Human G3BP1 interacts with β-F1-ATPase mRNA and inhibits its translation

The bio-energetic phenotype of eukaryotic cells is adjusted upon changes in environmental and/or physiological conditions by regulating the activity of its mitochondria (Rossignol et al., 2004). Changes in the molecular composition of mitochondria require the concerted expression of nuclear and mitochondrial genes (Scarpulla, 2008). Regulation of mitochondrial biogenesis is mainly exerted at the level of transcription in the nucleus and by replication and/or transcription in the mitochondrial genome (Scarpulla, 2008). The localization of mRNAs also exerts a fundamental role in the regulation of gene expression (Martin and Ephrussi, 2009). Mechanisms that control the localization (Lithgow et al., 1997; Marc et al., 2002) and translation (de Heredia et al., 2000; Izquierdo and Cuezva, 1997; Martinez-Diez et al., 2006) of nucleus-encoded mRNAs of mitochondria have been shown to define the bioenergetic phenotype of the cell. The sorting of mRNAs to the vicinity of mitochondria is a conserved feature in both yeast and mammalian cells (Lithgow et al., 1997; Sylvestre et al., 2003b) affecting those mRNAs that encode core components of mitochondrial complexes (Egea et al., 1997; Garcia et al., 2007; Marc et al., 2002). The targeting of mRNAs involves multiple steps and cis-acting elements placed in the 3′-untranslated region (3′UTR) of the mRNAs can confer mitochondrial targeting (Marc et al., 2002; Martin and Ephrussi, 2009).

Genetic screening and affinity purification methods have allowed the identification of several trans-acting factors involved in mRNA transport, anchoring, stability and translation (Martin and Ephrussi, 2009; Rodriguez et al., 2008). The Pumilio (PUF) family of RNA-binding proteins (RNABPs) binds 3′UTRs and modulates mRNA expression in a wide variety of eukaryotic species (Wickens et al., 2002). Studies aimed at identifying Puf-associated mRNAs in yeast, flies and human cells have shown that Puf targets display conserved binding sites and usually fall into the same functionally annotated pathway (Galgano et al., 2008; Gerber et al., 2004; Gerber et al., 2006). In this regard, Puf3p specifically associates with 256 mRNAs in Saccharomyces cerevisiae, 90% of which are nucleus-encoded mitochondrial proteins that are highly enriched in mitochondria-bound polysomes (Gerber et al., 2004; Saint-Georges et al., 2008). Consistently, yeast strains overexpressing Puf3p exhibit respiratory dysfunction, and abnormal mitochondrial morphology and motility (Garcia-Rodriguez et al., 2007; Gerber et al., 2004). A second class of 224 mitochondria-associated transcripts that lack Puf-binding sites, and whose expression and localization is not affected by PUF3 deletion, has been described (Saint-Georges et al., 2008), suggesting the existence of at least two pathways for mRNA sorting to mitochondria.

The subcellular localization and translation of mRNA of the catalytic subunit of the mitochondrial H⁺-ATP synthase subunit β (ATP5B, hereafter referred to β-F1-ATPase) has been studied both in yeast and in rat liver cells (Egea et al., 1997; Izquierdo and Cuezva, 1997; Lithgow et al., 1997; Margeot et al., 2002; Margeot et al., 2005). In yeast, β-F1-ATPase (ATP2) mRNA is preferentially sorted to the vicinity of mitochondria by the 3′UTR (Margeot et al., 2002; Margeot et al., 2005). Deletion of the 3′UTR in the ATP2 gene leads to deficient protein import, reduced ATP synthesis, depletion of mitochondrial DNA and respiratory dysfunction (Margeot et al., 2002; Margeot et al., 2005). Interestingly, ATP2 mRNA was not found as a Puf3p target and belongs to the above mentioned class of Puf3-independent mitochondrium-localized mRNAs for which the trans-acting factors remain to be identified (Gerber et al., 2004; Saint-Georges et al., 2008). In rat hepatocytes, mRNA of β-F1-ATPase (β-F1 mRNA) is present in a large ~150 nm ribonucleoprotein (β-F1–RNP) complex preferentially associated to the outer mitochondrial membrane (Egea et al., 1997; Lithgow et al., 1997) that contains components of the translational machinery (Ricart et al., 1997). The assembly and appropriate subcellular localization of β-F1–RNP requires two distal cis-acting elements, one placed in the ORF and the other in the 3′UTR (Ricart et al., 2002), and a common set of trans-acting proteins (Ricart et al., 2002). The 3′UTR of rat β-F1 mRNA has the activity of an internal-ribosome-entry sequence (IRES) (Di Liegro et al., 2000; Izquierdo and Cuezva, 2000) and controls the synthesis of the protein at G2-M during cellular proliferation (Martinez-Diez et al., 2006). The activity of the 3′UTR in translation is prevented by the binding of tissue-specific RNABPs, whose binding activity is regulated during development (Izquierdo and Cuezva, 1997) and in hepatocarcinogenesis (de Heredia et al., 2000). Remarkably, it has been consistently reported that the relative cellular expression level of β-F1-ATPase is significantly diminished in tumors compared with normal human tissues (Cuezva et al., 2002), resulting in a ‘bioenergetic signature’ of the biopsy with potential clinical applicability as an indicator of disease progression and response to therapy (for a review see Cuezva et al., 2009).

