In recent years, the list of tumor suppressor genes (or candidate TSG) that are inactivated frequently by epigenetic events rather than classic mutation/deletion events has been growing. Unlike mutational inactivation, methylation is reversible and demethylating agents and inhibitors of histone deacetylases are being used in clinical trails. Highly sensitive and quantitative assays have been developed to assess methylation in tumor samples, early lesions, and bodily fluids. Hence, gene silencing by promoter hypermethylation has potential clinical benefits in early cancer diagnosis, prognosis, treatment, and prevention. The hunt for a TSG located at 3p21.3 resulted in the identification of the RAS-association domain family 1, isoform A gene (RASSF1A). RASSF1A falls into the category of genes frequently inactivated by methylation rather than mutational events. This gene is silenced and frequently inactivated by promoter region hypermethylation in many adult and childhood cancers, including lung, breast, kidney, gastric, bladder, neuroblastoma, medulloblastoma, gliomas and it has homology to a mammalian Ras effector (i.e., Nore1). RASSF1A inhibits tumor growth in both in vitro and in vivo systems, further supporting its role as a TSG. We and others identified the gene in 2000, but already there are over a 150 publications demonstrating RASSF1A methylation in a large number of human cancers. Many laboratories including ours are actively investigating the biology of this novel protein family. Thus far, it has been shown to play important roles in cell cycle regulation, apoptosis, and microtubule stability. This review summarizes our current knowledge on genetic, epigenetic, and functional analysis of RASSF1A tumor suppressor gene and its homologues.
- epigenetic inactivation
- cell cycle arrest
The RAS-association domain family 1, isoform A (RASSF1A) tumor suppressor gene (TSG), is a member of a new group of RAS effectors thought to regulate cell proliferation and apoptosis. Formally known as 123F2, two laboratories independently cloned and sequenced RASSF1 ( 1, 2). Loss of heterozygosity (LOH) studies in lung, breast, and kidney tumors identified several loci in chromosome 3p likely to harbor one or more tumor suppressor genes, including 3p12, 3p14, 3p21.3, and 3p25-26. In 1993, the von Hippel-Lindau (VHL) TSG inactivated in familial and sporadic clear cell renal cell carcinomas was identified from region 3p25 confirming this hypothesis ( 3). This prompted the search for TSGs within the other regions of 3p. An important TSG was suspected to reside in 3p21.3 because instability of this region is the earliest and most frequently detected deficiency in lung cancer. Overlapping homozygous deletions in lung and breast tumor cell lines reduced the critical region in 3p21.3 to 120 kb and this region was found to be exceptionally gene rich. From this critical region, eight genes were identified, including CACNA2D2, PL6/placental protein 6, CYB561D2/101F6, TUSC4/NPRL2/G21, ZMYND10/BLU, RASSF1/123F2, TUSC2/FUS1, and HYAL2/LUCA2. However, despite extensive genetic analysis in lung and breast tumors, none of these candidate genes were frequently mutated. Meanwhile, Dammann et al. ( 2) identified RASSF1 in a yeast two-hybrid screen baited with XPA. Although the interaction was not formally shown, this novel gene was of interest because it mapped to the 3p21.3 entry in Genbank, it shared high-sequence homology with a known RAS effector in mouse (Nore1), and expression of one of the isoforms, RASSF1A, but not RASSF1C, was lost in most lung tumor cell lines. Subsequently, hypermethylation of the RASSF1A promoter region CpG island was identified as the major cause for loss of expression. Cells treated with demethylating agents reexpressed RASSF1A, which confirmed the role of DNA methylation in the inactivation of RASSF1A in tumor cell lines. Finally, overexpression of RASSF1A but not RASSF1C in non–small cell lung cancer (NSCLC) A549 cells reduced colony formation efficiency and suppressed growth of tumor cells in nude mice in vitro and in vivo growth assays, respectively.
Since the discovery of the first tumor suppressor gene, RB, Knudson's “two-hit” hypothesis ( 4) has been used to define this class of genes. The hypothesis states that inactivation of both alleles of a TSG, classically by genetic insult, are required for tumorigenesis. Recently, this model was extended to include epigenetically inactivated genes. Hypermethylated in cancer-1, HIC1, a gene on 17p13.3, is hypermethylated in a number of sporadic human tumors. In a study of Hic1+/− mice by Chen et al. ( 5), development was normal but the mice were predisposed to malignant tumors at 70 weeks. Examination of the tumors revealed loss of Hic1 expression in the context of an intact allele. Methylation analysis then showed that the remaining allele was inactivated by hypermethylation. Hence, inactivation of Hic1 in this mouse model satisfied Knudson's criteria for a TSG. The first “hit” being the genetic knockout of one allele and the second “hit” being hypermethylation of Hic1 promoters. Using a combination of combined bisulfite and restriction analysis and LOH analysis inactivation of RASSF1A in sporadic lung tumors was shown to obey Knudson's model for a TSG. In SCLC, hypermethylation was detected in 76% tumors with allelic imbalance at 3p21.3 indicating that genetic and epigenetic mechanisms provide the two “hits” to inactivate RASSF1A ( 6). Similarly, RASSF1A methylation was detected in 52% of NSCLC, 72% of bladder transitional cell carcinoma, and 70% of cervical squamous cell carcinoma (SCC) with allelic imbalance at 3p21.3 ( 7– 9). In primary medulloblastoma, however, in the absence of allelic imbalance at 3p21.3, there is evidence to suggest biallelic methylation is the major mechanism for RASSF1A inactivation in this tumor type ( 10).
