Poly(A)-binding protein-interacting protein 2, a strong regulator of vascular endothelial growth factor mRNA.

Expression of vascular endothelial growth factor (VEGF) is tightly regulated, particularly at the level of its mRNA stability, which is essentially mediated through the 3'-untranslated region (3'-UTR) of VEGF mRNA. To identify new protein partners regulating VEGF mRNA stability, we screened a cDNA expression library with an RNA probe corresponding to the entire VEGF mRNA 3'-UTR. We identified the "poly(A)-binding protein-interacting protein 2" (PAIP2) as a new VEGF mRNA 3'-UTR interacting protein. By RNA electromobility shift assays, we showed that PAIP2 binds to two distinct regions of a domain encompassing base 1 to 1280 of the VEGF 3'-UTR. Such in vitro interaction was confirmed using cell extracts in which PAIP2 expression is induced by tetracycline (Tet-on cells). Moreover, we demonstrated by RNA affinity purification as well as by ribonucleoprotein complexes immunoprecipitation, that PAIP2 interacts with VEGF mRNA in vivo. Using an in vitro RNA degradation assay, the half-life of VEGF 3'-UTR was found to be increased by overexpressing PAIP2. PAIP2 stabilizes endogenous VEGF mRNA in Tet-on cells, leading to increased VEGF secretion. Moreover, RNAi-mediated knock-down of PAIP2 significantly reduces the steady-state levels of endogenous VEGF mRNA. We also showed, by in vitro protein-protein interactions and co-immunoprecipitation experiments, that PAIP2 interacts with HuR, an already known VEGF mRNA-binding protein, suggesting cooperation of both proteins for VEGF mRNA stabilization. Hence, PAIP2 appears to be a crucial regulator of VEGF mRNA and as a consequence, any variation in its expression could modulate angiogenesis.

In the absence of sufficient oxygen delivery, organs induce production of angiogenic factors that function to recruit new blood vessels to hypoxic tissues (1,2). Vascular endothelial growth factor is one of the most important regulators of angiogenesis as inactivation of only one allele of its gene leads to embryonic lethality in mice (3). Expression of VEGF 1 is tightly regulated. At the transcriptional level, hypoxia-inducible factor-1 is recruited under hypoxia to the VEGF promoter (4). Oncogenic transformation also induces activation of VEGF transcription by recruiting AP-2 and Sp1 transcription factors to the proximal region of the promoter (5). VEGF expression is also regulated post-transcriptionally by stabilization of its mRNA (6 -9). Previous experiments have shown that the VEGF mRNA 3Ј-UTR contains nine copies of the nonameric consensus for AU-rich element (ARE) present in the 3Ј-UTR of many labile mRNAs (9). Consistent with their presence, the VEGF 3Ј-UTR can confer instability to a reporter mRNA in a cellular model (9 -11). Precedent reports have demonstrated the presence of binding sites for hypoxia or stress inducible proteins (7,9,11). Even if several proteins were shown to interact with the VEGF mRNA 3Ј-UTR, only two proteins have been identified: HuR, a member of the ELAV family (8) and the ribonucleoprotein hnRNPL (12). Moreover, the molecular mechanisms underlying VEGF mRNA stability, which is modified under physiological and pathological situations, remain poorly understood. The aim of our study was to identify VEGF mRNA protein partners that are essential for the modulation of VEGF mRNA stability. Our screening experiments identified 15 independent clones that interact with the VEGF mRNA 3Ј-UTR. The most strongly interacting clone was the poly(A)-binding protein-interacting protein 2 (PAIP2). At the same time, PAIP2 was identified as a translational repressor by using bicistronic vectors (13,14). Our study consisted in further analyzing the role of PAIP2 as a modulator of VEGF mRNA expression.
Cell Lines and Culture Conditions-The A431, C6, HeLa, LoVo, PROb, and Ras Val-12 cells (15) were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 7.5% heat-inactivated fetal calf serum and containing penicillin (50 units/ml) and streptomycin sulfate (50 g/ml). S19-R443 cells are a derivative of CCL39 Chinese hamster lung fibroblasts, which stably express a tetracycline repressor. They were stably transfected with an expression vector encoding a tetracycline inducible myc-tagged PAIP2 (cloned into the pCDNA4/TO/ myc-His A vector, Invitrogen) and cultured as described (16). Induction of myc-tagged PAIP2 was obtained by stimulating cells with tetracycline (1 g/ml).
cDNA Library Screening-We screened a mouse testis cDNA library constructed in the ZAP Express vector (Stratagene). Nitrocellulose filters soaked in 10 mM isopropyl-1-thio-␤-D-galactopyranoside were applied to the surface of plates previously inoculated with bacteriophages. Plates were then incubated overnight at 42°C. Filters were then peeled off the plates and immersed in a large volume of RNA binding buffer containing a final concentration of 10 mM HEPES (pH 7.5), 5 mM MgCl 2 , 50 mM KCl, 0.5 mM EGTA, 0.5 mM dithiothreitol, 10% glycerol, and 100 g/ml tRNA. After two washes in this binding buffer, radiolabeled RNA transcript (10 7 cpm) corresponding to the entire human VEGF mRNA 3Ј-UTR (NCB accession number AF024710), and heparin (5 mg/ml) were added to the filters. After incubation for 1 h with the probe at 30°C, RNase T1 was added to a final concentration of 1000 units/ml. The filters were then washed five times in RNA binding buffer and autoradiographed. The process of screening and plating was repeated until a homogenous population of positive recombinant bacteriophage was obtained. By using the ExAssist helper phage, a pBK-CMV phagemid vector containing the cDNA of interest was excised from the ZAP Express vector according to the manufacturer's protocol.
