SF2/ASF transforms immortal rodent cells

We then tested whether SR or hnRNP A/B proteins have oncogenic activity in several assays. First, we determined whether their overexpression is sufficient to transform immortal NIH 3T3 and Rat1 fibroblasts. By expressing T7-tagged human complementary DNAs in a retroviral vector, we generated stably transduced pools of cells that slightly overexpress (less than two-fold) selected SR proteins or hnRNP A1 (Fig. 2a and Supplementary Fig. 2 online). These tagged versions of the splicing factors have the same splicing, shuttling and translational activities as their untagged counterparts^4,22,23. NIH 3T3 cells that overexpressed SF2/ASF formed colonies in soft agar, whereas the overexpression of two other SR proteins, SC35 and SRp55, resulted in fewer colonies (50% and < 10%, respectively; Fig. 2b). Overexpression of SF2/ASF also promoted anchorage-independent growth of Rat1 cells, whereas three other overexpressed SR proteins did not (Supplementary Fig. 2). Finally, overexpression of hnRNPA1 in NIH 3T3 but not Rat1 cells also induced colony formation in soft agar (Fig. 2b and Supplementary Fig. 2). Thus, SF2/ASF is oncogenic in both cell lines, whereas the other splicing factors seem to have context-dependent or no oncogenic activity.

Open in a separate window

Figure 2

SF2/ASF transforms immortal cells and is tumorigenic in nude mice. (a) Total proteins from duplicate pools of NIH 3T3 stable cell lines transduced with retroviruses expressing LacZ or T7-tagged SF2/ASF, SRp55, SC35 or hnRNP A1 were analyzed by western blotting with anti-T7. The first four samples were also analyzed with anti-SF2/ASF (below) to compare the expression of endogenous and transduced SF2/ASF (b) Quantification of soft-agar colony formation by the stably transduced cell lines. The mean ± s.d. for each pair of pooled lines is shown. P-values in pairwise comparisons to the LacZ control: *P=4 × 10⁻⁸; **P = 10⁻⁷; ***P = 7 × 10⁻⁸. (c) Tumor growth curve in mice injected with 2 × 10⁶ cells from the indicated NIH 3T3 pooled lines. The number of tumors formed per number of injections is shown in parentheses. SRp55 and SC35 cell lines (six each) did not form tumors during the same time course (data not shown). Error bars, s.d. (d) Representative mice injected with LacZ-expressing control cells, or with SF2/ASF-overexpressing cells. (e) Light micrographs of formalin-fixed, paraffin-embedded tissue sections from tumors derived from NIH 3T3 cells overexpressing SF2/ASF of hnRNP A1, stained with hematoxylin and eosin. Scale bars, 100 μm.

NIH 3T3 cells that stably overexpressed the above factors were then injected into nude mice. The cells that overexpressed SF2/ASF gave rise to large tumors (Fig. 2c,d). Overexpression of hnRNP A1 also resulted in tumors, but they were smaller and developed more slowly (Fig. 2c). In contrast, neither SC35 overexpression nor SRp55 overexpression in NIH 3T3 cells gave rise to tumors (data not shown). Tumors derived from NIH 3T3 cells that overexpressed SF2/ASF or hnRNP A1 showed histological characteristics of high-grade sarcomas with spindleshaped morphology (Fig. 2e). The similar morphology of these tumors indicates that their different growth rates (Fig. 2c) might reflect differences at the molecular level.

Effects on apoptosis and proliferation

Inhibition of apoptosis is an important step in tumorigenesis, so we next investigated whether SF2/ASF overexpression can inhibit this process in primary cells. The overexpression of SF2/ASF had a protective effect against apoptosis induced by serum starvation or DNA damage in E1A-expressing mouse embryo fibroblast (MEF) cells, which are sensitized to apoptosis²⁴ (Fig. 3a,b). Overexpression of SC35 was also protective, whereas overexpression of SRp55 or hnRNP A1 had little or no protective effect. Overexpression of SF2/ASF did not prevent E1A-mediated induction of p53 and its target gene MDM2 (Fig. 3c). In addition to its antiapoptotic effects, SF2/ASF also cooperated with oncogenic H-Ras to enhance proliferation (Fig. 3d).

