Regulation of Escherichia coli Polynucleotide Phosphorylase by ATP*

Polynucleotide phosphorylase (PNPase), an enzyme conserved in bacteria and eukaryotic organelles, processively catalyzes the phosphorolysis of RNA, releasing nucleotide diphosphates, and the reverse polymerization reaction. In Escherichia coli, both reactions are implicated in RNA decay, as addition of either poly(A) or heteropolymeric tails targets RNA to degradation. PNPase may also be associated with the RNA degradosome, a heteromultimeric protein machine that can degrade highly structured RNA. Here, we report that ATP binds to PNPase and allosterically inhibits both its phosphorolytic and polymerization activities. Our data suggest that PNPase-dependent RNA tailing and degradation occur mainly at low ATP concentrations, whereas other enzymes may play a more significant role at high energy charge. These findings connect RNA turnover with the energy charge of the cell and highlight unforeseen metabolic roles of PNPase.

The monomeric subunit exhibits a five-domain structure that is widely conserved from bacteria to plants and mammals (12,13). The structural core of the subunit appears to be a duplication of an RNase PH domain, with the two RNase PH-like domains connected by a poorly conserved linker domain. In the doughnut-shaped homotrimeric protein, the subunits form a central channel where catalysis is thought to occur; the KH and S1 RNA-binding domains are located on top of the enzyme, with the former immediately above the central channel and the latter facing outward away from the channel (14).
PNPase was originally implicated in the synthesis of cellular RNA before the template-dependent RNA polymerase was discovered (2,15,16); later on, because of its phosphorolytic activity, it was implicated in RNA degradation (6). It has long been assumed that, because of the high P i intracellular concentration, PNPase would act in vivo mainly phosphorolytically (17). More recently, however, it was shown that PNPase can add heteropolymeric tails to RNA 3Ј-ends (18,19). Because in Escherichia coli PNPase-dependent heteropolymeric failing, like polyadenylation by polyadenyl polymerase (PAP), targets bacterial RNAs to degradation (20), both PNPase phosphorolytic and polymerization activities participate in PNPase-dependent RNA decay. Finally, it was suggested that PNPase plays a central role in the biosynthetic pathway of dCTP by providing the CDP precursor (21), thus linking RNA turnover to DNA replication.
Although widely conserved in Bacteria and Eukarya, the pnp gene does not seem to be essential for survival. In some bacteria such as E. coli and Yersinia enterocolitica, however, it is essential for bacterial growth in the cold (22)(23)(24). PNPase seems also to be directly or indirectly involved in the control of several processes such as cold shock response (25) in E. coli and virulence in Salmonella enterica and Yersinia spp. (26,27). In plant chloroplasts and mitochondria, PNPase directly controls RNA stability via both degradation and tail addition (19,28). In the last few years, human PNPase, which localizes in the mitochondrial inner membrane space and possibly in the cytoplasm, has been implicated in maintaining mitochondrial homeostasis and cell differentiation and in cellular senescence (29).
In E. coli, PNPase has been localized both in the cytoplasm and associated with the cell membrane and ribosomes (30,31). In addition, PNPase may exist as a single homotrimeric enzyme, associated with the RNA helicase RhlB (32), and in a multiprotein machine, the RNA degradosome, together with the endoribonuclease RNase E, which also serves as a scaffold for the assembly of the complex, RhlB, and enolase. In such heteromultimeric associations, PNPase can degrade otherwise refractory double-stranded RNA regions in an ATP-dependent manner (32)(33)(34)(35)(36), a property that has been used as a degradosome functional assay (37).
In experiments aimed at assessing the functionality of RNA degradosomes from pnp mutants isolated in our laboratory (38), we realized that ATP concentrations higher than 5 mM retarded the appearance of the cleavage products and, in control reactions with purified PNPase, inhibited RNA degradation. Here, we report that ATP binds to PNPase and inhibits both its phosphorolytic and polymerization activities. Such a direct control by ATP highlights an unforeseen metabolic role of PNPase that connects RNA turnover and the energy charge of the cell.

