INTRODUCTION

Gene expression depends on the concentration of mRNA, which in turn is determined by the ratio of transcription and turnover. Each mRNA has its own specific half-life, which can sometimes be altered in response to internal or external signals. Rapid degradation is particularly important for mRNAs encoding proteins that are only transiently required, like those of the early response to serum stimulation (1, 2). Whereas it is obvious that mRNA instability is essential for a quick shut-off of protein synthesis after transient transcription, a closer look reveals that a short mRNA half-life also favors rapid induction (1).

In yeast, a pathway of decay has been described that is common to many mRNAs, both stable and unstable (3-5). Degradation is initiated by shortening of the poly(A) tail. Only after the poly(A) tail has been shortened to 15 nucleotides or less is the 5-cap removed. Cap removal exposes the mRNA to a 5-3-exonuclease. Certain sequences within the RNA determine the rate of poly(A) shortening (6, 7). Mutations in these sequences lead to altered rates of deadenylation and often to corresponding changes in mRNA half-life (5). Thus, deadenylation is not only the first step in mRNA decay, but can be rate-limiting (5, 8).

In yeast, poly(A) tail removal requires the cytoplasmic poly(A)-binding protein I (PAB I).¹ In strains lacking PAB I, the steady state population of mRNA has longer poly(A) tails compared with wild type cells (9). The PGK1 and MFA2 mRNAs are deadenylated only after a lag phase, and 3- and 6-fold more slowly, respectively, in these mutants (10). Extracts prepared from pab1 strains do not degrade poly(A) (11). Addition of recombinant PAB I restores deadenylation. A Mg²⁺-dependent poly(A)-specific 3-exonuclease (poly(A) nuclease (PAN)) requiring PAB I for its activity has been purified. The enzyme contains two polypeptides of 76 and 127 kDa (11-14). At low salt concentration, PAN does not attack naked poly(A). The requirement for a RNA-binding cofactor can be satisfied not only by PAB I but also by spermidine (12). At high salt concentration, PAN has a PAB I- and spermidine-independent activity. It is not yet clear whether PAN is the enzyme responsible for deadenylation in vivo (14). Enzymes that catalyze decapping and 5-exonucleolytic decay have also been purified, and genes encoding them have been cloned (5, 8).

In higher eucaryotes, the mechanism of mRNA decay is less clear. However, for many mRNAs it has been shown that poly(A) removal precedes the decay of the mRNA body (1, 2, 15). As in yeast, mutations in the 3-untranslated region or the coding region can change the rate of deadenylation and the mRNA half-life (8). While deadenylation is usually the prelude to degradation, this is not necessarily the case. In animal oocytes, deadenylation is used to stop the translation of certain mRNAs. The deadenylated RNAs are stored inactive but intact and may even be readenylated and used again at a later time (16, 17).

The enzymology of mRNA decay, and in particular of deadenylation, in higher eucaryotes is unknown. It is also uncertain which role PAB I plays in deadenylation. Depletion of PAB I from a cell extract promoted rapid deadenylation and decay of mRNA (18). Thus, it was suggested that, in contrast to the situation in yeast, PAB I in mammalian cells inhibits mRNA decay by protecting the poly(A) tail. However, since the experiments were carried out in vitro, it is not certain whether the nuclease degrading the poly(A) tail was the same that is responsible for this process in vivo. Overexpression of PAB I in Xenopus oocytes protected mRNA from deadenylation, again suggesting that PAB I may inhibit poly(A) removal rather than promote it (19). However, since deadenylation is used for the purpose of translational control in oocytes and is not followed by complete degradation of the RNA, it is not clear whether the deadenylation reaction in oocytes is the same as in somatic cells. So far, nucleases responsible for poly(A) tail removal have not been identified unequivocally in higher eucaryotes. A poly(A)-specific 3-exoribonuclease has been partially purified from HeLa cell nuclear extract (20, 21). A mammalian, polysome-associated 3-exoribonuclease of 33 kDa has been purified to apparent homogeneity. This nuclease is able to degrade poly(A), but it is not poly(A)-specific. The enzyme is inhibited by yeast PAB I (22).

