Structure and mechanism of Mycobacterium smegmatis polynucleotide phosphorylase.

Polynucleotide phosphorylase (PNPase) catalyzes stepwise phosphorolysis of the 3'-terminal phosphodiesters of RNA chains to yield nucleoside diphosphate products. In the reverse reaction PNPase acts as a polymerase, using NDPs as substrates to add NMPs to the 3'-OH terminus of RNA chains while expelling inorganic phosphate. The apparent essentiality of PNPase for growth of M. tuberculosis militates for mycobacterial PNPase as a potential drug target. A cryo-EM structure of Mycobacterium smegmatis PNPase (MsmPNPase) reveals a characteristic ring-shaped homotrimer in which each protomer consists of two RNase PH-like domains and an intervening α-helical module on the inferior surface of the ring. The C-terminal KH and S1 domains, which impart RNA specificity to MsmPNPase, are on the opposite face of the core ring and are conformationally mobile. Single particle reconstructions of MsmPNPase in the act of poly(A) synthesis highlight a 3'-terminal (rA)4 oligonucleotide and two magnesium ions in the active site and an adenine nucleobase in the central tunnel. We identify amino acids that engage the 3' segment of the RNA chain (Phe68, Arg105, Arg112, Arg430, Arg431) and the two metal ions (Asp526, Asp532, Gln546, Asp548) and we infer those that bind inorganic phosphate (Thr470, Ser471, His435, Lys534). Alanine mutagenesis pinpointed RNA and phosphate contacts as essential (Arg105, Arg431, Lys534, Thr470+Ser471), important (Arg112, Arg430), or unimportant (Phe68) for PNPase activity. Severe phosphorolysis and polymerase defects accompanying alanine mutations of the enzymic metal ligands suggest a two-metal mechanism of catalysis by MsmPNPase. Published by Cold Spring Harbor Laboratory Press for the RNA Society.

INTRODUCTION

Polynucleotide phosphorylase (PNPase) is a versatile enzyme that plays synthetic and degradative roles in bacterial RNA metabolism (Mohanty and Kushner 2000). In its degradative mode, PNPase catalyzes processive phosphorolysis of the 3′-terminal phosphodiesters of RNA chains to yield nucleoside diphosphate (NDP) products. In its polymerase mode, PNPase uses NDPs as substrates to add NMPs to the 3′-OH terminus of RNA chains while expelling phosphate. The polymerase activity of PNPase was a crucial enabling reagent in the synthesis of mRNAs that led to the breaking of the genetic code (Lengyel 2012). The RNA polymerase reaction is the microscopic reversal of the RNA phosphorolysis reaction and both require a divalent cation, which can be either magnesium or manganese. We reported previously that Mycobacterium smegmatis (Msm) PNPase also acts on DNA polynucleotides as a DNA 3′-phosphorylase and a dADP-dependent DNA polymerase (Unciuleac and Shuman 2013a). The DNA modification reactions of PNPase depend on manganese and are not sustained effectively by magnesium.

Insights into the structural organization and catalytic mechanism of bacterial PNPase have emerged from crystal structures of exemplary enzymes (Symmons et al. 2000; Shi et al. 2008; Nurmohamed et al. 2009; Hardwick et al. 2012; Wang et al. 2017). PNPases are ring-shaped homotrimers. Each protomer is composed of five domain modules. Two RNase PH-like domains comprise the core of the trimeric ring. A metal-binding site is located within the distal PH domain (Nurmohamed et al. 2009). An α-helical module separating the PH domains is disposed on the inferior surface of the ring. The carboxy-terminal KH and S1 domains are on the opposite face of the ring and are conformationally mobile and thus usually not visualized in many of the available PNPase crystal structures. In the Caulobacter PNPase structure, the three KH modules of the trimer form a narrow aperture that contacts the RNA and through which the RNA chain is threaded into the central cavity of the ring (Hardwick et al. 2012).

M. smegmatis PNPase is a homotrimer of a 763-aa polypeptide (Unciuleac and Shuman 2013a). The Mn2+•DNA and Mg2+•RNA end modifying activities of MsmPNPase were coordinately ablated by mutating the metal ligand Asp526 to alanine, signifying that the catalytic magnesium and manganese ions bind to the same site on PNPase. Deletions of the carboxy-terminal S1 and KH domains of MsmPNPase exerted opposite effects on the RNA and DNA modifying activities. Subtracting the S1 domain diminished RNA phosphorylase and polymerase activity; simultaneous deletion of the S1 and KH domains further crippled the enzyme with respect to RNA substrates. In contrast, the S1 and KH domain deletions enhanced the DNA polymerase and phosphorylase activity of MsmPNPase (Unciuleac and Shuman 2013a). Two distinct modes of nucleic acid binding by MsmPNPase were observed: (i) metal-independent RNA-specific binding that required the S1 domain, and (ii) metal-dependent binding to RNA or DNA that was optimal when the S1 domain was deleted (Unciuleac and Shuman 2013a). The metal effect on RNA binding by the MsmPNPase core trimer is not dictated by the metal catalyst of phosphoryl transfer in the active site, insofar as alanine mutations of the predicted catalytic metal-binding residues Asp526 and Asp532 did not affect Mg/Mn-dependent RNA binding to the full-length and ΔS1 MsmPNPases (Unciuleac and Shuman 2013b). These findings added a new dimension to our understanding of PNPase specificity, whereby the carboxy-terminal S1 and KH modules serve a dual purpose: (i) to help capture an RNA polynucleotide substrate for processive 3′ end additions or resections, and (ii) to provide a specificity filter that selects against a DNA polynucleotide substrate.