EXPERIMENTAL PROCEDURES
PNPase Kinetic Assays-PNPase was purified as described (39). The phosphorolytic activity of purified PNPase using poly(A) as a substrate was measured in a coupled pyruvate kinase/lactate dehydrogenase assay as described (40) at the indicated MgCl 2 concentrations. The assay, which measures the ADP produced as NADH oxidation, was performed in a 1-ml reaction mixture at 28°C, and absorbance was continuously recorded at 340 nm. One unit of enzyme activity is defined as the amount of enzyme that releases 1 mol of ADP/min under the working conditions.
PNPase polymerization activity was detected by measuring the P i released during the polymerization reaction as described (41) using the EnzChek phosphate assay kit (Invitrogen). Assays were performed at 25°C in a 1-ml volume containing 50 mM Tris-HCl (pH 7.5), 0.2 mM ADP, 20 g/ml poly(A), and MgCl 2 at the indicated concentrations.
Degradation and 3Ј-Tailing of PNPL1 RNA-For degradation and polymerization assays with a specific RNA substrate, 32 P-labeled PNPL1 (pnp leader) RNA was synthesized by in vitro transcription with T7 RNA polymerase and [␣-32 P]CTP using a DNA template obtained by PCR amplification of plasmid pAZ101 (11) with oligonucleotides FG676 (5Ј-ctaatacgactca-ctatagggATGAATGATCTTCC-GTTGC, where lowercase letters indicate the T7 promoter, and uppercase letters indicate E. coli) and FG1387 (5Ј-AATGTAATATCCTT-TCTCTTTCTTAG). PNPL1 RNA encompasses the first 158 nucleotides of the non-processed pnp transcript from the pnp-p promoter.
Phosphorolysis was performed at 26°C in 30-l reaction buffer (10 mM Tris-HCl (pH 7.4), 0.75 mM dithiothreitol, 4.5 mM magnesium acetate, and 10 mM KCl) containing 0.2 M PNPL1 (35,000 cpm), 8.6 nM PNPase, and 10 mM P i . Polymerization was performed under the same conditions with 1 mM ADP and omitting P i . Reactions were stopped by adding 3-l samples to 5 l of stop solution (95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.02% xylene cyanol) and heating for 5 min at 95°C. The samples were analyzed by denaturing 6% acrylamide gel electrophoresis, and the autoradiographic image was acquired by phosphorimaging the exsiccated gel.
Photoaffinity Labeling-Photocross-linking assay was performed by UV irradiating (2 J cm Ϫ2 , 254 nm) 2 pmol of purified proteins preincubated for 10 min at 21°C with 3.3 pmol of either [␣-32 P]ATP or [␥-32 P]ATP in 10 l of 50 mM Tris-HCl FIGURE 1. PNPase inhibition by ATP. The assays were performed with a uniformly radiolabeled PNPL1 RNA as described under "Experimental Procedures." A, secondary structure of PNPL1 computed using Mfold (48) showing a 99-nucleotide-long hairpin structure at the 5Ј-end and a looser conformation of the 3Ј-region. The two bases preceding the 5Ј-end of pnp-p RNA are lowercase. B, degradation (10 mM P i ) and polymerization (1 mM ADP) reactions with or without 10 mM ATP. Lanes a-e, reaction times of 0, 2.5, 5, 10, and 30 min, respectively. The arrowheads point to PNPL1 (P) and the ϳ100-base-long stem-loop degradation (SL) and ϳ1-kb-long elongation (E) products. C, polymerization reaction performed in the presence of 1 mM ADP and increasing P i concentrations and quenched after 30 min. Ϫ, no PNPase added. D, degradation (10 mM P i ) and polymerization (1 mM ADP) reactions at increasing ATP concentrations.
(pH 7.5) and 0.4 mM dithiothreitol in a UV Stratalinker 2400 apparatus (Stratagene). 10 mM MgCl 2 and 33 M ADP were also present where indicated. Before UV irradiation, the samples were transferred in a 96-well microtiter plate on ice. The UVirradiated samples were then run on 10% SDS-polyacrylamide gel, blotted onto a nitrocellulose filter, and visualized by phosphorimaging.

