamination, and time consumption [33]. Recently, synthetic prenyl substrate analogs and inhibitors have been shown to be incorporated into cellular proteins. In particular, AGOH is a pro-drug version of AGPP and is incorporated into normally prenylated cellular proteins in an FTase-dependent manner [34-39]. To exploit this incorporation, antibodies have been developed against anilinogeranyl (AG) epitopes and used for detecting proteins modified with the AG moiety [8, 39]. Ras modified with farnesyl analogs has also been microinjected into Xenopus laevis oocytes to examine the effects of unnatural prenyl groups on signaling. Oocytes were monitored for downstream Ras effector functions and included germinal vesicle breakdown and MAPK activity [8]. In this method, it was found that hydrophilic farnesyl analogs p-NO2-AGPP, p-CN-AGPP, and Isox-GPP could function as H-RFIs. This procedure requires 3 days for incorporation and multiple steps that include acclimatizing animals, anesthesia, oocyte extraction, purification of H-Ras, modification with FPP analogs, microinjection, and a gel shift assay [8]. Such an elaborate protocol is very difficult to adopt for high throughput assays.

The Dictyostelium discoideum genome contains a protein prenyl transferase α subunit (Gene IDDDB_G0287077), CAAX prenyl protease (Gene IDDDB_G0290849), and isoprenylcysteine carboxyl methyl transferase (Gene ID-DDB_G0272799). These enzymes encompass the post-translational machinery for localization and activation of prenylated proteins. The D. discoideum genome also contains eighteen Ras GTPases (http://dictybase.org). With its simple media requirement for growth, its fast doubling time, rapid signaling responses, and genetic tractability, D. discoideum is a versatile model system for screening Ras function inhibitors. Here, we report a simple screening procedure based on live cell imaging of cells expressing Ras-binding domain of mammalian Raf1 fused to GFP (RBD_Raf1-GFP), a biosensor for activated RasG and likely other Ras GTPases [40, 41]. By employing this technique, we successfully identified m-CN-AGOH as a novel Ras function inhibitor. This technology has the potential for use in high throughput screening to identify potential pharmaceutical agents that target Ras and Ras pathways in cancer.

Materials and methods

Chemicals and devices

All analogs (Table 1) were synthesized as described [36, 42, 43]. All compounds were dissolved in DMSO (Sigma) and stored at −20 °C until used. HL5 medium in powder form was purchased from For Medium. Agar and G418 are from Research Products International Corps., and Hygromycin is from Cellgro. Folic acid was obtained from Fisher scientific, cAMP from Sigma chemicals, and the anti-AGPP polyclonal antibody was developed in the Spielmann lab [39]. Blocking buffer and secondary IR dye 680 coupled antibody were purchased from LiCor. Femtotip microinjection needles, micromanipulator, and pump are from Eppendorf International. One-well glass chambers were purchased from Lab-Tek (Nalge Nunc, Naperville, PA) and 0.1-cm electroporation cuvette is from Fisher scientific.

Table 1.

Potential Ras functional inhibitors

Compound #	Sample Code	Structure	Ras inhibition
1	p-tBu-AGOH		No
2	p-iPr-AGOH		No
3	p-CF3O-AGOH		No
4	AGOH		No
5	p-PhO-AGOH		No
6	m-CF3O-PGOH		No
7	p-CF3S-PGOH		No
8	m-CN-AGOH		Yes
9	p-CF3O-PGOH		No
10	BzDMAOH		No GGTI
11	AFOH		No

Dictyostelium transformation

Wild-type (A×2) cells were transformed with the plasmids expressing RBD_Raf1–GFP [40, 44] and LimE-RFP [45]. Wild-type (A×2) cells transformed with RBD_Raf1–GFP alone were used as the control. On the following day, the antibiotic G418 (30 μg/ml) was added to wild-type cells expressing RBD_Raf1-GFP alone, while G418 and Hygromycin (35 μg/ml) were added to other wild-type cells expressing both RBD_Raf1-GFP and mRFPmars-LimEΔcoil.