Whereas MsmPNPase has been well characterized biochemically, the genetics and physiology of mycobacterial PNPase are under-developed. The Rv2783c gene encoding M. tuberculosis PNPase is located within a cluster of cooriented open reading frames in the order: rpsO (encoding ribosomal protein S15), lppU (a lipoprotein), Rv2783c (PNPase), pepR (a zinc protease), and Rv2781c (a putative oxidoreductase). The upstream lppU gene and the downstream pepR and Rv2781c genes are deemed inessential for M. tuberculosis growth, as defined by the recovery of viable bacteria with transposon insertions within these respective open reading frames (Sassetti et al. 2003). In contrast, no transposon insertion within the PNPase gene was recovered, raising the prospect that PNPase might be essential for growth of M. tuberculosis under laboratory conditions (Sassetti et al. 2003). This appears to be the case insofar as direct attempts to delete the M. tuberculosis gene encoding PNPase were unsuccessful and a CRISPR/Cas9 knockdown approach that reduced PNPase mRNA levels by 20-fold resulted in slowed cell growth (Płociński et al. 2019). Accordingly, mycobacterial PNPase merits consideration as a target for inhibitor discovery. This idea is given credence by the reports that Asp67Asn and Lys702Gln mutations in MtuPNPase confer resistance of clinical M. tuberculosis isolates to the anti-TB pro-drug pyrazinamide (PZA) and that pyrazinoic acid (POA, the active form of the antibiotic) inhibits MtuPNPase activity in vitro (Njire et al. 2017; He et al. 2019).

In order to gain further insight into the mycobacterial PNPase mechanism, we set out to determine the structure of the enzyme. We grew large hexagonal crystals of recombinant full-length MsmPNPase but found that they diffracted X-rays to no better than 10 Å resolution. Alternatively, as described in this report, we obtained a structure of the MsmPNPase via single particle cryo-electron microscopy with an RNA 3′ terminus and two magnesium ions visualized in the active site. Guided by the structure, we conducted a mutational analysis of MsmPNPase that illuminates enzymic functional groups essential for its phosphorylase and polymerase activities.

RESULTS AND DISCUSSION

Cryo-EM structure of mycobacterial PNPase in the act of poly(A) synthesis

Poly(A) synthesis was initiated during a 3 min reaction of purified MsmPNPase with ADP, magnesium, and oligo(rA)12 primer at 22°C. The mixture was then quenched on ice and immediately applied to a Superdex-200 analytical gel filtration column. Aliquots of the peak PNPase fractions were applied to EM grids and imaged with a Titan Krios G2 transmission electron microscope equipped with a Gatan K3 camera. Discrete MsmPNPase particles (Supplemental Fig. S1A) appeared as toroidal rings with apparent threefold symmetry after 2D and 3D classification (Supplemental Fig. S1B,C). A population of 840,000 particles was used for ab initio 3D reconstruction. The workflow for refinement of the MsmPNPase cryo-EM map to 3.07 Å resolution is described in Materials and Methods and summarized in Supplemental Figure S1D–J. Two noncontiguous masses of density were evident in the cryo-EM maps at all stages of refinement (Supplemental Fig. S1E–H). The larger component of the map comprises the core PNPase trimer ring, into which we built three continuous polypeptide chains, spanning MsmPNPase residues Asp8 to Ile594. Model refinement statistics are compiled in Supplemental Table S1. We were unable to model the carboxy-terminal S1 domain or the preceding KH domain into the weaker cloud of density adjacent to the PNPase core. We surmise that the KH-S1 portion of MsmPNPase is conformationally mobile, as found previously in crystal structures of other PNPases.

The refined model of the MsmPNPase fold is shown in Figure 1A, colored by protomer and oriented so that the inferior/bottom surface of the trimer is facing forward. Figure 1B shows a surface model in the same orientation, thereby highlighting the central pore of the PNPase ring. The structure is rotated in Figure 1C to feature the component domain modules of the core MsmPNPase protomer, wherein the PH domains are colored cyan and blue and the intervening α-helical module is colored beige. (For the purpose of orientation, the missing KH-S1 module is shown as a trio of gray ovals hovering above the carboxyl termini of the modeled PNPase core structure.) Figure 2A depicts the core trimer with the top surface of the trimer (i.e., from which RNA exits the central tunnel) facing forward.

FIGURE 1.