RESULTS
ATP Inhibits Both PNPase Enzyme Activities-While performing functional assays with RNA degradosomes assembled with mutant PNPases isolated in our laboratory (38), we observed that ATP concentrations higher than 5 mM retarded the appearance of degradation products (data not shown). To discriminate whether ATP directly inhibits the catalytic activity of PNPase rather than other properties of the entire degradosome, we tested the effect of ATP on purified PNPase by monitoring the shortening and elongation of a specific RNA (PNPL1) substrate in the presence of P i and ADP, respectively. PNPL1 is predicted to form a long stable stem-loop at the 5Ј-end, followed by a less structured 3Ј-tail (Fig. 1A). As shown in Fig. 1B, incubation of 32 P-labeled PNPL1 with PNPase in the presence of P i led to the accumulation of an ϳ100-nucleotidelong degradation product, whereas the same substrate in the presence of ADP was elongated up to ϳ1 kb after a 30-min incubation. Both phosphorolysis and polymerization were, however, completely inhibited by 10 mM ATP.
To rule out that P i and ADP, which are spontaneous ATP hydrolysis products, could be responsible for these inhibitory effects by driving in the opposite direction the reversible reactions, we performed the PNPase assays in 1 mM ADP and varying P i concentrations. As shown in Fig. 1C, P i up to 0.3 mM did not reverse the polymerization reaction with 1 mM ADP, whereas 1 mM ADP did not reverse degradation with 10 mM P i . Thus, contamination of P i and ADP up to 3 and 10%, respectively, of the ATP concentration used in the previous experiment would be not be influential on the PNPase reactions under our assay conditions. By varying the ATP concentration, we observed that, under our assay conditions, 4 mM ATP substantially inhibited both phosphorolytic and polymerizing activities (Fig. 1D).
PNPase Binds ATP-To test whether PNPase can directly bind ATP, we performed affinity labeling experiments by UV cross-linking radiolabeled ATP to PNPase. As shown in Fig. 2A, both [␣-32 P]ATP and [␥-32 P]ATP (0.33 M) were cross-linked to PNPase. This indicates that ATP, rather than its spontaneous hydrolysis product ADP, AMP, or P i , was cross-linked to PNPase. No cross-linking could be detected with carboxypeptidase used as a control, whereas a weak signal (barely visible in the original scan) in the absence of MgCl 2 was detectable with ribosomal protein S1. 10 mM MgCl 2 reduced but did not abolish cross-linking. Affinity labeling was not competed by 33 M ADP. Interestingly, however, cross-linking was essentially not competed by unlabeled ATP up to ϳ100 M (Fig. 2, B and C), thus suggesting that 0.33 M ATP was far from saturation, in keeping with the I 0.5 values determined for the nucleotide (see below).
ATP Allosterically Inhibits PNPase-To perform kinetic analysis of PNPase phosphorolytic activity at different ATP concentrations, we used poly(A) as a substrate and measured the appearance of ADP in a coupled pyruvate kinase/lactate dehydrogenase assay at constant saturating poly(A) (30 g/ml) and at varying P i concentrations. The results (Fig. 3A) were consistent with a mixed-type inhibition (42), with a scanty increase in K m and a substantial decrease in V max upon increasing ATP concentrations. Thus, the pattern observed is close to pure noncompetitive inhibition. In particular, V max dropped from ϳ0.7 to 0.2 units/mg when ATP was raised from 0 to 5 mM. Similar results were obtained at fixed P i and varying poly(A) concentrations. However, at poly(A) concentrations close to or lower than the K m (ϳ2 g/ml), measurements were subject to large dispersions, which prevented us from accurately determining the kinetic parameters (data not shown). In any case, the activity was completely abolished even at 30 g/ml poly(A) (data not shown), thus ruling out competitive inhibition. We therefore conclude that the ATP-binding site is distinct from that of both substrates.
According to the kinetic model, replots of slopes and 1/v axis intercepts as a function of inhibitor concentration should yield straight lines with intercepts on the abscissa that equal the inhibition constant (42). However, following this procedure, we surprisingly observed large and systematic deviations from linearity of the experimental points. We thus conclude that PNPase does not conform to a classical mixed-type or noncompetitive inhibition model. Moreover, when plotting the percentage of PNPase inhibition versus ATP concentration at saturating substrates, a sigmoidal curve was obtained (Fig. 3B), which clearly points to cooperative allosteric binding of the nucleotide. Based on the best fitting curve, a Hill coefficient of 2.3 and an I 0.5 (i.e. a concentration that gives 50% inhibition) of 3.3 mM were estimated. The experiments were performed at 5 and 13 mM MgCl 2 , and the results obtained under either condition were superimposable (Fig. 3B). This rules out the hypothesis that the inhibition observed depends simply on depletion of Mg 2ϩ (an ion required for PNPase activity) complexed with ATP. A similar profile was obtained with purified FLAG-Rne degradosome (43) rather than trimeric PNPase (data not shown). We also checked the effect of other nucleoside triphosphates on PNPase activity. Our results (Table 1) show that the purine nucleotides GTP, dGTP, dATP, and non-hydrolyzable ATP were also inhibitory, whereas the pyrimidine nucleotides CTP and UTP did not exert any appreciable effect.
To detect the effect of ATP in the synthetic direction, we measured the released P i using a commercially available enzymatic assay for phosphate detection that allows continuous measurements of PNPase activity. The data shown in Fig. 3C indicate that ATP also inhibits polymerization at both 5 and 13 mM with I 0.5 values of ϳ5 mM.

