Yeast mRNA splicing in vitro.

Synthetic actin and CYH2 pre-mRNAs containing a single intron are accurately spliced in a soluble whole cell extract of yeast. Splicing in vitro requires ATP. The excised intron is released as a lariat in which an RNA branch connects the 5' end of the molecule to the last A in the "intron conserved sequence" UACUAAC. Two other discrete RNA species produced during splicing in vitro may represent reaction intermediates: free, linear exon 1 and a form of the intron lariat extending beyond the 3' splice site to include exon 2. Both lariat forms correspond to molecules previously shown to be produced during yeast pre-mRNA splicing in vivo.

gous to the intron-exon 2 species described for the mammalian system. Nevertheless, the introns in yeast messenger RNA precursors are distinguished in certain respects from their mammalian counterparts. Firstly, all Saccharomyces cereuisiae introns sequenced to date contain a sequence TACTAAC near their 3' end, and this sequence is essential for splicing (Langford and Gallwitz, 1983;Pikielny et al., 1983). The third A in this conserved sequence is the site of the branch in the intron lariat (Domdey et al., 1984;Rodriguez et al., 1984).
Secondly, the 5' end of yeast introns is characterized by the almost invariant sequence GTATGT. The strict sequence conservation of these splicing signals in yeast is in contrast to the variability of their counterparts in mammalian introns (Mount, 1982;Keller and Noon, 1984). Accordingly, introns of higher eukaryotes are generally refractory to splicing in yeast (Langford et al., 1983;Watts et al., 1983).
There are compelling technical advantages for studying pre-mRNA splicing in yeast. Not least is the opportunity to use genetic approaches to identify and analyze the components of the splicing machinery. The genetic approach, however, demands the development of an i n vitro system, capable of splicing messenger RNA precursors, and the biochemical fractionation of the system. Here we describe the development of an i n vitro system for yeast mRNA splicing. We present a detailed analysis of the products and presumed intermediates in the splicing of synthetic actin pre-mRNA i n vitro.
Plasmids and Plasmid Construction-Modification of the exon 2 sequence of the yeast CHY2 gene (KaLifer et al., 1983) entailed the following manipulations. A 592-bp' XhoI-BglII fragment including parts of the intron and exon 2 was removed from YEp CYH2 (a 2-p vector carrying the yeast URA3 gene and a 5.4-kb fragment of yeast DNA including the CYHB structural gene; kindly provided by Dr. J. Warner) and recloned in M13. As vector, we used a derivative of M13mp8 carrying a fragment of the yeast actin gene which fortuitously included sites for XholI and BglII in the correct orientation. Single-stranded phage DNA from this M13-actin-CYH2 clone was used as the template in a mutagenesis reaction primed with the synthetic 24-mer 5' CTTACCGATTCTACCCTTACCGGC 3'. This introduced three conservative single-base changes into exon 2 of the CYH2 fragment. The mutagenesis procedure is described in more detail in Newman et a1. ' The XhoI-BglII fragment carrying the mutations was then reinserted into YEp CYH2, reconstructing the intact CYHZ gene with a modified exon 2 sequence (designated CYH2").
The SP6-CYH2" transcription plasmid was constructed using Ml3mpll as an intermediate vector (Fig. 4). YEP CYH2" was digested with HinfI, and the HinfI ends were filled in using the Klenow fragment of DNA polymerase I. The digest was then recut with XhoI to yield a 241-bp fragment extending from the middle of exon 1 into the intron (Kaufer et al., 1983); this fragment was gel purified. A 608bp fragment extending from the XhoI site to the unique EcoRI site in exon 2 (Kaufer et al., 1983) was also gel purified, and the two fragments were inserted together into M13mpll cut with SmaI and EcoRI. Subsequently the CYH2" insert was transferred to the SP6 vector pSP64 using the flanking HindIII and EcoRI sites.
The SP6-actin transcription plasmid was made by inserting a 544bp AluI fragment of the yeast actin gene, including the intron and flanking portions of exon 1 and exon 2 (Gallwitz and Sures, 1980;Ng and Abelson, 1980), into the SmaI site of the SP6 vector (Fig. 4).
Preparation of Synthetic CYHP and Actin Pre-mRNA"CYH2" transcripts for in uitro splicing and subsequent oligonucleotide/SI assay were synthesized in 50-pl SP6 transcription reactions containing the following: 20 mM NaCl, 40 mM Tris-C1, pH 7.5, 6 mM MgC12, 2 mM spermidine, 10 mM DTT, 0.5 mM each of CTP, UTP, and ATP, 50 p~ GTP, 0.5 mM m7GpppG (P-L Biochemicals), 100 pCi/ml [a-3ZP]UTP, 1000 units/ml RNAsin (Promega Biotec), 50 pg/ml linearized template, and 400 units/ml SP6 RNA polymerase (Promega Biotec or Boehringer-Mannheim). Transcription was at 37 "C for 2 h, and full-length transcripts were purified by polyacrylamide gel electrophoresis. Using this ratio of "7GpppGGTP, the majority of transcripts initiate with m7GpppG. This was shown by two-dimensional thin layer chromatography analysis (Saneyoshi et al., 1972) of T2 digestion products from a 38-base long model transcript, labeled with [cu-~'P]ATP, synthesized using BamHI-linearized pSP64 DNA as template (data not shown). Actin transcripts for in vitro splicing reactions were synthesized and purified as above, but the UTP concentration was reduced to 25 p~, and [a-32P]UTP was present at 5 mCi/ml. To prepare [a-32P]CTP-labeled actin transcripts, the CTP concentration was 25 p~. Likewise, for [m3'P]ATP labeling, ATP concentration was 50 pM; for [a-32P]GTP labeing, GTP concentration was 100 p~. Unlabeled NTP concentration was kept at 500 ~L M in all cases. In actin transcripts, the cap dimer was not included in the SP6 transcription reaction.
