Regulation of Cyclin D1 RNA Stability by SNIP1

Cyclin D1 expression represents one of the key mitogen-regulated events during the G₁ phase of the cell cycle, whereas Cyclin D1 overexpression is frequently associated with human malignancy. Here, we describe a novel mechanism regulating Cyclin D1 levels. We find that SNIP1, previously identified as a regulator of Cyclin D1 expression, does not, as previously thought, primarily function as a transcriptional coactivator for this gene. Rather, SNIP1 plays a critical role in cotranscriptional or posttranscriptional Cyclin D1 mRNA stability. Moreover, we show that the majority of nucleoplasmic SNIP1 is present within a previously undescribed complex containing SkIP, THRAP3, BCLAF1, and Pinin, all proteins with reported roles in RNA processing and transcriptional regulation. We find that this complex, which we have termed the SNIP1/SkIP–associated RNA-processing complex, is coordinately recruited to both the 3′ end of the Cyclin D1 gene and Cyclin D1 RNA. Significantly, SNIP1 is required for the further recruitment of the RNA processing factor U2AF65 to both the Cyclin D1 gene and RNA. This study shows a novel mechanism regulating Cyclin D1 expression and offers new insight into the role of SNIP1 and associated proteins as regulators of proliferation and cancer. [Cancer Res 2008;68(18):7621–8]

Cyclin D1, an important regulator of the G₁ phase of the cell cycle, is controlled at multiple levels, including gene transcription, protein stability, and subcellular localization (1). Moreover, its aberrant expression is frequently associated with human malignancy (1). Previously, Cyclin D1 expression has been shown to be regulated, at least in part, by SMAD nuclear interacting protein-1 (SNIP1; ref. 2), a widely expressed and evolutionarily conserved factor originally identified by a yeast two-hybrid screen searching for SMAD interacting proteins (3). SNIP1 is a 396-amino-acid (aa) protein that contains a nuclear localization sequence, coiled-coil motif and, within its COOH terminus, a forkhead-associated (FHA) domain, which are recognized phospho-serine/threonine binding motifs (4). The majority of nucleoplasmic SNIP1 exists in a high molecular weight complex (2) of which neither the function nor composition has been previously described.

A number of functions have been ascribed to SNIP1, suggesting an ability to interact with multiple cellular partners. Although the NH₂ terminus of SNIP1 shares homology with the RNA binding and splicing factors, SRrp86 and p54 (BLAST-link database), the majority of functions described to date have involved transcriptional regulation. When overexpressed, SNIP1 inhibits transactivation by SMADs and the RelA(p65) nuclear factor-κB subunit, by preventing their interaction with p300 (5). Conversely, SNIP1 has a positive effect on c-Myc activity, facilitating p300 recruitment to c-Myc–regulated promoters, while also increasing c-Myc stability through inhibition of Skp-2 mediated ubiquitination (6). An additional role for SNIP1 has also been suggested as a regulator of ATR (ATM and Rad3-related) checkpoint kinase pathways (7). SNIP1 expression is up-regulated in a wide range of tumors and, when overexpressed, SNIP1 can enhance c-Myc and H-Ras–induced cell transformation (6). This suggests that SNIP1 could function as an oncogene.

Consistent with this putative oncogenic function, SNIP1 levels accumulate after serum stimulation and promote cell cycle progression through G₁, which, at least in part, is accomplished through regulation of Cyclin D1 expression (2). We previously showed that, in a variety of cell lines, small interfering RNA (siRNA)-mediated down-regulation of SNIP1 inhibits both Cyclin D1 mRNA and protein levels (2). As SNIP1 has been shown to regulate transcription factor activity, it was previously assumed that this was a transcriptional effect. However, in this study, we have further analyzed the ability of SNIP1 to regulate the endogenous Cyclin D1 gene and, surprisingly, find that the primary mechanism of SNIP1 control of Cyclin D1 expression results from specific regulation of Cyclin D1 RNA stability. We identify components of a novel SNIP1-containing complex that all have reported or suspected roles in RNA processing and transcription and show these also have a role in Cyclin D1 regulation. This study therefore represents both a novel function for SNIP1 as well as a new mechanism of regulation of this key proto-oncogene.