RASSF1 Gene Locus and Protein Structure
The RASSF1 gene locus spans about 11,151 bp of the human genome and is comprised of eight exons. Differential promoter usage and alternative splicing generates seven transcripts (RASSF1A-G; Fig. 1 ). Isoforms A and C are ubiquitously expressed, whereas isoform B is mainly expressed in cells of the hemopoietic system. Isoforms D and E are specifically expressed in cardiac and pancreatic cells, respectively. They are very similar to RASSF1A except for slight differences in the splice sites used in exons 2αβ (RASSF1D) and 3 (RASSF1E) providing each protein with four additional amino acids. Two CpG islands are associated with the promoter regions of RASSF1. The smaller of the two (737 bp, 85 CpGs, 71.5% GC and OE:0.89) spans the promoter region of RASSF1A (and RASSF1D, RASSF1E, RASSF1F, and RASSF1G). The second CpG island (1365 bp, 139CpG, 67.9% GC and OE:0.88) encompasses the promoter regions for RASSF1B and RASSF1C. The entire first exon of each RASSF1 transcript is contained within the CpG islands.
A Ras association or RalGDS/AF-6 domain encoded by exons 4 and 5 defines the RASSF1 gene and is located at the COOH terminus of isoforms A-E. This domain mediates interactions with Ras and other small GTPases. Despite the absence of any sequence homology, it has a similar structure to the RasGTP binding domain of Raf1 (kinase), the most studied RasGTP effector. The Ras association domain of RalGDS is comprised of a five-stranded mixed β-sheet and three α helices. Ras association domains homodimerize via intermolecular disulfide bonds, formed using two cysteine residues from each monomer. The interaction between RalGDS homodimers and Ras is mediated mostly by two antiparallel β-strands within the Ras association domain and Ras. RASSF1A and RASSF1D-G isoforms have a NH2-terminal protein kinase C (PKC) conserved region 1 (C1) domain encoded by exons 1α and 1β. In PKC, this is a diacylglycerol (DAG)/phorbol ester binding domain that regulates kinase activity. Overlapping the C1 domain is a putative zinc-binding domain, ZnF_NFX. This domain is present in a transcriptional repressor of HLA-DRA, NK-X1, and in Drosophila shuttle craft protein, vital for the late stages of embryonic neurogenesis. A SARAH domain, (Sav/RASSF/Hpo) is present at the COOH terminus of RASSF1A-E. SARAH domains mediate heterotypic interactions between proteins as shown for Hpo/Sav and homotypic interactions as shown with Mst1. In vitro studies have identified a peptide sequence within exon 3 that is a substrate for ataxia telagectasia mutant (ATM) kinase. ATM is vital for the activation of p53 following exposure to ionizing radiation. Phosphorylation levels of the RASSF1 peptide sequence WETPDLSQAEIEQK were comparable with levels of the ATM consensus sequence in p53 suggesting that RASSF1 might also be a substrate for ATM ( 11). The RASSF1 ATM site is present in isoforms A, C, D, and E.
RASSF1 Homologues and Orthologues in Model Organisms
RASSF1A homologues with 38% to 85% identity are present in rodents, fish, and nematodes ( Fig. 2A ). Each has a COOH-terminal Ras association domain and SARAH domain and a NH2-terminal C1 domain. In Xenopus, a RASSF1C homologue (MGC82975) has been identified but the genomic sequence is not available to verify the presence of RASSF1A. In Drosophila, an 89.6-kDa protein (LD40758p) has the closest homology with RASSF1A. However, although it has a Ras association domain at the COOH terminus, a LIM domain not a C1 domain is predicted at the NH2 terminus. Homology searches within the human genome have identified other Ras association domain containing genes called RASSF2 (20pter-p12.1), RASSF3 (12q14.1), AD037 (10q11.21), NORE1 (1q32.1), and RASSF6 (4q21.21; Fig. 2B). In common with RASSF1, these genes code for multiple transcripts and CpG islands are associated with their promoter regions. The Ras association and SARAH domains of these genes are located at the COOH terminus indicating that they share similar structural organization with RASSF1. NORE1 which has the most sequence identity with RASSF1 (49%) also has a putative NH2-terminal C1 domain.
Tumor-Associated Methylation of RASSF1A
Since the discovery of RASSF1A inactivation in lung tumors, a wealth of literature has been published implicating RASSF1A in the pathogenesis of a wide spectrum of tumors ( Table 1 and references within).
A direct correlation between promoter methylation and loss of RASSF1A expression has been shown in many tumor cell lines, including SCLC, NSCLC, breast, kidney, nasopharyngeal carcinoma, prostate, ovarian, hepatocellular, Hodgkin's lymphoma, melanoma, thyroid, bladder, neuroblastoma, medulloblastoma, rhabdomyosarcoma, and retinoblastoma ( 2, 10, 12, 13, 15, 16, 18, 21, 22, 24, 25, 27, 31– 35, 38). In most studies, reverse transcription-PCR for the expression of RASSF1C was used as a control for RNA integrity and loading. This served a duel purpose because it emphasized that methylation of the RASSF1A promoter region does not affect expression of the C-isoform. Expression of RASSF1B and RASSF1F have also been examined ( 2, 12, 13, 25). Northern blotting showed RASSF1B is expressed most strongly in hemopoeitic cells and expression was lost in two of four tumor cell lines of lymphoid origin ( 2). Loss of RASSF1B expression coincided with loss of RASSF1A expression in this study. Similarly, loss of RASSF1B expression occurred in seven of eight bladder cancer cell lines with loss of RASSF1A expression and was recovered along with RASSF1A following treatment with demethylating agent ( 13). Whereas the reasons for the loss of RASSF1B expression are not yet understood, none of the samples analyzed exclusively down-regulated RASSF1B. Conversely, expression of RASSF1F is intimately connected with RASSF1A expression because they share a common promoter region ( 15, 25). Hence, reexpression of RASSF1A in 5-aza-2′ deoxycytidine–treated neuroblastoma, rhabdomyosarcoma and retinoblastoma cell lines is coincident with RASSF1F reexpression ( 12) Although not reported, expression of RASSF1E, RASSF1D, and RASSF1G are also likely to be linked with expression of RASSF1A because they also share the same promoter region. Expression of RASSF1A has been investigated in primary tumors and matched normal tissue. Whereas contaminating normal cells can be problematic, down-regulation of RASSF1A was shown to correlate with promoter methylation in breast, renal, thyroid, and bladder tumors ( 13, 21, 24, 35, 41).
