RASSF1C, an Isoform of the Tumor Suppressor RASSF1A, Promotes the Accumulation of β-Catenin by Interacting with βTrCP Free

Since the discovery of RASSF1A epigenetic inactivation in lung tumors (1, 2), numerous reports have implicated the inactivation of the RASSF1A promoter by hypermethylation in tumorigenesis across a wide spectrum of human tumors (3, 4). The RASSF1 gene has eight exons and is located within a 120-kb region of chromosome 3p21.3 that frequently undergoes allele loss in human tumors (3, 4). Differential promoter usage and alternative splicing generate seven different transcripts (RASSF1A–RASSF1G). Isoforms A and C are ubiquitously expressed, whereas the expression of the other isoforms is tissue specific (4). A direct correlation between promoter methylation and loss of RASSF1A expression has been shown in many tumor cell lines, including small-cell lung, non–small-cell lung, breast, kidney, nasopharyngeal, prostate, and hepatocellular carcinoma, and retinoblastoma (reviewed in ref. 4). In contrast, the expression of another isoform of RASSF1, RASSF1C, has been shown not to be affected, and no epigenetic inactivation of the RASSF1C promoter has been reported in these tumors (3, 4). RASSF1A null mice that still express RASSF1C had a higher incidence of spontaneous tumorigenesis and a lower survival rate than wild-type mice (5). The mechanism of action of RASSF1A at the molecular level remains poorly understood. RASSF1A and RASSF1C differ at their amino termini and share an identical sequence at their carboxyl termini. This common region includes a Ras association or RalGDS/AF-6 domain (RA domain; ref. 6) that may bind RasGTP similar to that of the Ras effector for apoptosis Nore1 (7–9). RASSF1A and RASSF1C both associate with microtubules, but only RASSF1A stabilizes microtubules (10, 11). RASSF1A has been implicated in the control of cell migration (10) and in the control of cell cycle progression (4, 12, 13). In proapoptotic pathways, RASSF1A is involved in activation of Bax (14) and in a complex with the BH3 protein MAP1, linking death receptor signaling to the apoptotic machinery through Bax conformational change (15). We found that RASSF1A and RASSF1C had opposite effects on the accumulation of the β-catenin oncogene, a key effector of the Wnt signaling pathway (reviewed in ref. 16). Our results suggest that the inhibition of β-catenin accumulation could be one of the mechanisms by which RASSF1A exerts its tumor suppressor function.

Plasmids. Vectors for HA-βTrCP, HA-βTrCPΔF, HA-ATF4, and βTrCP-GFP were described in Lassot et al. (17). The plasmid Top-tk-Luci expressing luciferase was kindly provided by H. Clevers (Department of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, NC). The RSV-Renilla luciferase vector is from Promega (Madison, WI). Myc-RASSF1A, Myc-RASSF1C, and Myc-RA plasmids were kindly provided by M. White (Department of Cell Biology, UT Southwestern Medical Center, Dallas, TX). Vectors for Flag-RASSF1A and Flag-RASSF1C were obtained by transferring RASSF1A and RASSF1C into pSG (Sigma, St. Louis, MO). RASSF1ΔN (deletion of amino acids 1–49) was obtained by PCR from Myc-RASSF1C vector.

Single point mutations in Myc-RASSF1C were obtained by PCR-directed mutagenesis. Plamids for yeast two-hybrid have been described previously by Margottin et al. (18) and Hart et al. (19), except plex-βTrCP Nter 1 to 260, 192 to 569, and 260 to 569, which have been done by PCR using specific primers and plex-βTrCPΔF, which was done by internal deletion of residues 148 to 192.

Yeast two-hybrid screening. Yeast two-hybrid screening was done according to Formstecher et al. (20), using the NH₂-terminal domain (1–291) of βTrCP as bait, against a random primed cDNA library constructed from human placental poly(A)+ mRNA. Two hybrid assays were done as previously described (19).

