CCG-1423: a small-molecule inhibitor of RhoA transcriptional signaling

Cancer metastasis is a significant medical problem responsible for thousands of deaths every year (1). Metastases arise when dysregulation of one or more cellular processes allows malignant cells to escape the confines of the tissue of origin and establish themselves in alternate sites. These processes include cell adhesion, migration, invasion, extravasation, survival, and proliferation (1). Multiple members of the Rho family of small GTPases play important roles in these cellular processes and in some human tumors (e.g., colon, esophageal, lung, pancreatic, and inflammatory breast cancers), up-regulation of RhoA or RhoC is associated with a poor clinical outcome (2, 3).

Rho GTPases are best known for their effects on the actin cytoskeleton. The three main Rho GTPase subfamilies, RhoA, Rac, and Cdc42, control stress fiber formation, lamellipodia, and filopodia (4), respectively, which are structures important for cell motility. Rac and Cdc42 have been implicated in tumor growth, migration, and invasion in both mouse skin and human breast tumors (5–8). Within the RhoA family (RhoA, RhoB, RhoC, etc.), there is evidence for involvement of both RhoA and RhoC in cancer, with the latter being clearly implicated in multiple malignancies. RhoC is essential for inflammatory breast cancer cell growth, invasion, and survival (9, 10) and more recently was found to be critical for invasion by PC-3 prostate cancer cells (11). Similarly, RhoC is critical for in vivo metastasis of polyoma T antigen–induced mammary tumors (12). Clearly, the RhoA family GTPases play important roles in multiple cellular processes central to tumor growth and metastasis.

Heterotrimeric G protein–coupled receptors (GPCR), especially those activating the G12/13 family of Gα subunits, are upstream regulators of the Rho pathway and are also implicated in metastasis. Lysophosphatidic acid (LPA), thrombin, and bombesin, acting on their respective GPCRs, stimulate Rho signaling and migration of various cancer cell lines (13–15). A family of three rhoGEFs containing a regulator of G protein signaling homology (RH) domain serves to couple receptors and Gα_12/13 subunits to RhoA (16). The three RH domain-containing rhoGEFs are p115rhoGEF, PDZ-rhoGEF, and leukemia-associated rhoGEF (LARG). The latter was initially found as a fusion to mixed lineage leukemia in a patient with acute myelogenous leukemia (17). Binding of activated Gα subunits (i.e., Gα₁₂ and Gα₁₃) to the RH domain of the rhoGEF stimulates the DH-PH domain-mediated guanine nucleotide exchange on the small GTPase RhoA, leading to its activation (18–20). Another G protein–activated rhoGEF, p63-rhoGEF, is activated by Gα_q but does not contain an RH domain (21). Furthermore, the downstream Rho effector Rho-associated coiled coil–forming protein kinase (ROCK) is implicated in migration and invasion of cancer cells (22). Thus, Gα_12/13-coupled receptors can stimulate activation of RhoA as well as downstream cellular processes involved in cancer metastasis.

In addition to its established effects on motility, RhoA-dependent actin polymerization in response to activation of Gα_12/13-coupled GPCRs leads to changes in gene expression. The loss of free G-actin resulting from actin polymerization leads to its dissociation from the transcriptional coactivator megakaryoblastic leukemia 1 (MKL1), which then translocates into the nucleus where it collaborates with serum response factor (SRF) to induce gene expression (23, 24). SRF participates in many cellular processes, including cell growth and differentiation, apoptosis, and angiogenesis (25, 26). Although the role of gene transcription in Rho-related cancer biology is poorly understood, both RhoA and RhoC regulate genes important for cell growth and proliferation, such as c-fos and cyclin D1 (25, 27). RhoC overexpression has been linked to vascular endothelial growth factor-C, CXCL1 chemokine, and fibronectin up-regulation, which are important for angiogenesis and formation of the extracellular matrix (27). In addition, several serum-stimulated genes shown to be MKL1 dependent have been implicated in cancer, including SRF, adrenomedullin, epiregulin, interleukin-6, hexokinase 2, and zyxin (28). These MKL1-dependent genes participate in various cancer-linked processes, including cell growth, migration, invasion, and survival (25, 26, 29–33). Thus, alterations in gene expression are likely to be an integral part of Rho effects on cancer metastasis.

