The Ras homologous (=Rho) sub-family of low-molecular-mass (approx. 21 kDa) GTP-binding proteins encompasses Rho (e.g., RhoA, B, C), Rac and Cdc42 proteins. Among the Rho GTPases, Rho proteins specifically have been shown to be ADP-ribosylated by exo-enzyme C3 from Clostridium botulinum, resulting in their inactivation (Aktories et al.,1988; Chardin et al.,1989; Rubin et al.,1988). This type of analysis, also the micro-injection of constitutively activated V14RhoA, have shown the involvement of Rho proteins in regulating the organization of the actin cytoskeleton (Chardin et al.,1989). Rho proteins are reported to be associated with various kinases, such as protein kinase N(PKN) (Watanabe et al.,1996), phosphoinositide-kinase (Tolias et al.,1995) and Rho-binding kinase (RocK) (Leung et al.,1995). Notably, Rho-kinase interferes with the actin cytoskeleton through phosphorylation and activation of the myosin light chain (Amano et al.,1996). Despite the high homology of the different Rho isoforms, RhoA, RhoB and RhoC (Chardin et al.,1988; Yeramian et al.,1987), it is unlikely that they have completely identical physiological functions, since (i) RhoB differs from RhoA in its C-terminal isoprenylation and intracellular localization (Adamson et al.,1992); (ii) RhoB-specific regulatory proteins have been identified (Zalcman et al.,1996); (iii) rhoB, but not rhoA or rhoC, is inducible by growth factors and genotoxic stress (Fritz et al.,1995). Whereas RhoA is thought to interfere mainly with stress-fibre formation, Rac and Cdc42 are believed to regulate formation of lamellipodia and filopodia respectively (Nobes and Hall, 1995). In addition to their interference with the microfilamental network, Rho GTPases are involved in regulation of cell-cycle progression (Olson et al.,1995), cellular differentiation (Fritz et al.,1994), genotoxic stress-induced signaling (Coso et al.,1995) and, notably, malignant transformation (Avraham and Weinberg, 1989). Transforming capacity has been reported for RhoA (Avraham and Weinberg, 1989), and RhoA, Rac as well as RhoB have been shown to be essential factors for Ras-mediated transformation (Qiu et al.,1995). Furthermore, Rho GTPases have been shown to interfere with cadherin-dependent cell-cell contacts and integrin function (Braga et al.,1997; Laudanna et al.,1996), and the Rac GDP-dissociation stimulator Tiam 1 has been shown to induce invasion, at least in the T-lymphoma-cell system (Habets et al.,1994). Altogether, these data are indicative of a possible role of Rho GTPases in tumor development and progression.
In contrast to ras, which is well known to contribute to malignant transformation in man by over-expression or constitutive activation due to specific mutations (Marshall, 1984), convincing evidence for the involvement of Rho GTPases in human carcinogenesis is still missing. In particular, no data are available regarding the expression of Rho GTPases in human tumors. Therefore, we have analyzed the expression of proteins of the Rho family (RhoA, Rac, Cdc42) in normal human tissues and tumors from colon, breast and lung. In order to take into account possible inter-individual variations in the expression level of Rho proteins, tumorigenic tissue and the corresponding normal tissue obtained from the same patient were compared. Here we show that Rho GTPases, in particular RhoA, are over-expressed in different types of human tumors, as compared with the corresponding non-tumorigenic tissues. The data indicate a role of Rho-family GTPases in human carcinogenesis.
MATERIAL AND METHODS
The origins of normal and tumor tissues from colon and breast were described by Preuss et al. (1996). Normal and tumorigenic samples from lung were obtained from Dr. K. Strebhardt (Frankfurt, Germany). 32P-NAD was obtained from NEN-Dupont (Bad Homburg, Germany), RhoA, Rac1 and Rho-GDI antibodies were purchased from Santa Cruz Biotechnology (San Diego, CA). Polyclonal antibody directed against Cdc42 was generated by immunization of rabbits with a C-terminal peptide encompassing amino acids 167 to 183. Biotinylated anti-mouse IgG as well as avidin-biotin were from Vector (Peterborough, UK) 3,3′-diaminobenzidine tetrahydrochloride (DAB) was purchased from Sigma (Deisenhofen, Germany).
