The structure of a protein primer–polymerase complex in the initiation of genome replication

Two different crystal forms of the complex between the FMDV 3D polymerase, the primer protein VPg, UTP and metal ions, in the absence and in the presence of an RNA oligonucleotide of 10 adenine bases (A10) were obtained and analyzed by X‐ray crystallography. The structures of the complexes were obtained at 3.0 and 2.9 Å, respectively (Table I). In the two crystal forms, the quality of the electron density maps, after several rounds of manual modeling and refinement, allowed the accurate positioning of 15 of the 23 amino acids of the VPg protein (Figure 1). Weak and discontinuous density was also visible to accommodate one or two additional VPg residues poorly ordered, exiting from the polymerase. The remaining six carboxy‐terminal residues of VPg appeared completely disordered.

The 15 VPg amino acids determined in both structures showed almost the same conformation with little secondary structure. The N‐terminal portion of the protein is located close to the NTP entry cavity and projects the side chain of residue Tyr3 into the active site. Then, the peptide chain snakes through the large RNA binding cleft towards the thumb domain of the 3D protein, following a similar trajectory to that taken by the RNA primer and duplex product in the 3D–RNA complex (Ferrer‐Orta et al, 2004; Figure 2).

In the crystals obtained when the oligonucleotide A10 was absent, no density was observed to accommodate the UTP substrate. However, in the complex obtained in the presence of A10 and MgCl₂, the difference in electron density maps (2∣F_o∣−∣F_c∣ and ∣F_o∣−∣F_c∣) showed the product of the VPg uridylylation (VPg‐UMP) and revealed the presence of two metal ions in the polymerase catalytic site (Figures 1A and B). To distinguish whether Mg²⁺ or Mn²⁺ was at the active site was not obvious because both cations were present during the complex formation and the crystallization processes. We have recently reported the ability of FMDV 3D polymerase to catalyze the uridylylation of VPg using poly(A) as a template in the presence of Mn²⁺ ions (Arias et al, 2005). In contrast, the VPg uridylylation in the presence of Mg²⁺ was undetectable (A Arias and R Agudo, unpublished results). In order to provide an experimental evidence of the presence of Mn²⁺ in the active site of our complex, we collected a new data set of the 3D–VPg–UTP–metal complex cocrystals using the wavelength of the radiation tuned to the Mn²⁺ K edge, 1.89 Å (see Materials and methods section and Table I). The anomalous scattering difference Fourier map showed one peak (4σ level), indicating the presence of a single Mn²⁺ ion bound to the active site of the polymerase. The second strong peak seen in conventional difference Fourier maps was then interpreted by the presence of one additional Mg²⁺ cation (Figure 1). No ordered density was visible in the template entry channel to position the oligonucleotide that was expected to act as a template in the uridylylation. Therefore, it seems that uridylylation reaction took place before crystal formation. The pyrophosphate product was also absent. The position of the oligo(A) template was further modeled using the FMDV 3D–RNA template/primer complex (Ferrer‐Orta et al, 2004) as a guide (Figure 1B). The resulting model shows how the first five nucleotides of the oligo(A) template traverse the template entry channel pointing the base, A5 towards the catalytic site in a correct position, to be paired with the UMP molecule covalently attached to VPg (Figure 1B).

The structures determined show how VPg accesses the polymerase catalytic site from the front part of the molecule through the RNA binding cleft (Figure 2). No important conformation changes of the 3D polymerase conformation were induced by the VPg binding; only local rearrangements in the side chain of residues Arg179 and Asp338, which participate directly in the uridylylation reaction, were seen (see below). The largest network of VPg–3D interactions was observed between residues Glu166, Ile167, Arg168, Lys172 and Arg179, in motif F of the fingers domain, which, together with residues Thr407, Ala410 and Ile411 of helix α13 of the thumb domain, contacted the N‐terminal moiety of VPg, stabilizing the conformation of Tyr3 in the active site cavity (Figure 3). Two main chain–side chain hydrogen bonds were observed between the backbone oxygen atoms of Pro2 and Tyr3 of VPg, and the amino groups of the side chains of Lys172 and Arg168 of the polymerase. The guanidinium group of Arg179, which forms a double salt bridge with Glu166 within the same motif F of 3D is also hydrogen bonded with the hydroxyl group of Tyr3 in the nonuridylylated VPg. In the complex with the uridylylated form of VPg, the side chain of Arg179 changes slightly its conformation, maintaining the interaction with Glu166 and forming an additional salt bridge with the phosphate oxygen O2 of UMP (Figures 1B and 3). The main chain oxygen atoms of Pro6 and Leu7 of VPg were hydrogen bonded to the side chain of the 3D residue Lys387, located in motif E of the palm domain. The side chain of Leu7 is also located in a hydrophobic cavity formed by the side chains of residues Glu166 and Ile167 of motif F and the side chains of Thr407, Ala410 and Ile411 of helix α13 (Figure 3). Additional contacts were observed between Arg388 of motif E and residues Glu8, Arg9 and Gln10 of VPg. The side chain of Tyr336, within the catalytic motif C, also participates in interactions with VPg forming a hydrogen bond with the guanidinium group of Arg11 of the peptide primer. Finally, the 3D residues Gly216, Cys217 and Pro219, located at the beginning of the helix α8 in the fingers domain, also establish hydrophobic contacts with VPg residue Arg11 at the exit of the polymerase cavity (Figure 3).

