Alternative initiation of translation at three CUG and one AUG start codons leads to the synthesis of four isoforms of fibroblast growth factor 2 (FGF-2) that have distinct intracellular localizations and affect the cell phenotype differently. We show here that the expression of FGF-2 CUG-initiated isoforms decreases in a cell-density-dependent manner in normal human skin fibroblasts (HSFs) concomitantly with the FGF-2 mRNA level. In contrast, CUG-initiated FGF-2 expression is constitutive in SK-HEP-1 cells and in HSFs transformed with SV40 large T antigen. Cell transfection using a plasmid containing the FGF-2 mRNA leader fused to chloramphenicol acetyl transferase demonstrated that up-regulation of the CUG codons depends on cis-elements located in this leader. Furthermore, UV cross-linking experiments revealed a correlation between CUG codons utilization and the binding of several proteins to the mRNA leader. On the basis of the presence of an internal ribosome entry site (IRES) in the FGF-2 mRNA, we used bicistronic vectors to transfect normal and transformed cells. The density-dependent regulation in normal HSFs was cap-dependent, whereas the constitutive CUG-initiated FGF-2 expression in transformed cells occurred essentially by an IRES-dependent mechanism. Unexpectedly, the use of the AUG start codon occurred exclusively by internal entry, which suggests the presence of a second independent IRES in the FGF-2 mRNA that would be constitutive. A study of the eIF-4E levels and of the 4E-BP1 phosphorylation state at increasing cell densities showed a decrease of the eIF-4E level, concomitant with 4E-BP1 dephosphorylation in normal cells but not in transformed cells. These data point out a complex mechanism for the regulation of FGF-2 isoforms expression involving both the cap-dependent and the cap-independent initiation of translation and favor a positive role of CUG-initiated FGF-2 in cellular proliferation and transformation.
Tumor growth is partly determined by the proliferative features of tumor cells as well as by the ability to induce neovascularization. Human FGF-2,3 the prototype member of the large family of FGF genes, is a potent angiogenic factor and also a potential oncoprotein able to modulate both of the tumor formation parameters (1, 2). The involvement of FGF-2 in tumor growth has been demonstrated in numerous works: (a) FGF-2 is expressed in a broad range of cancerous cells, its expression and its release being modified by the tumor stage (2, 3, 4); (b) the use of DNA or RNA antisense results in the reversal of the transformed phenotype (5, 6, 7, 8, 9); (c) neutralizing antibodies against FGF-2 are able to inhibit tumor growth in syngenic mice (10, 11); and (d) the overexpression of FGF-2 leads to the transformation of NIH3T3 and endothelial cells (1, 12). In addition to its role in tumor growth, FGF-2 is involved in development and tissue repair and also in retinopathies, neuromuscular degenerative pathologies, atheromatous plaque formation, and so forth (13).
FGF-2 is synthesized as four molecular species. Such diversity is the result of a translation initiation at four alternative codons, i.e., one AUG and three upstream in-frame CUG codons on the same mRNA (14). These FGF-2 factors have distinct subcellular localizations and affect the cell phenotype differently. Constitutive expression of the Mr 18,000 AUG-initiated FGF-2 leads to the transformation of NIH 3T3 cells or adult bovine aortic endothelial cells, whereas that of the HMW CUG-initiated isoforms results in the transformation of NIH 3T3 cells with a unique phenotype and the immortalization of adult bovine aortic endothelial cells (1, 12). The Mr 18,000 FGF-2 is mostly cytoplasmic but is also secreted via an unknown pathway independent of the golgi apparatus (15, 16, 17); therefore, this isoform acts as a para- and autocrine factor. In contrast, the HMW isoforms are nuclear and have an intracrine mode of action (17, 18, 19, 20).
Human FGF-2 isoforms expression is regulated in a tissue-specific manner in transgenic mice (21) and also by activators of protein kinase C and cAMP pathways (22). Furthermore, their expression profile varies according to the cell type: normal human cell types such as skin fibroblasts, retinal pigment epithelium cells, and aortic endothelial cells mostly express the Mr 18,000 FGF-2 at confluence, whereas the cancerous cell lines, when expressing FGF-2, produce all isoforms (23). Expression of the HMW isoforms is also induced by stress in nHSFs (23).