Lysis, immunoprecipitation, and Western blot. Antibodies used were as follows: mouse monoclonal anti-Flag (Sigma) and rabbit anti-Flag (Sigma), rat anti-hemagglutinin (anti-HA; High Affinity, Roche), mouse monoclonal anti-Myc (Roche), rabbit anti-βTrCP (gift from Klaus Strebel, Laboratory of Molecular Genetics, National Institute of Child Health and Human Development, Bethesda, MD) or goat anti-βTrCP (Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti–α-tubulin (Sigma), mouse anti–β-catenin (Transduction Laboratories, Lexington, KY), goat anti-actin (Santa Cruz Biotechnology), and mouse anti-IκBα (Santa Cruz Biotechnology). A rabbit polyclonal RASSF1C antiserum was produced. The immunogen was RASSF1C recombinant protein expressed as a (His)₆-tagged protein. Antibody specificity was assessed by probing Western blots of 6-His–tagged RASSF1C recombinant protein (data not shown), lysates of Myc-RASSF1C–transfected HeLa cells (Fig. 1C,, lane 4), and lysates of untransfected cells (Fig. 1C,, lane 3). Anti-RASSF1C antibodies recognized both endogenous RASSF1C as well as transfected Myc-RASSF1C (Fig. 1C). Untransfected cells or cells at 24 h after transfection were harvested and lysed in 1% NP40, 150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 7.5), and 1 mmol/L EDTA. For immunoprecipitations, cell lysates were precleared with nonimmune antibodies and protein A or G agarose for 90 min, and supernatants were incubated overnight with the appropriate antibodies and then with protein A or G agarose beads (Sigma) for 1 h. Immunocomplexes were eluted with Laemmli buffer and were separated by SDS-PAGE.

Immunostaining of cells. We used the protocol of Lassot et al. (17). HeLa cells were transfected, fixed, permeabilized, and incubated with mouse anti–β-catenin antibodies; washed in PBS; and incubated with anti-mouse Cy2 antibodies (Jackson Immunoresearch) or directly incubated with anti–Myc-TRITC (Santa Cruz Biotechnology). Confocal microscopy was carried out under fluorescent light.

Pulse-chase experiments. HeLa cells were transiently transfected with 10 μg of empty vector, Myc-RASSF1A, or Myc-RASSF1C expression vectors, and pulse-chase analysis of β-catenin was done as previously described for ATF4 by Lassot et al. (17), but using mouse anti–β-catenin antibodies.

Ubiquitination assay. HeLa cells cotransfected with expression vectors and HA-ubiquitin plasmid were treated 4 h with MG132 (20 μmol/L). Cell pellets were resuspended in 1 volume of denaturing lysis buffer (50 mmol/L Tris, 10% SDS, 5 mmol/L NEM, 1 mmol/L NaOV, 10 mmol/L NaF, and 20 μmol/L MG132 with protease inhibitors), boiled for 10 min, and diluted in 9 volumes of 50 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 5 mmol/L EDTA, 1% NP40, 5 mmol/L NEM, 1 mmol/L NaOV, 10 mmol/L NaF, and 20 μmol/L MG132 with protease inhibitors; then, cells were immunoprecipitated with anti–β-catenin antibody. Ubiquitin species were revealed by Western blot using anti-HA antibodies.

Small interfering RNAs. RNA oligonucleotides were purchased from Eurogentech (Seraing, Belgium). We used the oligonucleotide small interfering RNA A (siRNA A; ref. 12) to silence RASSF1A. The oligonucleotides siRNA A/C (forward, 5′-gcugagauugagcagaagatt; reverse, 5′-ucuucugcucaaucucagctt) were designed to silence both RASSF1A and RASSF1C. We used the siRNA luciferase as described in Bres et al. (21). After annealing, siRNAs were transfected into HeLa cells or A549 cells using Oligofectamine. In Fig. 5A, 24 h later, cells were cotransfected with both siRNA and plasmid vectors using LipofectAMINE (Invitrogen, Carlsbad, CA). Cells were harvested and lysed 24 h after transfection.

βTrCP binds RASSF1C but not RASSF1A. βTrCP is the receptor subunit of the SCF^βTrCP ubiquitin ligase that recruits key signaling proteins, such as the oncogene β-catenin (19, 22), IκBα (23, 24), or ATF4 (17), and cell cycle regulators, such as Emi1 (25, 26) for proteasomal degradation (reviewed in refs. 27–29).

