|Home | About | Journals | Submit | Contact Us | Français|
The GTPase RhoA participates in a number of cellular processes, including cytoskeletal organization, mitogenesis and tumorigenesis. We have previously shown that the transforming activity of an oncogenic version of RhoA (Q63L mutant) was highly dependent on the transcriptional factor c–Myc. In contrast to these positive effects in the RhoA route, we show here that c–Myc affects negatively the F–actin cytoskeleton induced by RhoAQ63L and its downstream effector, the serine/threonine kinase Rock. This effect entails the activation of a transcriptional program that requires synergistic interactions with RhoA–derived signals and that includes the upregulation of the GTPase Cdc42 and its downstream element Pak1 as well as the repression of specific integrin subunits. The negative effects of c–Myc in the F–actin cytoskeleton are eliminated by the establishment of cell–to–cell contacts, an effect associated with the rescue of Pak1 and integrin levels at the post–transcriptional and transcriptional levels, respectively. These results reveal the presence of a hitherto unknown signaling feed–back loop between RhoA and c–Myc oncogenes that can contribute to maintain fluid cytoskeletal dynamics in cancer cells.
RhoA is a GTP–binding protein that belongs to the Rho/Rac GTPase subfamily (Bustelo et al., 2007; Etienne–Manneville & Hall, 2002; Jaffe & Hall, 2005). This protein plays critical roles in general cell functions such as F–actin cytoskeletal dynamics, mitogenesis and cell survival as well as in cell type–specific biological processes that include, among others, arterial contractility, axogenesis and phagocytosis (Bustelo et al., 2007; Etienne–Manneville & Hall, 2002; Jaffe & Hall, 2005). To carry out those functions, RhoA engages a wide collection of downstream effectors that work as protein kinases, lipid–related enzymes or scaffolding proteins (Bustelo et al., 2007; Jaffe & Hall, 2005). Among those effectors, RockI and RockII are perhaps the best characterized at the structural, regulatory, and biological levels (Mueller et al., 2005; Riento & Ridley, 2003). These serine/threonine kinases mediate the formation of stress fibers and focal adhesions, thereby participating in cell–to–cell and cell–to–substratum adhesion, cell migration and invasiveness, neurite retraction and phagocytosis (Mueller et al., 2005; Riento & Ridley, 2003).
In addition to the canonical regulation of RhoA by guanosine nucleotide exchange factors (GEFs), GTPase activating proteins (GAPs), and Rho GDP dissociation inhibitors (RhoGDIs) which is common to most Rho/Rac proteins (Bos et al., 2007; Dransart et al., 2005; Olofsson, 1999), the signaling output of RhoA can be controlled by unique mechanisms. Those include its proteosomal degradation via the ubiquitin ligase Smurf1 (Sahai et al., 2007; Wang et al., 2006; Wang et al., 2003), the inactivation of Rock proteins by the binding to either the cell cycle inhibitor p21WAF (Lee & Helfman, 2004) or RhoE, Gem and Rad GTPases (Hatzoglou et al., 2007; Riento et al., 2003; Ward et al., 2002), and the stimulation of Rock activity by dephosphorylation by the Shp2 phosphatase (Lee & Chang, 2008). The signal output from the RhoA route can be also influenced by more distal, transcriptional–based mechanisms. Hence, it has been shown that the tumor suppressor p53 activates RhoE gene transcription, leading to the inhibition of Rock–dependent effects (Ongusaha et al., 2006). Proteomic experiments have revealed that c–Myc can regulate the activity of RhoA–dependent routes by lowering the levels of RhoA, Cdc42, Rock and a subset of cytoskeletal–related proteins (Shiio et al., 2002). It is important to note that these regulatory influences are usually bidirectional, a property that facilitates the establishment of feed–back loops that can provide further plasticity to GTPase–regulated processes. Consistent with this view, it has been shown that RhoA and Cdc42 can stimulate and repress c–Myc (Berenjeno et al., 2007; Watnick et al., 2003) and p53 (Park et al., 2009), respectively. In this work, we present evidence indicating that there exists another cross–talk between c–Myc and RhoA that contributes to the dowmodulation of the RhoA/Rock cytoskeleton in mouse fibroblasts. Interestingly, this transcriptional program is inhibited by the establishment of cell–to–cell contacts both at the transcriptional and posttranslational level, a property that gives further flexibility to the modulation of F–actin cytoskeletal dynamics in vivo.
