Purification and Characterization of Human Cleavage Factor I m Involved in the 3 (cid:42) End Processing of Messenger RNA Precursors*

Six different protein factors are required for the spe- cific cleavage and polyadenylation of pre-mRNA in mammals. Whereas four of them have been purified and most of their components cloned, cleavage factor I m (CF I m ) and cleavage factor II m (CF II m ) remained poorly characterized. We report here the separation of CF I m from CF II m and the purification of CF I m to near homo- geneity. Three polypeptides of 68, 59, and 25 kDa copu-rify with CF I m activity. All three polypeptides can be UV cross-linked to a cleavage and polyadenylation substrate in the presence of a large excess of unspecific competitor RNA, but not to a splicing-only substrate. No additional protein factor is required for the binding of CF I m to pre-mRNA. Gel retardation experiments con- firmed the results obtained by UV cross-linking. In addition, we could show that CF I m stabilizes the binding of the cleavage and polyadenylation specificity factor (CPSF) to pre-mRNA and that CPSF and CF I m together form a slower migrating complex with pre-mRNA than the single protein factors. Cleavage stimulation factor (CstF) and poly(A) polymerase (PAP) had no detectable effect on the binding of CF I m to pre-mRNA. Further- more, the CstF (cid:122) CPSF (cid:122) RNA as well as the CstF (cid:122) CPSF (cid:122) PAP (cid:122) RNA complex are supershifted and stabilized upon the addition of CF I m . 3 ends

The 3Ј ends of almost all mRNAs in eukaryotes are generated posttranscriptionally in two tightly coupled steps. The primary transcript is first cleaved endonucleolytically at the polyadenylation site in the 3Ј-untranslated region. The upstream cleavage fragment is subsequently polyadenylated, whereas the downstream cleavage fragment is rapidly degraded. The two reaction steps can be uncoupled experimentally and assayed separately. Investigation of the cleavage reaction alone is possible by suppression of the Mg 2ϩ -dependent polyadenylation of the upstream cleavage fragment. No polyadenylation occurs, when the reaction is performed either in the presence of EDTA and ATP or in the presence of MgCl 2 and cordycepin 5Ј-triphosphate (3Ј-dATP) which acts as a chain terminator. The polyadenylation reaction can be investigated with "precleaved" RNA as substrate that ends at or near the natural cleavage and polyadenylation site.
Both cleavage and polyadenylation are dependent on cisacting elements in the pre-mRNA and on trans-acting protein factors (for reviews, see Refs. [1][2][3][4]. Two essential cis-acting elements have been described: one of them, the highly con-served polyadenylation signal AAUAAA, lies 10 -30 nucleotides upstream of the cleavage and polyadenylation site, and the second one, a less conserved GU-or U-rich sequence element, is located 10 -30 nucleotides downstream of the cleavage and polyadenylation site. Six different protein factors have been shown to be required for specific cleavage and polyadenylation of pre-mRNA in vitro (5,6). Three of them, cleavage stimulation factor (CstF), 1 cleavage factor I m (CF I m ) and cleavage factor II m (CF II m ; abbreviations for CF I m and CF II m according to (3)) are involved in the cleavage reaction only, while the cleavage and polyadenylation specificity factor (CPSF) and poly(A) polymerase (PAP) are necessary for both steps. Poly(A)-binding protein II (PAB II) acts as a stimulatory factor for poly(A) tail elongation. CstF, CPSF, PAP, and PAB II have been studied extensively, are well characterized, and most of their components have been cloned.
CstF consists of three subunits. Its 64-kDa subunit was shown to interact with the GU-or U-rich downstream sequence element (7)(8)(9)(10)(11). CPSF consists of three or four subunits (12)(13)(14)(15) and binds specifically to the polyadenylation signal AAUAAA (12,16). CstF and CPSF, together with a pre-mRNA substrate, form a stable complex (17) and are thought to confer specificity to the 3Ј end processing reaction. The approximate region in which cleavage will occur is defined by the relative positions of the AAUAAA and the GU-or U-rich elements, and the precise site of cleavage is then determined by a preference for a local nucleotide sequence within this region (18). CPSF remains bound to the hexanucleotide AAUAAA after cleavage has occurred and enables PAP to elongate specifically the upstream cleavage fragment to a poly(a) tail length of 250 nucleotides in the presence of PAB II (19).
