Expression of Human Fibroblast Growth Factor 2 mRNA Is Post-transcriptionally Controlled by a Unique Destabilizing Element Present in the 3′-Untranslated Region between Alternative Polyadenylation Sites*

Fibroblast growth factor 2 (FGF-2) belongs to a family of 18 genes coding for either mitogenic differentiating factors or oncogenic proteins, the expression of which must be tightly controlled. We looked for regulatory elements in the 5823-nucleotide-long 3′-untranslated region of the FGF-2 mRNA that contains eight potential alternative polyadenylation sites. Quantitative reverse transcription-polymerase chain reaction revealed that poly(A) site utilization was cell type-dependent, with the eighth poly(A) site being used (95%) in primary human skin fibroblasts, whereas proximal sites were used in the transformed cell lines studied here. We used a cell transfection approach with synthetic reporter mRNAs to localize a destabilizing element between the first and second poly(A) sites. Although AU-rich, the FGF-2-destabilizing element had unique features: it involved a 122-nucleotide direct repeat, with both elements of the repeat being required for the destabilizing activity. These data show that short stable FGF-2 mRNAs are present in transformed cells, whereas skin fibroblasts contain mostly long unstable mRNAs, suggesting that FGF-2 mRNA stability cannot be regulated in transformed cells. The results also provide evidence of a multilevel post-transcriptional control of FGF-2 expression; such a stringent control prevents FGF-2 overexpression and permits its expression to be enhanced only in relevant physiological situations.

teins (1)(2)(3)(4)(5). FGF-2 is produced in various cell types and tissues and has many biological roles. It is involved in embryogenesis and morphogenesis, especially in the nervous system and bone formation (6,7). FGF-2 is a major angiogenic factor, playing a crucial role in wound healing and in cardiovascular disease (8). It is also involved in cancer pathophysiology, notably in tumor neovascularization coupled with intrinsic oncogenic potential (9 -11).
FGF-2 expression is regulated both at the transcriptional level and more especially at the translational level. Its mRNA constitutes a complex example of alternative initiation of translation with five start codons that include four CUG codons (12)(13)(14). The five FGF-2 isoforms resulting from the alternative initiation process have different localizations and functions within the cell (10, 14 -17). The regulation of their expression is also different: the largest 34-kDa isoform is exclusively translated in a cap-dependent manner, whereas the other isoforms are translated by a process of internal ribosome entry mediated by an internal ribosome entry site located between the first and second CUG codons (14).
In addition to its GC-rich 5Ј-region that contains an internal ribosome entry site as well as other elements able to regulate alternative initiation of translation (18,19), the 6775-nt-long FGF-2 mRNA exhibits a very uncommon feature: it is 90% composed of untranslated regions due to the presence of a huge 3Ј-untranslated region. This AU-rich 5823-nt-long 3Ј-untranslated region is not only the longest 3Ј-UTR described to date, but also contains eight potential polyadenylation sites. The existence of more than three FGF-2 mRNA species reported in the literature (from 1 to 7 kilobases) would suggest that several of these poly(A) sites can be functional (20,21).
No information was available from the literature about the putative role of the huge and multiple-sized 3Ј-UTR present in the FGF-2 mRNA. However, the 3Ј-UTRs of many mRNAs are now known to play a pivotal role in the post-transcriptional regulation of gene expression by controlling mRNA subcellular localization, stability, or translation initiation (22). Interestingly, such regulation essentially concerns the messengers coding for proteins that have potent growth or developmental effects or whose function is temporarily restricted, for instance, during a particular phase of the cell cycle. The genes expressing unstable mRNAs include proto-oncogenes, cytokines, growth factors, hormones, receptors, and cell cycle-regulated genes (for review, see Ref. 23). The stability of most of these mRNAs is regulated by AU-rich elements (AREs) present in the 3Ј-UTR (23).