During a previous study (Berenjeno et al., 2007), we generated a number of NIH3T3 cell derivatives expressing RhoAQ63L, c–Myc, RhoAQ63L plus c–Myc, and RhoAQ63L plus either a c–Myc dominant negative mutant (MadMyc) or a c–Myc–specific short hairpin RNA (shRNA). Using those cell lines, we demonstrated that c–Myc was important for both the loss of cell contact inhibition and cell transformation induced by the RhoAQ63L oncogene in fibroblasts (Berenjeno et al., 2007). To further characterize the effect of the c–Myc network in the transformation mediated by this GTPase, we decided to check the status of the F–actin cytoskeleton in these cells using microscopy techniques. As previously described (Berenjeno et al., 2007), we observed that RhoAQ63L–transformed cells contained robust stress fibers when compared with the parental cell line (Fig. 1A). However, this cytoskeletal phenotype was lost in cell lines co–expressing RhoAQ63L and c–Myc (Fig. 1A and data not shown) but not in those co–expressing RhoAQ63L with either MadMyc (Fig. 1A) or a c–Myc specific shRNA (Fig. 1A). The comparison of parental and c–Myc expressing NIH3T3 cells indicated that the overexpression of c–Myc alone also induced the disruption of stress fibers (Fig. 1A). This effect, however, was less conspicuous than that found in the case of RhoAQ63L–transformed cells because of the lower levels of stress fibers present in the parental NIH3T3 cells (Fig. 1A). In agreement with the confocal microscopy data, we found using flow cytometry experiments that cell lines co–expressing RhoAQ63L and c–Myc had lower F–actin levels than those expressing exclusively RhoAQ63L (Fig. 1B,C). These analyses also indicated that the co–expression of MadMyc further elevated the total levels of F–actin induced by RhoAQ63L (Fig. 1B,C), suggesting that the increased levels of endogenous c–Myc protein induced by RhoAQ63L also contribute to tuning down the RhoAQ63L–dependent F–actin cytoskeleton in fibroblasts. Western blot experiments indicated that negative effect of c–Myc in the F–actin cytoskeleton was not due to alterations in the total amount of actin present in fibroblasts (Fig. 1D).
To extend these observations to other cellular structures, we investigated the status of focal adhesions and microtubules in those cells lines. The overexpression of c–Myc eliminated the numerous focal adhesions present in RhoAQ63L–transformed NIH3T3 cells (Fig. 2A,B). The negative effect of c–Myc overexpression in the RhoAQ63L–dependent F–actin cytoskeleton was not constitutive, because we observed a restoration of both stress fibers and focal adhesions when cell lines co–expressing RhoAQ63L and c–Myc established extensive cell–to–cell contacts (Fig. 2A,B). The effect of c–Myc in the F–actin cytoskeleton was specific, since it did not have any effect on the microtubule network when overexpressed in either parental or RhoAQ63L–expressing NIH3T3 cells (Supplementary Fig. 1 available online). We have also previously demonstrated that cell lines overexpressing RhoAQ63L and c–Myc are more transforming than those expressing the GTPase alone (Berenjeno et al., 2007). Consistent with the F–actin–specific inhibitory effect of c–Myc overexpression, we observed that the overexpression of c–Myc did not antagonize other RhoAQ63L–mediated responses such as the loss of cell polarity in NIH3T3 cells (Supplementary Fig. 2 available online). These results indicate that the antagonistic function of c–Myc seems to be limited to RhoA–dependent cytoskeletal pathways and cannot be generalized to other unrelated pathways that contribute to cell transformation or cell polarity loss.
