The Invasion Inhibitor Sarasinoside A1 Reverses Mesenchymal Tumor Transformation in an E-Cadherin–Independent Manner

This article is featured in Highlights of This Issue, p. 441

Tumor cell invasion of the extracellular matrix (ECM) is an important driver of metastasis (1). Although target-based approaches have been used to identify a number of compounds with anti-invasive properties (2), we and others have developed phenotypic screens to identify structurally diverse invasion inhibitors that most often act by modulating cellular interactions with the ECM (3). In this report, we describe a newly identified invasion inhibitor that is also capable of facilitating cell–cell interactions.

The overwhelming majority of solid human tumors arise within epithelial tissues that consist of cohesive, stationary cells. Normally, highly regulated epithelial-to-mesenchymal transformations (EMT) generate noncohesive migratory cells that facilitate morphogenetic tissue rearrangements during organ development. However, as solid tumors progress, this process is often co-opted to generate noncohesive, mesenchymal tumor cells that are not subject to normal regulatory controls and are thus highly invasive (4).

Most approaches to controlling tumor invasion have focused on controlling the motile phenotype in relative isolation. Thus, highly selective inhibitors of potential motility factors such as TGF-β1, EGF, and endothelin have been developed (5–7). In addition, compounds that specifically disrupt integrin-dependent interactions with the ECM have been used with widely varying degrees of preclinical and clinical efficacy (2). In an effort to identify compounds that inhibit invasion by novel means, we developed a cell-based phenotypic screen that eliminates false-positives due to cytotoxicity (8). We then used this assay to identify several structurally diverse invasion inhibitors that act by modulating cellular interactions with the ECM without directly affecting integrin–ligand interactions (8–13). In this report, we describe a newly identified invasion inhibitor, the triterpenoid saponin sarasinoside A1.

During normal development, an EMT is initiated by a transcriptionally mediated downregulation of E-cadherin expression that leads to a dismantling of cell–cell junctions. Thus, considerable effort has been expended to develop strategies that can specifically reverse this change in gene expression in tumor cells (4). However, in a number of tumor contexts, E-cadherin is permanently lost due to mutation or stable epigenetic silencing. Thus, identifying compounds that can suppress the mesenchymal phenotype in the absence of E-cadherin could facilitate the development of new antimetastatic therapeutic strategies. Here, we report that sarasinoside A1 may be just such a compound as it re-epithelializes mesenchymally transformed tumor cells and initiates the formation of nascent cell–cell junctions in an E-cadherin–independent manner.

The marine sponge Melophlus sarassinorum was collected by hand, using scuba, from Papua New Guinea. Samples were deep frozen on site, transported on dry ice and methanol extracts were prepared. Sarasinoside A1 was isolated from the sponge extract by assay-guided fractionation, using the invasion assay described below. Briefly, the aqueous fraction from the original methanol extract was concentrated, taken up in H₂O, and extracted with nBuOH. The resulting active fraction was subjected to LH20 chromatography in MeOH to give a cream solid which was further purified by elution through a C18 sep-pak (gradient of CH₃CN in 0.05% trifluoracetic acid, TFA). Active fractions were combined and subjected to reverse-phase high-performance liquid chromatography (HPLC; Inertsil C18 column, 40% CH₃CN in 0.05% TFA). This generated sarasinoside A₁ (Fig. 1A) as a white amorphous solid that was soluble in dimethyl sulfoxide (DMSO) and which generated NMR data consistent with literature values (14–16).