The unusual length and AU-rich composition of the FGF-2 mRNA 3Ј-UTR, together with the existence of eight potential  1 The abbreviations used are: FGF-2, fibroblast growth factor 2; nt, nucleotide(s); UTR, untranslated region; ARE, AU-rich element; RT-PCR, reverse transcription-polymerase chain reaction; CAT, chloramphenicol acetyltransferase; GM-CSF, granulocyte/macrophage colonystimulating factor; PBS, phosphate-buffered saline. alternative length-modifying polyadenylation sites, prompted us to determine its regulatory role in FGF-2 isoform expression. In this report, we show that poly(A) site utilization varies with cell type, and we identify a destabilizing element between the first and second polyadenylation sites of FGF-2 mRNA. These observations suggest that regulation of FGF-2 expression occurs at the level of RNA stability, conditioned by use of the poly(A) site.
pCAT-dest⌬A and pCAT-dest⌬S were obtained by digestion of plasmid pCAT-dest by AflIII and SpeI, respectively, followed by religation (see Fig. 4). pCAT-A8⌬dest was obtained by digestion of pCAT-A8 by AvrII plus BamHI, followed by Klenow treatment and religation.
In Vitro Transcription-DNAs were obtained either by plasmid lin-earization or by PCR amplification using T3 tail-containing primers (for antisense RNAs). Transcription was performed with T7 or T3 RNA polymerase using transcription kits provided by Ambion Inc.: the mMESSAGE mMACHINE TM kit was used for capped mRNAs, and the MAXIscript TM or the MEGAscript TM kit for uncapped and/or labeled RNAs. RNA transcripts were quantitated by absorbance at 260 nm and ethidium bromide staining on agarose gel, and their integrity was verified. The templates for T3 transcription of antisense RNAs against the dest element were obtained by PCR using primers RTA2 and PA2revT3 (Table I), followed by AflII, SpeI, or AvrII digestion (AS1, AS2, and AS3, respectively) (see Fig. 4). Primary Human Skin Fibroblast Cultivation-A piece of human skin obtained from the plastic surgery department of Rangueil Hospital (Toulouse, France) was washed for 15 min at 20°C in six antibiotic/PBS baths (15 ml): 1) penicillin/streptomycin (Life Technologies, Inc.) at a dilution of 1:50; 2) gentamycin (Life Technologies, Inc.) at a dilution of 1:750; 3) Bactrim (80 mg/ml; Roche Molecular Biochemicals) at a dilution of 1:660; 4) Cyflox (2 mg/ml; Bayer) at a dilution of 1:1000; 5) Fortum (333 mg/ml; Glaxo Wellcome) at a dilution of 1:1000; and 6) amphotericin (Life Technologies, Inc.) at a dilution of 1:100. The skin was cut into 1-mm 2 pieces, which were left to dry in a Petri dish for 10 min. Dulbecco's modified Eagle's medium plus 10% fetal calf serum and penicillin/streptomycin were added, and cultivation was pursued for 3-4 weeks until fibroblasts grew from the skin pieces, before trypsinization and seeding into new dishes (P0). The fibroblasts could then be used for seven passages.
Cell Transfection by RNA-COS-7, HeLa, and SK-Hep-1 cells (see Ref. 24) were transfected by the DMRIE-C method (Life Technologies, Inc.). Briefly, 100 l of serum-free medium containing 2 pmol of RNA (1-10 g depending on RNA size) was mixed with 100 l of serum-free medium containing 10 l of DMRIE-C reagent. After addition of 0.8 ml of serum-free medium, the mixture was added to PBS-washed cells and incubated for 14 h at 37°C.
Primary human skin fibroblasts were electroporated using a Bio-Rad apparatus. PBS-washed cells were scraped, resuspended in 10% fetal calf serum-containing medium, and centrifuged at 1000 rpm for 10 min. The pellet was washed with PBS and centrifuged twice. Cells were resuspended in serum-free medium at a final concentration of 2.5 ϫ 10 6 cells/ml. 400 l of cell suspension was mixed with 2 pmol of RNA (1-10 g) and transferred to a 4-mm electroporation cuvette. An electric shock of 260 V/950 microfarads was applied, and then the cells were seeded in medium-containing dishes and incubated at 37°C for 12-16 h.
