This Article Abstract Full Text (PDF) Supplemental material Other Versions of this Article: MCB.00272-07v1 27/20/7315    most recent Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Barbier, J. Articles by Auboeuf, D. Search for Related Content PubMed PubMed Citation Articles by Barbier, J. Articles by Auboeuf, D.

 Previous Article  |  Next Article 

Molecular and Cellular Biology, October 2007, p. 7315-7333, Vol. 27, No. 20
0270-7306/07/$08.00+0     doi:10.1128/MCB.00272-07
Copyright © 2007, American Society for Microbiology. All Rights Reserved.

Regulation of H-ras Splice Variant Expression by Cross Talk between the p53 and Nonsense-Mediated mRNA Decay Pathways ^,

INSERM U685, Equipe AVENIR, Hôpital Saint Louis, 1 Avenue Claude Vellefaux, Paris F-75010, France

Received 14 February 2007/ Returned for modification 6 March 2007/ Accepted 26 July 2007

	ABSTRACT

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
When cells are exposed to a genotoxic stress, a DNA surveillance pathway that involves p53 is activated, allowing DNA repair. Eukaryotic cells have also evolved a mechanism called mRNA surveillance that controls the quality of mRNAs. Indeed, mutant mRNAs carrying premature translation termination codons (PTCs) are selectively degraded by the nonsense-mediated mRNA decay (NMD) pathway. However, in the case of particular genes, such as proto-oncogenes, mutations that do not create PTCs and therefore that do not induce mRNA degradation, can be harmful to cells. In this study, we showed that the H-ras gene in the absence of mutations produces an NMD-target splice variant that is degraded in the cytosol. We observed that a treatment with the genotoxic stress inducer camptothecin for 6 h favored the production of the H-ras NMD-target transcript degraded in the cytosol by the NMD process. Our data indicated that the NMD process allowed the elimination of transcripts produced in response to a short-term treatment with camptothecin from the major proto-oncogene H-ras, independently of PTCs induced by mutations. The camptothecin effects on H-ras gene expression were p53 dependent and involved in part modulation of the SC35 splicing factor. Interestingly, a long-term treatment with camptothecin as well as p53 overexpression for 24 h resulted in the accumulation of the H-ras NMD target in the cytosol, although the NMD process was not completely inhibited as other NMD targets are not stabilized. Finally, Upf1, a major NMD effector, was necessary for optimal p53 activation by camptothecin, which is consistent with recent data showing that NMD effectors are required for genome stability. In conclusion, we identified cross talk between the p53 and NMD pathways that regulates the expression levels of H-ras splice variants.

	INTRODUCTION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
A large majority of human genes produce several mRNAs with different exon combinations by using alternative promoters, alternative splice sites, and alternative polyadenylation sites. The differential exon content of gene transcripts may change the nature of encoded proteins and considerably change the biological function of gene products (3). In addition, alternatively spliced exons may introduce premature translation termination codons (PTCs) that result in the degradation of the splice variants containing such PTCs through the nonsense-mediated mRNA decay (NMD) pathway. The discrimination between natural and premature stop codons is made by proteins like Y14, MAGOH, elf4A3, RNPS1, and Barentsz (BTZ), which interact with the NMD effectors Upf1, Upf2, and Upf3 (14, 28, 30).

The NMD pathway is a quality control step that degrades aberrant splicing transcripts and is a mechanism that regulates gene expression (20, 38, 47). One model, called regulated unproductive splicing and translation (RUST), hypothesizes that the modulation of the expression levels of translatable (or productive) transcripts may be achieved by the synthesis of unproductive splice forms that are PTC-containing splice variants whose fate is to be degraded without being translated (25). It was estimated that one-third of alternatively spliced exons introduce PTCs, which suggests a widespread coupling of alternative splicing and NMD (25). However, microarray profiling demonstrated that NMD-target splice variants are produced at uniformly low levels across diverse tissues, even when their degradation is inhibited (34). This shows that the coupling of alternative splicing and NMD may only participate in the fine regulation of specific classes of genes. If NMD-target splice variants play a role in gene expression regulation, it can be anticipated that their production is regulated. In this context, several reports demonstrated an important cross talk between mRNA and DNA surveillance machineries (2, 4). Interestingly, although ionizing radiation had no effect by itself on the levels of a mutated p53 mRNA, which is an NMD target in Calu6 cells, it slightly increased the levels of this NMD target in the absence of SMG-1, a regulator of Upf1 (4). These observations suggest that the NMD pathway could be involved in the cellular response to genotoxic stress.

The H-ras gene is one of the most frequently mutated or overexpressed oncogenes in cancers and its product (p21^H-ras) mediates the effects of extracellular signals that affect cell proliferation and apoptosis (29). By studying the expression regulation of the H-ras gene, we observed that one H-ras splice variant containing a small supplementary exon was highly enriched in the nuclear RNA population. The low basal level of this H-ras splice variant in the cytosol was mainly due to its degradation through the NMD pathway. We then observed that camptothecin (CPT), a DNA topoisomerase I inhibitor that induces genotoxic stress, favored the production of the H-ras NMD-target splice variant rather than the translatable H-ras splice variant. We demonstrated that production of the H-ras NMD-target splice variant required the p53 protein, which was induced by CPT. Upf1, a key NMD effector, was required for optimal p53 activation and for the accumulation of the H-ras NMD-target splice variant in response to CPT. We propose that the production of the H-ras NMD-target splice variant participates in the p53-mediated cellular response to CPT by limiting the production of potentially altered translatable p21^H-ras mRNA.

	MATERIALS AND METHODS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Cell culture, treatment, and transfection. MCF7 cells grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) were plated in 10-cm dishes at 50 to 75% of confluence 24 to 48 h before the treatments with the following agents: CPT (1 µM), BN80765 (hCPT [5 µM]), etoposide (50 µM), doxorubicin (2 µg/ml), cisplatin (CDDP [5 µg/ml]), wortmannin (5 µM), actinomycin D (5 µg/ml), and cycloheximide (10 µg/ml). MCF7 cells were transfected by reverse transfection in six-well plates with small interfering RNAs (siRNAs [see the supplemental material]) at a final concentration of 20 nM using Lipofectamine RNAiMax (Invitrogen). Cells were harvested 48 h after transfection or incubated for 6 more hours in the presence of CPT. MCF7 cells were also transfected in six-well plates with 1 µg of HA-SC35 or p53 expression vectors using Lipofectamine LTX (Invitrogen). Cells were harvested 24 h after transfection or incubated for 6 more hours in the presence of CPT. One or four milliliters of TRIzol (Invitrogen) was added to a well of a six-well plate or on a 10-cm dish, respectively, to prepare RNA from whole cells. Alternatively, cells were harvested and used to prepare different cellular fractions.

Cell fractionation. Cells, grown in a 10-cm dish (5 x 10⁶ MCF7 cells), were washed once with 10 ml phosphate-buffered saline (PBS [1x]), and then 1 ml of PBS (1x) was added. Cells were scraped and transferred into an Eppendorf tube. Cells were centrifuged at 4°C for 3 min at 3,000 rpm. The cell pellet was washed once with 1 ml of PBS (1x).

Cytosolic (C) and nuclear (N) fractions were prepared as follows. The cell pellet was carefully suspended in 200 µl of cold buffer RNLa (10 mM Tris HCl [pH 8], 10 mM NaCl, 3 mM MgCl₂, 0.5% NP-40, 1 mM dithiothreitol [DTT], 300 U/ml of RNaseOUT from Invitrogen) and incubated on ice for 5 min. The lysate was centrifuged at 4°C for 2 min at 3,000 rpm to pellet the nucleus. The supernatant (SN1) containing the cytosol was transferred into a new tube. The pellet containing the nucleus (P1) was kept. SN1 was centrifuged for 1 min at full speed to pellet potential contaminant nucleus. The resulting SN2 was transferred into a new tube. One milliliter of TRIzol (Invitrogen) was added. This was the C fraction. P1 was washed once with 200 µl of RNLa. The lysate was centrifuged at 4°C for 2 min at 3,000 rpm. The supernatant was eliminated. Two hundred microliters of RLNa was added to P2 to check the integrity of the nucleus. One milliliter of TRIzol was added to the nucleus. The nucleus with TRIzol was passed five times through a 20-gauge needle (0.9-mm diameter) and vortexed vigorously. This was the N fraction.