Since the formation of stress fibers and focal adhesions by RhoA is regulated by Rock (Amano et al., 1997; Riento & Ridley, 2003), we next investigated the status of this pathway in the presence or absence of overexpressed c–Myc. To this end, we evaluated the phosphorylation levels of two well–known Rock downstream elements, the myosin light chain (MLC) and the myosin binding subunit of MLC phosphatase (MYPT1) (Fig. 3A) (Amano et al., 1996; Kimura et al., 1996; Riento & Ridley, 2003). In addition, we evaluated MLC, MYPT1, RockI and RockII protein levels in the indicated cells (Fig. 3A). Whereas we did not observe any significant change in the total levels of those four proteins among RhoAQ63L– and RhoAQ63L/c–Myc expressing cells (Fig. 3B), we found that the phosphorylation levels of both MLC and MYPT1 were severely decreased in cells co–expressing RhoAQ63L and c–Myc (Fig. 3B,C), indicating that c–Myc overexpression leads the inhibition of the Rock pathway in RhoAQ63L–transformed cells. Additional experiments indicated that the inhibition of Rock signaling was not due to indirect effects in the membrane localization of RhoAQ63L (data not shown) or to increases in the cytosolic distribution of p21WAF1 (Fig. 3D), a cell cycle inhibitor that can bind to Rock and inhibit its catalytic activity (Fig. 3A) (Lee & Helfman, 2004). We could not see either any elevation in the mRNA levels of RhoE (data not shown, but see below, Fig. 5), a GTPase that inactivates Rock (Riento et al., 2003). These results indicate that c–Myc affects negatively the activity of the Rock/MLC pathway in RhoAQ63L–transformed cells.
Given the above results, we next investigated whether an increase in the levels of either RhoAQ63L or RockII signals could restore the F–actin cytoskeleton in c–Myc– and RhoAQ63L/c–Myc–expressing NIH3T3 cells. To this end, we transiently expressed in cells either an enhanced green fluorescent protein (EGFP) fused to RhoAQ63L or a chimeric protein composed of EGFP, the RockII kinase domain, and the estrogen receptor hormone binding domain (EGFP–RockIIKD–ER protein). The latter protein only becomes activated upon treatment of cells with hydroxytamoxifen (4–OHT) (Croft et al., 2004). The transient expression of EGFP–RhoAQ63L restored stress fibers in both c–Myc– and RhoAQ63L/c–Myc–expressing cells (Fig. 4A). Instead, its expression in RhoAQ63L–transformed cells did not enhance the numerous stress fibers already present in those cells (data not shown). The EGFP–RockIIKD–ER protein induced the formation of stress fibers in both parental and RhoAQ63L–expressing NIH3T3 cells when derepressed by 4–OHT binding (Fig. 4B). The stress fibers induced by the EGFP–RockIIKD–ER chimera showed an aster–like distribution and not the typical parallel network of F–actin fibers, an effect probably derived from the lack of activation of other RhoA downstream elements that contribute to proper stress fiber orientation (i.e., mDia) (Nakano et al., 1999). Similar, radially oriented stress fibers were also induced by the 4–OHT–stimulated EGFP–RockIIKD–ER in RhoAQ63L/c–Myc–expressing cells (Fig. 4B). However, in this case the transfected cells spread less efficiently than parental and RhoAQ63L–expressing NIH3T3 cells (Fig. 4B). The number of stress fibers per cell induced by the inducible RockII kinase domain was also reduced when compared to those found in the latter cell lines (Fig. 4B). The cytoskeletal effects of the EGFP–RockIIKD–ER chimera were due to its kinase activity, as demonstrated by the lack of stress fiber induction when a catalytically inactive version of this protein was used (Fig. 4B). Taken together, these results suggest that the effect of c–Myc on the F–actin cytoskeleton is probably mediated by two independent routes, one affecting the total signaling output of the RhoA/Rock signaling route (which can be recovered by the overexpression of RhoA or RockII) and another one probably linked to cytoskeletal components related to spreading and focal adhesions that cannot be fully restored by enhanced RhoA/Rock signals.