CryoEM structure of MsmPNPase. The MsmPNPase homotrimer is depicted as a cartoon model (A) and a surface model (B), colored by protomer and oriented so that the inferior/bottom surface of the trimer is facing forward. The r(A)4 RNA is shown in panel A as a stick model with a gold cartoon trace through the phosphate backbone. The structure is rotated in panel C to feature the component domain modules of the core PNPase protomer, wherein the PH domains are colored cyan and blue, and the intervening α-helical module is colored beige. The KH-S1 modules that could not be modeled into density are shown as a trio of gray ovals hovering above the carboxyl termini of the PNPase core structure.

FIGURE 2.

MsmPNPase in the act of poly(A) synthesis. The MsmPNPase homotrimer is depicted as a cartoon model, colored by protomer and oriented so that the top surface of the trimer is facing forward (A) or rotated to provide a side view (B). The r(A)4 RNA in the active site of each protomer is shown as a stick model with a gold cartoon trace through the phosphate backbone. The 5′ and 3′ ends of the RNA are labeled in panel A. The 5′ end is pointed toward the central channel; the 3′ end is directed toward the active site. Two magnesium ions in each active site are depicted as green spheres and labeled in panel B. The adenine nucleobase (Ad) in the central channel is rendered as a ball-and-stick model. (C) Close-up stereo view of the central channel looking down from the top of the PNPase trimer, highlighting how the adenine nucleobase (stick model with gray carbons) makes a π-stack on Phe89 (stick model, colored by protomer). The cryo-EM densities for adenine (blue mesh) and Phe89 (beige mesh) are shown (contoured at 2 σ, with 2 Å carve radius). Arg95 projecting into the central channel below Phe89 is also shown as a stick model.

A DALI search (Holm 2020) of the Protein Data Base affirmed a high degree of tertiary structure homology between mycobacterial PNPase and the PNPase enzymes from six different bacterial genera representing three phyla (Supplemental Table S2). As one might expect, the “top hit” was to the PNPase from the fellow Actinobacteria species Streptomyces antibioticus (Symmons et al. 2000) with respect to overall fold (Z score 54.8) and amino acid identity (73%). Whereas the fold similarity was only slightly less than PNPases from Proteobacteria (Coxiella, Caulobacter, Escherichia, and Acinetobacter; Z scores 50.7 to 47.7) and the Firmicutes taxon Staphylococcus aureus (Z score 45.1), the amino acid identities were clearly lower (43%–45%). MsmPNPase was more distantly related to human PNPase (Lin et al. 2012), a mitochondrial protein, with which it shares just 36% amino acid identity (Supplemental Table S2).

The cryo-EM maps for the MsmPNPase core trimer indicated that RNA and magnesium were bound to the enzyme. By virtue of the threefold symmetry of the PNPase core, the densities (and ensuing models) for the RNA moieties are observed in association with each protomer (Fig. 2A), notwithstanding that only one protomer can be engaged in catalysis at any time and only one RNA chain can traverse the central pore. With respect to the pore, we observed clear density for an adenine nucleobase engaged in a π-stack on the Phe89 aromatic side chain (Fig. 2C). The adenine is positioned near the center of the RNA channel (Fig. 2A; ball and stick model). A side view of the core trimer highlights that the adenine is situated high up in the channel near the exit site (Fig. 2B; ball and stick model). The crystal structure of E. coli PNPase contained electron density for a lone purine nucleoside at the same position in the channel and making a π-stack on the equivalent Phe77 side chain (Nurmohamed et al. 2009; pdb 3GCM).

RNA density for an (rA)4 tetranucleotide was apparent adjacent to the active site and in contact with several amino acid side chains of the two PH domains, for example, Phe68, Arg105, Arg112, and Arg430 (Fig. 3). The 5′-OH of the (rA)4 is pointing toward the center of the pore (Fig. 2A; stick model), whereas the 3′-OH of the (rA)4 is directed toward a pair of magnesium ions (Fig. 2B; green spheres) bound to the second PH domain of each protomer. The (rA)4 5′-OH is 19 Å away from the adenine near the tunnel exit, from which we construe that a poly(A) chain is present and threading through the tunnel but is either disordered, or obscured by the threefold symmetry, and hence not amenable to modeling. An ordered single nucleotide or nucleobase situated high up in the central channel, disconnected from a short oligonucleotide in the active site, has been noted previously in crystal structures of Sulfolobus and Pyrococcus exosome/PNPases (Lorentzen et al. 2007, pdb 2JEA; Navarro et al. 2008, pdb 2PO1).

FIGURE 3.

Active site RNA ligand modeled into cryo-EM density. Stereo view of the (rA)4 tetranucleotide (stick model with gray carbons and yellow phosphorus atoms) modeled into surrounding density (green mesh, contoured at 1.25 σ, with 2 Å carve radius). Adjacent amino acids Phe68, Arg105, Arg112, and Arg430 are depicted as stick models with beige carbons modeled into their density (blue mesh, contoured at 2 σ, with 2 Å carve radius).