MDA-MB-231 cells were originally acquired from Erik Thompson (University of Melbourne, Melbourne Australia) and were routinely tested for an absence of epithelial markers, including E-cadherin, and mesenchymal invasiveness as shown here under control conditions (see Figs. 1–3). A375 cells were from American Type Culture Collection (ATCC). Both of these cell lines were maintained in monolayer culture in Dulbecco's Modified Eagle's Medium (DMEM; Invitrogen) containing 10% FBS (Hyclone). MCF7 cells were grown in DMEM/F12 (Sigma) with 5% FBS and SKOV3 cells were maintained in 199/105 medium (Sigma) with 5% FBS. All cells were grown at 37°C in a humidified 5% CO₂ incubator. The pIRM2.1-IRES-dsFP593 vector encoding FLAG-Rap1V12 was from Michiyuki Matsuda (Kyoto University, Kyoto, Japan). The pEGFP-N1-Nectin2 construct was from Julian Guttman (Simon Fraser University, Burnaby, Canada), actin-GFP was from Invitrogen, and paxillin-GFP was kindly provided by Robert Nabi (University of British Columbia, Vancouver, BC, Canada). Cells were transfected transiently using Lipofectamine 2000 (Invitrogen).

Invasion inhibition assays were conducted as previously described and validated (8). Briefly, tumor cells were plated on reconstituted basement membrane gels (Matrigel; BD Biosciences) along with sarasinoside A1 or DMSO as a vehicle control. After 2.5 hours, noninvasive cells were recovered from the top surface of the Matrigel, and the number of viable cells was quantified using a MTT assay.

To measure cell viability, cells were plated in 96-well plates to near confluency. Sarasinoside A1 was added at various concentrations in triplicate, and cell viability was measured 24 hours later by MTT assay. To determine cell proliferation, cells were plated in 96-well plates at 5 × 10³ cells/well and incubated with different concentrations of sarasinoside A1, with daily change of medium and drug. Cell proliferation was measured at 0, 24, and 72 hours using an MTT assay.

For basement membrane invasion, cells were plated on top of, or were embedded within, reconstituted basement membrane gels (Matrigel, BD Biosciences) as previously described (17) in the absence or presence of sarasinoside A1 as indicated and then imaged, live, by phase microscopy. For collagen invasion, CMFDA-labeled cells (Invitrogen) were plated on collagen I (2 mg/mL in PBS; BD Biosciences) that was polymerized on Transwell filters (BD Falcon). The lower chamber contained media with 55% FBS as a chemoattractant. After 20 hours, filters were fixed with paraformaldehyde, embedded in optimal cutting temperature compound (OCT), cryosectioned vertically, stained for actin (see below), and z-axis images were acquired as shown.

Cells were plated on coverslips in serum-containing medium and maintained for 48 hours. Medium containing sarasinoside A1 or DMSO was then added to the cells for the specified lengths of time before fixation. Cells were fixed with 3% paraformaldehyde for 15 minutes at room temperature, permeabilized with 0.3% Triton X-100 in PBS for 15 minutes, and blocked with 3% bovine serum albumin (BSA; Sigma) in PBS for 30 minutes, followed by incubation with the following primary antibodies: mouse clone 14 β-catenin antibody (BD Transduction) at 1:100 dilution, rabbit polyclonal pan-cadherin antibody (Abcam ab16505) at 1:250, mouse clone 36 E-cadherin antibody (BD Transduction) at 1:100, and rabbit polyclonal ZO-1 antibody (Invitrogen) at 1:100, incubated overnight at 4°C. An Alexa Fluor 488–conjugated goat secondary antibody (Molecular Probes) was used at a dilution of 1 in 150. To visualize F-actin, cells were stained with a 165 nmol/L solution of rhodamine-phalloidin (Invitrogen). Nuclei were stained with a 500 ng/mL solution of 4′,6-diamidino-2-phenylindole (DAPI) dihydrochloride (Sigma). All images were obtained using the Uplan Apochromat 60×/1.35 numerical aperture objective on an Olympus FV1000 confocal microscope and captured using Olympus FluoView v1.6 software.