The translational regulation of FGF-2 expression requires cis-elements located in the 5′ untranslated region or in the alternatively translated region of its messenger (24). In particular, an IRES has been described in the FGF-2 mRNA leader (25). This allows FGF-2 mRNA translation to occur according to a mechanism different from the classical cap-dependent scanning mechanism (26), which is blocked under stress conditions or during viral infection. The internal ribosome entry mechanism, first described in picornavirus, has been discovered in a few other cellular messengers that mostly encode regulatory proteins (27, 28, 29, 30). Furthermore, the viral IRES activity has been proved to be controlled by cellular trans-acting factors, which suggests that cellular mRNA IRESs could also be trans-regulated by specific factors (31, 32, 33, 34). Consistent with this, activation of the CUG-initiated forms of FGF-2 is correlated with the binding of a p60 protein to the mRNA leader in stressed nHSFs (23). In cancerous cells, the expression of CUG-initiated FGF-2 also coincides with the binding of several proteins. This suggests that one or several of these proteins may be involved in the control of FGF-2 translation and may be responsible for IRES activation in stressed and cancerous cells. Kevil et al. (35) have also shown that expression of the FGF-2 isoforms from the rat FGF-2 mRNA is controlled by the level of the cap-binding protein eIF-4E, which suggests that the control of FGF-2 mRNA translation can involve mechanisms both of internal entry of ribosome and of cap recognition followed by ribosome scanning.
The literature describes several examples of growth factors and cytokines that are modulated as cell density increases (36, 37, 38). FGF-2 is regulated in a density-dependent manner in several cell types (39, 40, 41, 42, 43). The density-dependent regulations reported up to now mostly concern mRNA and occur at the level either of transcription or of mRNA stability. Very little information is available about translational control. However a well-described example is that of the c-myc proto-oncogene, which is regulated at the level of alternative initiation of translation. Expression of the CUG-initiated c-myc1 protein is activated as cell density increases (44).
The FGF-2 mRNA, like c-myc, expresses several alternative translation products the expression of which is modulated in normal cells in response to stress. These different data prompted us to look for a possible density-dependent regulation of FGF-2 isoforms expression at the translational level. We provide evidence here that expression of the CUG-initiated forms of FGF-2 is translationally down-regulated when cell density increases in nHSF cultures. In contrast, the HMW FGF-2 expression is constitutive in cancerous SK-HEP-1 cells as well as in HSFs transformed by SV40 large T antigen. This translational control requires elements within the FGF-2 mRNA leader and is correlated with the binding of cell proteins to the leader. Furthermore, we show that the density-dependent regulation in normal cells is cap-dependent, whereas the constitutive expression in transformed cells involves an IRES-dependent translational deregulation.
MATERIALS AND METHODS
nHSFs were obtained from the “Laboratory of Human Skin Cultivation” (Centre Hospitalier Regional Rangueil, Toulouse, France). SK-HEP-1 cells (human liver adenocarcinoma) and COS-7 cells (SV40 transformed African green monkey kidney) were obtained from the American Type Culture Collection (Manassas, VA). tHSFs were obtained after electroporation of nHSFs with the pAS plasmid that encodes the early functions of SV40 (45). Electroporated cells were reseeded every 15 days at 150 cells/mm2 (this constitutes a passage in Fig. 2) for the study of cell density effects on FGF-2 expression. At passage 30 after transfection, the cells showed features of the transformed phenotype. The cells were cultivated at 37°C in a 5% (nHSF, tHSF, and COS-7 cells) or 10% (SK-HEP-1) CO2 incubator in DMEM supplemented with 10% (nHSF, tHSF, and SK-HEP-1 cells) or 5% (COS-7 cells) FCS, 0.5% gentamicin, 1% glutamin, and 1% amphotericin (Seromed).