The aim of this study was to identify those proteins that interact with mRNA ATP5B (hereafter also referred to as β-F1 mRNA), in order to contribute to the knowledge of its biology because of the relevant implication β-F1-ATPase has in human pathology. For this purpose, we have implemented a method for the purification of β-F1-mRNA-binding proteins (β-RNABPs) and report here the identification of nine RNABPs that interact in vitro with β-F1 mRNA. Moreover, we show the in-vivo association of Ras-GAP SH3 binding protein 1 (G3BP1) with the 3′UTR of human β-F1 mRNA and demonstrate that G3BP1 specifically represses the translation of the transcript. Overall, these findings provide a step-forward in the characterization of the molecular determinants that accompany the alteration of the bioenergetic phenotype of the cancer cell.

The biology of β-F1 mRNA isolated from rat liver has been partially characterized (de Heredia et al., 2000; Izquierdo and Cuezva, 1997; Martinez-Diez et al., 2006). However, our knowledge of its human counterpart is missing. A first aim of this study was to verify the putative interaction of the 3′UTR of human β-F1 mRNA (hβ-3′UTR) with cellular proteins. Fig. 1A illustrates that the hβ-3′UTR forms RNA-protein complexes with cellular proteins derived from the human embryonic kidney (HEK) 293T cell line. UV crosslinking assays revealed the specific interaction of hβ-3′UTR with at least six human proteins with apparent molecular masses of 80, 61, 57, 49, 42 and 40 kDa (Fig. 1B), strongly resembling the set of β-RNABPs previously described in extracts of rat liver (Izquierdo and Cuezva, 1997; Ricart et al., 2002) and hepatoma (de Heredia et al., 2000).

To identify the β-F1 mRNA interacting proteins we implemented an affinity-purification technique for RNABPs (Fig. 2A) (Bardwell and Wickens, 1990; Zhou et al., 2002). The approach is based on the high affinity that the bacteriophage MS2 coat protein (CP) has for a short hairpin of its genomic RNA (MS2h; red stem loops in Fig. 2A) (Bardwell and Wickens, 1990; Zhou et al., 2002). A hybrid RNA molecule (hRNA, the bait), consisting of the RNA target of interest fused to MS2h, is incubated with cellular extracts (Fig. 2A). The hRNA-RNABPs complexes are pulled down on amylose beads by binding to a fusion protein (MBP-CP) made up of the MS2-CP fused to the Escherichia coli maltose-binding protein (MBP) (Fig. 2A) (Zhou et al., 2002). Three different hRNAs were prepared: hβ-3′UTR-MS2 and hβfl-MS2, containing the 3′-UTR and full-length human β-F1 mRNA (Fig. 2B), respectively, and RNA-C, which was used as a negative control because it contains a region of the ORF of β-F1 mRNA (Fig. 2B) that, following UV crosslinking, does not interact with cellular proteins (Fig. 1B). An affinity His-tag was inserted upstream the MBP-CP, so that the recombinant His-MBP-CP protein could be purified through Ni²⁺-agarose chromatography (Fig. 2C) in a competent conformation for binding amylose with high efficiency. The recombinant His-MBP-CP binds efficiently to the different synthesized hRNAs at low concentrations (Fig. 2D) and, consistently, is unable to bind RNAs that lack the MS2h tag (Fig. 2D). Titration experiments of the binding activity of His-MBP-CP revealed the formation of three RNA-protein complexes (see asterisk in Fig. 2D).

Other proteins contaminated the RNABPs purification by non-specific binding to the recombinant His-MBP-CP (see lack of differences in protein profiles in M lanes in Fig. 2E; data not shown). Therefore, we implemented an additional step in the purification procedure that consisted in the digestion of the anchored hRNA bait with RNase (see d in Fig. 2A,R; lanes in Fig. 2E). This allowed elution of the proteins bound to the hybrid RNA, and a lesser release rate of His-MBP-CP and other contaminating proteins (Fig. 2E).

In brief, the purification procedure for β-F1 mRNA binding proteins consisted of three sequential binding steps (a-c in Fig. 2A) and the final RNase elution (d in Fig. 2A). The eluted proteins were analyzed by SDS-PAGE (Fig. 2E) and the specifically enriched bands of the hβfl-RNA lane (Fig. 2E) were further processed for its identification by MALDI-TOF and/or HPLC-MS/MS (supplementary material Fig. S1). Sequencing of the bands with the same electrophoretic migration in the RNA-C lane (Fig. 2E) was carried out as negative control of unspecific association. We identified nine proteins that presumably form part of the RNP complex that contains β-F1 mRNA in HEK293T cells (Fig. 2E, Table 1, supplementary material Fig. S1). Importantly, all of the identified proteins are bona fide RNABPs, emphasizing the specificity of the improved affinity method (Fig. 2A). The identified proteins (Table 1) have a modular architecture, displaying the same type of RNA binding domain (RRM, RGG or dsRBM) and are involved in several steps of RNA metabolism (transcription, splicing, export, localization, stability and translation), suggesting their versatility both in their binding and functional activities.

To characterize the biological relevance that some of the identified β-mRNABPs could have in β-F1-ATPase biology, we focused in our study on three proteins: IMP1, G3BP1 and NPM1 (Table 1) which, by their apparent molecular weight (Ricart et al., 2002) and their involvement in cancer (Grisendi et al., 2006; Guitard et al., 2001; Ross et al., 2001; Tessier et al., 2004), could represent putative post-transcriptional regulators of β-F1 mRNA expression. For this purpose, FLAG-tagged versions of IMP1, G3BP1 and NPM1 were transfected into HEK293T cells and processed for RNA immunoprecipitation (RNA-IP) assays by using anti-FLAG antibodies. Western blot analysis confirmed the expression and immunoprecipitation (IP) of the three tagged proteins (see Fig. 3A-C). The reverse-transcriptase (RT) quantitative (q) PCR analysis of the IPs indicated that IMP1 and NPM1 do not interact or only interact marginally with β-F1 mRNA, respectively (Fig. 3B,C). However, cells that overexpressed G3BP1 (Fig. 3A) illustrated the enrichment of β-F1 mRNA in the IPs when compared with control non-transfected cells. The enrichment of β-F1 mRNA in the IPs was similar to that of Myc mRNA, which is a known RNA target of IMP1 (Bernstein et al., 1992) and G3BP1 (Gallouzi et al., 1998; Tourriere et al., 2001), thus suggesting the interaction of G3BP1 and β-F1 mRNA within the cellular context.