Currently, there are relatively few drugs or pharmacologic tools that target Rho GTPase family signaling pathways. Much of the effort to date has focused on inhibiting the COOH-terminal isoprenylation of the Rho GTPases. This lipid modification is necessary for membrane localization and function of the activated Rho proteins. The most widely used inhibitors of this modification include farnesyltransferase and geranylgeranyltransferase inhibitors and the cholesterol-lowering statin drugs (34). However, these compounds are not specific for the Rho family of small GTPases so their effects are difficult to interpret mechanistically. There have also been significant efforts aimed at inhibiting the Rho effector molecule ROCK. The ROCK inhibitors Y-27632 and Wf-536 have shown promising antimetastatic activity both in vitro and in vivo (35–37). To date, the only specific and direct inhibitor of Rho GTPases is the Rac1 inhibitor NSC23766 (38). Although it is relatively specific, it is not very potent (IC₅₀, ∼50 μmol/L). Thus, there is a clear need and opportunity for specific inhibitors of Rho GTPase signaling pathways.

In this study, we took advantage of a modified serum response element (SRE)-luciferase reporter to undertake a high-throughput screen aiming to identify novel small-molecule inhibitors of the RhoA family signaling pathway. We chose to use a firefly luciferase expression vector driven by a mutant SRE (SRE.L) lacking the ternary complex factor binding sites because this construct is a selective probe of RhoA-induced gene transcription (39). We initiated Rho pathway signaling through the upstream signals Gα₁₃ and LARG so that inhibition at any step from Gα₁₃ to the SRE could be detected. By this approach, we identified two novel and structurally similar small-molecule inhibitors of RhoA-stimulated transcription. Mechanistic analysis showed that the more potent compound (CCG-1423) elicits its effects downstream of RhoA and actin polymerization by a mechanism targeting MKL/SRF-dependent transcriptional activation. Interestingly, we find that CCG-1423 inhibits LPA receptor-stimulated DNA synthesis, cell growth, cell survival, and Matrigel invasion for several cancer cell lines.

The Rho-responsive SRE.L luciferase reporter construct and myc-tagged human LARG expression plasmid are described in ref. 39. The p(GAL4)₂-Luc luciferase reporter was generated by ligation of a 50-bp double-stranded oligonucleotide bearing two idealized GAL4 sites into the BamHI and BglII sites of pΔODLO2 (40). Control expression vectors driving Renilla luciferase, pRL-cytomegalovirus and pRL-thymidine kinase (TK), were from Promega. Human Gα₁₂Q231L, Gα₁₃Q226L, RhoA-G14V, and RhoC-G14V expression plasmids were from the UMR cDNA Resource Center. Expression plasmids for MKL1 (41), SRF-VP16 (42), GAL4-MKL1 (amino acids 601–931; ref. 23), and C3 exotoxin (43) were kindly provided by Dr. Michael Parmacek (University of Pennsylvania, Philadelphia, PA), Dr. Li Li (Wayne State University, Detroit, MI), Dr. Ron Prywes (Columbia University, New York, NY), and Dr. John Williams (University of Michigan, Ann Arbor, MI), respectively. The cytomegalovirus-driven NH₂-terminally HA-tagged GAL4-VP16 expression vector was generated by subcloning the coding sequence from a Rous sarcoma virus–driven version (44) into pcDNA3.1. The pcDNA3.1-zeo expression plasmid and mouse laminin were from Invitrogen. The marine toxin latrunculin B as well as daunorubicin, LPA, Igepal, antimouse IgG peroxidase conjugate, and antirabbit IgG peroxidase conjugate were from Sigma. Pertussis toxin (PTX) was from List Biological Laboratories, Inc. BD BioCoat Matrigel invasion chambers (8 μm) were from Becton Dickinson. 5-Bromo-2′-deoxyuridine (BrdUrd), anti-BrdUrd-peroxidase, BM blue peroxidase substrate, and cell proliferation reagent WST-1 were from Roche Diagnostics. The caspase-3 fluorescent peptide substrate, rhodamine 110, bis-N-CBZ-l-aspartyl-l-glutamyl-l-valyl-l-aspartic acid amide was from Biotium. The chemical compounds CCG-977 [N1-(4-([3,5-di(trifluoromethyl)anilino]sulfonyl)-phenyl)-4-chlorobenzamide] and CCG-1423 [N-[2-(4-chloroanilino)-1-methyl-2-oxoethoxy]-3,5-bis(trifluoromethyl)benzamide] were from Maybridge. The HEK293T, PC-3, SKOV-3, and A375M2 cell lines were kind gifts from Drs. J. Menon, Kenneth Pienta, and Kathleen Cho (University of Michigan) and Dr. Richard Hynes (Massachusetts Institute of Technology, Cambridge, MA). The A375, WI-38, and SW962 cell lines were obtained from the American Type Culture Collection. The SK-Mel-28 and SK-Mel-147 cell lines have been described previously (45, 46).

