Learn more: PMC Disclaimer | PMC Copyright Notice
Caffeine Regulates Alternative Splicing in a Subset of Cancer-Associated Genes: a Role for SC35▿
Abstract
Alternative splicing of pre-mRNA contributes significantly to human proteomic complexity, playing a key role in development, gene expression and, when aberrant, human disease onset. Many of the factors involved in alternative splicing have been identified, but little is known about their regulation. Here we report that caffeine regulates alternative splicing of a subset of cancer-associated genes, including the tumor suppressor KLF6. This regulation is at the level of splice site selection, occurs rapidly and reversibly, and is concentration dependent. We have recapitulated caffeine-induced alternative splicing of KLF6 using a cell-based minigene assay and identified a “caffeine response element” within the KLF6 intronic sequence. Significantly, a chimeric minigene splicing assay demonstrated that this caffeine response element is functional in a heterologous context; similar elements exist within close proximity to caffeine-regulated exons of other genes in the subset. Furthermore, the SR splicing factor, SC35, was shown to be required for induction of the alternatively spliced KLF6 transcript. Importantly, SC35 is markedly induced by caffeine, and overexpression of SC35 is sufficient to mimic the effect of caffeine on KLF6 alternative splicing. Taken together, our data implicate SC35 as a key player in caffeine-mediated splicing regulation. This novel effect of caffeine provides a valuable tool for dissecting the regulation of alternative splicing of a large gene subset and may have implications with respect to splice variants associated with disease states.
Alternative splicing of pre-mRNA is an essential process by which eukaryotes generate functionally diverse isoforms from a single gene through the selective joining of different exons. This process responds to developmental and environmental cues, contributing substantially to cell-specific and tissue-specific gene expression; it is estimated that over 60% of human genes are alternatively spliced (36). Given the pervasive and profound involvement of alternative splicing in multiple cellular processes, it is clear that the choice of splice site can have major phenotypic consequences, and defects in the splicing pathway are associated with a variety of disease states. Indeed, as many as 50% of human genetic diseases arise from mutations affecting splicing choices (32). For example, it has recently been demonstrated that many cancer-associated genes are regulated by alternative splicing and that a number of splice variants appear to be unique to the malignant state (53, 57). Whether these splice variants play a role in carcinogenesis or tumor maintenance remains to be determined, but their consistent association with the malignant state suggests, at the very least, a clear value as diagnostic indicators and a potential as therapeutic targets (8, 22). Therefore, a better understanding of the etiology of these cancer splice variants has become an imperative.
Decades of elegant studies have defined the basic splicing mechanism, which requires the interaction of spliceosomal core components with splicing signals resident within the pre-mRNA sequences. However, the regulation of spliceosome recruitment and splice site recognition is highly complex and only partly understood. Current models suggest that splice site recognition is under the control of both specific RNA regulatory motifs and at least two classes of RNA binding proteins. The first class includes the heterogenous nuclear ribonucleoproteins (hnRNPs) (33), which generally function as splicing repressors through their recognition of exonic and intronic silencer elements. Interactions of hnRNPs with silencer elements repress proximal exon recognition by inhibiting the use of adjacent splice sites, either by antagonizing the function of positive regulatory elements or by recruiting factors that directly interfere with the splicing machinery (9, 58). Indeed, the human genome is rich in silencing elements (10), suggesting a prevalent role of negative regulation in exon definition (54, 55).
The second class of RNA binding proteins is the serine-arginine-rich (SR) protein family (31). SR proteins are essential for multiple steps of spliceosome assembly at both the 3′ and 5′ splice sites and can function in both constitutive and alternative splicing (14), although how they accomplish this remains largely unclear. What has been shown is that SR proteins bind to either exonic or intronic splicing enhancers, promoting “exon recognition” by directly recruiting the splicing machinery and/or by antagonizing the action of nearby silencer elements (58). Many alternatively spliced exons, small exons, or exons with weak splice sites rely upon the activity of enhancer-bound SR proteins for their inclusion (56). Therefore, SR proteins such as ASF/SF2 and SC35 are generally considered splicing activators (14).
Both SR and hnRNP proteins recognize a large repertoire of target sequences. They have also been shown to modulate alternative splice site selection in a concentration-dependent manner (2, 15); competition between SR proteins and hnRNPs for interaction with regulatory sequences and/or constitutive spliceosomal components, achieved through either qualitative or quantitative changes in these splicing factors, can result in a subtle shift in the balance between positive and negative regulation (3). Adding to this subtle rheostat for splicing control are a number of auxiliary proteins that help to regulate the regulators. Finally, under some circumstances, SR proteins can act as repressors and hnRNPs can activate splicing, adding an additional layer of complexity to this already elaborate process (6, 18).
Despite these complexities, considerable progress has been made in the identification of trans-acting regulators and cis-acting elements involved in the choice of splice sites for a few specific genes, and recent genome-wide computational annotations on expressed sequence tags (35, 43, 54) as well as SELEX analyses (16, 25, 49) have begun to provide a broader view of splice site regulation. Nevertheless, the detailed molecular mechanism of alternative splicing regulation remains largely unclear, hampered by a loose definition of splice site sequences in higher eukaryotes, a high degeneracy of RNA regulatory elements, a limited understanding of RNA secondary structure, the extreme complexity of spliceosomal components and functions and, particularly, a paucity of means to manipulate and control alternative splicing in an experimental system.
In this report, we identify a subset of alternative splicing events in cancer-related genes that are coordinately regulated by the highly consumed psychoactive drug caffeine. The particular splice variants of these genes appear to be tumor specific, in that the alternate transcripts have not been identified in normal cells. Caffeine-mediated induction is rapid, time and concentration dependent, and reversible. Focusing on the tumor suppressor gene KLF6 as a prototype for this caffeine-regulated subset, we show that the SR protein SC35 is required for alternative splicing of this gene. Importantly, caffeine induces SC35 levels, implicating this protein in caffeine-mediated alternative splice site selection. That a small highly diffusible molecule widely present in the human diet can have such profound effects on alternative splicing is an entirely novel and highly significant observation. We propose that caffeine is mimicking an endogenous pathway that is activated in certain cancer cells to induce alternative splicing of a specific group of genes; thus, these studies identify both a unique tool and a unique subset of genes to investigate the mechanisms underlying alternative splicing associated with the cancer phenotype.
MATERIALS AND METHODS
Cell culture and reagents.
The human colon cell lines SW620 and HCT116, the fibrosarcoma cell line HT1080, and the erythroleukemia cell line K562 were maintained in RPMI 1640 medium. The human placental trophoblast cell line 3A was maintained in α-minimal essential medium. Human cervical carcinoma HeLa cells, human fetal lung fibroblast WI38 cells, the rat glioma cell line RG-2, and the mouse melanoma cell line B16-F0 were maintained in Dulbecco's modified Eagle's medium. In all cases, medium was supplemented with 10% (vol/vol) fetal bovine serum, 2.0 mM glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin. One day prior to treatment, 3A cells were transferred from 37°C to 40°C to induce a trophoblast-like phenotype.