We did two-hybrid screening using a βTrCP fragment from residues 1 to 291 as bait, and we found that βTrCP interacts with RASSF1C (data not shown). This interaction was confirmed in HeLa cells cotransfected with HA-βTrCP and Myc-RASSF1C, and coimmunoprecipitated with anti-Myc antibodies (Fig. 1A,, lane 6, top). HA-βTrCP was not coimmunoprecipitated with Myc-RASSF1A (Fig. 1A,, lane 8). Similar results were found by using Flag-tagged RASSF1A and RASSF1C isoforms (Fig. 1B,, lanes 1 and 2, respectively). Deletion of a 49-amino-acid NH₂-terminal region specific to RASSF1C (and not present in RASSF1A) abolished its interaction with HA-βTrCP (Fig. 1B,, lane 3), and the COOH-terminal RA domain alone was not able to associate with βTrCP (Fig. 1A,, lane 4). This indicated that the NH₂-terminal fragment of 49-amino-acid residues specific to RASSF1C was required for its interaction with βTrCP. The association of RASSF1C with βTrCP was also detected for the endogenous proteins. RASSF1C coimmunoprecipitated with βTrCP when anti-βTrCP antibodies were used for the immunoprecipitation (Fig. 1C,, lane 1), but not when anti-Flag antibodies were used for the immunoprecipitation (negative control Fig. 1C , lane 2).

RASSF1C promotes β-catenin accumulation. We overproduced Myc-RASSF1C in HeLa cells and found that this led to an accumulation of β-catenin, as detected by immunoblot in lysates of transfected cells (Fig. 2A, compare lanes 2–5 with lane 1). Overproduction of Myc-RASSF1A at similar levels had either no such effect (Fig. 2A, compare lanes 7-10 with lane 6) or resulted in less β-catenin expression (Fig. 2A,, right). In Myc-RASSF1C–transfected cells, there was substantial accumulation of β-catenin as assessed by indirect immunofluorescence with anti–β-catenin antibodies (Fig. 2B,, left), whereas in neighboring untransfected cells or in Myc-RASSF1A–transfected cells, β-catenin was barely detectable (Fig. 2B).

We did pulse-chase experiments to investigate the mechanism of β-catenin accumulation. RASSF1C-mediated accumulation of β-catenin was due to stabilization of β-catenin, as detected in Myc-RASSF1C–transfected cells after ³⁵S metabolic labeling (Fig. 2C,, top). After quantification of the labeled bands, we estimated that the stability of β-catenin was enhanced upon RASSF1C overexpression, with a half-life of 30 min in nontransfected cells and a half-life of 90 min in RASSF1C-transfected cells (Fig. 2C,, bottom). Transfection with RASSF1A had no effect on the degradation rate of β-catenin (Fig. 2C). This accumulation of β-catenin correlates with a strong inhibition of β-catenin ubiquitination promoted by overexpression of RASSF1C in cells cotransfected with HA-ubiquitin (Fig. 2D, compare lane 2 with lane 1).

The RASSF1C-mediated accumulation of β-catenin led to an ∼10-fold increase in β-catenin transcriptional activity based on experiments using a luciferase reporter gene under the control of the β-catenin–dependent TOP-flash synthetic promoter (ref. 19; Supplementary Fig. S1). This increase in β-catenin–dependent transcriptional activity was comparable with that obtained with βTrCPΔF, the βTrCP-negative transdominant mutant previously reported to inhibit the degradation rate of β-catenin (19).

We studied the effect of RASSF1C overexpression on two other known substrates of βTrCP, IκBα and ATF4, and we found that the degradation rate of these βTrCP substrates was also inhibited by overexpression of RASSF1C, whereas RASSF1A had no effect (Supplementary Fig. S2 for IκBα; data not shown for ATF4). Thus, the effect of RASSF1C is not specific for β-catenin but is a more general effect on the inhibition of βTrCP-mediated degradation of several protein substrates of the SCF^βTrCP E3 ubiquitin ligase.

We further considered the question of whether RASSF1A could play a role in RASSF1C-mediated β-catenin accumulation by first determining whether RASSF1A could interfere with the interaction between RASSF1C and βTrCP. We studied the interaction between transfected HA-βTrCP and Myc-RASSF1C in the absence or in the presence of increasing concentrations of RASSF1A-Flag, and we found that the level of expression of RASSF1A did not significantly affect the βTrCP/RASSF1C interaction analyzed by coimmunoprecipitation (Supplementary Fig. S3, compare lanes 3 and 4 with lane 2).

