NORE1B Is a Putative Tumor Suppressor in Hepatocarcinogenesis and May Act via RASSF1A

Hepatocellular carcinoma (HCC) is among the most frequent types of cancer worldwide accounting for approximately half a million deaths per year (1). The molecular mechanisms underlying the development of this disease are not well-understood (2, 3). In sharp contrast to other types of human malignancies, activating mutations in one of the Ras genes occur rarely in HCC (3–5). We recently found epigenetic inactivation of the Ras effector genes NORE1B and RASSF1A in the majority of HCC due to CpG-site hypermethylation in the promoter regions. As a result, 97% of the 28 tumors studied revealed epigenetic inactivation of NORE1B, RASSF1A, or both (6). However, the Ras effector NORE1A was not affected by hypermethylation. This raised our interest in the significance of NORE1B and RASSF1A for the development of liver cancer.

Ras effectors are defined as proteins with strong affinity to GTP-charged forms of Ras, Rap-1, and several other Ras subfamily GTPases (5). Currently, there are >10 different Ras effectors, such as RAF, RalGDS, phosphatidylinositol 3-kinase, and the RASSF gene family comprising RASSF1 and NORE1 (RASSF5; refs. 5, 7–11). The interaction of these effectors with GTP-charged Ras may trigger various signaling cascades, which induce not only growth, survival, and migration of cells but also cell cycle arrest, differentiation, senescence, and apoptosis (5, 7–10). These diverse biological effects may serve for the fine tuning of the cellular homeostasis. However, details in the interplay between the various Ras effector-mediated pathways and the biological consequences are still unknown.

Members of the RASSF family are deleted or repressed selectively by gene loss and/or epigenetic mechanisms in a considerable fraction of epithelial cancers and cell lines derived thereof. Re-expression usually suppresses the proliferation and tumorigenicity of these cells (7, 8–14). Accordingly, RASSF1A knockout mice show enhanced tumor susceptibility, and loss of function of the NORE1 gene is linked to a familial form of kidney cancer (15, 16). Thus, RASSF proteins act as tumor suppressors, which contrasts to the oncogenic effects of most other Ras effectors known, such as Raf or PI3-K.

The predominant RASSF family members are all characterized by the presence of a Ras association domain (RA) of the RalGDS/AF6 type (7–10). As a result of alternative splicing and/or differential promoter usage RASSF1 codes for RASSF1A and six further isoforms (RASSF1B-G), whereas NORE1 gives rise to NORE1A and NORE1B (also called RAPL). RASSF1A, NORE1A, and NORE1B all contain a COOH-terminal SARAH-domain, which may mediate interactions with bax-activating protein kinases (9, 17–19). NORE1B harbors a RA-domain like NORE1A, but binding to GTP-charged Ras proteins has never been proven. Because NORE1B transcription starts from an alternative exon 2β, located ∼48 kb downstream of exon 1, the product lacks the NH₂-terminal diacylglycerol-binding domain (DAG), the presumed linking site for RASSF1A. RASSF1A associates not directly with Ras-like GTPases but heterodimerizes with NORE1A, serving as link to GTP-charged Ras (9, 20, 21). Thus, the actual binding partners and functions of NORE1B have not been identified.

To understand the significance of the frequent NORE1B silencing in hepatocarcinogenesis, we studied the wild-type gene and several deletion mutants lacking the known functional domains. The results show that NORE1B antagonized cell transformation and delayed the cell cycle as effectively as the tumor suppressor RASSF1A. These effects require the RA and SARAH-domain, which serve to interact with RASSF1A and not with GTP-charged Ras. In conclusion, our data indicate a close cooperation of NORE1B and RASSF1A in the maintenance of cell homeostasis. Thus, the functional loss of NORE1B and/or RASSF1A, as observed in almost all HCCs, may contribute significantly to uncontrolled growth and transformation of hepatocytes.