Several authors have reported a periodic distribution of poly(A) tail lengths in vivo with prominent species differing in length by 25-30 nucleotides. Such a distribution was described in the steady state population of globin mRNA (23-25) and, less pronounced, also in mRNAs for epidermal growth factor and glyceraldehyde-3-phosphate dehydrogenase (26, 27). Similar molecules were identified as deadenylation intermediates of the gro- cytokine mRNA (27). A regular periodic distribution of intermediates of poly(A) tail shortening can also be seen in an analysis of c-fos mRNA (15), although the particular type of experiment carried out precluded an assignment of poly(A) tail lengths. PAB I covers about 27 nucleotides when bound to poly(A) (28). Therefore, the periodicity most probably reflects the binding properties of PAB I.

We report here that, in a cytoplasmic extract of HeLa cells, the poly(A) tail is shortened from the 3-end. The population of deadenylation intermediates has a length periodicity of about 30 nucleotides. We have purified a poly(A)-specific 3-exoribonuclease from calf thymus. Deadenylation of mRNA fragments by the purified nuclease in the presence of PAB I results in the same periodic pattern of intermediates as seen in cytoplasmic extracts and in vivo.

EXPERIMENTAL PROCEDURES

Poly(A), poly(U), poly(dA) (Boehringer Mannheim), and poly(C) (Sigma) were 5-labeled with [-³²P]ATP and T4 polynucleotide kinase (29). 3-Labeling of poly(A) was carried out as follows; 8.6 pmol of size-fractionated poly(A) (about 350 nucleotides long) was incubated with 50 µCi of [-³²P]ATP (3000 Ci/mmol) and 1.9 µg of yeast poly(A) polymerase for 30-40 min under reaction conditions for unspecific polyadenylation (30). For the preparation of homogeneously labeled poly(A), 156 pmol of oligo(A) primers, about 30 nucleotides in length, were incubated with 25 nmol of ATP, 50-100 µCi of [-³²P]ATP, and 1.9 µg of yeast poly(A) polymerase for 30-40 min. 5-Labeled poly(A) with either a 3-OH group or a mixture of 2- and 3-phosphates was prepared as follows; 5-labeled poly(A) was partially hydrolyzed in 0.2 N NaOH for 5 and 10 min at 37 °C. The samples were neutralized by the addition of 1/3 volume of 1 M Hepes, pH 5.5, combined, and ethanol-precipitated. Cyclic phosphates were opened in 0.2 N HCl. The RNA was ethanol-precipitated, and one half (about 3-5 pmol) was dephosphorylated with the 3-phosphatase activity of T4 polynucleotide kinase (20 units) in the buffer supplied by the manufacturer. These substrates were loaded onto a polyacrylamide gel, and poly(A) smaller than the starting material was gel-purified.

Other RNAs were derived from the plasmids pSP6L3pre and pSP6glob. pSP6L3pre contains the L3 polyadenylation signal of the adenovirus 2 major late transcription unit, which is truncated one nucleotide upstream of its natural poly(A) addition site (31). The plasmid pSP6glob is a pGEM3 transcription-vector (Promega) carrying a 187-base pair insert from the 3-end of the human -globin gene. The insert starts at the EcoRI site in the coding region and stops at the polyadenylation signal (AAUAAA) in the 3-untranslated region (32). This insert was derived from the plasmid pbE (kindly provided by N. J. Proudfoot, University of Oxford) by polymerase chain reaction amplification as follows. Primers were constructed converting the EcoRI site of the third -globin exon into a BamHI site and introducing an EcoRI site at the polyadenylation signal (AAUAAA). This fragment was ligated into the BamHI and EcoRI sites of pGEM3.

Plasmid pSP6L3pre was linearized with RsaI and pSP6glob with EcoRI. Capped, uniformly -³²P-labeled RNAs were obtained by in vitro transcription of the linearized template DNAs with SP6 RNA polymerase in the presence of 7-methyl-G(5)ppp(5)G and 50 µCi of [-³²P]ATP as described (33), except that the total ATP concentration was 0.05 or 0.1 mM and the UTP concentration 0.5 mM. Polyadenylated transcripts were prepared by the addition of an unlabeled poly(A) tail with yeast poly(A) polymerase. Enzyme (1.1 µg) was used for 1-2.5 pmol of RNA at 30 °C for 30-60 min under conditions of nonspecific polyadenylation (30).