Semiquantitative RT-PCR assays.
Cells were seeded in six-well plates at a density of 2.0 × 105 to 2.5 × 105 cells/well. After 18 to 24 h, cells were either untreated or treated with caffeine (8 to 14 mM) for 1 to 24 h as indicated in the figure legends. For inhibition of protein synthesis, 40 μg/ml of cycloheximide was added to cells in the presence or absence of 14 mM caffeine. Either 75 ng or 300 ng of total RNA, prepared using TRIzol reagent (Invitrogen), was analyzed using SS one-step reverse transcription-PCR reverse transcription-PCR (RT-PCR) reagents (Invitrogen) according to the manufacturer's recommendations. The number of amplification cycles used for each reaction was predetermined to ensure that transcript amplification was within a linear range. Primers used for KLF6 RT-PCR are shown below. The primers for β2-microglobulin have been described previously (42). Primers for microarray data validation are available upon request.
Real-time RT-PCR.
Total RNA was extracted from either caffeine-treated or untreated cells and reverse transcribed using MultiScribe reverse transcriptase (Applied Biosystems, CA). TaqMan real-time PCR was performed using primers and probes either specific to total KLF6 transcripts (primer F, 5′-CCTCCACGCCTCCATCTT-3′; R, 5′-GCACGCAACCCCACAGT-3′; probe, 5′-FAM [6-carboxyfluorescein]-TCCGGAACTGAGCAGGGAACCTTCT-BHQ1 [Black Hole Quencher 1])-3′), or specific to SpKLF6 (F, 5′-GCTAGAAAGTTTGAAATGGTGGG-3′; R, 5′-CGTTCCAGCTCTAGGCAGG-3′; probe, 5′-FAM-TGCTGATCTCGGCAGTTGTAGCTTCTTG-BHQ1-3′) using the ABI Prism 7700 sequence detection system (Applied Biosystems, Foster City, CA). Each sample was thermocycled in duplicate at three different concentrations of cDNA (10, 1, and 0.1 ng) to verify the linearity of the assay. Each experiment was repeated twice. Data were analyzed using ABI Prism SDS software.
RNA interference assay.
A total of 3 × 104 HeLa cells in 0.5 ml of Dulbecco's modified Eagle's medium without antibiotics were seeded into each well of a 24-well plate. After 18 h, 25 to 100 nM small interfering RNA (siRNA; siGENOME SMARTpool reagent; Dharmacon) was mixed with Oligofectamine reagent (Invitrogen), incubated at room temperature for 20 min, and added drop-wise to cells. At 48, 72, and 96 h post-siRNA transfection, cells were harvested for either whole-cell lysates or RNA isolation. The concentration and time required for optimal downregulation were determined empirically by both RT-PCR and Western blot analysis.
Western blot analysis.
Cells were washed twice with cold phosphate-buffered saline and lysed in radioimmunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1.0% Triton X-100, 0.1% sodium dodecyl sulfate, 1% deoxycholate, Na salt) plus protease inhibitor cocktail (Roche Diagnostics), 1 mM sodium fluoride, 1 mM sodium orthovanadate, and 100 mg/ml phenylmethylsulfonyl fluoride. The protein concentration of cell lysates was determined using the Pierce bicinchoninic acid protein assay. Equal amounts of total protein (15 to 25 μg) were analyzed by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by immunoblotting using goat polyclonal antibody against hUpf1 (1:1,000; Abcam), mouse monoclonal antibody against SC35 (1:20; provided by Cyril Bourgois and James Stevenin), mouse anti-c-Myc monoclonal antibody (1:1,000; Abcam), or goat polyclonal anti-β-actin antibody (1:1,000; Santa Cruz Biotechnology). The secondary antibody was either horseradish peroxidase-conjugated donkey anti-goat immunoglobulin G (IgG; Santa Cruz Biotechnology) or horseradish peroxidase-conjugated goat-anti-mouse IgG (1:2,500; Amersham). Immunoreactive bands were visualized using an enhanced chemiluminescent system (Amersham) according to the manufacturer's recommendations.
Microarray analysis.
The Cancer-Specific Splice Variant array was purchased from ArrayIT (Sunnyvale, CA; this specific product has been discontinued but similar products can be obtained from ExonHit Therapeutics [Gaithersburg, MD]). This array contains 524 pairs of oligonucleotides homologous to the intron-exon boundary sequence of the 524 pairs of isoforms; one member of each pair has been identified in normal cells, while the second member of the pair was found in tumor cells. The sequences of these oligos can be found on the website of the UCLA Alternative Splicing Annotation Project (http://bioinfo.mbi.ucla.edu/ASAP/). HeLa cells were seeded in 100-mm dishes, incubated overnight, and then treated with 14 mM caffeine. Total RNA was harvested 24 h later in TRIzol reagent (Invitrogen) following the manufacturer's protocol until the chloroform extraction step. The postextraction aqueous phase was mixed with 1 volume of 70% ethanol and immediately subjected to the RNeasy mini column (Qiagen), purified, and labeled as follows: 15 μg of total RNA from either untreated or caffeine-treated HeLa cells was labeled with either Cy5 or Cy3 and subjected to array hybridization. The probe labeling, array hybridization, washing, scanning, and data retrieval steps were performed by the Gene Expression Core Facility (RWJMS—Cancer Institute of New Jersey). Positive controls were included on the array to establish criteria for data mining, and signal differences of <2-fold were considered background.
Minigene and minigene deletion/mutation constructs.
The wild-type KLF6 minigene was created by inserting a PCR-generated fragment containing exon 1 through exon 2 of the KLF6 gene between the KpnI and BamHI sites of pcDNA 3.1(+) (Invitrogen). A PCR was performed using genomic DNA purified from HeLa cells; the primers for this PCR are available upon request. To prepare deletion constructs, the KLF6 minigene plasmid DNA was digested with ClaI and PmlI to remove a large portion of introns 1 and 1a and exon 1a, leaving 146 nucleotides (nt) downstream of exon 1 and 24 nt upstream of exon 2 within the pcDNA 3.1 backbone (pKLF6-Δintron 1/1a/exon 1a) (see Fig. Fig.5,5, below). Various sizes of fragments containing exon 1a and different lengths of flanking intronic sequences (indicated in Fig. Fig.5B,5B, below) were generated by PCR and cloned into pKLF6-Δintron 1/1a/exon 1a at the ClaI/PmlI sites. Point mutations were introduced into the deletion construct (De-4) using the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's recommendations. All the constructs were verified by sequencing.