Preparation of tissue extracts
Frozen normal and tumorigenic tissues originating from the same patient were dissected with a microtome. Extraction of proteins from cut slices was done as described (Preuss et al.,1996). Soluble fraction was obtained by centrifugation (10 000g, 4°C, 10 min). Protein determination was performed according to Bradford (1976). Extracts were frozen in liquid nitrogen and stored at −80°C.
32P-ADP-ribosylation with 25 μg of protein from cytosolic extracts was performed as described (Fritz et al.,1994). Reaction products were precipitated by methanol/chloroform and separated by 12.5% SDS PAGE. After staining, de-stained gels were dried and subjected to autoradiography. For quantitation, densitometrical analysis was performed. The relative amount of 32P-ADP-ribosylated proteins was calculated by referring the level of 32P-ADP ribosylation in tumorigenic tissue to that of the corresponding normal tissue, which was set to 1.0. To determine the amount of 32P-label incorporated (pMol ADP-ribose/mg protein), labeled bands were cut off from the dried gel and radioactivity was determined by scintillation counting. For statistical evaluation, Student's t-test was used.
For immunological detection of Rho proteins, 50 μg of cytosolic proteins were separated by SDS-PAGE (12.5% gel). After blotting to nitrocellulose, proteins bound to the membrane were stained by Ponceau S in order to confirm that identical amounts of protein had been transferred. RhoA-protein expression was analyzed using RhoA-specific antibody (1:2000, Santa Cruz). After incubation with peroxidase-coupled anti-rabbit IgG, RhoA protein was visualized by chemiluminescence. Subsequently, the same filter was re-hybridized with anti-Rac1 (1:800), anti-Cdc42 (1:2000), anti-Rho-GDI (1:2000) and/or anti-ERK2 (1:1000) antibody. Quantitation of the autoradiograms was done densitometrically.
Frozen, non-fixed tissue from breast carcinoma was dissected using a microtome. Sections (diameter 5 μm) were fixed with acetone and stored at −20°C. After thawing, sections were rinsed with hydrogen peroxide (1%, 20 min) and subsequently twice with PBS. After an incubation period of 1 hr at room temperature (RT) with mouse anti-RhoA antibody (1:200 in PBS/1% BSA), sections were rinsed twice with PBS and biotinylated secondary antibody (1:250; in PBS/1% BSA) was added (30 min, RT). After application of avidin-biotin staining (30 min, RT), RhoA protein was visualized with DAB. For nuclear staining, hemalaun was used.
RESULTS AND DISCUSSION
Over-expression or activation/inactivation of oncogenes and tumor-suppressor genes is well established as being related to tumor formation in man and thus useful as a prognostic marker (Bos, 1988). In particular, point mutations or amplification of members of the ras gene family have been found in a variety of human tumors (Bos, 1988). Whether the expression level of Ras-homologous Rho proteins is also related to human carcinogenesis has not yet been investigated. In the present study, we addressed this question by analyzing the amounts of various GTPases of the Rho family in different types of human tumors and compared them with those in the corresponding normal tissues originating from the same patient. To investigate the relative amounts of Rho GTPases (e.g., RhoA, B and C) in tumors, we made use of the Clostridium botulinum exo-enzyme C3-mediated 32P-ADP-ribosylation assay which specifically detects Rho proteins (Aktories et al.,1987, 1988; Rubin et al.,1988). The extent of C3-mediated ADP-ribosylation of Rho is affected both by the amount of Rho proteins and by their association with regulatory proteins (Bourmeyster et al.,1992) as well as their phosphorylation state (Fritz and Aktories, 1994). Thus, the level of ADP ribosylation of Rho can be taken as an overall indication of differences in the amount and/or physiological activity of Rho proteins between normal and tumorigenic tissues. Based on densitometrical analysis of the autoradiograms, 10 out of 11 extracts from colon tumors revealed approximately 2- to 10-fold enhanced levels of 32P labeling, as compared with the corresponding non-tumorigenic tissue (Fig. 1a). The inter-individual variation in the level of 32P-ADP-ribosylation was analyzed by quantitating the amount of 32P-ADP-ribose incorporation in extracts from normal and malignant tissues. The C3-catalyzed incorporation of ADP-ribose using extracts from normal tissue was 3.57 ± 1.06 pmol ADP ribose/mg protein, and 32P-labeling in extracts from tumors was 7.10 ± 2.37 pmol ADP-ribose/mg protein (Fig. 1b). Thus, as compared with normal tissues, extracts from tumors revealed, on average, a statistically significant (p < 0.05) higher level of ADP-ribosylation of Rho proteins.