Fractions C, R, and N were prepared as described above, except that 200 µl of cold buffer RNLb (10 mM Tris HCl [pH 8], 10 mM NaCl, 40 mM EDTA, 0.5% NP-40, 1 mM DTT, 300 U/ml of RNaseOUT) was added to P2. P2 was carefully suspended, and the lysate was centrifuged at 4°C for 2 min at 3,000 rpm. The resulting supernatant was transferred into a new tube. P3 containing the nucleus was kept. The supernatant was centrifuged for 1 min at full speed to pellet potential contaminant nucleus. The resulting supernatant was transferred into a new tube. One milliliter of TRIzol was added. This was fraction R, which contained ribosomes bound to the nuclear membranes (48). Two hundred microliters of RNLa was added to P3 to check the integrity of the nucleus. Then, 1 ml of TRIzol was added to the nucleus. The nucleus with TRIzol was passed five times through a 20-gauge needle (0.9-mm diameter) and vortexed vigorously. This was fraction N (with no ribosomes).

Fractions C (with ribosomes) and N (without ribosomes) were prepared as follows. The cell pellet was carefully suspended in 200 µl of buffer A (10 mM Tris HCl [pH 8], 140 mM NaCl, 1.5 mM MgCl₂, 10 mM EDTA, 0.5% NP-40, 1 mM DTT, 300 U/ml of RNaseOUT) and incubated for 5 min on ice. The lysate was centrifuged at 4°C for 5 min at 3,000 rpm. The resulting SN1 was transferred into a new tube. P1 was kept. SN1 was centrifuged at full speed for 1 min to pellet contaminant nucleus. The resulting SN2 was transferred into a new tube. One milliliter of TRIzol was added. This was fraction C, which contained ribosomes bound to the nuclear membrane (48). P1 containing the nucleus was washed once with 200 µl of buffer A and centrifuged at 4°C for 3 min at 3,000 rpm. The resulting supernatant was eliminated. Two hundred microliters of buffer A was added to P2 to check the integrity of the nucleus. One milliliter of TRIzol was added. The lysate was passed five times through a 20-gauge needle (0.9-mm diameter) and vortexed vigorously. This was fraction N (without ribosomes).

The soluble nuclear (sN) and nonsoluble nuclear (nsN) fractions were prepared as follows. The preparation was the same as described above, except that P2 was suspended in 200 µl of buffer B (10 mM Tris HCl [pH 8], 420 mM NaCl, 3 mM MgCl₂, 0.5% NP-40, 1 mM DTT, 300 U/ml of RNaseOUT). P2 with buffer B was incubated on ice for 10 min by flicking slowly every 2 min. The lysate was centrifuged at 4°C for 5 min at 3,000 rpm. The resulting SN3 was transferred into a new tube, and P3 was kept. One milliliter of TRIzol was added to SN3. This was the sN fraction. One milliliter of TRIzol was added to P3, and the lysate was passed five times through a 20-gauge needle (0.9-mm diameter) and vortexed vigorously. This was the nsN fraction.

RNA preparation and RT-PCR. RNAs were prepared using TRIzol, and 1 µl of Glycoblue (Ambion) was added before RNA precipitation. Polyadenylated mRNAs were prepared from total RNA (isolated from C or N fractions) using oligo(dT) beads (NucleoTrap mRNA; Macherey-Nagel). Each RNA preparation was treated with DNase I (DNAfree, Ambion). Reverse transcription (RT) was performed with between 0.1 and 1 µg of total RNA using Superscript II (Invitrogen) and random primers. Before performing PCR, all of the RT reaction mixtures were diluted to contain in fine 2.5 ng/µl of initial RNA. PCR was performed using 5 µl of this dilution and GoTaq (Promega). qPCR was performed using Master SYBR green I (Roche) on a Roche LightCycler. Primers and PCR conditions are described in the supplemental material. The quantification of p21^waf1 mRNA was performed using p21S (AGACTCTCAGGGTCGAAAAC) and p21AS (TTCCAGGACTGCAGGCTTC). The PCR product size was 121 bp.

ChIP. For chromatin immunoprecipitation (ChIP), MCF7 cells, treated or not with CPT for 2 h, were cross-linked with 1% formaldehyde. After sonication, DNA bound to proteins was immunoprecipitated with p53 antibody (Sigma). The purified DNA was used for PCR or quantitative PCR (qPCR) to amplify the region of the first intron of the H-ras gene using the sense primer CGCTCAGCAAATACTTGTCG and the antisense primer TGAGGTTACCGTCCTCCAGAAC.

WB analysis. Protein extracts were obtained using cells harvested from a 10-cm dish or from a pool of two wells of a six-well plate using 100 or 50 µl, respectively, of M-PER (Pierce) or NP-40 buffer (50 mM Tris [pH 8], 400 mM NaCl, 5 mM EDTA, 1% NP-40, 0.2% sodium dodecyl sulfate [SDS], 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride). After incubation of the cell lysate for 30 min on ice, cellular debris were pelleted by full-speed centrifugation for 10 min. Protein (12.5 µg) extracted with NP-40 and protein (20 µg) extracted with M-PER buffer (Pierce) were used for p53 and SC35 Western blotting (WB) analyses, respectively, using either p53 (1/1800 [Sigma]) or SC35 (1/1,000 SC-4F11 [Euromedex]) antibodies. Twenty micrograms of protein extracted with NP-40 was used for H-ras and SR protein WB analyses, using either H-ras (1/200, F235 [Santa Cruz]) or mAb104 (1/300) antibodies.

Bioinformatic analysis. Using the FAST DB MySQL database (9) and Perl script, we recovered cassette exons defined by human mRNAs. Only single-cassette exons occurring in CDS were considered (i.e., consecutively skipped exons and cassette exons in untranslated regions were excluded from this study). In addition, exons that were associated with only a specific promoter or a specific alternative terminal exon were not considered. We recovered 6,561 cassette exons from 4,640 distinct genes. Then, three kinds of cassette exons were considered for generation of a PTC: (i) case 1 included exons that insert an in-frame PTC, (ii) case 2 included exons that lead to a frameshift with the insertion of a PTC, and (iii) case 3 skipped exons that lead to a frameshift with the insertion of a PTC. For cases 2 and 3, we compared the open reading frame (ORF) within exons following the cassette exon with the ORF within the same exons of transcript not targeted by NMD. When the ORFs were the same, the considered cassette exon did not generate an NMD candidate. In 1,072 (16%) of the cassette exons, the corresponding transcripts were targeted by NMD. This result is consistent with a previous study (25), which reported that 18% of cassette exons ("perfect exon skip" and "exon inclusion"), supported by at least two expressed sequence tags, generated NMD candidates.

The p53MH algorithm (19) was used to predict p53-responsive genes. This algorithm was run using 93 as a cutoff score, on a 13-kb-long region: 3 kb upstream of the first transcription initiation site and 10 kb downstream of this site (or less when the gene length was smaller than 10 kb). Twelve percent (563 out of 4,640) of the considered genes were predicted to be p53-responsive genes in agreement with a previous report (19). We used matrices and default thresholds from ESEfinder (6) to find motifs of SC35, SF2/ASF, SRp40, and SRp55 in cassette exons that generate or do not generate a PTC. For each category, a sequence was constituted by assembling the sequences of each corresponding cassette exon after each other. A score was calculated by dividing the number of sites by the length of the analyzed sequence, as previously described (46, 49). By considering the four exon splicing enhancer (ESE) motifs analyzed, the average number of ESEs per nucleotide was 0.1460. This result is consistent with 0.1466 ESE/nucleotide reported elsewhere (46).

Statistical analysis. Statistical analysis was performed using the one-sided Wilcoxon rank sum test (nonparametric test for continuous data).