Preparation of Whole Cell Extracts of Yeus-Yeast strain EJlOl was grown, shaking at 30 "C in YPD medium (1 liter) to Am = 2-4. We use this strain because it is deficient in protease: but we have made active extracts from other strains of S. cereuisiae. The cells were harvested by centrifugation at 3000 rpm for 5 min (Sorvall GS3 rotor) and resuspended in 15 ml of 1 M sorbitol, 50 mM Tris-HC1, pH 7.8, 10 mM MgC12, 30 mM DTT. After 15-min incubation at 25 "C without shaking, the cells were collected by centrifugation at 3000 rpm for 5 min and resuspended in 15 ml of 1 M sorbitol, 50 mM Tris-HCl, pH 7.8, 10 mM MgC12, 3 mM DTT (called Buffer S): 90 p1 of 20 mg/ml zymolyase 60,000 (Seikagoku Kogyo Co., Ltd.) were added and the cell suspension was shaken very gently at 30 "C for 40 min. All subsequent steps were carried out at 0-4 "C. The spheroplasts were collected by centrifugation at 3000 rpm for 5 min, washed by gentle resuspension in 15 ml of Buffer S, collected by centrifugation at 3000 rpm for 5 min, and resuspended gently in 8 ml of 10 mM HEPES-K+ (pH 7.0 at 4 "C), 1.5 mM MgC12, 10 mM KC1, and 0.5 mM DTT (Buffer A). The spheroplasts were lysed in this hypotonic buffer by five strokes with a tight-fitting pestle in a glass Dounce homogenizer. Loose-fitting pestles have consistently given inactive extracts and we therefore judge the tightness of the pestle to be an important factor in this step. 2.0 M KC1 was added to give a final KC1 concentration of 0.2 M. The lysate was stirred gently on ice for 30 min and debris was removed by centrifugation at 17,000 rpm for 30 min (Sorvall SS34 rotor). The supernatant was then centrifuged at 37,000 rpm for 60 min (Beckman Ti60 rotor) and dialyzed for 3 h against 1 liter of KC1,20% (v/v) glycerol. Finally, the extract was centrifuged at 17,000 rpm for 20 min (Sorvall SS34) to remove small amounts of insoluble material. The supernatant was frozen at -70 "C in small aliquots. Extracts made by this procedure had protein concentrations of about 25-30 mg/ml and were stable for at least 6 months at -70 "C.
In Vitro Splicing Reactions-Standard splicing reactions were carried out at 25 "C for 10 min in a volume of 10 pl. They contain 4 pl 20 mM HEPES-K+, pH 7.0, 0.2 mM EDTA, 0.5 mM DTT, 50 mM E. Jones, personal communication.
of whole cell extract, 1 pl each of 20 mM ATP, 25 mM MgC12, 30% (w/v) PEG 8000, 0.6 M potassium phosphate, pH 7.0, and 10 fmol of synthetic pre-mRNA. The whole cell extract was in 20 mM HEPES-K+, pH 7.0, 20% (v/v) glycerol, 50 mM KCl, 0.2 mM EDTA, and 0.5 mM DTT. Hence, the final concentrations of the various components present were as follows: 2 mM ATP, 2.5 mM MgClz, 3% (w/v) PEG 8000, 60 mM potassium phosphate, 20 mM KC1, 8 mM HEPES, 8% (v/v) glycerol, 80 p~ EDTA, 0.2 mM DTT, and 1 nM actin or CYH2" pre-mRNA. Yeast proteins were present at about 10-12 mg/ml. The reaction was stopped by the addition of 2 p1 of stop solution (1 mg/ ml proteinase K, 50 mM EDTA, 1% SDS) and the mixture was incubated at 37 "C for 15 min. 200 pl of a mixture containing 50 mM sodium acetate, pH 5.3,1 mM EDTA, 0.1% SDS, 25 pg/ml Escherichia coli RNA were then added and proteins were removed by extraction with an equal volume of phenobch1oroform:isoamyl alcohol (50:50:1). The nucleic acids were precipitated with 2.5 volumes of ethanol, rinsed with 70% ethanol, and dried in uacuo. The RNA was then either analyzed directly by fractionation on thin polyacrylamide-urea sequencing gels, or was subjected to oligonucleotide/Sl analysis.
