Specificity of the Methylation-Suppressed A Isoform of Candidate Tumor Suppressor RASSF1 for Microtubule Hyperstabilization Is Determined by Cell Death Inducer C19ORF5

Introduction

Epigenetic inactivation specifically of the RASSF1A isoform by hypermethylation of promoter-specific CpG islands has been reported in at least 20 types of primary human tumors at a frequency that exceeds most other candidate tumor suppressor genes (1). This and induction of cell death and reduction in anchorage-independent growth and tumorigenicity by overexpression in cultured cells (2–4) suggest a role of RASSF1A in tumor suppression. However, the feature that distinguishes RASSF1A from RASSF1C, which would explain its specific role in tumor suppression, is unclear. Because of the Ras-association domain (RA) that is common to all isoforms, RASSF1 gene products have been proposed to play a role in ras-mediated cell growth suppression (5, 6) . A difference in A and C isoforms linked to ras signaling pathways has not been clear. When overexpressed RASSF1A is a powerful microtubule-stabilizing agent similar to the anticancer drug paclitaxel and thus a strong inhibitor of DNA synthesis, mitosis, and cytokinesis (7–11) . As expected of a strong paclitaxel-like microtubule stabilizer that blocks the cell cycle, RASSF1A stabilizes cyclins A, B, and D1 (11, 12) and inhibits the anaphase-promoting complex (11). A weak interaction in vitro with Cdc20, an activator of anaphase-promoting complex, has been reported (11) but the significance of the interaction remains to be established. Cells deficient in RASSF1 continue to exhibit normal Mad2- and Emi1-related spindle checkpoint controls (11).

RASSF1A (39 kDa) and RASSF1C (32 kDa) are distinguished by their unique N termini ( Fig. 1A ). Downstream are shared common structural domains comprised of an ATM domain defined by a consensus ATM kinase phosphorylation site, the RA domain, and a common C terminus. A minor isoform, RASSF1B, lacks the RASSF1A or RASSF1C N terminus and the ATM domain (1). Because of the selective suppression of RASSF1A in human tumors, studies of function have focused on RASSF1A without a rigorous comparison of RASSF1A with RASSF1C and other isoforms. In several reports where both isoforms have been studied in model test systems, they exhibit identical properties including induction of cell death and tumor suppressor activity (6, 9, 13) . The microtubular association and effect of RASSF1C on stabilization has been reported to be undetectable (11), weaker than (9), or equal to RASSF1A (14). A unique property that could underpin tumor suppression by only the RASSF1A but not the RASSF1C isoform remains to be established.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Overexpression of RASSF1A and RASSF1C, but not RASSF1B, isoforms similarly disrupts cell cycling by a paclitaxel-like hyperstabilization of microtubules. A, sequence homology domains of RASSF1 isoforms and mutant constructs and PCR primers used for mutant constructs. RASSF1A and RASSF1C exhibit unique N termini upstream of the ATM domain. Numbering of amino acid residues is based on RASSF1A and lengths in amino acid residues are indicated at top and right. RASSF1A151-340 is a natural splice variant RASSF1B. C1, phorbol ester/diacylglycerol binding domain that exists in protein kinase C and other “nonkinase” phorbol ester receptors (1, 30) ; ATM, ataxia-telangiectasia mutated domain that contains an ATM kinase concensus phosphorylation site; RA, Ras-association (RalGDS/AF-6) domain. B, comparison of association with and impact on microtubules of GFP-RASSF1A, GFP-RASSF1C, and GFP-RASSF1B in transiently transfected COS7 cells. COS7 cells transfected with cDNA coding for the GFP labeled construct indicated at the top were left untreated (a-e) or treated with 10 μmol/L of paclitaxel (f-j) or nocodazole (k-o) at the time of transfection. Tubulin was detected by mouse anti-β1-tubulin and Texas red–conjugated anti-mouse antibody (red) and merged with the green GFP signal in a single panel. Panels are representative of more than 95% of at least 2,000 cells examined with a positive GFP signal from three independent transfection experiments. Yellow, colocalization of GFP and β-tubulin (red). Bars, upper left, 10 μm. C, cellular distribution and impact of RASSF1C is identical to RASSF1A independent of fusion tag or cell type. The distribution of GFP-RASSF1C transfected into the indicated cell type (a and d) was observed directly and those tagged with HA (b) and c-Myc (c) indirectly with antibody against HA or c-Myc and secondary Texas red–conjugated anti-mouse antibody. D, percentage of cells with three or more nuclei in cultures treated with 10 μmol/L paclitaxel or expressing the products after transfection. Nuclei were visualized by 4′,6-diamidino-2-phenylindole stain for assessment of multinucleation and one hundred 1,250 μm² images containing about 10 cells were randomly captured and the percentage of cells with three or more nuclei was counted directly. In transfected cultures only cells exhibiting the reporter for expression products were scored. Data from three independent experiments in which a total of about 1,000 cells within 100 images were counted (columns, mean; bars, ± SD). Example of GFP-RASSF1A transfected cell (b) with multiple nuclei as stained with 4′,6-diamidino-2-phenylindole (DAPI; c) is also shown.

We first captured RASSF1C by interaction in the yeast two-hybrid system with the 393-amino-acid-residue C terminus of the microtubule-associated protein 1B homologue C19ORF5 (15) and the interaction of full-length C19ORF5 was subsequently confirmed to occur with RASSF1A in vitro and in vivo (10). Both steady-state and overexpressed recombinant C19ORF5 are normally dispersed in the cytosol and do not associate with steady-state arrays of microtubules. However, C19ORF5 is recruited specifically to paclitaxel- or RASSF1A-stabilized microtubules (10). ¹ In addition to its specific recruitment to only hyperstabilized microtubules as C19ORF5 accumulates in transfected mammalian cells, it becomes cytotoxic causing perinuclear aggregation of mitochondria that surround and invade the nucleus coincident with genome destruction (7). ¹ These dual functions that reside in distinct unrelated sequence domains within the C terminus of C19ORF5 suggest that C19ORF5 may be a link between hyperstabilization of microtubules and cytotoxicity. However, the function of the interaction between C19ORF5 and isoforms of RASSF1 that seem to occur equally with A and C isoforms in vitro remains unclear.