Cells and DNA/siRNA transfection. U-2 OS osteosarcoma, HEK-293 embryonic kidney, and HeLa cervical carcinoma cell lines were grown in 10% FCS (Life Technologies)/DMEM (Lonza) for no more than 30 passages. DNA transfections were performed using PEI (Polysciences), and siRNA duplex oligonucleotides were synthesized by MWG and transfected using Interferin (Polyplus) as per manufacturer's instruction.

Mass spectroscopy. HEK293 cells were transfected with glutathione S-transferase (GST) or SNIP1-GST expression plasmids, GST pull downs were performed, and bound proteins were eluted by boiling in LDS-sample buffer (Invitrogen) before separation on 4% to 12% precast gels (Novex) and colloidal Coomassie staining (Invitrogen). Bands of interest were excised and proteins subjected to in-gel trypsin digestion. Protein mass fingerprint data were obtained by MALDI-Tof-Tof (MS/MS) analysis using a 4700 Proteomic Analyser (Applied Biosystems) or by nano-LC-ESI (MS/MS) using the Ultimate 3000 nLC system (Dionex) coupled to a 4000 QTrap (Applied Biosystems). Peak list files generated from resultant data analysis were submitted to a MASCOT search engine for protein identification.

Plasmids. YFP-SNIP1 and SNIP1-YFP expression plasmids were constructed by inserting SNIP1 cDNA into pEYFPN1/C1 backbones (Clontech). Mammalian expression vectors for THRAP3, BCLAF1, SkIP, and Pinin were constructed by the University of Dundee, College of Life Sciences cloning service, by reverse transcription from cellular mRNA and insertion into pCMV5-HA. THRAP3 (aa 345–509), BCLAF1 (aa 287–450), and Pinin (aa 1–132) fragments were cloned into pQE30 (Qiagen) for bacterial expression, his-tag purification, and antibody production.

Antibodies. The SNIP1 rabbit polyclonal antibodies used for immunoprecipitation and mouse monoclonal SNIP1 antibodies used for Western blotting were described previously by Roche and colleagues (2). Polyclonal rabbit antibodies against THRAP3 (aa 345–509), BCLAF1 (aa 287–450), and Pinin (aa 1–132) were raised by Diagnostics Scotland using purified, recombinant His-tagged protein fragments. Other antibodies used in this article were as follows: anti–Cyclin D1 (PharMingen), anti–β-actin (Sigma), anti-hemagglutinin (12CA5; Cancer Research UK), anti-GST (Amersham-Pharmacia), anti-SkIP (ab23331; Abcam), anti-U2AF65 (Zymed), anti–γ tubulin (Sigma), anti-HDAC1 (sc-6298; Santa Cruz), anti-B23 (Zymed), and anti-Cyclin B1 (V152, Cancer Research UK). Secondary antibodies used for immunofluorescence were α-rabbit Rhodamine Red-X conjugate (Jackson ImmunoResearch) and α-mouse Cy5 conjugate (Jackson ImmunoResearch).

Rapid amplification of cDNA ends-poly(A) test PCR. Rapid amplification of cDNA ends–poly(A) test PCR (RACE-PAT) was performed essentially as described in (8). Briefly, RNA was prepared from U2-OS cells, genomic DNA was removed, RNA was precipitated, and cDNA was synthesized using avian myeloblastosis virus reverse transcriptase and an oligo(dT) anchor primer. PCRs were then performed using the reverse oligo(dT) anchor primer and a forward Cyclin D1 primer sitting 200 nucleotides upstream of the transcript end. PCR products were then run on 2% agarose gels.