Cell lines were normally maintained in DMEM containing 10% fetal bovine serum (FBS), 100 units/mL penicillin, and 100 μg/mL streptomycin at 37°C in 5% CO₂. HEK293T or PC-3 cells were seeded into 96-well plates (3 × 10⁴ to 4 × 10⁴ per well) 24 h before transfection. Cells were transiently transfected by incubation with the indicated amounts of DNA plasmids plus 1 μL of LipofectAMINE 2000 (Invitrogen) per μg of DNA in antibiotic-free culture medium. For dual-luciferase measurements, various activator plasmids (i.e., Gα₁₂Q231L, Gα₁₃Q226L, RhoA-G14V, RhoC-G14V, MKL1, SRF-VP16, GAL4-VP16, and GAL4-MKL1) were included along with the SRE.L or p(GAL4)₂-Luc luciferase reporter and PRL-TK or PRL-cytomegalovirus Renilla control plasmids. The total amount of DNA was kept constant by inclusion of the appropriate amount of pcDNA3.1. Transfection efficiencies determined with pcDNA3.1-eGFP were approximately 90% to 100% and 30% to 50% for HEK293T and PC-3 cells, respectively. Five to 6 h after transfection, the transfection mixture was removed and cells were starved overnight in DMEM containing 0.5% FBS and 1% penicillin-streptomycin. Firefly and Renilla luciferase activities were determined 18 to 19 h later using the dual-luciferase assay kit (Promega) according to the manufacturer's instructions. Luminescence was read on a Victor plate reader with dual injectors (Perkin-Elmer). (The effects of CCG-1423 on serum- and LPA-induced stress fibers were minimal. There were some LPA-stimulated cells showing a change in the morphology of phalloidin staining with 5 μmol/L CCG-1423, but most samples showed nearly normal stress fiber formation.) In the site of action studies, firefly luciferase activity was normalized to Renilla luciferase activity. For testing chemical compounds, 1 μL of compound or DMSO was added to the cells (final concentration, 1% DMSO) at the beginning of the serum starvation step.

HEK293T cells (6 × 10⁶ per dish) were seeded into 10-cm dishes 24 h before transfection. Cells were cotransfected with 5 ng of Gα₁₃Q226L, 150 ng of LARG expression plasmids, and 3 μg of the SRE.L reporter plasmid. Two thousand chemical compounds (Maybridge) were prespotted onto white 96-well Costar tissue culture plates (Corning) in 1 μL DMSO (final concentration of compounds, 10 μmol/L) using a Biomek FX Workstation (Beckman Coulter). Cells were trypsinized 5 to 6 h after transfection and transferred to the compound-containing assay plates (3 × 10⁴ per well) in 100 μL DMEM containing 0.5% FBS and 1% penicillin-streptomycin. After an additional 18 to 19 h of incubation, the robotic workstation was used to remove 70 μL of medium from each well and to add 30 μL of Steady-Glo reagent (Promega). Plates were incubated for 30 min at room temperature with rocking and luminescence was read for 1 s/well on the Victor plate reader. The statistical Z′ factor (47) for high-throughput assays was calculated by using the following formula: Z′ = 1 − [(3σ_c+ + 3σ_c−) / (∣μ_c+ − μ_c−∣)] (σ = SD, μ = mean, c+ = latrunculin B, and c− = DMSO-negative control).

NIH3T3 mouse fibroblast cells (5.0 × 10⁵ per well) were plated onto coverslips in six-well plates. After attachment, cells were serum starved overnight in serum-free DMEM (0% calf serum). Cells were pretreated with compounds for 1 h before the addition of LPA (30 μmol/L) or calf serum (10%). On addition of the stimuli, the cells were incubated for an additional 1 h. Wells were then washed with PBS and fixed for 15 min with 4% paraformaldehyde. Cells were then washed again with PBS and then lysed in buffer containing 0.3% Triton X-100 in PBS for 15 min. After washing in buffer containing 0.5% Igepal in PBS, cells were incubated for 1 h with rhodamine-phalloidin (1:200) stain in PBS containing 0.5% Igepal and 1% bovine serum albumin. Cells were then washed twice in 0.5% Igepal in PBS and then mounted onto slides with Gel Mount antifade solution (Biomeda). Cell images were obtained with an Olympus FluoView 500 microscope with a 60× oil objective.

PC-3 human prostate cancer cells (1.2 × 10⁴ per well) were plated onto 96-well plates coated with laminin. After attachment, cells were serum starved overnight in serum-free (0% FBS) DMEM and treated with LPA (100 μmol/L) or FBS (10%) for 27 h. BrdUrd (10 μmol/L final) was added to the wells during the final 4.5 h of the incubation. Wells were then washed with PBS and fixed for 20 min with 100 μL of 70% ethanol in 2.3 mol/L HCl. After washing with 10% FCS in PBS, cells were incubated for 1 h at room temperature with anti-BrdUrd-peroxidase. Wells were washed with PBS and then BM blue peroxidase substrate was added, and after 30 min, the reaction was stopped by the addition of 2 N H₂SO₄. The absorbance was read at 450 nm using a Victor plate reader. To determine cell viability, the tetrazolium salt WST-1 was added to the wells during the last 1 h of incubation with BrdUrd. The mitochondrial metabolite of WST-1 cleavage was detected spectrophotometrically at 450 nm just before washing and fixing the cells for BrdUrd measurements.