RNA and magnesium contacts in the MsmPNPase active site

Figure 4 shows a detailed stereo view of the active site, with the RNA—5′-A4pA3pA2pA1OH—depicted as a stick model. Two magnesium ions, separated by 4.7 Å, are rendered as green spheres within their respective cryo-EM densities (green mesh, contoured at 2 σ). The RNA tetramer is partitioned into two base-stacked dinucleotide segments by virtue of intercalation of the Phe68 and Arg112 side chains between A3 and A2, such that the Phe68 phenyl ring makes a π-stack on the A2 nucleobase and Arg112 contacts the A2 ribose ring and the A3 ribose 2′-OH (Fig. 4). Arg105 make bidentate hydrogen bonds to the A3pA and A2pA1 phosphates. Arg431 packs against the pA1 nucleotide. Arg430 packs against the pA3p nucleotide. Arg424 from the neighboring PNPase protomer makes van der Waals contacts with the A4 nucleobase. MsmPNPase RNA-binding residues Phe68, Arg112, Arg105, Arg431, and Arg430 are strictly conserved in the six structurally characterized bacterial PNPases cited in Supplemental Table S2 and also in human PNPase. In contrast, Arg424 is not conserved in the Coxiella, E. coli, Staphylococcus, and human PNPases.

FIGURE 4.

Stereo view of the MsmPNPase active site. PNPase protomer A is depicted as a cartoon model colored according to secondary structure (cyan α helices; magenta β strands). The (rA)4 tetranucleotide is rendered as a stick model with gray carbons and yellow phosphorus atoms. Magnesium ions M1 and M2 are depicted as green spheres within their density (green mesh, contoured at 2 σ, with 2 Å carve radius). A phosphate anion imported from the superimposed structure of E. coli PNPase is included as a semitransparent stick model with a green phosphorus atom, situated to the left of the metal ions. Amino acid side chains of protomer A that interact with the active site ligands are shown as stick models with beige carbons. The Arg424 side chain of a neighboring protomer that is situated adjacent to the 5′-terminal adenosine of (rA)4 is shown as a stick model with blue carbons.

The presence of two magnesium ions in the PNPase active site (labeled M1 and M2 in Fig. 4) is a novel finding vis-à-vis previous crystallographic evidence for one manganese in E. coli PNPase (Nurmohamed et al. 2009; pdb 3GME). The M1 cation is coordinated between Asp526 and Asp532 and corresponds to the manganese binding mode (to Asp486 and Asp492) observed in E. coli PNPase. M1 is imputed to play a catalytic role, insofar as a D526A mutation of MsmPNPase abolished its phosphorylase activity (Unciuleac and Shuman 2013a). The M2 cation in the MsmPNPase structure is coordinated between Asp532, Gln546-Oε, and Asp548. The latter two contacts might be water-bridged, based on interatomic distances, though waters are not apparent in the density map. (Note that no densities at the M1 and M2 sites were seen in a 2.8 Å cryo-EM map of MsmPNPase that had not been preincubated with ADP and magnesium prior to preparation of the EM grids; not shown.) The M2-binding residues Gln546 and Asp548 are conserved in the structures of E. coli PNPase (as Gln506 and Asp508), Caulobacter PNPase (as Gln508 and Asp510; pdb 4AIM; Hardwick et al. 2012), and human PNPase (as Gln558 and Asp560; pdb 3U1K; Lin et al. 2012). The equivalent of the M2-binding residues in the Sulfolobus exosome PNPase are Gln204 and Asn206 (Rrp41 subunit; pdb 4BA2; Lorentzen and Conti 2012).

There was no interpretable density for ADP substrate or PO4 product in the MsmPNPase structure, which might reflect the heterogeneity of active site occupancy in a nonsynchronous population of PNPase enzymes engaged in poly(A) synthesis. (Alternatively, ADP or PO4 might have dissociated from MsmPNPase during the gel filtration step.) As a proxy for the PO4 product site (equivalent to the PO4 substrate site in the phosphorylase reaction), we superimposed the RNA-bound MsmPNPase structure on the crystal structure of Caulobacter PNPase in a binary complex with inorganic phosphate (pdb 3AIM) and then imported the phosphate anion (rendered as a semitransparent stick model with a green phosphorus atom) into the MsmPNPase active site depicted in Figure 4. From such a position, the phosphate could be coordinated by MsmPNPase side chains Thr470, Ser471, His435, and Lys534.

