The motuporamines are macrocyclic alkaloids that inhibit tumor cell invasion by an, as yet, unknown mechanism. A structure–activity study recently identified dihydromotuporamine C (dhMotC) as a highly active and readily synthesized analogue. Here, we show that dhMotC causes subtle cytoskeletal alterations in highly invasive MDA231 breast tumor cells that include an increase in the thickness and number of cytoplasmic actin stress fibers. Experiments with serum-starved Swiss 3T3 fibroblasts showed that micromolar concentrations of dhMotC that inhibit tumor cell invasion induce the formation of new stress fibers and large focal adhesion complexes that are dispersed around the entire cell periphery. dhMotC treatment of Swiss 3T3 cells also initiates a strong, long-lived activation of the small GTP-binding protein Rho, and it stimulates Rho kinase-dependent sodium–proton exchanger activity. Liposome-mediated cell loading of C3 exoenzyme prevents dhMotC-mediated Rho activation and stress fiber formation in 3T3 cells. C3 exoenzyme loading also reestablishes elongated MDA231 breast tumor cell invasion in the presence of dhMotC. Taken together, these results indicate that the ability to activate Rho is one important determinant of the anti-invasive activity of dhMotC.
Cancer cell invasion of adjacent tissues and angiogenesis are critical contributors to metastatic tumor progression (1) . Because metastasis is responsible for the great majority of cancer deaths, drug development strategies have recently expanded to include the search for anti-invasive and antiangiogenic compounds that need not be directly cytotoxic to tumor cells (2 , 3) . Although molecular modeling and combinatorial chemistry are fueling a rapid escalation in the de novo generation of synthetic anticancer drugs, natural products continue to be an important reservoir of lead compounds that can be subsequently modified and improved upon (4) . Using a phenotype-based screen for inhibitors of invasion of metastatic MDA231 breast carcinoma cells, we recently identified a family of marine sponge alkaloids, the motuporamines, as highly active invasion inhibitors with low cytotoxicity (5 , 6) .
In three-dimensional basement gel culture, extract-purified motuporamine C (MotC; Fig. 1 ⇓ ) blocks the formation of elongated cellular protrusions that help initiate MDA231 cell invasion (5) . Additionally, in two-dimensional monolayer culture, MotC blocks the formation of actin-rich leading lamellae, and it prevents directed migration into wounds. MotC has similar phenotypic effects on vascular endothelial cells and thus prevents vascular endothelial growth factor-mediated endothelial cell sprouting in vitro and angiogenesis in vivo (5) . The fact that MotC is able to inhibit both elongated tumor cell invasion and angiogenesis makes the motuporamines attractive candidates for drug development. A structure–activity study recently identified dihydromotuporamine C (dhMotC) as a promising analogue for further study based on its potent activity, ease of synthesis, water solubility, and stability (6) . In this study, we show that dhMotC’s ability to activate Rho is an essential component of its anti-invasive activity.
MATERIALS AND METHODS
dhMotC was synthesized as described previously (6) . Because of slight differences between batches, optimal anti-invasive activity was found to range between 5 and 10 μm. Thus, these two concentrations were used interchangeably throughout the current study.
Cell Culture and Invasion Assays.
MDA231 breast carcinoma cells were routinely maintained in DMEM/F12 medium (Invitrogen, Carlsbad, CA) supplemented with 5% fetal bovine serum (FBS; Invitrogen) and 5 μg/ml insulin (Sigma, St. Louis, MO). Mouse Swiss 3T3 fibroblasts were maintained in DMEM medium (Invitrogen) supplemented with 10% FBS. Invasion inhibition assays were carried out as described previously (5) . Briefly, MDA231 cells were plated on reconstituted basement membrane gels (Matrigel; BD Biosciences, Mississauga, Canada) in the absence or presence of various compounds for 2–3 h, and elongated, invasive morphology was assessed by phase microscopy. The number of viable cells that did not invade the gel was then determined quantitatively; noninvasive cells were removed from the top of the gel by light trypsinization, allowed to reattach to tissue culture plastic for 18 h in separate wells, and viable cells were quantified using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Thus, unlike “classical” Transwell/Boyden chamber assays, this quantitative assay uses a positive readout for invasion inhibition that eliminates false positives caused by cytotoxicity. For experiments involving serum starvation, cells in subconfluent monolayer culture were maintained in unsupplemented medium for 18 h before the initiation of the described procedures.