RASSF1C-mediated β-catenin accumulation requires binding to RASSF1C via a SSGYXS motif. We isolated mutants of RASSF1C deficient for interaction with βTrCP and examined their ability to promote β-catenin accumulation. We mutated two serine residues, S18 and S19, present in a SS₁₈GYCS₁₉ motif in the NH₂-terminal region specific to RASSF1C. This SSGYCS motif is reminiscent of the canonical phosphorylation motif DSGXXS, previously characterized as the binding site of βTrCP (18, 27, 28). Constructs expressing wild-type Myc-RASSF1C or its mutants were cotransfected with HA-βTrCP to study the association between these proteins. The S18N, S19N, and S18-19N mutants did not immunoprecipitate βTrCP (Fig. 3A,, lanes 3-5), in contrast to wild-type RASSF1C (Fig. 3A,, lane 2), and the mutants did not promote β-catenin accumulation, as shown by immunofluorescence using anti–β-catenin antibodies (Fig. 3B). Interaction with βTrCP is therefore required for RASSF1C to mediate β-catenin accumulation.

A hypothesis that could account for this ability of RASSF1C to mediate β-catenin accumulation is that, by binding to βTrCP, RASSF1C inhibits the βTrCP-β-catenin interaction, thus impairing the proteasomal degradation of β-catenin. Consistent with this hypothesis, we found that increasing the amount of transfected RASSF1C inhibited the amount of endogenous β-catenin that could be coimmunoprecipitated with transfected HA-βTrCP (Fig. 3C,, top, compare lanes 3-5 with lane 2), although the total expression level of β-catenin had increased as expected from RASSF1C transfection (Fig. 3C,, top). As a control, overexpression of RASSF1A had no effect on coimmunoprecipitation of β-catenin with transfected HA-βTrCP (Fig. 3C,, bottom). From these data, we hypothesized that interaction with RASSF1C would require the substrate binding domain of βTrCP, comprising the WD40 repeats in the COOH-terminal moiety of the protein, and not the NH₂-terminal domain comprising the F-box, which connects to the proteasome through Skp1 binding (30, 31). Surprisingly, RASSF1C still interacted with the NH₂-terminal moiety of βTrCP after removal of the all the WD40 repeats (βTrCP NH₂-terminal 1–260; Fig. 3D,, lanes 1 and 2). No interaction was found between RASSF1C and the COOH-terminal substrate binding region, from residues 260 to 569 of βTrCP, which, by contrast, interacted with the HIV-1 Vpu protein used as a control (Fig. 3D,, lanes 5 and 11). βTrCP deleted from the F-box motif and the NH₂-terminal region upstream of the F-box (βTrCP 192-569) or βTrCP deleted only from the F-box (βTrCPΔF) still interacted with RASSF1C (Fig. 3D,, lanes 4 and 3), whereas, as a control, βTrCPΔF does not interact with Skp1 (Fig. 3D , lane 8). Altogether, the binding data indicate that neither the WD40 repeats nor the F-box is required for RASSF1C binding. Further mapping showed that a fragment encompassing residues 192 to 260 present between the F-box and the WD40 repeats conserved its ability to interact with RASSF1C (data not shown).

RASSF1C delocalizes βTrCP from nucleus to cyctoplasm. βTrCP is mostly localized to the nucleus in nonmitotic cells (17, 32), whereas RASSF1A and RASSF1C have been described as cytoplasmic, microtubule-associated proteins (10, 12). Cells transfected with βTrCP-GFP alone showed fluorescence in the nucleus (Fig. 4, top panels), in agreement with previous studies. In cells cotransfected with βTrCP-GFP and RASSF1C, there was substantial redistribution of βTrCP-GFP toward a mostly cytoplasmic staining (bottom panels), and this staining colocalized with that of RASSF1C on the cytoskeleton. As expected, Myc-RASSF1A, like Myc-RASSF1C, was also associated with the cytoskeleton in the cytoplasm (Fig. 4, middle panels). However, when βTrCP-GFP was cotransfected with Myc-RASSF1A, the staining of βTrCP remained mostly nuclear, similar to that of βTrCP-GFP transfected alone (middle panels), indicating that RASSF1A, unlike RASSF1C, did not change the subcellular localization of βTrCP from the nucleus to the cytoplasm.