It was instructive to compare the structure of RNA-bound MsmPNPase to the crystal structure of a homologous Sulfolobus exosome PNPase enzyme in complex with a (rA)4 RNA and a phosphate anion that approximates a substrate/product state of the phosphorylase/polymerase reactions, albeit absent a metal cofactor (Lorentzen and Conti 2012). Figure 5A,B depicts superimposed and horizontally offset images of the archaeal and mycobacterial PNPase active sites and their enzymic contacts to substrates. The core ring of the archaeal exosome PNPase is composed of distinct subunits Rrp42 and Rrp41 that are homologous to the tandem PH domains of bacterial PNPase. The amino acids that engage the RNA 3′ end in the active site of MsmPNPase (Phe68, Arg105, and Arg112 from the proximal PH domain; Arg430 and Arg431 from the distal PH domain) are conserved in the Sulfolobus enzyme (as Tyr75, Arg112, and Arg119 from Rrp42; Arg98 and Arg99 from Rrp41). In both structures, the conserved Arg105/Arg112 side chain coordinates adjacent phosphates of the RNA ligand (Fig. 5). The salient point from the comparison is that the conformation and register of the oligo(A) RNAs differ in the two complexes. Using the 3′-terminal A1 nucleotide of MsmPNPase as a reference point, we see that the Sulfolobus RNA ligand has an extra 3′-terminal adenylate (labeled A−1 in Fig. 5A) that corresponds to the nucleotide that will undergo phosphorolysis by the adjacent phosphate anion. (The A−1 nucleotide also represents the product of the polymerization step in which the ribose O3′ of the A1 nucleotide attacks the α phosphate of ADP and expels PO4.) Whereas the equivalents of the A1, A2, and A3 nucleotides in the Sulfolobus enzyme form a continuous base-stacked array that is “capped” by a π-stack of Tyr75 on the A3 nucleobase (Fig. 5A), the A2 and A3 bases in MsmPNPase are splayed apart when Phe68 and Arg112 inserts between them and Phe68 stacks on A2 (Fig. 5B). It is conceivable that these structures exemplify distinct conformations along the PNPase reaction pathway in which changes in base-stacking triggered by the conserved aromatic and basic amino acids abet a ratcheting mechanism of 3′ RNA end translocation in sync with cycles of nucleotide addition or phosphorolysis.

FIGURE 5.

Comparison of RNA-bound mycobacterial and archaeal PNPase structures. The MsmPNPase•RNA complex (B) was superimposed on the structure of a Sulfolobus exosome(PNPase)•RNA•PO4 complex (A) and then offset laterally. The RNA ligands are shown as stick models with gray carbons and yellow phosphorus atoms. The adenylate nucleotides in the MsmPNPase complex are numbered from the 3′ end. The same numbering scheme is applied to the RNA in the Sulfolobus exosome structure, which has an additional 3′-terminal nucleotide (labeled A−1 in blue font). The phosphate anion in the archaeal exosome active site is shown as a stick model. Amino acids are depicted as stick models with cyan carbons. Amino acids from the Rrp42 subunit of the archaeal exosome and the proximal PH domain of MsmPNPase are labeled in blue font. Amino acids from the Rrp41 exosome subunit and the distal PH domain of MsmPNPase are labeled in black font. Contacts of the homologous Arg112/Arg105 side chains to adjacent phosphates of the RNA are indicated by dashed lines.

Three amino acids engage the PO4 anion in the active site of the archaeal exosome, two of which (Arg99 and Ser138) are conserved in MsmPNPase (as Arg431 and Thr470). The archaeal Arg139 equivalent in MsmPNPase is Ser471, which is poised to make a hydrogen bond to the phosphate. In MsmPNPase, the Lys534 and His435 side chains (which are not conserved in the Sulfolobus enzyme but are conserved in E. coli and Caulobacter PNPases) are likely to make additional contacts to the phosphate (Fig. 5B).

Structure-guided alanine scan of MsmPNPase

We constructed seven alanine mutants of MsmPNPase targeting active site amino acids imputed to coordinate the phosphate anion (T470A-S471A, K534A, R431A) and amino acids that engage the (rA)4 RNA (R105A, R112A, R430A, R431A, F68A). The recombinant His-tagged MsmPNPase-Ala mutants and wild-type MsmPNPase were produced in E. coli and purified by sequential Ni-affinity, DEAE-Sephacel anion exchange, and Superdex-200 gel filtration chromatography steps. SDS-PAGE analysis of the peak homotrimer fractions from the gel filtration step affirmed similar purity of the wild-type and mutant MsmPNPase polypeptides (Supplemental Fig. S2A).

As an initial test of the MsmPNPases for RNA phosphorylase activity, we reacted them for 15 min with a 5′ 32P-labeled 24-mer RNA (1 pmol) in the presence of 5 mM Mg2+ and 0.5 mM phosphate. The reaction products were analyzed by urea-PAGE and the labeled RNAs visualized by autoradiography. We saw that 1 pmol of wild-type MsmPNPase trimer sufficed to convert most of the input 24-mer RNA to a mixture of trinucleotide and dinucleotide end-products (Fig. 6A). The T470A-S471A, K534A, R431A, and R105A proteins were severely defective with respect to RNA end resection (Fig. 6A). These results signify that the contacts of these amino acids with the phosphate nucleophile and the 3′-terminal RNA segment are essential for the phosphorylase reaction. Whereas mutants R112A and R430A resected nearly all of the input 24-mer RNA, the size distribution of the products seen after 15 min was longer and more heterogeneous than that of wild-type PNPase (Fig. 6A). In contrast, the products formed in 15 min by the F68A mutant were similar to that of the wild-type enzyme.

FIGURE 6.