Cells in normal culture medium were plated (2,000 per well) in a 96-well plate coated with laminin. After attachment, the medium was replaced with serum-free medium (0% FBS) with 30 μmol/L LPA with or without 300 nmol/L CCG-1423. Fresh LPA with or without CCG-1423 was added at day 5 to ensure that LPA and compound were present throughout the experiment. On day 8, WST-1 reagent was added to the wells for 1 h and absorbance at 450 nm was read using a Victor plate reader.

Cells in normal culture medium were plated (2 × 10⁴ per well) in a black 96-well plate coated with laminin. After overnight attachment, the medium in the wells was replaced with serum-free DMEM (0% FBS) with CCG-1423 or daunorubicin for 25 h. An equal volume of 2× reaction buffer [20 mmol/L PIPES (pH 7.4), 2 mmol/L EDTA, 0.2% CHAPS, 10 mmol/L DTT] containing 50 μmol/L of the fluorogenic caspase-3 substrate Z-DEVD-R110 was then added. After an additional 90 min of incubation at 37°C, fluorescence was measured in a Victor plate reader using excitation at 485 nm and emission detection at 520 nm.

PC-3 or SKOV-3 cells (5 × 10⁴) were transferred to 24-well Matrigel inserts in serum-free medium with or without CCG-1423. Serum-free medium (0% FBS) with or without 30 μmol/L LPA as chemoattractant was added to the lower well, and the invasion chambers were returned to the incubator for 24 h. Inserts were fixed in methanol for 10 min and then stained 30 min with 0.5% crystal violet in 20% methanol. After wiping the top surface of the filter to remove noninvaded cells, the number of remaining cells was determined by counting four nonoverlapping 20× fields. For PTX experiments, cells were treated with 100 ng/mL PTX overnight before being transferred to the Matrigel inserts and 100 ng/mL PTX was maintained in the medium.

To assess the suitability of the mutant SRE.L luciferase reporter (which lacks the ternary complex factor–binding sites present in the normal SRE) as a high-throughput assay for inhibitors of GPCR-regulated Rho-dependent pathways (Fig. 1A), we tested a series of Rho pathway activators beginning with Gα₁₃ and LARG. As shown in Fig. 1B, cotransfection of HEK293T cells with expression vectors for the constitutively active mutant Gα subunit Gα₁₃Q226L together with LARG activates the SRE.L-driven reporter synergistically (i.e., 1.6-fold above the sum of the independent stimulation by Gα₁₃Q226L and LARG). In Fig. 1C, effective stimulation is also observed for a constitutively active Gα₁₂ guanine nucleotide-binding protein, Gα₁₂Q231L, as well as the active Rho GTPases RhoA-G14V and RhoC-G14V. In addition, the SRE.L-driven reporter can be directly stimulated by expression of either the SRF coactivator MKL1 or the chimeric activator SRF-VP16 in which the potent VP16 viral transactivation domain is fused to full-length SRF (Fig. 1C). Thus, the pathway can be activated using signals either upstream (Gα₁₃ and/or LARG) or downstream (MKL1 or SRF-VP16) of RhoA and RhoC. As shown in the signaling pathway diagram (Fig. 1A), Rho-mediated SRE.L luciferase stimulation is dependent on actin polymerization and the ensuing monomeric G-actin depletion. Therefore, we examined the effect of the marine toxin latrunculin B, which binds to G-actin and inhibits actin polymerization, for its ability to inhibit Gα₁₃Q226L-stimulated SRE.L-driven luciferase expression. As shown in Fig. 1D, 0.5 μmol/L latrunculin B completely abrogates the Gα₁₃Q226L stimulation, confirming that our SRE.L system is indeed dependent on actin polymerization. Lastly, to ensure that the SRE.L reporter response is Rho dependent, we cotransfected cells with expression vectors for both Gα₁₃Q226L and the Clostridium botulinum exotoxin C3. ADP-ribosylation of Rho by the C3 exotoxin results in Rho inactivation. Gα₁₃Q226L-stimulated SRE.L luciferase expression is nearly abolished by the C3 exotoxin (Fig. 1D), showing the Rho dependence of the response.

