The mRNA cap structure stimulates rate of poly(A) removal and amplifies processivity of degradation.

Poly(A)-specific ribonuclease (PARN) is an oligomeric, processive, and cap-interacting 3' exonuclease. We have studied how the m7G(5')ppp(5')G cap structure affects the activity of PARN. It is shown that the cap has four distinct effects: (i) It stimulates the rate of deadenylation if provided in cis; (ii) it inhibits deadenylation if provided at high concentration in trans; (iii) it stimulates deadenylation if provided at low concentration in trans; and (iv) it increases the processivity of PARN when provided in cis. It is shown that the catalytic and cap binding sites on PARN are separate. The important roles of the 7-methyl group and the inverted guanosine residue of the cap are demonstrated. An active deadenylation complex, consisting of the poly(A)-tailed RNA substrate and PARN, has been identified. Complex formation does not require a cap structure on the RNA substrate. The multiple effects of cap are all accounted for by a simple, kinetic model that takes the processivity of PARN into account.

The mRNA poly(A) tail plays several important roles in eukaryotic gene expression. In yeast, it acts in synergy with the mRNA cap structure to stimulate initiation of translation. This requires a specific interaction between the poly(A) binding protein (PABP) 1 and the eIF4G factor of the eIF4F complex, where eIF4E interacts with the cap (reviewed in Refs. [1][2][3]. Poly(A) removal is an important and critical step during mRNA decay in yeast (reviewed in Ref. 4). The major pathway of mRNA degradation in yeast, i.e. the deadenylation-dependent pathway, is initiated by degradation of the poly(A) tail (5). When less than 10 adenosine residues remain, the decapping enzyme Dcp1 and associated factors remove the mRNA cap structure (6 -8), which triggers degradation of the mRNA by the action of a 5Ј33Ј exonuclease (9) encoded by the xrn1 gene (10,11). Taken together, it is believed that the poly(A) tail through its interaction in cis with the cap structure controls gene expression both at the level of initiation of translation and during mRNA decay. The poly(A) tail plays most likely similar roles in mammalian cells (e.g. Refs. 12,13), although there is less experimental evidence than for yeast. The recent development of in vitro mRNA decay systems based on mammalian cell free extracts is a significant achievement (14 -18). Using these systems, stimulation of mRNA decay by AU-rich elements (17), cap-dependent stimulation of poly(A) degradation (19), and mRNA decapping (20) can now be studied at the molecular level.
Poly(A) tail synthesis and degradation have been extensively studied in eukaryotic systems. Although poly(A) tail synthesis is now comparatively well understood (reviewed in Refs. 21,22), the mechanism behind poly(A) removal has remained obscure. Some nuclease activities that degrade poly(A) have been found (23-27, reviewed in Refs. 28 -30), and among these the poly(A)-specific ribonuclease (PARN) has been particularly well studied. PARN was initially described as a poly(A)-specific 3Ј exonuclease in HeLa cell free extracts (23,31). Recently, PARN was purified to apparent homogeneity and cloned (32)(33)(34). It was shown that its enzymatic activity in calf thymus cell free extracts is either associated with a 74-kDa polypeptide (32,33) or with a 54-kDa fragment thereof (34). Also in Xenopus oocytes, two isoforms of PARN, 74 and 62 kDa in molecular sizes, have been identified (33). The two isoforms differ in nuclear-cytoplasmic distribution, the 74-kDa form being exclusively nuclear while the 62-kDa form is cytoplasmic (33). So far it is not known how the subcellular localization of PARN is regulated, although it appears likely that proteolytic cleavage of PARN could play a role. In support of this, a putative nuclear localization signal has been found in the C-terminal part of 74-kDa PARN, a region that is absent in the 54-kDa PARN fragment. 2 Active PARN is oligomeric, highly processive, and stimulated by the presence of an m 7 G(5Ј)ppp(5Ј)G cap structure at the 5Ј-end of the RNA substrate (19,34,35).