The comparison of the 3D–VPg and the 3D–VPg–UMP structures reveal how the side chain of Tyr3 is oriented towards the active site close to the catalytic aspartic acids 245 of motif A and 338 of motif C. In the 3D–VPg–UMP complex, the hydroxyl group of Tyr3 side chain was found covalently attached to a UMP molecule by a phosphodiester linkage (Figure 1B). Two metal ions, Mn²⁺ and Mg²⁺, participate in the uridylylation reaction that appears to follow a similar mechanism to that described for the nucleotidyl transfer reaction in other polymerases (Steitz, 1998). Mn²⁺ bridges the carboxylate group of Asp338 and the O⁻ of tyrosine side chain, now covalently bonded to phosphate α of UMP. The side chain of Asp338 changes its rotamer conformation to optimize the described interactions. Mg²⁺ coordinates the carboxylic group of Asp245, the O1 oxygen of phosphate α and the hydroxyl group of Ser298 side chain. Tyr336 of motif C, which is hydrogen bonded to the VPg residue Arg11, also participates in hydrophobic contacts with the UMP molecule covalently attached to VPg (Figure 3). Finally, the positively charged residues of motif F (Arg168, Lys172 and Arg179) also participate in the uridylylation process (Figure 1B). These basic amino acids of motif F, together with Tyr336 of motif C, seem to play a role stabilizing Tyr3 and UMP in a proper conformation for the catalytic reaction.

To investigate the effect of amino‐acid replacements in 3D and VPg in the uridylylation reaction, five single replacements (Glu166‐Ala, Glu166‐Arg, Arg168‐Ala, Arg179‐Ala, Asp338‐Ala) and two double replacements (Lys387‐Ala/Arg388‐Ala, Thr407‐Ala/Ile411‐Ala) were introduced in 3D by site‐directed mutagenesis of plasmid pET‐28a3Dpol (Materials and methods). The uridylylation activity on VPg1 was tested and compared with that of the wild‐type 3D. The results (Figure 4A) show at least a 10‐fold reduction for all mutants examined, except for 3D mutant Arg168‐Ala that showed about 60% of the activity relative to wild‐type 3D, and the double mutation Thr407‐Ala/Ile411‐Ala, which did not produce any detectable effect on the uridylylation activity.

VPg replacements Tyr3‐Phe, Glu8‐Ala, Arg9‐Ala, Arg9‐Glu and, to a lesser extent, Pro6‐Ala resulted in significant decreases in uridylylation activity, as compared with wild‐type VPg1, VPg2 or VPg3 (Figure 4B). Thus, modification of 3D and VPg residues, predicted to be involved in critical interactions according to the 3D structure of the uridylylation complex, have remarkable effects in the uridylylation reaction.

The 3D structures of RdRPs, unbound to ligands, are available for four other members of the Picornaviridae family (PV, HRV1B, HRV14 and HRV16) (Love et al, 2004; Thompson and Peersen, 2004; Appleby et al, 2005). The structure‐based alignment of this five picornaviral polymerases shows that 12 out of the 16 residues of 3D polymerase involved in VPg binding are strictly conserved (Figure 5A). Sequence alignments of the VPg proteins from these picornaviruses also revealed a partial conservation of the 3D‐interacting residues (Figure 5B). Mapping of the 12 VPg‐interacting residues onto the surface of the FMDV polymerase is shown in Figure 6A. Structural comparisons between FMDV–3D–VPg complex with the other four picornaviral RdRPs showed that the size and the shape of the central cavity, where VPg binds, is almost identical in all five polymerases (Figure 6B), and FMDV VPg protein can be modeled in the central cavity of all picornaviral polymerases, without important steric hindrances. The modeled VPg in the central cavity of PV and rhinovirus RdRPs retained almost the same interactions described in the FMDV–3D–VPg complex. In particular, those relevant to the uridylylation reaction would be strictly conserved (Figures 5 and 6B).