Silencing of RASSF1A together with RASSF1C impairs β-catenin accumulation resulting from RASSF1A silencing. HeLa cells express both RASSF1A and RASSF1C, as previously reported (refs. 1, 33; data not shown). We tested appropriate siRNAs to silence specifically either RASSF1A or RASSF1C, or both, to mimic the situation found in tumor cell lines in which RASSF1A expression has been selectively lost by epigenetic methylation of the RASSF1A promoter. The efficiency of siRNA inhibition of the expression of each isoform in HeLa cells transfected with Myc-RASSF1A or Myc-RASSF1C was assessed by quantitative reverse transcription PCR and by Western blot using anti-Myc antibodies. Using these assays, we could not find any siRNAs that specifically silenced only isoform 1C and not isoform 1A. This is probably due to limited sequence differences between the mRNAs coding for the RASSF1 isoforms. As previously reported (12, 33), siRNA A was effective and could silence the expression of transfected Myc-RASSF1A by over 75%. However, the expression of transfected Myc-RASSF1C was also slightly inhibited (Fig. 5A, top, compare lane 3 with lane 1, and lane 4 with lane 2). We found that in siRNA A–treated cells, the level of β-catenin expression was several times higher than in control cells treated with siRNA luciferase (Luc; Fig. 5B, compare lane 3 with lane 2). Another siRNA, siRNA A/C, which targets a sequence common to RASSF1A and RASSF1C mRNAs, almost completely silenced both isoforms (Fig. 5A, compare lane 5 with lane 1, and lane 6 with lane 2, top), without any detectable effect on the expression of an unrelated protein, Myc (Fig. 5A,, bottom). HeLa cells treated with siRNA A/C behaved similarly to those treated with siRNA Luc, with no detectable accumulation of β-catenin (Fig. 5B, compare lanes 2 and 4). Therefore, we conclude that the silencing of RASSF1A was sufficient, in the absence of overexpression of exogenous RASSF1C, to promote the accumulation of β-catenin. This effect was observed despite the silencing of RASSF1A being incomplete, and despite partial silencing of RASSF1C expression (Fig. 5A , lane 4). This effect of RASSF1A silencing is comparable with that observed upon siRNA silencing of βTrCP1 and/or βTrCP2 (25).

Silencing of RASSF1C in a cell line that does not express RASSF1A strongly decreases the expression of β-catenin. We further considered the question of whether RASSF1C could play a role in controlling the expression level of β-catenin in a tumor cell line defective for RASSF1A expression, such as the A549 cell line. To address this question, we compared the effects of the siRNA A/C, which very efficiently suppresses the expression of RASSF1C (see Fig. 5A), with that of siRNA A, which should have no effect because RASSF1A is not expressed in these cells. As shown in Fig. 5C, treatment of A549 cells with siRNA A/C greatly reduced the expression level of β-catenin, whereas siRNA A had no effect, compared with siRNA Luc used as a control. Therefore, we conclude that the silencing of RASSF1C, in the absence of RASSF1A expression, was sufficient to strongly reduce β-catenin expression.

We found that RASSF1C interacts with βTrCP and as a result promotes the accumulation of β-catenin by inhibiting its βTrCP-mediated proteasomal degradation. This accumulation of β-catenin correlates with an increase of the half-life of β-catenin and a strong decrease of its ubiquitination. The stabilizing effect of RASSF1C on β-catenin does not seem to be restricted to this substrate of βTrCP, but seems also more generally to affect the degradation of other βTrCP substrates such as IκBα and ATF4. Mutants of RASSF1C that did not associate with βTrCP had no effect on β-catenin. RASSF1C could inhibit the interaction of β-catenin with βTrCP and could change the subcellular localization of βTrCP from the nucleus to the cytoplasm, where both proteins colocalized. The isoform RASSF1A did not have these effects. The silencing of RASSF1A in HeLa cells, which express both RASSF1A and RASSF1C isoforms, was sufficient, without RASSF1C overexpression, to promote the accumulation of β-catenin. In contrast, the silencing of both RASSF1A and RASSF1C in these cells had no effect on β-catenin. In A549 lung cancer cells, which do not express RASSF1A, silencing of RASSF1A by siRNA A as expected had no effect, whereas silencing of RASSF1C by siRNA A/C decreased β-catenin expression. Altogether, these observations suggest that RASSF1A and RASSF1C isoforms have opposite effects in controlling the degradation of β-catenin mediated by the SCF^βTrCP ubiquitin ligase.