The interaction between PARN and the cap structure at the 5Ј-end of an mRNA affects mRNA decay and most likely also initiation of translation; therefore, it is fundamental for the control of gene expression in eukaryotes. In this report we have studied how the cap affects the activity of the 54-kDa PARN fragment, and four different effects have been identified. First, cap stimulates deadenylation when provided in cis. When provided in trans it inhibits deadenylation at high concentration but stimulates deadenylation at low concentration. Finally, cap amplifies the processivity of poly(A) degradation by PARN. How cap affects PARN activity is interpreted in a simple kinetic model that takes the processivity of the enzyme into account.

EXPERIMENTAL PROCEDURES
Preparation of RNA Substrate-RNA substrates L3(A 30 ) non capped or capped at the 5Ј-end with m 7 G(5Ј)ppp(5Ј)G or G(5Ј)ppp(5Ј)G, and L3(A 5 ) non capped or capped at the 5Ј-end with m 7 G(5Ј)ppp(5Ј)G, were synthesized by in vitro transcription using T3 RNA polymerase (Stratagene 600111) and plasmids pT3-L3 (A 30 ) or pT3-L3(A 5 ) (23) digested with NsiI, as DNA template. Plasmid pT3-L3(A 5 ) is identical to plasmid * This work was supported by the Swedish Strategic Research Foundation, the European Commission through its TMR program, the Swedish Natural Science Research Council, the Swedish Research Council for Engineering Sciences, and funds at Uppsala University. 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.: 46-18-471-4908; Fax: 46-18-471-4862; E-mail: anders.virtanen@icm.uu.se. 1 The abbreviations used are: PABP, poly(A) binding protein; PARN, poly(A)-specific ribonuclease; PCR, polymerase chain reaction; EMSA, electrophoretic mobility shift assay. pT3-L3(A 30 ) with the exception of the number of inserted A residues (i.e. 5 versus 30 residues). RNA substrate was labeled either in its 54nucleotide-long body or in its poly(A) tail by inclusion of radioactively labeled mononucleotides (UTP and ATP, respectively) during in vitro transcription as previously outlined (23,31). The specific radioactivities of the included radioactive mononucleotides were 40 Ci/mmol in the transcription mixture for body labeling or 5 Ci/mmol in the transcription mixture for poly(A) tail labeling. Transcribed RNA was purified according to Moore and Sharp (36).
One-dimensional TLC and Quantitation of Nuclease Activity-Deadenylation activity was quantified as follows: L3(A 30 ) RNA substrate labeled by inclusion of [␣-32 P]ATP during in vitro transcription was incubated in conditions for in vitro deadenylation as described above with the addition of 0.1 mg/ml bovine serum albumin. Reactions were analyzed by one-dimensional TLC using 0.75 M KH 2 PO 4 , pH 3.5 (H 3 PO 4 ), as the solvent. The resulting polyethyleneimine cellulose F plate (Merck, number 5579) was dried and scanned by a 400 S Phos-phorImager (Molecular Dynamics). The fraction of released [ 32 P]AMP was determined. Knowing the specific activity of [ 32 P]AMP in the RNA substrate, the amount of released AMP was calculated. One unit of deadenylation activity was defined as the release of 1 mol of AMP/min.
Electrophoresis Shift Mobility Assay-Deadenylation reactions were carried out as described above. After incubation, samples were placed on ice and incubated 15 min in the presence of 500 ng/l (final concentration) of heparin. Samples (6 l) were immediately loaded on native polyacrylamide gels (1ϫ TBE buffer, 3.5% acrylamide:bisacrylamide (79:1, v/v), agarose 0.5%) pre-run for 30 min at 100 V at 4°C. After electrophoreses, the gels were fixed in 10% acetic acid for 10 min, transferred to Whatman 3MM paper, and dried in a Bio-Rad gel dryer for 1 h. The gels were finally exposed, scanned, and quantified by a 400 S PhosphorImager (Molecular Dynamics). Elution of ES complexes from native polyacrylamide gels was done by cutting the region of the gel containing the complex and extracting the RNA by rotation over night at 4°C in a buffer containing 0.5 M NH 4 OAc, pH 7.5, 10 mM MgCl 2 , 0.1% SDS, 0.1 mM EDTA, and 10 mM Tris-HCl, pH 7.5. The eluted RNA was precipitated in ethanol, collected by centrifugation, and fractionated by denaturing polyacrylamide gel electrophoresis.