Plasmids and constructs. Human NORE1B, NORE1A, RASSF1A, pECFP-tagged Ha-Ras-wt, and pECFP-tagged Ha-Ras-G12V were kind gifts from J. Avruch (Massachusetts General Hospital, Boston, MA), A. Khokhlatchev (University of Virginia, Charlottesville, VA) R. Dammann (University of Halle, Germany), and A. Wittinghofer (Max-Planck Institute, Düsseldorf, Germany), respectively. NORE1B-deletion mutants lacking the NH₂ terminus or the SARAH-domain were generated by PCR, deletants without RA-domain by site-directed mutagenesis, as described (22). For a list of primers and PCR conditions, see Supplementary Data. The coding regions of NORE1Bwt, deletion mutants, RASSF1Awt, and NORE1Awt, were sequenced and subcloned in-frame into pEYFP or pECFP.

Cell-lines. The human hepatocarcinoma cell-line Hep3B (ATCC-No HB-8064) was maintained as described (6). The absence of mutations in exons 2 and 3 of N-, Ha-, and Ki-Ras was verified by sequencing. Primer sequences are given in the Supplementary Data and methodical details in a study by Macheiner and colleagues (6). MIM-1-4 immortalized murine hepatocytes, deriving from p19ARF−/− mice, and MIM-R hepatocytes, established from p19ARF−/− and transformed with oncogenic v-Ha-Ras, were generated and cultured as given elsewhere (23). Cells were retrovirally transduced with pMSCV-GFP (ClonTech) and in a second round with pBabepuro harboring NORE1Bwt, NORE1Awt, or RASSF1Awt. Cells were kept in 5 μg of puromycin/mL medium for ∼2 wk and resistant clones were isolated and propagated.

Reverse transcription-PCR. Total RNA was extracted, reversely transcribed applying random primers and Moloney murine leukemia virus reverse transcriptase, and PCR products were analyzed by electrophoretic separation, as described (6).

siRNA. Silencer Select siRNAs targeting human NORE1B with the sequences 5′-CACUGCUAAGACUACCUUUTT-3′ and 5′-AAAGGUAGUCUUAGCAGUGAA-3′ were custom-made by Ambion (“siNORE1B”). A nonsilencing Silencer Select scrambled siRNA (“siSCRAMB,” No AM4390843; Ambion) and Cy3-labeled Silencer glyceraldehyde-3-phosphate dehydrogenase (GAPDH; “siGAPDH,” No AM4649; Ambion) were used as controls. Cells were transfected with either 30 nmol/L (siSCRAMB and siNORE1B) or 10 nmol/L (siGAPDH) siRNAs using siLentFect (BioRad) according to the manufacturer's instructions and were incubated for 24 h until analyses. TaqMan-based “gene expression assays” for β2-microglobulin, GAPDH, and NORE1B (ABI) were applied on an ABI-Prism/7500 Sequence Detection System (ABI). Changes in expression levels were determined using β2-microglobulin as house-keeping gene according to previous descriptions (6). All data were analyzed in duplicates. For further details, see Fig. 1.

Immunoblotting. Cells were centrifugated for 5 min at 800 × g at 20°C. The pellet was washed with PBS and cells were suspended in lysis buffer. Lysates were incubated on ice for 30 min, sonicated, and centrifugated for 5 min at 13.000 × g at 4°C. Protein samples (25 μg per lane) were separated by SDS-PAGE (12%) and electroblotted to Hybond-P membranes (Amersham Bioscience). Immobilized proteins were detected using antisera, as listed in the Supplementary Data. The staining signal was detected by enhanced chemiluminescence (Amersham Bioscience) on Kodak-X-Omat sheets (Kodak). Due to high homology, anti-NORE detected NORE1A and NORE1B of human and rodent origin.

