Purification and Characterization of a Mammalian Polyadenylate Polymerase Involved in the 3’ End Processing of Messenger RNA Precursors*

A polyadenylate polymerase involved in the polyad- enylation of pre-mRNA has been purified 6,000-fold to apparent homogeneity from extracts of calf thymus. In the last purification step, anion exchange chromatography separates the enzyme into three major peaks that are indistinguishable by other physical or func-tional criteria. On denaturing polyacrylamide gels, the two predominant forms of poly(A) polymerase have molecular weights of 57,000 and 60,000. In solution, the enzyme is a monomer. It polymerizes exclusively ATP. The reaction is distributive and proceeds linearly without any lag phase. The requirement for a primer can be satisfied by any of a number of polyribonucleo-tides. A significantly higher activity in the presence of Mn2+ as opposed to Mgz+ is due to a hundredfold higher affinity for the primer terminus. In the presence of Mg2+ and of a specificity factor partially purified from HeLa cells, the enzyme specifically polyadenylates an RNA that ends at the natural adenovirus L3 polyade- nylation site. This reaction depends on the AAUAAA polyadenylation signal. The 3’ ends of mRNA in higher eukaryotes are generated by posttranscriptional processing rather than transcription termination. A primary transcript extending hundreds or thousands of nucleotides beyond the mature 3‘ end is cleaved endonucleolytically 10-30 nucleotides downstream of the highly conserved sequence AAUAAA and upstream of a less conserved GU- or U-rich sequence element. The newly generated 3‘

A polyadenylate polymerase involved in the polyadenylation of pre-mRNA has been purified 6,000-fold to apparent homogeneity from extracts of calf thymus. In the last purification step, anion exchange chromatography separates the enzyme into three major peaks that are indistinguishable by other physical or functional criteria. On denaturing polyacrylamide gels, the two predominant forms of poly(A) polymerase have molecular weights of 57,000 and 60,000. In solution, the enzyme is a monomer. It polymerizes exclusively ATP. The reaction is distributive and proceeds linearly without any lag phase. The requirement for a primer can be satisfied by any of a number of polyribonucleotides. A significantly higher activity in the presence of Mn2+ as opposed to Mgz+ is due to a hundredfold higher affinity for the primer terminus. In the presence of Mg2+ and of a specificity factor partially purified from HeLa cells, the enzyme specifically polyadenylates an RNA that ends at the natural adenovirus L3 polyadenylation site. This reaction depends on the AAUAAA polyadenylation signal.
The 3' ends of mRNA in higher eukaryotes are generated by posttranscriptional processing rather than transcription termination. A primary transcript extending hundreds or thousands of nucleotides beyond the mature 3' end is cleaved endonucleolytically 10-30 nucleotides downstream of the highly conserved sequence AAUAAA and upstream of a less conserved GU-or U-rich sequence element. The newly generated 3' end is then elongated by the addition of approximately 200 adenylate residues (for reviews, see Nevins, 1983;Birnstiel et al., 1985;Humphrey andProudfoot, 1988 Manley, 1988;Wickens, 1990). The entire processing reaction can be carried out i n vitro, independent of transcription, on premade precursor RNAs containing the appropriate sequences (Moore and Sharp, 1984;1985). I n vitro as well as i n uiuo, cleavage and polyadenylation are tightly coupled; cleaved RNA lacking a poly(A) tail cannot usually be detected. I n vitro, the accumulation of cleaved RNA can be observed when polyadenylation is inhibited by EDTA or by a chain-terminating ATP analogue (Moore and Sharp, 1985). Conversely, polyadenylation can be assayed independently of cleavage through the * 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. This work was supported by grants from the Schweizerischer Nationalfonds and the Kantons of Basel to Dr. Walter Keller.
Fractionation of HeLa cell nuclear extracts active in 3' end processing revealed that at least four factors are required for the reaction. One of these factors is a poly(A) polymerase (Takagaki et al., 1988;Keller, 1988, 1989;McDevitt et al., 1988). In the presence of manganese, this enzyme by itself nonspecifically polyadenylates a variety of RNAs, whereas in the presence of magnesium it is rather inactive. Under the latter conditions, a second factor (cleavage and polyadenylation factor (CPF);' also called SF or PF2) endows the polymerase with activity specific for precleaved RNAs that have an AAUAAA sequence close to their 3' ends Keller, 1988, 1989;Takagaki et al., 1988Takagaki et al., , 1989McDevitt et al., 1988). CPF itself is able to bind to AAUAAA (Gilmartin and Nevins, 1989).' Thus, two factors suffice for the AAUAAA-dependent polyadenylation of precleaved RNA precursors that end at or near the natural cleavage site. At least two more factors, in addition to CPF and poly(A) polymerase, are required for the in vitro reconstitution of the cleavage reaction (Christofori and Keller, 1988;Gilmartin et al., 1988;Gilmartin and Nevins, 1989;Takagaki et al., 1989).3 Poly(A) polymerase is believed to play a role in the cleavage reaction based on the copurification of polyadenylation and cleavage activities (Takagaki et al., 1988(Takagaki et al., , 1989Keller, 1988, 1989;Gilmartin and Nevins, 1989). The extent of the poly(A) polymerase requirement is not entirely clear; differences between different substrate RNAs have been observed (Takagaki et al., 1988;Ryner et al., 1989a). The participation of poly(A) polymerase in the cleavage reaction may explain the tight coupling between cleavage and polyadenylation.