Identification of a caffeine response element in the KLF6 gene. (A) A KLF6 minigene was generated and shown to recapitulate caffeine-mediated endogenous SpKLF6 induction. The minigene was transfected into HeLa cells, and minigene-directed alternative splicing was analyzed by RT-PCR. Vector-KLF6 junction region-specific primers detected two bands, one corresponding to a transcript lacking exon 1a (KLF6) and one corresponding to a transcript that included exon 1a (SpKLF6). KLF6-specific primers detected both endogenously and exogenously expressed KLF6 (233 bp)/SpKLF6 (∼300 bp). (B) Minigene deletion analyses delimited the essential region regulating exon 1a inclusion to 65 nt upstream to 63 nt downstream of exon 1a (−65 to +63; De-4). Further deletion identified a possible silencing role for −65 to −37 (De-5) and an enhancing role for +38 to +63 (De-6). Note that some minigene SpKLF6 transcript was detectable at relatively higher levels in the absence of caffeine. This is likely due to minigene overexpression. (C, panel 1) The −65 to + 63 sequence of KLF6 exon 1a, in the heterologous context of the human β-globin gene [minigene DUP4-1 (KLF6)], is sufficient to respond to caffeine and induce exon 1a inclusion. Dup4-1 alone is the negative control. (Panel 2). Insertion of the caffeine response element [minigene DUP(en)] is sufficient to induce inclusion of a heterologous 33-nt exon upon caffeine treatment.
Chimeric minigene constructs.
Minigene DUP4-1, generously provided by Douglas Black (UCLA), contains human β-globin exons 1 and 2, as well as a 33-nt chimeric exon comprised of a segment of the first β-globin exon fused to the second exon (34). Minigene DUP4-1 (KLF6) was generated by replacing the middle exon and its flanking intronic sequence with KLF6 sequence from 65 nt upstream to 63 nt downstream of exon 1a at the ApaI/BglII sites of DUP4-1. The DUP(en) construct was generated by inserting the identified caffeine response element 65 nt downstream of the 33-nt hybrid exon in DUP4-1 (see Fig. Fig.5).5). Constructs were verified by sequencing.
Minigene transfection and splicing assays.
A total of 2.0 × 105 to 2.5 × 105 Hela cells/well were seeded into six-well plates. After 18 h, 0.1 to 0.2 μg of minigene construct supplemented with salmon sperm DNA for a total amount of 2 μg was transfected into cells using Lipofectamine 2000 reagent (Invitrogen). Following a 24-h incubation, cells were shifted to caffeine-containing medium and incubated for an additional 24 h prior to RNA isolation using TRIzol reagent (Invitrogen). KLF6 splice variants were assayed by RT-PCR using either primers containing both vector and KLF6 sequences (specific to KLF6 minigenes) or β-globin primers located in exons 1 and 2 and specific to the minigene-generated transcript.
SC35 mammalian expression vector.
SC35 cDNA was amplified by RT-PCR from total RNA prepared from HeLa cells using the forward primer 5′-CGGGAATTCATGAGCTACGGCCGCCCCCC-3′ and the reverse primer 5′-CGCCGGATCCAAAGAGGACACCGCTCCTTCCTC-3′. This PCR fragment was purified and digested with BamHI and EcoRI and then cloned into pcDNA3.1/myc-His(−)-B at the BamHI/EcoRI sites. The c-Myc and His tags are in frame with SC35 at the C terminus of SC35 cDNA. The construct (pcDNA3.1-SC35-c-Myc/His) was verified by sequencing.
To generate the SC35 tet-on construct, the SC35-c-Myc-His DNA fragment was amplified by PCR from the pcDNA3.1-SC35-c-Myc/His plasmid while adaptive restrictive enzyme sites EcoRI/ClaI were introduced at the ends. This PCR fragment was purified, inserted into pTRE-tight vector (Clontech), and verified by sequencing. The HeLa SC35 tet-on inducible clone 6-4 and vector (pTRE-tight)-only clone 4-1 were established following instructions from the manufacturer (Clontech).
SC35 overexpression and minigene splicing assay.
HeLa cells were seeded at a density of 3.0 × 105 cells/well in six-well plates and maintained for 6 h prior to transient transfection with increasing concentrations (0.5 to 1.5 μg) of pcDNA3.1-SC35-c-Myc/His and 0.2 μg of KLF6 minigene. Transfection was performed using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's recommendations. Either whole-cell lysates or total RNA was prepared as described above at 26 h posttransfection. Whole-cell lysates were subjected to Western blot analysis using anti-c-Myc antibody (1:1,000; Abcam). Total RNA was assayed by RT-PCR using either primers containing both vector and KLF6 sequences (specific to KLF6 minigenes) or β-globin primers located in exons 1 and 2 and specific to the minigene-generated transcript.
Nuclear extract, [32P]RNA probes, and RNA electrophoretic mobility shift assay (EMSA).
The HeLa SC35 tet-on clone 6-4 and vector-only tet-on clone 4-1 were treated with 1 μg/ml doxycycline for 48 h, and nuclear extracts were purified from these cells as described previously (21). The protein concentration was determined using a Bio-Rad protein assay kit. Expression of SC35-c-Myc was examined by Western blot analysis (see Fig. Fig.8E).8E). Wild-type and mutant RNA probes were transcribed from single-stranded oligonucleotide templates containing the 22-nt T7 RNA polymerase promoter annealed to a complementary primer. The T7 MEGAshortscript kit (Ambion) was used to transcribe RNAs with the following changes: the reaction mixture contained 750 μM each of CTP, UTP, and GTP, 75 μM of ATP, and 50 μCi of [α-32P]ATP (GE-Amersham Biosciences) for labeling reactions. Transcripts were purified using G-25 Sephadex columns (Roche). Each 30-μl RNA-protein binding reaction mixture contained 20 μg of nuclear extract incubated with 5 × 105 cpm [32P]RNA probe at 25°C for 30 min in HEPES binding buffer (10 mM HEPES at pH 7.5, 50 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 16 units SUPERase·In RNase inhibitor [Ambion], 120 μg/ml yeast tRNA, 5% glycerol) in the presence or absence of an ∼10-fold excess of unlabeled transcript. For supershift assays, 4 to 8 μg of either control IgG or anti-c-Myc antibody was added to the reaction mixture and incubated on ice for 10 min prior to the binding reaction. Reaction mixtures were electrophoresed on 8% nondenaturing polyacrylamide gels at 200 V for 2 h. Radiolabeled complexes were detected by autoradiography on screen-K using the Molecular Imager FX (Bio-Rad).