Effects of alanine mutations on RNA phosphorylase activity. (A) Reaction mixtures (10 µL) containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 0.5 mM (NH4)2PO4, 1 pmol (0.1 µM) of 5′ 32P-labeled 24-mer RNA (depicted at bottom with the 5′ label denoted by a dot), and 1 pmol (0.1 µM) wild-type or mutant recombinant PNPase homotrimer as specified above the lanes were incubated for 15 min at 37°C. The reaction products were analyzed by urea-PAGE and visualized by autoradiography. (B–D) Reaction mixtures (70 µL) containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 0.5 mM (NH4)2PO4, 0.1 µM 5′ 32P-labeled 24-mer RNA, and 0.1 µM wild-type or mutant PNPase homotrimer as specified above the gels were incubated at 37°C. Aliquots (10 µL) were removed at the times specified above the lanes and quenched immediately with formamide/EDTA. The reaction products were analyzed by urea-PAGE and visualized by autoradiography. The positions of the input 24-mer substrate and a 5′ 32P-labeled trinucleotide reaction product are indicated at the left of each gel.

We then examined the kinetic profiles of RNA resection by the “partly active” MsmPNPase mutants R430 and R112A in parallel with wild-type PNPase controls (Fig. 6B,C). Wild-type PNPase rapidly resected the 24-mer RNA to a predominant trinucleotide product within the first minute of the reaction and then slowly degraded the trinucleotide during a 15 min incubation (Fig. 6B). This pattern is indicative of a biphasic reaction in which initial resection is processive, after which the short product is prone to dissociate and only slowly reengages with the enzyme for further processing. A switch from processive to distributive action when the RNA chain is short accords with “classic” enzymology of bacterial PNPase, whereby Maxine Singer's laboratory showed that long RNAs are resected processively, whereas phosphorolysis of short RNA oligonucleotides is not processive (Klee and Singer 1968; Chou and Singer 1970a). Comparing a series of short oligo(A) substrates, they found that the Km for RNA increased by a factor of 39 in going from a pentanucleotide p(Ap)4A to a trinucleotide p(Ap)2A (Chou and Singer 1970b).

The initial phase of resection by the R430A mutant is clearly slowed and multiple cleavage intermediates of 6 to 23 nt in length are evident at the 1 min and 2 min timepoints that are slowly shortened to 5 nt and then 4 nt during the 15 min reaction, with only scant formation of a trinucleotide by 15 min (Fig. 6B). A similar kinetic pattern is seen for the R112A mutant (Fig. 6C). These altered kinetics suggest that the inflection point for product dissociation is shifted to a tetranucleotide to hexanucleotide range by the R430A and R112A mutations, in keeping with the structural data that Arg430 and Arg112 contact the third, fourth, and fifth nucleotides from the 3′-terminal nucleotide that undergoes phosphorolysis. The kinetic profile of the “active” MsmPNPase mutant F68A was interrogated during a 1 min reaction in parallel with the wild-type PNPase (Fig. 6D). Both enzymes rapidly resected the 24-mer to a trinucleotide after 5 sec, which was followed by gradual resection of the trinucleotide over 60 sec (Fig. 6D). Thus, the loss of the Phe68 aromatic side chain had little discernable impact on phosphorolysis under the conditions tested.

We estimated the magnitudes of the kinetic defects of the partly active R430A and R112A PNPase mutants by comparing the time required for the wild-type enzyme to resect the 24-mer to a trinucleotide (5 sec) to the times taken by the mutants to convert a significant fraction of the input 24-mer to a tetranucleotide, which were 5 min and 10 min for the R430A and R112A enzymes, respectively (Fig. 6). By this criterion, we surmise that Arg430 and Arg112 enhance the rate of phosphorolysis by ∼60-fold and ∼120-fold, respectively.

The seven alanine mutants were also assayed for RNA polymerase activity by incubation for 15 min with the same 5′ 32P-labeled 24-mer RNA (1 pmol) in the presence of 5 mM Mg2+ and 2 mM ADP. The reaction products were analyzed by urea-PAGE and the labeled RNAs visualized by autoradiography. The wild-type enzyme catalyzed extensive AMP transfer from ADP to the labeled primer to generate high molecular weight RNA chains that migrated near the top of the gel (Supplemental Fig. S2B). The mutational effects on polymerase activity generally echoed those seen for the phosphorylase reaction. To wit: (i) the R105A, K534A, R431A, and T470A-S471A proteins were inert or severely impaired in primer extension; (ii) the R112A and R430A proteins were less active than wild-type and formed shorter products during the 15 min polymerase reaction; and (ii) F68 was as active as wild-type PNPase (Supplemental Fig. S2B).

Alanine mutations of metal ligands

Alanine mutations D532A and Q546A-D548A were targeted to amino acids that either bridge the M1 and M2 magnesium ions (Asp532) or uniquely coordinate the M2 magnesium ion (Gln546 and Asp548). The purified recombinant mutant PNPase homotrimers (Supplemental Fig. S3A) were assayed for RNA phosphorylase activity in parallel with wild-type PNPase (Fig. 7). Loss of the Asp532 metal ligand drastically slowed the rate and extent of RNA resection, such that conversion of 24-mer substrate to tetranucleotide product by the D532A mutant was achieved after 15 min, at which time there was still a residual fraction of unresected 24-mer (Fig. 7). A similarly slowed kinetic profile was seen for the Q546A-D548A mutant (Fig. 7). The ∼180-fold rate decrement vis-à-vis wild-type elicited by loss of Gln546 and Asp548 contacts to the M2 metal ion suggests that the second metal plays an essential role in phosphorylase activity. The metal ligand mutations exerted a similar effect on the polymerase activity of MsmPNPase, whereby the fraction of the input radiolabeled 24-mer primer RNA that was extended and the rate of AMP addition to the extended primer by the D532A and Q546A-D548A enzymes during a 15 min reaction was markedly less than that of wild-type PNPase (Supplemental Fig. S3B).