Oligonucleotide/Sl Assays-The oligonucleotide probe 5' TCTACCCTTACCGGCTGAGACGTGTGGAGA was 5'-end la- SDS were added, followed by 1000 units of nuclease S1. After 30 min at 37 "C, the S1 digestion was stopped by addition of 2.5 volumes of ethanol. The S1-resistant products were fractionated by electrophoresis on 16% acrylamide-8 M urea gels and visualized by autoradiography. RNA Structure Analysis-RNase digestions were performed according to Domdey et al. (1984). Fingerprinting procedures and secondary analyses were carried out according to Volckaert et al. (1976) and Volckaert and Fiers (1977), as applied by Gegenheimer and Apirion (1980). The first dimension of fingerprinting was cellulose acetate membrane electrophoresis, pH 3.5, and the second dimension was homochromatography on polyethyleneimine (PE1)-cellulose thinlayer plates. Secondary analysis was performed by elution of the oligonucleotides with 30% triethylamine bicarbonate, pH 10, appropriate RNase digestions, and two-dimensional thin-layer chromatography on PEI (for T1 and A) or cellulose plates (for T2) (Domdey et al., 1984).

RESULTS
Design of the pre-mRNA Splicing Assay-In order to detect pre-mRNA splicing activity in a crude yeast extract, we needed a sensitive assay which specifically and unequivocally detected the rearranged RNA sequence resulting from correct splicing of an input pre-mRNA substrate. We developed a technique based on the S1 nuclease procedure (Berk and Sharp, 1977), which uses a 5'-end labeled oligonucleotide probe, specific for the exon 1-exon 2 splice junction in CYHZ mRNA (Kaufer et al., 1983), to detect conversion of pre-mRNA to mRNA in uitro (Tabak et al., 1981). The basic principle of the assay is illustrated in Fig. 1.
The oligonucleotide probe is allowed to hybridize with CYH2 sequences in the RNA sample under analysis, and the hybrids are digested with nuclease S1. The S1-resistant products are then fractionated by polyacrylamide gel electrophoresis and visualized by autoradiography. CYH2 mRNA and pre-mRNA protect the end-labeled probe to different extents, yielding two characteristic sets of protected products. The probe is provided with a 6-base extension at the 3' end, which cannot pair with mRNA or pre-mRNA and so is always removed by nuclease S1. This feature allows a distinction to be made between any residual undigested full-length probe as opposed to probe protected by virtue of hybridization to CYHZ RNA sequences. Hybridization to mRNA and pre-mRNA is driven by the excess probe, and the hybrids plus unhybridized probe are then digested with nuclease S1. Pre-mRNA and mRNA each yield a characteristic set of labeled S1-resistant products, which are visualized after electrophoresis and autoradiography. The 3"terminal six nucleotides of the probe do not pair with either mRNA or pre-mRNA, and so are always removed by S1. This feature allows nonspecific and specific protection of the probe to be distinguished.

Ala Gly LYT Gly Arq
Ile Exon 2 Early attempts to detect splicing of CYHB pre-mRNA in uitro using this assay revealed a serious difficulty: many yeast extracts contained a small, but nevertheless readily detectable quantity of CYHB mRNA. This gave rise to a false positive, even in zero time on "extract alone" reactions. Any low level of real splicing of added CYH2 pre-mRNA was then obscured by this background signal.

G i C G t C C A i T C C C A T C T 3 2 P 5 ' PrOQe
Alteration of CYH2 Exon 2 Nucleotide Sequence-To solve this problem we took advantage of the ability of nuclease S1 to cut efficiently at single base mismatches in an otherwise perfect duplex (Shenk et al., 1975). We synthesized an oligonucleotide designed to introduce three single-base changes into the exon 2 sequence (Fig. 2). These changes do not alter the identity of the amino acids encoded by this region of the gene; they merely replace one frequently used codon with another. We used the mutagenic oligonucleotide to introduce these three changes into exon 2 and then reintroduced the modified CYHB gene (CYH2") into yeast by transformation, using a multicopy plasmid vector. We used a strain (5401) carrying the recessive conditional mutation rna2-1, a temperature-sensitive lesion in a gene involved in pre-mRNA splicing Lee et al., 1984;Last et al., 1984). Poly(A)+ RNA was prepared from cultures of strain 5401 harboring the CYH2" plasmid, after growth at 23 or 36 "C. Similar RNA samples were isolated from the same strain harboring 2-p plasmid carrying the unmodified CYH2 gene.
For the S1 analysis shown in Fig. 3, we used an oligonucleotide probe designed to pair with CYHZ" sequences. Three conclusions can be drawn from Fig. 3. First, this probe is indeed specific for CYH2" sequences, since no SI-resistant products survive digestion if only unmodified CYHZ sequences are present. Apparently S1 is able to cut efficiently at one or mo& of the mismatched bases in such hybrids. Second, the assay readily distinguishes between precursor and mRNA allowing assay of conversion of one to the other. Third, not only is the altered CYH2" pre-mRNA spliced normally at 23 "C in uiuo, this splicing also exhibits the normal sensitivity to the rna2-1 lesion, so that at 36 "C the steady state level of spliced CYHZ" mRNA is dramatically reduced. This behavior faithfully mimics that of normal CYH2 transcription products, and that of other intron-containing polymerase I1 transcripts in rna2-1 strains of yeast Teem et al., 1983).