To examine whether the observed increase in ADP ribosylation was accompanied by an increase in the amount of Rho protein, Western-blot analyses were performed. As shown for 4 pairs of normal vs. tumorigenic tissues from colon, malignancy was in most cases accompanied by an increase in RhoA protein, whereas Rac and Cdc42 protein were similar in normal and in tumorigenic tissues (Fig. 2a). The smaller band observed upon use of the anti-RhoA specific antibody (in samples T9 and T11) might be due to partial proteolytic degradation of RhoA. In contrast, slightly larger bands were observed after use of anti-Rac and anti-Cdc42 specific antibodies (in samples T8 and T11). Whether these low-mobility bands represent different post-translationally modified Rac (Cdc42) protein or an immunologically related cross-reacting protein is unclear. Out of 11 colon tumors assayed in the paired samples, 10 over-expressed RhoA protein, and 5 of them even exhibited a large (5- up to 50-fold) increase in the amount of RhoA (Fig. 2b). Comparison of the level of RhoA protein with the extent of 32P-ADP-ribosylation of a specific tumor showed no strict correlation. This is probably due to the fact that the level of ADP ribosylation of Rho proteins is affected not only by their quantity, but also by post-translational modification (Bourmeyster et al.,1992; Fritz and Aktories, 1994, 1992; Ohtsuka et al.,1998). Thus, the data suggest that, at the same time as quantitative changes of RhoA have occurred, alterations affecting the activity of RhoA may also have happened during the process of malignant transformation. Besides changes in the expression level of RhoA, tumor-specific changes in the amount of RhoB and RhoC proteins (which are substrate for C3-mediated ADP ribosylation as well) may be involved in affecting the level of 32P-ADP-ribosylation. Lacking suitable antibodies, we were unable to analyze the expression of these Rho species by immunological methods. Analysis of their mRNA expression will be the subject of future studies.
To see whether enhanced 32P-ADP-ribosylation and over-expression of RhoA protein is specific for colon tumors, we investigated normal tissues and tumors from breast. As shown in Figure 3a,32P-ADP-ribosylation in non-tumorigenic control tissue was at the border of detection. In contrast, the level of 32P-ADP-ribose incorporation in extracts from the corresponding tumors was dramatically enhanced (7- to 300-fold), as revealed by densitometrical analysis of the autoradiograms (Fig. 3a). Incorporation of ADP ribose in extracts from non-tumorigenic tissues was 0.96 ± 1.08 pmol ADP-ribose/mg protein, whereas the 32P-labeling of extracts from breast tumors was 5.11 ± 4.57 pmol ADP-ribose/mg protein (Fig. 3b). Thus, the level of ADP-ribosylation of tumor extracts revealed wide variation and was, on average, clearly higher than that observed in normal tissues. On the level of the protein, the amount of RhoA, Rac and Cdc42 in normal tissues was either below the limit of detection (RhoA and Rac) or only very poorly detectable (Cdc42) (Fig. 3c). In contrast to non-malignant samples, tumorigenic tissues exhibited high levels of expression of all these GTPases (Fig. 3c). Expression of the Rho guanine-dissociation inhibitor (Rho-GDI) as well as the ERK2 kinase was also enhanced in tumors (Fig. 3c). In opposition to the Rho GTPases, the expression of RhoGDI and ERK2 proteins was clearly detectable also in normal tissue (Fig. 3c). The observed over-expression of the MAP kinase ERK2 in breast tumors is in line with a report by Sivaraman et al. (1997) indicating that changes in signaling occur during breast-tumor development. A high level of RhoA protein in breast tumors was also seen by means of immunohistochemistry (data not shown), supporting the data obtained by ADP-ribosylation and Western blotting.