RNA-protein cross-linking. RNA-protein cross-linking was performed as described by Niranjanakumari and colleagues (9). Briefly, an actively growing U2-OS cell monolayer was cross-linked with 1% formaldehyde; cross-linking was stopped by 0.25 mol/L glycine; and cells were lysed in radioimmunoprecipitation assay buffer RIPA; (50 mmol/L Tris (pH 7.5), 1% NP40, 0.5% sodium deoxycholate, 0.05% SDS, 1 mmol/L EDTA, 150 mmol/L NaCl, 40 units RNAsin and protease inhibitors). Lysates were then sonicated; insoluble material was removed by centrifugation and precleared with protein-A Sepharose; and immunoprecipitations were performed. Immunoprecipitates were then washed five times in high-stringency RIPA buffer [50 mmol/L Tris (pH 7.5), 1% NP40, 1% sodium deoxycholate, 0.1% SDS, 1 mmol/L EDTA, 1 mol/L NaCl, 1 mol/L urea, 40 units RNAsin and protease inhibitors]. Protein was then eluted [50 mmol/L Tris (pH 7), 5 mmol/L EDTA, 10 mmol/L DTT, 1% SDS]; cross-links were reversed (70°C/45 min); RNA was extracted by Trizol and treated with DNase I (DNA-free kit; Ambion); and bound RNAs were analyzed by reverse transcription-PCR (RT-PCR).

Nucleocytoplasmic RNA distribution. The cellular distribution of RNA was analyzed as described by Rousseau and colleagues (10). Briefly, U2-OS cells were trypsinized, washed in cold PBS, and resuspended by slow pipetting in 10 mmol/L Tris (pH 8.4), 140 mmol/L NaCl, 1.5 mmol/L MgCl₂, 0.5% NP40, and 1 mmol/L DTT and RNAsin. After centrifugation (1,000 rpm per 3 min), the supernatant was taken as the cytoplasmic fraction. Pellets were then suspended in 1/10th volume of 3.3% SDS and 6.6% Tween 20, vortexed, incubated on ice, centrifuged as above, and the resulting supernatant was discarded. RNA from the remaining nuclear pellets (and the above cytoplasmic fraction) were than extracted with Trizol, treated with DNase, and analyzed by RT-PCR.

Other experimental procedures. Western blots, protein extractions, immunoprecipitations, immunofluorescence, and gel filtration procedures were performed as previously described (ref. 2 and references therein). For GST pull downs in Fig. 1A, whole-cell lysates were prepared and 30 units DNase I were added for 30 min, followed by solid urea to a final concentration of 1 mol/L. Extracts were then thoroughly vortexed, Triton-X was added to a final concentration of 1%, and left at 4°C for a further 30 min. This was then snap frozen in a dry ice bath, thawed, and centrifuged to generate a soluble extract, which was then diluted 10-fold into pull down buffer [20 mmol/L HEPES (pH 7.9), 20% glycerol, 75 mmol/L NaCl, 1 mmol/L DTT, protease, and phosphatase inhibitors]. Detailed protocols for chromatin immunoprecipitation (ChIP) assays, cell elutriation, and fluorescence-activated cell sorting (FACS) analysis are described elsewhere (11). RNA was extracted using either Trizol (Sigma) or a Nucleospin RNA II kit (Machinary-Nagel) and cDNA synthesized using a Quantitect RT kit (Qiagen) as per manufacturer's instruction.

Primers for PCR analysis. PCR primers used for RT-PCR analysis are as follows:

• SNIP1 (f): 5′-GAAGAAGCAAGTCTCCTCGC.

• SNIP1 (r): 5′-GGGCTTCACTCTTCGGCC.

• GAPDH (f): 5′-GGTCGTATTGGGCGCCTGGTCACC.

• GAPDH (r): 5′-CACACCCATGACGAACATGGGGGC.

• Cyclin D1 (exon f): 5′-GAGAACAAACAGATCATCCGCA 3′.

• Cyclin D1 (exon r): 5′-GCTTCGATCTGCTCCTGG 3′.

• Cyclin D1 (intron f): 5′-GAGCACATTTTCAGACCTTCGG.

• Cyclin D1 (intron r): 5′-CCTGAAAATGACCCTCGGGC.

• Cyclin D2 (f): 5′-CTACCTTCCGCAGTGCTCCTA.

• Cyclin D2 (r): 5′-CCCAGCCAAGAAACGGTCC.

• rpLP32a (f): 5′-TGTGAAGCCCAAGATCGTCA.

• rpLP32a (r): 5′-GTCAATGCCTCTGGGTTTCC.