Open in a separate window

Figure 3

SF2/ASF overexpression protects E1A-sensitized MEF cells against apoptosis and enhances the proliferation of Ras-transformed cells. Primary wild-type MEF cells were transduced with retroviruses expressing the indicated human splicing factor cDNAs, or with the empty vector; some cells were cotransduced with adenovirus E1A or activated Ras, as indicated. Error bars indicate s.d. (a) After drug selection, cells coinfected with E1A were plated and serum-starved, and cell death was measured by trypan blue staining. P-values in pairwise comparisons to the pBABE control: *P = 5× 10⁻⁸; **P = 5 × 10⁻⁵; n = 3. (b) Cells coinfected with E1A were plated and treated with the indicated adriamycin concentrations after 24 h, and cell death was measured as in a. *P = 0.003, **P = 0.01, ***P = 0.006; n = 3; for SF2/ASF at low adriamycin, n = 1. (c) MEF cells were infected with the indicated retroviruses, plated after single or double selection (10⁶ cells per 10-cm plate) and lysed in SDS after 24 h. Western blots were carried out using the indicated primary antibodies. The reduction in SRp55 in the presence of E1A was not observed with other cell lines (data not shown). (d) p53^−/− MEF cells were transduced with the empty vector or with retroviruses expressing SF2/ASF or SRp55, either alone or with oncogenic Ras. Cells were fixed at 48-h intervals and stained with methylene blue. Each point represents the mean relative absorbance from six wells.

Effects on alternative splicing of endogenous targets

SF2/ASF is essential for cell viability and embryonic development and has pleiotropic activities, though it has been characterized most extensively as a splicing factor that affects alternative splicing in a concentration-dependent manner²⁵. We therefore examined the splicing patterns of transcripts from selected endogenous genes with roles in transformation and apoptosis, for which alternatively spliced isoforms have been characterized (Supplementary Table 1 online). These genes include known or putative tumor suppressors, such as BIN1, BRCA1, MGEA6, TSC2, VHL and TP73, and proto-oncogenes, such as ERBB2, RPS6KB1 (S6K1), MYBL2 (B-myb) and MYB (C-myb)^7,11,12,26. The BIN1 protein interacts with the product of the MYC (c-Myc) proto-oncogene, suppressing its oncogenic activity, and is inactivated or deleted in various cancers^11,27. The inclusion of exon 12A abolishes the tumor-suppressor activity of BIN1 by interfering with MYC binding, and transcripts that include this exon are overexpressed in melanoma¹¹. Human, mouse and rat cells that stably overexpressed SF2/ASF showed increased inclusion of exon 12A of BIN1 (Fig. 4 and Supplementary Table 1). By contrast, overexpression of SRp55 or hnRNP A1 did not promote inclusion of exon 12A (Fig. 4a). Transient SF2/ASF knockdown by RNAi in HeLa cells induced skipping of exon 12A—seen as a reduction in the levels of the two BIN1 mRNA isoforms that carry this exon—without a marked effect on the two isoforms that lack exon 12A, one of which is abundant in these cells (Fig. 4b). Consistent with this, overexpression of SF2/ASF in an inducible NIH 3T3 mouse cell line increased the inclusion of exon 12A in a dose-dependent manner (Fig. 4c). These results indicate that BIN1 is an in vivo target of regulation by SF2/ASF and that SF2/ASF inactivates this putative tumor suppressor through alternative splicing.