Immunofluorescence and f-Actin Staining.
For tubulin and vimentin immunostaining, cells were fixed (3.7% formaldehyde in PBS, 15 min at room temperature), simultaneously permeabilized and blocked (0.1% Triton X-100 and 1% BSA for 30 min at 4°C), and then incubated with mouse monoclonal antibodies against either β-tubulin (1 μg/ml; Developmental Studies Hybridoma Bank, University of Iowa) or vimentin (10 μg/ml; Sigma). For vinculin immunostaining, cells were fixed (3.7% formaldehyde in PBS, 10 min at room temperature), permeabilized (0.6% Triton X-100 in PBS for 6 min at room temperature), blocked (10% FBS, 2% BSA in PBS for 30 min at room temperature), and then incubated with mouse monoclonal antivinculin (460 μg/ml; Sigma). Primary antibody binding was detected using CY3-conjugated goat antimouse IgG, F(ab′)2 fragment-specific antibodies (The Jackson Laboratory, West Grove PA). f-actin was visualized after fixation (3.7% formaldehyde in PBS, 10 min at room temperature), acetone extraction (100%, 5 min at −20°C), and air drying followed by staining with rhodamine-labeled phalloidin (0.5 units/ml; Molecular Probes, Eugene OR). Images were generated using a Nikon Eclipse E400 microscope equipped with an Imaging Microimager II digital camera.
Rho Activation Assays.
Activated, GTP-bound Rho proteins were isolated by coprecipitation with the Rho-binding domain of rhotekin and quantified by Western blotting according to the manufacturer’s instructions (Upstate Biotechnology, Lake Placid, NY) using reagents and protocols originally developed by Ren et al. (7) . Neither the coprecipitation with the recombinant rhotekin fragment nor the antibody used for Western blotting is specific for individual Rho isoforms. Thus, these experiments were designed to assess the aggregate activation state of all mammalian Rho isoforms. Blotting of unprecipitated whole cell lysates with the pan-Rho antibody was carried out to demonstrate that observed changes in activation state were not caused by changes in steady-state Rho protein levels.
C3 Exoenzyme Loading into Live Cells.
The Rho inhibitor C3 exoenzyme from Clostridium botulinum (Calbiochem, San Diego, CA) was introduced into cells using the method of Renshaw et al. (8) . Briefly, to load cells, 5–20 μg/ml C3 exoenzyme (from 500 μg/ml stock reconstituted in DMEM/F12 base medium) were mixed with 5 μg/ml lipofectin (Invitrogen) 1:1 in medium lacking FBS and incubated with either MDA231 or Swiss 3T3 cells for 14 h at 37°C before initiating the experiments described.
Measurement of Intracellular pH (pHi).
Serum-starved Swiss 3T3 cells on glass coverslips were loaded with the acetoxymethyl form of the fluorescent probe 2′,7′-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein (BCECF; 2 μm; Molecular Probes). After 30 min, the coverslip was mounted in a perfusion chamber, and cells were superfused at 2 ml/min for 15 min with standard, nominally HCO3−/CO2-free, perfusion medium containing 136.5 NaCl, 3 KCl, 2 CaCl2, 1.5 NaH2PO4, 1.5 MgSO4, 10 D-glucose, and 10 HEPES (all in mm) that was titrated to pH 7.35 with 10 m NaOH. Solutions containing 40 mm NH4Cl were prepared by equimolar substitution for NaCl. dhMotC and the Na+/H+ exchange inhibitor HOE 694 (a gift from Aventis Pharma, Frankfurt, Germany) were applied by superfusion. All pHi experiments were conducted at 36°C–37°C.