Expression of Recombinant p54 PARN-HeLa cells poly(A) mRNA was prepared using a QuickPrep Micro mRNA purification kit (Amersham Pharmacia Biotech). The 54-kDa fragment of PARN was amplified by reverse transcription-PCR using a First-Strand cDNA synthesis kit (Amersham-Pharmacia Biotech) followed by PCR with the following oligonucleotides: (5Ј) 5Ј-TCGCATATGGAGATAATCAGGAGCA-3Ј, containing the NdeI restriction site, and (3Ј) 5Ј-CGTGGATCCTCAGTTAC-CAAAGGCACTGAA-3Ј, containing the BamHI restriction site. The obtained PCR fragment was cloned into the pGEM-T vector (Promega). The inserted fragment, between the NdeI and BamHI sites, was purified and subcloned into the corresponding sites of the vector pET 33 (Novagen). The obtained plasmid was finally transformed into BL21(DE3) cells (Novagen) to express the recombinant protein. Colonies were grown over night at 37°C with agitation in TB medium containing kanamycin, 50 g/ml. The cultures were diluted (1:16) in the same medium and grown at 37°C to reach 1 OD, and then induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside (final concentration). After 1.5 h of induction, the cells were harvested and the expressed soluble protein was purified using a TALON metal affinity resin (CLONTECH) following the instructions from the manufacturer.
Kinetic Model for Processivity-We assume steady-state kinetics, meaning that the partitioning of the enzyme between its different states (in Fig. 7B, below) remains unchanged with time. Such conditions, commonly used in Michaelis-Menten kinetics carried out in vitro, can be obtained when the enzyme concentration is very small in relation to the initial substrate concentration S n , when back reactions from intermediates and final product are negligible and when a single enzyme molecule has processed several substrates so that initial phase relations are lost. Such conditions are fulfilled for the experiments in Fig. 3 and Table I below. The processivity model has been used to interpret K m and V max values obtained for the experiments described therein. For the experiments in Fig. 4, where the hydrolytic reaction goes to completion, a more elaborate kinetic analysis is necessary, where inflows to the enzyme over all intermediate steps as well as product rebinding must be taken into account. This situation defies analytic solutions and must be treated numerically. For the scheme in Fig. 7B the following steady-state relations can be derived for the rate of production of free adenosines A by putting up flow relations for all the steps in the scheme and taking conservation of mass into account, The processivity parameter q is defined as, q is the probability that the enzyme catalyzes removal of an adenosine, rather than dissociates from the poly(A) tail. Accordingly, q n is the probability that PARN removes all adenosines from a poly(A) tail with n bases without falling off. S i is a substrate with a poly(A) tail containing i adenosines. E 0 is the total enzyme amount. The derivation is for initial enzyme kinetics, where S n dominates over substrates with smaller poly(A) tails (S i for i smaller than n). For perfect processivity when q is very close to 1, we can put q ϭ 1 Ϫ x, where x Ͻ Ͻ 1 and Taylor expand q n to first order in x, It also follows by Taylor expansion that in this limit, so that k Ϫa /k c ϭ x. Using these expressions for q and q n in the above formulas for V max /K m and V max gives, in the high processivity limit, When n ϭ 1, the normal Michaelis-Menten expressions are recovered in all the above cases. It may also be noted that when q is so small that q n can be neglected, i.e. in the low processivity limit, the expressions for V max /K m and V max become instead, The factor q/(1 Ϫ q) in the first of these expressions is, in this limit, the average number of bases removed by PARN before dissociation, once it has bound its substrate.