The pFC1, phCFC, and pHCFC plasmids used for cell transfection have been described previously (25). Adherent SK-HEP-1 cells, COS-7 cells, and tHSFs were seeded at different cell densities and transfected with 2.5 μg of plasmids by using the lipofectin reagent (Life Technologies, Inc.) according to the manufacturer’s instructions. nHSFs (2.106 cells) were electroporated with 10 μg of plasmid as described previously (23). Cells in suspension were seeded at different densities, and the medium was replaced 16 h later. The cells were scraped 72 h after transfection for Western blot analysis.
Preparation of Cell Extracts and Western Blot Analysis.
The cell monolayers were scraped in ice-cold PBS and pelleted by centrifugation. The cell pellets were lysed in PBS containing 2% SDS + protease inhibitors [phenylmethylsulfonyl fluoride (1 mm), aprotinin (2 μg/ml), leupeptin (2 μg/ml), N-Tosyl-l-Lysine chloromethyl ketone (50 μg/ml), N-Tosyl-l-phenylalanine chloromethyl ketone 100 μg/ml, pepstatin (0.7 μg/ml), and soybean trypsin inhibitor (100 μg/ml; Boehringer Manheim]. Cell lysates were sonicated, and the total protein content was determined by bicinchoninic acid method (Pierce). Equal amounts of total proteins were resolved by SDS-PAGE and transferred onto nitrocellulose membrane as described previously (7). FGF-2 and CAT proteins were revealed as described previously (23). P53 and Tag were immunodetected with monoclonal anti-p53 (DO1 diluted to 1:10,000) and anti-Tag (Pab101 diluted to 1:1,000) monoclonal antibodies purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-β-actin antibody (AC-74, Sigma) was used at a 1:20,000 dilution. Antibodies were detected using the ECL enhanced chemiluminescence system (Amersham Corp., Arlington Heights, IL). For the detection of eIF-4E and 4E-BP1, cells were lysed by three freeze-thaw cycles in a Tris-KCl buffer [20 mm Tris (pH 7.5), 150 mm KCl, 1 mm DTT, 1 mm EDTA, 300 μm NaVO3, and 45 mm glycerophosphate] containing protease inhibitors. Cellular debris were spun down by a 5-min centrifugation in a microfuge. Protein concentration was determined as above, and 50 μg of proteins were loaded onto a 15% polyacrylamide SDS gel. After transfer onto nitrocellulose, the filters were blocked in Tris-buffered saline [10 mm Tris (pH 8) and 150 mm NaCl] containing 5% nonfat dry milk and 0.1% Tween 20 for 3 h at room temperature and then were incubated with a rabbit polyclonal anti-4E-BP1 antibody at a 1:500 dilution and a monoclonal anti-4E antibody at a 1:2000 dilution. The remaining steps of the procedure were carried out as described for the other antigens. 10C6 anti-eIF-4E (46) and 11208 anti-4E-BP1 (47) antibodies were generous gifts from Nahum Sonenberg (McGill University, Montreal, Canada).
RNA Purification and RT-PCR Quantification.
The cell monolayers were scraped in ice-cold PBS and pelleted by centrifugation. RNA was prepared from the cell pellets by the Trizol method (Life Technologies, Inc.) as described previously (23). RNA was quantitated by measuring the absorbance at 260 nm and was checked for integrity by electrophoresis on agarose gel followed by ethidium bromide staining. The cDNAs were synthesized using the Superscript preamplification system (Life Technologies, Inc.) according to the manufacturer’s instructions in a final volume of 20 μl with 1 μg of total RNA and 50 ng of random hexamers in the presence of variable amounts of internal standard RNA. The remaining steps of PCR amplification and calculations of FGF-2 mRNA levels were done as described previously (23).
UV Cross-linking Assays.
S10 cytoplasmic extracts were prepared as already described (23). 32P-labeled RNA (105 cpm) was incubated with 4 μg of S10 extract and was UV-irradiated as described previously (23). The samples were then treated with 2.5 units of RNase A and 10 units of RNase T1 (Sigma) at 37°C for 30 min before 10% SDS-PAGE analysis and autoradiography.