Artificial CYH2" Pre-mRNA-We have used the bacteriophage SP6 RNA polymerase transcription system  to produce synthetic pre-mRNA from the modified CYH2" gene. An internal fragment of the CYH2" gene was inserted downstream of the SP6 promoter in the vector pSP64. This fragment included 80 base pairs of exon 1, the entire 509-base pair intron, and 300 base pairs of exon 2. Run-off transcripts were synthesized using template DNA linearized at a unique EcoRI site at the 3' end of the CYHZ" insert ( Fig. 4). In some experiments, transcripts were capped by including m"GpppG as a priming dinucleotide at 10-fold excess over GTP.
Before using this synthetic precursor, we carried out an experiment in uiuo, t o demonstrate that the CYHB" sequences represented in the SP6 transcript included all of the information required for accurate splicing. The same CYH2" fragment was fused to the promoter from the yeast alcohol dehydrogenase gene ADC1, on a plasmid carrying the 2 -p replication control region (Ammerer, 1983). This construction was introduced into yeast by transformation, and poly(A)' RNA was purified from the transformed strain. The RNA was analyzed by primer extension with reverse transcriptase, using an oligonucleotide primer for CYH2" exon 2. The results of this analysis established that the chimeric ADC1-CYH2" pre-mRNA was efficiently spliced in uiuo (see Newman et al., 1985).
Hind III  tion of nuclear components, and of the reaction conditions. The addition of an ATP regenerating system (phosphocreatine and creatine phosphokinase from rabbit muscle) to the splicing reactions did not alter this result, although it did stabilize the ATP against otherwise rapid hydrolysis in most nuclear extracts (data not shown).

Intron
We surmised these nuclear extracts might lack one or more components involved in pre-mRNA splicing, and so we turned our attention to whole cell extracts made from yeast spheroplasts. Yeast spheroplasts were prepared by zymolyase treatment and lysed in a Dounce homogenizer under hypotonic conditions. Components were solubilized from the crude lysate by extraction with 0.2 M KCl. Cellular debris was removed by centrifugation, and the supernatant was dialyzed and fro-zen in aliquots a t -70 "C (see "Materials and Methods" for details).
In an early experiment, a yeast whole cell extract made in this way was incubated with or without synthetic unlabeled CYH2" pre-mRNA (Fig. 5A). Samples were withdrawn at intervals, deproteinized, and analyzed by the oligonucleotide/ S1 method, using the CYH2"-specific oligonucleotide (cf. Fig.  2). There is no protection of the oligonucleotide against S1 digestion in the absence of the synthetic CYH2" pre-mRNA. At the beginning of the reaction containing synthetic CYH2" pre-mRNA, only pre-mRNA-specific S1 products are visible, whereas, after a few minutes of incubation, a second set of S1 products appears. This set is identical to that produced after S1 analysis of poly(A)+ RNA containing authentic CYH2" with or without the presence of synthetic CYH2"' pre-mRNA. Samples were withdrawn from the reaction at intervals (time of incubation is given in minutes), deproteinized, and analyzed using the oligonucleotide/nuclease S1 assay. The sets of S1 products characteristic of authentic CYH2" pre-mRNA and mRNA are shown in the lane at the center (poly(A)+ RNA from an rnu2-1 strain harboring YEp-CYH2"' at 23 "C). B, CYH2" pre-mRNA was incubated in a yeast mRNA (cf. Fig. 5B). Although the pre-mRNA added to the reaction was degraded very rapidly, the S1 assay was sensitive enough to detect the presence of spliced CYH2" mRNA. The time course of the splicing reaction is better illustrated in Fig.  5B, as the splicing reaction was carried out under optimal conditions. Splicing can be readily detected as early as 6 min after the start of incubation. No significant lag was observed before the appearance of the spliced products. This is quite different from a 25-to 45-min lag observed in in vitro pre-mRNA splicing in mammalian systems (Hernandez and Keller, 1983;Hardy et al., 1984;. As the reaction proceeds the yield of the mRNA-specific S1 products gradually increases, while that of the pre-mRNA-specific S1 products decreases. Furthermore, the appearance of the CYHZ" mRNA-specific S1 products was absolutely dependent on the addition of both synthetic CYHZ" precursor and the whole cell extract of yeast (data not shown). These data suggest that the CYH2" pre-mRNA is being correctly spliced in vitro to give the authentic mRNA sequence spanning the CYH2" exon 1-exon 2 junction.
Optimization of Pre-mRNA Splicing Reaction Conditions-We have systematically varied the reaction conditions, one parameter at a time, in order to establish optimum conditions for splicing and to investigate the requirements for exogenous cofactors. Fig. 6 shows the results of several such experiments. The important features to emerge are summarized below. CYH2" pre-mRNA splicing in the whole cell extract requires ATP (2500 p~) .