With the exception of homogeneously labeled poly(A) used for the trichloroacetic acid precipitation assay, all RNA substrates were gel-purified as follows. RNA was separated in a polyacrylamide-urea gel. A gel piece containing the desired RNA was cut out and incubated in 500 µl of elution buffer (500 mM ammonium acetate, 10 mM magnesium acetate, 0.1 mM EDTA, 0.1% (w/v) SDS, 1% (v/v) phenol) at 37 °C overnight. The eluted RNAs were extracted once with phenol/chloroform (1:1) and ethanol-precipitated.

PAB I was purified from calf thymus as described (34). A monoclonal antibody (10E10) directed against human PAB I was a gift from M. Görlach and G. Dreyfuss (University of Pennsylvania, Philadelphia). Purified recombinant yeast poly(A) polymerase containing six histidines at the N terminus was a gift from P. Preker (Biozentrum, University of Basel), and HeLa cell cytoplasmic extract was kindly provided by U. Rüegsegger (Biozentrum, University of Basel). Cytoplasmic extract was prepared from HeLa cell homogenates made according to Wahle and Keller (35). After pelleting of the nuclei, the supernatant was mixed with 10% (v/v) glycerol, 0.02% (v/v) Nonidet-P40, and 0.5 mM dithiothreitol and used as cytoplasmic extract. SP6 RNA polymerase and calf intestinal phosphatase were from Boehringer Mannheim. T4 polynucleotide kinase and acetylated bovine serum albumin were from New England Biolabs. Terminal nucleotidyltransferase was from MBI Fermentas, S1 nuclease from U. S. Biochemical Corp., pepstatin from Fluka, and phenylmethylsulfonyl fluoride from Merck. Ammonium sulfate (ultraPURE^TM) was from Life Technologies, Inc. DEAE-Sepharose FF and FPLC columns were from Pharmacia Biotech Inc. Blue Sepharose was the same as described previously (36). Heparin-Sepharose was prepared according to Ref. 37, except that Sepharose CL 6B (Pharmacia) was used. Leupeptin hemisulfate, heparin (H-7005), and polynucleotide phosphorylase (P-1384) were from Sigma.

Poly(A) degrading activity was routinely assayed by the release of trichloroacetic acid-soluble products from homogeneously labeled poly(A) (modified from Ref. 12). In assays used to monitor the purification, up to 5 µl of protein sample was mixed at room temperature with 50 µl of reaction buffer 1 (5 mM Hepes, pH 7.5, 2 mM MgCl₂, 0.5 mM dithiothreitol, 10% (v/v) glycerol, 0.2 mg/ml bovine serum albumin, 0.02% (v/v) Nonidet P-40, 0.3 mM spermidine), containing 5 ng of homogeneously labeled poly(A) and 1 µg of unlabeled poly(A) of heterogeneous size. The reaction was started by the addition of the protein sample and immediate incubation at 37 °C for 15 min and quenched by the addition of 150 µl of cold 13.3% (w/v) trichloroacetic acid. The samples were centrifuged at 14,000 rpm for 15 min in a microcentrifuge. 100 µl of the supernatant was added to a mixture of 100 µl of 1 M Tris base and 2 ml of scintillation mixture (Quick Safe A, Zinsser Analytik, Frankfurt, Germany). The amount of radioactive nucleotides released was determined by scintillation counting. One unit was defined as the activity that releases 1 nmol of nucleotides/min. To ensure that the measurements were in the linear range of the assay, the nuclease was diluted such that no more than 10% of the substrate was degraded. After improvement of the reaction conditions for the purified nuclease, the following components of reaction buffer 1 were changed: 20 mM Hepes, pH 6.8, 1 mM MgAc, 2 mM spermidine (reaction buffer 2).

Deadenylation reactions with polyadenylated mRNA fragments were assembled at room temperature as follows; 25 µl of reaction buffer 2 lacking spermidine was mixed with 2-10 fmol of polyadenylated capped L3pre or globin RNA and 4 units RNA guard (Pharmacia). Salt concentrations were as indicated. PAB I or spermidine was added to final concentrations of 0.3-11.5 nM or 2 mM, respectively, as indicated, and the reaction mix was prewarmed at 37 °C. The reaction was started by the addition of nuclease or HeLa cell cytoplasmic extract. At different times reactions were stopped with 50 µl of 2 × proteinase K buffer (0.2 M Tris HCl, pH 7.9, 0.3 M NaCl, 25 mM EDTA, 2% (w/v) SDS), and 25 µl of water containing 20 µg of proteinase K, 5 µg of glycogen, and 2 µg of rRNA. After incubation for at least 30 min at 37 °C, the RNA was precipitated with ethanol, dissolved in formamide loading buffer, heated for 2 min at 95 °C, and analyzed on sequencing gels (29). The gels were either exposed to Kodak X-Omat AR film or analyzed with a phosphorimager (Fuji BAS 1000).