RASSF1A methylation has been reported in at least 37 tumor types. Normal tissue controls are used in the majority of studies to show methylation is tumor specific and therefore involved in tumorigenesis. Methylation of RASSF1A occurs rarely in normal tissues. Combined with the high frequency reported in cancer, this makes RASSF1A a candidate molecular marker for tumor diagnosis. Detection of methylation in premalignant breast cancer lesions and in sputum samples from patients who later developed lung cancer ( 45, 46) have highlighted the potential importance of RASSF1A in early diagnosis. Several studies have investigated the prospect of using DNA methylation as a tumor cell marker in samples obtained by noninvasive techniques such as in plasma, sputum, urine, throat washings, and nipple aspirates ( 47– 50, 53, 56). When comparing RASSF1A methylation in nasopharyngeal carcinoma with corresponding nasopharyngeal swab and mouth/throat washings, Chang et al. showed 100% concordance and 98% concordance with plasma and buffy coat samples ( 47). In ovarian cancer, detection of RASSF1A methylation in corresponding peritoneal fluid and serum from patients with all histologic types, grades, and stages of disease even in samples, where the CA-125 (serum marker) levels were low (<35), underscored the importance of having a reliable tumor marker when early detection is crucial to patient outcome ( 48). In addition, RASSF1A methylation in urine from kidney cancer patients corresponded with methylation of the tumors in 73% of cases ( 53). Compared with other methods of tumor surveillance, detection of methylated DNA by methylation-specific PCR is very sensitive and relatively cheap. Therefore, RASSF1A methylation has the potential to be used (along with a panel of other TSGs to ensure 100% coverage) as a marker for early detection and monitoring.
The value of RASSF1A methylation as a prognostic marker has been investigated most in NSCLC. Tomizawa et al. suggest that RASSF1A methylation correlated with poor survival rate in patients with stage I lung adenocarcinoma disease ( 7). RASSF1A methylation was associated with poorly differentiated tumors, predominantly with vascular invasion and pleural involvement. A similar correlation with poor survival was also reported by Burbee et al. ( 15). Conversely, three separate studies did not find any correlation with survival and RASSF1A methylation ( 57, 59, 60). Instead, Wang et al. suggested RASSF1A methylation was predictive of early relapse ( 60).
Correlation between RASSF1A Methylation and Other Oncogenic Events
Due to the evidence linking RASSF1A to Ras signaling pathways, several studies have looked for correlation between mutation of K-Ras and inactivation of RASSF1A by methylation. An inverse correlation was observed in colorectal cancers ( 61), pancreatic adenocarcinomas ( 62), and NSCLC ( 63). However, a different study of NSCLC ( 64) found no correlation between methylation at RASSF1A and activating mutations in K-Ras; however, individuals with both defects had a poorer outcome suggesting a synergistic mechanism of action. Synergy was also proposed in melanomas ( 65) where in addition to mutation screening of N-Ras, K-Ras, and H-Ras, B-Raf was also analyzed and whereas no tumors with B-Raf mutations had additional mutations in N- or K-Ras, the majority of tumors with RASSF1A methylation had mutations in either B-Raf or N-Ras. However, in thyroid cancer the reverse situation was observed such that no tumors in which RASSF1A was methylated had mutations in B-Raf ( 66). These inconstancies may reflect other alterations in Ras signaling ( 67).
Cervical cancer and head and neck squamous cell carcinoma (HNSCC) are associated with human papillomavirus (HPV) infection of the tumor cells. The HPV can only replicate in dividing cells; therefore, it encodes viral proteins (E6 and E7) to subvert control of the cell cycle by inactivating p53 and Rb, respectively. DNA from HPV types 16 and 18 are particularly associated with transformation. HPV DNA was never found in cervical carcinomas with RASSF1A methylation ( 68) suggesting that the presence of viral proteins abrogated any requirement for RASSF1A inactivation implicating them both in the same pathway. A similar correlation was observed in a study of HNSCC ( 69). However, other studies ( 9, 70) have not found an association between RASSF1A methylation and absence of viral DNA.