Processing reactions are carried out i n vitro on minute quantities of radiolabeled RNA that is then analyzed by gel electrophoresis. As the substrate is very sensitive to nucleases, most assays of this type are carried out in HeLa cell nuclear extracts and cannot be done in tissue homogenates. This limitation of in vitro assays to tissue culture cells has certainly hampered the purification of the apparently scarce processing components. With one possible exception (a cleavage factor called CF1 by Gilmartin and Nevins (1989)), none has been purified to homogeneity.
Poly(A) polymerases have been purified before from a large number of different sources based on the enzymes' ability to incorporate labeled ATP into acid-precipitable material in the The abbreviations used are: CPF, cleavage and polyadenylation factor; BSA, bovine serum albumin; Hepes, 4-(2-hydroxyethyl)-lpiperazineethanesulfonic acid; SDS, sodium dodecyl sulfate.
K. Lang, E. Wahle, and W. Keller, unpublished data. 3131 nonspecific elongation of various RNA primers (for reviews, see Edmonds and Winters, 1976;Jacob and Rose, 1978;Edmonds, 1982Edmonds, ,1990. I have used such a nonspecific assay that is fast, quantitative, and not very sensitive to nucleases to monitor the purification to homogeneity of a poly(A) polymerase from calf thymus. The purified enzyme was also active in the in vitro reconstitution of specific AAUAAA-dependent polyadenylation of RNA precursors.

EXPERIMENTAL PROCEDURES
Nucleic Acids-Poly(A) (Boehringer Mannheim) was dissolved in water, extracted with phenol/chloroform, ethanol-precipitated, dissolved, and dialyzed against water. Oligo(A) (A12.,B; Pharmacia LKB Biotechnology Inc.), tRNA and poly(U) (both from Boehringer Mannheim) were dissolved in water and used without further purification. Poly(C) (Pharmacia) and poly(G) (Boehringer Mannheim) had to be dephosphorylated before they could serve as primers. Approximately 400-500 pg of each polynucleotide were incubated with 220 units of calf intestinal phosphatase in 250 pl of buffer (Sambrook et al., 1989) for 30 min at 37 "C. The polynucleotides were purified by proteinase K digestion, phenol extraction, and ethanol precipitation (Sambrook et al., 1989). Concentrations of homopolynucleotides were determined photometrically with extinction coefficients given by the suppliers or taken from the Pharmacia catalogue. For heterogeneous RNA, an A2e0 of 1 was assumed to correspond to 40 pg/ml.
From the amount of radioactive phosphate incorporated into dephosphorylated poly(A) by polynucleotide kinase and [y3'P]ATP, the average chain length was calculated as 67 adenylate residues. Direct analysis by gel electrophoresis of the end-labeled material and scintillation counting of gel slices lead to an estimate of 88 residues. For calculations, a chain length of 80 was assumed. The chain length distribution of poly(U) was similar to that of poly(A) as judged by gel electrophoresis after exchange labeling with polynucleotide kinase. Poly(G) and poly(C) consisted predominantly of short chains of around 15 and 6 residues, respectively.
RNAs used as substrates for specific polyadenylation were as described (Christofori and Keller, 1988); L3pre was derived from the adenovirus-2 L3 polyadenylation site. It contained at its 5' end 24 nucleotides of vector sequence followed by 41 nucleotides of adenovirus sequence, including the AAUAAA signal and ending 1 nucleotide upstream of the natural polyadenylation site. L3preA was identical except for a U to G mutation in the AAUAAA sequence. Both RNAs were made with a GpppG cap by in vitro transcription of truncated plasmid DNA with SP6 RNA polymerase in the presence of [ C X -~'~] UTP as described (Frendewey and Keller, 1985;Kramer and Keller, 1985) except that the UTP concentration was 0.1 mM. 5' labeling of RNA with polynucleotide kinase either after treatment with calf intestinal phosphatase or by exchange labeling was done according to Sambrook et al. (1989).