Cloning of Full-length PAIP2 cDNA-First-strand cDNA was synthesized from 1 g of human colon carcinoma cell HT29 poly(A) ϩ RNA, using avian myeloblastosis virus reverse transcriptase with an oligo(dT) primer. This material was used as a template for polymerase chain reaction amplification. The following oligonucleotides derived from the human PAIP2 (NCB accession number NM_016480) were synthesized and used as primers for the polymerase chain reaction: oligonucleotide 1, 5Ј-GTG-GATCCAAAGATCCAAGTCGCAGCAGTACTAGCCC-3Ј; and oligonucleotide 2, 5Ј-TAGTCGACTCAAATATTTCCGTACTTCACCCCAGG-3Ј. An aliquot of cDNA was amplified in a 50-l reaction volume with 200 ng of each primer, 200 M dNTPs, and 2.5 units of Ampli-Taq (Roche Applied Science) in buffer containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl 2 , and 0.001% gelatin. Polymerase chain reaction amplification was performed in a DNA thermal cycler (Biometra) using the following parameters: 30 s at 95°C, 1 min at 55°C, and 1 min at 72°C for 30 cycles, followed by an extra cycle with a 10-min extension step at 72°C. An expected fragment of 381 bp was analyzed by restriction enzymes and sequenced.
Plasmid Constructs-The 381-bp DNA fragment corresponding to the full-length PAIP2 cDNA was cloned into the pGEX-6P1 and pCMV-Tag3B vectors (Amersham Biosciences and Stratagene, respectively) within BamHI and SalI restriction sites. Both restriction sites were artificially added to the cDNA extremities during the PCR (BamHI site 5Ј and SalI site 3Ј). A DNA fragment of 454 bp was excised from the pCMV-Tag/PAIP2 vector with NotI and ApaI and subcloned into the corresponding sites of the pCDNA4/TO/myc-His A vector (Invitrogen). The full-length human HuR cDNA (NCB accession number BC003376) was introduced into the pCMV-Tag3B vector within EcoRI/XhoI sites. The plasmids containing the VEGF 3Ј-UTR used in RNA electrophoretic mobility shift assays (REMSA) and RNA degradation assays contain the rat VEGF 3Ј-UTR (NCB accession number U22372) (6). The full-length, NsiI, and StuI constructs have been previously described (6,11). To generate the ⌬NsiI construct (nucleotides 1276 -2211 of the VEGF 3Ј-UTR), the full-length construct (in pSP64 poly(A) vector, Promega) was digested with NsiI, blunted with T4 DNA polymerase, then digested with HindIII (site of the vector), blunted with Klenow and ligated. To generate the ⌬StuI construct (nucleotides 864 -1276 of the VEGF 3Ј-UTR), the NsiI construct was digested with HindIII and StuI, blunted with Klenow, and ligated.
Preparation of RNA-Cells were lysed in TRIzol reagent buffer (Invitrogen). RNAs were prepared according to the manufacturer's protocol. The mouse tissue RNAs were from Clontech. 20 g of RNA was used for Northern blot analysis and hybridized with the PAIP2 probe (381 bp) or the VEGF probe (640 bp) corresponding to the coding region of mouse VEGF (NCB accession number NM 009505). The 36B4 probe corresponds to the region comprised between bases 683 and 842 of the cDNA coding for the human acidic ribosomal phosphoprotein P0 (NCB accession number M17885) (17).
Preparation of Cytosolic Extracts, REMSA, and in Vitro RNA Degradation Assay-S19-R443/PAIP2 cells were stimulated or not with tetracycline (1 g/ml) for 4 h. Then, the S100 fraction of cytosolic proteins was prepared as described previously (6). RNA transcripts were synthesized from linearized pSP64 poly(A) 3Ј-UTR templates (Promega) (6) using the SP6 bacteriophage RNA polymerase. For REMSA experiments, radiolabeled RNA transcripts (100,000 cpm/reaction) were combined with GST fusion proteins or GST alone, or S100 extracts (5 g of cytosolic proteins) from cells stimulated or not with tetracycline, in a previously described binding buffer (6). The reaction mixture was incubated for 30 min at 30°C and treated for 15 min at room temperature with 100 units of ribonuclease T1 (Roche). When specific or nonspecific competitors were used, they were incubated for 15 min at 30°C with the proteins in binding buffer before the addition of the radiolabeled tran-scripts. The reaction mixtures were resolved on 5% native polyacrylamide gels in 0.5 ϫ Tris borate/EDTA buffer. Gels were dried and autoradiographed. For in vitro RNA degradation assays, polyadenylated and radiolabeled RNA transcripts (100,000 cpm/reaction) were incubated with 1 g of S100 extract from cells stimulated or not with tetracycline. At each time point, an aliquot of the reaction was taken and treated as previously described (11).
RNA Affinity Purification-2 g of in vitro transcribed polyadenylated RNA, corresponding to the NsiI probe or to the full-length Chinese hamster ERK1 mRNA 3Ј-UTR (18) were incubated with polystyrenelatex beads with dC 10 T 30 oligonucleotides covalently linked to the surface (Oligotex, Qiagen), in a binding buffer (20 mM Tris-HCl, pH 7.5, 1 M NaCl, 2 mM EDTA, and 0.2% SDS). After 2 washes in a wash buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1 mM EDTA), the RNAcoated beads were incubated with 100 g of S100 extracts from cells stimulated or not with tetracycline, in the RNA binding buffer used for REMSA experiments, supplemented with 1% bovine serum albumin. After incubation for 10 min at 30°C, the beads were extensively washed with RNA binding buffer. Laemmli sample buffer was then added to the beads. Proteins bound to the RNA probe were resolved on SDS-PAGE and subjected to Western blotting using the monoclonal anti-myc antibody described below.
Anti-PAIP2 and Anti-HuR Antibodies-Antibodies were generated by Eurogentec (Liege, Belgium), by injecting four times two rabbits with 100 g of a GST-PAIP2 fusion protein or a GST-HuR fusion protein. Sera were affinity purified using an EAH-Sepharose 4B column (Amersham Biosciences) to which GST-PAIP2 or GST-HuR was coupled. Specific IgG were then eluted with 10 mM glycine, pH 2.8, and neutralized with 20 mM Tris, pH 11.