Cells were treated with either DMSO or 20 μg/mL sarasinoside A1 for the indicated times. Cells were lysed in radioimmunoprecipitation (RIPA) buffer containing protease inhibitors [50 mmol/L Tris, pH 7.5, 150 mmol/L NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 20 μg/mL aprotinin, 10 μg/mL leupeptin, 100 μmol/L phenylmethanesulfonylfluoride (PMSF), 10 μg/mL pepstatin A, 2.5 mmol/L EDTA] at 4°C for 30 minutes with agitation and then clarified through centrifugation at 10,000 × g for 15 minutes at 4°C. All samples for electrophoresis were quantified with a BCA protein assay (Pierce) and equalized for total protein before electrophoresis. The following primary antibodies were used: mouse clone AC-40 actin (Sigma) at 1:500 dilution, mouse clone 14 β-catenin (BD Transduction) at 1:500 dilution, rabbit polyclonal pan-cadherin (Abcam ab16505) at 1:500, mouse clone 36 E-cadherin (BD Transduction) at 1:1,000, and rabbit polyclonal ZO-1 (Invitrogen) at 1:500.

MDA-MB-231 cells on 1- μm pore culture inserts (Falcon 35-3104) were treated with 20 μg/mL sarasinoside A1 or DMSO for 24 hours. Cells were fixed (1.5% paraformaldehyde, 1.5% glutaraldehyde, 0.1 mol/L sodium cacodylate, pH 7.3) at room temperature for 3 hours, placed in buffer (0.1 mol/L sodium cacodylate, pH 7.3) overnight at room temperature, post-fixed for 1 hour on ice (1:1 mixture of 2% OsO₄ in ddH₂O and 0.2 mol/L sodium cacodylate, pH 7.3), washed 3× with ddH₂O at room temperature, stained en bloc for 1 hour (1% aqueous uranyl acetate), and washed again 3× with ddH₂O. Finally, samples were dehydrated through a graded series of ethyl alcohol, passed through propylene oxide into a 1:1 mixture of propylene oxide and EMbed-812 (Electron Microscopy Sciences), followed by overnight infiltration and embedding in EMbed-812. Thin sections were stained with uranyl acetate and lead citrate and photographed on a Philips 300 electron microscope operated at 60 kV.

Fluorescence recovery after photobleaching (FRAP) was conducted on MDA-MB-231 cells expressing actin-GFP, paxillin-GFP, or Nectin2-GFP seeded on chamber slides (Ibidi). The indicated regions of interest were photobleached using a 405-nm laser (100% intensity, 0.1 seconds) on an Olympus FV1000 confocal microscope and captured using Olympus FluoView v1.6 software. Fluorescence recovery was followed at 4- or 5-second time intervals for at least 1 minute. The fluorescence signal was normalized to the prebleach intensity, and single exponential fit curves of the data were generated as previously described (18).

Activated, GTP-bound Rho proteins were isolated by precipitation with the Rho-binding domain of rhotekin and quantified by Western blotting according to the manufacturer's instructions (Millipore). Rap1-GTP was precipitated from cell lysates using a glutathione S-transferase–RalGDS fusion protein and detected by immunoblotting with a Rap1 antibody as described (19).

We previously developed a cell-based phenotypic screening assay and used it to identify a number of tumor cell invasion inhibitors (8–13). In this report, we used this screen to identify the triterpenoid saponin sarasinoside A1 as an invasion inhibitor based on the assay-guided fractionation of a marine sponge extract (Fig. 1A). Specifically, sarasinoside A1 inhibited the invasion of highly metastatic and mesenchymal MDA-MB-231 breast cancer cells into a reconstituted basement membrane matrix (i.e., Matrigel) at a concentration of 20 μg/mL (Fig. 1B). The same concentration also inhibited MDA-MB-231 cell migration on a planar, rigid tissue culture plastic substratum in 2-dimensional (2-d) monolayer culture (Fig. 1C). At this dose, sarasinoside A1 was not cytotoxic, although it was reversibly antiproliferative (Supplementary Fig. S1A). It is important to note, however, that higher concentrations of sarasinoside A1 did cause some cytotoxicity (LC₅₀ = 53 μg/mL; Supplementary Fig. S1B) as has been reported previously in studies in which its effect on cell movement was not assessed (16, 20).

