Dimethyl Sulfoxide Affects the Selection of Splice Sites*

Depending on the cell lines and cell types, dimethyl sulfoxide (Me 2 SO) can induce or block cell differentia- tion and apoptosis. Although Me 2 SO treatment alters many levels of gene expression, the molecular processes that are directly affected by Me 2 SO have not been clearly identified. Here, we report that Me 2 SO affects splice site selection on model pre-mRNAs incubated in a nuclear extract prepared from HeLa cells. A shift toward the proximal pair of splice sites was observed on pre-mRNAs carrying competing 5 * -splice sites or competing 3 * -splice sites. Because the activity of recombinant hnRNP A1 protein was similar when added to extracts containing or lacking Me 2 SO, the activity of endogenous A1 proteins is probably not affected by Me 2 SO. Notably, in a manner reminiscent of SR proteins, Me 2 SO activated splicing in a HeLa S100 extract. More- over, the activity of recombinant SR proteins in splice site selection in vitro was improved by Me 2 SO. Polar solvents like DMF and formamide similarly modulated splice site selection in vitro but formamide did not activate a HeLa S100 extract. We propose that Me 2 SO im- proves ionic interactions between splicing factors that contain RS-domains. The direct impact of Me 2 SO on al- ternative splicing may explain, at least in part, the different and sometimes opposite effects of Me 2 SO on cell differentiation and apoptosis. Me polymerase Although

Me 2 SO 1 is a polar solvent used to promote cell differentiation of tumor cell lines. For example, the treatment of mouse erythroleukemic and neuroblastoma cells with 2% Me 2 SO induces morphological changes and differentiation in red blood cells and neurons, respectively (e.g. see Refs. 1,2). Me 2 SO also induces differentiation of the human U937 monoblast leukemia cell line into monocyte/macrophage (3) and stimulates the differentiation of a human ovarian adenocarcinoma cell line (4). Paradoxically, Me 2 SO prevents the terminal differentiation of myoblasts (5,6), inhibits the differentiation of adipocytes (7), blocks the differentiation of antibody-producing plasma cells (8), and interferes with the differentiation of chick embryo chondrocytes (9). Whereas Me 2 SO has been used to induce apoptosis in some cell lines (10,11), it inhibits cell density-dependent apoptosis of CHO cells (6). Thus, depending on the cell line, Me 2 SO can have completely different effects on differentiation and apoptosis.
The cellular mechanisms that are affected by Me 2 SO remain unclear. Because Me 2 SO facilitates DNA uptake during transfection procedures (e.g. see Ref. 12), Me 2 SO has been proposed to affect the integrity of cell membranes. Because Me 2 SO alters protein kinase C activity and the expression of integrin complexes (6,13), Me 2 SO may alter intracellular signaling processes, which in turn may have a broad impact on many aspects of gene expression. Me 2 SO treatment promotes changes in the abundance of certain mRNAs and in the ratio of spliced isoforms (14 -17). Among the genes reported to be affected in their alternative splicing is the NCAM pre-mRNA. A 2% Me 2 SO treatment of N2a cells promotes an increase in the inclusion of neuro-specific NCAM exon 18 (18,19). Me 2 SO alters the alternative splicing of other genes including the amyloid precursor protein (20), the serotonin 5-HT3 receptor-A mRNA (21), and p53 (22,23). Me 2 SO has also been associated with an effect on c-Myc mRNA elongation, maturation, and stability (23)(24)(25), and on the translation of some mRNAs (1). Whether any of the above changes result from a direct effect of Me 2 SO on RNA synthesis, maturation, and/or stability is currently unknown.
Because treating cells with Me 2 SO can have a strong effect on the alternative splicing of many pre-mRNAs and because the mechanism of action of Me 2 SO remains unclear, we performed a series of experiments in nuclear extracts to assess whether Me 2 SO directly affects the activity of the splicing machinery. We find that Me 2 SO can have drastic effects both on 5Ј-splice site and on 3Ј-splice site selection in vitro. Notably, other solvents of the same category (e.g. DMF and formamide) also perturb splice site selection.