To identify novel chemical inhibitors of the RhoA pathway, we first adapted the SRE.L luciferase assay to a high-throughput format using 293T cells coexpressing Gα₁₃Q226L and LARG. Statistical analysis to determine the robustness and reproducibility of the assay for high-throughput screening yielded a Z′ factor of 0.7 (see Materials and Methods), which indicates that it is well suited for our purposes. Using this assay, we screened a 2,000-compound subset of the Maybridge diverse chemical compound collection. The results of the screen are summarized in Fig. 2A. Applying a stringent cutoff of >75% inhibition and using the actin polymerization inhibitor latrunculin B as a positive control, we obtained 39 candidates. We used the dual-luciferase format (see Materials and Methods) as a follow-up assay to confirm the initial results and to test for nonspecific cellular toxicity or general transcriptional inhibition. Of the original 39 hits, we confirmed inhibition of the SRE.L luciferase expression for 18 compounds. Of these, 13 also inhibited cytomegalovirus Renilla luciferase expression, suggesting a generalized transcriptional inhibition or nonspecific cellular toxicity. This yielded five confirmed hits, of which four were available for resupply from Maybridge. A common problem in high-throughput, luciferase-based inhibitor screens is the potential for the recovery of direct luciferase enzyme inhibitors. Of the four compounds isolated in the screen, two inhibited firefly luciferase in cell lysates, indicating that they inhibited the reporter enzyme directly rather than its cellular expression. Thus, out of 2,000 compounds screened, we identified 2 lead compounds, CCG-977 and CCG-1423, as specific inhibitors of the RhoA pathway based on our criteria. Strikingly, these compounds share substantial structural similarity because they contain identical R1 (3,5-bis-trifluoromethylphenyl) and R2 (p-chlorophenyl) groups connected by distinct linkers. CCG-977 has an aromatic linker with eight atoms separating the two R groups, whereas CCG-1423 has an aliphatic linker with six atoms between R1 and R2 (Fig. 2B).

To further assess whether CCG-977 and CCG-1423 selectively inhibit transcription induced by the Rho pathway, we examined the effect of these compounds on Gα₁₃Q226L-stimulated firefly luciferase expression driven by the SRE.L response element and Renilla expression from the constitutively active TK promoter in PC-3 prostate cancer cells. Both compounds inhibited SRE.L luciferase expression with IC₅₀ values of 1 to 5 μmol/L while only modestly inhibiting TK-driven Renilla expression (Fig. 3A and B). The extent of inhibition of Renilla expression correlated with inhibition of cell viability as detected by WST-1 absorbance (data not shown). Similar to the effect of CCG-977 and CCG-1423, latrunculin B potently inhibited the luciferase signal and also showed a modest effect on the TK Renilla signal at the highest concentrations (Fig. 3C). In contrast, the general inhibitor of transcription 5,6-dichlorobenzimidazole-1-β-d-ribofuranoside, which functions as an inhibitor of kinases of the RNA polymerase COOH-terminal domain, inhibited firefly and Renilla luciferase expression equally (Fig. 3D). Thus, CCG-1423 and CCG-977 have selective effects on SRF-mediated transcription activated by Rho pathway signaling in comparison with TK promoter-mediated transcription.

To determine the site of action of CCG-1423, we activated the Rho signaling pathway in PC-3 cells at multiple steps. CCG-1423 (10 μmol/L) inhibited SRE.L activation by heterotrimeric G proteins (Gα₁₂Q231L and Gα₁₃Q226L; Fig. 4A). [Control experiments testing the effect of CCG-1423 on activator protein expression (e.g., LARG, RhoA, and MKL1) could not be interpreted as we were unable to detect the expression of the tagged proteins in PC-3 cells by Western blots with concentrations of plasmids relevant to the luciferase assay.] This result indicates that the inhibitor is not specific for Gα₁₃ because Gα₁₂ signals are also affected. Notably, CCG-1423 does not seem to interfere with Rho activation per se (by targeting LARG for example) because acute treatment with CCG-1423 did not inhibit LPA- or serum-stimulated stress fiber formation in NIH3T3 mouse fibroblasts (Fig. 4B). CCG-1423, however, inhibited the activity elicited by expression of RhoA-G14V and RhoC-G14V. Because activation by these proteins is not dependent on upstream activators, this further supports an action of the compound at a downstream step. The ability to block signals initiated by both RhoA and RhoC indicates that the step inhibited by CCG-1423 is engaged by both proteins. This does not seem to be the Rho kinase (ROCK) because CCG-1423 does not inhibit ROCK kinase activity in vitro nor does the known ROCK inhibitor Y-27632 fully inhibit Gα₁₃-stimulated SRE-luciferase expression (Supplementary Fig. S1).⁴