Here we report results of a comparative study of RASSF1A and RASSF1C isoforms and how they interact with C19ORF5 when coexpressed together in the same cell. The results show that when expressed alone, A and C isoforms exhibit identical cellular locations, paclitaxel-like hyperstabilization of microtubules, and paclitaxel-like interference with mitosis. However, when coexpressed with C19ORF5, RASSF1C fails to associate with and promote hyperstabilization of microtubules and both RASSF1C and C19ORF5 remain dispersed in the cytosol. Specifically RASSF1A causes hyperstabilization of microtubules and the accumulation of C19ORF5 on them. The difference is due to the short unique N-terminal sequence of RASSF1C rather than the unique features of RASSF1A.

Materials and Methods

Plasmid Construction. HA-RASSF1A in pcDNA3.1 was a gift from Dr. John Minna and was subcloned into pBluescript SK after BamHI-XbaI treatment. GFP-RASSF1A with an enhanced green fluorescent protein (EGFP) fused at the N terminus of RASSF1A protein was created by blunt-end ligation of a 1.2-kb EcoRI-EcoRV fragment of RASSF1A-SK into the pCMV-C3 vector (Clontech Laboratories, Inc., Palo Alto, CA) cut with SmaI. RASSF1C-pACT2 is a plasmid containing the full-length coding sequence of RASSF1C captured by the C19ORF5-pGBKT7 in a yeast two-hybrid screen of a human liver cDNA library (7). RASSF1A-pACT2 was generated by ligating a 900-bp EcoRI-SphI fragment from GFP-RASSF1A and a 1,000-bp SphI-XhoI fragment from RASSF1C-pACT2 into the pACT2 vector after cutting with EcoRI-XhoI. GFP-RASSF1C was the ligation product of EcoRI-SmaI–digested pEGFP-C3 vector with a 1.3-kb EcoRI-EcoRV fragment carrying the full-length RASSF1C cDNA from RASSF1C-pACT2. HA-RASSF1C and CMYC-RASSF1C with HA or c-Myc tag, respectively, fused to the N terminus of RASSF1C, were generated by ligating a 1.6-kb EcoRI-BglII RASSF1C fragment from RASSF1C-pACT2 with EcoRI-BglII–cut pCMV-HA or pCMV-MYC vectors (Clontech Laboratories), respectively. Coding sequence for RASSF1C in pcDNA3.1 was created by ligating a 2.2-kb XbaI fragment of HA-RASSF1C with XbaI-cut pcDNA3.1/zeo vector (Invitrogen Life Technologies, Carlsbad, CA). GFP-RASSF1A120-340 and other constructs were generated by ligating the HindIII-KpnI–cut pEGFP-C3 vector with HindIII-KpnI–cut PCR fragments using template RASSF1C-pACT2 and primers as shown in Fig. 1A. GFP-RASSF1A120-287 was created by ligation of a 500-bp fragment from GFP-RA120-340 that was initially digested with HinfI, end-filled, and recut with HindIII into HindIII-SmaI–treated pCMV-C3. GFP-RASSF1A120-186 was created in the same way with a 200-bp fragment cut with MspI, end-filled, and recut with HindIII. GFP-RASSF1A120-167 was made by ligation of a 150-bp HindIII-PvuII fragment from GFP-RASSF1A120-340 into pCMV-C3 cut with HindIII-SmaI. HA-RASSF1A52-340, HA-RASSF1A102-340, and HA-RASSF1A120-340 were the ligation products of the EcoRI-KpnI–cut pCMV-HA vector and EcoRI-KpnI–cut PCR fragments generated using GFP-RASSF1A as template and primers as shown in Fig. 1A. GFP-C19ORF5 has been described previously (7). HA-C19ORF5 was created by ligation of a 1.5-kb EcoRI-BglII fragment of C19ORF5-pACT2 (15) with the EcoRI- and BglII-treated pCMV-HA vector. All constructs were authenticated by DNA sequencing.

Cell Treatments, Immunohistochemistry, and Fluorescent Microscopy. Except where indicated in the text, materials were obtained, and transfections and expressions of reporter tagged fusion proteins, microscopy, and immunofluorescence were done as described (7). ¹ Mitochondria were tracked with either extranuclear 4′,6-diamidino-2-phenylindole DNA stain or MitoTracker Red CM-H₂XRos (Molecular Probes, Inc., Eugene, OR; ref. 7). Analysis of nocodazole and cold resistance followed the procedure reported (16, 17) and described in detail in the corresponding figure legend. Cotransfectants of GFP-tagged C19ORF5 and HA-tagged RASSF1A or RASSF1C, or HA-tagged C19ORF5 and GFP-tagged RASSF1A or RASSF1C, were identified by direct GFP fluorescence and indirect double antibody immunofluorescence with anti-HA and Texas red–conjugated anti-mouse antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Polyclonal antibody prepared against a bacterially expressed recombinant fragment, comprised of residues 120 to 340 of RASSF1, was a kind gift from Dr. Michael A. White (12). Specificity for RASSF1 expression products and experimental titers were determined by immunoblot of lysates of cells transfected with GFP fusions with anti-GFP (Santa Cruz Biotechnology) and anti-RASSF1 antibody simultaneously.