Immunofluorescence-S19-R443/PAIP2 cells expressing inducible myc-tagged PAIP2 were plated on glass coverslips. After tetracycline induction for 12 h, cells were fixed with 3.3% paraformaldehyde at room temperature for 30 min and permeabilized with 0.2% Triton X-100. Coverslips were washed with PBS containing 10% fetal calf serum, incubated for 2 h with the first antibody diluted in PBS, 10% fetal calf serum (monoclonal anti-myc, 1/1000), then washed with PBS and incubated with the second antibody (Alexa 594-coupled anti-mouse antibody, Molecular Probes, 1/250). 4Ј,6-Diamidine dihydrochloride (Roche) was added at the same time to a final concentration of 0.2 g/ml. Coverslips were washed with PBS and distilled water and analyzed with a DMR Leica microscope.
Western Blotting-For cell extracts, cells were washed twice with ice-cold PBS and immediately lysed in Laemmli sample buffer. For the extracts derived from mice, tissues were frozen and homogenized with a Polytron in lysis buffer containing 1% Triton X-100. Protein extracts were resolved on a 12% SDS-PAGE gel and transferred onto a polyvinylidene difluoride membrane (Immobilon-P; Millipore). Membranes were incubated with a polyclonal anti-PAIP2 antibody (1/2000) or a monoclonal or polyclonal anti-myc antibody (both 1/1000) and then with an anti-rabbit or anti-mouse horseradish peroxidase-conjugated secondary antibody. Bound antibody was revealed using an ECL system (Amersham Biosciences).
Immunoprecipitation of Ribonucleoprotein (RNP) Complexes and Analysis by Reverse Transcriptase-PCR-RNP complexes were immunoprecipitated from whole HeLa cells extracts as described below in "Co-immunoprecipitation Experiments." Briefly, pre-cleared lysates were incubated 2 h at 4°C with the anti-PAIP2, an irrelevant antibody, or their respective preimmune sera, prior to incubation with protein A-Sepharose saturated first with tRNA (final concentration of 100 g/ml). The IgG-protein A-Sepharose beads were spun out and the supernatant was collected. After extensive washes, total RNA was purified from equivalent amounts of the supernatants and input, as well as from all the immunoprecipitated material using guanidinium thiocyanate, phenol, and chloroform, followed by isopropyl alcohol precipitation. Each RNA, resuspended in the same volume of RNase-free water, was reverse transcribed with the OmniScript RT kit (Qiagen) according to the manufacturer's protocol. A PCR amplification of the transcripts was then carried out using the TaqPCR Master mixture kit (Qiagen) and the following primers (5Ј to 3Ј): VEGF, ATGGCAGAAGGAGGGCAGCAT and TTGGTGAGGTTTGATC-CGCATCAT; ␤-actin, GCCAACCGCGAGAAGATGACCCAG and CTC-GAAGTCCAGGGCGACGTAGC. Reaction conditions were as follows: 30 s at 94°C, 1 min at 55°C, and 1 min at 72°C followed by a final extension step at 72°C for 10 min. For VEGF, 35 cycles were used, for ␤-actin, 25 cycles. PCR products were analyzed on a 2% agarose gel and visualized by ethidium bromide staining.
Measurement of VEGF mRNA Half-life-S19-R443/PAIP2 cells expressing inducible myc-tagged PAIP2 were stimulated or not with tet-

FIG. 1. The GST-PAIP2 fusion protein interacts with the VEGF mRNA 3-UTR in vitro.
A, schematic map of the VEGF 3Ј-UTR illustrating the templates used for generation of riboprobes for REMSA. E refers to the AREs (pentanucleotide AUUUA) present on the VEGF 3Ј-UTR. 1 corresponds to the HuR binding site. Probes interacting (ϩ) or not (Ϫ) with GST-PAIP2 are mentioned. B, REMSA were performed by incubating the radiolabeled VEGF 3Ј-UTR transcripts with purified GST-PAIP2, GST, or GST-HuR. For the full-length and NsiI transcripts, 750 nM of each GST fusion protein was used (left panels). For the StuI and ⌬StuI transcripts, increasing concentrations of GST-PAIP2 (75 and 325 nM) and a fixed concentration of GST (325 nM) were used (right panels). Competition with specific competitor (ϫ100 molar excess) corresponding to unlabeled StuI and ⌬StuI transcripts, respectively, was also performed in the presence of 325 nM GST-PAIP2. The arrows point to RNA-protein complexes. The brackets encompass free and degraded probe. racycline for 4 h prior to the addition of DRB (25 g/ml). Total RNAs were prepared at 0, 30, 60, and 120 min after the addition of DRB.
Determination of Protein Synthesis-S19-R443/PAIP2 cells were seeded in 12-well plates, supplemented with [ 3 H]leucine (L-[2,3,4 5-3 H]leucine; 2 mCi/ml, 0.8 mM final concentration), and incubated with or without tetracycline for 4 h, in the absence or presence of cycloheximide (5 g/ml). Cells were then fixed and washed twice with ice-cold trichloroacetic acid (5%). The precipitated material was solubilized with 0.1 N NaOH, and the radioactivity was subjected to liquid scintillation counting.
ELISA of Secreted VEGF-S19-R443 control cells or S19-R443/ PAIP2 cells were seeded in 12-well plates and incubated with or without tetracycline for 48 h. 2 h before the end of the tetracycline stimulation, cells were treated with heparin (50 g/ml). Cell culture supernatants were then collected and the cells were counted. Determination of the VEGF concentration of all supernatants was carried out using an ELISA kit Quantikine Mouse VEGF Immunoassay (R&D Systems) following the manufacturer's guidelines. We tested that this assay recognizes both mouse VEGF and hamster VEGF, without cross-reactivity to various cytokines or growth factors.