We hypothesized that the negative effects of c–Myc in the F–actin cytoskeleton of RhoAQ63L–transformed fibroblast had to be transcriptome–based. To identify gene targets that could be involved in this response, we carried out Affymetrix microarray analysis using total RNAs obtained from subconfluent cultures of RhoAQ63L–, RhoAQ63L/c–Myc–, and of RhoAQ63L/MadMyc–expressing cells (Berenjeno et al., 2007). Using the selection criteria indicated in the Supplemental Materials and Methods Section, we found a total of 535 genes (239 upregulated, 296 downregulated) whose protein products could be potentially involved in the negative modulation of the RhoA–dependent F–actin cytoskeleton by c–Myc (Fig. 5A and Supplementary Table I available online). Of those 535 genes, 152 genes had opposed expressing profiles between RhoAQ63L/c–Myc– and RhoAQ63L/MadMyc–expressing cells (Fig. 5A and Supplementary Table I). Functional annotation of the proteins encoded by those genes indicated that they were involved in a large variety of cellular functions (Supplementary Table II available online). Most of those proteins did not have any obvious correlation with cytoskeletal regulatory events or cytoskeletal structures. However, we found a small subset of c–Myc targeted loci encoding proteins that could be assigned to three functional categories related to the regulation of cytoskeletal dynamics: (i) Direct upstream regulators of the RhoA pathway. (ii) GTPases that antagonize RhoA signaling routes. (iii) Structural elements of the cell cytoskeleton (Fig. 5B and Supplementary Text available online). Some of those proteins were dismissed since they were unlikely to block signals derived from the constitutively active version of RhoA expressed in the NIH3T3 cell lines used in this study (see Supplementary Text). As a consequence, we decided to focus our attention in the deregulated genes that encoded important cytoskeletal regulators such as Cdc42, Pak1, and integrin subunits (Itgβ1l, Itgβ5 and Itgα5) (Fig. 5B and Supplementary Text). To corroborate the microarray data on those genes, we performed quantitative RT–PCR experiments. As negative controls, we used oligonucleotide pairs to amplify cDNAs encoding integrins (β1, β4) not identified in our arrays. In order to maximize the information gathered from these experiments, we decided to carry out RT–PCR reactions using total RNAs from subconfluent cultures of c–Myc–, RhoAQ63L–, RhoAQ63L/c–Myc– and RhoAQ63L/MadMyc– expressing cells. Using this approach, we wished to verify whether the expression of those genes was dependent exclusively on the expression of c–Myc or, alternatively, on synergistic interactions between the transcriptional programs of c–Myc and RhoAQ63L. These experiments indicated that the Cdc42 (Fig. 6A) and Pak1 (Fig. 6B) mRNAs were indeed upregulated in RhoAQ63L/c–Myc–expressing cells when compared to the rest of cell lines used in the study. The upregulation of Pak1 protein in RhoAQ63L/c–Myc–expressing NIH3T3 cells was also demonstrated by immunoblot experiments (Supplementary Figs. 3A and 4A,B). Consistent with the microarray data, we also observed reduced levels of expression for the Itbl1 (Fig. 6D) and Itgb5 (Fig. 6E) mRNAs but not for the Itgb4 transcript (Fig. 6F) in RhoAQ63L/c–Myc–expressing NIH3T3 cells. A significant reduction, not picked up in the microarray experiments, was also observed for the Itgb1 transcript in that cell line (Fig. 6C). This repression was also observed at the protein level in RhoAQ63L/c–Myc–expressing cells (Supplementary Fig. 3B). Most of the transcript tested were found differentially expressed at statistical significant levels only in RhoAQ63L/c–Myc–expressing cells, indicating that they are regulated by synergistic interactions between RhoAQ63L–dependent signals and c–Myc. Interestingly, the Itgb5 mRNA was the only one showing a specific variation in RhoAQ63L/MadMyc–expressing cells, suggesting that it may be one of the key elements promoting the increased numbers in stress fibers observed in this cell line (see above, Fig. 1A,B).