Screening strategy for Ras function inhibitors

Mammalian Raf-1 is a MAP kinase kinase kinase, which is a downstream effector of the Ras family of GTPases [2]. In D. discoideum, Raf-1 directly interacts with GTP-bound RasG, but can likely bind to other Ras proteins, as has been shown when RasG is deleted [40, 41]. By taking advantage of these interactions, it is possible to image Ras activation in live cells by fusing the Ras-binding domain of Raf-1 to a fluorescent reporter, the green fluorescent protein (GFP). Here, we used the genetically encoded RBD_Raf1-GFP as a probe for detecting activated Ras. We treated RBD_Raf1-GFP- and LimE-RFP-expressing cells with the AGOH analogs. LimE-RFP, which fluoresces red, is a biomarker for F-actin [45]. F-actin polymerization results from Ras signaling activation and is a critical component of cell migration. Cells expressing RBD_Raf1-GFP alone were used as DMSO-treated control and would be lacking the LimE-RFP expression protein. To insure that both cell lines were equally competent for folic acid stimulation, the control and treated cells were prepared for the assay at the same time and under the same conditions. Cells were stimulated with a uniform stimulus of folic acid and their RBD_Raf1-GFP responses measured. Only experiments where the control cells responded were subsequently scored. In no case did the treated cells respond while the control cells did not. Translocation of LimE-RFP to the plasma membrane was also monitored in the inhibitor-treated cells.

Processing and imaging of cells

Wild-type cells expressing RBD_Raf1-GFP and LimE-RFP or LimE-RFP (1 × 10⁷ cells) in a Petri dish were treated with 30 μM AGPP analogs overnight (16 h). The following day, these cells were removed from the dish and shaken in 50-ml flask for 3 h at 110 rpm and pulsed with 50 μM folic acid for 3 h. Cells were then washed 3X with development buffer (5 mM Na₂HPO₄, 5 mM KH₂PO₄, 1 mM CaCl₂, 2 mM MgCl₂ pH to 6.5), seeded on the cover glass of a one-well chamber microscope slide, and allowed to adhere for 15 min. 10 μl of 100 μM folic acid was loaded into a Femtotip micropipette, connected to a microinjector, and lowered to the coverslip using a micromanipulator. Positive pressure (50 hPa) was applied by the microinjector. 30 frames were captured with an interval between frames of ~1.5 s. Cells were stimulated with folic acid at the 5th frame of the movie by bringing the micropipette into close proximity of the cells being imaged. The cells were imaged and the translocation of RBD_Raf1-GFP and LimE-RFP to the plasma membrane was monitored by a Zeiss Axiovert microscope using a 40X (1.35 NA) oil objective lens and GFP and CY3 filters (Chroma).

Development on agar plate

Cells were treated with a putative RFI or DMSO as described above. A 1 % Agar solution was prepared and poured into a 35-mm Petri dish and allowed to solidify. Cells were washed with developmental buffer (DB) 3X and resuspended in DB containing 30 μM RFI or 10 μl DMSO at a concentration of 1 × 10⁹ cells/ml. For the wash experiments, cells were resuspended in DB alone. 100 μl of a 0.7 × 10⁸ cell suspension was spread onto a DB gar plate. The plate was incubated at 22 °C. Cells were imaged by bright field microscopy at indicated time points.

Early cell polarization and chemotaxis assay

Wild-type cells were developed on the DB agar plate as described above. At the 6th hour of development, cells were resuspended in DB by mechanically pipetting up and down several times. Cells were then positioned on the cover glass of a one-well chamber by micropipetting, allowed to adhere, and then imaged as described above. The chemotaxis assay was performed as described previously [46].

Eccentricity

Polarity of the cell can be measured in terms of eccentricity, which is a parameter associated with conic section. It is a measure of how much the conic section deviates from being circular. The eccentricity of a circle is zero. The eccentricity of an ellipse is >zero, but <1[47]. The formula used was √1-B2/A2, where A is the length and B is the width of the cell [48]. Length and width were measured by drawing straight lines across the cell using Slidebook-5 software.