The endonuclease could not be identified so far. Good candidates are the two poorly characterized cleavage factors CF I m and CF II m , which are required only for the first step of the reaction.
We report here the separation of CF I m from CF II m and the purification of CF I m to near homogeneity. Each of the three polypeptides copurifying with CF I m activity can be UV crosslinked to a cleavage and polyadenylation pre-mRNA substrate and gel retardation experiments showed that CF I m binds to pre-mRNA, even in the absence of any of the other protein factors involved in 3Ј end processing of pre-mRNA.

EXPERIMENTAL PROCEDURES
RNAs-RNAs used as substrates for cleavage reactions, UV crosslinking, and gel retardation assays were prepared from the following plasmids: plasmid pSV-L contains the polyadenylation signal of the SV40 late transcription unit (20), and plasmids pSP6L3 and pSP6L3⌬1 contain the L3 polyadenylation signal of the adenovirus 2 major late * This work was supported by grants from the Kantons of Basel and the Schweizerischer Nationalfonds. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 41-61-2672060; Fax: 41-61-2672079; E-mail: keller2@ubaclu.unibas.ch. transcription unit carrying either the wild-type polyadenylation signal AAUAAA or the mutated polyadenylation signal AAGAAA, respectively (21). Plasmid pSP6L3pre was derived from pSP6L3 by truncation of the L3 polyadenylation signal one nucleotide upstream of its natural cleavage site (22). Plasmid pBSAd1 contains the first two leader exons, the shortened form of the first intron, and 27 nucleotides of the second intron of the adenovirus 2 major late transcription unit (23).
16 and 23 S rRNA and tRNA from Escherichia coli MRE 600 were purchased from Boehringer Mannheim GmbH.
Proteins-CstF was purified from HeLa cell nuclear extracts as described (26). CPSF was prepared from calf thymus according to a procedure described earlier (12) and was kindly provided by E. Wahle. Recombinant bovine poly(A) polymerase containing six histidines at the N terminus was a gift of G. Martin and was prepared according to Ref. 26. Crude CF II m was a peak fraction of an 8-ml Mono S HR 10/10 FPLC column (Pharmacia Biotech Europe GmbH) similar to the one described here, which was dialyzed against 20 mM ammonium sulfate buffer (see "Purification of Cleavage Factor I m ", below).
HeLa cell nuclear extracts were prepared from 240 liters of freshly harvested or frozen HeLa suspension cells as described (27) except that HeLa cells were grown to 4 -6 ϫ 10 5 cells/ml, and buffers A and C contained Tris⅐HCl instead of Hepes⅐KOH. In addition, buffer C contained KCl instead of NaCl and 0.5 mM phenylmethylsulfonyl fluoride, 0.4 g/ml leupeptin hemisulfate, and 0.7 g/ml pepstatin. High salt buffer C was buffer C containing additionally 800 mM ammonium sulfate. After the high salt extraction of the nuclei, the homogenate was centrifuged for 1 h at 35,000 rpm (135,000 ϫ g max ) in a Centrikon TFT 65.38 rotor (Kontron Instruments). The nuclear extracts were stored frozen without dialysis.