SC35 is a mediator of caffeine action in the regulation of alternative splicing. (A) Ectopic overexpression of SC35 resulted in concentration-dependent exon 1a inclusion without caffeine treatment. RT-PCR using primers covering the junction sequences revealed increased levels of minigene-derived SpKLF6 transcript with SC35 overexpression in a given cell (lanes 2 to 5). Western blot analysis demonstrated an increasing level of SC35-c-Myc fusion protein in transient cotransfection of the SC35 expression plasmid and KLF6 minigene. β-Actin was a loading control. (B) Overexpression of SC35 mimicked caffeine action in that it induced exon 1a inclusion in the wild-type minigene and De-4, but not in De-6 and Mu(4). (C) Insertion of the caffeine response element into a human β-globin minigene (DUP4-1) is sufficient to induce inclusion of a heterologous exon in response to SC35 overexpression. (D) RNA EMSA analysis demonstrates increased complex formation at the CafRE in extracts overexpressing SC35 (left panel, solid arrow). Either wild-type or mutant RNA transcripts, radiolabeled or unlabeled, were incubated with nuclear extracts from either SC35-overexpressing cells or control cells. Excess cold wild-type RNA and nonspecific RNA served as competitors. Incubation with anti-c-Myc antibody resulted in a decrease in the specific band and the appearance of a supershifted band (empty arrow), indicating the presence of SC35 in the CafRE complex (right panel). IgG was used as a control. The binding complex was detected by autoradiography. The sequences of wild-type and mutant RNA probes are shown on the top with the mutated sites underlined. Wt, wild-type RNA probe; Mut, mutant RNA probe; NE-SC35, nuclear extract from SC35-overexpressing cells; NE-MOCK, nuclear extract from mock-transfected cells; NS, nonspecific RNA. (E) Western blot analysis demonstrated the overexpression of SC35-c-myc in SC35 tet-on HeLa cells but not in the vector-only control cells.
RESULTS
Caffeine induces a splice variant of the tumor suppressor gene KLF6.
KLF6 is a member of the Kruppel-like transcription factor family. It was first identified as a tumor suppressor in prostate cancers (40) and has been shown to be deleted or mutated in various other cancers (5, 20, 26, 44). In a preliminary analysis of the effect of caffeine on gene expression in SW620 cells, we observed an impact of caffeine on KLF6 gene expression. RT-PCR was carried out on total RNA from SW620 cells using primers located within exons 1 and 2 of the KLF6 gene. As shown in Fig. Fig.1A,1A, we observed a novel PCR product (∼300 bp; hereafter designated SpKLF6) created at the expense of the KLF6 transcript (223 bp). This result was confirmed by independent real-time RT-PCR analysis (Fig. (Fig.1A,1A, right panel). Sequence analysis of the upper band indicated that it was derived from a transcript identical to KLF6 except for the inclusion of an additional 66 nucleotides that mapped to the intron between exons 1 and 2 (Fig. (Fig.1B).1B). Examination of this sequence revealed the presence of consensus splice sites at both the 5′ and 3′ borders (Fig. (Fig.1B);1B); while the 5′ splice site fit the consensus, the 3′ sequence delimited a “weak” splice site, consistent with a cryptic exon (46) (hereafter designated exon 1a).
Caffeine induces a novel splice variant of KLF6, termed SpKLF6. (A) Caffeine-induced expression of an unexpected KLF6 transcript. Semiquantitative RT-PCR and real-time RT-PCR were performed on total RNAs from either untreated or caffeine-treated SW620 cells. Primers from exons 1 and 2 used to detect a 233-nt expected PCR product revealed a new PCR product (∼300 bp). Different PCR cycles were compared to ensure linear range representation. β2-Microglobulin (β2M) was used as a loading control. Real-time RT-PCR specifically quantified KFL6 and SpKLF6 transcripts, and data are presented as the log2 of change after caffeine treatment. (B) The KLF6 locus and the sequence of exon 1a and surrounding regions. The 66-nt sequence of exon 1a and the corresponding amino acids are shown within the rectangle. The underlined bases match those of the splice consensus sequences. Arrows indicate locations of primers used in the RT-PCR analysis in panel A. (C) Sequence alignment of exon 1a (uppercase letters) and the neighboring intronic sequences (lowercase letters) in human, rat, and mouse genes demonstrates a high level of conservation. The nucleotide differences are indicated by stars below the sequences. Premature stop codons are underlined. (D) Induction of exon 1a inclusion in rat (RG-2) and mouse (B16-F0) cell lines. Cells treated with increasing concentrations of caffeine for 24 h were collected for total RNA purification. RT-PCR assays were performed using either rat- or mouse-specific primers.
Interestingly, exon 1a is in frame with exon 1 but contains a premature termination codon (Fig. (Fig.1B).1B). A sequence comparison between the human KLF6 and rodent homologues revealed a striking homology within exon 1a (Fig. (Fig.1C),1C), including a totally conserved intron-exon boundary, greater than 90% homology within the coding region, and the premature stop codon (Fig. (Fig.1C).1C). Given these similarities, it was not surprising to find that caffeine also induced SpKLF6 in both rat RG-2 cells and mouse B16-F0 cells, although the induction was less robust than what had been observed in human SW620 cells (Fig. (Fig.1D).1D). Further analysis of total KLF6 mRNA by RT-PCR using a series of primers that span the entire coding region revealed no additional alternative splicing events induced by caffeine (data not shown).
The induction of SpKLF6 by caffeine was concentration dependent, with modest induction observed at 8 mM and significant expression induced at 14 mM (Fig. (Fig.2A).2A). A time course analysis indicated that caffeine-mediated SpKLF6 induction occurred rapidly (as early as 1 hour following caffeine exposure) and was reversible, since SpKLF6 levels were reduced following the removal of caffeine from the medium (Fig. (Fig.2B).2B). To determine the generality of this observation, experiments were repeated using several other human tumor cell lines as well as two nontransformed “normal” cell lines, with similar results (Fig. (Fig.2C2C).
Induction of SpKLF6 is rapid, reversible, and concentration dependent. (A) Caffeine-mediated splice variant induction is concentration dependent and 14 mM induces the highest level of SpKLF6 as revealed by RT-PCR analysis. (B) Caffeine rapidly and reversibly induces the expression of SpKLF6 in SW620 cells. SW620 cells were either treated with caffeine for the times indicated prior to RNA preparation or treated with caffeine for 24 h (time zero) and then washed with drug-free medium and incubated for the times indicated prior to RNA preparation and RT-PCR analysis. (C) Caffeine induces SpKLF6 in various cancer cell lines and nontransformed (“normal”) cell lines. SW620, HT1080, HCT116, and HeLa cells, as well as a normal human placental trophoblast cell line, 3A, and WI38 were treated with increasing concentrations of caffeine and analyzed for SpKLF6 induction by RT-PCR as described for Fig. Fig.1A1A.
Inhibition of NMD is not a major contributor to caffeine-mediated SpKLF6 induction, suggesting regulation at the level of splice site selection.