RESULTS
The Cap Structure Stimulates Deadenylation When Provided in Cis-We recently showed that a purified 54-kDa fragment of PARN deadenylates capped RNA substrates more efficiently than uncapped ones (34). Here, the role of the cap structure is further clarified by experiments where the same PARN fragment is used to deadenylate m 7 G(5Ј)ppp(5Ј)G capped, mRNA Cap Effects on Poly(A)-specific Ribonuclease G(5Ј)ppp(5Ј)G capped or uncapped L3(A 30 ) RNAs. Fig. 1 indicates that the m 7 G(5Ј)ppp(5Ј)G capped RNA is the most efficient, the G(5Ј)ppp(5Ј)G capped RNA the second most efficient, and the uncapped RNA the least efficient substrate for PARN. The effects of the cap structures on PARN activity were quantified by determination of Michaelis-Menten parameters for all three substrates, and the results are given in Table I. Deadenylation was in this case monitored by detection of the release of 5Ј-AMP mononucleotides with one-dimensional TLC ("Experimental Procedures"). It is seen that the efficiency parameter V max /K m is increased 1.5-fold by G(5Ј)ppp(5Ј)G and 3-fold by m 7 G(5Ј)ppp(5Ј)G in relation to the uncapped substrate. The maximal enzyme rate, V max , goes up more than 10-fold by addition of G(5Ј)ppp(5Ј)G to uncapped RNA but increases only about 3-fold by addition of methylated cap. This interesting result means that, although methylation of cap increases V max /K m 2-fold, it reduces at the same time V max about 3-fold (Table I, see "Discussion").
Taken together these data show that the cap structure stimulates in vitro deadenylation when provided in cis, suggesting that capped and polyadenylated mRNAs are the preferred substrates for PARN in vivo.
The m 7 G(5Ј)ppp(5Ј)G and G(5Ј)ppp(5Ј)G Cap Analogues are Non-competitive Inhibitors of PARN-It is known that a m 7 G(5Ј)ppp(5Ј)G cap analogue provided in trans inhibits PARNdependent poly(A) degradation of m 7 G(5Ј)ppp(5Ј)G capped RNA substrates (19,34,35). Here, we studied if deadenylation of uncapped RNA substrates is also inhibited by m 7 G(5Ј)ppp(5Ј)G provided in trans. Fig. 2 shows that PARNdependent deadenylation of both m 7 G(5Ј)ppp(5Ј)G capped and uncapped RNA substrates was inhibited by a trans-acting m 7 G(5Ј)ppp(5Ј)G cap analogue present at 10 M concentration. The mechanism of inhibition was studied further by Michaelis-Menten analysis of the rate of release of 5Ј-AMP from m 7 G(5Ј)ppp(5Ј)G capped L3(A 30 ) RNA substrates in the presence of different concentrations of m 7 G(5Ј)ppp(5Ј)G (i.e. 0, 0.5, 3, and 10 M) or G(5Ј)ppp(5Ј)G (i.e. 0, 50, 200, and 500 M) cap analogues provided in trans (Fig. 3). Because the K m values did not change by the presence of the cap-analogues when the V max values were reduced, the inhibition of PARN by trans-acting cap structures must be non-competitive. This suggests that the binding site for the cap structure is separate from the active site of PARN. A K i value, defined as the concentration of cap analogue that reduces PARN activity to half, was calculated for each of the two analogues acting on deadenylation of each of the three substrates. Table II shows that methylated cap analogue is an inhibitor that is about 50-fold more efficient than its unmethylated counterpart and that the K i values do not depend on the 5Ј-end of the RNA substrates. Surprisingly, when a methylated cap analogue is provided in trans at low concentrations (Ͻ0.5 M) it stimulates, rather than inhibits, PARNcatalyzed deadenylation of uncapped substrates or substrates with unmethylated cap. A similar activation is also visible in Fig. 2. There is, at the same time, no such stimulation for substrates with a methylated cap structure.