The FMDV mutants at the conserved polymerase residues that strongly interact with VPg—Glu166 and Arg179 of motif F, Asp338 of motif C and Lys387/Arg388 of motif E—show a drastic defect in VPg uridylylation (Figure 4). Substitution of Arg168 resulted in ∼40% reduction in uridylylation. This residue contacts VPg through a hydrogen bond with the backbone oxygen atom of Tyr3. The lack of the Arg168 side chain can be, at least in part, compensated by the presence in the vicinity of other positively charged side chains, like that of Lys172 (Figure 3). The double mutation Thr407/Ile411 to alanine shows no defect in VPg uridylylation (Figure 4). These residues form part of a hydrophobic cavity that accommodates Leu7 of VPg (Figure 3). It seems that alanine residues do not disturb this hydrophobic cavity. Both residues are not conserved among picornaviruses.

Mutational analysis in PV has identified two groups of residues at the surface of the polymerase whose substitutions affected either the rate of VPg uridylylation or the 3AB binding affinity and the ability of membrane‐bound 3AB recruitment. Uridylylation was also partially affected by this second group of mutants (Lyle et al, 2002; Boerner et al, 2005; Figure 6C). Substitution of the first group of residues (Tyr326, Asp358 and Lys359) corresponding to FMDV (Tyr336, Asp368 and Lys369) results in a drastic loss of uridylylation activity (Lyle et al, 2002). These residues are located at or near the VPg binding site (Figure 6C). The second group of residues (Phe377, Arg379, Glu382 and Val391) that would correspond to His389, His391, Tyr394 and Val402 in FMDV are located on the back side of the polymerase and lie far from VPg to allow direct contacts with this primer protein (Figure 6C). Substitution of most amino acids of this second group resulted only in a moderate decrease of VPg uridylylation activity, perhaps because this major effect is on binding of the 3AB precursor rather than of VPg (3B). Additional studies are needed to clarify this point. It can also be considered that these second group of residues, which lie within and around motif E at the hinge region of the palm and thumb domains, could play a role in maintaining the structural integrity and proper positioning of key polymerase residues during VPg uridylylation.

Replacement of the conserved VPg residues interacting with 3D (Tyr3, Pro6, Glu8 and Arg9) resulted in a dramatic reduction in the rate of VPg uridylylation (Figure 4B). Previous mutational studies in PV, using an inducible yeast two‐hybrid system, identified three positively charged residues of VPg (Lys9, Lys10 and Arg17; Figure 5B) as interacting with the 3D polymerase (Xiang et al, 1998). The mutational and structural results obtained in the present work (Figures 3 and 4) correlate well with this previous observation with PV.

In light of the structural and mutational results, we suggest a general mode of VPg binding and uridylylation through the front side of the picornaviral polymerases. A similar VPg binding mode was previously hypothesized for HRV16, on the basis of the crystallographic structure of the RdRP of this virus and molecular modeling (Appleby et al, 2005).

Comparisons among different RNA‐dependent RNA polymerases whose structures have been solved show that those viruses that follow a primer‐dependent mechanism of initiation of replication, such as picornaviruses and caliciviruses, have a more accessible active site than viruses with a de novo initiation mechanism, such as flaviviruses and bacteriophage ϕ6 (van Dijk et al, 2004; Ferrer‐Orta et al, 2004, 2006). An extra C‐termini domain has been determined for polymerases with a de novo initiation of RNA synthesis (Butcher et al, 2001; van Dijk et al, 2004). The C‐terminal protrusions of flaviviruses and ϕ6 polymerases partially occlude the active site, resulting in a more compact molecule where two narrow positively charged tunnels allow the access of RNA template and NTP substrate to the active site. The template tunnel is wide enough to accommodate ssRNA but not dsRNA (van Dijk et al, 2004; Ferrer‐Orta et al, 2006). Structural and biochemical studies indicated that the protruding extensions of the thumb domains of phi6 and flavivirus polymerases play two distinct functions: acting as priming platforms that stabilize the initiation complexes, and preventing undesirable back‐priming reaction by physically separating the template binding site from the room reserved for the daughter RNA chain. Furthermore, these initiation platforms block the path of the elongating RNA product at the level of two or three nucleotides and large conformational rearrangements are required to accommodate longer product chains (Van Dijk et al, 2004; Ferrer‐Orta et al, 2006).

In contrast, the structure of FMDV 3D in complex with RNA determined recently shows how the wide central cleft of picornavirus polymerases is able to accommodate a template–primer duplex during the phase of RNA elongation (Ferrer‐Orta et al, 2004), and the structure presented in this work demonstrated that this cavity can also accommodate the protein primer during the initiation step of picornavirus replication.

