Substitution of pre-mRNA with phosphorothioate linkages reveals a new splicing-related reaction.

In vitro transcription in the presence of a nucleoside 5'-O-(1-thiotriphosphate) has been used to prepare pre-mRNA analogues of the small intron of a rabbit beta-globin gene and flanking exon sequences. Incubation of transcripts prepared with adenosine 5'-O-(1-thiotriphosphate) in a HeLa cell nuclear extract showed that the presence of the thionucleotide in a transcript inhibited splicing, but a novel product was formed by cleavage three nucleotides upstream of the 3' splice site. This product was formed with the same kinetics as the intermediates of a normal splicing reaction, and its formation depended on the presence of intact small nuclear RNAs U1, U2, and U6. We conclude that activation of 3' splice site-proximal sequences need not be linked to exon ligation.

Splicing of labeled mammalian pre-mRNAs in vitro gives rise to novel RNA structures other than the precursor and ligated exons. Time courses of the reaction and detailed characterizations of the novel structures have shown that splicing can be represented as taking place in two steps, with the formation of conspicuous intermediates (1)(2)(3)(4). In the first step, a 2'-5'-phosphodiester linkage (or branch) is formed between an adenosine upstream of the 3' splice site and the first nucleotide of the intron, with displacement of the first exon. In the second step the two exons join and the intron is released as a lariat. The same reaction course has been demonstrated in vitro with extracts from Saccharomyces cerevisiue (5,6), and there is evidence from mammals and yeast that the same process takes place in viuo (7-9).
The first step in the mechanism involves reactions at the 5' splice site, which is set within consensus sequences (10, ll), and at the branch site, which is an adenosine within a very weakly conserved sequence 18-37 nucleotides upstream of the 3' splice site (12-15). The second step involves reactions at the 3' splice site which is adjacent to an absolutely conserved AG dinucleotide sequence preceded by a polypyrimidine tract (10,11).
It is not known whether each of the above steps is a discrete reaction. Mutagenesis of the sites would be expected to reveal partial reactions and to define the involvement of the sites in each step. Some mutations in metazoan sequences involved in step 1 reactions led to the use of cryptic sites (1,13,14) or to the inhibition of, principally, step 2, which indicated that the correct branchpoint structure must be formed before step 2 can take place (16)(17)(18). In yeast, mutations in the essential branch site region (19-21) have affected both steps (5,(22)(23)(24)(25). Mutations in metazoan 3' splice sites inhibited step 2 (12, 16,26) and have impaired step 1 (16), whereas deletion of the polypyrimidine tract prevented both steps (12,26). In yeast the AG sequence is required mainly for step 2 (25,28,29), and the upstream sequences are comparatively unimportant (27)(28)(29). The role of the polypyrimidine tract and the AG sequence in metazoan introns can be correlated with their importance for assembly of the spliceosome in which intermediates are found (30)(31)(32)(33)(34), whereas the branch site may play the equivalent role in yeast (25). None of these mutations resolved either step 1 or 2 into further discrete reactions.
Step 2 has been prevented by a variety of treatments, such as truncation of the 3' exon (26,30,35,36), heat treatment (37), or partial purification of components (37)(38)(39), which still permitted a full step 1 but provided no evidence for further partial reactions.
A more rigorous demonstration that each step was a discrete reaction could be achieved using substrates with chiral analogues of phosphodiester linkages if the configuration at phosphorus was followed (40,41). For this reason, we have investigated the splicing reactions of pre-mRNA substituted during transcription (42) with internucleotidic phosphorothioates. Certain of these substrates produced an unexpected and novel exon 2 product.

RESULTS
I n vitro transcripts were prepared from a portion of rabbit P-globin gene, from nine nucleotides preceding the normal transcription start to +310 (i.e. all of exon 1, IVS-1, and 39 nucleotides of exon 2), which had been cloned into mICE 10 ( Fig. lA; 36,43). Transcription with T7 RNA polymerase was terminated by a Hind111 cleavage of the replicative form, 50 nucleotides to the 3' side of the 3' splice site on the RNAsense strand. In addition single-stranded viral DNA was prepared for transcription by extension of a complementary strand from an oligonucleotide primer with a 5' terminus 50 nucleotides beyond the 3' splice site (36,43). Experiments with templates which lack the insert in mICE 10 have shown that transcripts will extend to the 5' terminus of the staggered end of the cleavage site in HindIII-cut template and to the 5' terminus of the oligonucleotide when primed viral DNA is used as a template (36). Therefore, the two templates will give rise to transcripts ending 54 and 50 nucleotides, respec- The "Experimental Procedures" are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

3' Splice
Site tively, beyond the 3' splice site. These transcripts will be referred to as RNA 54 and 50, respectively.