Expression of CUG-initiated FGF-2 Isoforms Is Regulated by Cell Density in nHSFs.
Two human cell types were used to study the effects of cell density on FGF-2 expression: (a) nHSFs that mostly produce the Mr 18,000 FGF-2; and (b) hepatoma SK-HEP-1 cells that express the four FGF-2 isoforms at confluence (23). Cells were seeded at different cell densities and harvested 72 h later. The status of FGF-2 mRNA was analyzed in nHSFs by quantitative RT-PCR and showed a 10-fold decrease of FGF-2 mRNA as a function of cell-density increase (Fig. 1 A) in agreement with the literature (43).
The FGF-2 isoforms expression was analyzed for both cell types by Western blotting using a polyclonal anti-FGF-2 antibody (Fig. 1, B and C). In nHSFs cultivated at a low density, CUG-initiated FGF-2 constituted nearly 50% of the total FGF-2, whereas it drastically decreased as cell density increased (Fig. 1,B, left panel, and Fig. 1,C). At high density (420 cells/mm2), expression of the HMW FGF-2 isoforms was hardly detectable as shown previously (23). In contrast, expression of the Mr 18,000 AUG-initiated FGF-2 was not influenced significantly by cell density (Fig. 1 B).
Unlike nHSFs, the SK-HEP-1 cells did not show any variation in their FGF-2 expression, at low or high cell density (Fig. 1 B, right panel). Such an absence of regulation was also observed in other cancerous cell lines such as SK-OV3 or HeLa cells (data not shown). These results show that CUGs utilization is regulated by cell density in nHSFs but not in cancerous cell lines, suggesting a relation between cell transformation and constitutive expression of the HMW FGF-2.
Cell Density-dependent Regulation of FGF-2 Isoforms Expression Is Lost during the Cell Transformation Process.
To check the hypothesis of a direct relationship between the absence of density-dependent regulation of HMW FGF-2 expression and cell transformation, we looked for a putative change in the regulation of FGF-2 expression during the transformation process of nHSFs. nHSFs were electroporated with the pAS plasmid that encodes Tag because of its strong oncogenic properties. Cells expressing Tag were naturally selected by the proliferative advantage conferred by the oncoprotein. Confluent cells were harvested at different passages after electroporation. Analysis of FGF-2 expression by Western immunoblotting showed that high-density-cultivated HSFs became able to produce the HMW FGF-2 isoforms in proportion to their Tag expression, as shown by immunodetection of Tag (Fig. 2). The activity of Tag was detected by following the accumulation of p53 that was stabilized and inactivated by Tag Fig. 2, bottom panel). The FGF-2 expression profile at passage 30 after transfection was similar to that observed in SK-HEP-1 cells [Fig. 2 (P30); compare with Fig. 1 B].
These cells displayed features of transformed cells as follows: (a) their saturation density was higher than that of nHSFs; (b) their doubling time in serum-rich medium was lower than that of nHSFs; (c) their doubling time was not affected by low-serum growth conditions unlike nHSFs; and (d) they formed colonies in soft agar (data not shown). FGF-2 expression was not modified by cell density in Tag-transformed HSFs (Fig. 3), which suggests in concordance with the results of Fig. 1 B that cellular transformation resulted in the loss of the ability to regulate the expression of the HMW FGF-2.
cis-Acting Elements in the FGF-2 mRNA Leader Are Sufficient to Ensure the Cell Density-dependent Regulation of FGF-2 Isoforms Expression.