The Cap Structure Contributes to the Processivity of PARN-We have previously reported that the 54-kDa fragment of PARN deadenylates m 7 G(5Ј)ppp(5Ј)G capped RNA substrates processively (34). However, the degradation pattern observed in Fig. 1 suggested that the cap structure influenced the processivity of PARN. Therefore, we asked if the cap structure is an indispensable processivity factor for PARN or whether it tunes processivity without being essential. An important prediction for a processive mode of degradation is that for enzymes working close to their V max values the earliest time point when fully deadenylated products appear is independent of the amount of RNA substrate added to the reaction. If, in contrast, the enzyme is distributive, the appearance of completely deadenylated substrates will be increasingly delayed as more sub-  b The K m value refers to the RNA substrate concentration. c The release of 5Ј-AMP was monitored. 0.07 ng of 54-kDa PARN fragment was used in a total reaction volume of 10 l. Tabulated K m and V max values (mean value Ϯ S.D.) were obtained from three or more independent experiments. Each experiment contained 13 or more substrate concentrations and estimates of K m and V max were obtained from Lineweaver-Burke plots.  Fig. 4 reveals that the m 7 G(5Ј)ppp(5Ј)G capped RNA substrate was deadenylated in a highly processive mode, as previously observed (34). The uncapped L3(A 30 ) RNA substrate was also deadenylated in an apparently processive mode, because the earliest time point when fully deadenylated products appear was not affected by the amount of RNA substrate added to the reaction. However, in this case it was not possible to detect both L3(A 30 ) RNA substrate and fully deadenylated RNA product at the same time, suggesting that the processivity of PARN is reduced in the uncapped case. This would mean that the average number of adenosine residues removed before the PARN⅐RNA substrate complex dissociates is smaller for uncapped than for capped RNA substrates. We therefore repeated the experiment, now using an RNA substrate contain-ing only 5 adenosine residues in its poly(A) tail. Fig. 4B shows that in this case both the capped and uncapped L3(A 5 ) substrates were deadenylated in a processive mode, because fully deadenylated products were obtained at the same time independently of the amount of the L3(A 5 ) RNA substrate used. Furthermore, it was possible to detect the L3(A 5 ) RNA substrate and fully deadenylated products simultaneously when the uncapped L3(A 5 ) RNA substrate was used. We conclude that the contribution of the cap structure is to drive PARN into a highly processive mode of deadenylation. Thus, the cap structure amplifies but is not necessary for processivity.

Recombinant 54-kDa PARN Fragment Forms a Cap-independent Complex with a Poly(A)-tailed RNA Substrate-A conse-
quence of the processive mode of deadenylation by PARN is that the exonuclease must stay associated with the RNA substrate during multiple rounds of catalysis. This prompted us to investigate if such a complex could be detected by a electrophoretic mobility shift assay (EMSA). A time course analysis of the deadenylation reaction was therefore performed using homogeneously purified recombinant 54-kDa PARN and m 7 G(5Ј)ppp(5Ј)G capped RNA as substrate. Reacted RNA was analyzed by EMSA or purified and subjected to denaturing gel electrophoresis. Fig. 5 shows that the deadenylation reaction was in a processive mode (Fig. 5A), because both substrate and  b The inhibition constant, K b was defined as the concentration of cap analogue that reduced PARN activity to half. The release of 5Ј-AMP was monitored.
c Activation refers to the highest concentration of cap analogue where activation of deadenylation could be observed. Increasing the concentration over the listed value inhibited the reaction. d NA, activation was not detected. Tabulated inhibitor constants (mean value Ϯ S.D.) and highest concentrations where activation was observed were obtained from at least three independent experiments, each using at least nine different concentrations of added cap analogue. mRNA Cap Effects on Poly(A)-specific Ribonuclease product coexisted during the incubation. The figure also shows that an RNA⅐protein complex was formed simultaneously (Fig.  5B) and that its formation did not require a cap structure at the 5Ј-end of the RNA substrate (Fig. 5C). Fig. 6 shows that the detected complex was indeed an active deadenylation complex, because it contained the expected RNA species. The presence of PARN in the retarded complex is supported by the observation that the complex could be immunoprecipitated by PARN-specific antibody (data not shown). The complex was specific for polyadenylated RNA substrate, because it was barely detected using non-polyadenylated RNA and it could be outcompeted by increasing amounts of poly(A) added in trans (data not shown). We therefore conclude that PARN forms a stable and active deadenylation complex with the RNA substrate during its reaction cycles both in the presence and absence of a cap structure at the 5Ј-end of the RNA substrate. DISCUSSION We have investigated how the m 7 G(5Ј)ppp(5Ј)G cap structure located at the 5Ј-end of mRNA affects the activity of PARN. Four different effects have been identified. The cap structure (i) stimulates deadenylation activity if provided in cis, (Fig. 1 Table II), (iii) stimulates deadenylation activity in trans if provided in low amounts ( Fig. 2 and Table II), and (iv) increases the processivity of PARN if provided in cis (Fig.  4). Fig. 7A summarizes our proposal, based on the various cap effects observed here, for how PARN operates. Fig. 7B shows a simple model for the processive behavior of PARN, which is used to interpret the kinetic data. We also consider previously published properties of PARN (19,34,35).