Transcripts were also prepared from a mutant gene in which the sequence AG/G at the 3' splice site (the slash indicates the splice site) had been changed to AG/A. Transcription of the mutated gene in the presence of ATPaS' would place a phosphorothioate linkage at the 3' splice site, whereas this would be absent after similar transcription of the wild-type sequence.
Transcripts were prepared in reactions containing, in each case, one of the four nucleoside a-thiotriphosphates. These transcripts were used as substrates for splicing reactions in vitro, and the reaction products were analyzed by polyacrylamide gel electrophoresis. Normal splicing products were assigned on the basis of comparisons between substrates with different lengths of 3' exon (the mobilities of exon 1 intermediate and lariat product are unaltered) and on the shifts in mobility, relative to markers, of lariat-containing species on gels with different concentrations of acrylamide. Unsubstituted RNA was spliced efficiently, but splicing of thio-substituted RNA was severely reduced. Strikingly, Fig. 1B shows that splicing of the wild-type RNA with phosphorothioate linkages 5' to adenosine ([sA]RNA 54) gave rise to novel products with lengths of about 54-55 nucleotides, and these were absent in the corresponding reaction with the mutant transcript ([sA]RNA 54A). Formation of these products requires ATP (Fig. 1C). Further experiments showed that two bands (and sometimes a faint third band) were often produced in reactions with transcripts derived from HindIII-cut templates, whereas a single strong band and, occasionally, a fainter upper band were produced when the substrate was transcribed from primed viral DNA (as, for example, in Fig.  3). The multiple bands in the former case were gradually replaced by a single, stronger band with increasing incubation times. The major band was of the same length whether it was derived from [sA]RNA 50 or [sA]RNA 54. In some experiments both wild-type and mutant [sA]RNA can be seen to undergo also one or both steps in normal splicing.
Transcripts prepared with phosphorothioate linkages 5' to guanosine, uridine, and cytidine were also tested. It can be seen from Fig. 2  The novel products have been detected in every nuclear extract preparation that is competent in splicing. In order to  determine whether the 54 nucleotide bands appeared with the same characteristic lag phase as the splicing reactions a comparatively inefficient extract was used. The new bands appeared with the same time dependence as did normal splicing intermediates (Fig. 3); the lag phase has been correlated with the assembly of the spliceosome, and this time course suggests that a similar process is required for generation of the new molecules.
Characterization of the RNA Fragment-For several reasons the new fragments were thought to be derived from the 3' exon. The lengths of the fragments appeared to correspond with that of the 3' exon rather than of other products which might conceivably arise from extensions of known reactions. Furthermore, the [sA]RNA fragment did not appear when a phosphorothioate linkage was placed at the 3' splice site (Fig.   1B). One or two principal bands were seen with [sA]RNA. If the fragments represented the portion of a transcript protected against degradation by the binding of a site-specific factor (13), such a clean result would not be expected. However, a site-specific cleavage a t or near the 3' splice site would be expected to give rise to doublet of major bands, the heterogeneity of which would result from the template-independent addition by T 7 RNA polymerase of (predominantly) one nucleotide to about half of the transcripts which reached the expected 3' end during run-off transcription (36  tion enzyme-cut template appear to run with a mobility of 56 and 57 nucleotides, although the CAG/3' exon sequence would predict fragments of 53 and 54 (based on comigration of the major [sA]RNA 54 bands with the fragments derived from [sA]RNA 50). However, on such gels with high percentages of acrylamide, increases are seen in the apparent molecular weight of RNA molecules (Ref. 2 and data not shown). The 5' termini of the [&]RNA fragments appear to be at the fourth and/or fifth nucleotides preceding the 3' splice site and at the fourth nucleotide following the site.