The ubiquitous human herpes virus, EBV, can transform B cells into continually growing lymphoblastoid cell lines in vitro. Viral DNA and transcripts are associated with a number of neoplasia, including nasopharyngeal carcinoma, endemic Burkitt's lymphoma, Hodgkin's disease, and gastric carcinomas. Whereas RASSF1A methylation is detected in the majority of nasopharyngeal carcinomas (67%), it is difficult to draw any correlations regarding virus infection of the tumor cells because virtually all cases are EBV positive. However, compared with other head and neck tumors, RASSF1A methylation occurs at a much higher frequency in nasopharyngeal carcinoma. Comparisons between virus infection and RASSF1A methylation have been possible with Hodgkin's lymphoma and gastric carcinomas because only a subset of these are EBV associated. In a cohort of 52 Hodgkin's lymphomas, 27 were EBV positive and 34 showed RASSF1A methylation ( 18). However, RASSF1A methylation did not correlate with EBV infection. In gastric carcinoma, a correlation between EBV infection and RASSF1A methylation was detected ( 71). RASSF1A was methylated in 14 of 21 (66.7%) of EBV-positive gastric carcinomas, whereas only 2 of 56 (3.6%) of EBV-negative gastric carcinomas were methylated. In this study, EBV-associated methylation of several other TSGs, including PTEN, p16, THBS1, and MINT12 was also examined and in general the frequency of CpG island methylation was higher in virus infected gastric carcinomas. Methylation of the viral genome occurs shortly after infection to restrict expression of viral transcripts which enable EBV to evade immune surveillance. This data suggest that EBV is associated with a methylator phenotype in gastric carcinoma and the virus may be involved in disrupting normal DNA methylation mechanisms. Coincidentally, another study on gastric carcinomas found RASSF1A methylation was associated with poorly differentiated gastric carcinomas ( 72), which is also the histologic subgroup most associated with EBV.
Malignant mesothelioma arises from mesothelial cells lining the peritoneum, pleura, and pericardium. About 50% of malignant mesotheliomas are associated with SV40. In a study comparing the methylation of profile of 66 malignant mesotheliomas and 40 lung adenocarcinomas, Toyooka et al. showed methylation of RARβ, CDH13, MGMT, p16, and APC was significantly greater in lung adenocarcinomas (22-52%) than in malignant mesotheliomas (0-10%; ref. 73). However, methylation of RASSF1A was not significantly different between the two tumor types, 32% malignant mesothelioma versus 43% lung adenocarcinoma. SV40 Tag was present in 48% (32 of 66) of malignant mesotheliomas and was shown to correlate with a higher methylation index than in SV40-negative malignant mesotheliomas. Furthermore, RASSF1A methylation was significantly higher in SV40-positive malignant mesotheliomas (48%) than SV40-negative cases (16%). The authors investigated the connection between aberrant methylation and SV40 by infecting mesothelial cells with virus and examining promoter methylation in early (8-30) and late (15-86) passaged cells. No promoter methylation was detected in early passaged cells but in two of six foci of late-passage cells. RASSF1A promoter methylation was detected with concomitant down-regulation of RASSF1A. These cells also displayed morphologic signs of transformation ( 74). Expression was recovered using 5-aza-2′ deoxycytidine but not with the histone deacetylase inhibitor trichostatin A indicating that SV40 induces methylation rather than histone deacetylation of the RASSF1A promoter.
Epigenetic Inactivation of Other Ras Association Domain Family Members
NORE1A but not NORE1B was shown down-regulated in lung tumor cell lines. Down-regulation of NORE1A expression correlated with promoter CpG island hypermethylation ( 75) and was found methylated in 24% of NSCLC tumors. Two missense changes were identified in a mutational screen of 20 lung tumor cell lines ( 75). Hypermethylation of the NORE1A CpG island in NSCLC has an inverse correlation with K-RAS mutations suggesting that they have a cooperative role in carcinogenesis ( 76). Chen et al. ( 77) identified two breakpoint-spanning genes, LSAMP and NORE1 in a familial clear cell renal cell carcinoma (CCRCC)–associated translocation t(1;3)(q32.1;q13.3). NORE1A was found methylated in 32% of sporadic CCRCC. RASSF2, like NORE1, is widely expressed (with particularly high levels in the brain and blood), isoform A of RASSF2 (RASSF2A) is frequently methylated and silenced in colorectal and lung tumors and tumor cell lines. 1 ADO37 seems expressed in most tissues and is methylated in lung and breast tumors but rarely in nasopharyngeal carcinomas or glioma tumors ( 78, 79). RASSF3 is unmethylated in human cancers ( 78), whereas RASSF6 methylation status in tumors has not been reported.
Methylation and RNA Interference
During embryogenesis, tissue-specific DNA methylation patterns are established for the correct regulation of imprinted genes such as insulin-like growth factor II and X chromosome inactivation. The majority of the rest of the genome is methylated with the exception of CpG islands, most of which are associated with the promoter regions of housekeeping genes. As well as transcriptional regulation, methylation of the genome is thought to suppress the activity of transposable elements (e.g., LINE-1 and Alu) that can disrupt gene function. Some of the DNA methyltransferases (DNMT1, DNMT2, DNMT3a, DNMT3b, and DNMTL) involved in these processes have been identified. DNMT1 ensures maintenance, whereas DNMT3b is involved in de novo DNA methylation. However, in tumor cells, a paradox exists with global hypomethylation of the genome, hypermethylation of CpG islands, and overexpression of DNMTs. Hypomethylation contributes to tumorigenesis through loss of imprinting, genomic instability, and activation of transposable elements. Whereas the mechanisms that lead to aberrant CpG island methylation are not fully understood, recent data by Kawasaki et al. may provide clues to unraveling this mystery ( 80). Many organisms use RNA interference (RNAi) to regulate gene expression post-transcriptionally. MicroRNAs (miRNAs), 21 to 23-nucleotide ds-RNAs, are the effector molecules for RNAi. Exploitation of this mechanism using chemically synthesized small inhibitory RNAs (siRNAs) is proving very fruitful for functional genomics, especially for the study of tumor suppressor gene inactivation in cell lines. However, if siRNA is designed to target CpG islands within gene promoters, sequence-specific methylation of DNA is induced with concomitant down-regulation of transcription. siRNA-directed methylation of the E-cadherin promoter has been shown in breast a tumor cell line ( 80). How this is achieved is not yet understood, but recent microarray data from 27 prostate cancers suggests that endogenous antisense transcripts required to generate siRNAs are synthesized in vivo from intragenic promoters including RASSF1 ( 81). It is tempting to speculate, therefore, that RNAi may lie at the heart of epigenetic inactivation of RASSF1A and other commonly inactivated TSGs.