DISCUSSION
Here, we present clear-cut evidence, using both a model RNA and poly(A) as substrates, that E. coli PNPase binds to and is inhibited by ATP. Although this effect is likely to play a major regulatory role, no such observation was reported to date despite over 5 decades of investigations on this enzyme.
Our results show that the ATP-binding site is distinct from that of the substrates, as none of them could abolish or reduce the inhibitory effect even at saturating concentrations. In par-ticular, our kinetic analysis demonstrates a mixed-type inhibition of ATP toward P i . Furthermore, replots of slopes and 1/v axis intercepts as a function of inhibitor concentration did not yield straight lines as predicted by the classical mixed-type inhibition model (42). In contrast, the sigmoidal profile of PNPase inhibition as a function of ATP concentration indicates a cooperative allosteric binding of the nucleotide. In particular, we assessed a Hill coefficient of 2.3, well in agreement with the trimeric structure of the enzyme, and an I 0.5 of 3.3 mM. Similar results were obtained at a poly(A) concentration close to the K m , i.e. 2 g/ml, and also using the whole degradosome instead of PNPase alone (data not shown). This latter observation implies that ATP may also modulate PNPase function at the level of the multiprotein complex. The nucleotide also exerted a comparable inhibition on the synthetic activity, with a Hill coefficient of 2.3 as in the phosphorolytic direction and an I 0.5 of 5.0 mM. Our kinetic assays may not fully reproduce the physiological conditions, as ADP and P i are promptly removed by the auxiliary reactions in the phosphorolytic and synthetic tests, respectively. This should, however, not affect the mode of ATP inhibition because the ATP-binding site is distinct from that of the substrates.
Taken together, these results strongly support the idea that the inhibition exerted by ATP on PNPase is a phenomenon of physiological relevance. In particular, enzymes displaying a sigmoidal dependence of their activity on metabolite concentrations are believed to exert major roles in metabolic regulation. However, the interpretation of our observations is not straightforward. Classical studies conducted by Atkinson (44,45) several decades ago established that high values of energy charge, and thus high ATP and low ADP and AMP, inhibit the catabolic pathways in both eukaryotes and prokaryotes. Unfortunately, accurate assessment of the enzyme behavior under physiological conditions is not feasible, as the local intracellular concentration of P i and nucleotides can hardly be predicted. Nevertheless, our finding may help shed light on the physiological significance of different RNA degradation pathways operating in E. coli.
It has been pointed out that the phosphorolytic degradation of mRNA by PNPase, which operates close to equilibrium, is more energy saving than the hydrolysis performed by RNase II, which releases nucleoside monophosphates and is far from equilibrium (46). Moreover, 3Ј-end tailing of RNA may be performed both by PAP at the expense of ATP and by PNPase, which uses NDPs. Thus, PNPase-dependent degradation of RNA would be energetically more favorable than RNase II/PAPdependent degradation and would be activated at a low energy charge. Additional regulatory levels might occur in the RNA degradosome and in the PNPase-RhlB complex, where the

TABLE 1 Inhibitory effect of different nucleoside triphosphates on PNPase phosphorolytic activity
Assays were performed at 28°C as described under "Experimental Procedures" in the presence of the nucleotide to be tested (5 mM) and MgCl 2 (13 mM).

ATP dATP ␤,␥-Imido-ATP GTP dGTP CTP UTP
Residual activity (%) a 30 39 50 65 40 102 98 a The activity (average of at least three independent measurements) is expressed as a percentage of the control assays without nucleotide added.
RNA helicase unwinds RNA secondary structures and hydrolyzes ATP (35). This could lower the local ATP concentration with concurrent PNPase activation. Conversely, a low helicase activity may allow a local increase of ATP concentration, thus inhibiting RNA degradation. Predictions of the above speculations are that RNase II, the PNPase-RhlB complex, and the RNA degradosome may perform RNA degradation at high energy charge, whereas free PNPase activity would be more relevant at low ATP concentrations. Although information on RNA decay mechanisms at different energy charge is scanty, an interesting observation that supports our model has been obtained by Andrade et al. (47). These authors found that PNPase strongly contributes to the decay rate of the ompA transcript in stationary phase, whereas RNase II seems to minimally affect the half-life of this mRNA. A systematic study of the role of different phosphorolytic and hydrolytic exoribonucleases in the mRNA decay at different energy charge will aid in understanding the physiological role of PNPase inhibition by ATP.
The above interpretation does not rule out other possible scenarios that implicate PNPase in metabolic processes other than the mere RNA decay. For instance, the release of ADP and other NDPs at low energy charge with concomitant formation of high energy phosphoanhydride bonds might be a mechanism that exploits RNA degradation to reconstitute ATP via adenylate kinase. Finally, as PNPase may have a role in providing the dCTP precursor CDP (21), inhibition by ATP might play a role in coordinating DNA replication, RNA degradation, and cell energy charge.
Whatever their interpretation may be, our findings imply that a reconsideration of the precise physiological role(s) of this enzyme is required. The accurate determination of the intracellular concentrations of the chemical species affecting PNPase activity (NTPs, NDPs, and P i ) and the prevalence of alternative RNA degradation pathways under different physiological conditions such as metabolic stress and cold shock might provide clues to understanding the role of the ATP inhibitory effect reported here. Moreover, it will be interesting to assess whether this property is shared by other PNPases and whether it controls the pleiotropic effects associated with this enzyme in bacteria, plants, and humans.