Proteins-CPF was partially purified from HeLa cell nuclear extracts through the Mono Q step as described (Christofori and Keller, 1988) with an additional Blue Sepharose column between heparin Sepharose and Mono Q. SP6 RNA polymerase and creatine kinase were from Boehringer Mannheim. Polynucleotide kinase was from New England Biolabs, N-glycanase from Genzyme, potato acid phosphatase from Sigma, and calf intestinal phosphatase from Biofinex (Praroman, Switzerland). Proteins used as markers in gel filtration were from Sigma (horse heart cytochrome c and yeast alcohol dehydrogenase) and Pharmacia (all others). Stokes radii were taken from Siege1 and Monty (1966) and information supplied by Pharmacia. Sedimentation constants were those given by Sober (1970).
Purified poly(A) polymerase was diluted in 50 mM Tris-HCI, pH 8.1, lo@ mM KCl, 1 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol, 0.05% Nonidet P-40, 200 pg/ml bovine serum albumin. All buffers were adjusted as 1 M stock solutions at room temperature. Only buffers used for the determination of the pH optimum of poly(A) polymerase were measured after dilution to 25 mM and at 37 "C with a pH meter calibrated at 37 "C.
Enzyme Assays-The standard poly(A) polymerase assay contained, in 25 MI, 25 mM Tris-HC1, pH 8.3 (measured at 25 mM and 37 "C), 40 mM KC1, 0.5 mM MnC12, 0.05 mM EDTA, 0.5 mM dithiothreitol, 0.1% Nonidet P-40, 10% glycerol, 200 pg/ml bovine serum albumin, 0.5 mM [w3'P]ATP (15-100 cpm/pmol), and 6.5 pg of poly(A). Enzyme was added to the mixture assembled on ice and the reaction was started by transfer to 37 "C. After 10 min, 100 pl of 10% trichloroacetic acid (w/v), 0.1 M sodium pyrophosphate was added. After mixing and chilling, the precipitate was collected by filtration through glass fiber filters, washed four times with 1% trichloroacetic acid, 0.01 M sodium pyrophosphate and once with ethanol and counted in a scintillation counter. One unit of poly(A) polymerase incorporated 1 pmol AMP in 1 min. When oligo(A) was substituted for poly(A), it was used at 1.25 pglassay. In this case, the reaction was terminated by application of the complete reaction mixture to a piece of DEAE paper. The paper was pretreated and washed essentially as described (Stayton and Kornberg, 1983). MgC1, replacing MnC& was used at 6 mM.
For kinetic experiments, reactions were initiated by the addition of enzyme to prewarmed reaction mixture. Incubations were for 5 or 10 min. K M and VmaX were obtained from Lineweaver-Burk or Eadie-Hofstee plots (Fersht, 1985).
Conditions for polyadenylation of the adenovirus-derived RNA (Lapre) were essentially as described (Christofori and Keller, 1988). Creatine kinase (100 or 10 pg/ml) was included. Reactions were terminated by the addition of proteinase K mix (20 pg of proteinase K and 5 or 10 pg of tRNA in 180 pl of 100 mM Tris-HC1, pH 7.5, 12.5 mM EDTA, 150 mM NaCI, 1% SDS). After 20 min at 30 "C, RNA was precipitated by the addition of 500 pl of ethanol, pelleted in a refrigerated microcentrifuge, washed with 70% ethanol, and analyzed on denaturing polyacrylamide gels (Sambrook et al., 1989).
Other Methods-Protein concentrations were determined according to Bradford (1976) with BSA as a standard. SDS-polyacrylamide gels were run according to Laemmli (1970) and stained with Coomassie Brilliant Blue.

RESULTS
Purification of Poly(A) Polymerase-Homogenates of calf thymus prepared at 50 mM KC1 contained a poly(A) polymerase activity that could be readily assayed by the primerdependent incorporation of labeled ATP into acid-precipitable material. Higher concentrations of KC1 in the homogenization buffer (up to 300 mM) did not increase the yield of enzyme activity. Poly(A) in the presence of Mg2+ or Mn2+ was a 2-7-fold better primer than total RNA from yeast or HeLa cells under the conditions used in these preliminary assays. With various primers, incorporation was 8-20-fold higher with Mn2+ than with Mg2+. Even in the crude extract, AMP incorporation was almost entirely dependent on the addition of primer RNA; addition of poly(A) in the presence of Mn2+ stimulated AMP incorporation 40-fold. The activity of the crude extract was completely stable during storage on ice for at least 5 h (data not shown).