UTP, GTP, and CTP can substitute poorly for ATP, perhaps by inefficient phosphate transfer to endogenous ADP. We note that the addition of an ATP regeneration system will support splicing in the absence of added ATP or ADP (data not shown). Nonhydrolyzable ATP analogs (P-7 methylene ATP (AMPPCP) and

CU-P methylene ATP
(AMPCPP), either by itself or in combination) cannot substitute for ATP, implying that splicing has a requirement for ATP hydrolysis.
Splicing was abolished by the addition of 5 mM EDTA, indicating a divalent cation requirement, and displayed a broad optimum for Mg2+ ion concentration centered at about 2-3 mM. A monovalent cation is also required K+, Na+, and NH: all support splicing over a broad concentration range, with an optimum of 80-100 mM. The addition of PEG 8000 to 3% (w/v) enhanced the yield of spliced product, presumably by an excluded volume effect. Optimum temperature for CYH2" pre-mRNA splicing was 25 "C, and optimum pH was 7.0 (data not shown). The placental ribonuclease inhibitor RNAsin (from Promega Biotec) at 1 unit/pl does not inhibit splicing in vitro. On the other hand, another ribonuclease inhibitor, vanadyl ribonucleoside complex (from Bethesda Research Laboratories) completely inhibits splicing in vitro at a concentration of 10 mM (data not shown).
Caps and Cap Analogs-The splicing reactions shown in Fig. 5 used synt.hetic CYH2" pre-mRNAs carrying 7mGpppG caps, as a result of the inclusion of m7GpppG as initiating dinucleotide in the SP6 polymerase transcription reactions. Nevertheless, uncapped transcripts, initiated with pppG, were spliced with apparently equal efficiency (data not shown). We have not ruled out the possibility, however, that such transcripts might be capped in the whole cell extract during the reaction.
We also show in Fig. 6 that the addition of the cap analog dinucleotide m7GpppG to the splicing reaction did not affect the yield of CYHZ" mRNA, even at concentrations as high as 250 PM. This is in contrast to the splicing of the adenovirus major late leader RNA in whole cell extracts from HeLa cells  where even 10 p~ cap analog dramatically inhibited splicing. Analysis of RNA Splicing Produced during Pre-mRNA Splicing in Vitro-The experiments outlined above allowed us to establish optimum conditions for splicing synthetic CYH2" pre-mRNA in uitro. We have subsequently used 32Plabeled synthetic pre-mRNA as substrates in order to isolate and characterize the products of splicing in uitro. Here we present the analysis of the RNA species generated in the whole cell extract from a synthetic actin pre-mRNA. This substrate was synthesized by SP6 transcription from linearized plasmid DNA carrying an internal fragment of the yeast actin gene (Gallwitz and Sures, 1980;Ng and Abelson, 1980) inserted downstream of a bacteriophage SP6 promoter (Fig.  4). The yeast DNA fragment included the 309-bp intron, 73 base pairs of exon 1, and 162 base pairs of exon 2.
32P-labeled run-off transcripts were produced from the SP6actin clone after linearization of the template with EcoRI or HpaII which cut within exon 2 of the actin gene. Synthetic actin pre-mRNAs of two different sizes were then synthesized, differing only in the size of the exon 2 portion of the transcript. These pre-mRNAs were then separately incubated with the whole cell extract under conditions optimized for splicing and samples were withdrawn at intervals. The RNA was deproteinized and fractionated by electrophoresis on an 8% polyacrylamide-8 M urea gel. Fig. 7 shows the result.
Four novel RNA species arise during the course of each reaction. The smallest of these, produced from both substrates, has the mobility expected for free, linear exon 1. HpaII substrate (short exon 2) also generates a product of mobility 135-140 nucleotides, exactly that predicted for the correctly spliced product from this pre-mRNA. Moreover, EcoRI substrate (long exon 2) gives rise to a product of mobility 250-260 nucleotides, again corresponding to that expected for the correctly spliced product. Both of these putative spliced RNA species display some size heterogeneity, with a tendency for the more slowly migrating forms to predominate as the reaction proceeds. This presumably reflects some covalent modification, probably polyadenylation, apparently peculiar to the putative spliced products.
Two other discrete RNA species also appear in each splicing reaction. These molecules share the striking characteristic that they migrate more slowly than the synthetic actin pre-mRNA from which they arise. Furthermore, the apparent mobility of these RNA species varies, relative to DNA markers, according to the concentration of the acrylamide gel on which the samples are fractionated: their electrophoretic behavior becomes more anomalous as the gel matrix concentration is increased (data not shown). This sort of behavior is characteristic of molecules which contain an element of circularity (Bruce and Uhlenbeck, 1978;Sanger et al., 1979) and in the present context suggests that these species might be RNA lariats Grabowski et al., 1984).