RNA synthesized by DNA primases is involved in the initiation of cellular DNA replication (Alberts et al, 2002). RNAs act as primers for replication of some DNA viruses, and for transcription of several RNA viruses; capped RNA structures, captured from cellular mRNAs, serve to initiate orthomyxovirus and bunyavirus replication by ‘cap snatching’ (Alberts et al, 2002; Flint et al, 2004).

The position of VPg in complex with the FMDV 3D polymerase is remarkably similar to the position of the primer and RNA duplex product found in the complex with the same enzyme (Figure 2B) (Ferrer‐Orta et al, 2004). Most of the amino acids of 3D seen in contact with the RNA primer and duplex product are also involved in interactions with VPg (Figures 1B and 3). In fact, the structure shows how the VPg protein accesses the active site cavity from the front of the molecule through the large RNA binding cleft mimicking, at least in part, the RNA molecule (Figure 2). The N‐terminal position of VPg projects into the active site where the hydroxyl moiety of the residue Tyr3 is in good proximity to the catalytic aspartates 245 of motif A and 338 of motif C. In this position, Tyr3 essentially mimics the 3′OH of the primer strand during the RNA elongation. All the observed structural features suggest a conservation of the catalytic mechanism described for all polymerases (Steitz, 1998).

For viruses, whose survival must rely in the efficient production of many progeny copies, stability and velocity of replication are essential, and these features may have favored, in prolonged evolutionary times, protein versus RNA, as an alternative to other chemical modifications to protect the genome termini.

FMDV polymerase (3D) was obtained and purified as described in Ferrer‐Orta et al (2004) and stored in a buffer containing 40 mM Tris–HCl (pH 7.5), 0.5 M NaCl, 0.8 mM DTT, 0.8 mM EDTA and 8% glycerol at a concentration of 4.6 mg/ml. The VPg used for crystallization was prepared by solid‐phase peptide synthesis of the FMDV C‐S8c1 VPg1 sequence (GPYAGPLERQRPLKVRAKLPRQE), purified by G25 Sephadex chromatography and HPLC, and analyzed by mass spectrometry.

The 3D–VPg complex was obtained and crystallized as follows: 9.5 nmol of VPg were mixed with 2 mM UTP and then the 3D polymerase was slowly added in an equimolar proportion in the presence of 2 mM MnCl₂. The solution was left 12 h at 4°C before the crystallization trials. Crystals were grown by vapor diffusion at 20°C from a solution containing 33% PEG 4K, 0.2 M ammonium acetate, 0.1 M sodium citrate (pH 5.6) and 4% γ‐butyrolactone. Trigonal P3₂21 crystals (a=b=95.09 Å, c=100.63 Å) typically appeared after 2 days.

The 3D–VPg–UMP–oligoA10 complex was obtained and crystallized as follows: firstly, 5 nmol of 3D polymerase were mixed with an equimolar proportion of the A10 oligonucleotide in the presence of 2 mM MgCl₂. A second solution was prepared containing 5 nmol of VPg and UTP (2 mM) in the presence of 2 mM MnCl₂. Then, both solutions were slowly mixed and incubated for 12 h at 4°C. Trigonal crystals (P3₂21, a=b=94.38 Å, c=99.71 Å) were grown from a solution containing 32% PEG 4K, 0.2 M ammonium acetate, 0.1 M MES (pH 6.0) and 4% γ‐butyrolactone.

The two crystal forms were flash frozen in liquid nitrogen, using cryoprotectants that contained 20% glycerol in the crystallization buffers. Three data sets were collected at 100K using synchrotron radiation at the ESRF (Table I). The X‐ray diffraction data for the 3D–VPg complex were collected at 3.0 Å resolution on beam line ID14 EH2 (λ=0.93 Å). For the 3D–VPg–UMP crystals, two different data sets were obtained. The first data set was collected at 2.9 Å on beam line ID14 EH4 (λ=0.98 Å) and the second X‐ray data (3.2 Å resolution) were collected on BM16 beam line at the K edge of manganese, 6547 KeV (λ=1.89 Å). All X‐ray data were processed and reduced using DENZO/SCALEPACK package (Table I) (Minor and Otwinowski, 1997).