RNA Interference Experiments-The following 21-mer oligoribonucleotides and their reverse sequence were synthesized by Eurogentec (Liege, Belgium). Two independent small interfering RNAs (siRNA) were designed in homologous regions of hamster/human/mouse PAIP2 cDNA. The first one is 5Ј-GGCUCUUCUCUGGAAGAUCTT-3Ј and corresponds to the coding region 295-313 relative to the first nucleotide of the human PAIP2 complete cDNA (NCB accession number NM_016480). The second one is 5Ј-GAUCUUGUGGUCAAGAGCAT-T-3Ј and corresponds to the coding region 310 -328 of the human PAIP2 complete cDNA. The siRNA sequence targeting the coding region 477-495 relative to the first nucleotide of the EGFP complete cDNA and used as an irrelevant siRNA is the following: 5Ј-GAACGGCAUC-AAGGUGAACTT-3Ј. The two RNA strands were mixed in equimolar ratios and annealed by heating to 95°C for 1 min and then to 37°C for 1 h. A431, HeLa, and S19-R443 cells were transiently transfected twice at a 24-h interval, with 50 nM siRNA against PAIP2, using the calcium phosphate method. 48 h following the second transfection, cells were lysed and analyzed either by Western blotting for PAIP2 expression or by Northern blotting for VEGF mRNA expression.
In Vitro Protein-Protein Interactions-1 g of pCMV-Tag3B/HuR was used for in vitro T3-driven transcription and translation performed in rabbit reticulocyte lysates using the TNT-coupled transcription/translation kit (Promega) with [ 35 S]methionine as the radioactive tracer amino acid in a final volume of 50 l. For binding experiments, 20 l of the reaction was incubated overnight at 4°C with 10 g of immobilized GST or GST-PAIP2 fusion protein in the presence of 0.1 mM 4-(2aminoethyl)benzenesulfonyl fluoride. After centrifugation, the supernatant containing the unbound fraction was removed and the beads were washed five times with a buffer containing 0.2% Triton X-100, 10 mM Tris, pH 8.0, 150 mM NaCl, and 2 mM EDTA. The bound proteins were then eluted by boiling in gel sample buffer and resolved by SDS-PAGE. 2 l of the in vitro translated reaction were loaded on the gel to evaluate the proportion of the radiolabeled protein that was "pull downed" with GST-PAIP2. The gel was fixed, dried, and autoradiographed.
Co-immunoprecipitation Experiments-Cells were washed twice with ice-cold PBS and immediately lysed in a buffer (0.1% Triton X-100, 20 mM Tris, 2 mM EDTA, 25 mM NaCl, 10% glycerol) containing protease and phosphatase inhibitors. 2 mg of protein extract per condition were pre-cleared with preimmune sera, specific for each antibody used, prior to incubation for 2 h at 4°C with 1 g of polyclonal anti-PAIP2 antibody, anti-HuR antibody, or an irrelevant antibody. For cell extracts treated with RNase A, they were incubated with RNase A (0.1 g/mg of protein extracts) for 1 h at 4°C, after the pre-clearing step and prior to incubation with the different antibodies. Protein A-Sepharose beads saturated first in PBS containing 2% bovine serum albumin were added to the reaction mixture for 1 h at 4°C. After extensive washes with lysis buffer, Laemmli buffer was added to the beads, and protein complexes were subjected to SDS-PAGE and Western blotting.

RESULTS
A GST-PAIP2 Fusion Protein Interacts with the VEGF mRNA 3Ј-UTR in Vitro-To identify proteins interacting with the VEGF 3Ј-UTR, we screened an expression mouse testis cDNA library with a [ 32 P]UTP-radiolabeled RNA probe corresponding to the full-length VEGF mRNA 3Ј-UTR. We obtained several clones, the hybridization of which resisted RNase T1 treatment. The relative affinity of the different clones was evaluated by the intensity of hybridization of the VEGF probe to the phage plaques. Moreover, total protein extracts from bacteria obtained with the rescued phagemid vectors (see "Experimental Procedures"), stimulated or not with isopropyl-1thio-␤-D-galactopyranoside, were used in REMSA. The intensity of complexes obtained with identical amounts of protein extracted from isopropyl-1-thio-␤-D-galactopyranoside-stimulated bacteria, also allowed us to measure the relative affinity for the VEGF RNA probe of the different proteins obtained  3. PAIP2 interacts with VEGF mRNA in vivo. A, S19-R443 cells were stably transfected with an expression vector encoding a tetracycline inducible myc-tagged PAIP2. Cells were stimulated for various times with tetracycline (1 g/ml). Inducible overexpression of myc-PAIP2 was then visualized either by Western blotting (upper panel), or by immunostaining (lower panel), using a monoclonal anti-myc antibody as described under "Experimental Procedures." B, S100 cytosolic extracts (5 g) derived from S19-R443/PAIP2 cells treated (ϩ) or not (Ϫ) for 4 h with tetracycline (1 g/ml) were incubated with the radiolabeled NsiI transcript in the absence or presence of ϫ100 or 1000 molar excess of cold competitor. The specific competitor corresponds to the unlabeled NsiI transcript, whereas the nonspecific competitor corresponds to the unlabeled full-length Chinese hamster ERK1 3Ј-UTR transcript (642 bp). The arrow points to the RNA-protein complex. The bracket encompasses free and degraded probe. This experiment is representative of three independent experiments. C, 2 g of unlabeled polyadenylated transcripts (NsiI for the specific probe and ERK1 3Ј-UTR for the nonspecific probe) were bound to oligo(dT) beads. S100 extracts (100 g) derived from S19-R443/PAIP2 cells treated (ϩ) or not (Ϫ) for 4 h with tetracycline (1 g/ml) were incubated with the RNA-coated beads. Proteins interacting specifically with the unlabeled transcripts were then subjected to Western blotting, using a monoclonal anti-myc antibody (upper panel). The Amido Black staining for total protein is shown as a loading control (lower panel). D, whole HeLa cell extracts were immunoprecipitated with no antibody (lane beads), with the preimmune serum, or with a polyclonal anti-PAIP2 antibody. Equal aliquots of purified total RNA isolated from the immunoprecipitates (P) (lanes 1, 3, and 5), and from the supernatants (S) (lanes 2, 4, and 6), were assayed by reverse transcriptase-PCR to detect the VEGF (upper panel) and ␤-actin (lower panel) transcripts. The