If the [sA]RNA fragments were produced by a specific processing event, rather than by an unusual activity of a degradative nuclease, a 5' phosphate would be expected on the 5"terminal C. This could be verified by ribonuclease T2 digestion of the [32P]pC universally labeled RNA fragment. Furthermore, identification of the 5' nucleotide as C would support the primer extension data and argue against very short capped 5' portions of the transcript being joined to the 3' exon in a splicing reaction.
RNA substrates were prepared in transcription reactions containing ATP& with [a-3ZP]UTP, [cI-~'P]CTP, or [a-"P] ATP as labeled nucleoside triphosphates. After incubation in a splicing extract, the 54 nucleotide RNA fragments (E2*) were purified by gel electrophoresis. Digestions of E2* with ribonuclease T2 were analyzed by homochromatography. Fig.  5A shows that a product other than [32P]Np was detected in the reaction labeled with [w3'P]CTP. This product disappeared when E2* was treated with calf intestinal phosphatase (Fig. 5B), indicating that the product was derived from a phosphorylated 5' terminus. Attempts to recover the product by subsequent kinase treatment failed because the RNA was degraded and many otherwise internal nucleotides were phosphorylated. The extra product did not appear when E2* was labeled with [cx-~'P]UTP or [w3'P]ATP. This indicates that the 5' terminus of E2* is homogeneous, and that the phosphorylated 5' nucleotide is not pU or PA. It could be p[S]A, which would not be labeled by [cx-~'P]ATP. However, the sequence prior to the 3' splice site is UCUJCAGJ (the first arrow indicates the site of termination by reverse transcriptase (Fig. 4) and the second arrow shows the normal 3' splice site). In this region, [c~-~'P]CTP would not label the ribonuclease T2 digestion product containing A. Thus, that novel product was inferred to be pCp[S]. Furthermore, because it was not labeled by [cx-~~PIATP, it is apparent that the formation of E2* requires p[S]A at the second nucleotide preceding the 3' splice site, and that, where [32P]pA was incorporated instead, E2* did not form.
In order to confirm that pCp[S] was being formed, ["PI pNp[S] markers were produced. RNA was transcribed in the presence of all four nucleoside thiotriphosphates and digested with ribonuclease T1. The products were end-labeled with [Y-~'P]ATP, purified by polyacrylamide gel electrophoresis, and digested with ribonuclease T2. This procedure separated free [T-~'P]ATP from the desired products, i.e. [3ZP]Np [S] where N is C, A, or U. [32P]pNp markers were produced likewise. Fig. 5A shows that the novel product migrated with pCp rather than pCp [S].
For further analysis of the ribonuclease T2 digestion products, two-dimensional chromatography on cellulose thin-layer plates was used. Fig. 5C shows that pCp[S] or pAp[S], rather than PCP, run in the same position relative to Cp as does the spot derived from the 5' terminus of [a-32P]CTP-labeled E2*. This result prompted us to reinvestigate the use of homochromatography. Analysis of ribonuclease T2 digestion products of RNA labeled with [w3'S]ATP and [a-32P]UTP during transcription led us to conclude that homochromatography caused almost 90% of the sulfur content of nucleoside 3'-thiophosphates to be deposited on the line of application to the plate (data not shown). We have not shown directly that the products are nucleoside 3'-monophosphates which migrate as normal, but this is highly likely. We do not understand why the pNp[S] markers in Fig. 5A appear to be unaffected. However, we conclude that the structure of E2* from [sA]RNA is almost certainly pCp[S]AG/3' exon.
Phosphorothioate Linkages Inhibit 3' Exonuclease-Substrates with different lengths of 3' exon sequence were produced in order to confirm that E2* included the 3' terminal sequences of the transcript. These substrates were produced by using various oligonucleotides as primers on viral DNA (43). The original [sA]RNA substrates (3' exon lengths of 50 or 54 nucleotides) gave rise to E2* as before, whereas shorter substrates (3' exon lengths of 42, 46, or 48 nucleotides) gave rise to no E2* products. Longer substrates gave rise to no E2* products, to fragments of the same size as did [sA]RNAs 50 and 54, or to longer fragments, varying with the extract (data not shown).
An explanation for these observations is that [sA]RNA 50 possessed a 3"terminal structure that was resistant to 3' exonuclease activity. Of the substrates tested above, only this substrate possessed a predicted 3'-terminal coded adenosine, and consequently a 3'-terminal phosphorothioate linkage. This explanation is consistent with the single E2* band produced from [ (Fig. U). If these linkages can be removed slowly, [sA]RNA 50 E2* will be fainter than [sA]RNA 54 E2*, which will appear as a doublet.