Because the data above indicate that CCG-1423 does not interfere with upstream components of the pathway, we probed steps more proximal to the transcriptional machinery. To this end, we tested effects of the compound on the transcriptional response induced by several transcription factors/coactivators. SRE.L-driven transcription is dependent on the ability of SRF to bind to this DNA element and nucleate the assembly of productive transcription complexes. CCG-1423 could therefore act by interfering with SRE-SRF recognition or by altering SRF-specific mechanisms of transcriptional activation. To distinguish between these possibilities, we examined the ability of CCG-1423 to inhibit activity elicited by an SRF-VP16 fusion protein. This chimera depends on the SRF-SRE interaction for its recruitment to the promoter but can activate transcription through the VP16 activation domain. As can be seen in Fig. 4A, CCG-1423 failed to inhibit the activity elicited by this protein, suggesting that the compound does not interfere with SRE-SRF interactions and implies that it does not affect transcriptional activation pathways used by the VP16 activation domain. This last point is confirmed by the observation that CCG-1423 does not inhibit the transcriptional response elicited by the chimeric activator GAL4-VP16 at a promoter-bearing GAL4 sites (Fig. 4A).

To examine the effects of CCG-1423 on transcriptional activation mechanisms used by SRF, we tested the ability of this compound to inhibit SRE.L-driven transcription stimulated by expression of the SRF coactivator MKL. As can be seen in Fig. 4A, despite the fact that expression of MKL1 led to a robust activation, reaching levels comparable with those achieved with the SRF-VP16 fusion (31,100 ± 17,500 and 28,500 ± 13,200, respectively), CCG-1423 was able to inhibit this response effectively, whereas it had no effect on SRF-VP16–mediated activity. Together with the lack of effect on GAL4-VP16 or TK-driven Renilla luciferase expression, this result suggests that the compound specifically interferes with SRF/MKL1-dependent transcriptional activation mechanisms. The ability of CCG-1423 to inhibit MKL1-stimulated activity could be due to alterations in the recruitment of MKL to SRF or to effects on MKL1-dependent postrecruitment transcriptional activation mechanisms. As an initial step to explore these possibilities, we bypassed the SRF-dependent MKL recruitment step by fusing a COOH-terminal region of MKL, which harbors a strong activation function to the GAL4 DNA-binding domain. Activity of this fusion protein at a promoter-bearing GAL4 sites was partially inhibited by CCG-1423 (Fig. 4A). The partial nature of the response argues that CCG-1423 may function by altering steps both upstream of MKL recruitment as well as by interfering with postrecruitment functions of MKL at the promoter. Thus, in the context of the current view of the Rho signaling pathway shown in Fig. 1A, our overall analysis indicates that the site of action of CCG-1423 lies at a common step downstream of RhoA and RhoC distinct from the SRF-SRE interaction. Actions on some aspect of MKL1/SRF function (e.g., nuclear translocation, posttranslational modifications, MKL/SRF interaction, or MKL interactions with the transcriptional machinery) seem most likely.

The above data show that CCG-1423 exerts selective effects on Rho-stimulated transcription. To determine whether the influence of this compound extends to Rho-mediated cellular responses central to malignant cell growth and metastasis, we have examined the effect of this compound on the growth and invasiveness of cancer cell lines that differ in their Rho pathway properties. Initially, we examined the effects of CCG-1423 on proliferation in response to activation of the Rho pathway. LPA is a major mitogen acting through GPCRs. Specific LPA receptors can activate signaling through at least three different G protein families (Gα_i, Gα_q, and Gα₁₂/Gα₁₃). Gα_i signaling is sensitive to PTX, whereas Gα_q and Gα_12/13 signaling are not. The downstream effects are also distinct because Gα_i activates ras (48), Gα₁₃ strongly activates Rho, whereas Gα_q activation leads to both Rho and ras signals (49, 50). Our laboratory has previously shown that LPA stimulates Rho in PC-3 cells through PDZ-rhoGEF (51). We therefore examined the effect of CCG-1423 on LPA-stimulated DNA synthesis in PC-3 cells using a BrdUrd incorporation assay. CCG-1423 specifically inhibited LPA-stimulated DNA synthesis in a dose-dependent manner and completely suppressed this response at 3 μmol/L (Fig. 5A). In this context, PTX had no effect on LPA-stimulated DNA synthesis (data not shown). This is consistent with a model in which LPA activates Rho through non-Gα_i pathways, such as Gα₁₂/Gα₁₃ or Gα_q, which in turn leads to Rho pathway signaling events that are blocked by CCG-1423. Notably, over the 24-h time course of this study, the compound did not affect cell viability as measured by WST-1 metabolism, showing that, at these doses, the compound does not have acute nonspecific toxic effects on PC-3 cells (data not shown).

To assess the antiproliferative effects of CCG-1423, we analyzed the sensitivity of a panel of cell lines that differ in their Rho pathway properties. The panel includes melanoma lines that overexpress RhoC (A375M2 and SK-Mel-147) as well as related low RhoC-expressing lines (A375 and SK-Mel-28). In addition, we also examined transformed (SW962, PC-3, and SKOV-3) and nontransformed (WI-38) cell lines. The A375M2 line was originally derived from the A375 melanoma line by selection for high metastatic potential by two passages through a mouse lung metastasis system (52). In this line, RhoC mRNA expression is 3- to 5-fold higher than in the nonmetastatic cells remaining in the primary tumor site (52). The SK-Mel-147 line shows substantially higher RhoC protein expression compared with SK-Mel-28 cells as assessed by Western blot (45).