A decrease in the FGF-2 mRNA level was observed concomitantly with the decrease in CUG-initiated FGF-2 expression (Fig. 1). However, we previously described (23) that the induction of CUG codons utilization occurred in stressed confluent nHSFs without any change in the mRNA level. This lack of correlation between HMW FGF-2 expression and the FGF-2 mRNA levels suggested a control mechanism at the translational level. To check this hypothesis for cellular density and cellular tranformation, we transfected nHSFs and tHSFs with a chimeric DNA containing the first 539 nucleotides (nt) of the FGF-2 mRNA leader fused to the CAT open reading frame under the control of the strong cytomegalovirus promoter. This construct allows expression of the four FGF-CAT chimeric proteins (24). Western immunoblotting was performed with transfected cell extracts, using either anti-CAT antibody or anti-FGF-2 antibody (Fig. 3). Expression of both CUG-initiated FGF-CAT (Fig. 3, nHSF, right panel) and endogenous HMW FGF-2 (Fig. 3, nHSF, left panel) decreased as cell density increased. These results show that the cell density-dependent regulation of FGF-2 expression occurs at the translational level. Moreover, cis-acting elements present in the FGF-2 mRNA leader are sufficient to mediate this regulation, independently of the coding region and of the 3′ untranslated region of this mRNA.
As expected, the leader of the FGF-2 mRNA mediated constitutive expression of CUG-initiated FGF-2 in tHSFs (Fig. 3, lower panels).
Interaction of Cellular Factors with the FGF-2 mRNA Leader Is Related to the Expression of HMW FGF-2.
The prominent role of FGF-2 mRNA leader in the regulation of FGF-2 isoforms expression prompted us to look for cellular factors able to interact with this 5′ region. S10 extracts from nHSFs and tHSFs as well as from SK-HEP-1 cells cultivated at increasing densities were incubated with a 32P-labeled RNA probe corresponding to the FGF-2 mRNA leader. After UV irradiation, the samples were RNase treated and subjected to SDS-PAGE analysis. Fig. 4 shows that the pattern of cross-linked proteins either from SK-HEP-1 or from tHSF cells was not influenced by cell density (Fig. 4, Lanes A–C and G–I). Interestingly, the cross-linked proteins observed with extracts from both SK-HEP-1 cells and tHSFs had the same apparent molecular weight (Fig. 4 Lanes A–C and G–I). Several of these proteins were also detected with extracts from proliferating nHSFs (Fig. 4, Lane D). However, one important difference was a p110 (Fig. 4, Lanes A–C and G–I, band a), the main bound product in both transformed cell types, that could not be detected in the nHSFs in which a p100 (Fig. 4, Lane D, band b) was specifically revealed. Another important difference was that the pattern of cellular factors from nHSFs cross-linked to the RNA probe varied as a function of cell density (Fig. 4, Lanes D–F). As cell density increased, most of these cross-linked factors disappeared and a few others appeared.
These results indicated a correlation between the use of the CUG initiation codons and the binding of several factors onto the FGF-2 mRNA leader. One or more of these proteins could be involved in the regulation of CUG codons utilization.
The Cell Density-dependent Regulation of FGF-2 Translation Occurs via a Cap-dependent Mechanism.
Different data suggest that the initiation of FGF-2 mRNA translation can involve both internal entry of ribosomes or cap recognition followed by ribosome scanning (25, 35).
To discriminate between cap-dependent and IRES-driven translation, we used the bicistronic vectors strategy: in such a test, the second cistron is expressed only in an IRES-dependent manner (28, 35, 48). Expression of the monocistronic and bicistronic plasmids with a medium or very stable 5′ hairpin structure (Fig. 5,A) was analyzed in low-density cultivated nHSFs, as well as in tHSFs, SK-HEP-1 cells, and COS-7 cells. For bicistronic plasmid, phCFC, expression of the first cistron was visualized by immunodetection of the CAT protein (Fig. 5, B–E, Lanes II). This cap-dependent expression was efficiently blocked by the addition of a very stable hairpin at the 5′ end of the messenger encoded by pHCFC (Fig. 5, B–E, Lanes III).