Mutation Analysis of RASSF1A Gene
As already mentioned, mutation of RASSF1A occurs rarely in human cancers. As far as we can ascertain, only one frame-shift mutation at codon 277 (within the Ras association domain) has been reported derived from a nasopharyngeal carcinoma ( 31). A number of missense changes and polymorphisms have been found in nasopharyngeal carcinoma, lung, breast, and kidney carcinomas ( 6, 15, 21, 31). Many of these localize to the functional domains of RASSF1A, three in the DAG binding domain, four in the ATM phosphorylation domain, and five in the Ras association domain ( Figs. 2B and 3 ). The functional significance of these changes has not been fully investigated, but data suggests that they are defective. For instance, unlike wild-type RASSF1A, a C65R/V211A mutant does not suppress growth of LNCaP prostate carcinoma cells or KRC/Y renal cell carcinoma cells, in vitro ( 34, 82). Phosphorylation of RASSF1A is reduced in the S131F mutation of the putative ATM site, which also results in less efficient inhibition of cell proliferation ( 83). Two missense changes, C65R and R257Q, relocate RASSF1A from the cytoplasm to the nucleus and have diminished ability to halt the cell cycle ( 84). This strongly suggests that cellular localization of RASSF1A is also vital to its function. Hence, study of naturally occurring tumor associated missense changes is proving fruitful for dissecting out the relative importance of the different putative functional domains of RASSF1A.
Frequent inactivation of RASSF1A gene in human cancers suggests that it must have a pivotal role in tumor prevention. This notion is supported by the phenotype of tumor cell lines with constitutive overexpression of RASSF1A. The observations in NSCLC, prostate, kidney, nasopharyngeal carcinoma, and glioma cell lines ( 2, 15, 21, 34, 78, 85) indicate that RASSF1A expressing cells are less viable, growth suppressed, less invasive, and have reduced anchorage/substrate independence. Using a tetracycline regulation system, RASSF1C has also been shown to inhibit colony formation in vitro ( 82). In addition, mutations detected in a gene inactivation test suggest that RASSF1C may also function as a TSG ( 82). Subsequent studies have indicated RASSF1A functions as a modulator of two pathways commonly deregulated in cancer, apoptosis, and cell cycle ( 83, 86). Microarray analysis of tumor cell lines (NSCLC and neuroblastoma) stably expressing RASSF1A identified expression changes consistent with the observed phenotypic effects such as reduced cell cycle effectors (Cyclins D1 and D3), increased cell adhesion molecules (N-cadherin) and alterations in the levels of a number of extracellular matrix modifiers (SPARC and ANPEP; ref. 87). Several of 66 RASSF1A-regulated genes identified by this study are also affected by activated Ras during cellular transformation. Interestingly, our data found that RASSF1A and Ras had the reciprocal effect on their expression, which might be further evidence to suggest that these molecules do indeed function in the same pathway. By identifying RASSF1A-interacting proteins ( Figs. 3 and 4 ), we are beginning to understanding how RASSF1A achieves growth suppression or induces apoptosis and may provide new therapeutic targets for the treatment of human cancers.
Microtubules, Adhesion, and Migration
The first mice lacking RASSF1 were made by Smith et al. ( 88) who generated a targeting vector to the mouse chromosome syntenic to the minimal region of human chromosome 3p21.3 deleted in lung tumors, this 370-kb deletion knocked out 12 genes including rassf1. Heterozygous mice with this contiguous gene deletion were viable and fertile; however, the homozygous null embryos died before 13.5d.p.c. Subsequently, mice were generated with a targeted deletion of the rassf1-coding region ( 89). These mice have been bred to homozygosity and shown to have no obvious phenotype; however, they have not been grown to old age nor have they been treated with tumor promoting agents. Interestingly mouse embryonic fibroblasts (MEFs) generated from rassf1 null mice are smaller than wild-type MEFs and show increased sensitivity to the tubulin depolymerizing agent nocodazole ( 89). We and others have shown that transfection of RASSF1A into RASSF1A-negative cell lines induced the microtubular structures radiating from the centrosome to become hyperstabilized circular structures and also protected the cells from the action of nocodazole ( 84, 89– 91). Microtubular association of tumor suppressor proteins is not without precedent, as both the adenomatous polyposis coli (APC) and von Hippel-Lindau (pVHL) proteins have been shown to bind to and stabilize microtubules suggesting that microtubule destabilization may be an important facet of tumorigenesis. The RASSF1A mutants C65R and R257Q, which fail to localize to the microtubules, were unable to protect against this nocodazole-induced depolymerization ( 84). Deletion analysis ( 89) showed that a region between amino acids 120 and 288 of RASSF1A was required for microtubule association, and a subsequent deletion analysis defined a smaller region amino acids 120 to 185 as the microtubule association domain ( 90). The RASSF1A peptide comprised of amino acids 120 to 185 destabilized microtubules and was termed dominant-negative RASSF1A for its ability to inhibit RASSF1A-induced cell death. Acetylation of α-tubulin is associated with microtubule stability, and transfection of NCI-H1299 cells with RASSF1A induced high levels of acetylated α-tubulin; however, the RASSF1A mutants C65R and R257Q were less competent at induction of microtubule acetylation ( 84). The microtubule stability induced by RASSF1A could have implications for cell adhesion and motility, especially in the light of data showing that genes for cell adhesion and motility such as tropomyosin I and CDH2 (N-CAD) were up-regulated in A549 cells stably expressing RASSF1A ( 87).