Treatment of Cells with Me 2 SO and in Vivo Alternative Splicing
Assays-Me 2 SO was purchased from various suppliers including EM Science and Fisher Scientific Inc. DMF and formamide were from Calbiochem. N2a and HeLa cells were cultured at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum. For treatment with Me 2 SO, medium containing 2% bovine calf serum was used. Following treatment, total RNA was isolated using the guanidinium-HCl protocol as described in Chabot (26). RNase T1 protection assay was performed according to Melton et al. (27) using a uniformly labeled 530-nt NCAM antisense RNA probe. Exon 17/exon 19 splicing yields a 303-nt protected fragment while the inclusion of exon 18 produces a 452-nt fragment. Products were resolved on a 5% denaturing acrylamide gel. The reverse transcriptase-PCR assay used to amplify products corresponding to exon 7B inclusion and exclusion has been described in Chabot et al. (28).
Substrate Pre-mRNAs and in Vitro Splicing Assays-pC5Ј Ϫ/Ϫ, pC5Ј 4/4 and pC3Ј Ϫ/Ϫ have been described in Blanchette and Chabot (29). pNCAM3Ј was constructed by substituting the 3Ј-splice site of exon 7B and exon 7B sequences in pC3Ј Ϫ/Ϫ for the equivalent regions of alternative exon 18 of the mouse NCAM gene (403 bp of exon 18 and 111 bp of upstream intron sequences). Splicing substrates were produced from plasmids linearized with ScaI, and transcribed with T3 RNA polymerase in the presence of cap analogue and [␣-32 P]UTP (Amersham Pharmacia Biotech). RNA purification was performed as described in Chabot (26). HeLa nuclear extracts and S100 extracts were prepared (30) and used in splicing reactions as described previously (28). Although Me 2 SO, DMF, and formamide were always added last, the order of addition did not affect the outcome.

Me 2 SO Affects Alternative
Pre-mRNA Splicing in Vivo-Me 2 SO can promote cell differentiation, a process that is often associated with a change in the alternative splicing profile of specific genes. One example of this effect is found in the mouse N2a neuroblastoma cell line. The treatment of N2a cells with 2% Me 2 SO induces neuronal cell differentiation and improves the frequency of inclusion of the neurospecific exon 18 in the NCAM pre-mRNA (Refs. 2, 31; Fig. 1A). A similar effect was observed on the hnRNP A1 pre-mRNA. In this case, we moni-tored the inclusion frequency of alternative exon 7B following the treatment of HeLa cells for 5 h with 5% Me 2 SO (Fig. 1B). Although the effect was less dramatic than for the NCAM pre-mRNA, Me 2 SO treatment significantly improved the inclusion of exon 7B.
Me 2 SO Affects Splice Site Selection in Vitro-To determine whether Me 2 SO can modulate splice site selection directly, we tested the effect of adding Me 2 SO to splicing reactions incubated in nuclear extracts prepared from HeLa cells. We used model pre-mRNAs derived from the hnRNP A1 alternative splicing unit (29). C5Ј Ϫ/Ϫ contains two competing 5Ј-splice sites and a unique 3Ј-splice site ( Fig. 2A). C5Ј Ϫ/Ϫ is spliced almost exclusively to the proximal 5Ј-splice site (Fig. 2B, lane 1). In contrast, the presence of A1 binding elements in C5Ј 4/4 promotes efficient splicing to the distal 5Ј-splice site (lane 5). The addition of Me 2 SO at a final concentration of 0.8, 1.6, and 2.4% did not affect the splicing efficiency of C5Ј Ϫ/Ϫ RNA, and 5Ј-splice site selection remained exclusively proximal (Fig. 2B, lanes 2-4). In contrast, Me 2 SO promoted a strong reduction in the use of distal 5Ј-splice site in C5Ј 4/4 pre-mRNA (lanes 6 -8).
The highest concentration of Me 2 SO (lane 8) produced a 5-fold decrease in the use of the distal 5Ј-splice site. In some experiments, the reduction in distal 5Ј-splice site use was accompanied by an increase in the production of lariat products derived from the proximal 5Ј-splice site (e.g. see Fig. 5A, lane 2).