For the RNA analysis experiments described below, pSP6 FIG. 7. Splicing of synthetic actin pre-mRNA in vitro. Radioactive actin pre-mRNAs synthesized as run-off transcripts from templates linearized with either HpaII or EcoRI (see Fig. 4) were incubated in the yeast whole cell extract under appropriate conditions. Samples were withdrawn at intervals (time of incubation is numbered in minutes), deproteinized, and analyzed by electrophoresis on an 8% polyacrylamide-8 M urea sequencing gel, followed by autoradiography. End-labeled DNA fragments (pBR322 cut with HpaII (MI) or EcoRII ( M 2 ) ) were run alongside as size mark- . actin plasmid DNA, which was linearized at the HpaII site ( Fig. 4) was used as template to produce run-off transcripts in vitro, labeled to high specific activity with one of the four [cL-~*P]NTPs. Preparative scale splicing reactions were performed, and the RNA species of interest were purified by gel electrophoresis. Unspliced transcripts from the splicing reactions, termed precursor RNA, were also recovered from the gel to serve as control RNA in the subsequent RNA analyses.
Exon 1 and Exon 2 Are Accurately Spliced-The [ c Y -~~P ] UTP-labeled spliced exon 1-2 RNA was digested with RNase A, and the products were fractionated by two-dimensional homochromatography (Volckaert et al., 1976). The [CL-~~P] UTP-labeled precursor RNA was fingerprinted for comparison. The sequence of the various spots was determined by comparing relative mobilities and nucleotide compositions with the known sequence (cf. Fig. 8). A pentanucleotide, GAGGUp, which spans the splice junction, was detected in the fingerprint of the spliced RNA (Fig. 9A). RNase A digestion of the precursor RNA does not produce an oligonucleotide of this sequence (Fig. 9B). Furthermore, this pentanucleotide will be produced only if splicing is accurate (cf. Fig. 8). An equivalent set of two-dimensional fingerprints was done with the spliced exon 1-2 and precursor RNA labeled with [ C L -~~P ] .,L.

c-Exon 1~'
ATP. The same pentanucleotide, GAGGUp, was detected in the fingerprint of the spliced RNA but not the precursor RNA ( Fig. 9, C and D). This result indicates that the phosphate at the splice junction of the mRNA comes from the 3' splice site. This pentanucleotide would not have been labeled if the phosphate was derived from the 5' splice site or from exogenous ATP.
This [32P]ATP-labeled pentanucleotide was eluted from the chromatograph and digested with RNase T1, and the products were fractionated by PEI thin-layer chromatography (Volckaert and Fiers, 1977). A labeled guanosine 3' monophosphate was the only product detected as expected (data not shown). We conclude that the intron was accurately excised and exon 1 and exon 2 were precisely joined together.
The Excised Exon 1 Has a 3' OH Terminus-One of the intermediates which appeared early in the splicing reaction was tentatively identified as exon 1 RNA. It presumably results from a cleavage at the exon 1-IVS junction. The putative exon 1 RNA was digested with RNase T1 and fingerprinted as shown in Fig. 1OA. An in vitro transcript of the   pSP6 actin DNA linearized at the XhoI site, which contains the exon 1 and a small amount of the IVS sequence (cf. Fig.  8), was fingerprinted for comparison (Fig. 10B). Both RNAs were labeled with [cY-~~PIUTP. The control fingerprint (Fig.  10B) contains all but one of the T1 oligonucleotides seen in exon 1 fingerprint (Fig. lOA), as well as several additional T1 oligonucleotides that are derived from the 5' portion of the IVS. Exon 1 ends with the sequence GAUUCUG (Fig. 8). Digestion of the control transcript produces the expected T1 oligonucleotide from this sequence, T17. This oligonucleotide is not present in the fingerprint of exon 1 (Fig. 1OA). Instead, a related oligonucleotide, T17', is found which moves slower in the first dimension and faster in the second dimension. This shift in mobility is expected if the difference between T17 and T17' is that T17' lacks a 3' phosphate. Both T1 oligonucleotides, T17 and T17', were eluted from the chromatographs and digested with RNase A, and the products were fractionated by PEI thin-layer chromatography. RNase A digestion of both T1 oligonucleotides produced 32P-labeled AUp and Cp as expected from the sequence (cf. Fig. 8; data not shown). Furthermore, T17' from the exon 1 RNA migrated more slowly than the corresponding oligonucleotide from the control RNA on a denaturing 20% polyacrylamide gel (cf. Padgett et al., 1984;data not shown). These data are consistent with T17' derived from the exon 1 RNA having a 3' hydroxyl group. The free exon 1, presumably an intermediate in the reaction, must, therefore, have a 3' hydroxyl group as is also the case in the HeLa system Padgett et al., 1984).