Wild-type RASSF1A was shown to associate with microtubules in immunofluorescence ( 84, 89) and cosedimentation assays, whereas mutant versions showed varying abilities to colocalize with microtubules. Fewer than 30% of cells transfected with C65R or R257Q showed RASSF1A localizing to microtubules, whereas ∼90% of wild type and 80% of mutants K21Q, S131F, A133S, R201H, V211A, Y325C, A336T localized to the microtubules ( 84). Bromodeoxyuridine incorporation assays showed that the mutants deficient in their ability to bind microtubules (C65R or R257Q) were also deficient in their ability to stop DNA synthesis in NCI-H1299 cells ( 84) suggesting a potential connection between competency to bind microtubules and ability to induce cell cycle arrest.
The localization of RASSF1A can be seen to alter during mitosis ( 92); cytoplasmic microtubular association was confirmed during interphase. However, as the cells progressed into prophase, RASSF1A relocated to the separated centrosomes, then to the spindle fibers and poles during metaphase and anaphase and finally to the midbody during cytokinesis in HeLa cells.
Microtubule association is likely to play an important role in the function of RASSF1. A yeast two hybrid screen ( 84) for RASSF1-interacting proteins showed binding to MAP1B and C19ORF5 (also called BPY2IP or VCY2IP1); in fact, 70% of interacting clones had homology to microtubule-associated proteins. The RASSF1A interacting protein C19ORF5 colocalized with RASSF1A to the microtubules, but overexpression of C19ORF5 unlike overexpression of RASSF1A could not protect against the depolymerizing effects of nocadozole. RASSF1C can also associate with microtubules ( 90); however, it is less effective at stabilizing them. RASSF1C has also been shown to interacts with C19ORF5 ( 93), which in turn can interact with the SEC-1 domain of leucine-rich pentapeptide repeat motif–containing protein (LRPPRC, also called gp130 and LRP130). LRPPRC also interacts with UXT ( 93), which in turn interacts with BUB3 that is involved in the kinetochore checkpoint that ensures cells containing misaligned chromosomes do not exit from mitosis prematurely. C19ORF5 and LRPPRC colocalize with β-tubulin in the cytoplasm; however, in apoptotic cells, they are found to colocalize in the nucleus ( 93). An interaction between C19ORF5 and the mitochondrial proteins (NADH-dehydrogenase subunit 1, cyclooxygenase-1) and LRPPRC associates RASSF1 with mitochondria, an organelle with pivotal functioning in control of apoptosis.
RASSF1 was predicted to form a soluble cytoplasmic protein with a Ras association domain ( 1). Ras is a small inner membrane–embedded GTPase that relays proliferative signals from cell surface receptors such as receptor tyrosine kinases. Ras becomes activated by binding to GTP causing a shift in conformation revealing cytoplasmic epitopes that mediate interactions with effector proteins such as Raf kinases; Ras signaling may be an important chemotherapeutic target. There are three Ras family members, H-Ras, K-Ras, and N-Ras, which are located at 11p15.5, 12p12.1, and 1p13.2, respectively. K-Ras has been shown to most strongly interact with RASSF1. RASSF1 has been shown to directly bind to Ras in a GTP-dependent fashion ( 86) and also to interact via heterodimerization with NORE1 ( 94). Ras is an important modulator of the apoptotic response and is thought to mediate cell survival in response to hyperproliferative signals in transformed cells through PI3-K– and Tiam-1–mediated pathways and to induce apoptosis in untransformed cells through as yet incompletely understood pathways that may be mediated by RASSF1 and NORE1.
The difficulty of generating NIH 3T3 cells stably expressing RASSF1C gave an initial clue that RASSF1 may play a role in apoptosis ( 86). Activated Ras-GTP can directly bind RASSF1C and can enhance RASSF1-induced apoptosis in 293 cells ( 86). Transient transfections were used to show that RASSF1C could induce cell death in 293 cells, and that activated Ras could augment this cell death, whereas dominant-negative Ras inhibited the cell death. The caspase inhibitor z-VAD-fmk reduced RASSF1C-induced cell death implicating an apoptotic mechanism. This information was augmented by demonstration in a yeast two hybrid assay that human RASSF1, murine NORE1, and Caenorhabditis elegans T24F1.3 all specifically bound to the Ste20-related proapoptotic kinase mammalian Sterile20-like (MST1) through their common conserved COOH-terminal tails ( 95). MST1 and MST2 (also called STK4 and STK3, respectively) are serine/threonine kinases that initiate apoptosis when overexpressed ( 96). The Drosophila loss of MST function mutant (hippo) fails to undergo accurate developmental apoptosis. MST becomes autoactivated by autophosphorylation of threonine (Thr183 of MST1 and Thr180 of MST2), and this autoactivation of MST1 is inhibited by cotransfection with RASSF1A or RASSF1C or NORE1A or NORE1C but is augmented by cotransfection of membrane-localized NORE1A-CAAX ( 97). MST1 contains caspase cleavage sites, cleavage with caspase 3 generates a 36-kDa fragment seen during apoptosis, and cleavage with caspase 6 or 7 generates a 41-kDa product.