The following primers were used in ChIP assays covering regions of the Cyclin D1 promoter and gene:

• 1f: 5′-GCAGAGCTGCCATAGCACGT.

• 1r: 5′-GTGCAGGGACCAGTCCCGA.

• 2f: 5′-CAACGTGGTTACCCTCCCGA.

• 2r: 5′-CCTCTGGGCTTTTGAGGGTC.

• 3f: 5′-GCTGTGAGCACCCTGGCGA.

• 3r: 5′-GTGTTCGTGGTTACATGAGAG.

• 4f: 5′-GTCTCCCCAGGCTGCGTGT.

• 4r: 5′-GCCCAAAAGCCATCCCTGAG.

• 5f: 5′-GGAAGATCGTCGCCACCTGG.

• 5r: 5′-GGCACTGGCTTCTCCCGAAG.

• 6f: 5′-GAGCACATTTTCAGACCTTCGG.

• 6r: 5′-CCTGAAAATGACCCTCGGGC.

• 7f: 5′-CTCCGTAGGTCTGCGAGGAAC.

• 7r: 5′-CCAGTGGTTACCAGCAGCTC.

• 8f: 5′-CCCTGATGGCCGCTCACC.

• 8r: 5′-CTTCCCAGACAGCACCCAG.

• 9f: 5′-GTGACTTACGGCTGCTTAAAG.

• 9r: 5′-CTCCAGGATTTGACCCTATCC.

• 10f: 5′-CGTGATAACATTCACAAGGCTCG.

• 10r: 5′-GGTATCTGAGGCTCCACATCC.

• 11f: 5′-CTGCCTCAGATGTGAAGTTCA.

• 11r: 5′-GCACCAGCCTCGGCATTTCC.

• 12f: 5′-CCTGGGATGACTGTGTCTTTC.

• 12r: 5′-AGCATCTGCTGAGACCCGTG.

• 13f: 5′-CTCAGGTTCAGAGGAGGCAG.

• 13r: 5′-CACGTCGGTGGGTGTGCAAG.

• 14f: 5′-CTCCAGAACACGGCTCACGC.

• 14r: 5′-GTCACAGGACAGACTCCGCTG.

siRNA oligonucleotides. The SNIP and scrambled siRNAs have been published previously (2). The sequences of other siRNAs used are as follows (sense strand only):

• THRAP3 (A): 5′-CAGGAGUUUCGUUCCAUUUTT.

• THRAP3 (B): 5′-GGAGGAGAUGGAUGAUCAATT.

• BCLAF1 (A): 5′-GCCCACUUAGGAUCAAAAUTT.

• BCLAF1 (B): 5′-UCAGGAAGCUCUAGAUUACTT.

• SkIP (A): 5′-GGGUAUGGACAGUGGAUUUTT.

• SkIP (B): 5′-UCCUUGAACUUCCAAUGAATT.

SNIP1 regulates Cyclin D1 RNA processing. Previously, we showed that SNIP1 regulates Cyclin D1 expression (2). SNIP1 has been shown to function as a regulator of transcription, although its NH₂ terminus shares homology with the RNA binding and splicing factors, SRrp86 and p54. We therefore wished to confirm whether its effect on Cyclin D1 was primarily transcriptional in nature. Consistent with our previous data, we found that an siRNA targeting endogenous SNIP1 caused a significant decrease in Cyclin D1 protein and mRNA (Fig. 1A and B). However, using primers specifically designed to amplify prespliced message or postspliced message, we surprisingly found that depletion of SNIP1 had no effect on the levels of prespliced Cyclin D1 RNA (Fig. 1B). This result did not represent an effect on alternative splicing because SNIP1 was also required for production of the Cyclin D1 “b” transcript in which the retention of intron 4 results in a protein with an alternate COOH-terminal sequence (Fig. 1C; ref. 12). A “no reverse transcriptase” control confirmed that PCR signals did not result from contaminating genomic DNA (Fig. 1D).