To get an overall view of the synergistic cross–talk among RhoAQ63L and c–Myc at the transcriptional level, we next compared the genes identified as “RhoAQ63L/c–Myc specific” in these experiments with the “RhoAQ63L specific” transcriptome previously described by us in NIH3T3 cells (Berenjeno et al., 2007). Consistent with expression patterns resulting from synergistic interactions between RhoAQ63L signals and overexpressed c–Myc protein, we found that the majority of the deregulated genes present in RhoAQ63L/c–Myc–expressing cells were not present in the transcriptome of RhoAQ63L–expressing fibroblasts (see further details in Supplementary Text). Since our previous results indicated that some of the Rock targets were downmodulated in RhoAQ63L/c–Myc expressing cells (see above, Fig. 3B,C), we also compared the present microarray data with the previously described Rock/Y27632–dependent transcriptome of RhoAQ63L–transformed cells (Berenjeno et al., 2007). We considered that, if c–Myc abrogated all Rock signaling, there had to be a significant overlap between the Rock/Y27632– and RhoAQ63L/c–Myc–dependent transcriptomes. We found that c–Myc overexpression only targeted 4.1% of the 97 Rock/Y27632–dependent genes previously identified as Rock/Y27632–dependent in RhoAQ63L–transformed cells, a result that further suggest that this transcriptional factor does not abolish all Rock–dependent functions in fibroblasts. This is consistent with previous data indicating that c–Myc overexpression, unlike the Y27632 treatments (Berenjeno et al., 2007; Sahai et al., 1999), does not affect RhoAQ63L transforming activity in NIH3T3 cells (Berenjeno et al., 2007). Taken together, these results indicate that the overexpression of c–Myc leads to synergistic interactions with RhoA–dependent signals that result in new patterns of gene expression. Moreover, they suggest that c–Myc overexpression targets the cytoskeletal signaling branch of Rock but not other Rock–related functions.
To further correlate the aforementioned transcriptomal changes with the conditions in which the cell cytoskeleton is inhibited in RhoAQ63L/c–Myc expressing cells, we evaluated the expression of Cdc42, Pak1, Itgb1, Itbl1 and Itgb5 genes in dense cell cultures. Quantitative RT–PCR experiments revealed that the expression of Cdc42 and Pak1 genes was independent on the density of RhoAQ63L/c–Myc–expressing cells in culture (Fig. 6, panels G and H, respectively). Despite this, we observed that the protein levels of Pak1 were severely diminished in dense cell cultures (Supplementary Fig. 4), suggesting that this kinase may be controlled by a post–transcriptional mechanism upon the establishment of cell–cell contacts. The repression of Itgb1, Itgbl1 and Itgb5 genes observed in RhoAQ63L/c–Myc–expressing cells was totally eliminated when these cells were cultured at increased densities (Fig. 6, panels I, J and K, respectively). A strong repression of Itgβ1 protein levels in RhoAQ63L/c–Myc expressing cells, which was rescued in high cell density conditions, was also demonstrated using Western blot analysis (Supplementary Fig. 3B). These experiments indicate that a subset of the RhoAQ63L/c–Myc–dependent transcriptome and proteome can be antagonized by the establishment of cell–to–cell contacts.
We hypothesized that if some of the above genes were involved in the disassembly of stress fibers in RhoAQ63L/c–Myc–expressing cells, the manipulation of the levels of activity and/or expression of their protein products should induce the disruption of the stress fiber network in RhoAQ63L–expressing cells in the absence of c–Myc overexpression or, alternatively, rescue the stress fibers defects in RhoAQ63L/c–Myc–expressing cells. To investigate this possibility, we first analyzed the influence of the Cdc42/Pak1 axis in the stability of stress fibers in RhoAQ63L–expressing NIH3T3 cells. To this end, we infected those cells with retroviruses encoding bicistronically the EGFP and different versions of Cdc42, Rac1, and Pak1 proteins. Those included a fast–cycling mutant (F28L) of Cdc42 that shows constitutive activity in vivo (Lin et al., 1997), a fast–cycling mutant (F28L) of Rac1 (a GTPase that also binds and activates Pak1) (Guo & Zheng, 2004), a Rac1F28L protein containing a point mutation in its switch region (F37A) that eliminates the binding and activation of Pak1 (Joneson et al., 1996), a Rac1F28L protein with a point mutation in the switch region (Y40C) that retains the ability of stimulating Pak1 (Joneson et al., 1996), and the wild type version of Pak1. The expression of Cdc42F28L (Fig. 7A, Supplementary Fig. 5A), Rac1F28L+Y40C (Fig. 7A, Supplementary Fig. 5A) and wild type Pak1 (Fig. 7A,B, Supplementary Fig. 5A,B) led to the disassembly of the F–actin stress fibers in RhoAQ63L–expressing cells. Instead, and consistent with the implication of Pak1 in this process, the expression of the Rac1F28L+F37A mutant had no effects on the cytoskeleton of RhoAQ63L–transformed cells (Fig. 7A, Supplementary Fig. 5A).