Detection of AG- and m-CN-AG-modified proteins by western blot

Wild-type cells were treated with 30 μM of m-CN-AGOH, AGOH, or DMSO, separately and overnight. Cells were removed from the plate and washed 3X with DB and suspended at a concentration of 2 × 10⁷ cells/ml in DB. 40 μl of cells were transferred into a microcentrifuge tube containing 10 μl of 5X SDS sample buffer. The sample was heated at 95 °C for 5 min in a hot block. 2.5 μl (4 × 10⁴ cells) cells were loaded into wells of 10 % SDS–polyacrylamide gel. After separation, protein was transferred to a nitrocellulose membrane at 30 V for an hour using the Invitrogen Xcell II blot module. The membrane was incubated with LiCor-Odyssey blocking buffer for an hour and then with anti-AGPP polyclonal antibody (1:2500 dilution) overnight at 4 °C [39]. The membrane was washed 3X for 10 min with TBST and incubated with IR dye 680 coupled secondary antibody for 1 h at room temperature. The membrane was washed three times with TBST, rinsed three times with TBS, and imaged using a LiCor Odyssey IR imager. Lane intensity of blot was determined by ImageJ.

Results

Synthesis and properties of anilinogeranyl diphosphate analogs

FTase-catalyzed transfer of FPP analogs to H-Ras depends on both the shape and size of the analog, [49] but not on analog hydrophobicity. However, Ras function can be inhibited by modification with less hydrophobic FPP analogs such as GPP [8]. We have previously shown that an aniline or phenoxy group is isosteric with the isoprene units of FPP and that analogs with a range of substituents on the aryl group are FTase-transferable substrates [50]. Based on these findings, we synthesized aniline and phenoxy FPP analogs with various aryl group substitutions that make each analog differ in shape, size, and hydrophobicity (Table 1) [36, 42, 43].

Identification of a Ras function inhibitor

Cells expressing RBD_Raf1-GFP alone (controls) or with LimE-RFP (treated) were imaged for Ras activation and actin polymerization, respectively, as described in the screening strategy. Inhibition of Ras farnesylation would be expected to inhibit translocation of the RBD_Raf1-GFP to the plasma membrane. Of nine putative RFIs (Compounds 1–3, 5–9, 11 in Table 1) screened, only m-CN-AGOH (Compound-8 in Table 1) was found to block translocation of RBD_Raf1-GFP and LimE-RFP (Fig. 1A, Supplementary Figure-1). Cells treated with the other putative inhibitors (Supplementary Figure-2) and the substrate analog AGOH (Compound-4 in Table 1; Fig. 1B) appeared to have normal Ras activity and actin polymerization responses.

Fig. 1.

m-CN-AGOH inhibits Ras activation. Cells expressing RBD_Raf1-GFP and LimE-RFP were treated with m-CN-AGOH (A) or AGOH (B) and stimulated uniformly with 100 μM folic acid. Putative inhibitor m-CN-AGOH (Compound-8, Table 1) inhibited RBD_Raf1-GFP and LimE-RFP translocation to membrane. C and T indicate control and treated cells, respectively. Note that treated cells show no Ras activity or actin response. Substrate analog AGOH did not inhibit the translocation of RBD_Raf1-GFP or LimE-RFP. Arrow indicates the recruitment of RBD_Raf1-GFP and LimE-RFP to the plasma membrane in response to folic acid stimulation (bar, 5 μm)

m-CN-AGOH delays cell polarization and development

Dictyostelium discoideum cells alter their morphology several hours after starvation and become elongated and polarized, with a distinct anterior and rear [51]. Cells naturally polarize in response to cAMP autocrine signaling and to cAMP gradients during cell migration [52-55]. Signaling proteins such as Ras, PI3K, and PI(3,4,5)P3 localize at the leading edge, while PTEN and Myosin-II localize at the rear and contribute to cell polarity and the migratory response [41, 56-59]. Cells were imaged at 6 h to visualize their ability to polarize in the presence of m-CN-AGOH or DMSO alone. Cells treated with DMSO polarized normally, while m-CN-AGOH-treated cells were still unpolarized at 6 h (Fig. 2).

Fig. 2.