Confocal laser scanning microscopy. Hep3B-cells were transfected with 1 μg of pEYFP- or pECFP-tagged constructs or with 0.5 μg of each plasmid DNA in case of double-transfections applying Fugene 6 (Roche Applied Bioscience); 24 h later, cells were fixed with formalin for 30 min, and the DNA was stained with Hoechst 33258 (0.8 μg/mL in PBS). Images were obtained by a LSM510 Meta-confocal microscope (Carl Zeiss).

Fluorescence resonance energy transfer. Cells were transfected as described above. After 24 h, images were obtained with a Nikon Eclipse TE300 microscope (Nikon), equipped with a SenSys camera, and applying a CFP (donor)-, a YFP (acceptor)-, and a RAW-fluorescence resonance energy transfer (FRET; acceptor emission at donor excitation)-filter. The FRET signals were calculated with MetaMorph 5.0 software as an intensity-modulated display image (24). After determination of the correction factor for CFP and YFP, FRET was calculated using the formula: FF − (d × DF) − (a × AF); FF, FRET filtersignal; d, donor correction factor; DF, donor filtersignal; a, acceptor correction factor; AF, acceptor filtersignal. For further details see ref. 25. The fusion construct pEYFP/CFP served as positive control (a kind gift from JA Schmid, Medical University of Vienna, Vienna, Austria).

Cell transformation and cloning efficiency. Primary rat embryo cells (REC) were isolated at day 15.5 of gestation by fractionated trypsinization and cultivated as described in detail (26); 24 h after plating cells were transfected via the calcium phosphate technique applying 20 μg of DNA containing 6 μg of pVEJ.B (encoding Ha-Ras), 6 μg of pSPmyc (encoding c-Myc), 8 μg of pBabepuro containing either NORE1Bwt, NORE1B deletion mutants, RASSF1Awt, NORE1Awt, or pBabepuro only. About 72 h later, cells were switched to a medium containing 5% FCS. After ∼10 d, the dishes were stained with Giemsa and the number of cell colonies was counted.

To test for cloning efficiency, RECs were transfected as described above and were selected with geneticin (200 μg/mL; Invitrogen) After 9 d, clones were isolated and maintained in selection medium.

Cell cycle analysis. Cells were fixed in 3% paraformaldehyd, RNA was digested with RNase A (10 mg/mL; Qiagen), and DNA was stained with propidium iodide (2.5 μg/mL; Sigma). Analyses were performed with a FACS Calibur Flow Cytometer (Becton Dickinson).

Statistics. Where indicated, statistical analysis were performed by the nonparametric Krukal-Wallis test. All tests were calculated as two-tailed.

Human NORE1B encodes a 265 amino acid (aa) protein, which consists of a unique 40 residue NH₂ terminus, followed by the CENTRAL-domain (aa 41–115), RA-domain (aa 116–207), and COOH-terminal SARAH-domain (aa 208–265; refs. 7, 9). To understand the functions of these domains we generated the following deletants: NORE1B_Nterm lacking aa 1-107, NORE1B_RA lacking aa 113-226, and NORE1B_SARAH without aa 222-265. For further information, see Supplementary Data. The effects of the wild-type form and the deletion mutants on growth, transformation, and on subcellular localization were studied.

NORE1B delays proliferation of hepatoma and hepatocyte lines. By fluorescence-activated cell sorting (FACS) analyses, we first checked the effect of NORE1B, NORE1A, and RASSF1A on the cell cycle of the human hepatoma cells Hep3B, a line with inherently low endogenous expression of these genes (6). The transfection stressed the cells as indicated by absence of the G₂-M phase. Expression of NORE1Bwt induced an increase of cells in G₀-G₁ by 32%. This effect was somewhat less pronounced for NORE1B_Nterm and NORE1B_RA, and was absent when transfecting NORE1B_SARAH (Fig. 1A). NORE1Awt and RASSF1Awt also elevated the G₀-G₁ fraction of Hep3B-cells. On the other hand, knockdown of NORE1B reduced the percentage of cells in G₀-G₁ by ∼12% and elevated the S- and G₂-M fraction together by ∼40% (Fig. 1B and C). This was first evidence that NORE1B may delay the cell cycle, as known for RASSF1A and NORE1A, and that mainly the SARAH-domain is of importance for this effect (27–33).