312 ml of HeLa cell nuclear extract were allowed to thaw in 250 ml of buffer without ammonium sulfate and subsequently diluted to the conductivity of buffer containing 40 mM ammonium sulfate. 1460 ml (2.4 g of protein) of diluted nuclear extracts were spun for 30 min at 8000 rpm (10,000 ϫ g max ) in a Sorvall GSA rotor, and the supernatant was applied to a DEAE-Sepharose fast flow column (Pharmacia) of 5 ϫ 12.5 cm equilibrated in 40 mM ammonium sulfate buffer. The column was washed with 1.4 column volumes of the same buffer and developed with a gradient (5 column volumes) of 40 -300 mM ammonium sulfate at 340 ml/h. Active fractions eluting between 90 and 150 mM ammonium sulfate were pooled (300 ml) and dialyzed for 16 h against 2 ϫ 1 liter of buffer containing 50 mM ammonium sulfate and 20 mM Hepes⅐KOH, pH 7.9, instead of Tris⅐HCl. Only 38% of the DEAE-pool were carried through the following steps of the purification. Therefore, 113 ml (147 mg of protein) were further dialyzed for 3 h against 1 liter of 25 mM ammonium sulfate, 20 mM Hepes⅐KOH buffer, spun for 30 min at 9000 rpm (10,000 ϫ g max ) in a Sorvall SS34 rotor, and loaded onto an 8-ml Mono S HR 10/10 FPLC column (Pharmacia) at 1.1 ml/min. The column was washed with 10 column volumes of the same buffer and eluted with a gradient (40 column volumes) of 25-380 mM ammonium sulfate at 2 ml/min. CF I m activity eluted between 90 and 190 mM ammonium sulfate and was pooled. 60 ml (45 mg of protein) of this CF I m pool were adjusted to 1 M ammonium sulfate by addition of 2 M ammonium sulfate, pH 7.9, stirred on ice for 85 min and centrifuged for 50 min at 11,500 rpm (16,000 ϫ g max ) in a Sorvall SS34 rotor. The pellets were resuspended in 4.5 ml buffer without ammonium sulfate and stirred on ice for 2.5 h. The mixture was spun for 30 min at 14,000 rpm (16,000 ϫ g max ) in an Eppendorf centrifuge rotor F-45-18-11, and the supernatant was adjusted to 400 mM ammonium sulfate by addition of 2 M ammonium sulfate, pH 7.9. The solution was stirred on ice for 40 min, dialyzed against 500 ml of 450 mM ammonium sulfate buffer for 4 h, and spun as above in an Eppendorf centrifuge. The supernatant (5.5 ml, 11 mg) was applied to a 1-ml phenyl-Superose HR 5/5 FPLC column (Pharmacia) equilibrated in the same buffer. The column was washed with 10 column volumes of the 450 mM ammonium sulfate buffer and developed with a gradient (15 column volumes) of 450 to 0 mM ammonium sulfate at 0.25 ml/min. CF I m fractions eluting between 340 and 170 mM ammonium sulfate were pooled (3.4 ml, 500 g), dialyzed against 3 ϫ 300 ml of 20 mM ammonium sulfate buffer for 3 h, and loaded onto a 1-ml Mono Q HR 5/5 FPLC column at 0.5 ml/min. The column was washed with 10 column volumes of the dialysis buffer and developed with a gradient of 20 -300 mM ammonium sulfate (40 column volumes) at 0.5 ml/min. CF I m eluted between 100 and 150 mM ammonium sulfate.
Cleavage Assays-Cleavage reactions (25 l) were assembled on ice as follows: 10 l of premix (5 mM dithiothreitol, 0.025% (v/v) Nonidet P-40, 50 mM creatine phosphate (Boehringer Mannheim GmbH), 6.5% (w/v) polyvinyl alcohol, 0.5 unit/l RNAguard (Pharmacia), 0.025 g/l creatine kinase (Boehringer Mannheim GmbH), 1.25 mM cordycepin 5Ј-triphosphate (Boehringer Mannheim GmbH), 3.75 mM MgCl 2 ) were mixed with 60 fmol of CstF, 50 fmol of CPSF, 240 fmol of PAP, 4 l of crude CF II m and the amount of CF I m indicated. The volume was adjusted to 23.5 l with buffer E containing 50 mM Tris⅐HCl, pH 7.9, 10% (v/v) glycerol and 0.1 mM Na 2 -EDTA. The final ammonium sulfate concentration was kept below 20 mM. The reaction was started on ice by the addition of 15 fmol of 32 P-labeled SV40 late pre-mRNA resuspended in 2.5 g of tRNA in 1.5 l of water. Reactions were incubated at 30°C for 75-90 min and stopped by the addition of 75 l of proteinase K mix (50 l of 2 ϫ proteinase K digestion buffer (0.2 M Tris⅐HCl, pH 7.9, 300 mM NaCl, 25 mM Na 2 -EDTA, 2% (w/v) SDS), 1 l of 10 mg/ml proteinase K (Boehringer Mannheim GmbH), 0.25 l of 20 mg/ml glycogen (Boehringer Mannheim GmbH), and 23.75 l of water). After incubation for 30 min at 30°C, the RNA was precipitated with ethanol and analyzed on a 6% (w/v) polyacrylamide, 8.3 M urea sequencing gel, which was run until the xylene cyanole marker had reached the bottom (27,28).