reactions in lanes 7 and 8 serve as negative controls. during the screening. Thus, we selected the clone interacting with the strongest affinity with the VEGF mRNA 3Ј-UTR. This corresponded to a partial cDNA encoding PAIP2. paip2, located on chromosome 5, is organized into 4 exons, 3 of which include coding sequences. To confirm the PAIP2/VEGF mRNA interaction in vitro, the full-length PAIP2 cDNA was cloned into the pGEX-6P1 vector to obtain a GST-PAIP2 fusion protein. GST-PAIP2 was incubated with the radiolabeled transcript corresponding to the entire rat VEGF mRNA 3Ј-UTR and the PAIP2-RNA complex formation was visualized by REMSA. Fig.  1B shows that GST-PAIP2 but not GST alone interacts with the entire VEGF mRNA 3Ј-UTR (full-length probe). HuR, a member of the ELAV protein family, previously shown to interact with the VEGF mRNA 3Ј-UTR (8) was used as a positive control (the HuR binding site on VEGF mRNA 3Ј-UTR is shown in Fig. 1A). It is noteworthy that, for the same amount of fusion protein used, protein-RNA complexes obtained with GST-PAIP2 and GST-HuR display equivalent intensities. The same results were obtained using an unpolyadenylated probe corresponding to the full-length VEGF 3Ј-UTR (data not shown). By deletion analysis (Fig. 1A), we have restricted the interacting region to a domain comprised between bases 1 and 1280 of the VEGF mRNA 3Ј-UTR (NsiI probe) ( figure 1B). The complementary region of the 3Ј-UTR (bases 1276 to 2211, ⌬NsiI probe) does not bind PAIP2 (data not shown). We then focused on the NsiI domain and divided it into two complementary regions: StuI and ⌬StuI. Fig. 1B shows that GST-PAIP2 interacts with both regions, in a dose-dependent manner. Moreover, this interaction can be completely inhibited by competition with a ϫ100 molar excess of unlabeled specific probe (Fig. 1B). The same molar excess of unlabeled nonspecific probe does not inhibit the PAIP2 interaction (data not shown), demonstrating the specificity of the binding.
Expression of PAIP2 in Different Tissues and Tumor Cell Lines-PAIP2 transcript expression was examined by Northern blotting of mouse and human tissues as well as of different tumor cell lines ( Fig. 2A). In human tissues, PAIP2 mRNA is mainly expressed in brain and lung, whereas in mouse tissues, the most abundant expression of the PAIP2 transcript is detected in testis and heart. These striking differences in PAIP2 mRNA levels between human and mouse, and particularly in the brain, might illustrate a differential transcriptional and/or post-transcriptional regulation of PAIP2 expression between both species. Among the different tumor cell lines examined, HeLa cells contain the highest level of PAIP2 mRNA. Using the GST-PAIP2 fusion protein, we generated a polyclonal antibody directed against PAIP2. We then analyzed PAIP2 expression at the protein level in mouse tissues and in tumor cell lines. Fig.  2B shows maximal expression of PAIP2 in liver and testis, after normalization to levels of ERK2, an ubiquitously expressed protein. Moreover, PAIP2 is also present in the mouse embryo. We note a correlation between the mRNA and proteins levels, except for liver where the PAIP2 protein is more abundant than the mRNA, compared with others tissues. For the tumor cell lines, there is also a good correlation between mRNA and protein levels, with the highest amounts in HeLa cells.
PAIP2 Interacts with VEGF mRNA in Vivo-To analyze the relevance of the PAIP2/VEGF mRNA interaction, we stably transfected S19-R443 cells (CCL39 derivatives) with an expression vector encoding a tetracycline inducible myc-tagged PAIP2 (16). S19-R443 cells were chosen because of their very low basal level of PAIP2 (data not shown). Fig. 3A shows that exogenous PAIP2 is detected after only 4 h of tetracycline treatment and reach a maximum after 12 h (4 -6-fold induction at 4 h to 45-fold induction at 24 h compared with the endogenous level of PAIP2). Moreover, overexpression does not affect its cytoplasmic localization (13) (lower panel). The effects of PAIP2 overexpression on RNA binding were tested after 4 h of tetracycline induction. REMSA experiments with the NsiI part of the VEGF mRNA 3Ј-UTR were performed with S100 extracts derived from cells treated or not with tetracycline. Fig. 3B shows a constitutive protein(s)-RNA complex, the intensity of which is further increased with S100 extracts from tetracycline-treated cells (compare lanes 1 and 5). This complex is completely inhibited by competition with an excess of unlabeled specific transcript corresponding to the NsiI region (lanes 2, 3, 6, and  7), but not competed out with an excess of unlabeled nonspecific ERK1 3Ј-UTR transcript (lanes 4 and 8). These results suggest that PAIP2 itself and/or protein(s) dependent on PAIP2 recruitment is(are) associated in this complex. We performed the FIG. 4. PAIP2 overexpression increases VEGF mRNA half-life in an in vitro RNA degradation assay. Radiolabeled NsiI was transcribed in vitro and incubated with S100 extracts (1 g) from S19-R443/PAIP2 cells treated (ϩ) or not (Ϫ) for 4 h with tetracycline (1 g/ml). An in vitro RNA degradation assay was then performed as described under "Experimental Procedures." All time points were performed in triplicate, and the experiment was repeated three times with similar results. A, representative autoradiograph of the in vitro RNA degradation assay. Times refer to incubation time of extracts with radiolabeled RNA transcripts. The remaining RNA was subjected to electrophoresis on a denaturing formaldehyde-agarose gel. After transfer to a nylon membrane, the amount of undegraded transcript was quantified by Phosphor-Imaging analysis. B, NsiI RNA decay curves of the in vitro degradation assay presented in A. The value at time 0 was taken as 100%. The half-life of the NsiI RNA transcript is 2.5 Ϯ 0.3-fold increased when incubated with S100 extracts from tetracycline-stimulated cells.
same REMSA experiments on the ⌬NsiI probe (bases 1276 to 2211). The RNA-protein complexes visualized with this probe are not modified upon tetracycline treatment (data not shown), which confirms our previous results using GST-PAIP2.