Open in a separate window

Figure 4

Specific alternative splicing changes induced by overexpression of splicing factors or knockdown of SF2/ASF. (a) Human primary cells (IMR90) were infected with the indicated retroviruses, and stable pools were selected. Total RNA was analyzed by radioactive RT-PCR, nondenaturing PAGE and autoradiography using the indicated primers (arrowheads). GAPDH mRNA was analyzed as a control. The exon structure of each BIN1 isoform is indicated. (b) HeLa cells were transfected in duplicate with siRNAs specific for SF2/ASF or PP2Cγ, or mock-transfected. After 72 h, total RNA was analyzed as in a. (c) NIH 3T3 cell lines with inducible expression of SF2/ASF were plated and induced 24 h later with 5 μM ponasterone A for the indicated times. RNA was analyzed by radioactive RT-PCR; quantification of the proportion of exon inclusion (% incl) by phosphorimaging analysis is shown below the ethidium bromide–stained agarose gel. Only two isoforms were detected in these mouse cells. Expressed T7-tagged SF2/ASF and endogenous β-catenin were analyzed by western blotting. (d) Total RNA and protein were extracted from IMR90 primary cells overexpressing the indicated splicing factors. For both panels, total RNA was analyzed by RT-PCR using the indicated isoform-specific primers (arrowheads). Alternatively spliced exons are shaded. Western blotting was carried out as in c. (e) Total RNA and protein were extracted from duplicate samples of HeLa cells treated with siRNAs targeting SF2/ASF or PP2Cγ, or mock-treated. PP2Cγ and SF2/ASF protein levels were analyzed by western blotting. RNA samples were analyzed as in d.

The Ras/MAP kinase (MAPK) and phosphatidylinositol-3-OH kinase (PI3K) pathways are hyperactivated in many human tumors²⁸. We therefore searched for possible splicing targets of SF2/ASF in these two pathways and identified the pre-mRNAs of MNK2 and S6K1 as SF2/ASF splicing targets. Human RPS6KB1 encodes S6K1, a kinase that phosphorylates the small-subunit ribosomal protein S6 and regulates translation²⁹. Only one human RPS6KB1 spliced mRNA has been reported, whereas the mouse gene has two isoforms (http://www.ensembl.org/, version 29). However, we detected similar patterns of S6K1 isoforms in human, mouse and rat cells (Fig. 4d, Supplementary Fig. 1 and Supplementary Table 1). In all three species, SF2/ASF overexpression promoted the expression of isoform-2 mRNA, which is identical to isoform-1 up to exon 6, but encodes a protein with a different C terminus (Supplementary Fig. 3 online). By contrast, overexpression of hnRNP A1 and SRp55 had no substantial effect on the S6K1 isoform distribution (Fig. 4d), whereas the effect of SC35 resembled that of SF2/ASF (Supplementary Table 1). Moreover, knockdown of SF2/ASF increased the expression of isoform-1 and decreased that of isoform-2 (Fig. 4e).

MNK1 and MNK2 are the only known kinases that phosphorylate eIF4E on Ser209 (^{ref. 30}). The MNK2 gene, MKNK2, is alternatively spliced, generating two catalytically active isoforms that are differentially regulated and localized³¹. Overexpression of SF2/ASF resulted in increased expression of the MNK2b isoform and a corresponding reduction in the MNK2a isoform (Fig. 4d and Supplementary Fig. 1), whereas SF2/ASF knockdown caused reduced expression of MNK2b (Fig. 4e).

The levels of SF2/ASF also affected the alternative splicing of other oncogenes, including C-myb and B-myb¹² (Supplementary Table 1). In addition, SF2/ASF overexpression promoted the inclusion of exon 5 in the transcription factor gene TEAD1 (TEF-1) and of exon 2 in the splicing factor gene HNRPA2B1 (Fig. 4d), whereas knockdown of SF2/ASF promoted skipping of these alternative exons (Fig. 4e). In the case of the tumor antigen gene MGEA6, knockdown of SF2/ASF promoted the inclusion of exon 19 (Fig. 4e). Knockdown of SF2/ASF also induced skipping of exons 9 and 10 of the tumor suppressor gene BRCA1, although SF2/ASF overexpression had no effect on this gene (Supplementary Table 1). Overexpression or knockdown of SF2/ASF had no effect on various other alternative splicing events. We conclude that changes in the levels of SF2/ASF result in changes in the alternative splicing of endogenous transcripts from many genes, including several important oncogenes and tumor suppressors. The effects are specific, in that several transcripts tested were not affected, and increasing the levels of other SR proteins did not result in the same spectrum of changes.