To quantitate cleavage reactions, gels were exposed to Phosphor Imager screens for 1 h, the screens were scanned with a Phosphor Imager 425 (Molecular Dynamics),and the ImageQuant program (version 3.3, Molecular Dynamics); the amount of precursor and upstream cleavage product was determined with the IPLab Gel software (version 1.5, Signal Analytics Corp.). The values obtained were corrected by subtraction of the background signal of the gel in a region where no radioactivity was detectable. The value for the upstream cleavage fragment was further corrected for the different uridine contents of precursor and product. Cleavage activity was calculated by dividing the amount of upstream cleavage product by the sum of upstream cleavage product and precursor and multiplying the ratio by the amount of precursor added to the reaction. 1 unit corresponds to 1 fmol of upstream cleavage product obtained during the incubation time and at the temperature indicated.
Gel Retardation Assays-Reactions (12.5 l) were assembled on ice as follows: 5 l of premix (5 mM dithiothreitol, 0.025% (v/v) Nonidet P-40, 2.5% (w/v) polyvinyl alcohol, 0.5 unit/l RNAguard (Pharmacia), 1.25 mM cordycepin 5Ј-triphosphate (Boehringer Mannheim GmbH), and 2.5 mM Na 2 -EDTA) were mixed with the protein fractions indicated and the volume adjusted to 10.5 l with buffer E as described for cleavage assays. After the addition of 7.5 fmol of 32 P-labeled L3WT pre-mRNA resuspended in 2 l of water containing 1.25 g of tRNA, the reactions were incubated for 30 -35 min at 30°C. 10 l of the reactions were loaded directly on a native polyacrylamide/agarose composite gel running at 300 V at 4°C. The gel (15 cm long, 1.5 mm thick) contained 3% (w/v) acrylamide/bisacrylamide (80:1) and 0.5% (w/v) agarose in 0.5 ϫ TBE buffer (28) and was prerun at 4°C for 20 min at 300 V with 1 ϫ TBE running buffer. A sample of bromphenol blue/xylene cyanole was loaded in a separate lane, and electrophoresis was carried out until the xylene cyanole dye had migrated to approximately 4 cm from the bottom. The gel was dried and exposed to a Kodak X-Omat AR film.
UV Cross-linking of 3Ј End Processing Factors-Reactions (25 l) were set up on ice as follows. To 10 l of premix (5 mM dithiothreitol, 0.025% (v/v) Nonidet P-40, 6.5% (w/v) polyvinyl alcohol, 1.25 mM cordycepin 5Ј-triphosphate (Boehringer Mannheim GmbH), and 3.75 mM MgCl 2 ) 60 fmol of CstF, 50 fmol of CPSF, 240 fmol of PAP, and 900 fmol of CF I m (2 l of Mono Q fraction 37, Fig. 3) were added as indicated. The volume was adjusted to 25 l with 20 mM ammonium sulfate buffer as described for the purification of cleavage factor I m . 15 fmol of 32 P-labeled RNA were resuspended in water containing 2.5 g of tRNA and added to the reactions on ice. For the samples containing rRNA, 2.5 l of water containing 0.1 g of heat-denatured rRNA were mixed with the 32 P-labeled substrate prior to addition to the reaction mixture. Reactions were incubated at 30°C for 15 min and exposed to UV light (2 ϫ 200 mJ, UV Stratalinker 1800, Stratagene) on a siliconized glass plate at room temperature at a distance of 15 cm. Samples were treated with 4 ng/l RNase A (Sigma) for 15 min at 37°C, proteins precipitated by addition of 0.18 volumes of 100% (w/v) trichloroacetic acid, and separated on a 10% SDS-polyacrylamide gel (29). The gel was stained with silver, dried, and exposed to a Kodak X-Omat AR film.