FIGURE 7.

Effect of alanine mutations of enzymic metal ligands. Reaction mixtures (70 µL) containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 0.5 mM (NH4)2PO4, 0.1 µM 5′ 32P-labeled 24-mer RNA, and 0.1 µM wild-type or mutant PNPase homotrimer as specified above the gels were incubated at 37°C. Aliquots (10 µL) were removed at the times specified above the lanes and quenched immediately with formamide/EDTA. The reaction products were analyzed by urea-PAGE and visualized by autoradiography.

Conclusions and mechanistic implications

The cryo-EM structure of MsmPNPase in the act of poly(A) synthesis provides new insights into the organization of the active site, especially the amino acids that contact the 3′ segment of the RNA chain (Phe68, Arg105, Arg112, Arg430, Arg431) and those that are inferred to contact inorganic phosphate (Thr470, Ser471, His435, and Lys534). Structure-guided mutational analysis identified particular RNA and phosphate contacts as essential (Arg105, Arg431, Lys534, Thr470+Ser471), important (Arg112, Arg430), or unimportant (Phe68) for PNPase activity. The structure underscores shared and divergent features of the RNA and phosphate interfaces of bacterial and archaeal RNA phosphorylases. The detection of two magnesium ions in the MsmPNPase active site is a novel finding and mutational analyses of the amino acid metal ligands (here for Asp532, Gln546, and Asp548 and previously for Asp526) is consistent with a two-metal mechanism in which both metals are critical for PNPase activity. (Although this is the most parsimonious view of our results, we cannot definitely exclude, based on the structural data, a scenario in which a single metal can occupy either of two positions in the PNPase active site.) We speculate that the metals promote catalysis in two ways: (i) by stabilizing the transition state on the terminal phosphodiester during the phosphorolysis reaction and on the ADP α-phosphate during the polymerase reaction; and (ii) by interaction with the RNA 3′-O leaving group in the phosphorylase reaction and the 3′-O nucleophile in the polymerase reaction. Validation of this model will hinge on capturing structures of Michaelis complexes of metal- and RNA-bound PNPase with either ADP or PO4 occupying the substrate site. The latter ligands are not visualized in the cryo-EM structure reported here, presumably because of heterogeneity of occupancy under conditions permissive for polymerization. The challenge will be to freeze PNPase in a single state along the reaction pathway or as a minimally perturbed on-pathway mimetic.

MATERIALS AND METHODS

M. smegmatis PNPase

The pET-MsmPNPase plasmids encoding wild-type or mutant PNPase proteins (Unciuleac and Shuman 2013a) were transformed into Escherichia coli BL21(DE3). Cultures (1 L) of E. coli BL21(DE3)/pET16b-MsmPNPase were grown at 37°C in LB medium containing 0.1 mg/mL ampicillin until the A600 reached 0.6. The cultures were chilled on ice, adjusted to 0.5 mM isopropyl-β-d-thiogalactopyranoside and 2% (v/v) ethanol, and then incubated at 17°C for 20 h with constant shaking. Cells were harvested by centrifugation and the pellets were stored at −80°C. All subsequent procedures were performed at 4°C. Thawed bacteria were resuspended in 25 mL of lysis buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 10% sucrose) and then sonicated for 10 min (alternating 1 min on and 1 min off) to achieve lysis. Insoluble material was removed by centrifugation. The supernatants were incubated for 1 h under continuous rotation on a nutator with 10 mL of Ni2+-nitrilotriacetic acid-agarose resin (Qiagen) that had been equilibrated with lysis buffer. The resins were then recovered by centrifugation and washed sequentially with 50 mL of lysis buffer, 50 mL of 50 mM Tris-HCl, pH 8.0, 3 M KCl, 1 mM DTT, and 100 mL of buffer A (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 10% glycerol) containing 20 mM imidazole. The resins were then poured into columns and bound PNPase was step-eluted with 300 mM imidazole in buffer A. The imidazole eluates were adjusted immediately to 0.5 mM EDTA and then dialyzed overnight against buffer B (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 0.5 mM EDTA, 10% glycerol). The dialysates were applied to 2-mL DEAE-Sephacel columns that had been equilibrated with buffer B, and the PNPase proteins were recovered in the flow-through fractions. The PNPase preparations were gel-filtered through a column of Superdex-200 equilibrated in 20 mM Tris pH 8.0, 100 mM NaCl, 10% glycerol. The PNPase elution profiles were monitored by A280 and affirmed by SDS-PAGE of the column fractions. The peak Superdex-200 fractions of wild-type PNPase and PNPase-Ala mutants were used for biochemical assays. The PNPase preparations were stored at −80°C. Protein concentrations were determined by using the Bio-Rad dye reagent with BSA as the standard.