Since over-expression of RhoA was observed in tumors from colon and from breast, we hypothesized that over-expression of this GTPase might be a marker of malignancy in different kinds of tumors. To provide support for this hypothesis, we determined the expression of RhoA protein using paired samples of normal and malignant tissues from lung (n = 4). Here too, normal tissues showed only very faint amounts of RhoA, whereas all tumor probes tested revealed largely enhanced levels (up to 50-fold) of RhoA protein (Fig. 4). Interestingly, in addition to RhoA a second protein with a slightly higher molecular weight was detected by the RhoA-specific antibody used. As discussed before, this larger band is supposed to be due to cross-hybridization of the antibody with a RhoA-related protein over-expressed in lung tumors. Alternatively, it may represent a RhoA-protein fraction with different post-translational modification (e.g., phosphorylation, isoprenylation).
Because RhoA is involved in the regulation of the microfilamental network and cadherin-dependent cell-cell contact (Braga et al.,1997; Nobes and Hall, 1995) and the metastatic potential of tumors is related to actin organization (Zachary et al.,1986), the question arose whether the amount of RhoA might be related to the state of malignancy. To address this, 32P-ADP-ribosylation experiments with extracts from breast tumors classified as WHO grade I and WHO grade III were performed. As shown in Figure 5a, the mean level of 32P-ADP ribosylation of extracts from the grade-III tumors was significantly higher (p < 0.05) than that of the grade-I tumors. Also, the mean expression level of RhoA protein was significantly higher in the grade-III tumors than in the grade-I tumors (p < 0.01) (Fig. 5b). Based on these findings and those of studies reporting that Rho proteins evoke metastatic properties in vivo (Del Peso et al.,1997), the expression level of Rho might be considered as a useful prognostic factor for tumor progression.
Taken together, our findings indicate that Rho GTPases, in particular RhoA, are related to malignancy in man. Since, under in vitro conditions, over-expression of RhoA results in transformation of mouse fibroblasts (Avraham and Weinberg, 1989), the data support the view that increase in RhoA is a critical event in malignant transformationn and/or progression of human cells. Overall, 20 out of 21 tumors derived from 3 different types of tissues (colon, breast, lung) showed over-expression of RhoA (as compared with the normal tissue of the same individual). This corresponds to a frequency of approximately 95%. The molecular basis for over-expression of RhoA remains to be elucidated. Also, the question of whether or not tumor-specific mutational activation of Rho GTPases (i.e., RhoA, RhoB, RhoC, Rac and Cdc42) can occur, will be subject of our forthcoming studies. As compared with mutations in p53 and ras and over-expression of myc, which are observed in about 30 to 60% of tumors (Alitalo et al.,1987; Bos, 1988; Hollstein et al.,1991), over-expression of RhoA appears to be a more general phenomenon in tumors. The data may be taken as indication that RhoA is suitable as a diagnostic marker for malignancy. To prove this, further studies with a larger number of different types of tumors are in progress.
We thank Drs. M. Kaufmann (Heidelberg, Germany), T. Beck (Mainz, Germany) and K. Strebhard (Frankfurt, Germany) for tissue samples from colon, breast and lung.