	RESULTS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Differential enrichment of H-ras splice variants in cytosolic and nuclear fractions. The H-ras gene produces two splice variants that contain or not an 82-nucleotide-long exon (IDX or exon 5 [see Fig. S1 in the supplemental material]) (16). In the MCF7 breast cancer cell line, the level of the splice variant containing E5 in total (T) cellular extract was low compared to that of the classic H-ras splice variant that does not contain this exon, as determined by RT-PCR using primers flanking E5 (lane 1, Fig. 1A). Nevertheless, after preparing cytosolic and nuclear fractions, whose quality was validated by looking at tRNA^tyr (mainly located in the cytosol) and U6 snRNA (mainly located in the nucleus), both H-ras splice variants were detected at similar levels in the N fraction, while the splice variant containing E5 was almost undetectable in the C fraction (lanes 2 and 3, Fig. 1A). Because the splice variant containing E5 was not enriched in nuclear membrane-bound ribosomes (lane 5, Fig. 1A), the C fraction was prepared using a higher concentration of EDTA to simultaneously extract the nuclear membrane-bound ribosomes (48). The relative level of the splice variant containing E5 was still higher in the N fraction deprived of nuclear membrane-bound ribosomes than in the C fraction (lanes 7 and 8, Fig. 1A). Therefore, the same preparation of cellular fractions was used in all subsequent experiments. The H-ras splice variant ratios were similar in sN and nsN fractions (lanes 10 to 11, Fig. 1A).

View larger version (50K): [in this window] [in a new window]   FIG. 1. Differential enrichment of H-ras splice variants in different cellular fractions. (A) Coamplification of H-ras splice variants containing or not containing E5. RNAs were purified from different MCF7 cellular fractions. PCR was performed to amplify tRNAtyr or U6 snRNA or to coamplify H-ras splice variants containing or not containing E5, using primers in H-ras exons 4 and 6. (B) Characterization of the expression levels of four H-ras splice variants. RT-PCR was performed using different RNA populations and different sets of primers to coamplify specifically two splice variants, as shown in Fig. S1 in the supplemental material. (C) Nuclear E5+ splice variants are polyadenylated mRNAs. Polyadenylated mRNAs were prepared from total RNA isolated from C or N fractions using oligo(dT) beads. RT was performed using random or oligo(dT) primers (RP or dT, respectively). (D) Coamplification of H-ras splice variants containing or not E5. Sense primers in exon 4, 3, or 2 and the E6/7 antisense primer that overlaps the last exons, 6 and 7, were used to coamplify E5+ and E5– splice variants. It was not possible to coamplify the E5+ and E5– transcripts by using a sense primer in exon 1 because of its high GC content. A forward primer overlapping exons 1 and 2 (E1/2) and reverse primers overlapping exons 5 and 6 (E5/6) or exons 4 and 6 (E4/6) were used to amplify the transcripts containing or not exon 5 in separate reactions. (E) Quantification of H-ras splice variants in different RNA populations. qPCR reactions were performed using different RNA populations [isolated from total cellular extract (T), C fraction, or N fraction] to quantify the levels of all the H-ras transcripts (hras), E5– splice variant, or E5+ splice variant as shown in Fig. S1 in the supplemental material. The value obtained for each sample in the qPCR assay was calculated per µg of RNA. Experiments were performed at least three times. Data are presented as the mean ± standard error of the mean (*, P < 0.005). (F) Characterization of the expression levels of four H-ras splice variants in MLS-1765 and HeLa cells. RT-PCR was performed using different RNA populations and different sets of primers to specifically coamplify two splice variants.

We then quantified the levels of all the H-ras transcripts (hras), as well as E5⁺ and E5^– splice variants, by qPCR. It was not possible to quantify E5⁺i6⁺ and E5^–i6⁺ due to their very low expression levels. While the concentrations of all the hras transcripts were similar in the cytosol and the nucleus, the E5^– splice variant was more concentrated in the cytosol than in the nucleus, and the E5⁺ splice variant was more concentrated in the nucleus than in the cytosol (Fig. 1E). The ratios of the concentration in the N fraction to that in the C fraction were 0.8 for hras, 0.3 for E5^–, and 5.3 for E5⁺. These results indicate that E5⁺ and E5^– were differentially enriched in cytosolic and nuclear RNA populations.

The E5⁺ splice variant was also enriched in the nuclear extracts prepared from the MLS-1675 and HeLa cells (Fig. 1F, lane E4-E6/7). Interestingly, while the E5⁺ splice variant was much more abundant than the E5⁺i6⁺ splice variant in MCF7 cells, both splice variants were detected at a similar level in the cytosolic extracts prepared from MLS-1675 and HeLa cells (Fig. 1F, lane E4/5-E7).

The H-ras E5⁺ splice variant is subject to cytoplasmic NMD. To measure the half-life of each H-ras splice variant, MCF7 cells were incubated for different times in the presence of actinomycin D, which blocks transcription. When transcription was inhibited, E5^– levels decreased at a constant low rate and E5⁺ levels decreased at a higher rate during the first 3 h (Fig. 2A, upper panel). Similarly, the cytosolic E5^– levels decreased at a constant low rate, while there was about half less cytosolic E5⁺ after only 3 h of treatment (Fig. 2A, lower panel). The decrease in the concentration of nuclear E5⁺ was slower than that of cytosolic E5⁺ [compare E5⁺(N) with E5⁺(C) (Fig. 2A)] and suggests that E5⁺ was rapidly degraded in the cytosol.

View larger version (57K): [in this window] [in a new window]   FIG. 2. Degradation of E5+ in the cytosol by the NMD pathway. (A) Half-life of H-ras splice variants. Different cellular fractions were prepared from cells treated with actinomycin D for different times (hours). qPCR was performed to quantify E5+ and E5– levels. The values at baseline (0 h) were designated 1. (B) Effect of cycloheximide on the E5+/E5– ratio. Different cellular fractions were prepared from cells treated or not with cycloheximide for 6 h. PCR was performed to amplify PTGS2 mRNA and to coamplify H-ras E5+ and E5– splice variants. (C) Quantification of H-ras splice variant levels in the presence of cycloheximide. Cells were treated with cycloheximide for different times (hours). qPCR was performed to quantify E5+ and E5– levels. The values at baseline (0 h) were designated 1. (D) Effect of cycloheximide on the cytosolic and nuclear levels of E5+ splice variant. C and N fractions were prepared from cells treated with cycloheximide for different times (hours). qPCR was performed to quantify the cytosolic and nuclear E5+ levels. The values obtained at baseline (0 h) in fractions C and N were designated 1. (E) Effects of siUpf1, siUpf2, siBTZ, siRNPS1, and the control siGAPDH on H-ras splice variant levels. MCF7 cells were transfected with different siRNAs as indicated. qPCR was performed to quantify E5+ and E5– levels. The values obtained in the control experiments (siGAPDH lanes) were designated 1. (F) Effects of siUpf1, siUpf2, siBTZ, siRNPS1, and the control siGAPDH on the cytosolic and nuclear levels of E5+ splice variant. C and N fractions were prepared from cells transfected with different siRNAs as indicated. qPCR was performed to quantify the cytosolic and nuclear E5+ levels. The values obtained in the control experiments (siGAPDH lanes) were designated 1. (G) Effect of wortmannin on E5+ splice variant levels. Cells were treated or not with wortmannin for 4 h. Different fractions were prepared, and qPCR was performed to quantify E5+ levels. Each value obtained in the presence of wortmannin was divided by the value obtained in the corresponding control experiment (i.e., cells not treated with wortmannin). Experiments were performed at least three times. Data are presented as the mean ± standard error of the mean (*, P < 0.05). (H) Effects of siUpf1, siBTZ, and the control siGAPDH on the p21H-ras protein level. Protein extracts were prepared from cells transfected with different siRNAs, as indicated, to perform a WB analysis of H-ras and actin protein levels with anti-H-ras (H-ras) and antiactin (Actin) antibodies.

We next inhibited the NMD pathway by using validated siRNAs (siUpf1, siUpf2, siBTZ, and siRNPS1) that target various NMD factors (see Fig. S2 in the supplemental material) compared with an siRNA targeting glyceraldehyde-3-phosphate dehydrogenase (siGAPDH) as a control. First, the global level of E5⁺ but not that of E5^– was increased upon transfection of siRNAs that target Upf1 and BTZ (Fig. 2E). This demonstrates that a subset of NMD factors participate in E5⁺ degradation. Second, the cytosolic E5⁺ levels increased in the absence of Upf1 and BTZ, while the nuclear E5⁺ levels did not (Fig. 2F). These results demonstrate that the mature E5⁺ splice variant is normally exported in the cytosol, where it is degraded by the NMD pathway. Reinforcing the role of Upf1 in this process, wortmannin, which inhibits SMG-1, a phosphatidylinositol 3-kinase-like kinase that activates Upf1 (33), increased E5⁺ levels in both the T and C fractions but not in the N fraction (Fig. 2G). Finally, upon knockdown of Upf1 but not BTZ, we observed a decrease of the p21^h-ras protein level (Fig. 2H). As both Upf1 and BTZ participated in H-ras E5⁺ degradation (Fig. 2E and F), it is plausible that Upf1 alters p21^h-ras protein levels independently of its NMD function, but this effect could be related to its function in translation or this could be due to an indirect effect of Upf1 knockdown on cell proliferation (1, 2).