The dual-excitation ratio method was used to estimate pHi, using a digital fluorescence ratio imaging system (Atto Instruments, Inc., Rockville, MD). Details of the methods used have been presented previously (9 , 10) . In brief, fluorescence emissions measured at 520 nm were collected from multiple regions of interest placed on individual cells. Raw emission intensity data at each excitation wavelength (488 and 452 nm) were corrected for background fluorescence before calculation of the ratio, photobleaching was reduced using electronic arc attenuation and neutral density filters, and ratio pairs were acquired at 1–12-s intervals with analysis restricted to those cells that retained BCECF throughout the experiment. The one point high-[K+]/nigericin technique was used to convert background-corrected BCECF emission intensity ratios (BI488:BI452) into pHi values using nonlinear least-squares regression fits after background subtraction (9) .
The effects of pharmacological treatments on both steady-state pHi and rates of pHi recovery from internal acid loads imposed by the NH4+-prepulse technique were examined. The recovery of pHi after an NH4+ prepulse was fitted to a single exponential function, and the first derivative of this function was used to determine the rate of pHi change (dpHi/dt). Instantaneous rates of pHi recovery under control and test conditions were then plotted against absolute pHi values, compared statistically at corresponding values of pHi, and the data points were fitted by weighted nonlinear least-squares regression (10) . Data are reported as means ± SE, with the accompanying n value referring to the number of cell populations (i.e., coverslips) from which data were obtained. Statistical comparisons were performed with Student’s two-tailed t tests (paired or unpaired, as appropriate); significance was assumed at the 5% level.
dhMotC Induces the Formation of Stress Fibers and Adhesion Complexes.
During an initial characterization of MotC’s ability to inhibit elongated tumor cell invasion and migration, we noticed that it had a subtle effect on the cytoskeleton of MDA231 breast carcinoma cells (5) . The structurally similar dhMotC analogue (Fig. 1A) ⇓ is more easily synthesized than the parent MotC compound, because it lacks the double-bond in the nonpolar head group, but it retains potent anti-invasive (6) and antimigratory (Fig. 1B) ⇓ activities. Thus, we began our characterization of the mode of action of dhMotC by examining its effects on the major components of the MDA231 cytoskeleton.
dhMotC treatment of MDA231 cells maintained in monolayer culture slightly decreased the density of the perinuclear microtubule cage (Fig. 2, A and B) ⇓ but otherwise had little observable effect on cytoplasmic microtubule distribution. Additionally, there were no readily discernible differences in the vimentin intermediate filament cytoskeleton between untreated and treated cells (Fig. 2, C and D) ⇓ . These observations suggested that dhMotC was not a general disruptor of the cytoskeleton, a tentative conclusion that was further supported by an examination of the actin cytoskeleton. Specifically, dhMotC treatment appeared to increase the number and thickness of cytoplasmic actin-containing stress fibers (Fig. 2, E and F) ⇓ . As stress fibers often terminate in adhesion complexes (11) , we also examined the latter structures. Indeed, dhMotC treatment appeared to increase both the number and size of adhesion complex-associated accumulations of vinculin (Fig. 2, G and H) ⇓ and paxillin (data not shown).
Dynamic changes to the actin cytoskeleton and regulated assembly and disassembly of adhesion complexes help mediate the advancement of the leading edge and the retraction of the trailing edge of motile cells in monolayer culture (12) . These two processes have been very well characterized in serum-starved mouse fibroblasts (13) . Therefore, to unequivocally determine whether dhMotC specifically affects these processes, we examined the effect of the compound in the Swiss 3T3 mouse fibroblast line.