SF2/ASF induces an oncogenic isoform of RPS6KB1

As shown above, the alternative splicing of RPS6KB1 (S6K1) is regulated by SF2/ASF, which promotes expression of the isoform-2 mRNA (Fig. 4d,e). We investigated whether this change in isoform mRNA ratios is reflected at the protein level. Mouse and human cells overexpressing SF2/ASF showed elevated levels of the isoform-2 protein (dubbed ‘p31’) (Fig. 5a and Supplementary Fig. 1). S6K1 is a substrate of the serine-threonine kinase mTOR and regulates cell growth and apoptosis³². Thus, we tested whether alternative splicing of mouse Rps6kb1 can affect transformation. We transduced NIH 3T3 cells with retroviruses encoding the two mouse S6K1 isoforms and measured their ability to grow in soft agar. Remarkably, cells that overexpressed isoform-2 formed colonies, whereas cells that overexpressed isoform-1 did not, even though isoform-1 was expressed at much higher levels (Fig. 5b,c). The cells that overexpressed isoform-2 formed fewer colonies than SF2/ASF-overexpressing cells, even though the former expressed higher levels of isoform-2 (Fig. 2b, Fig. 5c and data not shown), indicating that SF2/ASF might activate additional targets that also contribute to transformation. Nevertheless, the expression of S6K1 isoform-2 was sufficient to qualitatively recapitulate the oncogenic effects of SF2/ASF overexpression. We also investigated whether SF2/ASF expression correlated with the S6K1 isoforms in our panel of 50 lung tumors and found a positive correlation between SF2/ASF expression and the isoform-2/isoform-1 ratio (Supplementary Fig. 4 online).

Open in a separate window

Figure 5

SF2/ASF promotes expression of an oncogenic isoform of S6K1 and induces eIF4E phosphorylation. (a) MEF, NIH 3T3 and IMR90 cells were transduced with pBABE or SF2/ASF retroviruses and lysed in SDS. Western blotting was carried out using a monoclonal antibody against the N terminus of S6K1 to detect the endogenous isoforms and one against p-catenin as a loading control. (b) NIH 3T3 cells were transduced with pBABE or S6K1 isoform-1 or isoform-2 retroviruses, and lysates were analyzed by western blotting as in a. The S6K1 antibody reacts with both the endogenous and T7-tagged S6K1 isoforms, but the endogenous isoform-2 is not detected in this exposure. (c) Aliquots of cells in b were seeded in soft agar, and colonies were counted 14 d later. Means ± s.d. are shown; *P = 6×10⁻⁷ compared to the pBABE control. (d) MEF cells were transduced with the indicated retroviruses and analyzed by western blotting with the indicated antibodies.

SF2/ASF induces phosphorylation of eIF4E

The effect of SF2/ASF on the alternative splicing of MKNK2 (MNK2) (Fig. 4d) prompted us to investigate whether the induction of isoform MNK2b changes the phosphorylation state of eIF4E³³. The MAPKs ERK1, ERK2 and p38 are phosphorylated by upstream MAPK kinases, and they in turn phosphorylate MNK1 and MNK2, which is usually required for full activity of MNK1 and MNK2 kinases³¹. Remarkably, although SF2/ASF did not enhance phosphorylation of ERK1/2 or p38, it strongly enhanced phosphorylation of eIF4E on Ser209 (Fig. 5d and data not shown). Consistent with its effect on MKNK2 splicing (Fig. 4d), SRp55 also enhanced phosphorylation of eIF4E, although to a lesser extent than SF2/ASF (Fig. 5d).

Mechanism of SF2/ASF-driven transformation

Our working hypothesis has been that increased levels of SF2/ASF elicit concentration-dependent changes in the alternative splicing of certain key target genes and that the combined effects of the resulting isoforms drive the multistep transformation process.