Altogether, our data show that, despite a simple genomic organization (see Fig. S1 in the supplemental material), the H-ras gene produced four splice variants. Two of these splice variants contained intron i6 (i.e., E5⁺i6⁺ and E5^–i6⁺) and were poorly expressed in MCF7 cells (Fig. 1B), and two others did not contain i6 (i.e., E5⁺ and E5^–). The spliced and polyadenylated E5⁺ splice variant was abundant in the nuclear RNA population and was exported in the cytosol, where it was degraded by the NMD process.

Differential regulation of H-ras splice variant expression by topoisomerase inhibitors. To test whether E5⁺ plays a role in regulation of H-ras gene expression, we challenged H-ras gene expression by treating MCF7 cells for 6 h with CPT, an inhibitor of topoisomerase I, and with etoposide and doxorubicin, two inhibitors of topoisomerase II. These molecules induce genotoxic stress and mediate some of their effects by activating p53, which regulates H-ras promoter activity (8, 21). These molecules also have various effects on splicing (40). We looked at E5⁺ and E5^– levels in response to these molecules in both the cytosolic and nuclear fractions. The value obtained for each splice variant in the cytosolic fraction in the absence of treatment was designated 1 to highlight the differential effects of the drugs depending on the RNA populations. While etoposide had only minor effects on H-ras splice variant levels, doxorubicin decreased more the nuclear E5⁺ levels than the nuclear E5^– levels [compare E5⁺(N) with E5^–(N) (Fig. 3A)]. Interestingly, CPT increased the nuclear E5⁺ levels and had a minor effect on nuclear E5^– levels [3- versus 1.5-fold increase, respectively, comparing E5⁺(N) with E5^–(N) in Fig. 3A]. However, CPT had no effects on H-ras splice variant levels in the cytosolic RNA population [E5⁺(C) and E5^–(C), Fig. 3A], which correlated with an absence of effect of CPT on the levels of the p21^H-ras protein encoded by the E5^– splice variant (Fig. 3B; also see Fig. S1 in the supplemental material).

View larger version (70K): [in this window] [in a new window]   FIG. 3. Differential regulation of H-ras splice variant expression by topoisomerase inhibitors. (A) Effects of topoisomerase inhibitors on H-ras splice variant levels. MCF7 cells were treated or not with CPT, etoposide, or doxorubicin for 6 h. C and N fractions were prepared and used to quantify E5+ and E5– levels (upper and middle panels, respectively). The values obtained in the absence of treatment in fraction C were assigned 1 to highlight the differences between fractions C and N. PCR was performed to coamplify both E5– and E5+ splice variants (lower panel). (B) Effect of CPT on p21H-ras protein levels. MCF7 cells were treated or not with CPT for 6 h. Protein extracts were used for WB analysis of H-ras and actin protein levels. (C) Effect of CPT on the H-ras E5+i6+/E5+ ratio. Different cellular fractions were prepared from cells treated or not with CPT for 6 h. RT-PCR was performed using a sense primer (E4/5) overlapping E4 and E5 and an antisense primer in E7 to coamplify E5+i6+ and E5+ splice variants. (D) Kinetics of CPT effects in the cytosolic RNA population. C fractions were prepared from cells treated with CPT for different times or cells treated with CPT for 6 h followed by CPT withdrawal and cell incubation for different times (6 + 2 h, 6 + 4 h, and 6 + 18 h). The levels of E5– and E5+ splice variants were quantified by qPCR. The values at baseline (0 h) for each splice variant were designated 1. (E) Kinetics of CPT effects in the nuclear RNA population. Panel E is the same as panel D using N fractions (upper panel). PCR was performed to coamplify E5– and E5+ splice variants (lower panel). (F) Time course analysis of the E5+(C) and E5+(N) levels between 10 and 24 h of CPT treatment. Experiments were performed at least three times. Data are presented as the mean ± standard error of the mean (*, P 0.05).

View larger version (48K): [in this window] [in a new window]   FIG. 5. Role of SC35 in CPT effects on H-ras splice variant levels. (A) Roles of SC35, A1, and p53 in H-ras splice variant levels. qPCR was performed to quantify E5+ and E5– levels in the nuclear (N) RNA populations (upper panel). The values obtained in the control experiments (siGAPH lanes) were assigned as 1. PCR was performed on the nuclear RNA population to coamplify E5+ and E5– splice variants (lower panel). (B) WB analysis of SC35. Protein extracts obtained from cells transfected with siSC35, siP53, or siGAPDH were prepared and used for WB analysis of SC35, with actin as a control on the same membrane. SC35, anti-SC35 antibody; Actin, antiactin antibody. (C) Roles of p53 and SC35 in the CPT-mediated effects on H-ras splicing. Cells transfected with different siRNAs were treated or not with CPT for 6 h. qPCR was performed to quantify E5+ and E5– levels using RNAs prepared from N fractions. In each set of experiments (siGAPDH, siP53, and siSC35), the CPT effects on the E5+ splice variant were compared with the CPT effects on the E5– splice variant. (D) Opposite effects of siP53 and siSC35 on the nuclear E5+/E5– ratio. N fractions were prepared from cells transfected with different siRNAs and treated or not with CPT for 6 h. PCR was performed to coamplify E5+ and E5– splice variants using an antisense primer overlapping the last exon-exon junction and a sense primer in exon 4 (H-rasE4) or in exon 2 (H-rasE2). (E) Exogenous HA-SC35 protein was overexpressed in HeLa cells but not in MCF7 cells. Protein extracts were prepared from MCF7 and HeLa cells transfected with an HA-SC35 or a control expression vector. HA-SC35 protein was highly expressed compared to the endogenous SC35 protein in HeLa cells but not in MCF7 cells. (F) SC35 overexpression decreased the E5+/E5– ratio. Total RNAs were prepared from MCF7 and HeLa cells transfected with a HA-SC35 or a control expression vector. qPCR was performed to measure the levels of the H-ras E5+ and H-ras E5– splice variants. In each set of experiments, the effects on the E5+ splice variant were compared with the effects on the E5– splice variant. (G) Effect of CPT on SC35 protein levels. MCF7 cells were treated or not with CPT for 6 h and 18 h. WB analysis of SC35 was performed, using actin as a control. (H) Effect of CPT on the expression of SR proteins. MCF7 cells were treated or not with CPT for 18 h. WB analysis of SR proteins was performed, using the mAb104 antibody and actin as a control. (I) Effect of different genotoxic stress inducers on p53 and SC35 protein levels. MCF7 cells were treated or not with CPT, hCPT, cisplatin (CDDP), or doxorubicin (DOX) for 6 h. WB analysis of p53 and SC35 protein levels was performed. (J) Effect of different genotoxic stress inducers on H-ras E5+ and E5– nuclear levels. MCF7 cells were treated or not with CPT, hCPT, cisplatin, or doxorubicin for 6 h. qPCR was performed to measure the nuclear levels of the H-ras E5+ and H-ras E5– splice variants. PCR was performed to coamplify E5+ and E5– splice variants by using an antisense primer overlapping the last exon-exon junction and a sense primer in exon 4 (H-rasE4) or in exon 2 (H-rasE2). Experiments were performed at least three times. Data are presented as the mean ± standard error of the mean (*, P 0.05).

Role of p53 in H-ras splice variant expression regulation by CPT. To characterize the CPT effects on E5⁺ splice variant expression levels, MCF7 cells were preincubated or not with actinomycin D for 2 h before adding CPT for 2 additional hours. CPT alone increased nuclear E5⁺ levels, and this effect was completely blocked by actinomycin D pretreatment (Fig. 4A). This shows that the CPT effect on nuclear E5⁺ level required ongoing transcription.