The decay of poly(A) and poly(U) was analyzed as follows; 5-230 fmol of radioactively labeled homopolymer was added to 50 µl of reaction buffer 2. The reaction was started by the addition of nuclease and incubation at 37 °C. Aliquots were mixed with an equal volume of formamide loading buffer, heated for 2 min at 95 °C, and applied directly to a sequencing gel. In an analysis of the decay direction, aliquots of 5 µl were mixed with 1 µl of 0.5 M EDTA. 2 µl were mixed with loading buffer and applied to a sequencing gel. The residual 4 µl were mixed with 200 µl of 10% trichloroacetic acid, and trichloroacetic acid-soluble radioactive nucleotides were measured.

In tests of substrate specificity, the decay of poly(A) and other homopolymers was analyzed as follows; 60 fmol of 5-labeled homopolymers were added to 50 µl of reaction buffer 2. Reactions were started by the addition of nuclease and incubation at 37 °C. They were stopped with 50 µl of 2 × proteinase K buffer containing proteinase K, glycogen, and rRNA and analyzed on gels as described above.

Routine protein determinations were done by the method of Bradford (38). Purified protein was determined by SDS-polyacrylamide gel electrophoresis (39) with bovine serum albumin as a standard. Products of poly(A) decay reactions were analyzed by polyethyleneimine thin-layer chromatography as described (22).

The purification was carried out at or close to 4 °C. Between purification steps, the samples were frozen in liquid nitrogen and stored at 80 °C. The following buffer was used: 50 mM Tris-HCl, pH 7.9, 1 mM EDTA, 10% (v/v) glycerol, 1 mM dithiothreitol, 0.4 µg/ml leupeptin hemisulfate, 0.7 µg/ml pepstatin, 0.5 mM phenylmethylsulfonyl fluoride (added fresh before use), and 0.02% (v/v) Nonidet P-40. KCl concentrations indicated for individual steps always refer to KCl in this buffer. Extract was prepared from 1.2 kg of calf thymus in 50 mM KCl as described (40). The crude extract was applied to a 4-liter DEAE-Sepharose FF column, which was washed with 50 mM KCl and eluted with a gradient (2.5 column volumes) from 50 to 600 mM KCl at 3 liters/h. Active fractions, eluted from 75 to 200 mM KCl, were pooled, adjusted to 30% saturation with solid ammonium sulfate, and stirred for 1.5 h on ice. After centrifugation (30 min, 10,800 × g_max), the supernatant was adjusted to 50% saturation with solid ammonium sulfate and again stirred on ice and centrifuged. The 50% ammonium sulfate pellet was resuspended in 400 ml of 50 mM KCl, dialyzed for 10 h against 2 × 4.5 liters of 50 mM KCl and clarified by centrifugation. A 1.4-liter blue Sepharose column (7 × 36 cm) was equilibrated with 50 mM KCl, and the dialyzed 50% ammonium sulfate precipitate was applied. The column was washed with 1.5 bed volumes of 250 mM KCl and eluted with 1 bed volume of 1 M KCl at 2 liters/h. At this point, active fractions from two preparations, from 1.2 kg of tissue each, were combined and concentrated by precipitation with 60% ammonium sulfate. After centrifugation the pellet was resuspended in 200 ml of 50 mM KCl, dialyzed for 12 h against 3 × 4 liters of the same buffer, divided, and each half was applied separately to a 115-ml heparin-Sepharose column (2.5 × 37 cm), equilibrated with 50 mM KCl. The column was washed with 1.5 bed volumes of 50 mM KCl and developed with a gradient (10 bed volumes) from 50 to 500 mM KCl at 145 ml/h. DAN activity eluted in a broad peak from 80 to 150 mM KCl. Active fractions were pooled and dialyzed for 4 h against 30 mM KCl. The dialyzed pool was centrifuged, divided, and each half was loaded onto an 8-ml MonoQ FPLC column equilibrated with 50 mM KCl. The column was washed with two bed volumes 50 mM KCl and eluted with a 320-ml gradient from 50 to 500 mM KCl at 2.5 ml/min. DAN activity eluted in a sharp peak around 160 mM KCl. Active fractions from both columns were pooled (40 ml), dialyzed for 4 h against 2 liters of 30 mM KCl, centrifuged, and applied to a 1-ml MonoQ column, equilibrated with 50 mM KCl. DAN was step-eluted with 500 mM KCl at 0.9 ml/min. Four peak fractions were applied separately to a Superdex HR 10/30 FPLC column, equilibrated with 300 mM KCl, at 0.15 ml/min. Active fractions of three Superdex columns were pooled, adjusted to 25% saturation with solid ammonium sulfate and centrifuged. The supernatant was applied to a 1-ml phenyl-Superose FPLC column equilibrated with buffer containing 25% ammonium sulfate. The column was washed with 3 ml of buffer + 25% ammonium sulfate, 1 ml of buffer + 12.5% ammonium sulfate, and further developed with a gradient (15 bed volumes) from 12.5% to 0% ammonium sulfate, at 0.3 ml/min. DAN activity eluted around 8% ammonium sulfate (320 mM).