The presence of a conserved premature termination codon within SpKLF6 marks it as a potential substrate for the nonsense-mediated decay (NMD) pathway (37). Indeed, caffeine (3 mM) has been shown to inhibit phosphorylation of the essential NMD protein hUpf1 (50); this phosphorylation is required for hUpf1 function in NMD. Thus, we considered the possibility that induction of SpKLF6 following caffeine exposure was the result of increased stability of a preexistent transcript, due to caffeine-mediated inhibition of the NMD pathway. We reasoned that if NMD regulated SpKLF6 levels and caffeine blocked NMD, siRNA-mediated downregulation of a critical component of the NMD pathway would mimic the caffeine effect. To test this, HeLa cells were treated with siRNA specific for hUpf1 (Fig. (Fig.3A,3A, upper panel); this resulted in greater than 90% depletion of hUpf1 protein by 72 h posttransfection. A splice variant of the polypyrimidine transcript binding protein (PTB), known to be a target of NMD (48), was used as a positive control for NMD inhibition (Fig. (Fig.3A,3A, left panel). RT-PCR analysis of expression of the KLF6 gene in hUpf1-depleted cells revealed no increase in SpKLF6 expression relative to KLF6 (Fig. (Fig.3A,3A, lower panel, right), indicating that loss of function of hUpf1, and by inference inhibition of NMD, is not likely to be the mechanism for the induction of SpKLF6 by caffeine. Since NMD has been shown to be dependent on de novo protein synthesis (4), this result was confirmed by treating cells with the protein synthesis inhibitor cycloheximide in the presence or absence of caffeine, using PTB as a positive control. As shown in Fig. Fig.3B,3B, cycloheximide treatment had no effect on SpKLF6 levels (lane 3) compared to caffeine treatment (lane 2), consistent with the observation that inhibition of NMD does not mimic induction of this transcript by caffeine. This suggests that caffeine induces SpKLF6 by regulating alternative splicing of exon 1a. Interestingly, caffeine could not induce SpKLF6 in the presence of cycloheximide (Fig. (Fig.3B),3B), indicating that caffeine-mediated SpKLF6 induction was dependent on de novo protein synthesis.
SpKLF6 is not induced as a result of caffeine-mediated inhibition of the NMD pathway. (A) Downregulation of the NMD protein hUpf1 does not mimic the induction of SpKLF6 by caffeine. Western blot analysis indicates effective (> 90%) downregulation of hUpf1 levels in cells treated with hUpf1 siRNA compared to no treatment (none) or treatment with nonspecific siRNA (NS siRNA) (upper left panel). RT-PCR demonstrated that the levels of SpKLF6 were not elevated in cells treated with hUpf1 siRNA (lower panel). (B) Inhibition of protein synthesis by cycloheximide had no effect on SpKLF6 levels (compare lanes 1 and 3). SW620 cells were treated with or without caffeine and cycloheximide as indicated by + and −. RT-PCR was performed using total RNA isolated 6 h after treatment. Addition of caffeine in the presence of cycloheximide was not sufficient to induce SpKLF6 (compare lanes 2 and 4). Induction of the exon 11-skipping PTB transcript was used as a control for effective NMD inhibition.
Caffeine impacts alternative splicing of a subset of genes.
To determine whether the effect of caffeine was unique to KLF6 splicing, we next investigated the impact of caffeine on the regulation of other genes. We were particularly interested in the subset of splice variants that have been linked to the cancer phenotype. A recent analysis of more than 3 million transcripts in the EST database has demonstrated that 524 splice sites show statistically significant specificity to cancer tissues (57). These 524 cancer-specific splice sites derive from highly expressed splice isoforms, and many are being validated for use as reproducible molecular signatures of cancer (1, 17). The TransExpress Cancer Human Splice Variant microarray (offered by ArrayIT, Sunnyvale, CA, as a test platform) includes 524 pairs of oligonucleotides optimized for splice variant profiling of these 524 genes. Each pair consists of a splice site specific to normal cells and an alternate splice site reported to be specific to tumor cells.
Using RNA isolated from vehicle- and caffeine-treated HeLa cells, we performed an initial probe of this microarray (see Materials and Methods). Interestingly, caffeine induced alternative splicing in ∼40 of the 524 alternative spliced genes (unpublished data), indicating that caffeine does not impact the general splicing mechanism but is specific for a subset of genes. To validate this observation, semiquantitative RT-PCR was performed on total RNA samples harvested from both caffeine-treated and vehicle-treated HeLa cells, using primers designed to distinguish between the variants. The results for several genes are shown in Fig. Fig.4.4. Chaperonin-containing TCP1 subunit 3 (CCT3) expresses two transcripts containing either exon 2 or exon 2a, exclusively. While the exon 2a-containing transcript is predominant in untreated HeLa cells, exposure to caffeine resulted in an increase in levels of the exon 2-containing transcript accompanied by a decrease in the level of the exon 2a-containing transcript (this increase in the splice variant at the expense of the normal transcript was similar to what was observed for KLF6/SpKLF6 [Fig. [Fig.1A]).1A]). Similar changes were observed for the asparagine synthetase (ASNS), COMM-domain containing 5 (COMMD5), and ATP binding cassette subfamily F member 2 (ABCF2) genes. Changes were also observed in the levels of transcripts containing exclusively spliced exons for the yippee-like 5 gene (YPEL5, exons 2 and 3). Finally, similar to what was observed for SpKLF6, exon inclusion events (inclusion of exon 2 in the YPEL5 transcript and exon 2b in the SLC39A1/ZIRTL transcript) were also induced by caffeine.
Microarray analysis and semiquantitative RT-PCR validation demonstrate that caffeine regulates alternative splicing of a subset of cancer-associated genes. The list of caffeine-regulated alternative splicing events and the primers used to validate microarray data are available upon request. Several examples of RT-PCR analysis of caffeine-mediated alternative splicing are shown, such as CCT3, ASNS, COMMD5, ABCF2, SLC39A1/ZIRTL, and YPEL5. These genes express alternatively spliced isoforms, and the shaded exons are the ones promoted by caffeine to be included in transcripts. Primer locations for each RT-PCR assay are shown by arrows. Each experiment was repeated at least twice, with similar results.
Taken together, these results demonstrate that caffeine can impact alternative splicing of a subset of cancer-associated genes. To date, there have been no studies implicating caffeine in the regulation of splicing or the splicing machinery. Therefore, to begin to identify the mechanism by which caffeine exerts this effect, we used KLF6 as a prototype to identify both cis-acting regulatory elements and trans-acting splicing factors responsible for this novel caffeine action.
cis-acting elements located within and adjacent to exon 1a are essential for SpKLF6 induction.