When they were plated on top of prepolymerized Matrigel, mesenchymal MDA-MB-231 cells elongated as they began to invade the matrix (Fig. 2A, DMSO vehicle control). In contrast, sarasinoside A1–treated MDA-MB-231 cells remained rounded on top of the gel (Fig. 2A, SaraA image). This anti-invasive, morphologic effect was also observed when highly invasive A375 melanoma and SKOV3 ovarian carcinoma cells were plated on top of Matrigel in the presence of sarasinoside A1 (Supplementary Fig. S2).

When they were embedded within the Matrigel, MDA-MB-231 cells also elongated as they became invasive and mesenchymal (Fig. 2B, DMSO vehicle control). Interestingly, under these conditions, the sarasinoside A1–treated cells did not elongate. Instead, they clustered together and formed spheroidal structures which is a characteristic of nonmetastatic, epithelialized tumor cells (Fig. 2B, SaraA image; ref. 21). In our experience, this cell clustering in 3-D culture, which is suggestive of increased cell–cell adhesion, is not induced by any of the other tumor cell invasion inhibitors we have identified (data not shown).

To determine whether the anti-invasive cell clustering described above was restricted to highly flexible, low-compliance matrices such as Matrigel, MDA-MB-231 cells were also embedded within more rigid type I collagen gels (Fig. 2C). Under the latter conditions, sarasinoside A1 also inhibited invasive cell elongation and led to the formation of small cohesive cell clusters (Fig. 2C, arrowhead). When untreated MDA-MB-231 cells were instead plated on top of the collagen gels and their invasive behavior was directionally stimulated by a chemoattractant (FBS) delivered from below, the cells pulled apart from each other and moved into the gel as occurs at the invasive front of mesenchymally transformed tumor masses in vivo (Fig. 2D, DMSO vehicle, invasion occurred downward into the collagen gel represented by the bar). In contrast, sarasinoside A1–treated cells formed a cohesive cell mass that resembled a stratified epithelium (Fig. 2D, SaraA image), and they remained stationary and noninvasive on top of the gel.

When MDA-MB-231 cells maintained as 2-D monolayers on tissue culture plastic were treated with sarasinoside A1, the mesenchymal, spindle-shaped cells underwent a flattening and a morphologic epithelialization (i.e., they formed “cobblestone”-like monolayers; Fig. 3A). As a result, after 24 hours, the sarasinoside A1–treated monolayers resembled the monolayers that were formed by MDA-MB-231cells that had been reverted to an epithelial phenotype in response to the stable, forced expression of exogenous E-cadherin (Fig. 3A, top phase, MDA231/E-Cad stable transfectants).

In untreated MDA-MB-231 cells, considerable filamentous actin was concentrated at the terminal, lamellar edges of the migratory, mesenchymal cells (Fig. 3A, f-actin, left image, arrowheads). Sarasinoside A1 treatment eliminated these structures, and the majority of the F-actin instead relocalized to sites of cell–cell interaction (Fig. 3A, F-actin panels after 4- and 24-hour treatment). This F-actin relocalization was not associated with an increase or accumulation of endogenous E-cadherin in the sarasinoside A1–treated cells (Fig. 3A; E-cadherin; MDA231/ECad transfectants served as a positive control). This lack of endogenous E-cadherin upregulation in sarasinoside A1–treated cells was confirmed by Western blotting (Fig. 3B). Importantly, the sarasinoside A1–induced epithelialization of MDA-MB-231 cells also occurred in the presence of cycloheximide or actinomycin D, inhibitors of protein translation and transcription, respectively (data not shown). Thus, sarasinoside A1 appears to have caused a morphologic mesenchymal-to-epithelial reversion that did not require global changes in gene expression or a specific upregulation of E-cadherin.