Delayed polarization and development of m-CN-AGOH-treated cells. Cells were treated with either m-CN-AGOH or DMSO as a control. m-CN AGOH-treated cells immediately after starvation (0 h) and after 6 h. The treated cells did not polarize at 6 h, while the control cells were very polarized (bar, 5 μm). The m-CN-AGOH-treated cells also did not form fruiting bodies at 24 h, while the DMSO-treated control cells did develop in a timely manner and formed fruiting bodies (bar, ~50 μm)

Cells have typically aggregated and formed small mounds by 8 h and continue through development to form a multi-cellular fruiting body within 24 h [52-55]. We examined the treated cells by microscopy at 24 h (Fig. 2). Cells treated with DMSO had undergone all steps of the developmental process, culminating into fruiting bodies; however, m-CN-AGOH-treated cells did not develop past the slug stage. This result correlates with the above observation showing the m-CN-AGOH-treated cells did not polarize normally.

m-CN-AGOH inhibits random and directional migration

After incubation with m-CN-AGOH for 16 h, cells were developed on a DB agar plate as described above. The effects on random motility and directional migration under cAMP gradient were studied. m-CN-AGPP-treated cells did not chemotax after 6 h of development (Fig. 3). DMSO-treated cells developed and migrated as wild-type cells (Fig. 3). Random motility in the absence of a chemoattractant was slightly inhibited (Supplementary movie-1) when compared to DMSO-treated cells (Supplementary movie-2). Random motility of the DMSO-treated cells was comparable to wild type.

Fig. 3.

m-CN-AGOH inhibited directional migration. Cells were developed and treated with m-CN-AGOH or DMSO as described in the methods and subjected to cAMP chemotaxis. Cells in the continuous presence of m-CN-AGOH or washed did not polarize and migrate toward a cAMP-filled micropipette. DMSO-treated cells polarized and migrated directionally toward the cAMP-filled micro-pipette (bar, 10 μm)

The effect of m-CN-AGOH lasts at least 24 h

After m-CN-AGOH treatment and washing, cell polarity, development, and migration were monitored. Cells were not polarized after 6 h of starvation and development, and did not form fruiting bodies at 24 h (Fig. 4), while control DMSO-treated cells underwent a normal developmental process. Polarity of the 6 h developed cells was measured in terms of eccentricity as described in the methods (Fig. 5). The eccentricity of 0.59 for m-CN-AGOH-treated cells was significantly lower than 0.98 for DMSO-treated control cells (n = 15, P < 0.005). Similarly, the eccentricity of m-CN-AGOH-washed cells was 0.62, which is significantly lower than DMSO-washed cells (0.97), n = 15, P < 0.005. Cells treated with m-CN-AGOH before and after washing had random motility defects (Supplementary movies-1, 3) and were not able to chemotax in a cAMP gradient after starvation (Fig. 3). These findings clearly indicated that modification of Ras by m-CN-AGOH is irreversible and lasts at least 24 h.

Fig. 4.

The effect of m-CN AGOH lasts for at least 24 h. After treatment with m-CN AGOH and a subsequent wash, cells were imaged at 0, 6, and 24 h. m-CN-AGOH-treated cells did not polarize (bar, 5 μm) or develop, while control cells underwent normal developmental and fruited in 24 h (bar, ~50 μm)

Fig. 5.

m-CN-AGOH significantly inhibited the polarized morphology of cell. Wild-type cells were treated with m-CN-AGOH or DMSO and were developed with respective compounds in DB agar plate for 6 h. Another set of m-CN-AGOH- and DMSO-treated cells were developed in DB agar plate without compounds. Eccentricity was measured as described in the methods. The polarized morphology of m-CN-AGOH-treated cells was significantly inhibited (n = 15, P < 0.005) compared to control cells. Cells were not able to polarize even after being removed from m-CN-AGOH for 6 h (n = 15, P < 0.005)