Subcellular Fractionation. The heavy membrane pellet, light membrane pellet, and soluble fraction were prepared from Hep3B cells resuspended in isotonic HIM buffer as described (18). The heavy membrane and light membrane fractions were resuspended in HIM buffer equal to the volume of the soluble fraction. Fractions of 25 μL corresponding to 50,000 cells were loaded side by side on SDS-PAGE in triplicate gels and each replicate was subjected to immunoblot. Bands were visualized by reaction with antibody against RASSF1 described above, mitochondria-associated LRPPRC, or β-tubulin as described previously (7).

Ultracentrifugation of Paclitaxel-Stabilized Microtubules. COS7 cells transfected with GFP fusion constructs were lysed and broken, and the microtubule pellet was prepared from cleared lysates as described (19, 20) .

Protein-Protein Interaction Assays. The interaction between RASSF1 and C19ORF5 was first tested using the yeast two-hybrid system as described (7, 15) . For coimmunoprecipitation, transfected cells were solubilized in lysis buffer [50 mmol/L HEPES (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 2.5 mmol/L EGTA, 0.1% Triton X-100, 10% glycerol, 1 mmol/L NaF, 1 mmol/L phenylmethylsulfonyl fluoride, 2 μg/mL pepstatin, and 2 μg/mL aprotinin]. After clarification at 10,000 × g for 10 minutes, monoclonal anti-HA antibody and protein G beads (Amersham Biosciences, Stockholm, Sweden) were added to supernatant for 2 hours. Immunoprecipitates were collected by centrifugation, washed thrice, and resuspended in lysis buffer. The total cell lysate and the HA immunoprecipitates were analyzed in duplicate by SDS-PAGE and immunoblotted with anti-HA or anti-GFP antibody.

Results

RASSF1A and RASSF1C, but Not RASSF1B, Alone Similarly Cause a Paclitaxel-like Hyperstabilization of Microtubules and Disruption of Cell Cycling. To test for unique properties of RASSF1A, we compared the localization and effect of RASSF1A, B, and C isoforms on microtubule dynamics and cell cycling in mammalian cells expressing GFP-labeled products. Both green fluorescent GFP-RASSF1A and RASSF1C were strictly extranuclear and both colocalized (yellow) with pronounced bundles of stabilized microtubules (red; Fig. 1Ba,b). In cells treated with the microtubule-stabilizing drug paclitaxel before RASSF1 expression, both isoforms contributed to an even more intense bundling of hyperstabilized microtubules with which they were associated ( Fig. 1Bf, g). When microtubules were depolymerized before expression by treatment of cells with the microtubule destabilizing drug nocodazole, both RASSF1A and RASSF1C were associated with punctate foci, most of which still overlapped with tubulin ( Fig. 1 Bk, l). In marked contrast to the A and C isoforms, RASSF1B, a minor isoform of RASSF1, failed to colocalize with tubulin, but seemed solely in distinct punctate extranuclear reticular structures ( Fig. 1Bc, h, m). Notably normal microtubule arrays seemed diminished in cells transfected with RASSF1B, indicating that RASSF1B may actually exert a destabilizing rather than a stabilizing effect on microtubules observed with RASSF1A and RASSF1C ( Fig. 1Bc). The cellular distribution of all three GFP-tagged RASSF1 isoforms was distinct from that of GFP alone ( Fig. 1Bd, i,n), indicating that the distribution was indicative of the RASSF1 portion rather than the GFP tag. Distribution of the overexpressed RASSF1 isoforms was identical among different cell types independent of whether GFP, His, or c-Myc was employed as the fusion tag ( Fig. 1C). Overexpression of both RASSF1A and RASSF1C contributed equally to the hyperstabilization of preassembled microtubules in a manner similar to paclitaxel ( Fig. 1Be, j). Overexpression of both RASSF1A and RASSF1C alone, but not RASSF1B, exerted a paclitaxel-like disruption of mitosis resulting in up to 20% of cells with more than three nuclei ( Fig. 1Da) and in extreme cases up to 100 mini-nuclei ( Fig. 1Db,c). These results differ significantly from those reported by Song et al. (11) which showed that RASSF1C exhibited no microtubule association but was strictly nuclear in a pattern similar to GFP ( Fig. 1Bd). Our results are also in contrast with those reported by Liu et al. (8) which showed that the cellular distribution of RASSF1B was similar to control GFP.

To analyze whether the mechanism and extent of microtubule hyperstabilization induced by RASSF1A and RASSF1C were similar, we compared the two isoforms for resistance and reversibility of the hyperstabilized microtubule bundles in the presence of the common destabilizing drug nocodazole as well as cold treatment. Bundles of microtubules that resulted from overexpression of either RASSF1A or RASSF1C alone were resistant to breakdown by both nocodazole and cold treatment given subsequent to expression ( Fig. 2A and B ). Moreover, microtubule arrays failed to recover in the presence of either RASSF1A or RASSF1C 2 hours after removal of nocodazole that was added before transfection ( Fig. 2C). This indicates that both A and C isoforms at sufficiently high concentration are equally effective and both interfere with microtubule dynamics by prevention of breakdown of preformed microtubules as well as polymerization of new ones. Our results differ from those of Vos et al. (9) who reported that RASSF1C was less effective than RASSF1A in rendering microtubules resistant to breakdown by nocodazole.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

RASSF1A and RASSF1C similarly interfere with microtubule dynamics by prevention of both depolymerization and polymerization. A-B, at 24 hours after transfection, cells were treated with 10 μmol/L nocodazole (A) or washed and fixed at 0°C (B) for 2 hours. Untransfected cultures (a and b) were fixed at 37°C and stained with anti-β-tubulin (white). GFP fluorescence (white) was examined 24 hours after transfection with GFP-RASSF1A (c) or GFP-RASSF1C (e) and again in the same cells (d and f) 2 hours after the indicated treatment. C, nocodazole was washed out from cells in which microtubules were depolymerized before GFP-tagged proteins were expressed by replacement with culture media containing no nocodazole 24 hours after transfection. Microtubules were examined for polymerization 2 hours later. The indicated cells are representative of more than 95% of 2,000 cells examined that showed a positive GFP signal in three independent transfection experiments.