RESULTS

Upon incubation in a cytoplasmic extract of HeLa cells, polyadenylated L3pre RNA suffered a shortening of the poly(A) tail. Predominant decay intermediates differed in length by about 30 nucleotides, and the shortest oligoadenylated RNAs produced during the reaction carried a poly(A) tail of about 15 nucleotides (Fig. 1, panels A (lanes 7-11) and B). The amounts of fully deadenylated RNA that accumulated in these assays was variable but generally low. The same degradation pattern was observed at 10 or 100 mM KCl and with different concentrations of Mg²⁺. Upon fractionation of the extract by step elution from a DEAE-column (0.05, 0.17, 0.35, and 0.5 M KCl steps), the deadenylating activity was found in the 0.17 M KCl fraction. Gel electrophoresis showed that this fraction generated the same pattern of deadenylation intermediates as the complete extract. The 0.17 M KCl step contained 75% of the total poly(A) degrading activity recovered from the column as determined by a precipitation assay that measures the release of trichloroacetic acid-soluble material from homogeneously ³²P-labeled poly(A) (data not shown). A Western blot revealed that the 0.17 M step also contained most of the PAB I present in the extract (data not shown). Based on these results and those described below, we attribute the decay pattern to the cofractionation of a deadenylating nuclease and PAB I. 

Calf thymus extract was employed for the purification of the deadenylating activity. Fractions were monitored for poly(A) degrading activity with a trichloroacetic acid precipitation assay. To suppress unspecific nucleolytic activity in initial assays, a 400-fold excess of Escherichia coli rRNA was added to the homogeneously labeled poly(A) used as a substrate. Under these conditions, gradient elution of a DEAE-column revealed only a single activity peak eluted at around 0.12 M KCl (Fig. 2), likely corresponding to the activity identified in the 0.17 M step of the HeLa cell cytoplasmic extract. The poly(A) degrading activity could be slightly stimulated by spermidine (Fig. 2). A Western blot revealed that the PAB I peak was almost identical to the nuclease peak (data not shown).

During the subsequent purification the activity was monitored with only poly(A) as a substrate, no rRNA was added. The purification is summarized in Table I. The poor recovery on heparin-Sepharose is unexplained. Mixing of the nuclease peak fractions with other fractions from the same column gave no evidence for removal of a stimulatory factor. The poly(A) degrading activity was purified about 130,000-fold. Inasmuch as the total activity increased in the first purification step, the purification factor is probably overestimated. A profile of the final phenyl-Superose column is shown in Fig. 3A. A prominent band of about 74 kDa comigrated with the activity (Fig. 3B). The same band was also associated with the activity on a MonoQ column. The poly(A) degrading activity was named DAN, for deadenylating nuclease.