After they had been serum-starved for 18 h, Swiss 3T3 cells in subconfluent monolayer culture exhibited very few f-actin-containing stress fibers or vinculin-containing adhesion complexes (Fig. 3, A and B) ⇓ , but, as expected (13) , both were induced by the readdition of FBS (Fig. 3, C and D) ⇓ . As was the case with MDA231 cells, dhMotC treatment also induced stress fiber formation and focal adhesion formation in the serum-starved Swiss 3T3 cells. This induction was rapid (i.e., initiated within 15 min) and long lived (i.e., still maintained after 2 h of treatment). There were also qualitative differences between the serum- and dhMotC-induced structures: (a) dhMotC-induced stress fibers were generally thicker than those induced by serum (compare Fig. 3 ⇓ , E with C); and (b) although serum-induced adhesion complexes were concentrated at the leading (Fig. 3D ⇓ , arrows) and trailing (Fig. 3D ⇓ , arrowheads) cell edges, dhMotC-induced focal adhesions were rarely polarized and instead dispersed around the entire cell periphery (Fig. 3F) ⇓ . dhMotC-induced focal adhesions also tended to be thicker and blunter than those induced by serum (compare Fig. 3 ⇓ , D with F).
dhMotC Strongly Activates Rho.
A number of agents present in serum induce stress fiber and adhesion complex formation in Swiss 3T3 cells by transiently activating RhoA (13) , and various Rho isoforms have been implicated in cell migration (14) . Therefore, we assessed the effect of dhMotC on aggregate Rho activity using a rhotekin-based pulldown assay (see “Materials and Methods” for details). Clearly, dhMotC stimulated Rho activity in both MDA231 (Fig. 2I) ⇓ and Swiss 3T3 (Fig. 4A) ⇓ cells. In the latter cells, the activation occurred within 20 min, and it was sustained for ≥2 h.
The C3 exoenzyme from C. botulinum inhibits the activation of all three mammalian Rho isoforms by ADP ribosylation (15) . Using liposomal delivery, we preloaded Swiss 3T3 cells with C3 exoenzyme and found that it inhibited baseline Rho activity in serum-starved cells (Fig. 4B) ⇓ . It also strongly blocked the dhMotC-induced stimulation of aggregate Rho activity (Fig. 4B) ⇓ and stress fiber formation (Fig. 5) ⇓ . Therefore, the ability of dhMotC to alter the actin cytoskeleton in Swiss 3T3 cells is mediated, at least in part, by its ability to stimulate Rho activation.
dhMotC Increases Sodium–Proton Exchange.
The activation state of sodium–proton exchangers (NHEs), which can influence cell migration, is positively influenced by Rho signaling through the activity of the downstream effector Rho-kinase (ROCK; Refs. 16 and 17 ). Thus, in addition to its Rho-dependent effects on the actin cytoskeleton, we reasoned that dhMotC treatment might also activate Na+/H+ exchange activity in Swiss 3T3 cells.
Resting, steady-state pHi under nominally HCO3−/CO2-free conditions was 7 ± 0.06 (n = 6, a total of 38 cells), a value consistent with previous reports in Swiss 3T3 fibroblasts (18 , 19) . Under the same conditions, the recovery of pHi from imposed intracellular acid loads was reversibly inhibited by the selective NHE inhibitor HOE 694, applied at 50 or 100 μm (n = 7; Fig. 6A ⇓ ). These experiments established NHE as the principal mechanism involved in the recovery of pHi from internal acid loads imposed in Swiss 3T3 cells under our experimental conditions.