In nHSFs, the pattern of FGF-CAT isoforms expression of the monocistronic vector (Fig. 5,B, left panel, Lane I) was comparable to that of endogenous FGF-2 expression (Fig. 5,B, right panel, Lane I’). In contrast, the CUG-initiated FGF-CATs were very weakly expressed with the bicistronic vectors (Fig. 5,B, left panel, Lanes II and III) whereas all of the endogenous FGF-2 isoforms were still expressed (Fig. 5,B, right panel, Lanes II’ and III’). Interestingly, the AUG-initiated FGF-CAT protein was expressed with a similar efficiency from the monocistronic and bicistronic vectors and was not influenced by the stability of the 5′ hairpin (Fig. 5 B, Lanes I, II, and III). This indicates that translation initiation at this AUG codon is strictly cap-independent.
In tHSFs as well as in SK-HEP-1 and COS-7 cells, expression of the CUG-initiated isoforms efficiently occurred from the bicistronic vectors (Fig. 5, C–E, compare Lane I with Lanes II and III). In COS-7 cells, the CUG1-initiated protein is expressed at a slightly lower level than that observed with the monocistronic vector (Fig. 5 E). Thus, the initiation of translation at CUG codons seems to be partly cap-dependent in these Tag-transformed cells. This discrepancy may reflect a cell type-specific ability of Tag-induced transformation to modulate the two different mechanisms governing the utilization of CUGs.
We conclude from these results that in proliferating nHSFs, CUG-initiated form expression is not driven by internal ribosome entry. In contrast, in tHSFs, SK-HEP-1 cells, and COS-7 cells, an IRES-dependent mechanism seems to be mainly involved.
These results also show that AUG codon utilization occurs via an internal ribosome entry, even in the proliferating nHSFs. This would indicate the existence, in the FGF-2 mRNA, of an autonomous second IRES controlling the utilization of the AUG but not of the CUG start codons.
eIF-4E Expression Is Regulated in a Density-dependent Manner in nHSFs but not in Their Transformed Counterpart.
As the internal entry of ribosome cannot account for the presence of HMW FGF-2 in proliferating nHSFs, we hypothesized that cap-dependent initiation of translation could be enhanced in these cells. We, therefore, analyzed the levels of the cap-binding protein eIF-4E in nHSFs and tHSFs at different densities, shown to be limiting for translation initiation (49). Western immunoblotting using anti-4E antibody showed a drastic decrease in the eIF-4E level of high-density-cultivated nHSFs (Fig. 6,A, left panel). In tHSFs (Fig. 6 A, right panel), the level of eIF-4E expression was similar to that of untransformed cells at low density without overexpression of the factor. This observation is quite different from previous results that showed high levels of eIF-4E in cancerous tissues versus normal tissues (35, 50). However, the level of eIF-4E remained unchanged at high cell density.
These results suggest a relationship between HMW FGF-2 expression and eIF-4E levels in nHSFs.
4E-BP1 Phosphorylation State Is Cell Density-dependent in nHSFs and tHSFs.
Several proteins have been described for their ability to sequester eIF-4E by a direct interaction that depends on their phosphorylation state (51). 4E-BP1 is the best characterized. We analyzed the effect of cell density on 4E-BP1 phosphorylation in nHSFs or tHSFs by Western-blot analysis with anti-4E-BP1 antibody (Fig. 6,B). In low-density cultivated normal and transformed cells, 4E-BP1 was mostly detected in the slower migrating band, corresponding to the hyperphosphorylated form unable to sequester eIF-4E (Fig. 6,B, band HP). When cell density increased, a diminution of the hyperphosphorylated band and an increase of the phosphorylated and unphosphorylated 4E-BP1 bands was observed in both nHSFs and tHSFs (Fig. 6,B, bands P and HP, Lanes 2, 3, 5 and 6). However, at high density, the unphosphorylated form of 4E-BP1, which is able to sequester eIF-4E, was prominent in nHSFs (Fig. 6,B, Lane 3), whereas 4E-BP1 was mostly present in the phosphorylated form in tHSFs (Fig. 6 B, Lane 6). Despite these slight differences, the mechanisms controlling the phosphorylation state of 4E-BP1 are not completely deregulated in transformed cells.