RESULTS
Purification of Cleavage Factor I m -Six different protein factors have been shown to be required for specific cleavage and polyadenylation of mammalian primary mRNA transcripts. Whereas four of them, CstF, CPSF, PAP and PAB II, have been purified and most of their components cloned, the two others, CF I m and CF II m , remained poorly characterized. In order to analyze the cleavage factors further, we separated CF I m from CF II m on Mono S as described (5) and purified CF I m to near homogeneity. Column fractions were assayed for reconstitution of endonucleolytic cleavage of pre-mRNA derived from the polyadenylation site of the SV40 late transcription unit (SV40 pre-mRNA) in a reconstitution system containing CstF purified from HeLa cell nuclear extracts, CPSF isolated from calf thymus, recombinant bovine PAP and, after the separation of CF I m and CF II m , crude CF II m . The assays were performed in the presence of cordycepin 5Ј-triphosphate and MgCl 2 , since under these conditions endonucleolytic cleavage of the SV40 pre-mRNA was much more efficient than in the presence of ATP and EDTA. Although PAP was not essential for cleavage of the SV40 pre-mRNA to occur, it stimulated the reaction at least 2-fold (data not shown).
CF I m was purified from HeLa cell nuclear extracts by several fractionation steps, as shown schematically in Fig. 1. Nuclear extracts from HeLa cells turned out to be at least 50% more active when the nuclei were extracted with 200 mM am-monium sulfate instead of 300 mM KCl. Since this effect might be due to a stabilization of proteins by ammonium sulfate, all columns were eluted with ammonium sulfate, even though already small concentrations of ammonium sulfate (Ͻ30 mM) inhibit the cleavage reaction significantly (data not shown).
When diluted nuclear extracts were applied to a DEAE-Sepharose column, CstF and PAP did not bind to the column, whereas CF I m , CF II m , and CPSF bound and were eluted with a salt gradient. CF I m and CF II m were probably partially separated on this column, since a pool of the fractions showing cleavage activity was more active than the single fractions (data not shown). CF I m and CF II m were separated in the next purification step, a Mono S column, as described previously (5). The profile of the Mono S column is shown in Fig. 2. All column fractions were assayed for CF I m and CF II m activity in the presence of CstF, CPSF, PAP, and crude CF II m or crude CF I m , respectively. Crude CF I m and CF II m used for complementation in these assays were obtained from a 1-ml pilot Mono S column and had been identified before by testing all possible fraction combinations. The elution positions of CF I m and CF II m from the Mono S column described here are indicated by horizontal bars above the column profile (Fig. 2).
By means of ammonium sulfate precipitation and two additional chromatographic steps (see Fig. 1 and "Experimental Procedures"), CF I m was purified to near homogeneity. The purification is summarized in Table I. The quantitation of cleavage reactions proved to be difficult for the first steps of the purification. For obvious reasons, the calculation of activities gives reliable numbers only after the separation of CF I m and CF II m . The analysis of the separation of CF I m and CF II m on Mono S was complicated for two reasons: first, a nuclease interfered with the determination of CF I m activity and second, maximal cleavage activity depended on the optimal ratio of CF I m and CF II m at this stage of purification. Only after further purification of CF I m did cleavage activity depend linearly on the amounts of CF I m and CF II m (data not shown). Therefore, the elution positions of CF I m and CF II m are indicated only qualitatively in Fig. 2.