Crude extracts also contained an activity that stimulated purified poly(A) polymerase in the elongation of a poly(A) primer. Under standard assay conditions (see "Experimental Procedures"), the stimulation was 2-3-fold. Under other conditions, it was up to 10-fold, and a weaker stimulation was detected even with a 200-fold diluted extract that had no detectable poly(A) polymerase activity by itself. The stimulatory activity was not observed when an oligo(A) primer was used (data not shown). Therefore, although the faster and more economical poly(A) assay was used routinely, all purification steps were checked with the oligo(A) assay as well. By means of ammonium sulfate precipitation and six chromatographic steps, poly(A) polymerase was purified about 6,000-fold with a yield of 10% (Table I). Little more than 1 mg of pure enzyme was obtained from 2 kg of tissue. There was no indication that another poly(A) polymerase was lost during the purification; the recovery in any individual step was high, and single activity peaks were observed throughout the purification. Only the final Mono Q column separated the activity into at least three distinct peaks (Fig. 1A ). Polypeptides with molecular weights of 57,000-62,000 copurified with the activity in all peaks (Fig. 1B). The same set of proteins was also correlated with the activity peaks in three other kinds of columns (hydroxyapatite, gel filtration, and Mono S cation exchange chromatography) and in a number of independent preparations (data not shown).
All three poly(A) polymerase peaks had very similar specific activities in the extension of poly(A) or oligo(A) ( Table I and data not shown).
The good correlation between the abundance of the 60-kDa proteins and poly(A) polymerase activity strongly suggests that the minor bands detected in some of the fractions were unrelated contaminants.
Physical Properties-By SDS-polyacrylamide gel electro- Purification of poly(A) polymerase The purification from 2 kg of calf thymus was carried out at or close to 0 "C. Between purification steps, the samples were frozen in liquid nitrogen and stored a t -80 "C. Columns were run a t approximately 1 volume/h unless indicated otherwise. Fresh calf thymus was obtained from a local slaughterhouse, placed in ice for transport, and stored at -80 "C. One kg was allowed to thaw in 2 liters of buffer 1 and homogenized in a Waring Blendor for 30 s each a t low, medium, and high speeds. The homogenate was centrifuged for 1 h at 10,000 rpm (16,000 X gmax) in a Sorvall GSA rotor, and the supernatant was decanted through wide mesh gauze (resulting in fraction I). Solid ammonium sulfate was added (0.134 g for each ml of fraction I; 25% saturation), the mixture was stirred in ice for 2 h, and then it was centrifuged as above. Ammonium sulfate was added (0.115 g for each milliliter of supernatant; 45% saturation), and stiFring and centrifugation were repeated. The pellets were resuspended in a small volume of 50 mM buffer 2, and two such preparations were dialyzed together for 14 h against 4 X 4.5 liters of 50 mM buffer 2 (resulting in fraction 11). The two preparations were diluted separately to the conductivity of 50 mM buffer 2. Each was mixed with 4 volumes of 1 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol; one batch had to be further diluted with 3 volumes of the same solution containing 10 mM potassium phosphate, pH 7.2. The solutions were spun for 30 min a t 10,000 rpm in a Sorvall GSA rotor, and each of the two preparations was applied separately to a Bio-Rex 70 column of 7 X 28 cm (13-14 mg of protein loaded per ml of column bed volume) equilibrated in 50 mM buffer 2. The column was washed with 1 volume of 50 mM buffer 2 and developed with a gradient (5 column volumes) of 50-500 mM buffer 2. Active fractions around 150 mM potassium phosphate were pooled (resulting in fraction 111) and solid ammonium sulfate was added to 70% saturation (0.436 g added per ml). The mixtures were stirred overnight in ice and spun for 1 h a t 9,000 rpm (13,000 X g,,,) in a Sorvall GSA rotor. The pellets were resuspended in a small volume of buffer 3, dialyzed for 3 h against 2 liters of buffer 3 containing 20% glycerol, and clarified by centrifugation for 10 min a t 15,000 rpm (27,000 X g , J in a Sorvall SS34 rotor. Each of the two preparations was applied separately to a Sephacryl $300 H R gel filtration column (5 X 90 cm; approximately 35 ml loaded per run), equilibrated, and run in buffer 3 at 135 ml/h. Active fractions from both runs were combined (resulting in fraction IV), diluted with 1 volume of buffer 4 lacking KCl, and applied to a column of Blue Sepharose (2.6 X 26.5 cm, 2 mg of protein applied per ml of column volume) equilibrated in buffer 4 with 50 mM KCI. The column was washed with 1 volume of 50 mM KC1 and eluted with a gradient (10 column volumes) of 50-1,000 mM KC1 in buffer 4. Poly(A) polymerase was eluted around 200 mM KC1 (resulting in fraction V). CaC12 (1 mM final concentration) was added and the enzyme was pumped onto a hydroxyapatite column (Ultrogel HA; 1.5 X 13 cm; 2 mg of protein applied per ml of column volume) equilibrated in buffer 5. The column was washed