View larger version (46K): [in this window] [in a new window]   FIG. 4. Role of p53 in CPT effects on H-ras splice variant levels. (A) Effect of actinomycin D on CPT-mediated effects. Cells were pretreated or not with actinomycin D for 2 h and then treated or not with CPT for 2 more hours. qPCR was performed to quantify E5+ levels using RNAs prepared from N fractions. The value at baseline (no treatment) was designated 1. (B) Effect of cycloheximide on p53 protein levels in response to CPT. MCF7 cells were pretreated or not with cycloheximide for 30 min and then treated or not with CPT for 6 h. WB analysis was performed with p53 and actin as a control on the same membrane. P53, anti-p53 antibody; Actin, antiactin antibody. (C) Effect of cycloheximide on nuclear E5+ levels in response to CPT. MCF7 cells that were pretreated or not with cycloheximide for 2 h were treated or not with CPT for 2 h. Shown is the quantification by qPCR of E5+ levels using RNAs prepared from nuclear fractions. The value at baseline (no treatment) was designated 1. (D) Roles of p53 in CPT-mediated effects on nuclear E5+ splice variant levels. qPCR was performed to quantify E5+ in the nuclear RNA population. The value at baseline (lane siGAPDH in the absence of CPT) was designated 1. (E) Addition of fresh medium simultaneously with CPT stimulated the CPT effect on H-ras splice variants. qPCR was performed to quantify the E5+ and E5– levels in the nuclear extracts prepared from MCF7 cells plated 24 h before addition of CPT alone or CPT with fresh medium. (F) Recruitment of p53 on the endogenous H-ras gene. MCF7 cells were treated or not with CPT for 2 h. A ChIP assay using a control antibody and a p53 antibody was performed. The purified DNA was used in PCR (upper panel) and qPCR (lower panel). (G) CPT favored the production of the E5+ splice variant in MLS-1675 cells but not in HeLa cells. qPCR was performed to quantify the E5+ and E5– levels in nuclear extracts prepared from MCF7, MLS, or HeLa cells treated with CPT for 6 h. (H) CPT increased the E5+/E5– ratio in MLS-1675 cells but not in HeLa cells. PCR was performed to coamplify E5– and E5+ splice variants from nuclear extracts prepared from MCF7, MLS, or HeLa cells treated with CPT for 6 h. (I) P53 was poorly stabilized in HeLa cells treated with CPT. Protein extracts were prepared from MCF7 and HeLa cells treated or not with CPT for 6 h 1 day after transfection with a p53 or a control expression vector. p53 protein was almost not detected in HeLa cells treated or not with CPT but was strongly expressed after transfection with a p53 expression vector. (J) p53 overexpression increased the levels of the H-ras E5+ splice variant. Total RNAs were prepared from MCF7 and HeLa cells treated or not with CPT for 6 h 1 day after transfection with a p53 or a control expression vector. qPCR was performed to measure the levels of the H-ras E5+ splice variant. (K) p53 overexpression did not alter the levels of the H-ras E5– splice variant. qPCR was performed to measure the levels of the H-ras E5– splice variant. The same experimental conditions as those described in panel I were used. (L) p53 was stabilized in MCF7 cells treated with CPT or doxorubicin (Doxo) for 18 h. Protein extracts were prepared from MCF7 cells treated or not with CPT or doxorubicin for 18 h or from MCF7 cells transfected or not with a p53 expression vector. (M) Effects (fold) of topoisomerase inhibitors and p53 overexpression on H-ras splice variant levels. MCF7 cells were treated or not with CPT or doxorubicin for 18 h or transfected or not with a p53 expression vector. Nuclear fractions were prepared and used to quantify E5+ and E5– levels. (N) Recruitment of p53 on the endogenous H-ras gene following p53 overexpression or in response to CPT treatment but not in response to doxorubicin treatment. MCF7 cells were treated or not with CPT or doxorubicin for 18 h or transfected or not with a p53 expression vector. A ChIP assay using a control antibody and a p53 antibody was performed. The purified DNA was used in qPCR.

To test further this hypothesis, MCF7 cells were transfected with siRNAs that target p53 or GAPDH as a control (siP53 and siGAPDH, respectively; see Fig. S2 in the supplemental material) 48 h before a 6-h CPT treatment. As shown in Fig. 4D, a 6-h CPT treatment induced a sixfold increase in the nuclear E5⁺ levels in the siGAPDH control experiment and this effect was abolished in the absence of p53 (lanes siP53, Fig. 4D). We must point out that the approximately sixfold CPT effect was higher in this set of experiments than the threefold effect observed in Fig. 3, because the transfection medium was replaced by a fresh medium prior to addition of CPT in order to avoid cross-reaction between the transfection medium and CPT. As shown in Fig. 4E, the simultaneous addition of fresh medium with CPT compared to the addition of CPT alone, stimulated the H-ras gene transcriptional activity (increase in both E5⁺ and E5^– splice variant levels) probably due a stimulatory effect of growth factors. Importantly, this did not alter the differential effect of CPT, which still favored production of the E5⁺ splice variant of about twofold that of the E5^– splice variant (Fig. 4E). Finally, we showed by a ChIP assay that CPT treatment induced an increase in the p53 recruitment on the endogenous H-ras gene (Fig. 4F). Altogether, these results demonstrate that p53 activation was required to mediate the CPT effect on the H-ras E5⁺ splice variant expression level.

To potentially extend our observations in other cellular models, we used MLS-1675 cells, which express a wild-type p53 protein (5), and HeLa cells, in which p53 is degraded by the human papillovirus E6 (24). While a similar effect of CPT was observed in MLS-1675 cells compared to MCF7 cells, CPT had almost no effect in HeLa cells (Fig. 4G and H). Nevertheless, overexpression of p53 in HeLa cells (Fig. 4I) induced a strong increase in the H-ras E5⁺ splice variant levels and to a lesser extent the H-ras E5^– splice variant level (Fig. 4J and K). In MCF7 cells, overexpression of p53 also preferentially increased the levels of the H-ras E5⁺ splice variant and enhanced the CPT effect (Fig. 4I, J, and K). We must point out that in this set of experiments the expression levels of the H-ras transcripts were measured using total cellular extracts. Because a 6-h CPT treatment had no effect on the E5⁺ splice variant levels in the cytosol (Fig. 3D), the CPT effect on this variant was much lower using total cellular extracts (Fig. 4J) than nuclear extracts (Fig. 3E).

While our data indicated that accumulation of p53 protein in response to CPT and p53 overexpression in MCF7 cells resulted in an increase in the expression levels of the H-ras transcripts, in particular of the H-ras E5⁺ splice variant (Fig. 4C and J), a treatment with doxorubicin for 6 h that stabilized p53 did not (data not shown) (Fig. 3A). Therefore we performed a longer treatment with doxorubicin. A treatment with doxorubicin for 18 h that resulted in a similar effect on p53 protein expression levels compared to a treatment with CPT for 18 h in MCF7 cells (Fig. 4L) did not increase the levels of the H-ras transcripts in the N fraction compared to a treatment with CPT for 18 h or compared to p53 overexpression (Fig. 4M). By performing ChIP experiments using a p53 antibody, we demonstrated that the recruitment of p53 on the H-ras gene in MCF7 cells was strongly increased after both CPT treatment and p53 overexpression, but not after doxorubicin treatment. Altogether, our results demonstrated that the activation of the H-ras gene by the p53 transcriptional factor upon CPT treatment or p53 overexpression resulted in the preferential synthesis of the unproductive H-ras NMD-target transcript.

Role of SC35 in H-ras splice variant expression regulation by CPT. Because CPT favored the production of the E5⁺ splice variant compared with the E5^– splice variant (Fig. 3E), we investigated the potential role of hnRNP A1 (A1) and SC35, two splicing factors involved in H-ras pre-mRNA splicing (16). In addition, SC35 has been reported to be affected by topoisomerase I and CPT and A1 has been involved in the stress-mediated effect on splicing (12, 39, 42, 45). MCF7 cells were transfected with siRNAs that target A1 and SC35 (siA1 and siSC35, respectively [see Fig. S2 in the supplemental material]) and with siP53 to test their effect on H-ras splicing in the absence of CPT. While siA1 slightly decreased the levels of both H-ras splice variants without affecting their ratio (lane siA1, upper and lower panels, Fig. 5A), siSC35 increased the E5⁺/E5^– ratio, which suggests that SC35 favors H-ras E5 exclusion (lane siSC35, upper and lower panels, Fig. 5A).