Both RASSF1A and RASSF1C Are Normally Associated with Mitochondria. Notably overexpressed RASSF1A or RASSF1C that is concentrated on hyperstabilized microtubule bundles disperses into scattered but distinct punctate foci in the absence of intact microtubular arrays ( Figs. 1Bk, l and 2C). To determine the normal cellular location of steady-state levels of RASSF1, we examined the distribution in several cell types with an antibody (12) that reacts equally with RASSF1A, B, and C isoforms (data not shown). No distinct association with the microtubular cytoskeleton was apparent ( Fig. 3A ). Instead, RASSF1 was largely associated in distinct punctate foci similar to overexpressed RASSF1B ( Fig. 1Bc) and RASSF1A or RASSF1C in cells lacking intact microtubules because of treatment with nocodazole ( Figs. 1Bk, l and 2Cc-f). Cytosolic punctate foci are indicative of mitochondria and this was further suggested by the overlap of RASSF1 antigen with extranuclear DNA ( Fig. 3A). The mitochondrial association of steady-state levels of cellular RASSF1 in Hep3B cells was confirmed by direct subcellular fractionation and analysis. A 32-kDa band of RASSF1C was apparent in fractions enriched in mitochondria whereas none was detectable in the soluble fraction ( Fig. 3B). Consistent with its suppression in tumors and most long-term cultured cell lines with high proliferative potential, ² RASSF1C (32 kDa) was predominant whereas RASSF1A (39 kDa) was very low or absent. When Hep3B cells were cultured in the presence of DNA methylation inhibitor 5-aza-2′-deoxycytidine (5-Aza-CdR) for 5 days, reexpression of RASSF1A did not change the punctate distribution of RASSF1 proteins ( Fig. 3C). This suggested that both RASSF1A and RASSF1C exhibit a similar distribution. These results indicate that most, if not all, of steady-state RASSF1 is associated with mitochondria in interphase cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Association of RASSF1A, B, and C with mitochondria. A, subcellular distribution of endogenous RASSF1. The indicated cell types were exposed to polyclonal RASSF1 antibody and RASSF1 antigens were visualized with fluorescein isothiocyanate–labeled second antibody. White, positive green fluorescence. The indicated cells are representative of majority of 10,000 cells in three independent staining experiments. Yellow, colocalization of endogenous RASSF1 (red) with extranuclear mitochondrial DNA (green) in Hep3B cells. B, association of RASSF1 with the mitochondrial fraction of Hep3B cells. Cells were fractionated by sequential centrifugation (Materials and Methods). Heavy membrane (HM) is the fraction enriched in mitochondria, light membrane (LM) is the post-mitochondrial pellet, and soluble fraction (S) the supernatant. The predicted molecular masses of intact RASSF1A, C, and B are 39, 32, and 22 kDa, respectively. LRPPRC is a mitochondrial-associated protein (7, 31) . C, similar association of endogenous RASSF1 antigens with mitochondria in Hep3B cells in the presence and absence of DNA methylation inhibitor (5-Aza-CdR). The mitochondria structure labeled with RASSF1 antibody is identical to the structure that was routinely labeled with MitoTracker. D, association of overexpressed GFP-RASSF1A and GFP-RASSF1C with mitochondria in absence of intact microtubules. Cells overexpressing GFP-RASSF1A, B, and C (green) were stained with MitoTracker or 4′,6-diamidino-2-phenylindole (DAPI, red) to visualize mitochondria. Where indicated, cells expressing RASSF1A or RASSF1C were treated with 10 μmol/L nocodazole just before transfection. Treatment with nocodazole had no effect on distribution of RASSF1B as shown in Fig. 1Bm. Merge yellow, colocalization of RASSF1 isoforms with mitochondria.

Next we tested for a mitochondrial association of RASSF1 isoforms in cells overexpressing RASSF1A, B, and C. In cells exhibiting the extensive RASSF1A- or RASSF1C-induced microtubular bundling, punctate MitoTracker-labeled mitochondria (red) in their normal cytosolic reticular networks were distinct among the microtubular arrays decorated by RASSF1A or RASSF1C (green) without indication of overlap (absence of merge yellow; Fig. 3D). However, in cells treated with nocodazole before transfection that were devoid of intact microtubules, overexpressed levels of both RASSF1A and RASSF1C were associated with mitochondria (yellow, Fig. 3D). The overlap analysis also suggested that RASSF1B that failed to associate with microtubules in untreated cells resided with mitochondria ( Fig. 3D). The mitochondrial association in absence of microtubule stabilization and bundling was further confirmed by overlap of RASSF1A and RASSF1C with distinct foci of extranuclear mitochondrial DNA in nocodazole-treated cells ( Fig. 3D). These results taken together indicate that RASSF1A, B, and C are normally mitochondrial-associated proteins, but only RASSF1A or RASSF1C traffics to and contributes to hyperstabilization of microtubules. In absence of intact microtubules, both RASSF1A and RASSF1C revert to the mitochondrial location even at elevated levels.