We next determined which step in Cyclin D1 RNA processing is SNIP1 dependent. RACE-PAT PCR analysis showed that polyadenylation of Cyclin D1 RNA was not affected by SNIP1 depletion (Fig. 2A). Similarly, SNIP1 depletion equally reduced both nuclear and cytoplasmic levels of mature Cyclin D1 mRNA, indicating SNIP1 does not regulate nucleocytoplasmic mRNA shuttling (Fig. 2B). These results suggested that the primary effect on Cyclin D1 expression could be at the level of cotranscriptional splicing or the stability of the nascent transcript. Therefore, we examined the effect of inhibiting transcription using actinomycin D. A very rapid decrease in prespliced Cyclin D1 RNA was observed in the first hours of actinomycin D treatment, consistent with loss of signal as Cyclin D1 gene transcription is inhibited and prespliced message is converted into spliced mRNA (Fig. 2C). Depletion of SNIP1 by siRNA had no effect on this process. By contrast, analysis of the spliced Cyclin D1 mRNA revealed a relatively steady decline in its levels upon transcriptional inhibition, indicative of its stability within cells. Here, depletion of SNIP1 had a significant effect on Cyclin D1 mRNA degradation kinetics, with a rapid decline in postspliced message being seen within the first hours of actinomycin D treatment. However, after this point, the rate of Cyclin D1 mRNA depletion was identical to that seen in control cells. Consistent with our previous results (Fig. 1B), no effect of SNIP1 depletion was observed on Cyclin D2 mRNA stability. Taken together, these data reveal SNIP1 to be a critical posttranscriptional regulator of Cyclin D1 RNA within the nucleus.

SNIP1 exists in a high molecular weight complex containing multiple transcription and RNA processing factors. Previously, we have shown that the majority of SNIP1 is present within an unidentified high molecular weight protein complex of ∼1.5 MDa (2). To learn more about the cellular function of SNIP1, we wished to identify components of this complex. To achieve this, a GST-SNIP1 fusion protein was expressed in HEK 293 cells and affinity purified using glutathione-agarose. Several proteins reproducibly copurified with GST-SNIP1 (Fig. 3A). These were identified by in-gel trypsin digestion and MALDI-Tof-Tof (MS/MS) to be the thyroid hormone receptor–associated protein 150 kDa (THRAP3 or TRAP150), the THRAP3-related protein Bcl-2–associated transcription factor (BCLAF1 or Btf), Pinin, and the Ski-interacting protein (SkIP). Each of these proteins have known or suspected roles in RNA processing and transcription (13–21). Interestingly, THRAP3 has been reported to bind SRrp86 (17), a splicing factor sharing homology with NH₂ terminus of SNIP1. Additionally, SkIP has been previously reported to bind another FHA-domain containing factor, CHES1/FOXN3 (22), as well as the known SNIP1 interactors p300 (23) and the SMADs (24).

Endogenous SkIP, Pinin, THRAP3, and BCLAF1 all coeluted in the same high molecular weight gel filtration fractions as SNIP1, suggesting that they are components of the same complex (Fig. 3B). To confirm these interactions, HA-tagged SNIP1, THRAP3, or SkIP were expressed in 293 cells and α-HA immunoprecipitations were performed before Western blotting for endogenous SNIP1-interacting proteins (Fig. 3C). Importantly, interactions between all the identified SNIP1-associated proteins were detected, indicating these proteins are not simply isolated SNIP1-interacting factors, but rather form a larger complex of which SNIP1 is a constituent. This was further shown by coimmunoprecipitations of endogenous proteins (Fig. 3D). RNase treatment also had no effect on the ability to coimmunoprecipitate these proteins, indicating that complex formation is not a result of independent RNA interactions (Supplementary Fig. S1A). Taken together, these results reveal a previously undescribed, high molecular weight complex, consisting of SNIP1, SkIP, Pinin, THRAP3, and BCLAF. Due to the likely function of these proteins, we have named this complex the SNIP1/SkIP-associated RNA-processing (SNARP) complex. Many components of the SNARP-containing complex have known roles in RNA binding and splicing (14, 16–20), and SkIP has been previously identified in 35S and activated 45S spliceosomes (18). However, we conclude that SNARP is distinct from spliceosomal complexes as we are unable to show interactions between SNIP1 and other spliceosomal factors (data not shown). Furthermore, whereas SNARP components all uniformly elute in a high molecular weight fraction (Fig. 3B), spliceosome proteins were generally found distributed across a much wider molecular weight range (data not shown).