The PARN⅐RNA Substrate Complex-The two key features of the PARN⅐RNA substrate complex (Fig. 7A) are the oligomeric nature of PARN (34) and the simultaneous interaction between PARN and the mRNA 5Ј-and 3Ј-ends (19,34,35). We propose that the active site is separate from the cap binding site. This is based on the observation that the m 7 G(5Ј)ppp(5Ј)G cap analogue acts as a non-competitive inhibitor (Fig. 3), which is kinetic evidence for at least two separate sites. Furthermore, we suggest that three cap binding sites and three active sites are present within the oligomeric PARN. Neither site has been defined at the molecular level yet. However, Körner et al. (33) have suggested that PARN belongs to the RNase D family and conserved motifs characteristic of this class of exonucleases (37) are present within the PARN polypeptide. Thus, multiple and putatively active sites are present in an oligomeric form of PARN.
FIG. 6. Identification of an active deadenylation complex. 10 fmol of m 7 G(5Ј)ppp(5Ј)G capped 32 P-labeled L3(A 30 ) RNA substrate was incubated with 8 ng of the recombinant 54-kDa fragment of PARN during 0 s, 15 s, 1 min, 5 min, or 20 min, as indicated above the lanes, under standard conditions for in vitro deadenylation. A, reacted RNA was recovered and fractionated by electrophoresis using a 10% polyacrylamide:bisacrylamide (19:1, v/v)/7 M urea gel. The resulting fluorogram is shown. Arrows to the right marked with S, P(A), and P denote the locations of RNA substrate, partially deadenylated product, and deadenylated product, respectively. The appearance of the partially deadenylated products, P(A), is due to the high concentration of recombinant PARN (6 nM) added to these reactions. These products are not observed when smaller amounts of PARN are used (e.g. Fig. 5). B, reactions in A were analyzed by EMSA as described under "Experimental Procedures." Arrows to the right marked with O, C, and SP denote the locations of origin of electrophoresis, PARN⅐RNA complex, and unreacted RNA substrate/deadenylated product, respectively. C, RNA present in regions denoted C and SP of the EMSA analysis shown in B was purified and subjected to denaturing gel electrophoresis as outlined under "Experimental Procedures." The resulting fluorogram is shown. In lanes marked C and SP, RNA purified from the corresponding regions of the gel shown in B was fractionated. Arrows to the right marked with S, P(A), and P denote the locations of RNA substrate, partially deadenylated product, and deadenylated product, respectively.

mRNA Cap Effects on Poly(A)-specific Ribonuclease
The opposing roles of the cap structure, i.e. stimulating PARN when acting in cis ( Fig. 1 and Table I) Table II) can be explained by the presence of multiple cap binding and active sites. Thus, we propose that binding of one cap structure induces a conformational change of PARN, which activates the enzyme. We speculate further that this conformational change (i) inactivates the subunit that binds the cap structure and (ii) enhances the activity of the remaining subunits. According to this model, additional binding of a second and third cap structure would inactivate the active sites of the second and third subunits, thereby inhibiting the exoribonuclease activity completely. We emphasize that the m 7 G(5Ј)ppp(5Ј)G cap analogue, even if it is an efficient inhibitor of the reaction when provided in trans, has the interesting property of stimulating deadenylation of uncapped or G(5Ј)ppp(5Ј)G capped RNA substrates when provided in trans provided its concentration is sufficiently low ( Fig. 2 and Table II). This stimulatory effect is in keeping with the proposed activation/inactivation model based on multiple cap binding and active sites. We therefore suggest that the in trans-acting methylated cap analogue, provided at low concentration, activates deadenylation of uncapped or G(5Ј)ppp(5Ј)G capped RNA substrates by mimicking the role of the single cap structure provided in cis when m 7 G(5Ј)ppp(5Ј)G capped RNA substrate is used. In this activation step the 7-methyl group of the cap structure must play a crucial role (see also below). In vivo such an activation/inactivation model would ensure that the oligomeric PARN is only engaged in deadenylating one single polyadenylated mRNA molecule at the time and not multiple mRNA molecules. The 10 3 -to 10 4fold difference between the K i /activation values (micromolar range) and the K m values (nanomolar range) is, we suggest, because the K m value reflects the dissociation constant (K D ) of the PARN⅐RNA substrate complex, whereas the K i /activation values is the K D value for the PARN⅐cap interaction. The K D for the PARN⅐RNA substrate complex is in the 10 nM range. 3 If this assumption is correct it follows that PARN first interacts with the poly(A) tail and then intramolecularly with the 5Ј-endlocated cap structure of the RNA substrate.