Incorporation of AGPP analogs to prenylated proteins in Dictyostelium

m-CN-AGOH was the only molecule of 10 compounds screened that inhibited prenyl function as assayed by RBD_Raf1-GFP membrane translocation. To test whether we could detect incorporation in Ras, we performed Western blotting analysis (Fig. 6). Previously, a polyclonal anti-anilino geranyl antibody was used to detect AGPP- and AGPP analog-modified proteins in mammalian cells and in an in vitro system [39]. We used the same antibody to detect AGPP- and m-CN-AGPP-modified proteins in D. discoideum. We found that both AGPP and m-CN-AGPP were incorporated into proteins with molecular weights of approximately 50 and 60 kDa, respectively. Accordingly, whole-lane densitometry analysis indicated two bands in AGPP- and m-CN-AGPP-treated cells (lane-1 and 2) with slightly different intensity peaks, while no bands were observed in control cells (lane-3). However, incorporation was not detected with molecular weights in the range of Ras (19–27 kDa) or in other small molecular weight G-proteins. This was not completely surprising, at it has been reported that the farnesylated small molecular weight GTPases are present in relatively low abundance [32, 39]. Consequently, the farnesylated small molecular weight GTPases are likely difficult to detect. In another study, only faint bands corresponding to these molecules were found when tritium-labeled farnesol was incubated with glioma cells (C6) or green monkey kidney cells (CV-1) [39, 60]. Since natural levels of endogenous Ras are very low, detection only occurred when using extracts from HEK-293 cells over-expressing a GSTHRas grown in the presence of AGOH with the polyAG-Ab [39]. Interestingly, we also observed two unknown prenylated proteins in D. discoideum, which were a similar size to the mammalian proteins previously observed.

Fig. 6.

Detection of AGOH and m-CN-AGOH incorporation into cellular proteins by western blot. Cells treated with AGOH (lane-1) or m-CN-AGOH (lane-2) or DMSO (lane-3). AGPP and m-CN-AGPP were incorporated into proteins with molecular weight of ~50 and 60 kDa, respectively. No bands were observed in DMSO-treated control cells. Accordingly, two bands were observed in lane-1 and lane-2 and none was observed in lane-3 as supported by densitometry analysis

Discussion

Inhibition of Ras function is a promising approach for developing anticancer therapies. Various agents such as FTIs and the nontoxic farnesylcysteine analog farnesylthiosalicylic acid have been used as methods to restore regulation of Ras–GTP levels and to alter the interaction of Ras–GTP with downstream targets. In the past, RFI screening assays have relied on the monitoring of readouts downstream of Ras signaling pathway such as MAPK activation and germinal vesicle breakdown. This procedure, used in Xenopus oocyte, is time consuming and cannot analyze large numbers of samples in a short time frame [8, 39]. Therefore, designing an assay that can be simple to use for high throughput testing of these small molecule inhibitors is imperative. Live cell imaging of D. discoideum cells expressing RBD_Raf1-GFP and treated with various AGOH analogs can be readily performed.

We describe a technique that successfully identified the m-CN-AGOH analog as an RFI. This system provides several readouts which could not be detected by previously described anti-AG antibody-based detection strategies [39]. First, we can monitor Ras activation in real time by stimulating cells with a chemoattractant and monitoring the translocation of the Ras biosensor RBD_Raf1-GFP to the plasma membrane. Also, obvious phenotypic changes can be monitored during aggregation, development, and motility. Disruption of these events would be expected if Ras function was inhibited [41, 61]. Consistent with previous observations described in X. laevis oocytes, m-CN-AGOH appears to inhibit Ras function. Loss of Ras activity results in improper regulation of downstream effector molecules, and cells fail to polarize and develop on time. Determination of FPP analog m-CN-AGOH as an RFI was consistent with previous findings [8] showing that hydrophilic FPP analogs are RFIs. Moreover, this is the first time that an RFI has been tested for its in vivo stability and reversibility. In our screening strategy, it was possible to determine the long-term stability of the mCN-AGOH modification. We found that m-CN-AGOH exhibited an inhibitory effect on polarity, development, and motility even after 24 h of removal. It is possible that modification of Ras by m-CN-AGPP dislodges membrane localization and thereby ablates its function. Ras activity was easily visualized with the bright RBD_Raf1-GFP reporter and suggests that this biosensor would be useful for high throughput screening of other RFIs using Dictyostelium. Positive hits could then subsequently be tested for other phenotypes such as polarity, development, and motility before screening in a mammalian system.