The profile of the final Mono Q column is shown in Fig. 3A. Three polypeptides with apparent molecular masses of 68, 59, and 25 kDa copurified with CF I m activity (Fig. 3, B and C). The fractions were also tested for RNA binding by gel retardation (Fig. 3D, see below). No RNA component could be detected in CF I m fractions of the Mono Q column after proteinase K treatment and 3Ј end labeling with [␣-32 P]cordycepin 5Јtriphosphate and PAP (30) (data not shown). On this column, the 68-kDa polypeptide eluted slightly later than the 59-and 25-kDa polypeptides, and the staining was less intense. Independent purifications over the same or different columns showed either the same effect or resulted in exactly comigrating polypeptides with an approximate equimolar ratio. Twodimensional gel electrophoresis on a nondenaturing polyacrylamide gel (31) in the first dimension and a denaturing SDSpolyacrylamide gel in the second dimension showed that the three polypeptides comigrated on the native gel (results not shown). These and other data suggest that all three polypeptides are part of CF I m (see below).
CF I m Can Bind to RNA in the Absence of the Other 3Ј End Processing Factors-To test wether CF I m can bind RNA either alone or in the presence of other 3Ј end processing factors, UV cross-linking reactions were set up essentially with the same amounts of protein factors and under the same conditions as cleavage reactions. Various combinations of proteins were preincubated with different 32 P-labeled pre-mRNA substrates in the presence of tRNA as an unspecific competitor at 30°C and then irradiated with UV light at room temperature. After treatment with RNase A and precipitation of the proteins, the samples were analyzed on a 10% SDS-polyacrylamide gel. The autoradiograph and the silver-stained gel are shown in Fig. 4, A and B, respectively.
UV Cross-linking of the 64-kDa subunit of CstF and of the 30-and 160-kDa subunits of CPSF to RNA has been described previously (7,14,16,(32)(33)(34). Under the conditions used here, only the cross-link of the 64-kDa subunit of CstF to the cleavage and polyadenylation substrate L3 was detectable and only in the presence of CPSF and PAP, which were shown previously to stabilize the binding of CstF to RNA (13, 35) (compare lanes 2 with 3 and 6 with 7, respectively). The cross-linking efficiency of the 64-kDa subunit of CstF was slightly reduced in  (36) with bovine serum albumin as a standard or were estimated from polyacrylamide gels stained with Coomassie Brilliant Blue and comparison with bovine serum albumin standards run on the same gels. The purification was monitored by quantitation of cleavage assays with similar incubation times (80 -85 min), which were performed within 1 week. For technical reasons, activity, yield, and purification factor cannot be determined for nuclear extracts and the DEAE-Sepharose column. AS, ammonium sulfate, units, definition as described under "Experimental Procedures." Step  the presence of heat denatured rRNA, which was added to the reaction simultaneously with L3 pre-mRNA and tRNA as a less structured unspecific competitor (compare lanes 3 and 7). Weak cross-links of the 64-kDa subunit of CstF could also be detected to the L3⌬1 substrate that carries a point mutation in the AAUAAA polyadenylation signal and is thus no longer able to bind CPSF (lane 11), and to the L3pre RNA that ends one nucleotide upstream of the natural cleavage and polyadenylation site of the L3 pre-mRNA and lacks the natural binding site for CstF (L3pre; lane 15). The unspecific cross-linking of CstF to RNA can probably be explained by the observation that CstF alone has no strict sequence requirements for binding to RNA and that CstF/RNA complexes are stabilized by CPSF and PAP, respectively. No cross-link could be detected to the splicing substrate Ad1, which does not contain a polyadenylation signal (lane 19).
The same set of RNA substrates was used for UV crosslinking reactions with CF I m . The peak fraction of the Mono Q column shown in Fig. 3 (fraction 37) was irradiated with UV light either alone or in the presence of CstF, CPSF, and PAP. Surprisingly, all three polypeptides of CF I m were cross-linked to the L3 pre-mRNA in the absence of any other 3Ј end processing factor (lane 4). The cross-links were assigned to the three polypeptides by superimposing the autoradiograph and the silver-stained gel; the migration positions of the 68-, 59-, and 25-kDa polypeptides of CF I m on the silver-stained gel are indicated by arrowheads in Fig. 4. The 68-and 59-kDa polypeptides comigrated exactly with signals detected on the autoradiograph, whereas the 25-kDa polypeptide detected by silver staining migrated slightly faster than the signal detected by autoradiography. This is probably due to retardation of the cross-linked portion of the polypeptide by covalently bound residual RNA nucleotides. The difference in the migration behavior is detectable for small polypeptides like the 25-kDa polypeptide but not for larger ones such as the 59-and 68-kDa polypeptides of CF I m .