The effect of Me 2 SO on 5Ј-splice site selection was as strong on a pre-mRNA that was synthesized in the absence of cap analogue (data not shown). Thus, the reduction in distal 5Јsplice site usage was independent of the cap structure at the 5Ј-end of the pre-mRNA. Me 2 SO also affected 5Ј-splice site

FIG. 1. Me 2 SO promotes exon inclusion in vivo.
A, mouse N2a cells were treated 48 h with 2% Me 2 SO. Total RNA was analyzed for changes in the alternative splicing of NCAM exon 18. A RNase T1 protection assay was used to monitor the ratio of exon 18 inclusion (E18ϩ) or exclusion (E18Ϫ). B, HeLa cells were treated with 5% Me 2 SO for 5 h. Three dishes of cells were tested for each treatment and total RNA was analyzed for changes in the alternative splicing of exon 7B in the hnRNP A1 pre-mRNA. The percentage of exon 7B inclusion on endogenous A1 transcripts was determined by using a reverse transcriptase-PCR assay as described in Chabot et al. (28). Control PCR reactions were performed with plasmids containing the cDNA from A1 (lacking exon 7B, lane 1), or A1 B (containing exon 7B, lane 2). The values were plotted as percentage of inclusion on a histogram that shows standard deviations.

FIG. 2. Me 2 SO affects 5-splice site selection in vitro.
A, structure of the pre-mRNAs used to assay modulation of 5Ј-splice site selection. C5Ј Ϫ/Ϫ and C5Ј 4/4 have been described previously (29). The C5Ј 4/4 pre-mRNA contains two CE4 elements, which are binding sites for hnRNP A1. B, incubation of the pre-mRNAs in HeLa extracts was for 2 h in the presence of different percentages of Me 2 SO (0, 0.8, 1.6, 2.4%). Labeled RNA products were fractionated on a denaturing 11% polyacrylamide gel. The position and structure of the proximal and distal lariat products are shown. selection in a model pre-mRNA carrying two copies of the 5Ј-splice site of exon 7 (data not shown). Identical effects were seen with Me 2 SO solutions obtained from different suppliers, and the deionization of Me 2 SO did not change its activity on 5Ј-splice site selection (data not shown). Transient exposure of nuclear extracts to Me 2 SO (i.e. incubation in the presence of Me 2 SO followed by dialysis) did not affect 5Ј-splice site usage (data not shown). Thus, Me 2 SO needs to be present in splicing mixtures to affect splice site selection.
Because Me 2 SO had a strong effect on the alternative splicing of a pre-mRNA carrying A1 binding elements (C5Ј 4/4), we asked whether Me 2 SO compromised the activity of the hnRNP A1 protein. We have shown previously that hnRNP A1 promotes distal 5Ј-splice site utilization on this pre-mRNA (29). In nuclear extracts containing Me 2 SO, the addition of hnRNP A1 efficiently shifted selection toward the distal 5Ј-splice site (Fig.  3A, lanes 6 -10). The effect was specific because the addition of similar amounts of GST or gene 32 protein had no effect (data not shown). Notably, the profile of stimulation obtained with increasing amounts of recombinant A1 was similar to the profile obtained in a nuclear extract lacking Me 2 SO (Fig. 3A, lanes 1-5; compare the slopes in Fig. 3B). Because the activity of recombinant hnRNP A1 is not compromised by the presence of Me 2 SO, it is unlikely that Me 2 SO affects the activity of the endogenous A1 proteins.