MDA-MB-231 cells do not express the classical epithelial cell–cell junction molecule E-cadherin due to epigenetic silencing (22). They do, however, express other cadherins, including R-cadherin and cadherin-11. As often occurs in mesenchymally transformed tumor cells, these cadherins do not initiate cell–cell junction formation in MDA-MB-231 cells. Thus, as expected, immunofluorescent staining indicated that these nonepithelial cadherins were diffusely localized in control cells (Fig. 4A, DMSO vehicle control). In contrast, sarasinoside A1 treatment caused these endogenously expressed cadherins to relocalize to sites of cell–cell contact together with the cell junction–associated proteins β-catenin and ZO-1, none of which exhibited a demonstrable change in their steady-state protein levels (Fig. 4A and B). These observations reinforce the notion that the sarasinoside A1–mediated re-epithelialization does not rely on changes in gene expression.

Nectins are Ca²⁺-independent Ig-like cell adhesion molecules that facilitate cell–cell adhesion in diverse homotypic and heterotypic cell–cell interactions (23). Nectin-based adhesion complexes, which are often disrupted in tumor cells, promote the initiation and stabilization of cadherin-mediated adherens junctions by facilitating the homophilic binding of cadherins expressed on the surface of juxtaposed cells. To determine whether sarasinoside A1 could re-localize exogenous nectins, nectin2-GFP was expressed in MDA-MB-231 cells. Similar to endogenous cell junction proteins in these tumor cells, nectin2-GFP was diffusely localized in control cells but it was recruited to sites of cell–cell contact after sarasinoside A1 treatment (Fig. 4A). In addition, transmission electron microscopy indicated that neighboring sarasinoside A–treated cells had very closely apposed plasma membranes compared with controls (Fig. 4C), morphologic evidence of adhesion between cells. Taken together, these data suggest that sarasinoside A1 initiates nectin-based cell–cell junction formation in MDA-MB-231 cells.

The formation of nectin-based adhesion complexes can result in activation of the Rap1 GTPase, an event that is required for adherens junction formation (24). The activation of Rap has also been shown to stabilize both- cadherin (25) and integrin-based adhesions (18) in an actin-dependent manner, which reduces the invasive capacity of many types of tumor cells. In contrast, activation of Rho GTPases can drive tumor cell migration and invasion by increasing cellular contractility and focal adhesion dynamics (26, 27). We therefore asked whether treating MDA-MB-231 cells with sarasinoside A1 altered the activation of Rho and Rap GTPases. We found that sarasinoside A1 had no effect on Rho activation but substantially increased the amount of activated Rap1 in the cells (Fig. 5A, B).

To test the idea that sarasinoside A1 alters the dynamics of de novo adhesion formation in a Rap-dependent manner, we expressed nectin2-GFP and monitored FRAP at the membrane. Nectin-based contacts between DMSO-treated cells were rare and, when formed, were highly dynamic (Fig. 5C). These contacts often disassembled during the imaging period (data not shown) and those that were maintained showed a high recovery of nectin2-GFP that is indicative of either trafficking-mediated turnover or lateral mobility of nectin within the membrane (Fig. 5C, left). In contrast, in sarasinoside A1–treated cells, we frequently observed nectin2-GFP at cell–cell contacts and the nectin2-GFP at these adhesions was stable, that is, it exhibited a reduced mobile fraction when bleached compared with the nectin2-GFP in control cells (Fig. 5C, middle). Because sarasinoside A1 potently activated the Rap GTPases, we asked whether expressing a constitutively active form of Rap1 (Rap1V12) was sufficient for stabilizing nectin-based adhesions between MDA-MB-231 cells. Indeed, expressing Rap1V12 also led to a stabilization (i.e., reduced mobile fraction) of nectin2-GFP in the membrane at these contact sites, and this stabilization was not further enhanced by treating the cells with sarasinoside A1 (Fig. 5C, right). This suggests that sarasinoside A1 stabilizes nectin-based junctional complexes, at least in part, due to its ability to activate Rap1.