We confirmed the specificity of the in-vivo interaction of G3BP1 with β-F1 mRNA both by immunoprecipitation of the endogenous G3BP1 in non-transfected HEK293T cells (Fig. 3D) as well as by fluorescence in-situ hybridization (FISH) of mRNA (Fig. 4A). The obliteration of the partial colocalization of the endogenous G3BP1 with β-F1 mRNA in cells incubated with an excess of the unlabeled riboprobes (Fig. 4B), indicates the specific interaction of G3BP1 with the transcript in non-stressed cells.

β-F1 mRNA binding to G3BP1 was further analyzed by RNA-bridged TriFC (Stohr et al., 2006). In this system, the C-terminal domain of the yellow fluorescent protein (YFP-Ct) is tethered to MS2h-tagged β-F1 mRNA by the CP-MS2h interaction. The complementary N-terminal domain of YFP (YFP-Nt) is fused to G3BP1, so when the RNABP associates with β-F1 mRNA it brings together the two domains of YFP allowing the reconstitution of fluorescence and further revealing the subcellular localization of this interaction (Stohr et al., 2006). To this aim, we generated the C4 cell line that stably expresses MS2h-tagged β-F1 mRNA in 100-fold higher level than that of wild-type mRNA (Fig. 5A). Despite the overwhelming abundance of β-F1 mRNA in C4 cells, the steady-state expression level of β-F1-ATPase (Fig. 5B) and in-vivo synthesis rate of β-F1-ATPase (Fig. 5C) remained unchanged when compared with the parental cells. Interestingly, we confirmed the in-vivo interaction of G3BP1 with the tagged β-F1 mRNA as revealed by the specific cytoplasmic YFP fluorescence in C4 cells (Fig. 5D) and the absence of fluorescent complexes after equimolecular transfection of YFP-Nt–G3BP1 and YFP-Ct–CP fusion proteins in the parental cell line (Fig. 5E). The positive immunostaining of NRK cells with anti-FLAG antibodies confirmed its appropriate transfection (data not shown). Interestingly, YFP fluorescence in C4 cells was found clustered in both small and large cytoplasmic granules (see arrows and arrowheads in Fig. 5D), which might illustrate two different cytoplasmic compartments of the tagged β-F1 mRNA.

The overexpression of G3BP1 promotes the assembly of stress granules (SGs) (Tourriere et al., 2003). Consistently, we observed the partial colocalization of G3BP1 with TIA1, a prominent SG marker (Kedersha et al., 1999), in large cytoplasmic granules when using both anti-FLAG (arrowheads in Fig. 6A) and anti-G3BP1 (arrowheads in Fig. 6B) antibodies, suggesting the partial recruitment of β-F1 mRNA (Fig. 5D) into SGs. However, G3BP1 was also found in smaller granules that were TIA1 negative (arrows in Fig. 6A,B), consistent with the dual cytoplasmic localization of β-F1 mRNA observed in C4 cells (Fig. 5D), and with the colocalization images of endogenous G3BP1 and β-F1 mRNA (Fig. 4). In non-transfected cells the endogenous G3BP1 reveals a diffuse perinuclear cytoplasmic localization (Fig. 6C).

Next, we asked about the possible spatial association of the fluorescent granules containing β-F1 mRNA with mitochondria. For this purpose, C4 cells transfected with both YFP-Ct–CP and YFP-Nt–G3BP1 were incubated with Mitotracker Red (Fig. 7A). Both large (arrowhead in Fig. 7A) and small (arrows in Fig. 7A) β-F1 RNA granules were shown to colocalize with mitochondria. To further confirm this finding, C4 cells were transfected with a vector expressing a green fluorescent protein (GFP)-CP fusion protein that has a nuclear retention signal (Rook et al., 2000), so that the cytoplasmic localization of GFP is only possible when the fusion protein is exported from the nucleus tethering the MS2-tagged β-F1 mRNA. High-resolution immunoelectron microscopy of C4 cells by using an antibody against GFP revealed the specific labeling of the cytoplasm with both scattered and clustered gold signals (Fig. 7B,C). Interestingly, there was also specific and significant GFP labeling of mitochondria in its outer membrane (Fig. 7B,C), suggesting the spatial association between β-F1 mRNA and mitochondria in the C4 cell line.

We next analyzed whether G3BP1 could interact in vivo with the 3′UTR of human β-F1 mRNA (Fig. 8A). To this aim, we studied the reconstitution of YFP fluorescence in NRK cells that transiently express a chimeric reporter RNA made of the β-galactosidase sequence fused to seven MS2-CP binding sites containing or not (Fig. 8A or 8B, respectively) the 3′UTR of β-F1 mRNA. Remarkably, the β-galactosidase chimera bearing the 3′UTR of β-F1 mRNA essentially recapitulates the same cytoplasmic fluorescence pattern observed with the full-length MS2-tagged β-F1 mRNA (compare Fig. 8A with Fig. 5D), whereas the chimera that lacked the 3′UTR was unable to regenerate YFP fluorescence (Fig. 8B). It should be noted that numbers are fewer, and size and fluorescence intensity are less in foci reconstituted in NRK cells with the β-galactosidase reporter compared with foci of the tagged β-F1 mRNA in C4 cells (compare Fig. 8A with Fig. 5D). This might result from differences in the cellular abundance of the corresponding RNA probe and/or from the involvement of additional cis-acting elements and/or β-RNABPs that contribute to the stabilization of the interaction of G3BP1 in the β-F1–RNP complex. Overall, these results indicate that G3BP1 binds the 3′UTR of β-F1 mRNA and suggest that the determinants for the subcellular localization of β-F1 mRNA in the cytoplasm of human cells reside in the 3′UTR of the transcript.