To address whether Me 2 SO has a similar activity on 3Ј-splice site selection we tested a pre-mRNA (C3Ј Ϫ/Ϫ; Fig. 4A), which is spliced predominantly to the distal 3Ј-splice site (Ref. 29; Fig.  4B, lane 1). C3Ј Ϫ/Ϫ splicing was sensitive to increasing amounts of Me 2 SO (Fig. 4B, lanes 2-4). At the highest concentration of Me 2 SO (lane 4), more than 50% of splicing occurred at the proximal 3Ј-splice site. We also tested a derivative of C3Ј Ϫ/Ϫ in which the central portion was substituted for the 3Јsplice site region and a portion of NCAM alternative exon 18 (NCAM3Ј RNA). Although splicing of NCAM3Ј RNA was less sensitive to Me 2 SO than C3Ј Ϫ/Ϫ, Me 2 SO promoted a stronger reduction in the use of the distal 3Ј-splice site than of the proximal 3Ј-splice site (Fig. 4B, lanes 5-9). Alternative splicing of a ␤-globin pre-mRNA carrying duplicated 3Ј-splice sites was also affected by Me 2 SO (data not shown).
Me 2 SO Activates SR Proteins-The effect of Me 2 SO on splice site selection is reminiscent of the activity of SR proteins, which tend to activate splicing of the proximal pair of splice sites (32,33). Although Me 2 SO did not stimulate overall splicing activity in nuclear extracts (Fig. 5A, lanes 1 and 2), we asked whether Me 2 SO could mimic the generic splicing activity of SR proteins. This activity was initially defined by the capacity of SR proteins to activate splicing in a HeLa S100 extract, either as a mixture of SR proteins or individually (33,34). U2AF 65 also activates splicing when added to a HeLa S100 extract (35). Notably, the addition of Me 2 SO to a HeLa S100 extract stimulated splicing as efficiently as the addition of the recombinant SR protein ASF/SF2 (Fig. 5A, lanes 3-5). The

FIG. 4. Me 2 SO affects 3-splice site selection in vitro.
A, structure of the pre-mRNAs used to assay modulation of 3Ј-splice site selection. C3Ј Ϫ/Ϫ is derived from the hnRNP A1 gene (29). The NCAM 3Ј pre-mRNA is an hybrid pre-mRNA containing the 5Ј-splice site of exon 7, the 3Ј-splice site of NCAM alternative exon 18 and the 3Ј-splice site of adenovirus L2 exon. B, in vitro splicing assays of model pre-mRNAs. Incubation was for 2 h in HeLa extracts containing increasing percentages of Me 2 SO (0, 0.8, 1.6, 2.4% in lanes 1-4, and 0, 0.8, 1.6, 2.4, 3.2% in lanes [5][6][7][8][9]. Labeled RNA products were fractionated on denaturing 11% (for C3Ј Ϫ/Ϫ) or 6.5% (for NCAM3Ј) polyacrylamide gels. The position of the pre-mRNA and proximal and distal lariat products is shown. addition of Me 2 SO to a S100 extract also stimulated the formation of complexes, as judged by native gel analysis (Fig 5B,  lanes 7-9). These results suggest that Me 2 SO increases the activity of residual amounts of SR or U2AF proteins in the S100 extract. The level of splicing stimulation varied considerably in different preparations of S100 extract. Although Me 2 SO and recombinant ASF/SF2 restored splicing activity in a similar manner, splicing to the distal 5Ј-splice site was not detected, as is the case in a nuclear extract (lanes 1 and 2). We have shown previously that distal 5Ј-splice site selection on this pre-mRNA requires hnRNP A1 (29). The failure to activate distal 5Ј-splice site use is probably because of the fact that S100 extracts contain small amounts of hnRNP A/B proteins as compared with nuclear extracts (not shown).
The above result suggests that Me 2 SO may affect the activity of SR proteins. To further examine this possibility, we tested the effect of adding Me 2 SO to splicing reactions preincubated with a recombinant SR protein. At the concentrations used and in the absence of Me 2 SO, the recombinant SR protein GST⅐SRp30c had little effect on 5Ј-splice site selection when using the C5Ј 4/4 pre-mRNA (Fig. 6A, lanes 1-3). However, in the presence of Me 2 SO, the same amount of GST⅐SRp30c stimulated proximal 5Ј-splice site utilization (lanes 4 -12). Thus, the simultaneous addition of Me 2 SO and SR produced a shift toward proximal use that was greater than the sum of their individual contribution. Because recombinant SR proteins display more activity in the presence of Me 2 SO, a similar effect on the endogenous SR proteins may be responsible for the activity of Me 2 SO in nuclear extracts.