Although KLF6 splice variants have been identified, there have been no studies that investigated the alternative splicing regulation of this gene. To determine the cis elements regulating KLF6 and SpKLF6 splicing, particularly the element(s) responsive to caffeine, a KLF6 minigene and minigene deletion mutants were generated. These minigenes were transfected into HeLa cells either untreated or treated with caffeine. RT-PCR analyses were performed on total RNA harvested 24 h after caffeine treatment to monitor the splicing pattern of the KLF6 minigenes. Two sets of primers were used. The first set, the same as the one used in Fig. Fig.1A,1A, was complementary to the KLF6 sequence and detected both KLF6 and SpKLF6 transcripts but could not distinguish between endogenous and minigene-generated RNA. The second set of primers, complementary to the junction regions of the vector and KLF6 sequences, was designed to detect only transcripts generated from the minigene. As shown in Fig. Fig.5A5A (right panel, lane 3), the minigene accurately recapitulated the endogenous KLF6 gene, i.e., KLF6 was the predominant transcript generated from the minigene in untreated cells, and caffeine induced the expression of minigene-derived SpKLF6 transcripts. To delimit intronic sequences required for induction of SpKLF6, minigene deletion constructs that sequentially eliminated exon 1a flanking sequence were analyzed (De-1, -2, -3, -4, -5, and -6) (Fig. (Fig.5B).5B). Interestingly, even a deletion construct with only nt −65 to +63 adjacent to exon 1a (De-4) was permissive for SpKLF6 induction. Indeed, when exon 1a and the flanking −65/+63 sequence were inserted into a heterologous context (Dup4-1, human β-globin exons 1 and 2 [34]), this sequence was sufficient to support caffeine-mediated induction of exon1a inclusion [Fig. [Fig.5C,5C, DUP4-1 (KLF6)], suggesting that the cis-acting regulatory elements involved in exon 1a inclusion lie within this region. When the exon 1a flanking region was further delimited to −37/+63 (De-5), a higher basal level of SpKLF6 transcript was expressed (Fig. (Fig.5B,5B, compare lane 11 to lane 1), which was further induced by caffeine treatment (Fig. (Fig.5B,5B, compare lane 12 to lane 11). This result suggests the presence of exon1a silencing elements that are disrupted upon deletion of the sequences from −65 to −37. Notably, De-5 still supports alternative splicing in the presence of caffeine. In contrast, deletion of sequences from +38 to +63 (De-6) inhibited SpKLF6 expression, even in the presence of caffeine (Fig. (Fig.5B,5B, lane 14), suggesting the presence of an enhancing element(s) within this region that is likely to be responsive to caffeine.
To test whether the +38 to +63 region contains a caffeine response element (CafRE) that is capable of promoting exon inclusion, we took an approach that was successfully employed in studies of the c-src N1 exon enhancer, using an engineered DUP4-1 minigene (34). The DUP4-1 minigene contains exons 1 and 2 of the human β-globin gene as well as an embedded 33-nt chimeric exon comprised of a segment of the first β-globin exon fused to the second exon (Fig. (Fig.5C).5C). DUP4-1 was initially constructed to test a relationship between exon length and constitutive exon skipping; the study showed that the middle 33-nt exon was constitutively skipped until it reached a length of >51 nt (7). Taking advantage of this system, Black and colleagues inserted a c-src N1 enhancing element in triplicate downstream of the 33-nt exon in DUP4-1 and demonstrated the activity of the N1 enhancer in promoting inclusion of the 33-nt middle exon in a complete heterologous context (34). Using the same strategy, the putative CafRE was inserted into the DUP4-1 minigene to generate a heterologous minigene, DUP(en) (Fig. (Fig.5C).5C). DUP(en) was transfected into HeLa cells, and minigene splicing assays were performed using total RNA from untreated or caffeine-treated cells. As shown in Fig. Fig.5C,5C, panel 2, caffeine induced inclusion of the heterologous 33-nt hybrid exon in a subset of β-globin transcripts derived from the DUP(en) minigene (7.6%). Although this rate of exon inclusion was modest, it was highly reproducible and is similar to what was observed in analyses of the role of the c-src exon N1 enhancer; in that study, DUP4-1 containing three copies of the c-src exon N1 enhancer supported a low level of exon inclusion when analyzed in HEK cells (34). Overall, our data strongly suggest that activation by caffeine occurs via the CafRE.
To further delineate the CafRE, scanning mutations were performed in the region between +38 and +63 in the context of De-4. As shown in Fig. Fig.6A,6A, mutations in Mu(2) led to dominant expression of SpKLF6, while those in Mu(4) resulted in a significant reduction of exon 1a inclusion in the presence of caffeine, suggesting one silencing motif and one enhancing motif in close proximity within this region. Essential nucleotides for caffeine-mediated exon 1a inclusion were localized within the TTGAG motif. Interestingly, sequence elements with high homology to this motif were found in other caffeine-regulated genes identified in the microarray; notably, all of these elements were found within the flanking introns of the exons regulated by caffeine (Fig. (Fig.6B).6B). Extensive experimental analyses are currently under way to test the functional relevance of these putative caffeine response elements.
Delineating the caffeine response element and caffeine response motif. (A) Scanning mutation analyses within the caffeine response element in the context of De-4 revealed one silencing motif and one enhancing motif in this region. Mutations in Mu(2) resulted in exon 1a inclusion even without caffeine treatment (lanes 5 and 6), suggesting a silencer element within this motif. Mutations in Mu(4) resulted in substantially reduced SpKLF6 induction in the presence of caffeine (lane 10 [24.7%] versus lane 2 [76.2%]), indicating an enhancer site, possibly TTGAG, as a caffeine response motif. (B) Intronic elements containing TTGAG were located in the flanking neighbors of the array-identified alternatively spliced exons regulated by caffeine. The gene structure and approximate location of those elements are illustrated on the right.
The SR protein SC35 is modulated by caffeine and regulates exon 1a inclusion.
It has been well established that a major mechanism for regulating alternative splicing involves directed silencing of particular exons (10); indeed, many alternatively spliced or cryptic exons rely upon the activity of enhancers to counteract silencers and permit their inclusion into mRNA (56). Thus far, our data support a model in which the cryptic exon 1a of KLF6 is repressed under normal conditions, and the addition of caffeine activates a positive regulatory mechanism. Having identified both silencing and enhancing cis elements regulating the alternative splicing of exon 1a, we have begun to search for splicing factors that work through the CafRE to promote exon 1a recognition. Using an RNA interference approach, we performed individual knockdowns of several SR proteins. Downregulation of SR proteins such as SRp20 and SRp40 did not affect caffeine-mediated exon 1a inclusion (data not shown); however, when HeLa cells were transfected with siRNAs specific to the splicing factor SC35, an effect on SpKLF6 induction was observed.