We therefore investigated the expression of several endogenous genes that are involved in transformation and apoptosis and that are known to be regulated by alternative splicing. Many of these genes were not affected by overexpression or knockdown of individual SR or hnRNP A1 proteins, at least within the transcript regions we probed (Supplementary Table 1). However, we identified several cancerrelated genes whose patterns of isoform expression were strongly affected by SF2/ASF levels, including the putative tumor suppressor BIN1, the transcription factor TEF-1, the kinase MNK2 and the ribosomal-protein kinase S6K1 (Figs. 4 and ​and6,6, Supplementary Fig. 1 and Supplementary Table 1). These observations markedly expand the number of known endogenous alternatively spliced gene targets of SR proteins^20,35–38.

The Ras-MAPK and PI3K-mTOR pathways are deregulated in many cancers, contributing to the establishment and maintenance of the transformed phenotype^28,39. Therefore, we investigated whether SF2/ASF can modulate these signaling pathways. We found that alternative splicing of two downstream components of these pathways, MNK2 and S6K1, is deregulated by SF2/ASF overexpression, leading to activation of these enzymes independently of upstream signaling activity.

The effects of SF2/ASF on the alternative splicing of MKNK2 link SF2/ASF to the Ras-MAPK pathway. Notably, activation of eIF4E phosphorylation by SF2/ASF bypasses the proximal steps of this pathway, so that SF2/ASF activates MNK2b splicing and eIF4E phosphorylation downstream of MAPK activation (Fig. 5d). As eIF4E phosphorylation contributes to its oncogenic activity⁴⁰, modulation of MNK2 splicing is probably an important branch of SF2/ASF's transforming activity (Fig. 8).

Open in a separate window

Figure 8

A model for transformation by SF2/ASF. SF2/ASF (which can be blocked by specific shRNAs) activates several key targets that contribute to the transformed phenotype: induction of MNK2b promotes phosphorylation of eIF4E, which enhances cap-dependent translation; induction of the oncogenic S6K1 isoform-2 directly leads to transformation; and induction of inactive isoforms of the putative tumor suppressor BIN1 inhibits apoptosis. Other SF2/ASF targets are also likely to contribute to transformation.

Overexpression of SF2/ASF also led to altered isoform expression of S6K1, which is involved in cell growth and cell death³² (Fig. 4d,e, Fig. 5a and Supplementary Fig. 1). SF2/ASF had a strong effect on the expression of mouse, human and rat S6K1 isoforms with distinct C termini (Fig. 4d,e and Supplementary Figs. 1 and 3). SF2/ASF increased the expression of isoform-2, which includes three alternative cassette exons between exons 6 and 7 that are skipped in isoform-1, as well as an alternative poly(A) site within intron 7 (Supplementary Fig. 3). This mRNA isoform encodes a 316-residue polypeptide that lacks the regulatory domain at the C terminus of isoform-1, which is phosphorylated by mTOR^29,41, and that has half of its kinase domain altered. The sequences of the two isoforms are identical for the first 5 of 12 conserved kinase subdomains, and they diverge within subdomain VIA and thereafter⁴² (Supplementary Fig. 3). Thus, isoform-2 has the same ATP-binding site as isoform-1 but lacks a conserved aspartate in the presumptive active site, so its functional properties probably differ from those of the previously studied isoform-1 of S6K1, and it is probably inactive as a kinase.

Remarkably, we found that isoform-2 transforms NIH 3T3 cells, whereas isoform-1 does not, even when present at much higher levels (Fig. 5b,c). Thus, p31 has potent transforming activity even as a minor isoform. This result provides one possible molecular explanation for SF2/ASF-mediated transformation: the induction of an oncogenic isoform that is a specific splicing target of SF2/ASF (Fig. 8). The fact that SF2/ASF expression in lung tumors correlates with the S6K1 isoform-2/isoform-1 mRNA ratio (Supplementary Fig. 4) further supports the potential relevance of this alternative splicing event in human tumors. A stronger correlation might be precluded by the effect of other splicing factors on S6K1 splicing (Supplementary Table 1), differences in translational efficiency and protein turnover between the two isoforms, stromal cell contamination and tumor-type heterogeneity. However, we found a clear correlation between SF2/ASF levels and S6K1 isoform mRNA and protein levels in human lung and breast cancer cell lines (Supplementary Fig. 1).