A U C U G / U A A U A A C C A C G / A U A U U A U U G / G / A A U~U A G / G / G / G / C U U G /~U U U G /
The IVS* and IVS* Exon 2 RNA Contain a Branch Point-T h e [~~~~P I U T P -l a b e l e d IVS* RNA (the IVS RNA in a lariat form) was digested with RNase T1, and the products were fractionated by two-dimensional homochromatography. The [~~-~~P]UTP-labeled precursor RNA was fingerprinted for comparison. This control fingerprint contains all of the unmodified T1 oligonucleotides of the IVS, as well as additional TI oligonucleotides that are derived from the flanking exon sequences ( Fig. 11B; cf. Fig. 8). A comparison of the T1 fingerprints reveals two oligonucleotides present in excised IVS* which are not present in the precusor RNA. First, the T1 oligonucleotide UpUpUpApGp derived from the 3' end of the intron (T14 in Fig. 11B) is not present in the fingerprint of IVS* and instead a related oligonucleotide, T14', is seen which moves slower in the first dimension and faster in the second dimension. Oligonucleotide T14' also has a slow mobility on a 20% polyacrylamide-8 M urea gel when compared to T14 (data not shown). These data suggest that T14' does not contain a 3' phosphate and that, as in the case of exon 1, the endonuclease cleavage must generate a 3' hydroxyl. Second, a unique oligonucleotide, T27', containing the TAC-TAAC sequence of the IVS, moves faster in the first dimension and slower in the second dimension when compared to the related T27 from the control RNA (Fig. 11). Both T27 and T27' were eluted from the chromatographs and digested with RNase A, and the products were fractionated by PEI thin-layer chromatography (Fig. 12). The RNase A digestion products of the T27 from the control RNA are ACp and AUp, as expected from the sequence (Fig. 12B; cf. Fig. 8). Digestion of the T27' derived from the IVS* RNA produced the same two dinucleotides, as well as a unique oligonucleotide that barely moved from the origin (Fig. 12A). This unique RNase A product of T27' must contain the branch point observed previously in the in vivo study (Domdey et al., 1984) and by other groups in the mammalian system Padgett et al., 1984). Indeed, this unique oligonucleotide contains an RNase T2-resistant component as determined by RNase T2 digestion and cellulose thin-layer chromatography (Saneyoshi et al., 1972) (Fig. 13A). The same RNase T2resistant species, which migrated slowly in the second dimension, was present in the RNase T1 oligonucleotide T27' (Fig.  13B). We have previously suggested that one of the last two, A's in the UACUAAC sequence is the site of a branch point in the IVS (Domdey et al., 1984). In other words, the unique RNase A oligonucleotide in the T27' has a sequence AACp linkled 2'-5' to a G, presumably the G at the 5' end of the IVS. The fact that the RNase T2-resistant oligonucleotide can be labeled with [ c Y -~~P I U T P is consistent with this assumption, because the sequence of the 5' end of the IVS starts with GU. The location of the branch in T27' was unambiguously determined by analysis of the IVS* RNA labeled with each of the four [CY-~~PINTPS. The 32P-labeled IVS* RNA was digested with RNase T1 and RNase A, and the products were fractionated by PEI thin-layer chromatography. The tetranucleotide, A@, was further analyzed by RNase T2 digestion. The results are summarized in Table I. RNase T2 digestion of the [32P]ATP-labeled tetranucleotide, AAE, produced labeled Ap as well as the RNase T2-resistant component (Fig.  130). This result indicates that the G is linked to the second A in the AACp. If the G is linked to the first A in the AACp, RNase T2 digestion would have produced labeled Cp. An equivalent set of experiments was done with the putative splicing intermediate, IVS*exon 2 RNA (the IVS-exon 2 in a lariat form), and similar results were obtained (data not shown). We conclude that both IVS* and IVS*exon 2 RNA are lariat structures with the G at the 5' end of the IVS linked to the last A in the TACTAAC sequence via a 2'4' phosphodiester bond.

DISCUSSION
We have shown that a soluble whole cell extract of yeast will accurately splice synthetic yeast pre-mRNAs containing a single intron. Splicing i n vitro generates the correctly spliced mRNA, and three other discrete RNA species: free exon 1 bearing a 3'OH terminus; a molecule consisting of the intron and exon 2; and the intact, excised intron. The last two of these contain an intramolecular branch at a specific site near Vitro C.  Fig. 9 legend for details). The spots on the fingerprints are given numbers, corresponding to the numbered T1 oligonucleotides in Fig. 8. The T14', derived from the IVS* RNA, migrated slower in the first dimension and faster in the second dimension than the related T14. On the other hand, the T27', derived from the IVS* RNA, migrated faster in the first dimension and slower in the second dimension than the related T27. A predicted A-rich 17-mer from the sequence in Fig. 8, AlrCUG, which streaks in the first dimension, and a large %-mer, which is hardly transferred from the first dimension cellulose acetate membrane onto the second dimension PEI plate, may not be observed on the fingerprints. were eluted from the fingerprint chromatographs, digested with RNase A, and the products were analyzed by two-dimensional thin layer chromatography on PEI plates as described under "Materials and Methods." A, T27' derived from IVS* RNA; B, T27 derived from pre-mRNA, both were [32P]UTP-labeled. Arrows indicate the directions of chromatography. The origin of chromatography is marked with an X. The nucleotide sequence of each 3ZP-labeled spot is indicated. A unique oligonucleotide derived from T27', migrating very slowly in both dimensions, contains the branched nucleotide. Several RNase A digestion products, visualized by ultraviolet light illumination, were circled for alignment. The proposed nucleotide sequence of T27' is shown below the chromatograms. Nearest neighbor nucleotides are given in parenthesis.
the 3' end of the intron, corresponding to the final A of the essential "intron conserved sequence" UACUAAC (Langford and Gallwitz, 1983;Pikielny et al., 1983). This A residue is linked by a 3'-5' phosphodiester bond to the adjacent C, and by a 2'-5' phosphodiester bond to a G, the first nucleotide of the intron.