To analyze the functional role that G3BP1 could have in β-F1-ATPase expression we overexpressed G3BP1 in HEK293T cells. The overexpression of G3BP1 (Fig. 9A) did not induce changes in the steady-state levels of β-F1 mRNA (Fig. 9B), suggesting that G3BP1 is not influencing the stability of the transcript. Interestingly, overexpression of G3BP1 induced a specific inhibitory effect in the in-vivo rate of β-F1-ATPase synthesis (Fig. 9C) whereas it had no influence in the synthesis of Hsp60 (Fig. 9C) – a mitochondrial protein used as control – which suggests a role for G3BP1 in the control of β-F1 mRNA translation.

The specific repression of β-F1 mRNA translation exerted by G3BP1 was further confirmed in an in-vitro system (Fig. 10). Addition of G3BP1 (Fig. 10A) to cell-free translation assays primed with β-F1 mRNA triggered a profound repression of the synthesis of the β-F1-ATPase precursor (pβ-F1) when compared to translations that lacked the inclusion of any protein (control) or of equivalent amounts of a non-relevant protein (PK) or of HuR, a 3′UTR β-F1 mRNA binding protein (Fig. 10B) (Ortega et al., 2008).

We next analyzed the interaction of the transcript with the translational machinery by using sucrose-density gradient centrifugation (Fig. 10C) under the same conditions as for in-vitro translations (Fig. 10B). The presence of G3BP1 prevented the recruitment of β-F1 mRNA into the 80S complex (Fig. 10C). By contrast, the absence of G3BP1 or presence of PK (Fig. 10C) allowed the assembly of the transcript into the 80S translation initiation complex. Overall, these results suggest that G3BP1 represents a negative regulator of β-F1-ATPase expression by blocking the initiation step of mRNA translation.

Here we show that human β-F1 mRNA is able to form an RNP complex by recruiting RNABPs from cellular extracts. In particular, the 3′UTR of the transcript is a relevant element for protein recruitment and for its sorting to cytoplasmic RNA granules. By developing an improved affinity-purification method for RNA-binding proteins we have identified nine proteins that putatively interact with β-F1 mRNA. Moreover, we show that G3BP1 interacts with the 3′UTR of β-F1 mRNA and functionally exerts the activity of a repressor of β-F1 mRNA translation, providing a component of the β-F1–RNP complex that might control the bioenergetic phenotype of the cell.

Genetic screens in yeast have identified the hnRNP-like Npl3p as a suppressor of a defect in the mitochondrial targeting sequence of β-F1-ATPase (Ellis and Reid, 1993). Interestingly, this serine/arginine (SR)-type RNABP shuttles between the nucleus and the cytoplasm, and functions as a negative regulator of translation (Lei et al., 2001; Windgassen et al., 2004). However, no human Npl3p homologue has been identified. In our hands, i) genetic screens of mammalian proteins that interact with 3′UTR β-F1 mRNA by using the three-hybrid system as previously described by SenGupta et al. (SenGupta et al., 1996) and ii) our in-vitro experiments failed to provide evidence of a direct interaction of some of the identified proteins with the 3′UTR of β-F1 mRNA (data not shown). This suggests that the assembly of the β-F1–RNP complex is a multifaceted and assisted process (Abdelhaleem et al., 2003; Ricart et al., 2002).

We approached the identification of proteins that interact with β-F1 mRNA by using the improved MS2-based affinity chromatography purification procedure described in this study. We used the full-length β-F1 mRNA as bait to increase the chances of effectively assembling the β-F1–RNP complex in vitro and identified nine RNABPs that interact with β-F1 mRNA. Most of these proteins are known to be involved in virtually all stages of mRNA life cycle and have roles in more than one RNA-processing step, which suggests that they are multifunctional proteins (Huttelmaier et al., 2005; Martin and Ephrussi, 2009; Mongelard and Bouvet, 2007; Moore, 2005). In addition, they are present in other nuclear and cytoplasmic RNP complexes, such as the spliceosome (Chen et al., 2007), the Ago1 complex (Hock et al., 2007), IMP1 granules (Jonson et al., 2007) and FMRP granules (Khandjian et al., 2004), as well as in viral RNPs (Jorba et al., 2008; Mayer et al., 2007; Pacheco et al., 2008). The finding that these proteins built an RNP complex of oxidative phosphorylation highlights for the first time their relevance as post-transcriptional regulators in the biogenesis and function of mitochondria.