with 1 volume of buffer 5 and eluted with a gradient (10 column volumes) of buffer 5 to buffer 6. Poly(A) polymerase was eluted around 120 mM phosphate (resulting in fraction VI). Only two-thirds of fraction VI were carried through the remaining two steps of the purification. All numbers in the table are corrected as if all of fraction VI had been used. Nonidet P-40 was added to 0.05%, and the solution was dialyzed for 6 h against 1 liter of buffer 7 containing 20 mM KCl. The conductivity was adjusted to that of the dialysis buffer by dilution with buffer 7 lacking KCI, and the enzyme was loaded on a 1-ml Mono S fast protein liquid chromatography column. It was eluted with a 25-ml gradient of 20-250 mM KC1 in buffer 7 at 0.5 ml/min. Poly(A) polymerase, desorbed a t 90 mM KC1 (resulting in fraction VII), was dialyzed for 5 h against 1 liter of buffer 8 lacking KC1 and loaded on a 1-ml Mono Q fast protein liquid chromatography column. The column was developed with a 20-ml gradient of 0-200 mM KC1 in buffer 8 a t 0.5 ml/min. Poly(A) polymerase was found in three main peaks as shown in Fig. 1   polymerase. An aliquot of the hydroxyapatite pool (fraction VI) was taken before the detergent Nonidet P-40 was added to the rest. The detergent-free material, purified to near homogeneity by Mono S chromatography, contained mostly the 57-kDa form of poly(A) polymerase and was used for the determination of the enzyme's native molecular weight. The Stokes radius as determined by gel filtration and comparison with known proteins (Siegel and Monty, 1966) was 31 A (Fig. 2).
A sedimentation coefficient of 3.4 S was obtained from centrifugation through glycerol gradients along with markers (Fig. 3). From these numbers and an assumed partial specific volume of 0.73 ml/g, a native molecular weight of 45,000 was calculated (Siegel and Monty, 1966). Thus, under the conditions used (50 mM Tris-HC1, pH 8.1, 100 mM KCI; initial enzyme concentrations 0.2-0.7 mg/ml), the enzyme was a monomer.
Poly(A) polymerase appeared not to be glycosylated. In one experiment, aliquots taken from the various Mono Q peaks ( Fig. 1) were transferred from an SDS-polyacrylamide gel to a membrane and then oxidized with sodium periodate. Dialdehydes generated from sugars by the oxidation were detected by reaction with a hydrazide derivative of a hapten (digoxigenin), followed by incubation with alkaline phosphatase-conjugated antibody and phosphatase staining (Haselbeck and Hosel, 1990). In this assay, poly(A) polymerase was clearly negative. In another set of experiments, the proteins were separated on SDS gels, transferred to membranes, and incubated with biotinylated lectins (concanavalin A or wheat germ agglutinin). Lectin binding was detected with a biotinavidin-alkaline phosphatase complex and phosphatase staining. No reaction of poly(A) polymerase was found beyond a quite pronounced lectin-independent binding of the detection reagent. Finally, treatment with N-glycanase (which removes almost all N-linked carbohydrates) or with 0.1 N NaOH (which removes all 0-linked sugars) did not result in an altered electrophoretic mobility of any of the poly(A) polymerase bands found in Mono Q fraction 20 (data not shown).
Similarly, treatment of aliquots of all Mono Q peak fractions with either potato acid phosphatase or calf intestinal phosphatase did not change the electrophoretic mobility of poly(A) polymerase (data not shown). While this result does not exclude phosphorylation of the enzyme, it suggests that the electrophoretic heterogeneity is not caused by phosphorylation. Stability-The enzyme prepared in the presence of the detergent Nonidet P-40, as described in Table I, was somewhat unstable. Fraction 24 of the Mono Q column (Fig. 1) did not lose significant activity during 8 days at -80 "C, 10 cycles of freezing (in liquid nitrogen) and thawing, or during storage on ice for 30 h. After 20 days of storage at -80 "C, Mono Q fractions 9, 19, 23, and 33 had lost 44, 22, 7, and 5% activity, respectively. After 100 days of storage, fraction 24 had been inactivated by 32%. An aliquot of the hydroxyapatite pool (fraction VI) of the same preparation, removed before the addition of Nonidet P-40 and further purified in the absence of detergent (see above) was more unstable; storage at -80 "C for 5 days, including two cycles of freezing and thawing, resulted in 35% inactivation. After storage on ice for 28 h, 20% of the remaining activity was lost. The protein concentration of this preparation was comparable to that of the Mono Q peak fractions of Fig. 1. An independent detergentfree preparation that had a lower protein concentration showed even greater instability. This preparation was also stabilized by detergents (data not shown). Reaction Requirements-The pH optimum of poly(A) polymerase, assayed with a poly(A) primer in the presence of Mn2+, was 8.3. At pH 7.15 and 9.16, the activity was 55 and 62%, respectively. With oligo(A) as a primer in the presence of Mg", the pH optimum was near 8.0. At pH 8.3, the activity was 30% lower. The pH of the Tris-HC1 buffers used in these experiments was measured under reaction conditions (see "Experimental Procedures").