DMF and Formamide Also Modulate Splice Site Selection-To understand the chemical basis for the activity of Me 2 SO in alternative splicing, we tested other solvents. At equivalent percentages, both DMF and formamide were at least as active as Me 2 SO at modulating 5Ј-splice site selection (Fig. 7A). Surprisingly, although DMF and formamide shared with Me 2 SO the ability to modulate 5Ј-splice site selection, formamide was unable to activate splicing in a HeLa S100 extract (Fig. 7B). DISCUSSION We have observed that the addition of Me 2 SO to nuclear extracts can have strong effects on splice site selection while having minimal effects on the efficiency of splicing. In contrast, the addition of Me 2 SO to a splicing-deficient HeLa S100 extract stimulated splicing in a manner reminiscent of the activity of SR proteins. The effect of Me 2 SO on splice site selection was also similar to the activity of SR proteins because Me 2 SO shifted selection toward the proximal pair of splice sites. Consistent with the notion Me 2 SO stimulates the activity of SR proteins, we found that the combination of Me 2 SO and SRp30c produces a shift that is greater than the sum of their individual contribution. Thus, a general stimulation in the activity of all endogenous SR proteins most probably explains why Me 2 SO influences splice site choice in vitro. Likewise, the addition of Me 2 SO to a S100 extract may stimulate the residual amounts of SR proteins present in this extract.
An additional mechanism by which Me 2 SO may affect splice site selection is through the inactivation of the hnRNP A/B proteins, which are known to antagonize the activity of SR proteins in splice site selection (36,37). However, we have FIG. 5. Me 2 SO rescues splicing in a HeLa S100 extract. A, splicing reactions were performed in a HeLa nuclear extract (NE) and in a HeLa S100 extract. The extracts were incubated in the absence or in the presence of 3.2% Me 2 SO. The S100 extract was also supplemented with 0.5 g of recombinant ASF/SF2 protein (lane 5). The pre-mRNA substrate used was C5Ј 4/4. B, a HeLa nuclear extract (NE), a HeLa S100 extract or S100 extract supplement with 3.2% Me 2 SO(S100 ϩ DMSO) were incubated with a model pre-mRNA derived from the adenovirus major late transcription unit. Following incubation for the time indicated (in minutes), samples were fractionated in a native 5% polyacrylamide gel. The position of the nonspecific complex H and of splicing complex A is indicated.
FIG. 6. Me 2 SO activates SR proteins. A, using the C5Ј 4/4 pre-mRNA, the activity of the recombinant SR proteins GST⅐SRp30c was tested in the absence and in the presence of increasing concentrations of Me 2 SO. The GST⅐SRp30c protein (0.5 and 1 g) was preincubated in nuclear extract 15 min at 30°C before adding the pre-mRNA and Me 2 SO. B, diagram depicting the SRp30c-mediated reduction in distal 5Ј-splice site usage in extracts lacking or containing different concentrations of Me 2 SO. observed that the activity of recombinant hnRNP A1 proteins in splice site selection is not affected by the presence of Me 2 SO. This result suggests that the activity of endogenous A1 proteins is probably not affected by Me 2 SO.
Because Me 2 SO, DMF, and formamide decrease the melting temperature of DNA and RNA duplexes, it is possible that changes in 5Ј-splice site selection may be caused by the weakening of the base pairing interactions between U1 snRNA and 5Ј-splice sites. However, this explanation is unlikely for the following reasons. First, a reduction in U1 snRNP binding should lead to more distal 5Ј-splice site utilization, as is the case when U1 activity is reduced through oligonucleotide-targeted RNase H digestion (data not shown). Moreover, increased selection of a proximal 5Ј-splice site is associated with improved U1 snRNP binding (38). Second, splice site selection on a transcript that is normally spliced to a distal 5Ј-splice site because of a duplex structure in the pre-mRNA was insensitive to Me 2 SO. 2 Thus, the effect of Me 2 SO, formamide and DMF on 5Ј-splice site selection cannot be explained by a reduction in the stability of base pairing interactions. However, it is possible that this denaturing activity contributes to the reduction in overall splicing activity observed at higher concentrations of solvents.