Only Part of the ATM Domain Common to RASSF1A and RASSF1C Is Required for Microtubule Hyperstabilization. To determine the structural determinants for the association and hyperstabilization of microtubules, a series of site-directed mutations summarized in Figs. 1A and 4B were constructed and the products of constructs were examined for cellular location and induction of microtubule hyperstabilization. As predicted by the identical effects of RASSF1A and RASSF1C when expressed alone, the stable association with and induction of hyperstabilization of microtubules were independent of either of the unique N termini that distinguished RASSF1A and RASSF1C but required all or part of the common ATM domain that was missing in RASSF1B ( Fig. 4A). The analysis indicated that the signature RA domain of RASSF1 gene products was also required for the hyperstabilization. A more detailed analysis of the requirement for the ATM sequence domain revealed that the N-terminal 8 residues, including the consensus serine phosphorylation site within the 14 consensus residues for ATM serine phosphorylation (underlined W125-K138) that defines the ATM domain ( Fig. 4B; ref. 21), were not required for the microtubule hyperstabilization activity ( Fig. 4Ca, b, c). Moreover, specific point mutations, including a serine phosphorylation site (S131) that occurs in nature in the RASSF1 ATM consensus sequence (1), did not diminish the effect ( Fig. 4Cd-g). In summary, these results indicate that the C-terminal 6 residues (AEIEQK) of the W125-K138 ATM consensus sequence downstream of serine-131 and presumably the rest of the putative ATM domain upstream of the N terminus of RASSF1B confer the stable association with and hyperstabilization of microtubules by RASSF1A and RASSF1C. The deletion mutant 1A139-340 beginning at isoleucine 139 exhibited the punctate distribution indicative of association with mitochondria ( Fig. 4Cc) similar to RASSF1B.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Sequence requirements for microtubule association and hyperstabilization by RASSF1A and RASSF1C isoforms. A, requirement of the RA domain but not N-terminal sequences upstream of the ATM domain for microtubule association and stabilization. The constructs indicated at top were tagged at the N terminus with GFP and spanned the residues indicated according to RASSF1A numbering ( Fig. 1A). B, sequence of the ATM domain containing a consensus ATM phosphorylation site (S131). Serine 131 within a peptide comprised of the underlined sequence is phosphorylated by the ATM kinase (21). The four indicated naturally occurring point mutants in the sequence (1) were constructed and transfected into COS7 cells. Numbering is according to RASSF1A ( Fig. 1A). C, the influence of artificial and naturally occurring mutants on microtubule association and stabilization. The N terminus of the consensus ATM hosphorylation site including serine 131 is not required for microtubule association and stabilization (a-c). Mutants beginning at residues 125, 133, and 139 of RASSF1A are indicated. The naturally occurring point mutants in the ATM consensus sequence do not affect microtubule association and stabilization (d-g). COS7 cells were transiently transfected with the indicated constructs ( Fig. 1A) tagged with GFP at the N terminus and examined for location of GFP signal and hyperstabilization of microtubules. The presence of predicted full-length expression products was confirmed by separate extraction and immunoblot of transfected cell extracts with anti-GFP or RASSF1 antibody. D, analysis of microtubule association GFP-tagged constructs in vitro. Microtubules from about 10⁶ transfected cell extracts (150 μL) were stabilized with 10 μmol/L paclitaxel, collected by ultracentrifugation, and proteins in the pellet (P) and supernatant (S) were analyzed by immunoblot with antibody for β-tubulin (Tub) or GFP as indicated. Bottom, densitometric estimation of the ratio of antigen in the pellet to that in the pellet and supernatant. The tubulin band was at apparent molecular mass of 55 kDa and the recovered GFP-tagged products indicated from left to right migrated at the expected full-length molecular masses of 64, 58, 52, 45, 48, 34, 31, and 26 kDa, respectively.

Next we compared the association of GFP-tagged RASSF1A, B, C or mutant constructs with paclitaxel-stabilized microtubules recovered from cell extracts free of mitochondria and other cellular organelles ( Fig. 4D). As expected RASSF1A, RASSF1C, and constructs 1A120-340 and 1A120-287 sedimented with the microtubules. Surprisingly, RASSF1B, of which association with microtubules at overexpressed levels in intact cells was below detection thresholds, exhibited an association with microtubules in vitro in absence of mitochondria. Moreover, construct 1A120-186 devoid of the RA domain and comprised of only the ATM and sequence residues comprising the spacer sequence between the ATM and RA domains still associated with microtubules in vitro. Microtubule association in vitro was lost by further truncation of the inter-ATM/RA domain sequence in construct 1A120-167. We interpret these results to indicate that a simple RASSF1 microtubule-association domain resides within residues 167 to 186. However, this domain is insufficient to induce microtubule hyperstabilization that requires both the C terminus of the ATM domain and the RA domain.