Using an engineered version of GFP with a nuclear-retention signal we document the presence of GFP-tethered β-F1 mRNA on the mitochondrial outer membrane. This finding suggests that the human β-F1–RNP complex is transported to and anchored within the periphery of the mitrochondrium – a finding that is in agreement with similar findings in rat hepatocytes (Egea et al., 1997; Lithgow et al., 1997), and also with subcellular fractionation studies in yeast and mammalian cells (Margeot et al., 2002; Ricart et al., 2002; Sylvestre et al., 2003a). Consistent with the sorting of β-F1 mRNA in the cytoskeleton (Jansen, 1999), most of the β-RNABPs identified in this study are known to shuttle from the nucleus to the cytoplasm (Borer et al., 1989; Gallouzi et al., 1998; Huttelmaier et al., 2005; Parrott et al., 2005; Tourriere et al., 2001). Moreover, some of the β-RNABPs are also present in a kinesin-propelled motile RNA granule of neurons (helicases of the DDX family, SFPQ and NonO) (Kanai et al., 2004) and in dynein-transported RNA granules of developing brain (DHX9, ILF3, NCL, IMP1, G3BP1 and NPM) (Elvira et al., 2006). The subcellular distribution and localization of mitochondria in cells of mammals is exerted through the microtubule network (Hirokawa et al., 1991; Nangaku et al., 1994; Rojo et al., 1998); we therefore suggest that the molecular composition of the β-F1–RNP mirrors the structural components that are transported by the two motor proteins kinesin and dynein along microtubules in order to fast-track β-F1 mRNA to the organelle (Lithgow et al., 1997). Interestingly, G3BP1 is also present in RNP granules that contain the microtubule-associated protein tau (Atlas et al., 2004), which participates in the remodeling of the cytoskeleton.

We observed the localization of G3BP1 in the large TIA1-positive granules that contain β-F1 mRNA (Figs 5, 6), consistent with its recruitment into SGs (Kedersha et al., 2005), as a result of G3BP1 overexpression (Tourriere et al., 2003) and/or because of the high expression levels of β-F1 mRNA in the C4 cells. However, we also show that the endogenous G3BP1 in non-stressed cells directly interacts with β-F1 mRNA (Figs 3, 4), indicating the specificity of the G3BP1-β-F1 mRNA interaction in the normal cellular context. Moreover, G3BP1 is also present in the smaller TIA1-negative granules (Figs 4, 5 and 6). We thus suggest that the smaller β-F1 mRNA granules represent the sites of its natural cytoplasmic fate. Consistent with the diffuse cytoplasmic presentation of G3BP1 in human cells (Figs 4, 6) (Parker et al., 1996; Tourriere et al., 2001), we show that the interaction of G3BP1 with the 3′UTR of β-F1 mRNA occurs in the cytoplasm, predominantly in the perinuclear region and after nuclear export of the transcript has occurred (Fig. 8). This suggests that G3BP1 can act as a repressor of β-F1 mRNA translation while the RNA granule is in transit to mitochondria, where its spatially restricted translation will take place (Huttelmaier et al., 2005; Martin and Ephrussi, 2009).

We did not observe the in-vivo interaction of β-F1 mRNA with NPM1 or IMP1. By contrast, we show that G3BP1 binds β-F1 mRNA in vivo, specifically its 3′UTR, and inhibits β-F1 mRNA translation both in vivo and in vitro, supporting its role as a relevant factor of the β-F1–RNP complex in the control of β-F1-ATPase expression. G3BP1 is a conserved and widely expressed protein that is involved in Ras signaling and proliferation, RNA metabolism and SGs assembly (Tourriere et al., 2003; Tourriere et al., 2001). In addition, it was shown to promote the decay of myc mRNA by regulating its endoribonuclease activity in a phosphorylation-dependent way (Tourriere et al., 2001). However, downregulation of β-F1-ATPase synthesis in G3BP1-overexpressing cells is exerted in the absence of relevant changes in the cellular availability of β-F1 mRNA. Moreover, we show that the in-vitro synthesis of β-F1-ATPase is repressed in the presence of G3BP1 (preventing the recruitment of mRNA into the translational machinery), and suggest that G3BP1 inhibits expression of β-F1-ATPase at the level of translation initiation. Binding of RNABPs to the 3′UTR of β-F1 mRNA has been shown to inhibit the translation of β-F1 mRNA (de Heredia et al., 2000; Izquierdo and Cuezva, 1997), probably by masking the intrinsic enhancer activity of the 3′UTR in translation (Izquierdo and Cuezva, 1997; Izquierdo and Cuezva, 2000; Willers et al., 2010). In this scenario, G3BP1 bound to the 3′UTR might sterically interfere in the efficient association between β-F1 mRNA and the translational machinery (Fig. 10). However, we cannot exclude that post-translational regulatory events that influence the activity of G3BP1 (Gallouzi et al., 1998; Irvine et al., 2004; Pazman et al., 2000; Tourriere et al., 2001) also participate in the control of β-F1 mRNA translation.

Remarkably, β-F1-ATPase expression has consistently been reported to be downregulated in most human carcinomas (Cuezva et al., 2004; Cuezva et al., 2002; Isidoro et al., 2005; Lopez-Rios et al., 2007). The downregulation of β-F1-ATPase expression in chronic myeloid leukemia is exerted at the level of mRNA by hypermethylation-mediated silencing of the ATP5B promoter (Li et al., 2010). By contrast, downregulation of β-F1-ATPase expression in human lung, breast and colon cancer is exerted at the level of translation (Willers et al., 2010), in agreement with similar findings in rat hepatocarcinomas (de Heredia et al., 2000) and in fetal rat liver (Izquierdo and Cuezva, 1997). Translation masking of β-F1 mRNA in these situations is recapitulated in in-vitro assays and is mediated by the binding of trans-acting factors to the mRNA (de Heredia et al., 2000; Izquierdo and Cuezva, 1997; Willers et al., 2010). In contrast to the diminished β-F1-ATPase expression in human cancer, G3BP1 is overexpressed in several tumors and cancer cell lines (Barnes et al., 2002; Guitard et al., 2001; Zhang et al., 2007). Although it could be argued that inhibition of β-F1 mRNA translation in response to overexpressed G3BP1 may result from the formation of SGs, the findings that endogenous G3BP1 interacts in vivo with β-F1 mRNA and strongly inhibits its translation support the speculation that G3BP1 has a relevant role in defining the bioenergetic phenotype of the cancer cell. In this regard, we suggest that the control of β-F1-ATPase expression is regulated by specific interactions of the RNABPs with the transcript and other RNABPs within the β-F1–RNP complex (Lal et al., 2004; Langland et al., 1999; Lunde et al., 2007) and that translation masking of β-F1 mRNA in cancer (de Heredia et al., 2000; Willers et al., 2010) is mediated by signaling through G3BP1. In other words, G3BP1 could have a relevant role in the glycolytic switch that occurs in cellular transformation, thereby contributing to the establishment of the Warburg phenotype that characterizes the cancer cell.