Activated Ras can bind NORE1 and RASSF1. RASSF1 and NORE1 can form both homodimers and heterodimers ( 94), which complex with Mst-1 regulating its seriene/threonine kinase activity. The MST1 binding site is at the COOH-terminal end of RASSF1A (amino acids 358-413 of NORE1; ref. 95). Whereas overexpression of MST1 could induce apoptosis in NIH 3T3 cells, and coexpression with NORE1 had a small effect on increasing apoptosis, the addition of a CAAX motif to NORE1 to induce membrane targeting enabled NORE1 alone to induce apoptosis, and NORE1 together with MST1 had a greater proapoptotic effect than either protein alone ( 95). This cell death was inhibitable by the caspase inhibitor z-VAD-fmk ( 95). A model was proposed ( 97) whereby NORE1 and MST1 constitutively form a complex maintaining MST in an inactive reservoir until activation by serum stimulation is able to induce association with Ras.
CNK1 (CNKSR1) is a scaffold protein that in Drosophila is required for Ras to activate Raf kinase. Transfection of CNK1 into 293 cells can induce apoptosis ( 98), and this can be abrogated by dominant-negative inhibitors of MST1/2. CNK1 has been shown to bind to RASSF1A or RASSF1C but not to NORE1 ( 98). RASSF1 has previously been shown to interact with MST1/2 ( 95); therefore, it was thought that RASSF1 may be providing the mechanism linking the two proteins; this was shown by deletion of the C-terminal (RASSF1 interacting) region of MST1 that prevented MST1 from interacting with CNK1. Despite the fact that both RASSF1A and RASSF1C can interact with CNK1, only RASSF1A could augment CNK-induced apoptosis ( 98).
Deregulation of cell cycling is an essential prerequisite for tumorigenesis. In normal cells, cycling is exquisitely controlled by a number of protein complexes whose activity is required for the cell to pass through specific checkpoints. For example, mitotis-promoting factor, a cyclinB/Cdc2 complex that phosphorylates mitotic regulators allowing the cell to progress from G2 to M.
Anaphase promoting complex/cyclosome (APC/C) is a protein complex that interacts with ubiquitin-conjugating and activating enzymes to catalyze the poly-ubiquitylation of proteins destined for degradation ( 99) to allow the cell cycle to progress. APC/C is activated by complexing with Cdc20 or cdh1 (proteins that contain WD40 repeats). These WD40 repeat proteins are required for APC/C to interact with target proteins. During S, G2, and prophases, APC/C is inhibited by sequestration of Cdc20 by Emi1; later during prometaphase, RASSF1A takes on the role of regulator as it in turn sequesters Cdc20, and coimmunoprecipitation of RASSF1A with Cdc20 was shown by Song et al. ( 92). To allow progression this repression by RASSF1A must be released allowing APC/C to become active inducing the polyubiquitylation and degradation of cyclin A. Coexpression of Cdc20 with RASSF1A suggests that it is the relative levels of these proteins that is important. During prometaphase, RASSF1A and Cdc20 are colocalized, but during mitosis, only a small proportion coprecipitates ( 92). Overexpression of RASSF1A leads to accumulation in prometaphase with raised levels of cyclins A and B ( 92); this arrest is before the metaphase to anaphase transition. The spindle checkpoint keeps the cells in metaphase until chromosomes are correctly aligned; the kinetochores induce assembly of the inhibitory proteins Mad2 and Mad3 on incorrectly tensioned spindles. These inhibit the APC/C until alignment is correct when the repression is relieved allowing the APC/C-Cdc20 to ubiquitinate securin (a separase inhibitor) targeting it for degradation and initiating sister chromatid separation and anaphase.
During telophase, an APC/C-cdh1 complex targets mitotic cyclins for destruction resulting in exit from mitosis. The APC/C complex is kept in its inactive form by phosphorylation of cdh1 inhibiting its ability to bind APC/C. cdk1 maintains cdh1 in its phosphorylated state, whereas Cdc14 induces dephosphorylation of cdh1 when the spindle becomes correctly orientated.
The cyclins D1 and D3 were also shown RASSF1A regulable, because they were down-regulated in A549 cells stably transfected with RASSF1A ( 87). Another cyclin, cyclin A (that regulates CDK2 thereby controlling progression through S phase) is regulated by the transcription factor p120E4F ( 100). We have shown that p120E4F can bind to RASSF1A ( 101) and this interaction was shown mediated by amino acids 1 to 119 of RASSF1A. Cotransfection of RASSF1A with p120E4F-induced G1 arrest of a greater magnitude than that induced by transfection with either construct alone. p120E4F provides a mechanistic link to other known tumor suppressor genes such as p14ARF, Rb and p53 that are known to interact with p120E4F. This suggests that RASSF1A maybe able to affect cyclin A expression.
Systematic studies were undertaken by two hybrid screens to determine possible RASSF1-interacting partners. Mst-1, C19ORF5, MAP1B, p120E4F, and CNK1 were identified as interacting partners (refs. 84, 95, 98, 101; Figs. 3 and 4). Additionally, the catalytic domain of plasma membrane calmodulin-dependent calcium ATPase (PMCA4b, also called ATP2B4) was found via a bacterial two hybrid screen to interact with RASSF1 through amino acids 74 to 123 of RASSF1C or amino acids 144 to 193 of RASSF1A ( 102). PMCA4b is involved in expelling calcium from cells and may also play a role in signaling because it interacts with proteins such as nitric oxide synthase I and calcium/calmodulin–dependent serine protein kinase. The interaction between RASSF1 and PMCA4b reduced epidermal growth factor (EGF)–dependent activation of Erk, a downstream target of the Raf, Ras, and mitogen-activated protein kinase cascade ( 102).