SC35-specific siRNA was transfected into HeLa cells, resulting in a greater-than-80% decrease in SC35 levels by 72 h posttransfection (Fig. (Fig.7A,7A, lower panel). Under these conditions of SC35 depletion, the induction of SpKLF6 by caffeine was markedly reduced (Fig. (Fig.7A,7A, upper panel), implicating this SR protein in the activation of exon 1a inclusion. To determine whether caffeine had a direct effect on SC35, Western analysis was performed on caffeine-treated cells. As shown in Fig. Fig.7B,7B, the baseline (untreated) level of SC35 remained constant for 24 h (lanes 1, 3, and 5); in contrast, treatment with caffeine resulted in a marked (>6-fold) increase in SC35 protein (lanes 2, 4, and 6). Notably, this increase occurred at a time and concentration commensurate with the induction of SpKLF6. Western blot analysis using a pan-anti-SR antibody (that recognizes most SR domain-containing proteins except ASF/SF2 and SC35), monoclonal antibody MAb104 (that recognizes most phosphorylated SR proteins), or an anti-ASF/SF2 antibody revealed no changes in these proteins following caffeine treatment (data not shown).
SC35 is involved in caffeine-mediated SpKLF6 induction. (A) Knockdown of SC35 in HeLa cells impaired induction of SpKLF6 by caffeine (compare lanes 2 and 4). Western blotting (WB; lower panel) demonstrated at least 80% knockdown of SC35 knockdown in cells treated with an SC35-specific siRNA. β-Actin was used as a loading control. The KLF6 transcript profile (SpKLF6/total) was analyzed by RT-PCR in SC35 siRNA-treated cells as well as in nonspecific siRNA-treated cells. β2-Microglobulin was used as a loading control. (B) Time course analysis revealed increased levels of SC35 (∼6-fold) upon caffeine treatment (lanes 2, 4, and 6). β-Actin was used as a loading control.
Increased SC35 expression is sufficient to mimic caffeine induction of SpKLF6.
To verify a role for SC35 in caffeine-mediated alternative splicing, exogenous SC35 was coexpressed with either wild-type or mutant KLF6 minigenes, and transfectants were assayed for their KLF6 splicing profile. Ectopic overexpression of SC35 in the absence of caffeine resulted in concentration-dependent exon 1a inclusion in the wild-type KLF6 minigene-derived transcripts (Fig. (Fig.8A),8A), indicating that increased SC35 is sufficient to induce exon 1a inclusion. Importantly, SC35 overexpression caused exon 1a inclusion in both wild-type and De-4 minigenes but did not induce significant exon 1a inclusion in De-6 and Mu(4) (Fig. (Fig.8B),8B), mimicking the effect of caffeine on these mutants (Fig. (Fig.5B5B and and6A).6A). Finally, like caffeine, overexpression of SC35 induced inclusion of the 33-nt middle exon in a subset of β-globin transcripts derived from the DUP(en) construct (Fig. (Fig.8C,8C, compare with 5C, map 2), validating a role for SC35 in caffeine-mediated SpKLF6 induction through the CafRE.
SC35 binds to the CafRE in vitro.
An in vitro EMSA analysis was performed to demonstrate interactions between the CafRE and SC35. As shown in Fig. Fig.8D,8D, when a radio labeled wild-type RNA probe containing the KLF6 CafRE was incubated with nuclear extract from either vector-only or SC35-overexpressing HeLa cells, a unique band was observed compared to what was observed with probe alone (Fig. (Fig.8D,8D, compare lane 3 to lane 2). Notably, this band was markedly increased using extracts from SC35-overexpressing cells compared to mock-transfected cell extract (compare lane 3 to lane 4; see Fig. Fig.8E8E for verification of overexpression). Excess cold wild-type RNA transcript competed effectively for the novel band (lane 5), while a nonspecific cold RNA transcript did not (lane 6), verifying the specificity of the binding. Moreover, when a probe containing the point mutations shown to inhibit induction of SpKLF6 [as in minigene Mu(4)] was incubated with SC35-enriched nuclear extract, no specific band was observed (lane 7). To determine whether SC35 was part of the novel complex, supershift assays were preformed using anti-c-Myc antibody. As shown in lanes 9 and 11, incubation with the specific antibody resulted in a decrease in the unique band and the appearance of a supershifted band, while the control IgG did not affect the specific complex. Taken together, these data indicate that overexpression of SC35 results in increased complex formation at the CafRE. Whether SC35 interacts directly or indirectly with the CafRE is currently under investigation.
DISCUSSION
In the present study, we show that caffeine can rapidly and reversibly regulate alternative splicing. This caffeine action is not via targeting of a general spliceosomal core component, since not every alternative splicing event is affected; indeed, microarray analysis of a small set of cancer splice variants indicates that processing of ∼8% of these genes is regulated by caffeine. To begin to evaluate the mechanism underlying caffeine-mediated splicing of this gene subset, we have focused on the induction of SpKLF6, an alternatively spliced form of the KLF6 tumor suppressor gene. Minigene deletion and mutation analyses identified both positive- and negative-acting elements regulating SpKLF6 induction. Notably, the activating element (CafRE) was able to support caffeine-mediated exon inclusion in a heterologous context, and sequences with high homology to the KLF6 CafRE were found in the regions flanking the caffeine-regulated exons of other genes in the subset. Using an RNA interference approach, we have further shown that the SR splicing factor SC35 is required for caffeine-mediated splicing. Strikingly, caffeine treatment led to an increase in SC35 levels commensurate with the induction of SpKLF6, and overexpression of SC35 mimicked the induction of SpKLF6 by caffeine. Furthermore, an RNA EMSA demonstrated increased complex formation at the CafRE with extracts from cells overexpressing SC35, validating a role for this factor in this splicing event. Taken together, our data inform a model of caffeine-regulated alternative splicing of KLF6 and, by inference, a subset of alternatively spliced genes.
Caffeine, alternative splicing of KLF6, and SC35.
KLF6, a member of the Kruppel-like family of transcription factors, has recently been identified as a key tumor suppressor gene in a variety of cancers, where it is often inactivated by allelic loss and/or somatic mutation (19, 20, 26, 40, 45). Notably, it has been shown that mice with targeted deletion of one KLF6 allele (KLF6+/−) display increased liver mass, suggesting that haploinsufficiency at the KLF6 locus may be sufficient to deregulate cell growth control (41). In the present study, we have observed a caffeine-mediated induction of the SpKLF6 variant at the expense of the wild-type KLF6 transcript, effectively reducing the KLF6 transcript by about half (Fig. (Fig.1A).1A). Moreover, while the function of the SpKLF6 splice variant is still under investigation, it is interesting that its novel exon1a includes an early termination codon, predicting an SpKLF6 protein product lacking the DNA binding domain. In all likelihood, this product would either lack transcription activator function or act as a dominant negative regulator. In either case, the resultant haploinsufficiency or loss of KLF6 function, respectively, could very well impact tumor cell growth.