For the creation of the hybrid RNAs, DNA encoding three copies of the MS2-phage hairpin loops (MS2h) were inserted downstream of the human full length β-F1-ATPase cDNA (hβfl, NM_001686.3: nucleotides 83-1857), an internal region of β-F1-ATPase cDNA (RNA-C, 946-1392) and the 3′ UTR (hβ-3′UTR, 1696-1852). The hairpin clone (MS2h) and the plasmid encoding the fusion protein MBP-CP were kindly provided by Robin Reed (Harvard Medical School, Boston, MA) (Zhou et al., 2002). For the generation of the His-tagged MBP-CP fusion protein we inserted a sequence coding for five His residues in-frame with the fusion protein at the N-terminus of MBP. The primers used were: HisMBP-CP-5′, 5′-ctgagaaccccgcataatctatggtccttgttggtgaagt-3′, and HisMBP-CP-3′, 5′-catcatcatcatcatggtatgaaaactgaagaaggtaaactggta-3′.

The prβ-MS2 plasmid for eukaryotic expression of rat β-F1 mRNA (Atp5b mRNA) tagged with MS2h was used for the generation of the C4 stable cell line. Plasmid for eukaryotic expression of β-galactosidase mRNA fused to seven MS2h hairpins (pβ-gal-MS2) and the MS2-CP fused to GFP and a nuclear localization signal (pGFP-CP-nls) were kindly provided by Keneth S. Kosik (University of California, Santa Barbara, CA) (Rook et al., 2000). The vector coding for β-galactosidase mRNA fused to the rat 3′ UTR of β-F1 ATPase and seven MS2h hairpins was generated from pβ-gal-MS2 by inserting the 3′UTR region at BglII-NotI sites. The plasmid encoding YFP-Ct fused to hemagglutinin (HA)-tagged CP (YFP-Ct–Ha-CP) was kindly provided by Stefan Hüttelmaier (Martin Luther University, Halle-Wittenberg, Germany) (Stohr et al., 2006). The plasmids encoding YFP-Nt fused to FLAG-tagged G3BP1 (YFP-Nt–FLAG-G3BP1), YFP-Nt fused to FLAG-tagged IMP1 (YFP-Nt–Flag-IMP1) and YFP-Nt fused to FLAG-tagged NPM1 (YFP-Nt–Flag-NPM1) derivate from plasmid YFP-Nt–Flag-ZBP1 (YFP-Nt fused to FLAG-tagged Zip code-binding protein 1), which was also provided by Hüttelmeier (Stohr et al., 2006). Plasmids encoding pyruvate kinase, HuR and G3BP used in vitro-translation assays were purchased from OriGENE (Rockville, MD).

Human embryonic kidney 293T (HEK293T) and normal rat kidney (NRK) cells were cultured as described previously (Di Liegro et al., 2000; Ortega et al., 2008) and plasmid transfections were performed using Lipofectamine and Plus reagent (Invitrogen). For the generation of the C4 cell line NRK cells were transfected with prβ-MS2 and selected with 500 μg/ml of geneticin (G418) (GIBCO/Invitrogen).

For purification of RNABPs and in-vitro binding assays cells were resuspended in RLN buffer (50 mM NaHPO₄-Na₂PO₄ pH 7.4, NaCl 140 mM, 1.5 mM MgCl₂ and 1 mM DTT). Lysis was performed with RLN-T buffer (RLN buffer plus 0.5% Triton X-100 and the complete protease inhibitors cocktail EDTA-free (Roche)) at 20×10⁶ cells/ml for 5 minutes on ice. RNA was extracted from post-nuclear extracts using the RNeasy mini kit (QIAGEN, Hilden, Germany).

Hybrid RNAs (hRNA) were generated by in-vitro transcription using the T7 MEGAscript kit (Ambion/Applied Biosystems). Riboprobes for in-vitro binding assays and sucrose-density gradients were generated using 0.5 U/μl of T7 RNA polymerase (Roche) and 20 μCi of α-[³²P]-UTP (GE Healthcare) for 20 minutes at 37°C (Ricart et al., 2002). UV crosslinking (UVXL) experiments were performed as described (Ricart et al., 2002). In electrophoretic mobility shift assays (EMSAs), the binding reaction conditions were the same as for UVXL experiments but the amount of cellular protein (Izquierdo and Cuezva, 1997) was varied.

MBP-CP was expressed and purified as described (Zhou et al., 2002). His-MBP-CP was produced in the E. coli BL21 strain (Studier, 2005) and tandem affinity purified using both amylose-based (New England Biolabs, Ipswich, MA) and Ni²⁺-based (GE Healthcare) affinity chromatography. For RNABPs experiments, His-MBP-CP aliquots (1.5 mg) were thawed and incubated with amylose resin (1 ml) for 3 hours at 4°C. The resin was washed three times with five volumes of first RLNT and second RLN.