Seven dishes (5 cm in diameter) were transfected with 1 g of RNA/dish to measure mRNA stability. After incubation for 2 h, the RNA-containing medium was removed; the cells were washed with PBS; and fresh medium was added. The cells were harvested either immediately or after increasing periods of time (7.5-360 min).
CAT activity determinations were performed as described previously (14). Luciferase activity was measured using the Promega luciferase assay system.
Purification of Cellular RNA-The cell monolayers (5 ϫ 10 6 cells) were lysed in 0.5 or 1 ml of RNABle (Eurobio) in 60-or 90-mm dishes, respectively. Extraction was continued by adding 100 l of chloroform/ 0.5 ml of RNABle and precipitating the aqueous phase with 1 volume of isopropyl alcohol (15 min at 4°C). After centrifugation, the pellets were rinsed with 75% ethanol. The RNA was quantitated by measuring the absorbance at 260 nm and checked for integrity by electrophoresis on agarose gel and ethidium bromide staining.
Quantitative Reverse Transcription-Polymerase Chain Reaction-The cDNAs were synthesized using the Superscript TM preamplification system (Life Technologies, Inc.) according to the manufacturer's instructions. The reverse transcription reaction was carried out using 1 g of total RNA and 50 ng of random hexamers in a final volume of 20 l. Variable amounts of internal standard RNAs synthesized from the pKS-PA1⌬, pKS-PA2⌬, and pKS-PA8⌬ plasmids (see above) were added to the reactions as described previously (24) to quantify the different regions of the FGF-2 mRNA 3Ј-UTR.
PCR was performed with the primer couple RTA1/PA1rev, RTA2/ PA2rev, or RTA8/PA8rev, hybridizing upstream from the first, second, or eighth poly(A) site, respectively. The resulting fragments corresponded as follows: nt 737-1040, 3183-3441, and 6455-6763 of the FGF-2 cDNA, respectively. The PCR reactions were carried out using 0.5 units of Goldstar Taq DNA polymerase (Eurogentec) in a final volume of 50 l, with variable amounts of cDNA (1 l or less). The reaction was performed on a TrioThermoblock apparatus (Eurogentec) under the following conditions: 94°C for 3 min and then 30 cycles of 94°C for 30 s, 63°C for 1 min, 72°C for 1 min, and finally 72°C for 5 min. Amplification results (one-fifth of the reactions) were analyzed on 6% polyacrylamide gels (Tris borate/EDTA), followed by ethidium bromide staining. The intensity of the ethidium bromide luminescence was measured by image acquisition on a UV max apparatus (OSI), followed by image treatment with NIH Image software as described previously (24).

RESULTS
Alternative Polyadenylation of the FGF-2 mRNA 3Ј-UTR Is Regulated According to Cell Type-Poly(A) site utilization was analyzed in different cell types by quantitative RT-PCR. Total RNAs were purified from three transformed simian or human cell lines (COS-7, HeLa, and SK-Hep-1 cells) as well as from primary human skin fibroblasts. RT-PCR was performed as described in our previous report (24) with internal standard RNAs and oligonucleotide primer couples specific to the region upstream from the first (A1), second (A2), and eighth (A8) poly(A) sites, respectively ( Fig. 1A; see "Experimental Procedures"). We were able from the results to divide the FGF-2 mRNAs into three groups of messengers: (i) cleaved at A1, (ii) cleaved between A2 and A7, and (iii) cleaved at A8. As shown in Fig. 1B, COS-7 and HeLa cells exhibited 100 and 92% short mRNAs cleaved at A1, respectively, whereas skin fibroblasts exhibited 95.5% long mRNAs cleaved at A8. SK-Hep-1 cells showed a more heterogeneous profile, with 28, 49, and 23% of each mRNA species, respectively. The presence of 100% A1cleaved mRNA was also checked by Northern blotting in COS-7 cells, thus ruling out the possibility of a bias in our quantification due to 3Ј-UTR species variability between monkeys and humans (data not shown). These data clearly indicate a regulation of the use of the poly(A) sites favoring the appearance of the longest 5823-nt-long 3Ј-UTR in skin fibroblasts and of the shortest and intermediary 3Ј-UTRs in the three transformed cell lines.