In addition, we found that alternative splicing of mouse Rps6kb1 is essential for SF2/ASF-mediated transformation, as knockdown of isoform-2, but not isoform-1, blocked SF2/ASF-mediated transformation (Figs. 7 and ​and88 and Supplementary Fig. 5). Although we obtained only one shRNA that was specific for S6K1 isoform-2, we observed knockdown of this isoform at the mRNA and protein level in human and mouse cells. The phenotypic effects of this knockdown on transformation were consistent with those of SF2/ASF knockdown and opposite to those of isoform-2 or SF2/ASF overexpression, arguing against off-target effects.

The 17q23 genomic region, which is commonly amplified in breast and other tumors, comprises several putative proto-oncogenes (RPS6KB1, PPM1D and TBX2)^7,21. Our results show that SFRS1 is a novel proto-oncogene that also resides in this region and modulates the oncogenicity of S6K1, the product of RPS6KB1. The physical and functional linkage between SFRS1 and RPS6KB1 indicates that when both genes are coamplified, they might synergize to enhance tumorigenicity, similar to oncogenes that are part of other amplicons^43,44.

In addition to the modulation of MNK2 and S6K1 splicing, we found that SF2/ASF regulates the expression of isoforms of BIN1, which counteracts the MYC proto-oncogene^11,45. SF2/ASF induces the inactive isoforms of this putative tumor suppressor¹¹, indicating that SF2/ASF upregulation can modulate the apoptotic activity of MYC and that SF2/ASF and MYC might cooperate.

It is possible that the effects we observed on alternative splicing of endogenous genes are indirect. However, these effects are consistent with the known properties of SF2/ASF, based on previous biochemical and transfection experiments, including the concentration dependence, shown here by the reciprocal effects of overexpression and RNAi. In some cases, overexpression of one splicing factor might affect expression of another splicing factor through various feedback and cross-regulatory loops⁴⁶, potentially initiating a regulatory cascade. For example, overexpression of SF2/ASF partially abrogated expression of endogenous SF2/ASF (Fig. 5d and Supplementary Fig. 2) and affected the alternative splicing of hnRNP A2/B1 (Fig. 4). Some of the phenotypic effects of SF2/ASF and the other splicing factors could involve other aspects of gene expression besides alternative splicing, given their reported effects on post-transcriptional processes such as mRNA turnover, export and translation²⁵.

Knockdown of SF2/ASF can induce genomic instability, including double-stranded DNA breaks and a high rate of mutations⁴⁷. However, it is unlikely that SF2/ASF overexpression triggers a similar response, as we did not observe induction of p53—reflecting DNA damage—although the p53 pathway remained intact, as shown by the ability of E1A to induce p53 and its target gene, MDM2 (Fig. 3c and data not shown). In addition, we did not observe mutations in several cDNAs from cells overexpressing SF2/ASF (data not shown).