Hybrid RNA (0.1 nmol) was incubated with 1 nmol of His-MBP-CP bound to the resin in RLN buffer at a resin:RLN ratio of less than 1:10 (v/v) for 1-2 hours at 4°C. The RNA-protein complexes bound to the resin were washed three times with ten volumes of RLN buffer and incubated with cellular extracts (12-16 mg of protein) for 2 hours at 4°C. The resin was subsequently washed five times with 50 volumes of RLN buffer. Elution of the RNA-binding proteins (RNABPs) bound to the hybrid RNA was carried out by incubating the resin with 20 μg/ml of RNase A in RLN at 37°C for 30 minutes. Then, a further elution with 0.1 M maltose or 0.5 M imidazol in RLN buffer was carried out. Eluted proteins were acetone precipitated and protein samples were resolved by SDS-PAGE. Gels were fixed and stained with AgNO₃ (GE Healthcare). Each purification experiment consisted of three parallel assays in which we compared the profile of eluted proteins obtained when using hβfl-RNA as bait (Fig. 2C) with the profiles obtained using the non-relevant β-F1 mRNA portion as control (RNA-C in Fig. 2C) and when not adding any hRNA (−RNA in Fig. 2B). Protein spots were excised manually and then digested automatically (Shevchenko et al., 1996) using a Proteineer DP protein digestion station (Bruker-Daltonics, Bremen, Germany).

Aliquots of protein digestions were analyzed by mass spectrometry as described (Suckau et al., 2003). MALDI-MS and MS/MS data were combined through MS BioTools program (Bruker-Daltonics) to search the NCBInr database using Mascot software (Matrix Science, London, UK) (Perkins et al., 1999).

HEK293T cells (~10⁶ cells) were transfected with 3 μg of Flag-G3BP, Flag-IMP1 or Flag-NPM1 and 48 hours post-transfection were resuspended in lysis buffer (50 mM Tris-HCl pH 8.0, 140 mM NaCl, 1.5 mM MgCl₂, 0.5% Triton-X-100) supplemented with complete protease inhibitors cocktail EDTA-free (Roche) and RNase inhibitor (Applied Biosystems). The pre-cleared cell lysate was incubated over night at 4°C with monoclonal anti-FLAG-M2 (Sigma) or monoclonal anti-G3BP1 (BD Bioscience) antibodies. Thereafter, the lysate was mixed with protein-G–sepharose beads and incubated at 4°C for 2 hours. The beads were washed with lysis buffer, the final wash containing 1 M urea. Finally, the resin was resuspended in a buffer containing 50 mM Tris-HCl pH 8.0, 140 mM NaCl, 1.5 mM MgCl₂, 0.25% Triton-X-100, 0.1% SDS supplemented with proteinase K (Roche) and incubated for 30 minutes at 50°C. The RNA was extracted using TriPure reagent (Roche Applied Science).

1 μg of RNA was reverse-transcribed (RT) using the High-Capacity cDNA Archive Kit (Applied Biosystems). Quantitative PCR (qPCR) was performed using the LightCycler FastStart DNA MasterPlus SYBR Green Mix (Roche). The following forward (F) and reverse (R) primers were used to amplify human β-F1-ATPase, G3BP1, c-Myc and GAPDH cDNAs: F-βF1: 5′-cagcagattttggcaggtg-3′, R-βF1: 5′-cttcaatgggtcccaccata-3′; F-G3BP1: 5′-ctttggtgggtttgtcactg-3′, R-G3BP1: 5′-tgctgtctttcttcaggttcc-3′; F-cMYC: 5′-cacaaacttgaacagctacgg-3′, R-cMYC: 5′-ggtgattgctcaggacatttc-3′ and F-GAPDH 5′-agccacatcgctcagacac-3′, R-GAPDH: 5′-gcccaatacgaccaaatcc-3′. Relative expression levels of β-F1 mRNA and MYC mRNA were determined using the comparative ΔΔCt method with GAPDH as a control and relative to the not-transfected cells. The primers for quantifying MS2, β-F1 and β-actin of rat origin were F-β-act: 5′-tgaaccctaaggccaaccg-3′, R-β-act: 5′-aggtctcaaacatgatctgggtc-3′; F-rat-β-F1: 5′-aaagctggtgcccctgaag-3′, R-rat-β-F1: 5′-ggagatggtcatagtcacctgct-3′; F-MS2-Dra: 5′-cgctttaaaggatccgatatccgt-3′ and R-MS2-Dra: 5′-cgctttaaagaattccgtaccctg-3′.

The TriFC assay (Stohr et al., 2006) was used to visualize and localize the β-F1 mRNA-G3BP1 interaction in living cells. C4 and NRK cells were cultured in glass-bottom dishes coated with poly-D-lysine (MatTek Corporation, Ashland, OR), and were co-transfected with equal amounts of YFP-Ct–Ha-CP and YFP-Nt–FLAG-G3BP1 plasmids. NRK cells were cultured and transfected as described above, but additionally co-transfected with the reporter mRNA plasmid βGal-3′β-MS2 or βGal-MS2. In this system, YFP-Ct–HA-CP is tethered to the RNA as a result of the high affinity of the MS2 coat protein (CP) for its MS2 RNA hairpins. If G3BP1 associates with the β-F1 mRNA, the two YFP domains come in close proximity and the fluorescent complex is reconstituted. Cells were visualized using an inverted confocal LSM510 META microscope using a 63×1.4 oil Plan-Apocromate objective.