Given the clear role of KLF6 in cancer etiology, and the probability that SpKLF6 will impact that role, KLF6 was chosen as the prototype to investigate the regulation of alternative splicing of cancer-related variants. We have identified two classes of cis elements that regulate expression of the SpKLF6 variant. One class functions as silencers, since mutation of these elements led to constitutive overexpression of SpKLF6. The other element, which we have termed the CafRE, acts as an enhancer mediating caffeine-regulated induction of SpKLF6 (Fig. (Fig.6A).6A). In light of these results, we propose a model in which the inclusion of exon 1a is normally repressed through protein interactions at the silencer element(s). Upon caffeine treatment, this repression is relieved by a quantitative or qualitative change in protein interactions at the CafRE. This appears to be a relatively common mechanism for the regulation of alternative splicing; for example, there have been several reports in which splicing activator proteins antagonize the silencing activity of repressor proteins, resulting in the inclusion of relatively weak or cryptic exons (56, 58). In the majority of these cases, the choice between activation and silencing of exon inclusion is most often determined by the relative abundance of the activator and silencer proteins, although posttranslational modifications may also play a role (2, 15).
Our observation that the SR protein SC35 is also induced by caffeine (Fig. (Fig.7B)7B) is unique and is consistent with our model. SC35 is a prototypical SR protein initially identified as an essential splicing factor involved in multiple steps of spliceosome assembly (12) and thereafter shown to affect alternative splicing as well (13, 28). Regulation of alternative splicing by SC35 is both complex and context dependent. Like other SR proteins, the binding of SC35 to enhancer elements generally promotes exon splicing, often by counteracting the activity of negatively acting hnRNPs. Our observations that caffeine induces SC35 (Fig. (Fig.7B),7B), SC35 overexpression leads to CafRE-dependent cryptic exon inclusion in the context of the KLF6 gene as well as in a heterologous globin gene construct (Fig. (Fig.8),8), and overexpression of SC35 results in increased complex formation at the CafRE in vitro (Fig. (Fig.8D)8D) allow us to extend our model to state that caffeine regulates the CafRE-dependent induction of SpKLF6, at least in part, by inducing the expression of SC35, thereby altering the activator/repressor ratio. While computational analysis did not identify an SC35 consensus binding site (29, 49) within the CafRE, alternative splicing regulatory elements are known to be highly degenerative. Therefore, we are currently investigating whether SC35 acts directly by binding to a nonconsensus site within the CafRE or indirectly by modulating or interacting with a CafRE binding protein complex. In either case, we propose that the CafRE complex antagonizes the silencing elements. In this regard, preliminary studies have identified two hnRNPs that are involved in exon 1a repression (data not shown). Future studies will be directed at determining whether these repressor proteins interact with the KLF6 silencing elements and whether their activity is impacted by caffeine.
Caffeine regulates alternative splicing in a subset of genes.
Our data from individual gene analysis (KLF6) and splice variant array analysis/validation have clearly demonstrated that caffeine can regulate alternative splicing in a subset of genes. Whether caffeine accomplishes this through activation of SC35 alone or whether caffeine has additional effects on other factors remains to be determined. However, as a first step toward identifying a common mechanism for regulation of this gene subset, we have examined a number of caffeine-regulated genes for the presence of sequences with high homology to the enhancing motif identified within the KLF6 CafRE. Given the high degeneracy of splicing factor binding sites in general, we were encouraged to find matches to the sequence motif within close proximity to the caffeine-regulated exons, although more extensive analyses will be needed to determine the functional relevance of these putative CafREs.
Alternative splicing, SC35 expression, and cancer.
Within the past several years, considerable interest has been paid to the role of aberrant alternative splicing in cancer (52). Isoform imbalance is a common occurrence in tumors, and several splice variants are serving as cancer biomarkers due to their correlation with malignancy (47); cancer-specific splice variants may also provide targets for tumor-targeted toxin-conjugated antibodies (24). It has recently been shown that alternative splicing produces oncogenic variants of KLF6 (SV1, SV2, and SV3); these variants are associated with a germ line single-nucleotide polymorphism downstream of exon 1a and have been found to be overexpressed in prostate cancers (38, 39). While these KLF6 splice variants do not appear to be regulated by caffeine (data not shown), these studies do support the notion that KLF6 splice variants can impact the cancer phenotype and mark the introns surrounding exon1a as critical sites for alternative splicing regulation.
Factors involved in alternative splicing pathways are also under investigation as potential therapeutic targets (51). Indeed, SF2/ASF has been recently identified as a proto-oncogene that activates several key targets contributing to particular transformed phenotypes (23). Aberrant increases in SC35 have also been associated with the cancer phenotype (27). For instance, transformation of mouse fibroblasts with either large T antigen or c-erbB2 caused an increase in SC35 levels when compared to untransformed cells (30), and a marked induction of SC35 has been observed in malignant ovarian tissues relative to normal ovarian tissue in certain malignancies (11). It is intriguing to speculate that SC35 is also a proto-oncogene whose transforming potential lies in its ability to regulate some of the genes identified in this study.
Conclusion.
In this report we show, for the first time, that the splicing of a subset of genes is regulated by caffeine and identify SC35 as both a mediator of caffeine-regulated splicing and a target of caffeine action. We hypothesize that caffeine is mimicking an endogenous pathway that is aberrantly activated in certain cancer cells to induce alternatively expressed splicing variants of a subset of genes. Thus, our studies have identified both a unique tool and a unique cadre of genes to investigate the mechanisms underlying alternative splicing associated with the cancer phenotype. Given the dearth of information regarding the regulation of alternative splicing, particularly for the many genes whose alternative splice variants have been implicated in the cancer phenotype, we believe that caffeine, and the subset of genes that it targets, provides us with a novel and powerful system by which to gain essential insights into this complicated and potentially clinically relevant process.
Acknowledgments
We thank Cyril F. Bourgeois and James Stevenin at Institut de Génétique et de Biologie Moléculaire et Cellulaire, France, for generously providing SC35 monoclonal antibody. We thank Douglas Black (UCLA) for generously providing the DUP4-1 clone. We acknowledge Todd Martinsky (ArrayIT, Inc.), Eric Olson (Genesifter), and Jonathan Bingham and Subha Srinivasan (Jivan Biologics Inc.) for facilitating the microarray data analyses and validation. We are grateful to Rebecca Cimildoro and Julia Pimkina for preliminary studies on caffeine-mediated KLF6 splicing and to other members of the Scotto laboratory as well as our CINJ colleagues for insightful discussions.
This work was supported by NIH-CA06927 (FCCC), NIH-P30-CA072720 (Cancer Institute of New Jersey), NCI-R01 CA 122573 (K.W.S.), and the UMDNJ Foundation grant program (J.S.).
Footnotes
▿Published ahead of print on 19 November 2007.