Whether RASSF1C is a βTrCP substrate, and as such could compete for the binding of other substrates such as β-catenin or IκBα, remains an open question. However, several lines of evidence suggest that it is not the case: First, the association of RASSF1C with βTrCP seems to be independent of the substrate binding domain of βTrCP; second, the association is very stable and does not require the addition of proteasome inhibitors to be unraveled as for most βTrCP substrates; last, knockdown of βTrCP by siRNA hardly affects RASSF1C levels whereas it dramatically increases β-catenin levels (data not shown). Altogether, these observations suggest that RASSF1C acts more likely as a βTrCP pseudosubstrate like hnRNP-U (32) and/or as a negative modulator of the SCF^βTrCP complex capable to inhibit the βTrCP-mediated degradation of several substrates, although the association of RASSF1C to βTrCP depends on a SSGYCS motif.

β-Catenin accumulation was promoted either by the overexpression of RASSF1C or by the silencing of RASSF1A, and this implies that the balance between the two isoforms is crucial for the βTrCP-mediated degradation of β-catenin. When the equilibrium between the RASSF1A and RASSF1C isoforms is disrupted, either by lack of RASSF1A expression or by RASSF1C overexpression, the function of SCF^βTrCP is impaired and β-catenin accumulates. Thus, the ratio between RASSF1A and RASSF1C expression levels may be crucial to tight regulation of β-catenin expression levels via constitutive degradation of β-catenin mediated by the SCF^βTrCP ubiquitin ligase. In the presence of normal levels of RASSF1A expression, the inhibitory activity of RASSF1C on the degradation of β-catenin mediated by the SCF^βTrCP seems to be prevented. In the absence of RASSF1A expression, RASSF1C could inhibit β-catenin degradation, thus promoting β-catenin accumulation. The decrease in β-catenin expression we observed upon silencing of RASSF1C in RASSF1A-deficient cell line A549 strongly supports this interpretation. Furthermore, our observations agree with previous reports which showed that the expression of cyclin D1, a target of β-catenin, was high in cells depleted of RASSF1A expression, and that cyclin D1 was then down-regulated upon reexpression of RASSF1A (4, 12). Both βTrCP and RASSF1A act at the same phase of the cell cycle: βTrCP controls the degradation of the APC^Cdc20 inhibitor Emi1 (26); and RASSF1A has been implicated in prometaphase arrest by also inhibiting the APC^Cdc20 complex and by subsequently stabilizing cyclin A (34). As discussed by Jackson (35), it is unclear why both RASSF1A and Emi1 would regulate the levels of cyclin A. One possibility is that RASSF1A may control the activity or destruction of Emi1, perhaps involving an interaction between RASSF1C and βTrCP.

βTrCP controls the stability of key signaling proteins and cell cycle regulators that are believed to be involved in carcinogenesis. Expression defects or mutations in the βTrCP gene have been found in some human tumors (18, 27–29, 36, 37). Inactivation of βTrCP in transgenic mice, by overexpression of the negative transdominant βTrCPΔF, also resulted also in more frequent tumors (38).

Additional investigations will be needed to better understand the role of RASSF1C in tumorigenesis mediated by the absence of expression of RASSF1A in numerous tumors. Two previous reports indicated that RASSF1C could have, together with RASSF1A, tumor suppressor activity (39, 40). This hypothesis cannot be ruled out today and RASSF1C could have multiple functions depending on the cell type. However, in many tumor cell lines, loss of RASSF1A expression has been observed (see ref. 4 for review), whereas changes in RASSF1C expression have not been reported in these tumors. Our present observations led us to postulate that RASSF1C, through inhibition of β-catenin degradation, may be involved in the tumorigenesis events linked to the inactivation of RASSF1A, illustrating the opposite effects that two isoforms of the RASSF1 gene may have in carcinogenesis. Consistent with this view, it was recently reported that RASSF1C could activate osteoblast cell proliferation through interaction with IGFBP-5, confirming that RASSF1C could play a role in carcinogenesis (41).

Grant support: Association pour la recherche sur le cancer, La Ligue Nationale Contre le Cancer (I. Lassot), Association nationale de recherche contre le sida, Sidaction (G. Blot), La Mairie de Paris, French Ministry of Education (I. Lassot, E. Estrabaud, and G. Blot), and Institut National de la Sante et de la Recherche Medicale (E.L. Le Rouzic).

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.

We thank Michael White for the kind gift of plasmids and for fruitful discussions, and Klaus Strebel and Geoff Clark (Department of Cell and Cancer Biology, National Cancer Institute, NIH, Rockville, MD) for the kind gift of antibodies.