We next examined the effects of dhMotC on steady-state pHi and rates of pHi recovery from internal acid loads. The application of 5–10 μm dhMotC evoked, after a 3–5-min delay, an increase in steady-state pHi of 0.14 ± 0.02 pH units (n = 24), which reversed on washout of the compound (Fig. 6B) ⇓ . When internal acid loads were imposed after 5–8-min pretreatment with dhMotC, rates of pHi recovery were increased, compared with control rates of pHi recovery established in the same cells before dhMotC application (Fig. 6B) ⇓ . In six paired experiments of the type shown in Fig. 6B ⇓ , pHi recoveries from acid loads imposed in the presence of 5 μm dhMotC were compared with control rates of pHi recovery at the same absolute values of pHi. The resulting plots of the pHi dependence of rates of pHi recovery are presented in Fig. 6B ⇓ (inset); instantaneous rates of pHi recovery were increased significantly by 5 μm dhMotC at all absolute values of pHi. Similar results were obtained in six additional experiments with 10 μm dhMotC (data not shown). The increases in pHi evoked by NH4+ (quantified by taking the difference between the steady-state pHi immediately before the application of NH4+ and the maximum pHi observed during its application; see Ref. 10 ) were similar before and during dhMotC application (0.6 ± 0.03 and 0.56 ± 0.04 pH units, respectively; n = 12 in each case; P > 0.05), suggesting that marked alterations in intracellular buffering power are unlikely to contribute to the increase in rates of pHi recovery observed during dhMotC application. Furthermore, the augmented rates of pHi recovery from acid loads imposed in the presence of dhMotC continued to be sensitive to 50 or 100 μm HOE 694 (n = 3 and 6, respectively; Fig. 6C ⇓ ), confirming that dhMotC increased pHi recovery by stimulating NHE activity.
In the final series of experiments, Swiss 3T3 cells were pretreated for 60 min with the specific ROCK inhibitor Y-27632 (20) . In three experiments, 30 μm Y-27632 prevented 5 μm dhMotC from increasing rates of pHi recovery from imposed acid loads, e.g., at a common test pHi of 6.45, the rate of pHi recovery obtained in the presence of dhMotC and Y-27632 was significantly slower than the rate of pHi recovery obtained in the presence of dhMotC alone and was not significantly different from the rate of pHi recovery observed in the absence of both compounds (Fig. 6D) ⇓ . Interestingly, however, pretreatment with Y-27632 failed to affect the rise in steady-state pHi evoked by 5 μm dhMotC. Specifically, dhMotC increased steady-state pHi by 0.16 ± 0.05 pH units (n = 3) after pretreatment with 30 μm Y-27632, an increase that was not significantly different from that induced by dhMotC alone (see above).
The data described above indicate that dhMotC causes a Rho/ROCK-dependent increase in NHE activity and ROCK-independent increase in steady-state pHi in Swiss 3T3 cells. Although the mechanism responsible for the ROCK-independent increase in steady-state pHi is unknown, the mechanism responsible for increased ROCK-dependent NHE activity after an acid load is well known, and it involves changes to the actin cytoskeleton that regulate the subcellular localization of the actin-binding ezrin/radixin/moesin proteins, which also bind to and functionally modulate the NHE complex at the membrane (21) . Therefore, the increase in NHE activity observed here likely occurs as a consequence of the dhMotC-mediated increase in Rho activity.
dhMotC-Mediated Inhibition of Elongated Tumor Cell Invasion Is Rho Dependent.
When they are plated on top of Matrigel, highly invasive MDA231 breast carcinoma cells rapidly elongate and send out processes (i.e., “invadopodia”) that move into the gel within a few hours (Fig. 7A) ⇓ . The formation of these elongated invadopodia, which helps facilitate the movement of MDA231 cells into the basement membrane matrix, was not affected by preloading with the C3 exoenzyme (Fig. 7B) ⇓ . This finding supports the recent observation by Sahai and Marshall (22) that Rho inhibition does not block elongated cellular invasion in other tumor lines.
In contrast to the ineffectiveness of C3 exoenzyme, elongated cellular invasion is prevented by the parent MotC compound (5) , and it was prevented by dhMotC as well (Fig. 7C) ⇓ . As a result, MDA231 cells treated with dhMotC remained rounded and noninvasive on top of the basement membrane matrix. Strikingly, this morphological effect of dhMotC was prevented by preloading the cells with C3 exoenzyme (Fig. 7D) ⇓ . The ability of C3 exoenzyme to abrogate the anti-invasive effects of dhMotC was confirmed in our quantitative anti-invasion assay (Fig. 8) ⇓ . Taken together, these data demonstrate that dhMotC’s ability to activate Rho is a critical component of its anti-invasive activity.