To obtain stable expressions close to physiologic levels, we inserted NORE1Bwt, NORE1Awt, and RASSF1Awt retrovirally into murine hepatocyte lines without (MIM-1-4) and with mutant Ha-Ras (MIM-R). In all sublines generated, the inserts were transcribed (Fig. 2A), whereas proteins remained below the detection level (data not shown). The vector alone had nearly no effect on the growth of the cells (Fig. 2B). However, replication was significantly reduced in MIM-1-4 and MIM-R cells expressing NORE1B, NORE1A, or RASSF1A as manifested by significantly decreased cell numbers after a cultivation period of 3 days. When calculating the mean duration of each cycle phase, NORE1Bwt prolonged the G₀-G₁ phase fraction in both cells types and the G₂-M phase in MIM-1-4 cells (Fig. 2D). A similar effect was observed with NORE1A and RASSF1A.

NORE1B antagonizes cell transformation. RECs were transfected with plasmids encoding for c-Myc/Ha-Ras, which produced numerous transformed cell clones (Fig. 3). Cotransfection with NORE1Bwt or NORE1B_Nterm significantly lowered focus formation by 24%. NORE1B_RA and NORE1B_SARAH exerted no decrease indicating that the RA- and SARAH-domain may be important for interference with c-Myc/Ha-Ras induced cell transformation. The number of foci was not significantly affected by RASSF1Awt or NORE1Awt. However, the combination of RASSF1Awt and NORE1Bwt was even more effective than NORE1Bwt alone and suppressed focus formation by ∼50% (Fig. 3). This indicates that NORE1B with an intact RA- and SARAH-domain lowers cell transformation and that this effect becomes even more evident in the presence of RASSF1A.

We isolated six clones per group, all of which expressed the ectopic constructs at mRNA and protein levels (Fig. 4). Unlike the control clones with a fibroblastoid morphology and 0.4 cell doublings per day, NORE1Bwt-clones showed an epitheloid morphology (Fig. 3B) and a somewhat reduced population doubling time of 0.3 per day (Fig. 4). RASSF1Awt, however, did not alter morphology (data not shown) but significantly reduced the population doublings when compared with control- and NORE1Bwt-clones. We conclude that NORE1B impairs the development and growth of c-Myc/Ha-ras transformed cells and changes the morphology of the transformed clones.

Subcellular localization of NORE1B, NORE1A, and RASSF1A. In Hep3B cells, YFP-tagged NORE1Bwt was localized in the cytoplasm and excluded the nucleus (Fig. 5). RASSF1Awt impressed with a meshwork-like pattern in the cytoplasm (Fig. 5). Dot-like structures close to the nucleus were formed by NORE1Awt (Fig. 5). The subcellular distribution of NORE1B_Nterm was still cytoplasmic, whereas NORE1B_RA was localized punctiform around and in the nucleus and NORE1B_SARAH predominantly nuclear. NORE1B harbors a nuclear export signal (NES) in the RA-domain close to the SARAH-domain. When blocking the nuclear export by leptomycin B (34), NORE1Bwt was localized predominantly in the nucleus (Fig. 5). Thus, the lack of cytoplasmic occurrence of the NORE1B_RA or NORE1B_SARAH was probably due to lack or steric hindrance of the NES, respectively. The unaltered cytoplasmic location of NORE1B_Nterm under leptomycin B-treatment may be due to the deletion of a hitherto unidentified nuclear localization signal close to the NH₂ terminus. These data indicate that NORE1B may be active in the cytoplasmic and nuclear compartment of the cell.