Table I. Purification of the deadenylating nuclease The purification was monitored by the trichloroacetic acid precipitation assay with buffer 1. Unit definition was as described under "Experimental Procedures." Step Protein Activity Specific activity Yield Purification mg units units/mg % -fold Extract 92,350 46,030 0.5 100 DEAE-Sepharose 24,900 130,600 5.2 284 10 Ammonium sulfate precipitation 11,250 133,500 12 290 26 Blue Sepharose 703 138,800 197 302 427 Heparin-Sepharose 104 23,240 224 50 486 MonoQ 12 13,450 1,121 29 2,432 Superdex gelfiltration 0.7 6,420 9,171 14 19,900 Phenyl-Superose 0.06 3,640 60,670 8 131,600

The highest DAN activity, determined with the trichloroacetic acid precipitation assay, was found at pH 6.8 with 2 mM spermidine and 1 mM magnesium acetate in the absence of additional salt. At pH 5.9 and 7.7, the activity decreased to 30 and 50%, respectively. The dependence on Mg²⁺ was absolute, and Mg²⁺ could not be replaced by Mn²⁺ or Ca²⁺. In the absence of KCl, DAN activity was almost completely dependent on spermidine with an optimum of 2 mM. At this concentration of spermidine, KCl or potassium acetate was inhibitory at all concentrations (Fig. 4). Spermidine-independent activity was observed in the presence of salt. The optimum was 120 mM with either KCl or potassium acetate (Fig. 4). Under these conditions, the activity was approximately 10-fold lower than at 2 mM spermidine in the absence of salt. At 120 mM potassium acetate, the activity could be increased 2-3-fold by 7 mM spermidine.

3- or 5-labeled poly(A) was incubated with DAN, and the decay products were analyzed on a denaturing 15% polyacrylamide gel (Fig. 5). With the 3-labeled substrate, mononucleotides were detected shortly after DAN addition, and no shortened decay intermediates were seen. In contrast, with 5-labeled poly(A) there was no radioactivity in the mononucleotide fraction even after 90 min, but shortened intermediates were present. Nevertheless, after prolonged incubation with higher concentrations of DAN, poly(A) was completely degraded to mononucleotides (data not shown). A quantitative analysis of the same assay by trichloroacetic acid precipitation showed that, with 3-labeled poly(A), the trichloroacetic acid-soluble nucleotides were released at 8.5 fmol/min, and the reaction was over within 5 min. With 5-labeled poly(A), radioactivity was released at 0.14 fmol/min, and the release continued throughout the time of the experiment (data not shown). Taken together, these results demonstrate that DAN is a 3-5-exonuclease.

3-5-exonuclease activity of DAN.

Mononucleotides released by DAN from homogeneously labeled poly(A) were analyzed by thin-layer chromatography (Fig. 6). 3- and 5-AMP were used as markers. The released mononucleotides comigrated with 5-AMP.

Release of 5-AMP.

As expected for a nuclease degrading poly(A) from the 3-end and releasing 5-AMP, DAN requires a free 3-hydroxyl group. A substrate with a phosphorylated 3-end was not degraded at all. After treatment of this substrate with T4 polynucleotide kinase, which has a 3-phosphatase activity, poly(A) was degraded completely (data not shown).

With a 400-fold molar excess of substrate over DAN, a large portion of the poly(A) was shortened in a synchronous manner (Fig. 7).² Thus, DAN dissociates after removal of one or a few nucleotides and then binds and degrades another RNA molecule. A distributive mode of action of DAN was also suggested by experiments in which a constant amount of poly(A) was incubated with varying concentrations of DAN. The length of degradation intermediates varied with the amount of DAN (Fig. 8A and data not shown).

A fraction of the substrate was not degraded at all. This fraction was eluted from the gel and incubated with yeast poly(A) polymerase (which needs a free 3-OH group for elongation) and ATP under reaction conditions for unspecific polyadenylation. The poly(A) was not elongated, whereas control substrates did receive poly(A). Thus, the 3-end of the eluted poly(A) was blocked, and this was presumably the reason for its resistance to DAN.

The spermidine-dependent DAN activity degraded poly(A) and poly(U) but not poly(C) and poly(dA) (Fig. 8A). Controls showed that all of the substrates used for this analysis could be elongated either by poly(A) polymerase with ATP, in the case of polyribonucleotides, or by terminal nucleotidyltransferase and dNTPs, in the case of poly(dA). Thus, blocked 3-ends were not responsible for the resistance of poly(C) and poly(dA). Since poly(dA) was not degraded, DAN is a ribonuclease. With spermidine poly(A) or poly(U) were degraded equally well (Fig. 8, A and B).