To further verify the involvement of Pak1 in the c–Myc–dependent inhibition of the cytoskeleton, we infected RhoAQ63L–, RhoAQ63L/c–Myc– and RhoAQ63L/MadMyc–expressing cells with retrovirus encoding bicistronically the EGFP and a dominant negative Pak1 protein (K298R mutant). The overexpression of that mutant promoted the re–establishment of stress fibers in cells co–expressing RhoAQ63L and c–Myc, indicating that the endogenous Pak1 protein does contribute to the downmodulation of their stress fibers (Fig. 7C, Supplementary Fig. 5C). In agreement with the lack of upregulation of Pak1 in RhoQ63L– and RhoAQ63L/MadMyc–expressing cells, Pak1K298R did not induce any significant effect in the stress fibers of those two cell lines (Fig. 7C, Supplementary Fig. 5C). We observed that the inhibitory effect of Pak1 on the stress fiber network was lost when RhoAQ63L/Pak1–expressing cells established large numbers of cell–to–cell contacts (Fig. 7D, Supplementary Fig. 5D). These experiments indicate that Pak1 overexpression mimics the effects of c–Myc overexpression in the F–actin cytoskeleton of RhoAQ63L–transformed cells.
We next used siRNAs for Itgb1 and Itgb5 transcripts to investigate the effects of integrin downmodulation in the stability of stress fibers of RhoAQ63L–expressing NIH3T3 cells. The knockdown of any of these two integrins induced a ≈50% reduction in the total number and thickness of stress fibers (Fig. 7E, Supplementary Fig. 5E). As control, we observed that a siRNA containing a scrambled nucleotide sequence did not affect the stress fibers present in RhoAQ63L–expressing fibroblasts (Fig. 7E). The knockdown of Itgb1 and Itgb5 mRNAs induced loss of adherence and the detachment of cells from the coverslip (data not shown). The knock–down levels of Itgb1 and Itgb5 transcripts were verified by RT–PCR experiments (Fig. 7E). Taken together, these results demonstrate that Pak1 and integrin subunits β1 and β5 are integral components of the transcriptional program that eliminates the RhoA/Rock dependent cytoskeleton in RhoAQ63L/c–Myc–transformed NIH3T3 cells.
To verify whether the RhoA/c–Myc cross–talk had any effect in the biological program of RhoA–transformed cells, we evaluated the invasion and adhesion properties of the cell lines used in this study. We observed that RhoA–transformed cells displayed lower invasive rates than the parental or the c–Myc expressing fibroblasts (Fig. 7F). However, this defective invasiveness was rescued in RhoAQ63L/c–Myc–expressing cells (Fig. 7F). We also observed that RhoA–expressing cells displayed enhanced adhesion to fibronectin, collagen IV, laminin, and fibrinogen when compared to either NIH3T3 or c–Myc expressing cells (Fig. 7G). By contrast, RhoAQ63L/c–Myc–expressing cells show lower levels of adhesion to fibronectin, collagen IV and fibrinogen when compared to RhoA–transformed cells (Fig. 7G). These data indicate that the c–Myc expression status does change at least some of the cytoskeletal–related biological responses mediated by RhoAQ63L.