Although our results suggest that Me 2 SO affects the activity of SR proteins, the mechanism by which SR proteins become activated remains unclear. Western analysis using an antibody that recognizes phosphorylated epitopes on SR proteins revealed no change in the overall and relative abundance of phosphorylated SR proteins upon incubation with Me 2 SO (data not shown). Moreover, Me 2 SO did not affect the binding of SR proteins to a purine-rich RNA splicing enhancer element (data not shown). Me 2 SO also did not modify the solubility of SR proteins when extracts were incubated with increasing concentrations of MgCl 2 (data not shown). Although Me 2 SO is regarded as a relatively inert solvent for pharmacological applications, it improves the solvation of cations and stimulates nucleophilic reactions. DMF and formamide share with Me 2 SO this chemical property. Thus, one possibility is that Me 2 SO improves the solvation of positive charges on proteins. This may influence the structure at the surface of proteins and facilitate ionic contacts between charged domains of interacting proteins. Consistent with this view, modulation of 5Ј-splice site selection in vitro is known to be sensitive to the ionic conditions of the reaction (39). Splicing proteins that carry charged domains include SR and U2AF proteins, which have RS-domains rich in positively and negatively charged amino acids (arginines and phosphorylated serines, respectively). Interactions between the RS-domain containing proteins ASF/ SF2, U1 snRNP-70 kDa, and U2AF 35 have been proposed to occur early during spliceosome assembly (40). Moreover, these interactions are sensitive to the phosphorylation state of ASF/ SF2 (41). Thus, Me 2 SO may activate splicing in a S100 extract by improving the quality of the ionic interactions between residual amounts of SR and U2AF proteins. Because the amount and activity of these proteins are in excess in a nuclear extract, this would explain why Me 2 SO stimulates generic splicing in a S100 but not in a nuclear extract.
Even though nuclear extracts contain sufficient amounts of SR proteins for generic splicing, their activity in splice site selection is not maximal because adding more SR proteins can have a strong effect on the selection of splice sites (32,33). Because a similar effect can be obtained by adding kinases that phosphorylate the RS domains of SR proteins (42), the profile of charged residues at the surface of SR proteins is crucial for their activity as splicing regulators. Moreover, the requirement for charged residues at the surface of SR proteins appears different for generic and alternative splicing because dephosphorylation of ASF/SF2 is essential for constitutive splicing, but is not required for the protein to function as a splicing regulator (41). Thus, Me 2 SO may affect the presentation of charged residues that are important for the activity of SR proteins in splice site selection. Because Me 2 SO and DMF activate splicing in a HeLa S100 extract, Me 2 SO and DMF may also affect the presentation of different residues that are important for generic splicing. In contrast, because formamide affects splice site selection but cannot activate a S100 extract, formamide may only affect the presentation of residues that play a role in splice site selection.
Our results raise the possibility that the documented effect of Me 2 SO on cell differentiation may be caused, at least in part, by changes in the activity of SR proteins, which in turn affect splice site selection. This conclusion is supported by the observation that DMF can mimic the effect of Me 2 SO both in differentiation assays (5,(43)(44)(45), and in splicing assays in vitro. Depending on the cell types, Me 2 SO can either promote or block differentiation or apoptosis. These opposite outcomes may be explained if different subsets of pre-mRNAs are expressed in different cell types. For example, alternative splicing is often used to control the production of proteins involved in programmed cell death such as Fas, Bcl-2, Bax, and Ced-4 (46). Hence, Me 2 SO may alter the alternative splicing of a pre-mRNA to favor the production of a repressor protein in one cell type, while an inducer may be produced from another gene in a different cell type. 2 6 and 12). The position of the pre-mRNA and proximal and distal lariat products is shown. B, splicing reactions with the C5Ј 4/4 pre-mRNA were set up in HeLa S100 extracts in the presence of increasing amounts of Me 2 SO, DMF, or formamide (0, 2.4, 3.2, and 4%). For comparison, a splicing reaction performed in a HeLa nuclear extract is shown (NE, lane 13).