The signal of the cross-link of the 68-kDa polypeptide is weaker than the signal of the 59-kDa polypeptide. The crosslinking efficiency of proteins to RNA largely depends on the amino acid composition of the RNA binding site and, since the RNA substrate was labeled with [␣- 32  CF I m Stabilizes the CPSF⅐RNA Complex-For gel retardation assays, aliquots from the Mono Q column shown in Fig. 3 were preincubated with L3 pre-mRNA for 30 min at 30°C and loaded directly onto a native polyacrylamide/agarose composite gel at 4°C. Reactions were done under similar conditions as cleavage assays, but contained less polyvinyl alcohol, and EDTA instead of MgCl 2 (see "Experimental Procedures"). The autoradiograph of the dried gel is shown in Fig. 3D. The protein-RNA complex comigrated exactly with CF I m activity, but only CF I m peak fractions gave rise to a complex detectable as a distinct band. Smaller amounts of CF I m resulted in apparently less stable protein-RNA complexes. The faint band detectable in almost all lanes migrating above the protein-RNA complexes most likely results from a RNA with a different structure.
In order to compare the ability of all known 3Ј end processing factors to bind RNA in gel retardation experiments, equal amounts (650 fmol) of CF I m , CstF, CPSF, and PAP were incubated separately with 7.5 fmol of 32 P-labeled L3 pre-mRNA and run on a native gel (Fig. 5). Even with this high excess of protein, only CPSF formed a distinct complex with the pre-mRNA substrate (lane 4), whereas CF I m ⅐RNA and CstF⅐RNA complexes formed short smears running above the free RNA (lanes 2 and 3).  Fig. 3). As expected, no complex of CstF or PAP with L3 pre-mRNA could be detected (lanes 6 and 10) and the signal of the CPSF⅐RNA complex was strongly reduced and less distinct than with larger amounts of protein (lane 8). Addition of CstF or PAP to CF I m did not change the CF I m ⅐RNA complex (lanes 7 and 11), but CF I m , CPSF, and L3 pre-mRNA formed a complex that migrated more slowly than the CPSF⅐RNA complex (lane 9). Furthermore, CPSF formed a slower migrating complex in the presence of CstF (lane 12, complex A) which was supershifted and stabilized upon the addition of PAP (lane 14, complex C). The CstF⅐CPSF⅐RNA complex as well as the CstF⅐CPSF⅐PAP⅐RNA complex were supershifted upon the addition of CF I m , and the signal was significantly enhanced (lanes 13 and 15, complexes  B and D, respectively). The CPSF⅐CF I m ⅐RNA complex was already detectable in the presence of 150 fmol of CF I m (0.5 l of the Mono Q fraction 34, Fig. 3), which corresponds to the amount used in the cleavage assay (Fig. 3C). 50 fmol of CF I m (0.15 l of the Mono Q fraction 34, Fig. 3) were sufficient to induce a supershift of the CstF⅐CPSF⅐RNA and CstF⅐CPSF⅐ PAP⅐RNA complexes. The signals increased in all three cases with increasing amounts of CF I m (result not shown). DISCUSSION We report here the purification of CF I m from HeLa cell nuclear extracts. Three polypeptides of 68, 59, and 25 kDa copurified with CF I m activity. Several lines of experimental evidence support the notion that these polypeptides are true subunits of CF I m . 1) All three polypeptides could be UV crosslinked to the cleavage and polyadenylation substrate L3. 2) The polypeptides formed a complex that was stable upon native gel electrophoresis.