With 120 mM potassium acetate, the substrate specificity was increased. Under these reaction conditions, DAN degraded only poly(A) but not poly(U) (Fig. 8B). A quantitative analysis showed that about 50% of poly(A) had been shortened after a 30-min incubation, whereas poly(U) was stable. In the same experiment, poly(U) was degraded in the presence of spermidine.

The substrate specificity of DAN was also investigated with polyadenylated L3pre RNA as a substrate (see "Experimental Procedures"). When the spermidine-dependent activity was measured, the poly(A) tail was almost completely destroyed within 2 min. The deadenylated body of the RNA accumulated and was degraded slowly during prolonged incubation (Fig. 9). Similar results were obtained with polyadenylated -globin RNA as a substrate, except that the deadenylated RNA remained stable even after prolonged incubation (data not shown). Thus, DAN is largely, but not entirely, specific for poly(A). DAN was also specific for the poly(A) tail when the spermidine-independent activity was monitored at 100 mM KCl (see below and Fig. 11, lanes 2-6).

Polyadenylated L3pre RNA was also used to explore the effects of PAB I on DAN activity. When, at 10 mM KCl, neither spermidine nor PAB I were added, poly(A) tails were completely stable during an extended incubation with DAN (Fig. 10). Addition of 11 nM PAB I, but not 0.3 nM, permitted deadenylation of the RNA (Fig. 10 and data not shown). 11 nM PAB I is a 8-fold molar excess over binding sites, but close to the K_d for the PAB I-oligo(A) interaction (41, 42). In the presence of PAB I, strongly populated species of decay intermediates differed in length by about 30 nucleotides, similar to those observed in cytoplasmic extract. The shortest reaction product still contained about 15 A residues at the 3-end. Little or no completely deadenylated RNA was detectable. The pattern of deadenylation intermediates in the reconstituted system is shown side by side with that in cytoplasmic extract in Fig. 1. The PAB I-dependent reaction was much weaker than the spermidine-dependent reaction. At a 3:1 molar ratio of DAN and RNA, deadenylation was complete within 2 min in the presence of spermidine (Fig. 9). In the presence of PAB I, deadenylation was not complete after 120 min, although an 18-fold excess of nuclease was used (Fig. 10). When the nuclear poly(A)-binding protein II (PAB II) (34) was used at the same concentration as PAB I, DAN was not stimulated.

When reactions were carried out at 100 mM KCl, spermidine-independent deadenylation occurred as expected. Again, the deadenylated RNA was relatively stable. Under these conditions, PAB I inhibited deadenylation in a concentration-dependent manner. At the concentration of PAB I used to detect PAB I-dependent activity under low salt conditions (11 nM), inhibition was complete (Fig. 11, lanes 7-11). At lower PAB I concentrations (0.4 nM), deadenylation was still possible and intermediates resembled those observed above. However, the shortest prominent deadenylation intermediate had about 25 A residues remaining, rather than 15, and completely deadenylated RNA accumulated.

As mentioned above, crude fractions of DAN were stimulated weakly by spermidine, whereas the purified enzyme was strongly spermidine-dependent. The spermidine dependence developed during the purification, mostly on the MonoQ column. MonoQ fractions outside the DAN peak were able to restore spermidine-independent activity when mixed with purified DAN. These fractions were much more efficient in the activation of DAN than PAB I at any concentration tested, and Western analyses showed that they contained only negligible quantities of this protein (data not shown). Thus, a different protein is probably responsible for this activity.

DISCUSSION

Ribonucleases cleave their substrates by one of two different mechanisms. Some enzymes activate the 2-OH of ribose, which cleaves the phosphodiester bond to form 5-OH and a cyclic 2-3-phosphodiester as an intermediate. These enzymes are generally metal-independent. Other enzymes use a divalent metal ion to activate water for an attack on the phosphodiester bond. The products of this reaction are two fragments terminating in a 3-OH and a 5-phosphate, respectively (43, 44). By three criteria (metal dependence, production of 5-phosphates, and requirement for a 3-OH at the chain end), DAN belongs to the second category.