An RNA-destabilizing Element Is Present between the First and Second Polyadenylation Sites of the FGF-2 mRNA-The observed variations in the 3Ј-UTR length due to cell typespecific alternative polyadenylation prompted us to look for regulatory elements in the 5823-nt-long 3Ј-UTR of FGF-2 mRNA. To avoid interference from the alternative polyadenylation process and to analyze the expression of one mRNA species at a time, cell transfection was performed using in vitro transcribed, capped, and polyadenylated mRNAs (25). This procedure has been previously validated for mRNA half-life analysis in a study of the GM-CSF ARE (26). COS-7 cells were transfected with CAT mRNAs bearing five of the different FGF-2 mRNA 3Ј-UTRs (measuring 83, 2390, 3073, 4446, and 5823 nt, respectively), fused or not to the leader region of FGF-2 mRNA (Fig. 2A). CAT mRNA expression was first analyzed by CAT activity measurement (Fig. 2B). CAT-A1 mRNA with the shortest 3Ј-UTR was efficiently expressed, in comparison to the CAT-A0 control, whereas CAT expression from A2, A3, A4 and A8 mRNAs was very inefficient. The same profiles were obtained in the presence of the FGF-2 mRNA leader (Fig. 2B, 5ЈCAT series). This suggested the existence of an inhibitory element in the FGF-2 mRNA 3Ј-UTR between the first and second poly(A) sites.
The half-lives of the different CAT mRNAs were measured to determine whether this inhibitory element was an RNA-destabilizing element or a translational silencer (Fig. 2C). This analysis revealed a CAT-A1 mRNA half-life (110 min) similar to the CAT-A0 mRNA half-life (126 min), whereas a drastic shortening of the half-life was measured for CAT-A2, CAT-A3, CAT-A4, and CAT-A8 mRNAs (26,25,18, and 14 min, respectively). This demonstrated that the inhibitory element located between the first and second poly(A) sites was a destabilizing element. The same destabilizing effect was observed in the presence of the FGF-2 mRNA 5Ј-region, even though the 5ЈCAT mRNAs were slightly more stable than their CAT counterparts (Fig.  2C).
CAT mRNAs half-lives were also measured in skin fibroblasts and HeLa and SK-Hep-1 cells (Fig. 2, D-F). The destabilizing element clearly shortened the mRNA half-life by 4-fold in all these cell types, as in COS-7 cells, thus indicating that the activity of the FGF-2 mRNA-destabilizing element was not cell type-specific.

The FGF-2 mRNA-destabilizing Element Involves a 122-nt Tandem Repeat Located Upstream from the Second Poly(A)
Site-Prior to further characterization of the FGF-2 mRNA-destabilizing element, the possibility of a nonspecific destabilizing effect due to the large size of the 3Ј-UTR was considered by analyzing the effect of the 5823-nt-long 3Ј-UTR in a reverse orientation (Fig. 3A, CAT-3Јinverted) on mRNA stability. The results showed that the CAT-A8-inverted mRNA, like the control CAT-A0, was four times more stable than the CAT-A8 mRNA, both in COS-7 cells and in skin fibroblasts, so the hypothesis of nonspecific effects generated by the unusual length of the 3Ј-UTR could be ruled out.
The efficiency of the FGF-2 mRNA-destabilizing element was also compared with that of the well characterized GM-CSF ARE. COS-7 cells and skin fibroblasts were transfected with CAT mRNAs bearing the 3Ј-UTR of the GM-CSF mRNA, with or without its ARE (Fig. 3A, CAT-3ЈGM-CSF and CAT-3ЈGM-⌬AU). Measurement of the mRNA half-life showed that the GM-CSF ARE was able to reduce the CAT mRNA half-life, under our conditions, by a factor of 3 (Fig. 3A), in accordance with the report of Rajagopalan and Malter (26). The FGF-2 mRNA-destabilizing element, which could shorten the mRNA half-life by a factor of 4, could thus be considered an efficient destabilizing element.