The optimal concentration of Mn2+ was equal to the concentration of ATP: 0.5 mM at 0.5 mM ATP (50% activity between 0.3 and 0.4 mM and between 0.7 and 1 mM) and 1 mM at 1 mM ATP. Mg2+ could substitute for Mn2+ with a broad optimum around 4-6 mM.
The requirement for KC1 depended on the primer and the divalent cation used. With poly(A) and Mn", the optimum was 40 mM. 2 and 80 mM KC1 resulted in 28 and 59% activity, respectively. With oligo(A) and Mn2+, the KC1 optimum was between 20 and 60 mM, and the activity varied less than 2fold between 2 and 80 mM. A similar salt dependence was observed in the presence of poly(A) and Mg2+. With oligo(A) and Mg", KC1 was inhibitory at all concentrations higher than 2 mM with 50% inhibition at 30 mM. In summary, both poly(A) and Mn2+ (at 0.5 mM) favored poly(A) polymerase activity at 40 mM KCl, whereas oligo(A) and Mg2+ (at 6 mM) caused higher activity in the absence of salt.
Interaction with Nucleoside Triphosphates-Titration of ATP in the presence of 6 mM Mg2+ and 0.05 mg/ml oligo(A) (far below saturation) revealed an apparent KM of 0.3 mM.
With a poly(A) primer and Mn2+ equimolar to ATP, the enzyme was saturated at 0.5 mM ATP. Under these conditions, the ATP dependence did not follow Michaelis-Menten kinetics, presumably due to the mutual dependence of Mn2+ and ATP concentrations (see above). In the presence of Mn2+, G T P was not incorporated a t measurable levels, whereas UTP, CTP, and dATP (all at 0.5 mM) were used a t 0.2, 0.5, and 4%, respectively, of the rate of ATP.
Interaction with the Primer-Kinetic data characterizing the use of poly(A) and oligo(A) primers in the presence of Mg2+ or Mn2+ are summarized in Table 11. Oligo(A) was used with a slightly lower affinity than poly(A) but a higher V,,,.
The much lower activity of poly(A) polymerase in the presence of Mg2+, as compared to Mn2+, was due to a 30-100-fold lower affinity for the primer. Reaction rates at primer saturation were comparable for both cations. The turnover number was 3-7/s with a poly(A) primer and 20-4O/s with an oligo(A) primer.
In a preliminary survey of different polynucleotide primers in the presence of Mg2+, poly(A) polymerase appeared to have a moderate preference for poly(A) ( Table 111). An accurate comparison of all primers can, however, not be made due to their different sizes and other complications (see the discussion of poly(U) below). The elongation of radiolabeled poly(A) and L3pre RNA (see "Experimental Procedures") by poly(A) polymerase alone in the absence of CPF was analyzed by gel electrophoresis. This experiment also showed a preference for poly(A) (data not shown). A 2-3-fold increase in primer activity of poly(A) after phosphatase treatment (Table  111) indicated that 50-60% of the 3' ends must have had a phosphate rather than a hydroxyl group. The kinetic data given for poly(A) in Table I1 are thus probably off by a factor of 2. This does not affect the major conclusions. The oligo(A) used was not tested after dephosphorylation. However, extension of 5' labeled primers with cold ATP indicated that essentially all chains could serve as primers (see below). With poly(A), oligo(A), or tRNA as primers, AMP incorporation proceeded a t a constant rate without any significant lag phase ( Fig. 4; data not shown). With a poly(U) primer, however, a short lag phase was visible (Fig. 4). With this primer, the rate of polymerization decreased again after the incorporation of less than 1 AMP/primer (data not shown). These kinetics suggest that the addition of the first AMP to a poly(U) chain improved its ability to act as a primer so that only a subset of all chains was elongated. Once several adenylate residues had been added, further elongation was probably inhibited by hybrid formation. Experiments in which the elongation of radiolabeled poly(U) was measured by gel electrophoresis confirmed this interpretation (data not shown).

Kinetic parameters of poly(A) synthesis
Poly(A) polymerase elongated an oligo(A) primer in a distributive manner. Within the limitations imposed by the heterogeneity of the primer used, AMP residues appeared to be added one at a time to the whole population of primers, as measured by gel electrophoresis (Fig. 5). Since in this experiment primers were present in a more than 1000-fold excess over enzyme, this means that poly(A) polymerase dissociated after every polymerization event to bind to a new primer. The same behavior was seen with Mg2+ or Mn2+. With smaller amounts of primer and higher concentrations of ATP and enzyme, poly(A) chains of several hundred nucleotides in length were obtained (data not shown).