Although we have previously used monoclonal antibodies raised against SNIP1 to investigate its subnuclear localization (7), we observed that these only poorly immunoprecipitated endogenous SNIP1, suggesting that we were not detecting the location of the majority of the endogenous protein (data not shown). Therefore, to determine SNIP1 localization, U-2 OS cell lines were constructed stably expressing SNIP1-YFP COOH- and NH₂-terminal fusion proteins. Importantly, using these cells, SNIP1 was found to colocalize with other members of the SNARP complex in nuclear speckles (Supplementary Fig. S1), which, although not thought to be sites of transcription or RNA processing, can couple these processes by serving as sites of splicing and transcription factor storage, assembly, and modification (reviewed in ref. 25). The identification of these bodies as nuclear speckles was confirmed by their characteristic increase in size and staining intensity after transcriptional inhibition (Supplementary Fig. S1D), together with the colocalization of both SNIP1 and THRAP3 with various nuclear speckle–associated splicing factors (data not shown). In contrast to other SNARP components, these cells also revealed nucleolar localization of SNIP1, a result confirmed with the endogenous protein by subcellular fractionation (Supplementary Fig. S2C). The function of nucleolar SNIP1 is unknown. However, the nucleolus can sequester nucleoplasmic proteins, and this has been previously reported for factors associated with cell cycle progression and stress responses (26). Nucleolar localization may therefore act as a reservoir of available SNIP1 or alternatively could contribute to other functions, such as its role as regulator of the ATR-dependent DNA damage responses (7).

SNIP1-associated proteins are cell cycle regulated and control Cyclin D1 expression. Previously, we have shown that SNIP1 contributes to progression through G₁ phase of the cell cycle and, in particular, to Cyclin D1 expression (2). This suggested that the SNARP complex might become active during G₁. To test this hypothesis, we prepared extracts from pools of cells enriched for different phases of the cell cycle by elutriation. We found significantly higher levels of SNIP1 and the other SNARP complex proteins, SkIP, BCLAF1, and Pinin, in early G₁ phase, coinciding with an increase of Cyclin D1 protein (Fig. 4A). By contrast, THRAP3 was found at all stages of the cell cycle, implying additional roles. FACS analysis and Western blotting for Cyclin B1 known to be up-regulated at later cell cycle stages confirmed the successful separation of cells. This result indicated that the activity of SNIP1 and its associated proteins is likely to be cell cycle regulated, with a specific role during G₁ phase.

Although SNIP1 can regulate Cyclin D1 expression, the ability of other SNARP complex components to perform this function was unknown. Consistent with the components of this complex functioning in a coordinate manner, the production of mature Cyclin D1 mRNA was strongly dependent on SNIP1, SkIP, and THRAP3/BCLAF1 (Fig. 4B). Importantly, depletion of these factors had little effect on the expression of other SNARP components, although a small generally inhibitory effect was noted with THRAP3/BCLAF1 reduction (Supplementary Fig. S2; data not shown). We then examined the relative levels of prespliced and postspliced Cyclin D1 mRNA after siRNA-mediated depletion of SNARP complex components by quantitative PCR. Significantly, when the effects of knocking down SNIP1, SkIP, and THRAP3/BCLAF1 were normalized by comparing the ratio of postspliced to prespliced Cyclin D1 RNA (mRNA to hnRNA), all SNARP complex components displayed similar inhibition of Cyclin D1 mRNA processing (Fig. 4B). Cyclin D1 specificity was shown by the failure of SNARP knockdown to affect Cyclin D2 mRNA production (Fig. 4B), as well as the transcript levels of many other genes examined, including Skp2, GADD45α, Rb, PUMA, and E2F4 (data not shown). The fact that SNIP1 depletion does not lead to the accumulation of prespliced Cyclin D1 RNA argues against a requirement for SNIP1 in Cyclin D1 mRNA splicing. Coupled with previous data regarding the effect of SNIP1 on Cyclin D1 degradation after transcriptional blockage (Fig. 2C), these data strongly support SNIP1 being required for stabilization of newly spliced mRNA within the nucleus, before export.