To further demonstrate the presence of PAIP2 in the RNAprotein(s) complex, we performed RNA affinity purification assays. The unlabeled polyadenylated transcript corresponding to the NsiI probe was bound to oligo(dT) beads. S100 extracts from cells treated or not with tetracycline were incubated with the RNA-coated beads. After extensive washing, proteins associated with the bound probe were resolved by Western blotting. Fig. 3C shows that PAIP2 induced by tetracycline strongly interacts with the NsiI transcript. This interaction appears to be specific because, in the same conditions, PAIP2 is not able to bind the polyadenylated ERK1 3Ј-UTR transcript, used as a negative control. Even if proteins are unspecifically trapped on the beads (see Amido Black), the intensity of an irrelevant protein (i.e. ERK2) is not modified by tetracycline treatment (data not shown). This result clearly demonstrates the presence of PAIP2 in the protein(s)-RNA complex described above and then, the ability of PAIP2 to bind VEGF 3Ј-UTR.
To confirm that PAIP2 not only interacts with an in vitro transcript corresponding to a region of the VEGF 3Ј-UTR, but also with the endogenous VEGF mRNA, we performed ribonucleoprotein complexed immunoprecipitation experiments with whole cell extracts from HeLa cells (a cell line expressing strongly PAIP2). Fig. 3D shows that immunoprecipitation of PAIP2 results in the co-precipitation of VEGF mRNA (lane 5, upper panel). Such co-precipitation is highly specific because immunoprecipitation with either preimmune serum (lane 3) or an irrelevant antibody (data not shown) does not co-precipitate VEGF mRNA. As expected, VEGF mRNA is also detected in total RNA prepared from the HeLa cell extracts used for immunoprecipitations (input, lane 7). Moreover, it is noteworthy that immunoprecipitation of PAIP2 does not lead to co-precipitation of a nonspecific transcript such as ␤-actin mRNA (lane 5, lower panel), which is used here as a negative control. The same results were obtained in A431 cells, another cell line that highly expresses PAIP2. These results provide convincing evidence that VEGF mRNA is specifically associated with PAIP2.
PAIP2 Overexpression Increases VEGF mRNA Half-life in an in Vitro RNA Degradation Assay-We then used a specific in vitro RNA degradation assay to evaluate the role of PAIP2 on VEGF mRNA stability, a protocol already used to demonstrate the stabilization of VEGF mRNA induced by the HuR protein or by stress-activated protein kinases (6 -8, 11, 19). S100 extracts were prepared from cells treated or not with tetracycline and incubated with the in vitro radiolabeled NsiI transcript. Fig. 4 shows that this transcript is 2.5 Ϯ 0.3-fold more stable (n ϭ 3) following incubation with extracts from tetracyclinestimulated cells, demonstrating that PAIP2 overexpression confers higher stability of the VEGF mRNA via its 3Ј-UTR, at least in vitro, when the protein is bound to the NsiI region.
PAIP2 Overexpression Increases VEGF mRNA Half-life in Vivo-To analyze the relevance of the PAIP2/VEGF mRNA interaction in vivo, we determined whether overexpression of PAIP2 increased endogenous VEGF mRNA stability. For this purpose, DRB chase experiments in stably transfected cells treated or not with tetracycline were performed. Four VEGF mRNA species ranging from 1.5 to over 4.0 kb are detected on the Northern blot (Fig. 5A). All the isoforms are stabilized upon tetracycline treatment, except form 2 with a long time exposure to DRB. The total amount of VEGF mRNA at each point corresponds to the sum of the global quantity of each isoform measured with a PhosphorImaging system. Hence, we show that the VEGF mRNA half-life is increased 4.4 Ϯ 0.4-fold (n ϭ 3) when PAIP2 is overexpressed (Fig. 5, A and B).
As PAIP2 was shown to act as a translational repressor (13,14), we decided to test whether increased VEGF mRNA stability was because of a general inhibition of protein synthesis, which is responsible for accumulation of many mRNA including oncogenes, growth factors, and VEGF mRNA itself (20). As compared with untreated cells, Fig. 5C shows no major modification of protein synthesis after 4 h of incubation in the absence or presence of tetracycline in S19-R443/PAIP2 cells, which are the previous experimental conditions. This result demonstrates that the PAIP2 stabilizing effect on VEGF mRNA does not reflect a nonspecific inhibitory effect on protein synthesis. PAIP2 specifically stabilizes VEGF mRNA by interacting with the VEGF mRNA 3Ј-UTR.
PAIP2-mediated VEGF mRNA Stabilization Contributes to Increased VEGF Secretion-To correlate VEGF mRNA stability with VEGF production, we measured VEGF protein levels

FIG. 5. PAIP2 overexpression increases VEGF mRNA half-life in vivo.