When cell–cell junctions form, networks of filamentous actin (F-actin) are stabilized at the cytoplasmic face of the junctional complex where they act to physically link neighboring cells across the junction (28). To assess F-actin localization and stability, we used FRAP. Actin-GFP was expressed in MDA-MB-231 cells and regions of actin at sites of cell–cell contact were bleached. In control cells, actin-GFP was rarely incorporated into F-actin filaments that accumulated at small punctate sites of cell–cell interaction (Fig. 6A, DMSO, vehicle control). In contrast, sarasinoside A1–treated cells exhibited long, continuous accumulations of F-actin at sites of cell–cell interaction (Fig. 6A). In addition, FRAP-based recovery of actin-GFP intensity was significantly reduced at these sites in sarasinoside A1–treated cells, which is indicative of increased stability due to the maintenance of the photobleached actin (Fig. 6A). Interestingly, sarasinoside A1 did not increase the stability of cell–ECM contacts (i.e., focal adhesions), as there was no difference in the fluorescence recovery after bleaching of transfected paxillin-GFP, an actin-binding structural component of the focal adhesions (Fig. 6B). Taken together, these data show that sarasinoside A1 selectively increased the stability of F-actin networks that underlie the de novo cell–cell adhesions that formed between cells that would normally have been mesenchymal and invasive.

The loss of E-cadherin initiates an EMT in many cancers that arise within the epithelia. In the specific case of lobular breast carcinoma, genetic mutation followed by LOH causes a loss of E-cadherin and the generation of invasive tumor cells that move into the surrounding tissue stroma in a single file pattern that is highly characteristic of mesenchymal invasion (29). In contrast, in ductal breast carcinoma, E-cadherin expression is more often downregulated by transcriptional repressors such as Snail and Slug followed by epigenetic silencing (30). In the mesenchymally transformed MDA-MB-231 breast carcinoma cells used in this study, E-cadherin silencing has been achieved through hypermethylation of the E-cadherin promoter (22). While it is possible to reverse this stable silencing by genetic manipulation (31), it has proven difficult to do so, particularly in a gene-specific manner, by pharmacologic means.

The natural product sarasinoside A1 was first isolated from a Palauan marine sponge more than 20 years ago when it was originally described as being mildly cytotoxic (14, 16, 20). We confirmed this cytotoxicity at high doses. However, at lower, noncytotoxic doses, sarasinoside A1 inhibited MDA-MB-231 tumor cell invasion, caused a morphologic epithelial reversion, and it initiated cell–cell junction formation. Importantly, all of these effects occurred in the absence of any upregulation of E-cadherin.

When epithelially derived tumor cells undergo a mesenchymal transformation, the expression of nonepithelial cadherins often increase as E-cadherin levels decrease. This process has been termed “cadherin switching: (32). Importantly, the upregulated cadherins rarely initiate stable junction formation between the tumor cells, and in some situations, they have been found to contribute to the invasive process by facilitating dynamic, migratory interactions with surrounding stromal and endothelial cells (32). Mesenchymally transformed MDA-MB-231 tumor cells express a number of nonepithelial cadherins (33) which we found to be diffusely localized throughout the cell cytoplasm under control conditions. Sarasinoside A1 treatment caused these cadherins to relocalize to sites of cell–cell interaction which, together with electron microscopic evidence, suggested that the compound causes nonepithelial cadherins to participate in cell–cell adhesion.

Cell–cell adhesion is initiated when the extracellular domains of transmembrane cadherins on the surface of neighboring cells form weak interactions at sites of cell–cell contact. These interactions are strengthened by the recruitment of nectin molecules which interact with ligand-engaged cadherins in the plane of the membrane and facilitate their stable clustering at the nascent cell–cell junction (34). Recruitment of nectin-based adhesion complexes are thought to enhance cadherin-mediated adhesion, in part by, binding F-actin via the afadin/AF-6 scaffolding protein (35) and by activating the Rap GTPases (24). We found that sarasinoside A1 facilitated the recruitment and stabilization of exogenously expressed nectin-2-GFP to these sites in MDA-MB-231 cells, suggesting that the compound acts, at least in part, by enhancing nascent junction stabilization. Sarasinoside A1 also selectively activated the Rap GTPases, which may have occurred because of an increase in nectin-based adhesion or the formation of more stable nectin-based cell–cell junctions, resulting in the activation of signaling pathways that promote Rap activation. Moreover, expressing an activated form of Rap1 was sufficient to stabilize nectin in the membrane of contacting MDA-MB-231 cells, suggesting a positive feedback loop in which nectin-based junctional complexes activate Rap1, which in turn acts to further stabilize these nectin-based adhesion complexes.