NORE1B interacts with RASSF1A but not with wild-type or mutant Ha-Ras. To elucidate the molecular target of NORE1B, we double-transfected Hep3B cells. The presence of RASSF1Awt altered the subcellular localization of NORE1Bwt and NORE1B_Nterm to a RASSF1A-like pattern but had no effect on the other NORE1B-deletants (Fig. 6A). As a result, NORE1Bwt or NORE1B_Nterm distributed differently when compared with single transfections and exhibited colocalization with RASSF1A. To check for interactions between the two molecules, we performed FRET analyses (Fig. 6B). The cotransfection of RASSF1A and NORE1A, a known RASSF1A-binding partner, produced a significant FRET signal. Transfecting RASSF1A with either NORE1Bwt or NORE1B_Nterm also resulted in considerable signals, whereas the combinations of NORE1Bwt/NORE1Awt, NORE1B_RA/RASSF1A, and NORE1B_SARAH/RASSF1A were ineffective (Fig. 6). This strongly suggests that NORE1B interacts physically with RASSF1A, which involves the RA- and the SARAH-domain.

NORE1A has been described to bind GTP-charged Ras proteins by its RA-domain (20). Accordingly, in double-transfected Hep3B-cells NORE1A altered the subcellular distribution and colocalized with mutant Ha-Ras only and not with the wild-type form (Fig. 6C). The absence of the FRET signal does not exclude interaction of NORE1A and mutated Ras because the energy transfer requires parallel orientation of donor and acceptor dipoles of the two fluorophores (35). These preconditions are often not met by the labels of the transfectants. When cotransfecting NORE1Bwt with either wild-type or mutant Ha-Ras, NORE1B neither changed its subcellular localization nor did it produce a significant FRET signal (Fig. 6D). Thus, a physical interaction of NORE1B with Ras in its wild-type or mutated form seems unlikely.

In the majority of HCCs investigated, NORE1B was found to be epigenetically silenced (6). The present report provides several lines of evidence supporting that NORE1B is a putative tumor suppressor gene and cooperates closely with RASSF1A, as outlined in the following.

NORE1B, RASSF1A, and cell cycle. In hepatocyte and hepatoma cell lines, NORE1B, NORE1A, and RASSF1A increased the percentage of cells in G₀-G₁ at the expense of the S-phase fraction. Retroviral insertion of NORE1B and RASSF1A in hepatocyte lines tended to increase additionally the G₂-M fraction. As a result, cell cycle progression was significantly delayed and growth of the lines suppressed. Identical effects were reported for RASSF1A and NORE1A in cells derived from melanoma, kidney, cervix, breast, lung, and prostate cancer (27–30, 32, 36). Data on hepatocytes or hepatoma cells have not been available. The molecular mechanisms of the G₁-S and G₂-M arrest by RASSF1A are largely identified. The protein complexes with the transcription factor p120^E4F, thereby delaying the G₁-S transition via interference with p14ARF, retinoblastoma, p53, and cyclin A (9, 10, 27, 31). During the interphase RASSF1A localizes to microtubules and is found in centrosomes and spindles during mitosis by binding to RABP1 (also called C19ORF5; ref. 37). Overexpression of RASSF1A causes a prometaphase arrest by preventing activation of the anaphase-promoting complex/cyclosome, whereas depletion of RASSF1A accelerates mitotic progression and causes mitotic defects (9, 10, 29, 33). Details of the molecular mechanisms underlying the reduced S-phase fraction by NORE1B, as observed in the present study, have yet to be defined. According to our data, the SARAH-domain and to some extent also the RA-domain of NORE1B are essential for growth suppression.

NORE1B and cell transformation. Under our experimental conditions, NORE1B antagonized c-Myc/Ha-Ras–induced transformation of embryonal cells. Only NORE1B constructs harboring the RA-/SARAH-domains and being capable of binding to RASSF1A, were effective. Interestingly, RASSF1A alone did not significantly antagonize cell transformation but enhanced greatly the NORE1B effect, which indicates cooperation of these genes. In the clones generated, the effects of NORE1B and RASSF1A seemed to persist as being evident by a reduced population doubling time over an observation period of several weeks.