A, S19-R443/PAIP2 cells were stimulated (ϩ) or not (Ϫ) with tetracycline (1 g/ml) for 4 h prior to incubation, for the indicated times, with DRB (25 g/ml). 20 g of extracted total RNA were analyzed by Northern blotting for expression of VEGF mRNA. The arrows point to the different VEGF isoforms. 18 S ribosomal RNA is shown as a loading control. This experiment is representative of three independent experiments. B, the amounts of VEGF mRNA were quantified with a Phos-phorImaging system. The values are normalized to 18 S rRNA and the values at time 0 were taken as 100%. VEGF mRNA half-lives were deduced from the regression lines: t1 ⁄2 for control unstimulated cells (ϪTet) ϭ 12.5 min; t1 ⁄2 for tetracycline-stimulated cells (ϩTet) ϭ 50 min. C, effect of PAIP2 overexpression on protein synthesis. S19-R443/ PAIP2 cells were supplemented with [ 3 H]leucine and incubated without (white box) or with (hatched box) tetracycline (Tet; 1 g/ml) for 4 h, in the absence or presence of cycloheximide (CHX; 5 g/ml). Incorporated radioactivity is presented as a mean of two independent experiments performed in quadruplicate (left panel). Protein extracts (20 g) prepared in parallel were analyzed, as a control, by Western blotting using a monoclonal anti-myc antibody (right panel). The Amido Black staining for total proteins is shown as a loading control.
in supernatants of cells overexpressing or not PAIP2. Secreted VEGF was measured by ELISA in supernatants of S19-R443 control cells or S19-R443/PAIP2 cells stimulated or not with tetracycline. Fig. 6 shows a 1.5 Ϯ 0.17-fold induction (n ϭ 3) in VEGF secretion when PAIP2 is overexpressed, whereas no induction is detected in control cells incubated with tetracycline. These results suggest that PAIP2-mediated VEGF mRNA stability directly correlates with an increase in VEGF production.
Silencing of PAIP2 by RNA Interference Reduces Steady-state Levels of Endogenous VEGF mRNA-To obtain independent verification of the conclusions raised from the overexpression of PAIP2, we turned to paip2 gene silencing by using siRNA duplexes. We performed these experiments in S19-R443 cells, the same cells used for PAIP2 overexpression assays, as well as in two tumor cell lines: A431 and HeLa. Cells were transfected twice with 50 nM of a 21-nucleotide duplex homologous to the PAIP2 sequence (first PAIP2-siRNA sequence described under "Experimental Procedures"). 48 h following the second transfection, cells were collected for protein and RNA analysis. In contrast to the unchanged level of ERK2 used as a loading control, PAIP2 is efficiently depleted with siRNA (91% of inhibition) in the three cell lines (Fig. 7A). Moreover, the steadystate levels of VEGF mRNA were strongly down-regulated in siPAIP2-transfected A431 and HeLa cells and to a lesser extent in S19-R443 cells. No modifications of steady state VEGF mRNA levels were obtained in cells transfected with the irrelevant GFP siRNA (Fig. 7B, upper panel). Quantification of these results shows an approximate 50% decrease of VEGF mRNA levels in A431 and HeLa cells and a 30% reduction in S19-R443 cells (Fig. 7B, lower panel). Although steady-state VEGF mRNA levels are affected by the knock-down of PAIP2, no change was detected in VEGF mRNA stability measured by a DRB chase experiment (data not shown; see "Discussion"). The same experiments were performed using an independent siRNA sequence targeting PAIP2 (see "Experimental Procedures"). An equivalent decrease of steady-state VEGF mRNA levels was obtained (data not shown), confirming the specificity of the PAIP2 silencing effect on VEGF mRNA expression.
PAIP2 and HuR Physically Interact-To explain the molecular mechanism leading to PAIP2-mediated VEGF mRNA stabilization, we analyzed potential interactions of PAIP2 with proteins that were previously described as VEGF mRNA stabilizing proteins. HuR interacts with the VEGF mRNA 3Ј-UTR and participates in VEGF mRNA stabilization under hypoxia. It is part of a protein complex recruited under hypoxia that is composed of HuR itself and two unknown proteins of 29 and 17 kDa (7,8). As the size of the 29-kDa protein was compatible with that of PAIP2 (13), we have first determined whether PAIP2 and HuR could interact in vitro. As shown in Fig. 8A, in vitro translated HuR interacts with immobilized GST-PAIP2 fusion protein but not with GST protein alone, demonstrating that PAIP2 and HuR interact directly, at least in vitro, without need of any partners. Moreover, this interaction is highly specific because an in vitro translated irrelevant protein does not interact with immobilized GST-PAIP2 (data not shown). To demonstrate an in vivo association, HeLa cells were transiently transfected with an expression vector encoding myc-tagged HuR or myc-tagged PAIP2. We showed that the overexpressed forms of both proteins interact, because they co-immunoprecipitate (data not shown). We then performed co-immunoprecipitation experiments directly on the endogenous proteins to confirm what we have observed in vitro and with overexpressed HuR and PAIP2. Fig. 8B shows that endogenous HuR can be co-immunoprecipitated with an anti-PAIP2 antibody (lane 4). Endogenous PAIP2 can also be co-immunoprecipitated (to a lower proportion than HuR) with an anti-HuR antibody (Fig. 8B, lane 2). Moreover, interaction between both endogenous proteins is specific because neither anti-HuR nor anti-PAIP2 preimmune sera (lanes 1 and 3), nor an irrelevant antibody (lane 5), are able to co-immunoprecipitate endogenous HuR and PAIP2. These results clearly demonstrate an interaction between both proteins in vivo. To evaluate whether this association could be RNA dependent, because PAIP2 and HuR bind to the same target mRNA, we performed coimmunoprecipitations with RNase A-treated cell extracts. Fig.  8C demonstrates that treatment with RNase A does not abolish the interaction between PAIP2 and HuR. Hence, PAIP2 and HuR physically interact without any partner RNA. Nevertheless, RNase A treatment seems to increase the pool of immunoprecipitated HuR, whereas it has no effect on the amount of co-immunoprecipitated PAIP2 (Fig. 8C, lane 4). DISCUSSION To better understand processes controlling VEGF mRNA stability, we cloned PAIP2 as a new protein partner of the VEGF mRNA 3Ј-UTR. At the same time, PAIP2 was shown to interact with the poly(A)-binding protein, which prevents poly(A)-binding protein binding to the poly(A) tail of mRNAs and inhibits cap-dependent and cap-independent translation (13). Here, we show that PAIP2 strongly interacts with the VEGF mRNA 3Ј-UTR both in vitro and in vivo, leading to VEGF mRNA stabilization. Furthermore, PAIP2-mediated VEGF mRNA stabilization correlates with an increase in VEGF production.