The SNARP complex is recruited to the Cyclin D1 gene. Previously, we found that SNIP1 can be recruited to the promoters of c-Myc–regulated genes (6). Significantly, ChIP analysis revealed that SNIP1 and the SNARP complex proteins SkIP, THRAP3, and BCLAF1 are all recruited to sites within the Cyclin D1 gene, downstream of the promoter region (Fig. 5A; data not shown). The recruitment of THRAP3 and BCLAF1 to genes has not previously been described, although SkIP recruitment to the 24-hydroxylase and HIV promoters has been reported (20, 27). It is noteworthy that high levels of THRAP3 were also seen binding to region 7 (Fig. 5A), where other SNARP proteins are only weakly detectable. This could indicate an additional non-SNARP role for THRAP3, a possibility also suggested by its cell cycle–independent expression (Fig. 4A). Significantly, serial immunoprecipitation and ChIP (re-ChIP) analysis showed corecruitment of SNIP1 with SkIP, THRAP3, and BCLAF1 (Fig. 5B), indicating the presence of the SNARP complex rather than individual protein components. Despite the reported interaction of the transcriptional coactivator p300 with both SNIP1 and SkIP (5, 23), the presence of p300 was only weakly detectable with these proteins in re-ChIP analysis (Fig. 5B). SNIP1 depletion by siRNA also reduced THRAP3 recruitment to the Cyclin D1 gene (Fig. 5C), providing further evidence for the recruitment of the intact SNARP complex to the Cyclin D1 gene.

SNIP1 regulates U2AF65 recruitment to the Cyclin D1 gene and RNA. In yeast, many RNA processing factors are associated with transcribed regions of genomic DNA, because of their participation in cotranscriptional mRNA processing, (28, 29). U2AF65, an RNA processing factor best characterized for its role in 3′ splice site selection, is one of the few such proteins currently shown to bind along the length of mammalian genes in ChIP assays (30). Moreover, U2AF65 has also been postulated to play additional roles in RNA processing indicated by its continued association with many cellular mRNAs, of which a high percentage represent known cell cycle regulators (31). SkIP, BCLAF, THRAP3, and Pinin have also been reported in similar postsplicing RNA/protein complexes (32). We therefore investigated the ability of SNIP1 to regulate the recruitment of U2AF65 to the Cyclin D1 gene. Significantly, we were able to detect a similar pattern of U2AF65 association with the 3′ region of Cyclin D1 gene to SNARP proteins (Fig. 5A). Moreover, siRNA-mediated depletion of SNIP1 strongly reduced recruitment of U2AF65 to the Cyclin D1 gene (Fig. 5D), indicating that the SNARP complex provides a link between transcription and RNA processing factors. This effect is likely to be indirect as we have not been able to show a direct SNIP1/U2AF65 interaction (data not shown).

We next determined whether SNIP1 and other SNARP complex components were also associated with Cyclin D1 RNA and whether the connection with U2AF65 binding was still observed. Importantly, we found that endogenous SNIP1 and THRAP3 both interacted with the 3′ untranslated region of endogenous Cyclin D1 RNA (Fig. 5D). Furthermore, in parallel with recruitment to the Cyclin D1 gene, depletion of SNIP1 also inhibited the recruitment of THRAP3 and U2AF65 to Cyclin D1 mRNA (Fig. 5D) to a far great extent than that simply attributable to decreased Cyclin D1 mRNA levels with SNIP1 siRNA treatment. Quantitation of these results (data not shown) revealed that even after adjusting for the reduction in Cyclin D1 mRNA levels seen with SNIP1 siRNA treatment, depletion of SNIP1 resulted in a 70% decrease in THRAP3 binding, whereas U2AF65 association decreased by >90%. Taken together, these data suggest a role for SNIP1 and the SNARP complex as a link between Cyclin D1 transcription and RNA processing, aiding the recruitment of other established RNA-processing factors, such as U2AF65, to stabilize Cyclin D1 RNA within the nucleus.