Riboprobes for β-F1 mRNA fluorescence in-situ hybridization (FISH) were synthesized using the mirVANA miRNA probe construction kit (Ambion/Applied Biosystems) and the following oligonucleotids: hβF1-1526-1575 5′-tggtgggacccattgaagaagctgtggcaaaagctgataagctggctgaacctgtctc-3′; hβF1-1188-1236 5′-agctgtggatcctctagactccacctctcgtatcatggatcccaacattcctgtctc-3′; hβF1-792-838 5′-ggtagcgctggtatatggtcaaatgaatgaaccacctggtgctcgtgcctgtctc-3′; hβF1-566-619 5′-gatctgctagctccctatgccaagggtggcaaaattgggctttttggtgcctgtctc-3′; hβF1-284-331 5′-tggttttggaggtggcccagcatttgggtgagagcacagtaaggactacctgtctc-3′, except that UTP was replaced by aminoallyl-UTP (Sigma). A riboprobe mixture containing 5 μg of each probe was labeled using the Cy5 dye (GE Healthcare).

HEK293T cells were transfected with either 1.5 μg of Flag-G3BP1 or βGal-MS2 as a control. At 48 hours post-transfection cells were fixed with 2% paraformaldehyde and processed for immunofluorescence (Martinez-Diez et al., 2006). Coverslips were incubated with monoclonal anti-FLAG M2 (1:1000, Sigma Aldrich) or anti-G3BP1 (1:200, BD Biosciences) and polyclonal anti-TIA1 (1:200, sc-1751, Santa Crux Biotechnology, Santa Cruz, CA) antibodies. Alexa-Fluor 488 and Alexa-Fluor 555 (1:500) conjugated antibodies (Invitrogen) were used as secondary antibodies. For combined immunofluorescence and FISH, the cells were fixed with 4% paraformaldehyde for 20 minutes at room temperature. Immunofluorescence of G3BP preceded FISH of β-F1 mRNA. For in-situ hybridization the labeled riboprobes were hybridized overnight at 37°C in a solution containing 2×SCC, 50% formamide, 100 μg salmon sperm DNA and 100 μg of yeast tRNA. For competition experiments, a 100-fold excess of the unlabeled β-F1 mRNA riboprobe mixture was added one hour before the addition of the Cy5-labeled probes. Cellular fluorescence was analyzed by confocal microscopy using a 63×1.4 oil plan-apocromate objective.

Antibodies and dilutions used in this study were: rabbit polyclonal anti-β-F1-ATPase (1:20,000) (Cuezva et al., 2002) and mouse monoclonal antibodies against α-tubulin (1:5000) (B-5-1-2, Sigma Aldrich), FLAG-epitope M2 (1:500) (M2, Sigma Aldrich), anti-G3BP1 (BD Biosciences) and anti-GFP (1:5000) (JL-8, Clontech, Mountain View, CA).

Rates of in-vivo synthesis of β-F1-ATPase and Hsp60 were determined as recently described (Ortega et al., 2008). Briefly, after 30 minutes of metabolic labeling, the incorporation of ³⁵S-Met into β-F1-ATPase and Hsp60 was determined by sequential immunoprecipitation of the proteins from cellular lysates (Ortega et al., 2008).

Cells were fixed 30 hours post-transfection in 4% paraformaldehyde in 0.1 M Sörensen phosphate buffer (pH 7.4). After freeze substitution the samples were infiltrated with Lowicryl HM20 (Egea et al., 1997). Ultrathin sections were incubated with anti-GFP antibody (1:50 dilution, JL-8, Clontech) and with protein A conjugated to 15 nm gold particles (Egea et al., 1997). Counterstaining was performed with 2% aqueous uranyl acetate and 1% lead citrate. Grids were examined at 80 KV using a JEM1010 electron microscope (Jeol, Japan).

Synthesis of PK, HuR and G3BP was performed using the TnT T7 (HuR and G3BP) or SP6 (PK) expression system (Promega, Madison, WI). In-vitro translation of β-F1-ATPase was carried out with the Rabbit Reticulocyte Lysate System (GE Healthcare) using 100 ng of in-vitro transcribed and capped β-F1 mRNA, 1.6 μCi/μl of [³⁵S]-Met and [³⁵S]-Cys labeling mix, 75 mM KCl and 0.5 mM magnesium acetate for 1 hour at 30°C. For sucrose-density-gradient analysis, the capped ³²P-labeled β-F1 mRNA (2×10⁶ cpm) was incubated under the same conditions for 20 minutes and the reactions stopped by addition of cyclohexamide (CHX; 100 μg/ml). Samples were further diluted in polysome buffer (10 mM Tris-HCl pH 7.4; 80 mM KCl, 5 mM MgCl₂, 1 mM DTT) supplemented with CHX (100 μg/ml). The samples were loaded onto a linear 15-40% (w/v) sucrose gradient and centrifuged at 247,000 g for 130 minutes at 4°C in a SW 41 rotor. Fractions were collected from the bottom of the tube and the radioactivity was determined by scintillation counting.

We thank Encarnación Martínez-Salas (CBMSO) for critical advice and reading of the manuscript and M. Chamorro for skillful technical assistance. We are indebted to Zhaolan Zhou and Robin Reed (both Harvard Medical School, Boston, MA), Keneth S. Kosik (University of California, Santa Barbara, CA) and Stefan Huttelmaier (Martin Luther University, Halle-Wittenberg, Germany) for supplying reagents. This work was supported by grants from the Ministerio de Educación y Ciencia (BFU2007-65253), the Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER) from the Instituto de Salud Carlos III and the Comunidad de Madrid (S-GEN-0269). A.D.O. and I.W. were recipients of predoctoral fellowships from the Plan de Formación de Profesorado Universitario from the Spanish Ministry of Education, Spain. The CBMSO receives an institutional grant from the Fundación Ramón Areces.