The double effect of PAIP2 on mRNA stability (this report and Ref. 21) and mRNA translation (13) has already been described for the iron regulatory proteins IRP1 and IRP2 as well as for CUGBP2, which binds to the cyclooxygenase-2 mRNA. IRP1 and IRP2 proteins tightly regulate ferritin and transferrin receptor mRNA stability and translation, to maintain steady-state levels of iron in the blood (22). In the case of CUGBP2, it positively regulates cyclooxygenase-2 mRNA stability but negatively regulates cyclooxygenase-2 mRNA translation (23). Hence, PAIP2 seems to be a multifunctional protein that controls two important aspects of post-transcriptional regulation of gene expression, such as mRNA stability and translation. VEGF mRNA could represent a particular class of mRNA where the main role of PAIP2 is devoted to mRNA stability.
This increase in mRNA stability is consistent with that mediated by other stabilizing proteins like HuR or hnRNPL (8,12). We also observed that PAIP2 is implicated in the stability of sugar transporters GLUT5 mRNA (21) and GLUT1 mRNA (data not shown), the later being also stabilized by Hu proteins (24). Our results suggest that RNA structures or sequences targeted by PAIP2 are present on several labile mRNAs (25).
tracts are shown as controls. Total RNA extracted from the same cell extracts before (Ϫ) or after (ϩ) RNase A treatment was loaded on a 1% agarose gel to visualize by ethidium bromide staining, as a control, the integrity of the RNA (lower panel). The brackets encompass degraded RNA following RNase A treatment. These experiments are representative of three independent experiments each.  1). B, HeLa cell extracts were immunoprecipitated with polyclonal anti-HuR or anti-PAIP2 antibody. Immunoprecipitation (IP) with no antibody (lane beads) or with anti-HuR and anti-PAIP2 preimmune sera, as well as an irrelevant antibody, are used as negative controls. Immunoprecipitated complexes were immunoblotted using anti-HuR or anti-PAIP2 antibody. Immunoblots with 50 g of total extracts are shown as controls. Asterisks represent co-immunoprecipitated proteins. C, HeLa cells extracts were treated (ϩ) or not (Ϫ) with RNase A prior to the incubation with polyclonal anti-PAIP2 or anti-HuR antibody. Immunoprecipitated complexes were immunoblotted using anti-HuR or anti-PAIP2 antibody (upper panels). Immunoblots with 50 g of total ex-The role of PAIP2 in mRNA stability correlates with its tissue distribution. We observed high levels of PAIP2 in testis, a tissue were mRNA stabilization is a crucial mechanism for gamete maturation (26,27). PAIP2 is also highly expressed in liver where the mRNA stability is crucial for the regulation of the expression of enzymes such as the cholesterol-7␣-hydroxylase (28,29).
From the results of PAIP2 overexpression experiments, it was argued that PAIP2 silencing in the same cells would lead to a decrease in VEGF mRNA stability. Surprisingly, PAIP2 knock-down by RNA interference does not induce such an effect on VEGF mRNA stability, not only in S19-R443 cells but also in two others cell lines: A431 and HeLa. However, following PAIP2 knock-down, we observe a strong decrease in steadystate levels of VEGF mRNA in all the cell lines we have tested. Similar results have been described in the case of HuR and two of its target mRNAs: urokinase plasminogen activator and its urokinase plasminogen activator receptor. Indeed, HuR silencing by using the siRNA approach does not affect urokinase plasminogen activator and urokinase plasminogen activator receptor mRNA stability (whereas HuR overexpression does), but induces a decrease in the steady-state levels of these mRNAs (30). Moreover, a decrease in steady-state levels of several labile mRNAs, whose stability is regulated by HuR, was also observed in HuR-depleted cells (31,32). One hypothesis to explain our results is that PAIP2 also affects the processing or export of VEGF pre-mRNA. In this context, it is noteworthy that PAIP2, although mainly cytoplasmic, is able to shuttle between the nucleus and the cytoplasm (data not shown). Further experiments are needed to definitely resolve this issue.
PAIP2 does not contain any RNA-binding motifs (33) that are common features of proteins that bind pre-mRNA, mRNA, preribosomal RNA, and small nuclear RNA. However, some proteins that were shown to interact with and to stabilize labile mRNA, such as the Wilms' transcription factor, do not contain such sequences (34). The AREs present in the 3Ј-UTR of many labile mRNAs are involved in the rapid turnover of these mRNAs (35). Four AREs corresponding to the pentamer AUUUA were previously described in the ⌬StuI region of the VEGF mRNA 3Ј-UTR (36), a region that we show to interact with PAIP2. A polypyrimidine sequence motif, which has also been demonstrated to mediate changes in mRNA stability (37,38), was described in the StuI region of the VEGF 3Ј-UTR (36). PAIP2 could also target this motif, as we show an interaction between PAIP2 and the StuI region. Therefore, these consensus sequences could represent potential targets for PAIP2.
We demonstrate in this study that PAIP2 directly interacts with HuR, a member of a protein complex recruited to the VEGF mRNA 3Ј-UTR following hypoxia (6). The molecular weight of PAIP2 matches that of one of the three proteins described in this complex. However, RNA as well as protein levels of PAIP2 and HuR are not modified under hypoxia (8) (data not shown). PAIP2 and HuR could be constitutively associated in mediating basal VEGF mRNA stability. HuR was shown to interact with a region downstream of the PAIP2 interacting domain (8,19). Interaction between both proteins could influence the overall RNA structure by bringing together disparate RNA sequences. Such stable structures would make RNA inaccessible to endonucleases.
All our experiments lead us to conclude on a crucial role of PAIP2 in regulating VEGF mRNA expression. Henceforth, we are investigating others target mRNAs for PAIP2 by differen-tial screening using the RNAi-mediated knock-down of PAIP2. Moreover, because PAIP2 may modulate angiogenesis, the regulation of PAIP2 expression upon growth factor or cytokine stimulation and stress situations, where angiogenesis is affected, is also under investigation.