Taken together, the data show that under physiological conditions, i.e. in the presence of Mg2+, poly(A) polymerase activity is limited by a low affinity for the primer terminus. The enzyme has no primer specificity beyond a slight preference for either poly(A) or a 3' terminal adenylate residue.
Sensitivity to a n SH-alkylating Reagent-After incubation of poly(A) polymerase on ice for 10 min with 10 mM Nethylmaleimide and subsequent addition of 50 mM dithiothreitol, 87% of the activity remained. The enzyme thus appears not to have any essential SH groups. In two controls, water was substituted for N-ethylmaleimide, or 50 mM dithiothreitol was added first. Recoveries were 103 and 105%, respectively.
Processing of mRNA Precursors-Specific polyadenylation was assayed by the extension of a 'lP-labeled RNA, LSpre, with unlabeled ATP. L3pre contained the adenovirus-2 L3 polyadenylation signal, including the AAUAAA sequence and ending 1 nucleotide upstream of the normal cleavage and polyadenylation site (see "Experimental Procedures"). All The samples were extracted once with phenol/chloroform, ethanolprecipitated after the addition of 3.5 pg of glycogen as carrier, and analyzed on a denaturing 12% polyacrylamide gel.
fractions of the Mono Q column of Fig. 1 that had detectable nonspecific poly(A) polymerase activity were, in combination with the auxiliary factor CPF (see the Introduction), also active in this specific polyadenylation assay (Fig. 6). Quantitation of the experiment by scintillation counting of gel slices revealed that there was no linear relationship between the amount of poly(A) polymerase and the number of precursor RNA molecules extended. However, similar amounts of polymerase from the various peaks extended similar quantities of precursor; i.e. the various forms of poly(A) polymerase had equivalent specific activities in this assay. Low concentrations of poly(A) polymerase resulted not only in a small fraction of the primer being extended but also in very short oligo(A) tails. Thus, as in the nonspecific poly(A) extension, the poly(A) polymerase appeared to behave distributively in the specific reaction as well. Polyadenylation of the L3 precursor by all forms of poly(A) polymerase was dependent both on a wild type AAUAAA sequence and the presence of CPF; polyadenylation was observed neither with an RNA that carried a point mutation in the AAUAAA sequence (AAGAAA) nor in the absence of the specificity factor (Fig. 7).

DISCUSSION
The purification of poly(A) polymerase reported in this paper was based on the nonspecific extension of a poly(A) primer. Nevertheless, the enzyme, when supplemented with the specificity factor CPF, also specifically elongated an RNA that contained the AAUAAA signal close to its 3' end. An RNA with a point mutation in AAUAAA was not elongated. Specific polyadenylation has recently been reconstituted with this poly(A) polymerase and CPF purified to near apparent homogeneity from calf t h y m~s .~ This confirms that two factors are sufficient for this reaction. A MF-dependent poly(A) polymerase purified from calf thymus (Winters and Edmonds, 1973) has also recently been shown to be active in the specific polyadenylation reaction (Bardwell et al., 1990). It has also been reported that an antiserum raised against a poly(A) polymerase preparation from rat hepatoma cells inhibits the 3'-processing reactions in HeLa cell nuclear extracts (Terns and Jacob, 1989). However, in these experiments it was not demonstrated that the inhibition could be overcome by the addition of excess enzyme. Moreover, the preparation of poly(A) polymerase (made according to the procedure of Rose and Jacob (1976)), of which a major band of 48 kDa was used as an antigen, may not have been homogeneous (see below).
In preliminary experiments, the calf thymus poly(A) polymerase also promoted the AAUAAA-dependent cleavage of pre-mRNA. A more thorough investigation of this aspect, however, requires a better characterization of the cleavage factors.
With respect to the nonspecific polyadenylation reaction, most characteristics of the poly(A) polymerase described here resemble those that have been reported for a number of similar enzymes (Edmonds, 1982(Edmonds, , 1990. However, many poly(A) polymerases investigated previously were found to start polyadenylation only after a pronounced lag phase (reviewed by Edmonds, 1982Edmonds, , 1990. This might be taken as additional evidence for the impurity of these enzymes (see below). However, the enzyme described here, which did not show any lag phase under standard assay conditions, also had a nonlinear behavior under two different conditions; in contrast to poly(A), all other primers are altered by the elongation reaction. This may lead to a change in the reaction rate, as was observed for poly(U). The rate of AMP incorporation also increased with time when Mn2+ was used at a level 2-fold above the ATP concentration. Since the Mn2+ optimum was rather sharp, this effect may have been due to the removal of excess Mn2+ by the pyrophosphate generated in the course of the reaction.