In conclusion, we showed a correlation between cell-density increase and 4E-BP1 dephosphorylation, which indicates a potential sequestration of eIF-4E in both nHSFs and tHSFs cultivated at high density, although 4E-BP1 occurred in a less phosphorylated state in the nHSFs than in their transformed counterpart. Thus, HMW FGF-2 expression is related to the desequestred eIF-4E level in normal cells but not in transformed fibroblasts: the density-dependent CUG utilization seems to be eIF-4E-dependent in nHSFs, whereas the constitutive CUG expression in tHSFs cannot be explained by 4E-desequestration, thus supporting the hypothesis of an IRES-dependent translational activation. However, it should be noted that we have not studied the other proteins reported to sequester eIF-4E.
In this report, we demonstrate that, in addition to a density-dependent decrease in the FGF-2 mRNA level, expression of the CUG-initiated FGF-2 isoforms is down-regulated whereas cell density increases in primary HSFs. This translational regulation is lost on cell transformation resulting in constitutive expression of all of the FGF-2 isoforms. In untransformed cells at low density and in transformed cells, expression of the HMW FGF-2 isoforms is regulated by cis-elements within the mRNA leader in correlation with the binding of several proteins to this mRNA leader. Furthermore, our data suggest that HMW FGF-2 induction at low density essentially involves a cap-dependent mechanism, whereas its constitutive expression in transformed cells mostly results from an IRES-dependent mechanism.
Our observation of a density-dependent regulation of the FGF-2 mRNA expression (Fig. 1 A) is in agreement with several studies described in the literature. Such a regulation has been observed in several normal cells, including human melanocytes, retinal pigment epithelium cells, umbilical vein endothelial cells, and astrocytes (39, 43). According to these reports, the cell density-dependent regulation occurs at the transcriptional level (43) or at the levels of mRNA polyadenylation and stability (39, 40).
In this study, we provide initial evidence that the density-dependent regulation of FGF-2 expression also occurs at the translational level (Fig. 1). Furthermore, it is not a global regulation of translation initiation but a modulation of the use of alternative CUG start codons. Moffett et al. (43) described an intense nuclear FGF-2 immunoreactivity in astrocytes growing at low cell density. This is in agreement with the induction of the nuclear HMW FGF-2 isoforms observed in our experiments. In the body, where most cells are necessarily at high density, such induction of HMW FGF-2 expression may occur in certain injured tissues in which the wound-repair process involves a transient cell proliferation. An intriguing question is that despite the strong down-regulation of FGF-2 mRNA accumulation, the global FGF-2 expression does not decrease in nHSFs. This contrasts with total FGF-2 protein expression following the mRNA level in astrocytes (43) and may reflect differences in the translational control process according to the cell type.
In contrast to that observed in normal cells, the density-dependent regulation of FGF-2 mRNA expression is not so clearly effective in transformed cells (this regulation is lost in tranformed cells such as U251MG glioma; Ref. 43), whereas it still occurs in human U87MG astrocytome and renal carcinoma cells (41, 42). The translational control shown here brings a new insight to this discrepancy; in fact, the key parameter associated with cell transformation may not be the global FGF-2 expression but the level of CUG-initiated forms expression, which cannot be controlled at the mRNA level but only at the translational level. Consistent with this are our previous observations—the CUG-initiated forms are mainly expressed in cancerous cell lines at confluence (23)—and the present observation—expression of these isoforms is constitutive in cancerous cells lines and becomes constitutive in Tag-transformed HSFs (Figs. 2 and 3). In fact, the deregulation of HMW FGF-2 translation could represent a critical step in the acquisition of the transformed phenotype. However, this notion must be restricted to certain cellular types expressing low levels of CUG-initiated FGF-2 (23) inasmuch as these isoforms can be efficiently expressed in some normal tissues (21).