A Kinetic Model for Processivity-An important effect of the methylated cap structure on the PARN activity is to drive the deadenylation reaction from a modestly to a highly processive mode (see Figs. 1 and 4). High processivity requires that PARN and its RNA substrate remain as a stable enzyme⅐substrate (ES) complex during multiple rounds of catalysis, and Figs. 5 and 6 provide direct evidence for such a complex. Formation of this complex does not require the cap structure (Fig. 5), in keeping with the observation that the cap structure amplifies but is not essential for processivity (Fig. 4). The data in Table  I together with the model in Fig. 7B immediately suggest how cap and the methylation of cap may tune the kinetics of PARN to amplify its processivity. According to this model (see "Experimental Procedures") the K m values for the different substrates for PARN only depend on the association (k a ) and dissociation rate constants (k Ϫa ) for substrate binding, i.e. K m ϭ k Ϫa /k a . The K m values for uncapped substrates and substrates with methylated cap are the same, and the K m value is about 7-fold higher for substrates with unmethylated cap (Table I). We propose, therefore, that k Ϫa increases 7-fold by the unmethylated cap and that k Ϫa decreases back to its initial, uncapped, value when the cap becomes methylated. For low and intermediate processivity, V max is determined by the catalytic constant k c ("Experimental Procedures"), which for unmethylated cap is about 12-fold higher than in the uncapped case (Table I). This suggests that the processivity parameter q ϭ 1/(1 ϩ k Ϫa /k c ), describing the probability that the enzyme cleaves off an adenosine rather than dissociates from the substrate, is closer to one for the unmethylated cap than without cap. The biologically relevant cap structure is methylated (38), and therefore it appears paradoxical that V max is more than 3-fold smaller for the methylated than the unmethylated cap (Table I). Superficially, this would mean that the efficiency of PARN is lower for its natural substrate than for substrate analogues. However, if it is assumed that methylation of cap considerably enhances the processivity of PARN by stabilizing its binding to both substrate and product, our model shows that for short poly(A) tails the V max parameter may become limited by the rate of product release (nk Ϫa ), rather than by the catalytic rate constant (k c ) ("Experimental Procedures"). One cannot take for granted that k Ϫa is exactly the same when PARN is bound to poly(A) tails as when it is bound to product, as in our simplified scheme. However, previous experiments (34) demonstrate that the K m values for PARN cleavage of different homopolymers are very similar, although the V max values differ radically. This indicates that a general enhancement of the binding of PARN to RNA by methylation of cap may, indeed, lead to rate inhibition by slow product release. It should be emphasized that we use substrates with n ϭ 30 adenosines in the tail, whereas the poly(A) tails in vivo normally contain about 200 bases (39). Because our model predicts that inhibition by product release would be virtually eliminated when the poly(A) tails increase in size from 30 to 200, the suggested effect of cap would give cells the advantage of higher PARN processivity accompanied with only marginal inhibition due to slow dissociation of products.
To sum up, our data and theoretical analysis suggest that methylation of cap enhances processivity of PARN by stabilizing its binding to RNA in a sequence-independent way. When the poly(A) tails are much shorter than in vivo, but not otherwise, this stabilization may lead to reduction in the maximal rate, V max , by which PARN removes adenosines.