In an 8-day growth assay in the presence of LPA (Fig. 5B), a low concentration of CCG-1423 (300 nmol/L) markedly inhibited the proliferation of RhoC-overexpressing melanoma cells (A375M2 and SK-Mel-147) while not affecting their related low RhoC-expressing counterparts (A375 and SK-Mel-28). The squamous cell cancer line SW962, shown by Sahai and Marshal (22) to exhibit Rho-independent invasion, in contrast to the Rho-dependent A375M2, was similarly unaffected. PC-3 prostate cancer cells, SKOV-3 ovarian carcinoma cells, as well as the nontransformed WI-38 human fibroblasts were inhibited only marginally at this low concentration. However, growth of PC-3 prostate cancer cells was inhibited in this 8-day study with an IC₅₀ of 1 μmol/L in good agreement with its potency in effects on DNA synthesis and SRE-mediated gene transcription. The spectrum of activity of CCG-1423 in this panel of cell lines is consistent with its role as a Rho/SRF pathway inhibitor because its antiproliferative effects display selectivity toward RhoC-overexpressing melanoma cell lines. One potential mechanism for reduced growth in the presence of CCG-1423 could be enhanced apoptosis. We therefore measured caspase-3 activation in response to either CCG-1423 or daunorubicin in the RhoC-overexpressing A375M2 and low RhoC-expressing A375 melanoma cell lines. Compared with the low RhoC-expressing A375 line, CCG-1423 induced a 2-fold higher caspase-3 activity in the RhoC-overexpressing (A375M2) derivative (Fig. 5C, left). In this cell line, the activity induced by CCG-1423 was nearly as large as the effect of daunorubicin. In contrast to CCG-1423, the effect of daunorubicin (Fig. 5C, right) is substantially higher in the low RhoC-expressing A375 cells compared with the RhoC-overexpressing A375M2 cells. These data indicate that CCG-1423 displays selectivity toward Rho-overexpressing lines and the reversal of specificity of the two compounds (CCG-1423 versus daunorubicin) between A375 and A375M2 cells suggests a distinct mechanism of action for CCG-1423 compared with daunorubicin.

A critical process in metastasis is the ability of malignant cells to invade heterologous tissues. Recently, the invasiveness of PC-3 prostate cancer cells was shown to be dependent on RhoC and Gα₁₂ (11, 53). Thus, we compared the effects of CCG-1423 on matrix invasion by PC-3 prostate cancer cells and SKOV-3 ovarian cancer cells. In a Matrigel invasion assay, PC-3 cells display high invasion rates and do not require an exogenous stimulus for invasion (Fig. 6A, left). In contrast, invasion by SKOV-3 cells is greatly stimulated in the presence of LPA (Fig. 6A, right). Invasion in these two cell lines relies on different pathways because LPA-stimulated invasion by SKOV-3 cells was completely eliminated by PTX (Gα_i pathway dependent), whereas PC-3 cell invasion was not affected (Fig. 6B). Consistent with a Rho pathway inhibitor function, CCG-1423 (3 μmol/L) strongly suppressed the Rho-dependent invasion by PC-3 cells. A similar effect was seen on PC-3 cells in the presence of LPA (data not shown). In contrast, CCG-1423 did not affect the Gα_i pathway-dependent invasion by SKOV-3 cells (Fig. 6B).

Recent developments in cancer therapeutics have highlighted the importance of molecularly targeted therapies with a current major emphasis on tyrosine kinases (54, 55). However, there is a strong body of evidence supporting efforts to disrupt signaling by oncogenic GTPases, such as ras and Rho family members. These include blockade of critical lipid modifications by farnesyltransferase and geranylgeranyltransferase inhibitors, statins, and bisphosphonates (34). Inhibitors of Rho kinase have also been evaluated in cancer models (35, 37). In this report, we describe a novel small-molecule inhibitor, which disrupts transcriptional signaling by the RhoA GTPase family. The compound described here (CCG-1423) has nanomolar to low micromolar potency as well as selectivity toward Rho-overexpressing and invasive cancer cell lines for inhibition of DNA synthesis, cell growth, and/or invasion.