Hence, the intron and intron-exon 2 molecules consist of a 5' circular component and a 3' linear component downstream of the intramolecular branch. These RNA species correspond precisely to the lariats produced during splicing of globin and adenovirus pre-mRNAs in mammalian cell extracts Padgett et al., 1984).
We have found that ATP is required for yeast pre-mRNA splicing in vitro. What could be the molecular basis for this requirement? ATP analogs such as AMPPCP and AMPCPP will not substitute for ATP itself, implying a need for ATP hydrolysis. When these analogs replace ATP, no specific cleavage or rearrangement of the pre-mRNA occurs (data not shown). One possibility is that the component responsible for making the 2'-5' phosphodiester linkage at the UACUAAC sequence needs ATP-dependent activation. There is a precedent for this in yeast: the tRNA splicing ligase is activated by ATP-dependent adenylylation (Greer et al., 1983). If cleavage at the 5' splice site and formation of the branch are in some way coupled, neither would occur in the absence of ATP.
Alternatively, ATP might be required to organize the precursor and the components of the splicing machinery into a precisely folded structure preparatory to splicing. Perhaps energy is needed to scan for the specific RNA sequences which define the splice sites and intron branch point to be used or to impose a specific three-dimensional organization on such sequences. Biochemical characterization of the splicing machinery may enable us to address these questions.  There are evidently extensive similarities in the way introns are removed from pre-mRNAs in yeast and higher eukaryotes, at least in respect to the intermediates observed and the fate of the excised intron RNA. Whereas in yeast introns, the 5' splice site and intramolecular branch point are rigorously defined by highly conserved sequences, in mammalian introns, and those of higher eukaryotes in general, the corresponding sequences exhibit considerable variability. Indeed, deletion analysis of the rabbit P-globin large intron failed to show a requirement for any specific internal sequence corresponding to the UACUAAC branch point motif from yeast introns (Wieringa et al., 1984). Nevertheless, the sequences surrounding those mammalian intron branch points which have been examined show clear similarities, not only among themselves but also with the yeast TACTAAC sequence Keller and Noon, 1984). Computer analysis of metazoan intron sequences revealed that similar potential branch-point sequences were invariably present near the 3' splice site (Keller and Noon, 1984). Presumably, then, a specific but more flexible branch-point sequence is also re-

TABLE I
Nucleotide composition of the RNase T2-resistant oligonucleotide The SP6-actin pre-mRNAs were synthesized in vitro using each of the four [32P]NTPs in separate reactions and incubated in the i n vitro splicing reactions, and the IVS* RNA was purified and digested with RNase T1 + A. The digestion products were fractionated by two-dimensional TLC on PEI plates. The unique oligonucleotide which migrated very slowly in both dimensions was eluted and digested with RNase T2, and the products were analyzed by twodimensional TLC on cellulose plates as described under "Materials and Methods." The presence or absence of a 32P-labeled product is indicated by + or -, respectively. A schematic diagram of the oligonucleotide containing the RNase T2-resistant component is shown below the table. The RNase T I , A (pancreatic RNase (Panc.)), and T2 cleavage sites are indicated. The 2'4' and 3'4' phosphodiester linkages of the branched nucleotide are also indicated.
1 RNAase I TI and Panc. I T I , P a m , and T2 1 t quired in metazoan introns. When the normal branch point is removed by deletion, the splicing machinery is able to recruit an alternative branch-point sequence elsewhere in the intron (unpublished data cited in Ruskin et al., 1984). Are the mechanistic similarities between yeast and mammalian splicing likely to be related in similarities in the components of the splicing machinery? In mammalian cells, there is molecular evidence that U1 snRNAs are involved in pre-mRNA splicing (Padgett et al., 198313;Kramer et al., 1984). Furthermore, there is a specific interaction between U1 snRNP and the 5' splice site in mouse P-globin pre-mRNA in vitro (Mount et al., 1983). Yeast also does contain small nuclear RNAs resembling the U class of mammalian snRNAs in respect to their size, 5' cap structure, and the presence of modified bases (Wise et al., 1983). These yeast snRNAs are present in low abundance, however, and their involvement in yeast pre-mRNA splicing is open to question. Fractionation and characterization of the components of the splicing apparatus should shed some light on this point.
Are the RNA species produced during yeast pre-mRNA splicing in vitro and the likely sequence of events inferred from their structure a fair reflection of splicing as it occurs in the cell? Lariat forms of the actin and rp51A introns are indeed produced by splicing in vivo (Domdey et aL, 1984;Rodriguez et al., 1984). Moreover, the intramolecular branch occurs at precisely the same position within the UACUAAC sequence in intron lariats generated in vitro and i n vivo. In a separate study: we have examined the effects of specific mutations in the intron sequences which define the 5' splice site and branch point on splicing of the yeast CYH2 pre-mRNA in vivo and in vitro. The results are again consistent with the belief that the RNA metabolism we observe in vitro is an accurate representation of these events as they occur in vivo.