PARN Compared with Other Processive Enzymes-Processivity is important for many polymerases and nucleases. For DNA polymerases, processivity is achieved by the action of  FIG. 7. A, a schematic drawing of the PARN⅐RNA complex. Oligomeric PARN consisting of three identical subunits interacts with the 3Ј-end-located poly(A) tail and the 5Ј-end-located cap structure (white square) of the RNA substrate. Two subunits interact with the poly(A) tail, whereas the third subunit interacts with the cap structure. The cap binding (black oval) and the active (white oval) sites are separate from each other. See "Discussion" for further details. B, a kinetic model for processivity. A reaction scheme for deadenylation is shown. See "Experimental Procedures" and "Discussion" for further details.
mRNA Cap Effects on Poly(A)-specific Ribonuclease external factors of which two classes are known: the sliding clamps and the globular accessory factors (reviewed in Refs. 40,41). The sliding clamp encircles the DNA and stays topologically linked to it, thereby keeping the DNA polymerase attached to the DNA template (42)(43)(44). For T7 DNA polymerase, processivity is achieved by formation of a stable complex between the polymerase and the Escherichia coli thioredoxin protein, a globular accessory factor (45,46). The -exonuclease, which degrades one strand of a double-stranded DNA molecule, is an example of a highly processive 5Ј33Ј exonuclease. In this case processivity is explained by its toroidal structure (47). There seems to be a fundamental difference between the mode of action of PARN on one hand and DNA polymerases and -exonuclease on the other. In contrast to these enzymes, PARN completely removes the substrate to which it is bound, although they always can remain attached to a DNA strand that is complementary to the one that is synthesized or degraded. The absence of an auxiliary, and yet unknown, processivity factor for PARN is strongly supported in that (i) a homogeneously purified preparation of PARN (34) was used in this study and (ii) recombinant PARN is processive (Fig. 5).
The processive mode of the -exonuclease was recently described with a model where the enzyme rotates and in this way brings a new site into action for each cleavage step (47). It is tempting to speculate that the molecular background to the processive action of PARN is to be found in its oligomeric form and that it rotates in much the same way as the -exonuclease.
A new catalytic site of the oligomer may in this way be brought into action at each step of cleavage and with the cap moving in synchrony from subunit to subunit.
Coupling Poly(A) Removal with mRNA Decay and Translation-For some classes of mRNAs removal of the cap, during mRNA decay, does not occur until the poly(A) tail is degraded (5, 12, 48 -50). It is therefore likely that the cap binding property of PARN prevents the action of the decapping enzyme during the deadenylation process. It has also been proposed (51) that dissociation of the poly(A) binding protein (PABP) from an mRNA, caused by a shortening of its poly(A) tail, makes the mRNA cap structure accessible to the decapping enzyme. This phenomenon could be due to the role of PABP in stabilizing the binding of eIF4E to the cap with eIF4G as mediator (reviewed in Refs. 1-3). The present and previously published results suggest a functional connection between mRNA decay pathways and initiation of translation (19,34,35). Thus, removal of the poly(A) tail by PARN may be a first step for mRNA degradation and, in addition, a control that translation does not begin on an mRNA targeted for degradation.
The effect of PABP on mRNA deadenylation has been controversial. Studies in yeast have shown that PABP stimulates mRNA deadenylation (24,27,52), whereas other studies in vertebrates show that mRNA degradation is inhibited by PABP (53,54). In the case of PARN it has been found that PABP, depending on conditions, can both stimulate and inhibit PARN activity in vitro (32). The reason for these discrepancies is unknown. However, it appears likely (32) that the RNA binding properties and relative concentrations of PABP and PARN will influence efficiency of deadenylation.
The K D values for the PARN⅐RNA 3 and PABP⅐poly(A) (55,56) complexes are both in the 10 nM range. The cap binding factor eIF4E binds the cap with an affinity in the 0.3-0.7 M range (57), close to the predicted cap binding affinity of PARN, and is increased at least 10-fold by the addition of eIF4G (58,59). The eIF4E⅐eIF4G complex is further stabilized by PABP interacting with the poly(A) tail (reviewed in Refs. 1-3). Taken together, these affinity data suggest that the binding of PABP and/or PARN to the mRNA poly(A) tail is a key event for regulation of both initiation of translation and mRNA decay in vivo. Consequently, we suggest that tuning the processivity of PARN may be an important way to regulate this control step.