FIG. 1. RT-PCR quantification of different FGF-2 mRNA 3-UTRs.
A, shown is a schema of the 5Ј-untranslated and coding regions of FGF-2 mRNA. The complete 6775-nt-long FGF-2 mRNA with positions of the eight poly(A) sites and the fragment obtained from RT-PCR quantification is shown below. B, quantitative RT-PCR was performed using total RNAs from COS-7, HeLa, and SK-Hep-1 cells and skin fibroblasts as described under "Experimental Procedures," but using three different primer couples, leading to the amplification of fragments I-III (each with its internal standard; see "Experimental Procedures") corresponding to regions upstream from poly(A) sites A1, A2, and A8. The percentage of mRNAs ending at A1, A2-A7, and A8 was calculated from the quantified amounts of fragments I-III: A1 mRNAs ϭ fragment I-II; A2-A7 mRNAs ϭ fragment II-III; A8 mRNAs ϭ fragment III.
The destabilizing element was then more precisely localized by a deletion approach (Fig. 3B). RNA transfection of human skin fibroblasts showed that removal of the 208 nt upstream from the second poly(A) site abolished the destabilizing effect (Fig. 3B, CAT-A2⌬1). However, this 208-nt-long fragment, when directly inserted 3Ј of the CAT gene, was not sufficient to generate mRNA destabilization (CAT-dest208). In contrast, a fragment corresponding to the 334 nt upstream from the second poly(A) site was sufficient to induce the destabilizing effect (Fig. 3B, CAT-dest334). Removal of this fragment (called DEST hereafter) from the complete FGF-2 3Ј-UTR also abolished RNA destabilization (Fig. 4C, A8⌬dest).
Examination of the DEST nucleotide sequence revealed the existence of two 122-nt-long direct repeats with 88% identity, 79% of which consisted of A and U residues. However, each of these repeats contained only a single AUUUA motif (Fig. 4A), whereas AREs described in proto-oncogene and lymphokine mRNAs always contain reiterated AUUUA motifs (23). Deletions were therefore performed within the DEST element, and RNA transfection was carried out in skin fibroblasts as described above. The deletion of a 57-nt-long AflIII fragment corresponding to the 3Ј-part of the upstream repeat did not affect RNA destabilization (Fig. 4, B and C, dest⌬A), whereas the deletion of a 126-nt-long SpeI fragment, which removed a complete repeat, abolished the destabilizing effect (Fig. 4, B  and C, dest⌬S).

FIG. 2. Expression and half-lives of transfected CAT mRNAs.
Capped and polyadenylated CAT mRNAs bearing different FGF-2 3Ј-UTRs were transcribed in vitro using T3 RNA polymerase. Different cell types were transfected with 2 pmol of each mRNA species. A, mRNA schemata. CAT-A0 is the control mRNA devoid of the 3Ј-UTR. CAT-A1, CAT-A2, CAT-A3, CAT-A4, and CAT-A8 have 3Ј-UTRs of 83, 2390, 3073, 4446, and 5823 nt, respectively. 5Ј-CAT mRNAs bear, in addition to the different 3Ј-UTRs A0 to A8, a fusion of the FGF-2 mRNA 5Ј-region with CAT sequence (giving rise to CAT fusion proteins initiated at the FGF-2 start codons). B, COS-7 cell transfection by the different constructs. CAT activity of the cell extracts was measured 14 h after transfection (see "Experimental Procedures") and is shown by histograms. C, mRNA half-life measurements in COS-7 cell extracts. For each mRNA species, seven dishes were transfected in parallel to allow cell harvesting at different times after RNA removal. Total RNAs were prepared from the transfected cells and analyzed by Northern blotting with a 32 P-labeled probe corresponding to the CAT coding sequence. D-F, to half-life measurements of CAT mRNAs transfected in human skin fibroblasts and HeLa and SK-Hep-1 cells, respectively, as described under "Experimental Procedures." The autoradiography (the period before harvesting is indicated for each point) and the half-life quantification (histograms and values) are given to the right of the name of each chimeric mRNA for each transfected cell type. These were obtained using a PhosphorImager (Molecular Dynamics, Inc.).