In order to test the proposal above, oligonucleotides with mismatched 5' termini were used as primers on viral DNA before transcription. One oligonucleotide directed a product with a 3' exon of 42 nucleotides ending in three As, another directed a transcript with a 3' exon of 56 nucleotides ending in three As, and another substituted C for A at the 3' end of the transcript with a 3' exon of 50 nucleotides. Fig. 6 shows the result of splicing after transcribing these templates with ATP&. Even without the substitution of AAA at the 3' end, the 56 nucleotide 3' exon transcript gave rise to appropriate products, although stronger bands are seen to comigrate with those from the 50-nucleotide exon transcript, and the substitution of AAA inverts the intensities of these bands such that the longer products are now dominant. We conclude that the presence of 3"terminal phosphorothioate linkages does substantially stabilize this product, and that the range of suitable substrates is less narrow than at first appeared. However, it should be noted that a substitution of AAA at the 3' end of the 42-nucleotide 3' exon transcript resulted in a loss of the very faint E2* band. The low yield of E2* with this exon length might be correlated with the relatively high level of normal splicing seen in Fig. 6. Substitution of C for A at the end of the 50-nucleotide 3' exon does result in a weakening of the signal, but there is also a one-nucleotide increase in length. In part this can be attributed to the removal by an exonuclease of full length products ending in PC, whereas template-independent addition of one nucleotide to the 3' terminus of a transcript (36) will give rise to a low proportion of molecules ending in p[S]A, which will be protected.
Thus, although there is no direct verification of the sequences present between those hybridizing with the +36 primer and position 42 of the exon, we conclude that beyond reasonable doubt the fragment produced from [sA]RNA extends from three nucleotides prior to the 3' splice site to the

Splicing-related Reaction Near 3' Splice Site
C. FIG. 6. The dependence of E2* formation on protection of the 3' terminus of the 3' exon by phosphorothioate linkages against 3' exonuclease activity. Transcripts synthesized in the presence of ATP& and [L~-~'P]CTP were incubated in splicing reactions for 1 and 2 h, as shown, and subjected to electrophoresis as in Fig. 1. Markers and descriptions of reaction products are as in Fig. 1; IVS-1 (tailless) describes the lariat intron product from which the 3' linear tail has been removed by 3' exonuclease activity. The distance of the 3' nucleotide of the transcript from the 3' splice site was specified by an oligonucleotide (see "Experimental Procedures") and is described by the number of the RNA (ix. 42, 50, or 56 lanes g-j), the three nucleotides at the 3' end of the transcript were changed to adenosine (see "Experimental Procedures"); in -A[sA]RNA 50, the 3'-most adenosine in the transcript was changed to cytidine. All transcriptions were performed in the presence of ATP& in place of ATP, except for the control reactions (lanes rn and n).
A. end of the transcript. A corresponding 5' fragment has not been observed in any of these experiments, and it was not detectable by S1 nuclease mapping (data not shown). Reaction Requires Ul, U2, and U6 RNAs-Selective degradation of snRNAs U1, U2, U4, and U6 in the nuclear extract by oligonucleotide-directed RNase H cleavage has been used to show that splicing requires these snRNAs (33, 37, 55, 57-59). This technique was used to establish whether the same snRNAs are required for production of the novel 3' exonderived fragment. In the first experiment, cleavage of U1 snRNA was incomplete (Fig. 7A), and splicing of a normal transcript was reduced but not eliminated (Fig. 8A). The same treated extract gave a &fold reduction in the level of E2* compared with untreated extract and an even greater reduction when compared with the activity of an extract treated without an oligonucleotide (Fig. 8A). In a second experiment, UlsnRNA cleavage was still incomplete but the production of E2* was eliminated (Figs. 7B and 8B) (Figs. 7 , A and  B ) , and these extracts were inactive in both splicing and in 3' splice site-proximal cleavage of the thiosubstituted transcript (Figs. 8, A and B ) . Residual anti-U4 or U6 oligonucleotides cleaved some of the substrate RNA, but this is only a small proportion of the substrate and this does not jeopardize our interpretation. Oligonucleotides directed against U4 RNA failed to cleave the snRNA, despite the inclusion of ATP in the reaction.