In this report, we identify an important new regulatory step controlling Cyclin D1. We show that the newly identified SNARP complex is required for the production of mature Cyclin D1 mRNA within cells but is not necessary for production of the prespliced message. Moreover, SNARP complex components associate with both the Cyclin D1 gene and mRNA and are required for the further recruitment of the RNA processing factor U2AF65. We found that SNARP complex proteins and U2AF65 bind to the 3′ end of the Cyclin D1 gene and mRNA (Fig. 5D). At present, the mechanism required for SNIP1 recruitment is unclear. However, analysis of genome data (University of California Santa Cruz Genome Bioinformatics Site)¹

reveals significant blocks of evolutionarily conserved sequence within the 3′ untranslated region of the Cyclin D1 gene, where SNARP and U2AF65 recruitment is strongest. We speculate that these may play a role in recruitment of the SNARP complex, although how this targeting might be accomplished is still unclear.

Inhibiting transcription using actinomycin D, followed by analysis of Cyclin D1 RNA degradation, indicated a role for the SNARP complex immediately after transcription (Fig. 2C). Furthermore, the observation that prespliced mRNA does not accumulate with SNIP1 depletion indicates that the splicing of the RNA itself is unlikely to be the stage at which SNIP1 is required. However, it has recently been shown that RNA polymerase II synthesized RNAs containing functional splicing sites are protected from nuclear degradation, presumably because the increased association of the splicing machinery will exclude nucleases (33). Our data are therefore consistent with the SNARP complex stabilizing Cyclin D1 mRNA by regulating the recruitment of RNA processing factors, such as U2AF65, to nascent mRNA transcripts, thus protecting them from premature degradation. In this study, U2AF65 was chosen to examine SNIP1-dependent recruitment as it colocalizes with SNIP1 in nuclear speckles (data not shown), recruited to target genes in mammalian cells (30), and remains associated with the mRNAs of a number of genes after splicing, including many cell cycle regulators (31). However, it is likely that the recruitment of other factors to Cyclin D1 RNA will also be SNIP1 dependent.

The RNA polymerase II COOH-terminal domain heptad repeat (CTD) is thought to play a significant role in recruiting RNA processing factors, although additional mechanisms can facilitate and regulate protein recruitment (reviewed refs. 34, 35). Whether the SNARP complex provides a further link between the CTD and RNA-processing or represents an RNA-polymerase II–independent mechanism is not currently known, although, interestingly, we have observed that SNIP1 coimmunoprecipitates with endogenous RNA polymerase II (data not shown). Although our results have revealed a new role for SNIP1 as being primarily a regulator of Cyclin D1 RNA stability, these do not override previous data that suggested a transcriptional role at this and other genes. Indeed, taken together, both the previous and new observations imply a dual role for SNIP1 as a regulator of gene expression, where its relative contribution to these processes may vary depending on the gene or cellular context.

Our results have important implications for the role for SNIP1 in tumorigenesis. SNIP1 is required for Cyclin D1 expression and augments c-Myc function in multiple cell lines (2, 6). SNIP1 and c-Myc are overexpressed in tumors, and SNIP1 cooperates with both c-Myc and H-Ras to cooperatively induce foci formation in an in vitro transformation assay (6). Interestingly, Cyclin D1 is also frequently overexpressed in cancer and, similar to SNIP1, also cooperates with c-Myc and H-Ras in cell transformation (36, 37). Our results imply that these SNIP1 effects are mediated, at least in part, by the SNARP complex. It will be of interest to investigate whether SNARP complex proteins are coordinately overexpressed in tumors and whether this correlates with Cyclin D1 and c-Myc target gene expression. Alternatively, if SNIP1 levels are limiting in normal cells, its overexpression alone might be sufficient to drive tumorigenesis.

No potential conflicts of interest were disclosed.

Grant support: Cancer Research UK program grant C1443/A5435 (C.P. Bracken) and a project grant from the Association of International Cancer Research (B. Barré).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Moto Ono for preparation of nucleolar extracts; the CLS cloning service for cDNA isolation and plasmid construction; Douglas Lamont and Kenneth Beattie for their assistance with mass spectrometry; all members of the NDP Laboratory for their help and assistance; and Angus Lamond, Julian Blow, and Sonia Rocha for their critical reading of the manuscript and helpful suggestions.