Specific Colocalization of RASSF1A and C19ORF5 on Hyperstabilized Microtubules. Unlike RASSF1 isoforms that are mitochondrial-associated and at elevated levels cause microtubule stabilization, the microtubule-associated protein homologue C19ORF5 is distributed in the cytosol. Neither steady-state nor overexpressed levels of C19ORF5 alone exhibit a distinct association with normal cellular arrays of microtubules or contribute to the hyperstabilization of microtubules. However, C19ORF5 seems specific to microtubules hyperstabilized by paclitaxel. ¹ Owing to both RASSF1A and RASSF1C exhibiting a paclitaxel-like hyperstabilization of microtubules when overexpressed alone, we examined the fate and colocalization of C19ORF5 in cells expressing C19ORF5 and RASSF1A or RASSF1C concurrently. The C terminus of C19ORF5 was tagged with GFP and transfected together with either HA-tagged RASSF1A or RASSF1C. In agreement with results using full-length C19ORF5 (10), instead of its initial dispersion in the cytosol in cells expressing only GFP-C19ORF5 ( Fig. 5Aa ), GFP-C19ORF5 was associated with hyperstabilized microtubules resulting from cotransfection with HA-RASSF1A in about 60% of total GFP-positive cells ( Fig. 5Ab,e). Analysis with anti-HA antibody revealed that RASSF1A was also associated with the microtubules in 100% of the cells in which GFP-C19ORF5 decorated bundles of stabilized microtubules ( Fig. 5B). In striking contrast to RASSF1A, when cultures were cotransfected with GFP-C19ORF5 and HA-RASSF1C ( Fig. 5Ac and B), C19ORF5 remained dispersed in the cytosol with no apparent association with microtubules, a pattern similar to control cells transfected with only GFP-C19ORF5 ( Fig. 5Aa) or cotransfected with an empty HA vector ( Fig. 5Ad and B). We failed to detect a single cell with microtubule-associated GFP-C19ORF5 in 10 independent cotransfection experiments with HA-RASSF1C. This indicated that coexpression with C19ORF5 prevented the association with and contribution to the hyperstabilization of microtubules specifically by RASSF1C.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Specific association of complexes of RASSF1A and C19ORF5 on hyperstabilized microtubules. A, hyperstabilization and association of GFP-C19ORF5 with microtubules occur in cells coexpressing RASSF1A, but not RASSF1C. The GFP fusion of the 393-amino-acid-residue light chain of C19ORF5 (GFP-C19ORF5C) was cotransfected into COS7 cells with the indicated constructs tagged with HA. White, distribution of GFP-C19ORF5. Data from three independent transfection experiments of the percentage of the total GFP-labeled cells in which GFP-C19ORF5 decorated hyperstabilized microtubules were estimated by direct count (columns, mean; bars, ± SD). About 700 total cells were observed per experiment. B, only RASSF1A and C19ORF5 colocalize on hyperstabilized arrays of microtubules. COS7 cells were cotransfected with GFP-C19ORF5 (green) and HA-RASSF1A, HA-RASSF1C, or empty HA vector (pcDNA3.1/zeo) as indicated. HA-tagged RASSF1 products (b) were visualized with anti-HA antibody and subsequent reaction with Texas red–conjugated anti-mouse antibody. Yellow, overlap. Positive cells expressing RASSF1A or RASSF1C increased four to five times in intensity over background (b). Numbers were the ratio of maximum fluorescent intensity of positive cells to a background area in the same field containing no cells (b). C, C19ORF5 prevents association and hyperstabilization of microtubules by RASSF1C. GFP-RASSF1A or GFP-RASSF1C was cotransfected with HA-C19ORF5 or empty HA vector (pCMV-HA) as indicated. Cells were scored positive or negative for HA and quantified as follows: untransfected cells, no GFP signal and background anti-HA stain; transfected with HA-C19ORF5 only, no GFP signal and strongly HA positive (see B for example of a positive); transfected with GFP-RASSF1A or GFP-RASSF1C only, GFP signal and background HA stain; coexpressing cells, GFP signal and strong HA stain. Only the latter cells that were clearly coexpressing both products were examined and used in quantification of impact of coexpression of RASSF1A/C and C19ORF5 on cellular location of each. On average 50% of total cells were positive for GFP when transiently transfected with a GFP-tagged construct. Because of background fluorescence from the anti-HA or anti-RASSF1 antibody, about 30% of total cells were scored as positive when transfected with HA-tagged constructs alone. Doubly labeled cell yields were about 10% of the total population. Data from three independent transfection experiments in which about 100 cells were examined (e): columns, mean; bars, ± SD.

The reciprocal analysis in which cells were cotransfected with GFP-RASSF1A and HA-tagged C19ORF5 confirmed that GFP-RASSF1A was associated with bundles of stabilized microtubules in 99% of cells expressing GFP independent of whether they were coexpressing HA-C19ORF5 ( Fig. 5Ca, c). In marked contrast to cultures transfected with RASSF1A or RASSF1C alone, the GFP signal was dispersed in the cytosol in about 20% of total cells expressing GFP-RASSF1C ( Fig. 5Cb). Moreover, this fraction of cells exhibited no distinct fibrillar arrays of hyperstabilized microtubules observed in cells expressing RASSF1C alone. Subsequent analysis with anti-HA antibody revealed that 100% of this fraction of cells stained positively for HA-C19ORF5 and the signal was also dispersed in the cytosol with no particular association with microtubular arrays ( Fig. 5B). A count of cells coexpressing both GFP-RASSF1A or GFP-RASSF1C and HA-C19ORF5 revealed that 100% of doubly labeled cells expressing RASSF1A exhibited hyperstabilized microtubules and RASSF1A was associated with them whereas in over 95% of doubly labeled cells RASSF1C was dispersed in the cytosol without evidence of microtubule hyperstabilization ( Fig. 5Ce). The fact that an N-terminal tag as long as the 26 kDa of GFP or as short as the 9 amino acid residues of the HA tag on either C19ORF5 or RASSF1 isoforms yielded similar results strongly suggests that the differences are encoded in RASSF1A and RASSF1C sequences and are unaffected by the tag sequence. Notably in the presence of overexpressed levels of C19ORF5, RASSF1C did not return to its default punctate pattern indicative of its normal mitochondrial association in the absence of association with microtubules. This may indicate that complexes of C19ORF5 and RASSF1C do not associate with mitochondria.