For a number of poly(A) polymerase preparations, specific activities near lo5 pmol of AMP/min/mg have been reported (Edmonds, 1982(Edmonds, , 1990. In contrast, the enzyme described here has a specific activity of lo7 pmol of AMP/min/mg. It seems unlikely that this discrepancy can be accounted for solely by differences in experimental conditions. Also, for a number of preparations, I have observed purification factors between 6,000 and 15,000. This again is significantly higher than reported by almost any other poly(A) polymerase. The one clear exception among the poly(A) polymerases described before is the enzyme purified from calf thymus by Tsiapalis et al. (1975). Specific activity and purification factor, as well as the molecular weight and a number of other properties of this enzyme, agree very well with the poly(A) polymerase described here. This suggests that most if not all of the other eukaryotic cellular poly(A) polymerases investigated so far may have been far from homogeneous. Claims of homogeneity for preparations purified a few hundredfold with yields of 1 S. Bienroth and W. Keller, unpublished data.
C P F -+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Specific polyadenylation reactions were carried out with either the wild type L3pre RNA or the mutant L3preA with or without CPF as indicated. Each of these reactions was done with poly(A) polymerase from Mono Q fractions 9, 19, 23, and 33 (see Fig. 1). These fractions were diluted by factors of 40, 200, 600, and 400, respectively, and 0.5 pl was used per reaction. The size of DNA markers (in nucleotides) is indicated on the left. PAP, poly(A) polymerase. mg of protein or more from 100 g of cells (Rose and Jacob, 1976;Nevins and Joklik, 1977) must be regarded with skepticism. The lack of purity of most earlier preparations may have been a reason for widely differing molecular weight estimates of poly(A) polymerase (Edmonds, 1982). A recent 20,000-fold purification of poly(A) polymerase from Saccharomyces cereuisiae to apparent homogeneity5 confirms that poly(A) polymerase is not an abundant enzyme. The poly(A) polymerase described here appears to be nearly homogeneous based on electrophoretic analysis (see "Results"). In addition, we have obtained amino acid sequence data of the amino J. Lingner, I. Radtke, E. Wahle, and W. Keller, submitted for publication.
terminus and several tryptic peptides. All these amino acid sequences were present in one clone isolated from a cDNA library, indicating that all were derived from the same protein: The relationship of the enzyme described in this paper to a poly(A) polymerase extracted from rat liver nuclear envelopes with detergents and purified by lectin affinity chromatography (Kurl et al., 1988) is not clear. No detergent was necessary for solubilization of the calf thymus enzyme, and a specific reaction with lectins was not observed.
The existence of multiple poly(A) polymerases has been suspected for a long time. One argument has been the purification of enzymes with quite different properties, mainly regarding the preferential use of M e or Mn2+ (reviewed by Edmonds, 1982). However, as has been pointed out before (Edmonds, 19821, the preference of a poly(A) polymerase fraction for a certain divalent cation may be related to a number of variables in the assay, including the kind of primer used and the contamination with nucleases or CPF and other factors that may influence primer utilization. Thus, a difference in ion requirements does not necessarily reflect a true difference of the enzymes. One well documented example of two poly(A) polymerase preparations from the same tissue differing in their properties are the Mg2c-dependent enzyme purified from calf thymus by Winters and Edmonds (1973) and the Mn2+-dependent enzyme purified by Tsiapalis e t al. (1975) from the same source. As pointed out above, the latter enzyme is most likely identical with the one described here that is active in the specific polyadenylation reaction. Since the preparation of Winters and Edmonds has recently also been found to be active in this specific assay (Bardwell et al., 1990), these two seemingly very different enzymes are most likely closely related if not identical. A second argument that has been used in favor of the existence of several poly(A) polymerases has been the observation of multiple chromatographic peaks during purification (reviewed by Edmonds, 1982). Multiple peaks of poly(A) polymerase were also observed in this preparation during Mono Q chromatography.
However, all peaks contained the same enzyme as demonstrated by similar molecular weights and indistinguishable E. Wahle, G. Martin, E. Schilz, and W. Keller, unpublished data. specific activities in the nonspecific as well as in the specific polyadenylation reactions. The difference in affinities for the anion exchange column suggests the presence of minor modifications of the same polypeptide. Their nature and possible physiological significance remain to be established. The small differences in molecular weight that were observed in SDSpolyacrylamide gels are not sufficient to explain the chromatographic heterogeneity since there was no clear correlation between molecular weight and chromatographic behavior. In HeLa nuclear extracts, multiple chromatographically distinct poly(A) polymerases have been observed, too.
All of them promoted AAUAAA-dependent cleavage and polyadenylation (Ryner et al., 1989b).7 The availability of milligram quantities of pure poly(A) polymerase should now permit a thorough examination of the enzyme's role in the metabolism of polyadenylate chains and facilitate the purification of the other components required for the generation of mRNA 3' ends.