Interestingly, we observed that siP53 had the opposite effect from siSC35 by decreasing the E5⁺/E5^– ratio (lane siP53, upper and lower panels, Fig. 5A). The same results were obtained using RNA purified from total and cytosolic extracts (data not shown). It is noteworthy that when we checked the effects of the various siRNAs on the different protein levels, we observed that the transfection of MCF7 cells by siP53 was reproducibly associated with an increase in SC35 protein levels (Fig. 5B). These results suggest that p53 depletion decreased the E5⁺/E5^– ratio due to an increase in SC35 protein levels.

MCF7 cells were then transfected with siGAPDH, siP53, or siSC35, 48 h before a 6-h CPT treatment. In the control experiment, the nuclear level of E5⁺ compared with E5^– was doubled by CPT treatment (lane siGAPDH, Fig. 5C). This effect was abolished by siP53 and enhanced by siSC35 (lanes siP53 and siSC35, respectively, Fig. 5C). Consequently, CPT increased the nuclear E5⁺/E5^– ratio and this effect was abolished in the absence of p53 and enhanced in the absence of SC35 (Fig. 5D, upper panel). The same results were obtained using a forward primer in exon 2 (Fig. 5D, lower panel). We must point out that the E5⁺/E5^– ratio was lower when using a forward primer in exon 4 than in exon 2, probably due to the size differences of the PCR products that may not be coamplified with the same efficiency. Nevertheless, using either a forward primer in exon 4 or in exon 2, CPT induced an increase in the E5⁺/E5^– ratio (lanes siGAPDH), siP53 decreased the E5⁺/E5^– ratio (compare lanes siP53 and siGAPDH in the presence of CPT), while siSC35 increased it (compare lanes siSC35 and siGAPDH in the presence of CPT).

To test further the role of SC35, MCF7 and HeLa cells were transfected with an hemagglutinin (HA)-SC35 expression vector. Nevertheless, in MCF7 cells, the expression of the HA-SC35 protein was low compared to the endogenous SC35 protein (Fig. 5E) and we did not observe any significant effect under these conditions on the E5⁺/E5^– ratio (Fig. 5F). As described before (Fig. 4J), the CPT effect on the E5⁺/E5^– ratio was much lower using total cellular extracts of transfected cells than nuclear extracts. Importantly, in HeLa cells, where a high level of HA-SC35 could be achieved (Fig. 5E), overexpression of SC35 clearly decreased the E5⁺/E5^– ratio (Fig. 5F). Altogether, these results indicate that SC35 promotes H-ras exon E5 exclusion.

Because SC35 depletion mediated a similar effect on splicing to CPT (Fig. 5A and C), we hypothesized that CPT altered SC35 expression. Interestingly, it was also shown that SC35 is phosphorylated by the kinase activity of topoisomerase I that is inhibited by CPT (39). We observed that treatment with CPT for 18 h decreased SC35 protein levels and treatment for 6 h altered the migration profile of the SC35 protein (Fig. 5G). Interestingly, using the mAb104 antibody, which recognized a common phosphor-epitope in several SR splicing factors (17), we observed that an 18-h CPT treatment resulted in the alteration of the signal at the levels of the SRp30 proteins, which include SC35, and a moderate increase in the SRp55 and SRp75 phospho-epitopes (Fig. 5H). These data suggest that the 18-h CPT treatment did not profoundly alter the expression of several SR proteins.

To test further our hypothesis that CPT modulated H-ras expression through the inhibition of topoisomerase I and alteration of SC35 splicing factor, we used another toposimerase I inhibitor (BN80765, or hCPT) that we compared to two other genotoxic stresses, including doxorubicin, which inhibits topoisomerase II, and cisplatin (CDDP), which acts by cross-linking DNA. As shown in Fig. 5I, while all three treatments resulted in the stabilization of p53 protein as expected, only hCPT altered the SC35 profile migration as CPT did (Fig. 5G). Finally, only hCPT increased the nuclear levels of the H-ras E5⁺ transcript levels and increased the E5⁺/E5^– ratio (Fig. 5J) as CPT did (Fig. 3E). In conclusion, our results indicate that CPT and hCPT selectively stimulated the production of the H-ras NMD-target splice variant, first by activating the p53 transcription factor (Fig. 4) and second by altering SC35 splicing factor (Fig. 5).

Because p53 overexpression had a similar effect on H-ras gene expression compared to CPT treatment in MCF7 cells (Fig. 4J, K, and M), we next tested whether p53 overexpression mediated its effects on H-ras splicing by also inhibiting SC35 expression. After the transfection of MCF7 cells with a p53 expression vector, we observed no significant effects of p53 overexpression on SC35 protein expression levels (data not shown). However, when we compared the effects on the nuclear E5⁺/E5^– ratio of a CPT treatment for 18 h and of p53 overexpression in MCF7 cells, we observed that CPT had a stronger effect than p53 overexpression (Fig. 6A). p53 overexpression resulted in increases of about three- and twofold in the nuclear levels of the E5⁺ and E5^– transcripts, respectively, while CPT treatment for 18 h resulted in increases of about six- and twofold in the nuclear levels of the E5⁺ and E5^– transcripts, respectively (Fig. 4M and 6A). This observation suggests that while both CPT treatment and p53 overexpression favored the production of the H-ras E5⁺ splice variant, this effect is more pronounced after a CPT treatment. This discrepancy could be apparently linked to an absence of effect of p53 overexpression on SC35 protein expression levels compared with the decrease in SC35 protein levels induced by CPT treatment for 18 h (Fig. 5G).

View larger version (65K): [in this window] [in a new window]   FIG. 6. (A) Effects of CPT and p53 overexpression on the E5+/E5– ratio. MCF7 cells were treated or not with CPT for 18 h or transfected with a p53 expression vector. qPCR was performed to quantify E5+ and E5– levels using RNAs prepared from N fractions. The CPT effects on the E5+ splice variant were compared with the CPT effects on the E5– splice variant. Similarly, the effects of p53 overexpression on the E5+ splice variant were compared with the effects of p53 overexpression on the E5– splice variant. (B) Cotransfection of p53 and SC35 expression vectors. Total RNAs were prepared from HeLa cells transfected or not with p53 and SC35 expression vectors as indicated. qPCR was performed to measure the levels of the H-ras E5+ splice variant. The values in the control experiments (transfection of control expression vectors) were designated 1. (C) p53 overexpression increased the levels of the H-ras E5+ splice variant in both the N and C fractions. N and C fractions were prepared from MCF7 and HeLa cells transfected or not with a p53 expression vector. qPCR was performed to measure the levels of the H-ras E5+ splice variant. (D) Cycloheximide treatment did not further increase the expression levels of the H-ras E5+ splice variant made in cells transfected with a p53 expression vector. Total RNAs were prepared from HeLa cells transfected or not with a p53 expression vector as indicated and treated or not with cycloheximide 3 h before harvesting the cells. qPCR was performed to measure the levels of the H-ras E5+ splice variant. The values in the control experiments (transfection of control expression vectors in absence of cycloheximide) were designated 1. (E) Effect of cycloheximide and p53 overexpression on the expression levels of various NMD-target splice variants in HeLa cells. Total RNAs were prepared from HeLa cells transfected or not with a p53 expression vector as indicated and treated or not with cycloheximide 3 h before harvesting the cells. (F) Effect of cycloheximide and CPT treatment on the expression levels of various NMD-target splice variants in MCF7 cells. Total RNAs were prepared from MCF7 cells treated or not with cycloheximide for 6 h or with CPT for 18 h as indicated.

Because we observed a very strong effect of p53 overexpression in HeLa cells on the H-ras NMD transcript levels in the total RNA population (Fig. 6B), we hypothesized that p53 did not alter the levels of this transcript in the nuclear RNA population only. Nuclear and cytosolic RNAs were prepared from HeLa cells transfected with a p53 expression vector. As shown on Fig. 6C, p53 overexpression resulted in a 10-fold increase in the H-ras E5⁺ NMD transcript in the C fraction, where this transcript is normally degraded by the NMD pathway (Fig. 2). To a lesser extent, this transcript was increased in the C fraction of MCF7 cells transfected with a p53 expression vector (Fig. 6C). This result correlated with the increase in the H-ras E5⁺ NMD transcript levels in the cytosol following CPT treatment of MCF7 cells for 18 h (Fig. 3F).