We have unveiled in this work a synergistic cross–talk between RhoAQ63L and c–Myc that, through modification of the cellular transcriptome, promotes the disassembly of stress fibers and focal adhesions in RhoAQ63L–transformed fibroblasts. This cross–talk entails both the upregulation and repression of genes encoding proteins involved in cytoskeletal architecture (i.e., integrin subunits) and regulation (i.e., Cdc42, Pak1) (Fig. 5B). This cross–talk exhibits three remarkable features. The first of them is that the transcriptional regulation of genes involved in this process requires synergistic signals from both RhoAQ63L and c–Myc. A second feature is that c–Myc influences negatively the RhoA/Rock–dependent cytoskeleton while it favors at the same time other RhoA–dependent routes essential for transformation status of RhoAQ63L–expressing cells. This signaling specificity can be applied even within specific branches of the Rock pathway, as inferred from the observations indicating that c–Myc reduces the phosphorylation levels of both MLC and MYPTP1 in RhoAQ63L–transformed cells without compromising the Rock–dependent transcriptome present in those cells. This is probably due to the fact that Pak1 and integrins are located downstream of Rock, a property that ensures the conservation of additional Rock downstream pathways. A third feature of this biological program is that it can be reversed rapidly by intercellular contacts or increased cell densities of cultured cells. Although the elucidation of this counterbalancing pathway remains to be elucidated, our data are consistent with the simultaneous assembly of two different regulatory steps upon the establishment of cell–cell interactions. One of those steps relies on transcriptional–and/or mRNA stability–linked processes, because we have found that the expression levels of Itgb1, Itbl1 and Itgb5 transcripts are restored in dense cell cultures. By contrast, another step occurs at the post–transcriptional level since we have observed that increased cell–to–cell contacts affecs negatively the protein levels of Pak1 while they had no effect on its mRNA levels. It is also possible that the establishment of extensive cell contacts could also favor indirectly stress fiber stability by the creation of nucleation sites for the growth of F–actin cables via, for instance, the formation of cadherin/α–catenin complexes at cell junctions.
Although counterintuitive at first sight, the negative regulation of specific downstream branches of the RhoA/Rock pathway by c–Myc may have advantages for the overall tumorigenic program of cancer cells. This is because this route represents a double edge sword for cell motility. On the one hand, it ensures the sequential adhesion and detachment of cells from the cell substrate, a process that favors cell survival, directional migration, motility, invasiveness and extravasation processes. On the other hand, the presence of high levels of RhoA and Rock activity can impair all those processes by favoring excessive and static adhesion of tumor cells to the substrate or, alternatively, by inducing cell rounding and detachment due to excessive F–actin contractility. In this functional scenario, a system that counteracts excessive levels of RhoA and Rock activity will favor the dynamic assembly and disassembly of stress fibers and focal adhesions. However, a system that eliminates permanently the possibility of proper stress fiber and focal adhesion formation will also severely impair cell motility. In this context, it is clear that a regulatory network that ensures a regulatable and selective inhibition of the RhoA/Rock–dependent cytoskeleton will have an obvious advantage to ensure fluid and dynamic F–actin structures and, at the same time, maintain other RhoA–dependent functions that are important for cell growth and tumorigenesis. Consistent with this regulatory model, we have seen that the overexpression of c–Myc does favor a change in the invasion and adhesion properties of RhoA–transformed cells. A foreseeable biological setting where this regulatory system could be highly beneficial for tumor dissemination and metastasis is the case of cancer cell subtypes associated with exacerbated signaling outputs from c–Myc and the RhoA pathway.
We thank M. Blázquez for technical help. X.R.B. work is supported by grants from the NIH (5R01CA73735), the Spanish Ministry of Science and Innovation (SAF2006–01789, GEN2003–20239–C06–01), the Red Temática de Investigación Cooperativa en Cáncer (RD06/0020/0001), the Castilla y León Autonomous Government (GR97), and the 7th Framework European Union Program (FP7–HEALTH–2007–A–201862). The authors declare that they do not have competing financial interests.
All NIH3T3 cell lines used in this work have been described previously (Berenjeno et al., 2007). In that reference and the Supplementary Information linked to it online (http://www.nature.com/onc/journal/v26/n29/suppinfo/1210194s1.html?url=/onc/journal/v2 6/n29/full/1210194a.html), readers can also find information about the proliferative and transforming properties of those cell lines, the culture conditions used, and the expression levels of the proteins ectopically expressed in them. Unless otherwise stated, we used in the experiments of this work cell lines expressing RhoAQ63L (IMB11–1–P cell line), c–Myc (15–2–7 cell line), RhoAQ63L plus c–Myc (16–2–P3 cell line), RhoAQ63L plus MadMyc (12–3–17 cell line) and RhoAQ63L plus a c–Myc–specific shRNA (15–6–35 cell line).
Additional information about other methods and reagents can be found in the Supplementary Information available online.