Lastly, we confirmed that both RASSF1A and RASSF1C isoforms similarly formed stable complexes with C19ORF5 although the three exhibited different cellular locations and impact on microtubules. The RASSF1C-C19ORF5 interaction was initially suggested by the yeast two-hybrid interaction analysis in which RASSF1C was trapped with the 393-amino-acid-residue C terminus of C19ORF5 (7) that exhibits the multiple activities of C19ORF5 described to date (7). An interaction between RASSF1A and full-length C19ORF5 in mammalian cells was subsequently confirmed (10). Fig. 6A indicates that RASSF1A and RASSF1C similarly interact with the active C terminus of C19ORF5 but not with another mitochondrial-associated interaction partner of C19ORF5 (LRPPRC) in the yeast two-hybrid system. Coprecipitation of both GFP-RASSF1A and GFP-RASSF1C with HA-C19ORF5 using anti-HA antibody indicated that both A and C isoforms were in stable complexes with C19ORF5 in lysates of cotransfected cells ( Fig. 6B). Together these results indicated that although both mitochondrial RASSF1A and RASSF1C alone at elevated levels similarly associate with and promote hyperstabilization of microtubules, the interaction with C19ORF5 differentially determines whether RASSF1C associates with microtubules and contributes to their hyperstabilization. This in turn determines the association and accumulation of C19ORF5 on hyperstabilized microtubules.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

The unique N terminus of RASSF1C prevents the association of RASSF1C-C19ORF5 complexes with microtubules. A, similar interaction with C19ORF5 between RASSF1A and RASSF1C in the yeast two-hybrid trap. cDNAs coding for the indicated factors were fused with the DNA binding or activation domains in vectors (pGBKT7 and pACT2, respectively) in the yeast two-hybrid complementation system as previously described (7, 15) . LRPPRC is also an interaction partner with C19ORF5 (15). B, recovery of complexes of intact RASSF1A and RASSF1C with C19ORF5 from cotransfected cells. HA indicates the 42-kDa HA-tagged 393-amino-acid-residue C19ORF5 whereas GFP indicates 65, 58, or 26 kDa of GFP-RASSF1A, GFP-RASSF1C, or GFP, respectively. About 2 × 10⁶ COS7 cells were transfected with 15 μg of HA-C19ORF5 and cotransfected with 15 μg of GFP-RASSF1A, GFP-RASSF1C, or GFP as indicated and solubilized in 400 μL of lysis buffer. Lysates from about 125,000 cells (25 μL) were mixed with immobilized protein G-anti-HA antibody (about 6.3 μL of packed beads and 0.25 μg of antibody) and the immunoprecipitated material bearing HA or GFP (HAIP) was analyzed with 1 μg/mL of anti-HA or anti-GFP antibody and compared with total lysate. The HA- or GFP-containing bands were visualized with alkaline phosphatase-conjugated anti-mouse or anti-rabbit immunoglobulin G antibody, respectively. C, the unique N terminus of RASSF1C prevents RASSF1C-mediated microtubule hyperstabilization and accumulation of C19ORF5 on microtubules. GFP-C19ORF5 (white) was cotransfected with the indicated HA-tagged RASSF1A deletion constructs beginning at residues 52, 102, and 120 (see Fig. 1A). The indicated cells represented about 50% of cells exhibiting a positive GFP signal among a total of 2,000 cells examined from three independent cotransfection experiments.

The Unique N terminus of RASSF1C Indirectly Determines the Specificity of RASSF1A in Hyperstabilization of Microtubules. RASSF1A and RASSF1C are distinguished by unique N termini of 119 and 49 residues, respectively ( Fig. 1A). To determine whether the unique N terminus of RASSF1A or that of RASSF1C was the determinant for selective association of RASSF1A and C19ORF5 on hyperstabilized microtubular arrays, we examined sequential deletions of the N terminus of RASSF1A down to the ATM domain (RASSF1A120-340) for induction of hyperstabilized microtubule-associated GFP-C19ORF5 in cotransfected cultures ( Fig. 6C). Surprisingly, the results indicate that it is the unique 49-residue N terminus of RASSF1C upstream of the ATM domain that is responsible for the lack of association of RASSF1C with microtubules in the presence of C19ORF5. This unique structural feature of RASSF1C indirectly underpins the selective ability of RASSF1A to promote accumulation of C19ORF5 on hyperstabilized microtubules.

Discussion

An increasing number of reports show that epigenetic inactivation specifically of the A isoform of RASSF1 through promoter-specific hypermethylation is a common molecular change associated with human cancers (1–3) . Based on this, it is a strong candidate for a tumor suppressor. RASSF1A has been observed in association with microtubules, causes a paclitaxel-like hyperstabilization of microtubules at sufficient levels (8–11, 14) , and induces numerous cellular and biochemical phenomena that are a consequence of microtubule hyperstabilization (11–14) . How RASSF1A might specifically suppress tumors compared with its closely related isoform RASSF1C of which promoter is not suppressed by hypermethylation in tumors (1) is unknown. In this study, we have identified a mechanism that may underlie the specific association of RASSF1A with microtubules, maintenance of microtubule hyperstabilization, and potentially how the hyperstabilization may be linked to cytotoxicity resulting in tumor suppression. Our results show that when examined in isolation, RASSF1A and RASSF1C are identical in respect to cellular location, mechanism of association with and contribution to hyperstabilization of microtubules, and impact of hyperstabilization on cell cycling. These results differ from reports that contend that RASSF1C is nuclear-associated and does not associate with microtubules (11) or that RASSF1C is less effective in microtubule stabilization than RASSF1A (9). However, our results are in agreement with reports showing that, similar to RASSF1A, RASSF1C in isolation inhibits tumor cell growth in vitro and in reconstituted tumors in animals (6, 13) .