Few 3-exoribonucleases have been purified from eucaryotic cells. A poly(A)-specific 3-exoribonuclease has been identified and partially purified from HeLa cell nuclear extract (20, 21). Similarities in catalytic properties and chromatographic behavior suggest that we may have purified this nuclease. A polysome-associated, Mg²⁺-dependent 3-exoribonuclease has been purified from human K562 erythroleukemia cells. This enzyme also releases 5-NTPs and requires a free 3-OH group (22). Since it has a mass of 33 kDa and is not poly(A)-specific, it is probably not related to DAN. In yeast, a 3-exoribonuclease activity is associated with the product of the essential RRP4 gene involved in 5.8 S rRNA processing (45). The nucleolytic activity requires a free 3-terminal hydroxyl group and releases 5-NTPs. Other eucaryotic 3-exoribonucleases described previously have been summarized by Astrom et al. and Mitchell et al. (21, 45). Their physiological roles remain unclear.

DAN appears to be most closely related to the yeast poly(A) nuclease, PAN (11, 12). Both DAN and PAN are Mg²⁺-dependent poly(A)-specific 3-exoribonucleases that require a free 3-OH group, release 5-AMP as sole product, and degrade poly(A) in a distributive fashion. The most unusual property, common to PAN and DAN, is the strong dependence on RNA-binding cofactors like spermidine. With PAN, the highest activity was observed in the presence of PAB I. A direct interaction between PAN and PAB I has been demonstrated (13). DAN was stimulated by PAB I only weakly. Spermidine and yet uncharacterized factors were much more efficient. Since the dependence of DAN on these various cofactors is most pronounced at very low salt concentration, the physiological relevance remains to be established.

Interest in poly(A)-specific 3-exoribonucleases stems from the observation that the first and rate-limiting step for the decay of many mRNAs in yeast as well as in higher eucaryotes is poly(A) shortening (8). The role of PAB I in this process is controversial. As outlined above, genetic data obtained in yeast suggest a PAB I requirement for deadenylation. Although the dependence of the yeast enzyme PAN on PAB I is in agreement with the genetic data, the properties of PAN2 and PAN3 mutants have so far not provided evidence that the enzyme is indeed involved in deadenylation of mRNA in vivo (13, 14). In vertebrates, evidence has been offered that PAB I may be an inhibitor of deadenylation (18, 19). The fact that DAN can be either inhibited or stimulated by PAB I, depending on the reaction conditions, suggests that the discrepancy between the apparent PAB I dependence of deadenylation in yeast and a possible PAB I inhibition of the same reaction in vertebrates may not be as fundamental as one might think.

Whether PAB I is inhibitory or not, it is very likely that the substrate for the deadenylation reaction is not naked poly(A) but, least in somatic cells, a complex of poly(A) and PAB I. This is suggested by the fact that PAB I is present in sufficient quantities to coat all poly(A) in the cytoplasm (41). Furthermore, the pattern of deadenylation intermediates observed in vivo (see Introduction) is most easily explained by the binding properties of PAB I (28).

We have reproduced a deadenylation reaction in cytoplasmic extract that proceeds through the same pattern of intermediates as observed in vivo. The reaction reconstituted with purified DAN and PAB I showed the same pattern. When conditions were chosen under which deadenylation was PAB I-dependent, the product of the reaction still had about 15 A residues attached. This is close to the smallest poly(A) tract able to bind PAB I (46). Little or no completely deadenylated RNA was produced. A straightforward explanation of this digestion pattern would be that PAB I has a two-fold effect. Binding is required for susceptibility to DAN, but the binding site itself is protected against digestion. At higher salt concentration, deadenylation was inhibited by PAB I. Under conditions of partial inhibition, a similar pattern of intermediates was generated. However, the shortest oligo(A) tail had a length of about 25 nucleotides, corresponding to the fragment protected by PAB I (28), and completely deadenylated RNA was the final product. Under these conditions, K⁺ suffices to make the RNA susceptible to DAN, so the digestion pattern appears to be due to simple protection of poly(A) by PAB I. That the protective effect is stronger under these conditions may reflect salt-dependent differences in affinity or binding mode of either PAB I or DAN. One may argue that the pattern of intermediates is just a PAB I footprint and would be observed with any 3-exonuclease. However, in control reactions with the processive 3-exonuclease polynucleotide phosphorylase, no such pattern was found.

DAN was the predominant poly(A) degrading activity found in whole cell extracts. The properties of the enzyme suggest that it plays some role in poly(A) tail metabolism in vivo. Further experiments will show whether this role is in the deadenylation reaction that leads to mRNA decay.