The initial maps were obtained after a rigid body fitting of the coordinates of 3D protein (PDB entry 1WNE; Ferrer‐Orta et al, 2004) to the new unit cells (Table I), using the program CNS (Brunger et al, 1998). In both structures, these initial 2∣F_o∣−∣F_c∣ and ∣F_o∣−∣F_c∣ difference maps showed an elongated extra density covering part of the RNA binding groove that was interpreted by the presence of the VPg primer protein occupying the cavity. The density was clear to interpret with confidence the first 15 amino acids of VPg (Figure 1A), but the carboxy‐terminal part of the oligopeptide was disordered and could not be interpreted.

In the structure of the complex obtained in the presence of UTP, A10, MnCl₂ and MgCl₂, the difference density maps showed the product of the VPg uridylylation (VPg–UMP), plus two strong peaks of density that would be interpreted by the presence of two metal ions in the polymerase catalytic site (Figures 1A and B). Data collected at the manganese K edge were used to detect one Mn²⁺ ion, via anomalous difference Fourier maps (Figure 1B). The second strong peak seen in conventional difference Fourier maps was then interpreted by the presence of one additional Mg²⁺ ion. No ordered density for the pyrophosphate product was seen in the vicinity of the polymerase active site. The A10 oligonucleotide added as a template in the uridylylation reaction was also disordered in the electron density maps. Several cycles of automatic refinement, performed with program CNS, were alternated with manual model rebuilding using the program TURBO (Roussel and Cambillau, 1991). The statistics of the refinement in both complexes are summarized in Table I. Coordinates have been deposited at the PDB with accession codes 2D7S and 2F8E for the 3D–VPg and 3D–VPg–UMP, respectively.

Mutant polymerases with substitutions Glu166‐Ala, Glu166‐Arg, Arg168‐Ala, Arg179‐Ala, Lys387‐Ala/Arg388‐Ala or Thr407‐Ala/Ile411‐Ala were obtained using expression and purification protocols previously described for the wild‐type 3D (Arias et al, 2005). Different vectors encoding for all mutant polymerases have been constructed by site‐directed mutagenesis by using the QuickChange Site‐Directed Mutagenesis Kit (Stratagene) as indicated by the manufacturer. Briefly, for each construction, a different pair of primers (sense and antisense) coding for an amino‐acid substitution (Glu166‐Ala, Glu166‐Arg, Arg168‐Ala, Arg179‐Ala) or two‐amino‐acid substitutions (Lys387‐Ala/Arg388‐Ala, Thr407‐Ala/Ile411‐Ala) was used. The vector that contains the wild‐type 3D‐coding region (pET‐28a3Dpol) was amplified with the appropriate pair of primers to introduce the desired mutation. The PCR product was digested with DpnI and used to transform XL1‐Blue Supercompetent Cells (Stratagene). Positive colonies were isolated, plasmid DNA was extracted and transfected into Escherichia coli BL21 cells. The DNA was extracted again, and the entire 3D‐coding region was sequenced to confirm the presence of the expected mutation(s). The cells were grown, IPTG‐induced, collected, lysed, and the mutant 3Ds purified as described previously (Ferrer‐Orta et al, 2004; Arias et al, 2005). All mutant polymerases were >95% pure by SDS–PAGE analysis.

All mutant peptides were prepared by solid‐phase peptide synthesis, purified by G25 Sephadex chromatography and HPLC, and analyzed by mass spectrometry as described previously (Arias et al, 2005). The peptides prepared are the FMDV VPg1 (GPYAGPLERQRPLKVRAKLPRQE), VPg2 (GPYAGPMERQKPLKVKARAPVVKE), VPg3 (GPYAGPVKKPVALKVKAKNLIVTE) and the VPg1 mutants Tyr3‐Phe, Pro6‐Ala, Glu8‐Ala, Arg9‐Ala and Arg9‐Glu.

The VPg uridylylation assay was performed as described in Arias et al (2005). The uridylylation mixture contained 30 mM MOPS (pH 7.0), 33 mM NaCl, 0.6 mM MnCl₂, 40 ng/μl poly(A) (300 residues on average), 150 μM VPg, 0.4 mg/ml BSA (Boehringer), 8% glycerol and 50 μM [α‐³²P]UTP (0.01 mCi/ml; 200 mCi/mmol and 0.4–0.8 μM 3D; 46 μl of the mixture were prewarmed at 37°C and the reaction was started by adding 4 μl of 3D). The standard reaction was carried out for 30 min at 37°C, and stopped by adding 10 μl of 500 mM EDTA; the product was analyzed by SDS–PAGE in a Tris–Tricine buffer (16% (w/v) acrylamide, 10% (v/v) glycerol).

Amino‐acid sequences of picornaviral 3D and VPg proteins were retrieved from GeneBank, aligned using the program ClustalW (Higgins et al, 1996) and rendered with ESPript (Gouet et al, 1999).