3) The three polypeptides copurified during different purification procedures, although they were partially separated on certain column matrixes. We, therefore, believe that all three polypeptides are part of CF I m , but we cannot rule out at this stage that the 59-kDa and/or 25-kDa polypeptide are degradation products of the 68-kDa polypeptide that are still able to bind to RNA and are at least partially active. Partial proteolysis of the 68-kDa polypeptide would be one possible explanation for the observation that the largest subunit was slightly less abundant than the other two polypeptides in some CF I m preparations. Attempts to reconstitute CF I m from the single polypeptides obtained by elution from a SDS-polyacrylamide gel were unsuccessful. Thus, the true composition of CF I m remains uncertain until cDNAs coding for the polypeptides will be cloned.
The question whether CF I m can bind to RNA either alone or in the presence of other 3Ј end processing factors was addressed by two different methods, UV cross-linking and gel retardation assays. Whereas by UV cross-linking even weak interactions of proteins and RNA can be detected, and the RNA binding polypeptide(s) can be identified on a SDS-polyacrylamide gel, gel retardation assays can reveal protein-RNA interactions in solution. Both methods allowed the detection of a CF I m ⅐RNA complex in the absence of any other 3Ј end processing factor. UV cross-links of CF I m could only be detected to the cleavage and polyadenylation substrate L3, but not to the splicing substrate Ad1. The preference of CF I m for L3 was in principle confirmed by gel retardation assays, but the CF I m ⅐L3 shift could be competed by increasing amounts of Ad1, although less efficiently than by L3 itself (data not shown). No cross-links of CF I m to L3pre mRNA, which ends one nucleotide upstream of the natural cleavage and polyadenylation site, could be detected. It is not possible to decide from the experiments shown here where precisely CF I m binds on L3 pre-mRNA. Further experiments are needed to map the region on the pre-mRNA substrate that is bound by CF I m .
Gel retardation assays with purified CstF, CPSF, PAP, and crude cleavage factors, either alone or in different combinations, and L3 pre-mRNA have been described before (13,16,17,19). It was shown that formation of the CPSF⅐RNA complex depends on AAUAAA, that the CPSF⅐RNA complex is stabilized by PAP, and that addition of CstF causes the formation of a slower migrating complex. Under the conditions used, neither a CstF⅐RNA nor a PAP⅐RNA complex could be detected. These results are in good agreement with those presented here, which were obtained with more highly purified components. We further demonstrated the binding of CF I m to L3 pre-mRNA in the absence of any other protein factor. Neither CstF nor PAP had an influence on the binding of CF I m to L3 pre-mRNA, but the addition of CPSF led to the formation of a more slowly migrating complex. Furthermore, CF I m stabilized the CstF⅐CPSF⅐ RNA as well as the CstF⅐CPSF⅐PAP⅐RNA complex and resulted in slower migrating complexes. In contrast, Gilmartin and Nevins (17) have reported previously that the addition of crude cleavage factors abolished the formation of the CPSF⅐RNA complex and destabilized the CstF⅐CPSF⅐RNA complex. The results presented here were obtained in the absence of CF II m , because this factor is presently only available as a relatively impure fraction. The destabilization of the CPSF⅐RNA and the CstF⅐CPSF⅐RNA complex by the cleavage factors (17) may thus be either an unspecific effect, since less purified cleavage factors were used in these experiments, or it may have been caused by CF II m .
The results obtained from the UV cross-linking experiments and gel retardation assays are consistent with the suggestion that CF I m binds to pre-mRNA with a certain degree of specificity and interacts with CPSF, but not with CstF and PAP upon binding to RNA.
It remains to be shown which of the protein factors essential for the cleavage reaction acts as the actual endonuclease. This question can only be resolved after the other cleavage factor, CF II m , has been characterized. The purification of all factors involved in cleavage and polyadenylation of primary mRNA transcripts should allow it to study in more detail the assembly of the components involved in 3Ј RNA processing, to investigate the sequence requirements further, and to identify regulatory factors that participate in the 3Ј end formation of pre-mRNA in a well defined, fully reconstituted system.