The role of RhoA family members, particularly RhoA and RhoC, in cancer cell growth and/or metastasis has become quite clear. Multiple studies correlate RhoA or RhoC overexpression and tumor aggressiveness (2, 3). Suppression of RhoC protein expression reduces invasiveness in vitro (11). Furthermore, breast tumors induced by the polyoma virus middle T antigen in a RhoC-deficient mouse show normal growth but a dramatic reduction in metastasis (12). The role of GPCR signaling to Rho is also relevant. A recent study found that a nonsynonymous polymorphism in the RH-containing rhoGEF, PDZ-rhoGEF, was associated with reduced lung cancer risk in a case-control study of Mexican-Americans (56). In addition, Yao et al. (11) report that RhoC plays an important role in Rho-dependent prostate cancer cell invasion. This result correlates with our observation that CCG-1423 inhibits PC-3 prostate cancer cell matrix invasion (Fig. 6B).

In the strongly Rho-overexpressing cell lines (A375M2 and SK-Mel-147), we see an effect of nanomolar concentrations of CCG-1423 on cell growth. Rho-mediated regulation of the cell cycle is complex, involving both Rho and ras pathway components, leading to effects on cyclins A and D1 and on the cyclin-dependent kinase inhibitors p21^Waf1/Cip1 and p27^kip1 (57–59). Interestingly, the induction of cyclin D1 by the Rho effector ROCK (59) could be due to Rho-mediated gene transcriptional events. Thus, differential dependence on these cell cycle regulators could underlie the functional differences in cell growth and DNA synthesis that we observe with CCG-1423 on the various cancer cell types.

Furthermore, the growth inhibition in the 8-day experiments could also involve enhanced cell death as well as inhibition of cell cycle progression. Consistent with this, caspase-3 activation in the highly metastatic RhoC-overexpressing A375M2 melanoma cell line was enhanced by CCG-1423 whereas a smaller increase was seen with the parental A375 cell line, whereas just the opposite pattern was seen with daunorubicin.

The precise molecular mechanism of action of CCG-1423 is not clear but our findings indicate that CCG-1423 may disrupt Rho signaling through functional inhibition of SRF transcriptional activity. The data suggest that effects on the coactivator MKL1 are likely, although effects on other SRF regulatory cofactors cannot be ruled out. Nuclear MKL1 function depends on Rho-mediated actin polymerization, which leads to dissociation from G-actin and nuclear translocation of MKL1. Free from G-actin, MKL1 can bind the transcription factor SRF and stimulate transcription (23, 24). MKL1 function is also regulated by posttranslational modifications, including covalent modification by members of the small ubiquitin-like modifier family. Consistent with the effects of sumoylation on sequence-specific transcription factors (60), small ubiquitin-like modifier modification of MKL1 attenuates its transcriptional activation potential (61). Thus, CCG-1423 could modify several aspects of MKL1 function, for example (a) preventing release from actin or blocking nuclear translocation of MKL1, (b) enhancing sumoylation causing transcriptional repression, (c) inhibiting the protein-protein interaction between MKL1 and SRF, or (d) disrupting MKL1 coactivator function. The ability of CCG-1423 to partially inhibit the GAL4-MKL1 transcriptional signal supports at least in part this latter point, although definitive conclusions will require additional experiments. Alternatively, CCG-1423 might disrupt some other aspect of SRF function, such as recruitment or activity of other SRF transcription partners (e.g., Nkx3.1 and GATA-4; refs. 62, 63).

The role of different Rho signaling events in cancer biology has been difficult to decipher because RhoA family members activate many effectors, such as Rho kinases (ROCK-I and ROCK-II), rhotekin, and mDia1. The downstream events from Rho signals are similarly diverse, with alterations in actin cytoskeletal organization and gene transcription being two prominent effects. Our identification of CCG-1423 as an inhibitor of Rho-mediated transcription provides a useful tool to further elucidate the role of this process in cancer biology. Given the strong inhibition of PC-3 cell matrix invasion and A375M2 cell growth, it is likely that Rho-stimulated transcriptional processes play a role in these phenomena. Indeed, enhanced Rho-dependent matrix metalloproteinase expression induced by the chemokine CXCL12 acting at CXCR4 receptors is involved in melanoma cell matrix invasion (64). It will clearly be of interest to identify which Rho-dependent genes show reduced expression with CCG-1423.

In summary, our identification of CCG-1423 provides a novel lead compound for the development of more potent and specific inhibitors of Rho-mediated transcription. CCG-1423 should serve both as a pharmacologic tool and as a potential lead for therapeutics for Rho-dependent cancers. Further exploration on the mechanism of action of CCG-1423 and identification of even more potent analogues will be important steps for the future.

We thank Dr. Peter Clapp (University of Michigan) for assistance with the actin staining studies; Sergey Chupetra for assistance with the GAL4 transcription studies; Drs. Li Li, Michael Parmacek, and Ron Prywes for kindly providing DNA plasmids; and Drs. J. Menon, Kenneth Pienta, Kathleen Cho, and Richard Hynes for kindly providing cell lines.