Functional Analysis of Other RASSF Family Members
NORE1 (novel Ras effector, also called RASSF5) was originally identified ( 103) from a mouse T-cell library by its ability to bind to Ras-GTP in a yeast two hybrid screen. Confirmation of this interaction by immunoprecipitaion showed that NORE1 only bound to Ras after stimulation with EGF or 12-O-tetradecanoylphorbol-13-acetate. NORE1 expression inhibited cell growth and this inhibition was augmented by expression of H-Ras, whereas cotransfection with the antiapoptotic agent Bcl2 blocked this growth inhibition implicating an apoptotic mechanism ( 104). Transfection of NORE1A or NORE1B into cell lines with low endogenous NORE expression suppressed colony formation in A549 and G361 lines (which have disrupted Ras signaling due to a Ras activating mutation or constitutively active B-Raf kinase respectively). However, when transfected into other lines with similar disruptions to Ras signaling (NCI-H460 and M14), there was no effect on colony formation ( 105). In the A549 cell line, NORE1A inhibited anchorage-independent growth and induced G1 arrest but failed to induce apoptosis ( 105). Deletion of the COOH-terminal MST-interacting domain and the Ras binding domain and/or mutation of the zinc finger domain had minimal effect on the growth suppressive activities of NORE1, leading Aoyama et al. ( 105) to conclude that the growth suppression of NORE1 was due to amino acids 188 to 250. However, binding to MST1/2 or Ras-like GTPases had previously been shown required for growth suppression ( 95, 104). NORE1 had been shown to induce apoptosis in 293T cells ( 104) rather than the cell cycle delay seen in A549 cells ( 105).
RASSF2 was first identified by Comincini et al. ( 106) and it seems up-regulated in radiation workers ( 107). RASSF2 (originally called Rasfadin or KIAA0168) can bind to K-Ras in a GTP-dependent manner through its Ras effector domain ( 108); however, it only weekly binds to H-Ras. Overexpression of RASSF2 inhibited growth of lung tumor cell lines, and Vos et al. ( 108) were unable to generate cell lines stably overexpressing RASSF2. In transient assays, RASSF2 inhibited cell growth, and this inhibition was augmented when cotransfected with K-Ras, whereas H-Ras had little additional effect. Cell death induced by RASSF2 coexpressed with activated K-Ras was shown mediated by caspase 3 and hence via an apoptotic mechanism; cell cycle analysis also showed a decreased proportion of cells in G2-M implicating G0-G1 arrest.
AD037 (also called RASSF4) has been shown to bind activated, but not wild-type K-Ras and for the two proteins to act synergistically to induce apoptosis in 293T cells. AD037 can also inhibit the growth of human tumor cell lines, and this inhibition can be enhanced by the addition of a tag (CAAX) to induce membrane localization thus mimicking the effect of activation ( 79). No functional studies on RASSF3 or RASSF6 have been reported.
RASSF1A tumor suppressor gene undergoes frequent tumor-specific epigenetic inactivation in a wide range of tumors. RASSF1A methylation has also been shown in preneoplastic lesions and in patient's DNA obtained using noninvasive procedures such as urine from kidney cancer patients and sputum from lung cancer patients. Hence, RASSF1A methylation can potentially be developed as a molecular biomarker for screening cancer patients and populations at risk. At least three other RASSF1 homologues (NORE1, AD037, and RASSF2) are also inactivated in tumors by promoter hypermethylation. Whereas somatic mutational inactivation is a rare event for all RASSF family members analyzed thus far. Although, several of the missense changes reported in RASSF1A have been shown functionally relevant. The inverse correlation reported for RASSF1A and NORE1A inactivation and K-ras alterations in some tumor types may provide alternative pathways for affecting Ras signaling.
Studies are in progress to elucidate the function of the protein products of the RASSF1 family of TSGs. Reintroduction of RASSF1A in tumor cell lines lacking endogenous expression decreased in vitro colony formation and in vivo tumorigenicity. It can induce apoptosis likely through interactions with the MST proteins. RASSF1A can also bind to and stabilize microtubules, a property that seems central to its function, and through modulation of APC/C activity can affect cell cycle regulation. Rassf1−/− mice are viable and fertile; it remains to be seen if they are prone to increased spontaneous and or induced tumor formation. There has been an explosion of reports on RASSF1A methylation in cancer; hopefully, the next few years will yield further insights into the biology of this important family of tumor suppressor genes.
After this article went to press, RASSF1A isoform-specific knockout mice were reported to demonstrate increased susceptibility to spontaneous and chemically induced tumors (Tomassi S, Denissenko MF, Pfeifer GP. Cancer Res 2005;65:92–8).
Grant support: Sport Aiding Medical Research for Kids, the Wellcome Trust, Breast Cancer Campaign, Association for International Cancer Research, and Birmingham Children's Hospital Research Fund.
Note: Due to space limitation, it has not been possible to include all references relating to RASSF1A methylation.
↵1 Hesson and Latif, unpublished data.
- Received November 15, 2004.
- Revision received January 19, 2005.
- Accepted February 1, 2005.
- ©2005 American Association for Cancer Research.