At the onset of metastasis, carcinoma cells must first break away from the primary tumor and move through the basement membrane that surrounds all epithelial tissues (23) . Thus, movement through Matrigel is often used to assess tumor cell invasion in vitro (24) . We developed previously a high-throughput, quantitative variation of the classical Matrigel/Transwell/Boyden chamber assay to guide the purification of MotC from a marine sponge extract (5) . Subsequently, we used the same assay to identify dhMotC as a potent, easily synthesized anti-invasive analogue (6) . Our original characterization of the motuporamines suggested that they cause subtle changes to cell adhesion and the cytoskeleton, but otherwise, these initial studies provided no clues about their mode(s) of action.
We have now demonstrated that dhMotC does not globally disrupt microtubules, intermediate filaments, or microfilaments but instead induces the formation of stress fibers caused by its ability to activate Rho. dhMotC treatment also stimulated NHE activity, which, at first glance, appears paradoxical because NHEs act as membrane anchors that physically interact with ezrin/radixin/moesin proteins to facilitate directed cell migration under some conditions (21) . However, NHE activation also facilitates the formation of stress fibers and focal adhesions through this ezrin/radixin/moesin-mediated linkage (16 , 21 , 25) , both of which were also induced by dhMotC treatment. Therefore, NHE activation, which was ROCK dependent, appears to represent another legitimate Rho-mediated end point of dhMotC action. Interestingly, an associated increase in steady-state pHi was not prevented by pharmacological ROCK inhibition. Thus, dhMotC may have additional effects on pH metabolism that are not mediated by its ability to activate Rho. This possibility is currently under investigation.
Rho activity, stress fibers, and focal adhesions are all required for contractility-mediated traction events that pull the cell forward during directed migration in monolayer culture (12 , 14 , 26) . However, hyperactivation of Rho, which can be initiated by directly manipulating Rho activity (27) , misregulating chemoattractant-mediated receptor activation (28) , altering integrin-mediated adhesion (29 , 30) , or knocking out the focal adhesion kinase (31 , 32) , can inhibit migration. Under these conditions, the cells rarely form leading lamella or trailing edges but instead often form extensive stress fiber networks and large, stable, peripherally located focal adhesion complexes (27, 28, 29, 30, 31, 32) , which is precisely the phenotype observed in dhMotC-treated cells maintained in monolayer culture (Fig. 9) ⇓ .
MDA231 breast carcinoma cells extend protrusive invadopodia on Matrigel. This is an example of the recently characterized “elongated” tumor cell invasion (22 , 33) . Like directed cell migration in two-dimensional monolayer culture, elongated tumor cell invasion in three-dimensional culture is prevented by dhMotC. As a result, dhMotC-treated cells remain rounded, but attached and viable, on top of the gel. Importantly, the abilities of dhMotC to prevent tumor cell elongation and quantitatively inhibit tumor cell invasion were both abrogated by preloading of cells with the Rho-inhibiting C3 exoenzyme. This indicates that dhMotC’s ability to activate Rho blocks invasion, at least in part, by preventing the tumor cells from extending elongated protrusions into the basement membrane matrix.
Is Rho activation desirable from a therapeutic perspective? In a number of cases, Rho overexpression and/or pathway activation has been positively correlated with malignant progression (34, 35, 36, 37, 38) , which suggests that Rho inhibitors, rather than Rho activators, might show antimetastatic activity. However, the efficacy of either approach may be tumor specific. As was described above, MDA231 breast carcinoma cells must elongate to invade. The same is true of BE colon and S2962 squamous carcinoma cells (22) . Importantly, all three of these tumor cell lines continue to invade when Rho signaling is down-regulated (Fig. 7 ⇓ and Ref. 22 ). Thus, metastatic tumors which use “elongated” invasion may be sensitive to Rho activation by pharmacological agents. This may explain the selective antitumor activity of farnesyl transferase inhibitors, which, although originally designed to attenuate the action of Ras proteins, are now believed to act, at least in part, by activating RhoB (39 , 40) . Conversely, metastatic tumors whose cells invade using a rounded, amoeboid movement are clearly susceptible to Rho inhibition (22) . Differences in invasion morphology also confer a differential sensitivity to protease inhibition (22 , 41) . Interestingly, the parental MotC compound was also able to block the invasion of PC-3 prostate carcinoma cells (5) , which invade with a rounded morphology. This suggests that the motuporamines mediate additional anti-invasive effects that are not mediated by Rho activation alone.