Potential roles of other splicing factors in cancer

Although SF2/ASF is the first splicing factor shown to be directly involved in cancer, the effects we observed for some of the other splicing factors in various oncogenic assays, together with their expression patterns, indicate that they might also contribute to cancer development in appropriate contexts. SR proteins share a high degree of homology and have similar splicing activities, although the degree of overlap in their in vivo targets is unknown². Our analysis of endogenous transcripts in cells that stably overexpressed individual SR proteins showed only limited overlap in the alternative splicing events affected by each protein (Fig. 4 and Supplementary Table 1). In some of the functional assays, particular SR proteins had effects similar to those of SF2/ASF: for example, SC35 enhanced colony formation in NIH 3T3 and Rat1 cells and slightly inhibited apoptosis (Fig. 2b, Fig. 3b and Supplementary Fig. 2); and SRp55, which is overexpressed and in some cases amplified in certain cancers (Fig. 1 and data not shown), cooperated with Ras to enhance proliferation (Fig. 3) and anchorage-independent growth (data not shown) and enhanced eIF4E phosphorylation (Fig. 5d). Although each of these oncogenic assays reflects effects on apoptosis and cell proliferation that contribute to transformation, multiple changes are required to achieve tumorigenicity in vivo²⁸. Of the splicing factors we tested, only SF2/ASF had strong transforming activity, and it was far more tumorigenic in nude mice than hnRNP A1 (Fig. 2), indicating that only SF2/ASF affects alternative splicing of the appropriate set of targets in NIH 3T3 cells in such a way as to promote rapid tumorigenesis. However, some of the related splicing factors might be tumorigenic in other cell types or genetic contexts.

RNA interference

For inhibition of SF2/ASF and PP2Cγ expression, HeLa cells were seeded (7 × 10⁴ cells per well) in six-well plates in antibiotic-free medium. After 24 h, cells were transfected with 200 pmol short interfering RNA (siRNA) per well (Dharmacon), using Oligofectamine (Invitrogen). After a further 72 h, cells were lysed, and protein and RNA were extracted as described below. siRNA target sequences were 5′-TTGACCACTGAAGAAGTCA-3′ (PP2Cγ) and 5′-ACGATTGCCGCATCTACGT-3′ (SF2/ASF); both siRNA strands had 3′ dTdT tails. shRNAs against mouse and human SF2/ASF (SF2-sh1, SF2-sh2), a control shRNA with two mismatches (SF2-sh1^m) and shRNAs against mouse and human S6K1 isoforms 1 and 2 (p70-sh1, p70-sh2, p31-sh1) (Supplementary Table 2) were cloned in the LMP retroviral vector. NCI-H460 and NIH 3T3 cells were transduced and selected as described above.

Immunoblotting

Cells were lysed in SDS and analyzed for total protein concentration as described⁵⁰. We separated 30 μg of total protein from each cell lysate by SDS-PAGE and transferred it onto a nitrocellulose membrane. The membranes were blocked and probed with antibodies, using enhanced chemiluminescence detection. Quantification was done with NIH-Image software (http://rsb.info.nih.gov/nih-image/download.html). Primary antibodies were β-catenin (1:5,000, Transduction Laboratories; or 1:2,000, Sigma); SF2/ASF (mAb AK96 culture supernatant, 1:100)²²; hnRNP A1 (mAb 4B10, 1:1,000); hnRNP A2 (mAb DP3, 1:1,000); SRp55 (mAb 8-1-28 culture supernatant); lamin A-C (1:1,000, Santa Cruz Biotechnology); T7 tag (1:5,000, Novagen); MDM2 (mAb 2A10 culture supernatant, 1:100); p53 (1:1,000 CM5, NovoCastra); S6K1 (N terminus; 1:200, Pharmingen); and phosphor–eIF4E Ser209, eIF4E, phospho-ERK (T202/Y204), ERK1/2 (1:1,000, Cell Signaling Technology). Secondary antibodies were HRP-conjugated goat anti-mouse or anti-rabbit IgG (H+L) (1:10,000, Pierce).

Immunoprecipitation and western blotting

Cells were lysed with RIPA buffer, the lysates were cleared by centrifugation at 12,000g for 15 min, and the supernatant was passed through a 0.45-μm filter. For each cell line, 50 μl of 50% (v/v) protein G–Sepharose beads was incubated with 2 μg of antibody against S6K1 (Pharmingen) for 1 h at 4 °C. The beads were then incubated with 1 mg of total protein from each lysate for 2–3 h. The beads were washed three times with RIPA buffer and boiled for 5 min in 50 μl of 2 × SDS sample buffer. After SDS-PAGE and transfer as above, the membranes were probed with antibodies against S6K1. In some experiments, TrueBlot secondary antibody (1:1,000; eBioscience) was used to avoid detection of the heavy and light chains.