DISCUSSION
The work described in this paper shows that phosphorothioate-substituted RNAs are poor substrates for splicing in vitro. However RNA products were characterized in detail and shown to begin three nucleotides prior to the 3' splice site. The reaction was found to require ATP and sn-RNAs U1, U2, and U6, which supports a role for the splicing apparatus in this process.   The relationship of this novel reaction to the normal processes of pre-mRNA splicing is not clear. The disappearance of the upstream portion of the substrate after E2* formation might suggest that E2* is formed by a 5'-exoribonuclease which is arrested by complexes bound at the 3' splice site and stops prior to a phosphorothioate linkage. The existence of a 5'-exoribonuclease activity was inferred from the production during splicing of 3"terminal fragments from uncapped RNA (34,60), and this activity has been characterized recently (61). However, there are several reasons why this is unlikely to be the explanation. The activity is known to be blocked by transcription of the RNA in vitro under conditions in which a cap analogue is incorporated (34), as in this work. The 5' termini of fragments produced by the endogeneous exonuclease during the splicing reaction are located 10-20 nucleotides upstream of the 5' splice site and 7-16 nucleotides upstream of the branch site (1,34,60), and no fragments have been seen derived from termini near the 3' splice site. Furthermore, studies of nuclease accessibility have shown that the principal region of the substrate protected in splicing reactions encompasses the branch site (34), and sequences extending up to the 3' splice site are involved in stable interactions which allowed them to be immunoprecipitated by anti-snRNP antibodies (32,33,58,62). The polypyrimidine tract upstream of the 3' splice site has been shown to be necessary for binding of a protein which may be associated with the U5 snRNP (49,63) and for binding of a factor which is required for U2 snRNP-branchpoint interaction (64). Thus, even if a 5'exoribonuclease had not been arrested on our substrates before the branch site, it is highly unlikely that it would be arrested by a factor binding to the 3' splice site after removal of the vital polypyrimidine tract which is required for such binding. Neither [sU]RNA, [sG]RNA, or [sAJRNA 54A produce E2*. It is most unlikely that this is due to a failure of spliceosome assembly, for the unsubstituted RNA can splice well (in the case of RNA 54A, better than wild type) and slow splicing can be seen with [sU]RNA. Finally, it should be observed that the presence of a phosphorothioate linkage to the 3' side of the 5' terminus does not imply that the nuclease hypothesis is sufficient. For The most probable explanation of our data is that the substrate has arisen by site-specific hydrolysis, possibly followed by local thiophosphate-arrested exonuclease activity. The hydrolysis reaction might be a result of activation of the 3' splice site such as would be expected to precede step 2 in normal circumstances, or even be derived from a normal component of splicing step 2 (which in this case would not be a simple transesterification). In either case it is apparent that the normal step 2 reaction at the 3' splice site has been blocked by the presence of phosphorothioate linkages in the substrate and that the reaction seen was misplaced. Altered reactions with phosphorothioate-substituted molecules have been reported with simpler systems. It has been shown that the presence of one configuration of a phosphorothioate linkage near, but not at, the site of cleavage of an oligonucleotide by EcoRI will inhibit the reaction (65). Furthermore, the presence of phosphorothioate linkages at the cleavage site of Ap4A by Ap4ase of Artemia caused a shift in the site of cleavage (66). The splice site-proximal reaction combines both elements, i.e. a phosphorothioate linkage may inhibit cleavage at a nearby site and cause the site of cleavage to shift. The time course of E2* production follows that of step 1, not step 2. For this reason we favor a model in which assembly of the spliceosome and the onset of step 1 normally results in activation (strain) at the 3' splice site, but that the presence of phosphorothioate linkages causes alterations in the site of strain such that it is exposed to solvent and prone to hydrolysis.
A number of important points are still unresolved. We do not understand the loss of sequences to the 5' side of the putative point of cleavage which is restricted to substrates which can form E2*. The inactivity of [sA]RNA 54A and [SUI RNA toward putative cleavage or nuclease activities is unexplained. Finally, it is very important to establish whether hydrolysis or 3' splice site activation takes place in normal splicing reactions and thus whether the reaction described here will provide a means of studying an aspect of step 2 events in isolation.