When Rho alone is activated by dominant active mutant expression in tumor cells that use the elongated mode of invasion, these cells undergo a phenotypic switching and begin to invade basement membrane matrices with a rounded morphology (22) . dhMotC only partially induced this phenotype switch. As mentioned above, elongated MDA231 cells become rounded but not invasive. This further implicates additional signaling events in the anti-invasive activity of dhMotC. One possibility is the small GTPase Rac, which is required for both elongated and rounded tumor cell invasion (22) . This idea is attractive because highly adhesive cell matrix interactions that lead to hypertrophic focal adhesion formation and sustained Rho activation in nontumorigenic cells can act to dampen Rac activity in normal cells (29) . Thus, we are currently investigating dhMotC-mediated changes in Rac activity in migratory monolayer and invasive Matrigel culture.
The original MotC parent compound is antiangiogenic in vivo (5) , and preliminary data suggest that both MotC and dhMotC also have antitumor activity in vivo. 5 The motuporamines show some structural resemblance to the antiangiogenic/antitumor agent squalamine (42 , 43) . However, dhMotC activates Na+/H+ exchange downstream of Rho, whereas squalamine inhibits it (44) . As is the case with tumor type-specific anti-invasion responses to Rho activation, the antiangiogenic response in the face of NHE activation may be based on the morphological target of the motuporamines. Specifically, MotC prevents vascular endothelial growth factor-dependent endothelial cell sprouting into fibrin gels (5) , which morphologically mimics elongated tumor cell invasion into Matrigel. Interestingly, a recent analysis demonstrated that four structurally dissimilar antiangiogenic compounds also induce hypertrophic stress fiber and focal adhesion formation (45) , which suggests that they may have biological modes of action in common with the motuporamines. Thus, elongated endothelial cell sprouting may be susceptible to Rho/ROCK-dependent NHE activation, whereas other aspects of angiogenesis may be prevented by an inhibition of the same pathway as occurs in response to squalamine treatment.
It is possible that dhMotC activates the Rho pathway by directly targeting Rho regulators. This does occur in nonmigratory focal adhesion kinase knockout fibroblasts where Rho signaling is activated because of an inhibition of the negative regulator p190RhoGAP (32) . Alternatively, dhMotC could impinge on downstream signaling end points of parallel pathways that then feedback to modulate Rho activity with functional consequence. An example of such a modulator is the Fos family member Fra-1, which is a mitogen-activated protein kinase pathway end point that, when lost, stimulates integrin signaling, which then activates Rho to inhibit tumor cell motility (30) . Interestingly, we have recently used a yeast screen to demonstrate that dhMotC confers chemical-genetic synthetic lethality on genes involved in sphingolipid metabolism. 6 These data suggest that dhMotC might target signaling pathways upstream of Rho by modulating plasma membrane dynamics. Regardless its direct target(s), it is clear from the present study that the ability of dhMotC to functionally activate Rho is a critical component of its anti-invasive properties in vitro and, presumably, its anticancer activity in vivo.
We thank Hilary Anderson for many helpful experimental suggestions and a critical reading of this manuscript.
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.
Requests for reprints: Calvin D. Roskelley, Department of Anatomy and Cell Biology, University of British Columbia, 2177 Wesbrook Mall, Vancouver, B.C. V6T 1Z3, Canada. Phone: (604) 822-0779; Fax: (604) 822-2316; E-mail:
↵5 A. Minchinton and R. Andersen, unpublished observations.
↵6 K. Baetz and M. Roberge, unpublished observations.
- Received August 29, 2003.
- Revision received December 1, 2003.
- Accepted December 16, 2003.
- ©2004 American Association for Cancer Research.