Anchorage-independent growth

Colony formation in soft agar was assayed as described⁵⁰. Plates were incubated at 37 °C and 6% CO₂. After 10–18 d, colonies from ten different fields in each of two wells were counted for each transductant pool, and the average number of colonies per well was calculated. The colonies were stained as described and photographed under a light microscope at 100 × magnification.

Tumors in nude mice

NIH 3T3 stable pools expressing the indicated splicing factors or lacZ, were injected (2 × 10⁶ cells per site in 200 μl) subcutaneously into each rear flank of (NIH nu/nu) nude mice, using a 26-gauge needle. Tumor growth was monitored as described⁴⁸. For NCI-H460 cells expressing SF2/ASF shRNA, nude mice were γ-irradiated (400 rad) before subcutaneous injection with 10⁶ cells per site as above.

Apoptosis experiments

Primary MEFs were infected with the indicated retroviruses, with or without hygromycin-resistant E1A retrovirus. After 72 h of puromycin and 72 h of hygromycin selection, cells were allowed to recover for 24 h and seeded (4 × 10⁴ cells per well) on 12-well plates (Nunc). After 24 h, the medium was replaced with fresh medium containing the indicated adriamycin concentrations, and the cells were incubated for 24 h. For serum starvation, cells were washed with PBS and grown for 24 h in medium containing 0.1% (v/v) FBS. Cells were then trypsinized, and dead cells were stained with trypan blue. These assay conditions induce apoptotic cell death²⁴.

Growth curves

Primary MEFs from wild-type or p53^−/− mice were infected with the indicated retroviruses. After selection, 5,000 cells per well were seeded in 96-well plates. Cells were fixed and stained with methylene blue as described⁵⁰, and the A₆₅₀ of the acid-extracted stain was measured on a plate reader (BioRad).

RT-PCR

Total RNA was extracted with Trizol reagent (Invitrogen) and 2 μg of total RNA was reverse transcribed with SuperScript II (Invitrogen). PCR was performed on 1/10 (2 μl) of the cDNA, in 50-μl reactions containing 0.2 mM dNTP mix, 10× PCR buffer with 15 mM MgCl₂ (Invitrogen), 2.5 units of TaqGold (Invitrogen) and 0.2 μM of each primer; 5% (v/v) DMSO was included in some reactions. For semiquantitative PCR, primers were end-labeled with [γ-³²P]ATP and polynucleotide kinase (NEB). PCR conditions were 95 °C for 5 min, then 35 cycles of 94 °C for 30 s, 57 °C for 30 s and 72 °C for 45 s, followed by 10 min at 72 °C. For radioactive PCR, 30 cycles were used. PCR products were separated on 1.5% or 2% agarose gels or on 6% nondenaturing polyacrylamide gels.

Quantitative PCR

DNA copy number was quantified using ABI 7700 or ABI 7900 Sequence Detection Systems (Applied Biosystems). We determined the mRNA levels of SF2/ASF, hnRNPA1, SRp55 and S6K1 isoforms in tumors and in the corresponding normal tissues (Clontech) by performing quantitative PCR with fluorogenic TaqMan probes, directly after reverse transcription. Absolute mRNA levels for SF2/ASF, hnRNP A1 and SRp55 were within 50% for five samples of each normal tissue. Actin was used as a reference probe.

Primers

For primers used, see Supplementary Table 2.

We thank G. Dreyfuss (University of Pennsylvania) for antibodies to hnRNP A1 and A2; D. Pant for advice on statistical tests; A. Rosenberg, M. Nolan and S. Muthuswamy (Cold Spring Harbor Laboratory) for the breast-cancer cell lines; and M. Hemann, L. Zender, R.-M. Xu, A. Neuwald, S. Powers and R. Lucito for helpful discussions. This work was supported by grant CA13106 from the US National Cancer Institute.