Cryo-EM of MsmPNPase in the act of poly(A) synthesis

A reaction mixture (740 µL) containing 3.58 mg of the DEAE fraction of wild-type PNPase (in 650 µL buffer B), 4.9 mM MgCl2, 1.9 mM ADP, and 0.1 mM oligo(rA)12 was incubated for 3 min at room temperature to allow poly(A) synthesis, then quenched on ice and loaded immediately onto a Superdex 200 Increase analytical gel filtration column in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl at a flow rate of 0.3 mL/min. The peak fraction was concentrated to 2 mg/mL PNPase. Aliquots (3 µL) were applied to glow-discharged UltrAuFoil 300 mesh R1.2/1.3 grids (Quantifoil). Grids were blotted for 1.5 sec at 100% humidity and 5.2°C, then plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI/Thermo Fisher Scientific). All data were collected on a Titan Krios G2 transmission electron microscope (FEI/Thermo Fisher Scientific) operating at 300 kV with a K3 camera (Gatan) at the Richard Rifkind Center for Cryo-EM at the Memorial Sloan Kettering Cancer Center. Using a set defocus range of −0.8 µm to −1.2 µm, 2890 raw movies were recorded at a super-resolution pixel size of 0.532 Å (physical pixel size of 1.064 Å). Exposures were acquired at an electron dose of 20 e−pixel−1 s−1 with a total exposure time of 3 sec, for an accumulated electron dose of 53 e−Å−2 per exposure. Intermediate frames were recorded every 0.05 sec for a total of 60 frames.

Dose-weighted drift correction of the movie frames was performed with MotionCor2 (Zheng et al. 2017). Drift-corrected sums were used for CTF determination and defocus estimation via CTFFIND-4.1 (Rohou and Grigorieff 2015). From these images, 3,057,420 particles were automatically picked using the Laplacian-of-Gaussian filter and extracted using a particle box size of 300 pixels in RELION-3 (Zivanov et al. 2018). Particles were transferred to cryoSPARC v2 (Punjani et al. 2017) for 2D classification, where data was initially split into 100 averaged classes. Of these, 13 classes constituting 840,466 particles with high-resolution particle features were selected for generating an ab initio 3D model. Excluded particles from the 2D classification were used to generate five classes of “bad” 3D models in another ab initio run. These “bad” 3D models were used, in conjunction with the first ab initio model of the selected particles and the original particle stack from RELION, as input in a 3D heterogeneous refinement job for the purpose of filtering out particles not associated with the sample of interest. This process yielded 890,249 particles for subsequent 3D refinement, or 5.9% more particles than the 13 original 2D classes. Due to apparent threefold symmetry in the 2D and 3D classes, threefold (C3) symmetry was applied in subsequent refinements. The filtered particles were then subjected to one round of 3D heterogeneous refinement, followed by one round of nonuniform refinement, resulting in a map at an overall resolution of 3.07 Å.

An initial MsmPNPase model generated in Phyre2 (Kelley et al. 2015) was fit to the experimental map and real space refinement was performed in Phenix (Adams et al. 2010). Iterative model building was performed using O (Jones et al. 1991). The final PNPase model contains 1761 protein residues, 15 nucleotides, and six magnesium ions. RMSDs for bond lengths and angles are 0.006 Å and 0.65°, respectively. 96.4% of the protein residues fall in the most favored regions of the Ramachandran plot. The correlation between the map and the model is 0.85.

RNA phosphorylase assay

A synthetic 24-mer RNA was 5′-radiolabeled by using T4 polynucleotide kinase and [γ32P]ATP and then purified by electrophoresis through a native 20% polyacrylamide gel. RNA phosphorylase reaction mixtures containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 0.5 mM (NH4)2PO4, 0.1 µM 5′ 32P-labeled 24-mer RNA and 0.1 µM recombinant MsmPNPase homotrimer as specified were incubated at 37°C. The reactions were quenched at the times specified by adding 10 μL of 90% formamide, 50 mM EDTA to 10 µL of phosphorylase reaction mixture. The samples were analyzed by electrophoresis through a 40-cm 20% polyacrylamide gel containing 7.5 M urea in 44.5 mM Tris-borate (pH 8.3), 1 mM EDTA. The radiolabeled RNAs were visualized by autoradiography.

RNA polymerase assay

Reaction mixtures containing 20 mM Tris-HCl (pH 7.5), 5 mM MgCl2, 2 mM ADP, 1 pmol (0.1 µM) 5′ 32P-labeled 24-mer RNA primer, and 10 pmol (0.1 µM) recombinant MsmPNPase homotrimer were incubated at 37°C. The reactions were quenched with an equal volume of 90% formamide, 50 mM EDTA. The samples were analyzed by electrophoresis through a 40-cm 12% polyacrylamide gel containing 7.5 M urea in 44.5 mM Tris-borate (pH 8.3), 1 mM EDTA. The radiolabeled RNAs were visualized by autoradiography.

DATA DEPOSITION

Atomic structure models of MsmPNPase have been deposited in the PBD and EMDB databases under ID codes 7LD5 and EMD-23282, respectively.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.