To test whether the accumulation of the H-ras E5⁺ transcripts in the C fraction following p53 overexpression in HeLa cells was due to a transient accumulation of such transcripts independently of an inhibition of the NMD pathway, we investigated the effect of cycloheximide in HeLa cells transfected by a p53 expression vector. As shown on Fig. 6D, treatment with cycloheximide for 3 h 1 day after transfection of HeLa cells with a p53 expression vector did not further increase the expression of the H-ras E5⁺ transcripts induced by p53. This result suggests that the degradation of the H-ras E5⁺ transcripts by the NMD pathway was already inhibited by p53 overexpression before the cycloheximide treatment.

To test whether the NMD pathway in general was inhibited, we looked at other NMD-target transcripts. As shown on Fig. 6E, p53 overexpression clearly resulted in an increase in the H-ras NMD-target transcript in total RNA population (compare lanes 1 and 2), while it did not increase the levels of CA150 and PTGS2 NMD-target transcripts (compare lanes 1 and 2). Meanwhile, these NMD targets are all stabilized by a cycloheximide treatment (compare lanes 1 and 3). Because the effects on the cytosolic levels of the H-ras E5⁺ transcripts of p53 overexpression corroborated with the effects of long-term treatment with CPT, which also resulted in an increase in the cytosolic levels of the H-ras E5⁺ transcripts, we tested whether the NMD pathway in general was inhibited in MCF7 treated with CPT for 18 h. As shown on Fig. 6F, cycloheximide treatment stabilized all of the NMD transcripts tested and CPT increased only the levels of the H-ras NMD splice variant. In conclusion, our results indicate that a short-term treatment with CPT resulted in an increase in the levels of the H-ras NMD splice variant only in the nucleus and not in the cytosol, where it is degraded by the NMD pathway. A long-term treatment, as well as p53 overexpression, resulted also in an increase in the levels of this transcript in the cytosol, which suggests that the degradation of this transcript by the NMD pathway was inhibited under such conditions.

p53 and SC35 regulate a subset of NMD-target splice variants in an opposite manner. Although CPT treatment did not increase the expression of all the NMD targets tested, we investigated whether our observations on H-ras gene expression by CPT could be extended to other genes. We selected the caspase 2 (Casp2), Clk/sty (Clk), and SC35 genes because they produce NMD-target splice variants and topoisomerase inhibitors have been reported to affect their expression (23, 35, 40, 44). We also selected CA150 gene, which produces a predicted NMD-target splice variant, and PIG3 gene as a control, which produces several splice variants that are not predicted to be NMD targets (see Fig. S3 in the supplemental material). The introduction of a PTC resulted from the inclusion of an exon in the cases of H-ras, Casp2, and CA150 and from the exclusion of an exon or part of an exon in the cases of Clk and SC35 (see Fig. S3 in the supplemental material). To facilitate the description of the results, the NMD-target splice variants were named SV^NMD and are marked by arrows in Fig. 7. The splice variants encoding protein isoforms were named SV^CDS. We designed several primer sets to coamplify the SV^NMD and SV^CDS of each of the selected genes, and we verified that the selected SV^NMD were subject to NMD by using siRNAs that target NMD effectors and by using cycloheximide treatment (Fig. 7 and see Fig. S3 in the supplemental material).

View larger version (42K): [in this window] [in a new window]   FIG. 7. p53 and SC35 regulate a subset of NMD-target splice variants in an opposite manner. (A) Effect of cycloheximide on various SVNMD/SVCDS ratios. Different cellular fractions were prepared from cells treated or not with cycloheximide for 6 h and used to coamplify SVNMD and SVCDS produced from different genes. The SVNMD are marked by an arrow on the right. (B) Effect of CPT on different nuclear SVNMD/SVCDS ratios. N fractions were prepared from cells treated or not with CPT for 6 h and used to coamplify SVNMD and SVCDS produced from different genes. (C) Effects of siP53 and siSC35 on different nuclear SVNMD/SVCDS ratios in the presence of CPT. N fractions were prepared from cells transfected with siP53, siSC35, or siGAPDH and treated or not with CPT for 6 h and used to coamplify SVNMD and SVCDS produced from different genes. (D) Effects of siSC35 on different nuclear SVNMD/SVCDS ratios in the absence of CPT. N fractions were prepared from cells transfected with siSC35 or siGAPDH and used to coamplify SVNMD and SVCDS produced from different genes. (E) SC35, SF2/ASF, SRp40, and SRp55 binding motifs in alternatively spliced exons. The search of the SC35, SF2/ASF, SRp40, and SRp55 binding motifs was done using ESEfinder within cassette exons generating or not a PTC (categories "NMD" and "no NMD," respectively). For each category, a sequence was constituted by assembling the sequences of each corresponding cassette exon after each other. A score was calculated by dividing the number of sites by the length of the analyzed sequence as described in Materials and Methods.

As previously reported (40), CPT slightly increased the level of the Casp2 splice variant that contains the 61-nucleotide supplementary exon 9 (Fig. 7B). Because this exon generates a PTC, CPT increased the nuclear Casp2^NMD/Casp2^CDS ratio, like in the case of the H-ras gene. However, CPT had no effect on CA150 and SC35, and it decreased the nuclear Clk^NMD/Clk^CDS ratio due to an increase in the nuclear Clk^CDS spice variant levels (Fig. 7B and see Fig. S3 in the supplemental material).

Interestingly, while p53 depletion inhibited the CPT effect in the cases of H-ras and Casp2, it enhanced it in the case of Clk (compare lanes 9, 10, and 11, Fig. 7C). In all three cases, p53 depletion decreased the SV^NMD/SV^CDS ratio (compare lanes 10 and 11, Fig. 7C). Strikingly, SC35 depletion had the opposite effect from that of p53 depletion (compare lanes 11 and 12, Fig. 7C). Overall, the splicing pattern obtained in the absence of p53 was very different from the splicing pattern obtained in the absence of SC35. Therefore, we identified a subset of genes whose SV^NMD/SV^CDS transcript ratio was inversely regulated by p53 and SC35.

As mentioned above, the introduction of a PTC resulted from the inclusion of an exon in the cases of H-ras and Casp2 and it resulted from the exclusion of an exon in the case of Clk (see Fig. S3 in the supplemental material). Remarkably, SC35 depletion increased the SV^NMD/SV^CDS ratio in all three cases even in the absence of CPT (compare lanes 13 and 14, Fig. 7D). Because the H-ras, Casp2, and Clk alternatively spliced exons contain several SC35 binding sites (see Fig. S3 in the supplemental material) predicted by ESEfinder (6), we tested by bioinformatic analysis whether alternatively spliced exons that generate a PTC are enriched in SC35 binding sites. We selected alternatively spliced exons that generate or don't a PTC within our FAST DB database (9). Comparing alternatively spliced exons that do or do not generate a PTC (1,072 and 5,489 exons, respectively), we showed that both families contained the same number of SC35 binding sites, using ESEfinder analysis (Fig. 7E). We found no differences by looking at other splicing factor binding sites (Fig. 7E).

Because of the role of p53 in the stimulated production of the H-ras NMD-target splice variant (Fig. 4 and 5C), we also tested whether p53-target genes produced more NMD-target splice variants. We selected predicted p53-target genes within our FAST DB database (9) as previously described (19). Among 774 alternatively spliced exons from predicted direct p53-target genes, 123 (16%) generated a PTC. Similarly, among 5,787 alternatively spliced exons from genes not predicted to be direct p53-target genes, 949 (16%) generated a PTC. This result suggests that predicted p53-target genes do not produce more NMD targets than other genes.

Upf1 is required for optimal cellular response to CPT. Because CPT selectively stimulated the production of the H-ras NMD-target splice variant in a p53-dependent manner (Fig. 3 and 4), we investigated further the interplay between the p53 and the NMD pathways. Thus, we tested CPT effects under conditions in which the NMD pathway was inhibited by transfecting siUpf1 and siBTZ 48 h before a 6-h CPT treatment. We observed that CPT and siBTZ had an additive effect on the global E5⁺ levels (comparing siGAPDH and siBTZ lanes, Fig. 8A). This additive effect was due to both an increase in the cytosolic E5⁺ levels induced by siBTZ (Fig. 8B) and an increase in the nuclear E5⁺ levels induced by CPT (Fig. 8C).