The FGF-2 mRNA, capped as are all cellular mRNAs, has two alternative cap-dependent and IRES-dependent mechanisms for its translational regulation: (a) HMW FGF-2 induction has been observed in response to stress, probably in an IRES-dependent manner (23); and (b) overexpression of eIF-4E is also able to activate the efficiency of at least the upstream CUG of rat FGF-2 mRNA (35). The results of Fig. 5 demonstrate that the density-dependent regulation of HMW FGF-2 expression is not IRES-dependent. Furthermore, high cell density coincides with the disappearance of eIF-4E and dephosphorylation of 4E-BP1. This suggests that a double mechanism locks the cap-dependent translation at high density and that HMW FGF-2 expression according to density simply follows the eIF-4E availability in cells.
Whereas CUGs utilization is not IRES-driven in low-density nHSFs, AUG utilization incontestably occurs by internal ribosome entry (Fig. 6 B). These data strongly suggest the presence of a second IRES in the FGF-2 mRNA which would drive the AUG-initiated form expression exclusively. This second IRES does not seem to be influenced either by stress or by cell density, or by the mRNA level. Although IRESs are present in the mRNAs of several growth factors, proto-oncogenes, or other regulatory proteins, none of them contains two of these translation-regulatory elements. However, except for c-myc mRNA that shows an alternative initiation of translation, the other mRNAs code for only one translation product. As regards FGF-2 mRNA, the presence of two IRESs would allow the cell to modulate independently the expression of the CUG- and AUG-initiated isoforms.
The last question to be addressed here is the mechanism governing FGF-2 translation in transformed cells. The literature describes a correlation between cancer and eIF-4E overexpression, which suggests that the deregulation of regulatory proteins depends on a cap-dependent mechanism (35, 50, 52). However, translational deregulation cannot always be explained by the cap-dependent hypothesis. For instance, in Bloom’s syndrome cells, c-myc is overexpressed in an eIF-4E independent-mechanism, probably mediated by the IRES present in the messenger (29, 30, 53). The data obtained here with FGF-2 show that abnormal translational activation in transformed cells results from the contribution of an IRES-dependent mechanism (Fig. 5,C and Fig. 6). However, the two hypotheses are not exclusive inasmuch as eIF-4E levels have been shown to control CUG codons utilization of the rat FGF-2 mRNA (35). The cross-linked p110, specific to transformed cells, may be involved in IRES activation (Fig. 4). Constitutive binding of such a factor leading to constitutive expression of HMW FGF-2 would result in cell transformation because of the ability of HMW FGF-2 to act as an intracrine factor to stimulate cellular proliferation (18). The identification of such a factor may provide some insight into the oncogenesis process.
In conclusion, the double translational unlocking in transformed cells probably also concerns other genes, for instance, c-myc or the platelet-derived growth factor (c-sis), whose expression is tightly controlled in normal cells by the presence of highly structured and IRES-containing leaders in their mRNAs (27, 29, 30). Thus, these data once again support the increasingly apparent relationship between abnormal translation control of gene expression and cancer.
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.
Supported by Grants from the Association pour la Recherche contre le Cancer, the Ligue Nationale contre le Cancer, the Conseil Régional Midi-Pyrénées, and the European Union BIOTECH program. B. Galy had a fellowship from the Ligue Nationale contre le Cancer.
The abbreviations used are: FGF, fibroblast growth factor; HMW, high molecular weight; HSF, human skin fibroblast; nHSF, normal HSF; tHSF, transformed HSF; IRES, internal ribosome entry site; RT-PCR, reverse transcription PCR; CAT, chloramphenicol acetyl transferase; Tag, SV40 large T antigen.
We thank T. Levade, F. Bayard, and S. Vagner for helpful discussions, N. Bajeon for excellent technical assistance, R. Couret and D. Paris for illustrations, and D. Warwick for English proofreading. We thank J. Feunteun (Centre National de la Recherche Scientifique URA 1967, Villejuif, France) for the plasmid pAS, A. C. Gingras and N. Sonenberg (McGill University, Montréal, Canada) for anti-eIF4E and anti-4E-BP1 antibodies, and J. Bonafé (CHU Rangueil, Toulouse, France) for primary HSFs.