Previous studies described a colony-suppressing activity of NORE1B, NORE1A, and RASSF1A in cancer cell lines deriving from kidney, lung, or melanoma (8, 12, 13, 21, 38). Colony formation in already transformed cell lines is due to altered cell-cell adhesion and anchorage independence, partly reflects the process of tumor progression, and differs greatly from the transformation of unaltered and mortal primary cells. Our data add to the profile of NORE1B and RASSF1A, indicating that these genes interfere with processes already at the very beginning of carcinogenesis.

Putative interaction partners of NORE1B. Harboring a RA-domain like NORE1A, NORE1B has been anticipated to bind GTP-charged Ras (20). However, after cotransfection of NORE1Bwt and mutant Ha-Ras, there was no evidence for any interaction. Furthermore, NORE1B affected the cell cycle in Hep3B cells harboring wt-forms of Ha-, Ki-, and N-Ras and in MIM-hepatocytes without or with activated Ras. Thus, NORE1B seems to act independently of Ras and loss of NORE1B function may render a growth advantage to hepatocytes with or without Ras mutation.

NORE1B lacks the NH₂-terminal DAG-domain, the putative RASSF1A binding site, and nevertheless interacts with this gene via the RA- and SARAH-domain. Cotransfections of RASSF1A with NORE1B-constructs containing the critical domains produced significant FRET-signals. Generally, efficient energy transfer of >50% through the resonant coupling of the dipole moments of the donor (CFP-tagged NORE1B) and acceptor (YFP-tagged RASSF1A) requires parallel orientation and proximity of <7 nm of the two fluorophores, and an interaction time beyond random molecular collisions (35). Thus, FRET proves the molecular proximity between the two macromolecules, NORE1B and RASSF1A, and strongly suggests close physical interaction of the two peptides. Furthermore, FRET analyses are superior to conventional biochemical binding studies in that they allow to study living cells (35).

Via heterodimerization NORE1B seems to be recruited to the microtubular binding sites of RASSF1A, as the colocalizing gene products seem to arrange as fibers. The molecular mechanisms of growth arrest by RASSF1A largely involve interactions with microtubules, as outlined above. Similar mechanisms may underlie the growth suppressing effects of NORE1B. Furthermore, as essential components of cell adhesion, intercellular contacts, and/or the cytoskeleton microtubuli affect cell morphology (39). Accordingly, by binding to the microtubuli, NORE1B may induce the observed shift from fibroblastoid to an epitheloid, cobblestone-like morphology in the embryonal cells.

Lack of NORE1A silencing and hepatocarcinogenesis. As part of a multiprotein complex RASSF1A controls the cell cycle and recruits protein kinases to induce apoptosis (9, 10, 17–19, 27–33). Our data show that RASSF1A heterodimerizes with both, NORE1A and NORE1B, but that only NORE1A seems to serve as link to GTP-charged Ras. Thus, it may be a critical point for a cell whether the RASSF1A-containing protein complex attaches to NORE1A or NORE1B and thus may be linked to active Ras or not. In human HCC NORE1B and RASSF1A are epigenetically silenced but not NORE1A (6). Impaired function of NORE1A by mutations of the gene has already been excluded (6). Because most of the HCCs lack Ras mutations, the adaptor function of NORE1A between RASSF1A and GTP-charged Ras-proteins may be of minor importance for hepatocytes to abrogate harmful Ras effects. This may explain that NORE1A is not frequently epigenetically silenced in the HCCs.

In conclusion, our data suggest that NORE1B and RASSF1A interact closely to antagonize cell transformation and to control growth of hepatocytes. Thus, functional integrity of both genes seems to be fundamental in preventing the development of liver cancer.

No potential conflicts of interest were disclosed.

Grant support: “Herzfeldersche Familienstiftung” and by the Austrian “Gen-Au Programme” (study No GZ 200.058/6-VI/2/2002).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.