As nascent cell–cell junctions begin to mature, cytoplasmic scaffolding proteins help generate a physical linkage between the transmembrane cadherins and F-actin at the junctional site (28). Two such scaffolding proteins, β-catenin and ZO-1, were all prominent at sites of cell–cell contact in sarasinoside A1–treated cells. Sarasinoside A1 treatment also stimulated the accumulation of stable F-actin at sites of cell–cell interaction, which is a biophysical indicator of functional cell–cell junction formation (28). Importantly, this stabilization was selective; sarasinoside A1 did not increase the stability of the actin-binding protein paxillin molecule at cell–ECM junctions (i.e., focal adhesions; ref. 36). Thus, the compound does not appear to act as a general F-actin stabilizer such as, for example, jasplakinolide which also inhibits tumor cell invasion but has other substantial deleterious effects (37, 38). This raises the possibility that sarasinoside A1 may act by selectively regulating the activity of F-actin modulators at cell–cell junctions. Interestingly, a number of anti-invasive compounds that act by selectively altering preexisting cell–ECM or cell–cell junction dynamics have been identified in other screens (9, 11, 39, 40).

The Src kinase inhibitor SKI-606 (bosutinib) inhibits mesenchymal breast tumor cell migration and invasion (41), perhaps by decreasing the phosphorylation of focal adhesion proteins (42, 43). Interestingly, SKI-606 also facilitates cell–cell adhesion in a manner that, like sarasinoside A1, does not require either the presence or upregulation of E-cadherin in MDA-MB-231 cells (43). This is very different than the natural compounds 3,3′-diindolylmethane and isoflavone, which also epithelialize tumor cells but do so by upregulating E-cadherin expression in a manner that depends on modulating microRNA levels (44). Thus, sarasinoside A1 may impinge upon Src signaling to re-epithelialize cells. However, this would likely be an indirect effect because, unlike SKI-606 (41, 43), sarasinoside A1 does not inhibit downstream ERK activation (data not shown).

Preventing tumor cell invasion is an important goal of many target-based antimetastatic drug development efforts. While compounds of potential therapeutic value have been identified using this approach, few have been shown to specifically reverse a stable EMT. However, like the targeted molecule SKI-606, the natural compound sarasinoside A1, appears to initiate such a reversion. Importantly, the fact that SKI-606 and sarasinoside A1 both initiate this reversion in an E-cadherin–independent manner opens the door to the further development of novel antimetastatic therapies that suppress the mesenchymal phenotype, at least in part, by inducing cell–cell adhesion using preexisting adhesion molecules on the surface of metastatic tumor cells.

No potential conflicts of interest were disclosed.

Conception and design: P. Austin, R.J. Andersen, M. Roberge, C.D. Roskelley

Development of methodology: P. Austin, S.A. Freeman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Freeman, C.A. Gray, A.W. Vogl, R.J. Andersen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Austin, S.A. Freeman, C.A. Gray, C.D. Roskelley

Writing, review, and/or revision of the manuscript: P. Austin, R.J. Andersen, M. Roberge, C.D. Roskelley

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.A. Gray

Study supervision: R.J. Andersen, C.D. Roskelley

Collected the electron microscopic images: A.W. Vogl

This research was supported by grant #018344 from the Canadian Cancer Society Research Institute (M. Roberge and C.D. Roskelley) and by funding from the infrastructure program of the Michael Smith Foundation for Health Research to the Solid Tumor Progression Research Unit (C.D. Roskelley, M. Roberge, R.J. Andersen).

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.