View larger version (46K): [in this window] [in a new window]   FIG. 8. Upf1 is required for optimal cellular response to CPT. (A) Effects of siUpf1 and siBTZ on E5+ levels induced by CPT. Cells transfected with siGAPDH, siUpf1, or siBTZ were treated or not with CPT for 6 h. The E5+ level was quantified by qPCR. (B) Effects of siUpf1 and siBTZ on the cytosolic E5+ levels induced by CPT. Panel B is the same as panel A using C fractions. (C) Effects of siUpf1 and siBTZ on the nuclear E5+ levels induced by CPT. Panel C is the same as panel A using N fraction. (D) Effects of siUpf1 and siBTZ on the nuclear E5+/E5+ ratio modulated by CPT. PCR was performed to coamplify E5+ and E5– using an antisense primer overlapping the last exon-exon junction and a sense primer in exon 4 (H-rasE4) or in exon 2 (H-rasE2). (E) Effects of siUpf1 on p53 protein and mRNA levels in response to CPT. MCF7 cells transfected with siGAPDH, siUpf1, siBTZ, or siP53 were treated or not with CPT for 6 h. WB analysis of p53 and actin (upper panel) and amplification of the p53 mRNA (lower panel) were performed. (F) Effect of wortmannin on p53 protein levels in response to CPT. MCF7 cells were pretreated or not with wortmannin for 30 min and then treated or not with CPT for 6 h. WB analysis of p53 and actin as a control was performed. (G) Effects of siUpf1 and siP53 on different nuclear SVNMD/SVCDS ratios modulated by CPT. N fractions were prepared from cells transfected with siUpf1, siP53, or siGAPDH and treated or not with CPT for 6 h and used to coamplify different SVNMD and SVCDS. (H) Effect of siUpf1 on p21waf1 mRNA expression levels in response to CPT. Quantification of p21waf1 mRNA levels by qPCR using an RNA preparation from MCF7 cells transfected with siGAPDH, siUpf1, or siP53 and treated or not with CPT for 6 h. The value obtained with siGAPDH in the absence of CPT was designated 1. Experiments were performed at least three times. Data are presented as the mean ± standard error of the mean. (I) Outcome of the activation of the H-ras locus by p53 in response to CPT. The H-ras gene produces different splice variants that have different fates and that result in the production of different protein isoforms. In this report, we demonstrated that CPT treatment increased the synthesis of the unproductive E5+ splice variant (degraded by the NMD process) through the stabilization of p53 protein, which activated the H-ras gene, and through the alteration of SC35 splicing factor, which promoted exon 5 inclusion. It can be anticipated that the impacts on the H-ras locus of p53 or other p53 family proteins may have different outcomes, depending on the nature of the genotoxic stress that activates the p53 pathway, on the cellular context, and, at the molecular levels, on alternative splicing regulation.

Because siUpf1 inhibited the CPT effect similarly to the inhibition by siP53 (comparing Fig. 4D and 8C), we tested whether Upf1 depletion affects the p53 protein levels. As already shown, CPT increased p53 protein levels (siGAPDH lanes, Fig. 8E); this effect was strongly affected by siUpf1 and to a much lower extent by siBTZ (siUpf1 and siBTZ lanes, Fig. 8E). While siP53 strongly decreased p53 mRNA levels, as expected, siUpf1 and siBTZ had no effect on p53 mRNA levels in the absence or presence of CPT (lower panel, Fig. 8E). Because the primers used to amplify the p53 mRNA covered the full ORF (see Fig. S3 in the supplemental material), this observation suggests that Upf1 effect on p53 protein level induced by CPT is not due to an alteration of p53 mRNA expression.

It is noteworthy that a wortmannin pretreatment of MCF7 cells also reduced the accumulation of p53 protein in response to CPT (Fig. 8F). This observation is in agreement with a recent report that demonstrated that SMG-1 was required for optimal p53 activation by ionizing radiation (4). Altogether, our data indicate that the activation of the p53 pathway by CPT is disturbed under conditions in which Upf1 is altered.

Supporting a general alteration of the p53 pathway in the absence of Upf1, siUpf1 had similar effects to those of siP53 on the CPT-mediated effects on H-ras, Casp2, and Clk gene expression (Fig. 8G). In addition, siUpf1 altered the CPT effect on the level of the p21^waf1 mRNA, which is a classic target of p53 (Fig. 8H). CPT increased about sixfold the p21^waf1 mRNA levels, and this effect was abolished by siP53, as expected, and was decreased by siUpf1 (approximately threefold). This demonstrates that Upf1 was required for the cellular response to CPT mediated by p53. In conclusion, modulation of the p53 pathway resulted in the modulation of the production of a subset of NMD targets and p53 activation was disturbed when the NMD pathway was altered, which demonstrate a complex link between the mRNA and DNA surveillance pathways.

	DISCUSSION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
The H-ras proto-oncogene produces a splice variant subject to cytoplasmic NMD. The H-ras gene belongs to a family of proto-oncogenes, which are the most frequently mutated or overexpressed genes in cancer (29). The ras proteins act as molecular switches of signaling pathways that modulate cell proliferation, survival, and apoptosis. In this study, we characterized the expression of four different H-ras splice variants that result from the alternative splicing of exon E5 and intron i6 in the MCF7 breast cancer cell line (Fig. 1 and see Fig. S1 in the supplemental material).

The splice variants that contained the last intron i6 (i.e., E5⁺i6⁺ and E5^–i6⁺) were poorly expressed in MCF7 cells (Fig. 1B). The splice variant that contained neither E5 nor i6 (i.e., E5^–) was particularly abundant in the cytosol and produced the p21^hras protein (Fig. 1C and 3B and see Fig. S1 in the supplemental material). The splice variant that contained E5 but not i6 (i.e., E5⁺) was abundant in the nuclear RNA population but was poorly detected in the cytosolic RNA population compared with the E5^– splice variant (Fig. 1). The nuclear E5⁺ splice variant was polyadenylated and did not contain intronic sequences (Fig. 1 and see Fig. S1 in the supplemental material). This PTC-containing splice variant (Fig. S1) was stabilized in the cytosol under conditions in which the NMD pathway was inhibited (Fig. 2). This demonstrates that this H-ras E5⁺ splice variant was a mature mRNA exported in the cytosol, where it was degraded by the NMD pathway. Thus, the H-ras E5⁺ splice variant is a novel example of unproductive splicing and a novel example of a mammalian cytoplasmic NMD target (11, 26, 37, 43). We verified that the production of the H-ras NMD-target splice variant occurred in the absence of mutation by sequencing the H-ras gene in MCF7 cells. Therefore, to our knowledge, H-ras is the first example of a major nonmutated oncogene that produces a "natural" NMD-target splice variant. In addition, we provide evidence that the production of this NMD target is modulated by the genotoxic stress inducer CPT and the tumor suppressor p53.

Regulation of the production of the H-ras NMD-target splice variant by CPT, p53, and SC35. Our observations that E5⁺ was as abundant as E5^– in the nuclear RNA population and that this polyadenylated mRNA did not contain intronic sequences (Fig. 1; also see Fig. S1 in the supplemental material) suggest that E5⁺ did not result from splicing errors and was produced as much as E5^–. We found evidence that the E5⁺ splice variant plays a role in H-ras gene expression regulation by observing that CPT selectively stimulated the production of this splice variant in a transcription-dependent manner and increased the nuclear E5⁺/E5^– ratio (Fig. 3 and 4). Because this effect was reversed by CPT withdrawal (Fig. 3) and was mediated by p53 and SC35 proteins (see below), our results demonstrate that the production of an NMD-target splice variant can be regulated by a cellular signaling pathway.

Several of our results suggested that the CPT effect on H-ras splicing was mediated in part by SC35 inhibition. First, CPT and SC35 depletion increased the E5⁺/E5^– ratio (Fig. 5), while the overexpression of SC35 promoted H-ras exon 5 skipping (Fig. 5F). Supporting a direct role of SC35 in H-ras splicing, H-ras exon 5, which contains several SC35 binding sites (see Fig. S3 in the supplemental material), was shown to be bound by SC35, which participated in H-ras splicing in a minigene assay (16). It is noteworthy that H-ras exon 5 and Casp2 exon 9 had more than 50% identity in an alignment of 62 nucleotides (see Fig. S3 in the supplemental material) and were regulated by CPT and SC35 in the same way (Fig. 7). CPT favored Casp2 exon 9 inclusion (Fig. 7B and C) as previously described (40, 41). SC35 depletion favored Casp2 exon 9 inclusion (Fig.