We show that A, B, and C products of the RASSF1 gene are in large part normal residents of mitochondria that are capable of association with microtubules. We identified a microtubule association domain as a relatively basic sequence that lies between the ATM and RA domains that is common to the A, B, and C isoforms. Owing to microtubule association as a prerequisite for microtubule stabilization, these results are in agreement with others who reported that sequences containing the region are required for microtubule hyperstabilization (8, 9, 14) . The simple microtubule association domain was necessary but insufficient to induce microtubule stabilization. Microtubule hyperstabilization required a part of the ATM domain and the RA domain flanking the basic microtubule association sequence. The phosphorylation site S131 that defines the ATM domain and is a candidate for ATM phosphorylation (21) failed to affect microtubule hyperstabilization. This indicated that the role of the sequence domain in hyperstabilization is structural rather than regulated by ATM phosphorylation. Our experiments that show microtubule hyperstabilization, but not microtubule association, requires the RA domain explains the “dominant negative” effect in respect to induction of cell death of the mutant construct devoid of the RA domain but containing the ATM and adjacent basic sequence that mediates microtubule association (9).

The results here and those of others show that RASSF1A and RASSF1C alone exhibit a default mitochondrial location and can both associate with and cause hyperstabilization of microtubules through separate, but shared, sequence domains to cause cellular consequences of paclitaxel-like microtubule hyperstabilization. What is the difference that specifically distinguishes hypermethylation-suppressed RASSF1A in the process and could qualify it specifically as a tumor suppressor relative to RASSF1C? Here we show that the 393-amino-acid-residue C terminus of microtubule-associated protein 1A/B homologue C19ORF5 that complexes with both RASSF1A and RASSF1C prevents the association and hyperstabilization of microtubules specifically by RASSF1C. Unexpectedly, it is the unique C terminus of RASSF1C that prevents the association of complexes of C19ORF5 and RASSF1C with microtubules and thus indirectly limits the specific microtubule association and hyperstabilization effects to RASSF1A.

C19ORF5 is widely expressed in tissues and normally dispersed in the cytosol unassociated with microtubules (7, 10, 15) . C19ORF5 is recruited specifically to microtubules stabilized by either paclitaxel ¹ or, as we specifically show here, RASSF1A. Similar to RASSF1A, B, and C, the C terminus of C19ORF5 exhibits a distinct and apparently independent basic amino acid-rich microtubule-binding domain. ¹ C19ORF5 differs from RASSF1A and RASSF1C in that in isolation it fails to associate with and cause hyperstabilization of microtubules even when expressed at high levels (7). Like C19ORF5, both RASSF1A and RASSF1C are also widely expressed in normal tissues (2) and, as we show here, largely reside on mitochondria without significant association with steady-state microtubular arrays in interphase cells. In mitotic cells, RASSF1A has been reported to associate with microtubule spindles, but no clear alignment with spindle microtubular arrays was shown (8, 11) . Whether the association reflects spindle-associated mitochondria or other non-microtubular spindle elements remains to be established. The mechanism and order in which RASSF1A and C19ORF5 converge and accumulate on microtubules to cause hyperstabilization is a subject requiring detailed future study. Despite the presence of distinct and independent microtubule-binding domains evident in vitro, RASSF1A, B, and C, except at high levels, are mostly associated with mitochondria in intact cells. C19ORF5 at all levels is dispersed in the cytosol. The independent association of RASSF1 isoforms and C19ORF5 with microtubules in steady-state dynamics is sufficiently transient to escape detection. Thus, a role of RASSF1A, B, and C, or C19ORF5 in steady-state microtubule dynamics (polymerization/depolymerization) remains to be established. However, when taken together, our data suggest that a cooperative interaction between cytosolic C19ORF5 and mitochondrial RASSF1A may result in the extended residence of both factors on microtubules, which cascades into a paclitaxel-like hyperstabilization and its cellular and biochemical consequences. In view of its independent but specific affinity for stabilized microtubules in absence of RASSF1A, C19ORF5 may be the key catalyst at the transiently stable ends of microtubules during steady-state dynamics. C19ORF5 may trap specifically RASSF1A of which prolonged residence on the microtubules inhibits depolymerization, attracting more C19ORF5, and thus more RASSF1A, resulting in the observed hyperstabilization. Unlike paclitaxel that acts only indirectly, RASSF1A acts both indirectly and directly to cause accumulation of C19ORF5 on hyperstabilized microtubules.

In summary, we have shown how C19ORF5 confers the specific ability of RASSF1A in the presence of other RASSF1 isoforms to cause hyperstabilization of microtubules. This function is lost in tumors by suppression specifically of RASSF1A expression by promoter-specific hypermethylation. Currently, the paclitaxel-like interference in microtubule dynamics is a candidate pathway for the tumor suppressor activity of RASSF1A. Most attention on the increasing number of both validated and candidate tumor suppressors, like RASSF1 isoforms that exhibit microtubule association and impact dynamics, has been on the potential role in spindle microtubule quality and prevention of inheritable genomic instability caused by chromosome missegregation at mitosis (22–24) . More work is needed to determine whether RASSF1A is an “endogenous taxoid” of which induction of microtubule hyperstabilization signals cell death and genome destruction by similar mechanisms to taxoid family antitumor drugs (25–29) . It should be noted that, in addition to its key role as a potential catalyst and the determinant of the specificity of RASSF1A in microtubule hyperstabilization, as C19ORF5 accumulates in transfected mammalian cells it becomes cytotoxic by causing perinuclear aggregation of mitochondria that surround and invade the nucleus coincident with genome destruction (7). This process will be documented in detail in a separate report. ¹ A sequence domain as short as 25 amino acids in the 393-residue C terminus of C19ORF5 that is distinct from the microtubule association domain independently mediates the cytotoxic effect of C19ORF5. ¹ The dual functions of C19ORF5 and specifically RASSF1A-mediated accumulation on hyperstabilized microtubules may be a link between microtubule hyperstabilization and cytotoxicity. Reversal of the isoform-specific epigenetic suppression of RASSF1A may be equivalent to restoration of a “natural taxoid” and provide a potential target for prevention of propagation of aneuploidy and genetic instability.