An alternative strategy was to cotransfect CAT mRNA with an excess of antisense RNAs targeting different parts of the DEST element. The results given in Fig. 4D show that an antisense RNA targeting the complete DEST element was able to prevent RNA destabilization, whereas an antisense RNA directed against the CAT sequence had no effect (Fig. 4, B and D, AS3 and ASCAT). Furthermore, shorter antisense RNAs directed against part or all of the downstream repeat were also able to abolish the destabilizing effect (AS1 and AS2).
These data show that the DEST element involves the two AU-rich direct repeats. Apparently both repeats, except for the 3Ј-part of the upstream repeat, are required and are responsible for the entire destabilizing effect of the long FGF-2 mRNA 3Ј-UTR. DISCUSSION We show in this report that the length of the FGF-2 mRNA 3Ј-UTR is conditioned by a process of alternative polyadenylation, specific to the cell type. The proximal poly(A) sites seem to be preferentially used in three transformed cell lines, whereas the eighth poly(A) site is mostly used in primary skin fibroblasts, giving rise to the huge 5823-nt-long 3Ј-UTR. This regulation of alternative polyadenylation has consequences for the regulation of FGF-2 expression, as a destabilizing element, corresponding to two AU-rich tandem repeated sequences, has been localized between the first and second poly(A) sites.
No study of FGF-2 mRNA half-life had been carried out up to now because of the technical difficulty of detecting the mRNA on Northern blots. Furthermore, the presence of several mRNA species resulting from alternative polyadenylation rendered interpretation very complex and prevented the expression of homogenous mRNAs after DNA transfection. The development and optimization of the RNA transfection procedure described here, novel for primary cells, were crucial in studying the expression of a single mRNA species in the absence of alternative polyadenylation. We provide the first evidence that a member of the everincreasing FGF family is regulated at the level of mRNA stability by a cis-acting element, the presence of which is controlled by alternative polyadenylation. Regulation of FGF-2 mRNA stability has been reported only in Xenopus, in which the gfg antisense RNA that induces FGF-2 mRNA destabilization by a process of RNA editing is involved (27). However, the mammalian gfg mRNA, although present in cells, has never been shown to affect FGF-2 mRNA stability (28). Our results suggest that mammalian FGF-2 mRNA stability is regulated by a process similar to that controlling other cytokines and involving AU-rich elements.
The AU-rich FGF-2 mRNA-destabilizing (DEST) element described here is unique. It does not contain the tandem AUUUA motifs described for most proto-oncogenes, cytokines, and growth factor mRNAs or the long U-rich region enhancing the destabilizing effect of AREs found in c-fos mRNA, for example  (23). In fact, sequence comparison revealed homology of the FGF-2 DEST element to the interleukin-2 and vascular endothelial growth factor AREs, which are also unusual (29,30). However, neither of these destabilizing elements contains the 122-nt-long direct repeat observed for FGF-2 (Fig. 4). Evidence is provided here that the presence of AUUUA motifs is not sufficient to destabilize FGF-2 mRNA: the 3Ј-UTR contains 15 AUUUA motifs outside the DEST element, which do not influence mRNA decay (Fig. 4) (13). Furthermore, the two AUUUA motifs located in the 5Ј-part of each repeat in the DEST element are unable to generate RNA instability by themselves and require the presence of tandem repeats for RNA destabilization.
This raises the question of protein involvement in the regulation of FGF-2 ARE activity. The ARE-mediated destabilization mechanism, although still mechanistically unclear, involves RNA/protein interactions. At least 10 ARE-specific binding proteins have been identified to date, with different affinities for specific RNA sequences (23). Nine of them are able to bind to the c-myc ARE, and six of them are able to bind to poly(U); two of these proteins, AU-B and AU-C, are unable to bind to the c-myc ARE or poly(U). Interestingly, the AU-B protein binds to the interleukin-2 ARE (31). Furthermore, the inactivation of the vascular endothelial growth factor ARE, involved in hypoxia-induced mRNA stabilization, is correlated with the binding of three proteins with molecular masses of 17, 28, and 32 kDa (30). This supports the hypothesis that the FGF-2 DEST element could also be regulated by cellular protein binding. The requirement of both elements of the repeat for RNA destabilization suggests that the potential regulatory protein(s) has two binding sites in the DEST element; one possibility would be the cooperative binding of two protein molecules, only active as a dimer, to the DEST element.
There are few examples in the literature of regulation involving processes of alternative polyadenylation coupled with RNA stability. One interesting case of inducible stability dependent on polyadenylation is provided by the glutaminase mRNA. This mRNA is expressed with two forms of 3Ј-UTR, the longer of which contains a pH-responsive stability element (32). Vascular endothelial growth factor mRNA stability is regulated by a hypoxia-controlled ARE located between two alternative polyadenylation sites (33). The FGF-2 DEST element, located between the first and second poly(A) sites, could ensure a rapid turnover of the FGF-2 mRNA in cells presenting polyadenylation at A2 or downstream. The coupled polyadenylation and destabilization of FGF-2 mRNA in such a case could enable the cell to provide a rapid regulation change in response to exogenous stimuli.
An interesting hypothesis is provided by the results in Fig. 1, which show that the shortest mRNA (cleaved at A1) is present in the three transformed cell lines (28 -100%), but constitutes only 4.5% of the FGF-2 mRNA in primary skin fibroblasts.  Fig. 3). The positions of the antisense RNAs used below are shown together with the SpeI, AflII, and AvrII restriction sites used either to obtain internal deletions or to cut the DNA templates before transcription of AS1, AS2, and AS3 RNAs, respectively. C, skin fibroblasts were transfected with CAT mRNAs containing either the complete 334-nt fragment (CAT-dest334) or deletions of 57 and 126 nt in this fragment (CAT-dest⌬A and CAT-dest⌬S, respectively). Transfections were also performed with a CAT-A8-derived mRNA in which the 334-nt fragment was deleted from the long FGF-2 mRNA 3Ј-UTR. Half-life measurements after transfection were obtained as described in the legend to Fig. 2. D, dest334 mRNA was hybridized with AS1, AS2, or AS3 RNA (in 10-fold excess) before transfection. RNA-RNA hybridization was obtained by denaturing the mixture of dest334 mRNA with antisense RNA for 2 min at 95°C in H 2 O and adding culture medium before returning slowly to 20°C. The RNA hybrids were used for skin fibroblast RNA transfection as described for B. Half-life measurements obtained as described above are shown on the right.
These results suggest that the FGF-2 mRNA can be destabilized in skin fibroblasts, but is stable in transformed cells, where it is devoid of the destabilizing element. We have shown in a previous report that FGF-2 expression is translationally regulated in normal skin fibroblasts, but constitutively expressed in transformed cell lines (SK-Hep-1 and HeLa) as well as in skin fibroblasts transformed by SV40 large T antigen (34). The present data once again suggest that FGF-2 expression cannot be regulated in transformed cells, a feature that could be a cause or a consequence of the transformed phenotype.
All these observations indicate that although FGF-2 expression undergoes transcriptional control, it is mostly regulated post-transcriptionally at three levels: mRNA polyadenylation, stability, and translation. The existence of several post-transcriptional regulations for a given mRNA renders the control of FGF-2 expression very stringent, which permits its expression only under relevant conditions. It also provides the possibility for untransformed cells to very rapidly modulate the level of FGF-2 expression in response to exogenous stimuli. The unlocking of these regulations may be one of the parameters responsible for cell transformation.