|Home | About | Journals | Submit | Contact Us | Français|
The flavivirus NS5 N-terminal RNA methyltransferase (MTase) enzyme is responsible for methylating the viral RNA cap structure. To increase our understanding of the mechanism of viral RNA cap binding we performed a detailed structural and biochemical characterization of guanosine cap binding pocket of the dengue (DEN) and yellow fever (YF) virus MTase enzymes. We solved an improved 2.1 Å resolution crystal structure of DEN2 MTase and new 1.5 Å resolution crystal structures of the YF virus MTase domain in apo form and new 1.45 Å structure in complex with GTP and RNA cap analog. Our structures clarify the previously reported DEN MTase structure, suggest novel protein–cap interactions, and provide a detailed view of guanine specificity. Furthermore, the structures of the DEN and YF proteins are essentially identical, indicating a large degree of structural conservation amongst the flavivirus MTases. GTP analog competition assays and mutagenesis analysis, performed to analyze the biochemical characteristics of cap binding, determined that the major interaction points are (i) guanine ring via π–π stacking with Phe24, N1 hydrogen interaction with the Leu19 backbone carbonyl via a water bridge, and C2 amine interaction with Leu16 and Leu19 backbone carbonyls, (ii) ribose 2′ hydroxyl interaction with Lys13 and Asn17, and (iii) α-phosphate interactions with Lys28 and Ser215. Based on our mutational and analog studies, the guanine ring and α-phosphate interactions provide most of the energy for cap binding, while the combination of the water bridge between the guanine N1 and Leu19 carbonyl and the hydrogen bonds between the C2 amine and Leu16/Leu19 carbonyl groups provide for specific guanine recognition. A detailed model of how the flavivirus MTase protein binds RNA cap structures is presented.
Flavivirus genomes consist of a 5′ capped 10.7–11.0 kilobase positive-strand RNA that lacks a poly(A) tail 1. Viral proteins are translated directly from the genomic RNA in a cap-dependent manner to produce a single polyprotein that is co-translationally cleaved into 10 mature proteins. Insect-borne flaviviruses require the 5′ RNA cap structure for efficient translation of the viral polyprotein. The endogenous mRNA capping machinery is located in the nucleus, while flavivirus RNA replicates in endoplasmic reticulum derived replication compartments in the cytoplasm. Because flaviviruses do not have access to the cellular capping machinery, they have evolved their own capping enzymes to form a cap structure that can be recognized by the cellular translation machinery. Flavivirus genomic RNA is modified at the 5′ end by a cap-1 structure (me7-GpppA-me2) generated by the virus encoded nucleotide triphosphatase (NS3), 2′-O-nucleoside methyltransferase (NS5), guanine-N7-methyltransferase (NS5), and as yet unidentified guanylyltransferase 2; 3; 4; 5; 6; 7; 8. The structure of the dengue, West Nile, Murray Valley encephalitis (MVEV), and Mebean virus NS5 N-terminal 2′-O- / Guanine N7-methyltransferase domain (from now on referred to as MTase) have been solved 5; 6; 9; 10; 11; 12; 13; 14. The MTase protein binds the guanosine nucleoside and phosphates of the viral RNA cap structure and sequentially transfers a methyl group from S-adenosyl methionine first to the N7 position of the guanosine cap and then to the 2′ hydroxyl position on the ribose of the first viral genomic RNA nucleotide to form the cap-1 structure 8; 9. Cap-1 modified mRNAs are specifically recognized by the translation initiation factor eIF4E in combination with eIF4A and eIF4G to form the eIF4F cap binding complex 15. Ribosome-mediated translation cannot occur on many mRNA species without specific recognition of me7-GpppN by eIF4E, and it has been suggested that efficient ribosome recruitment is enhanced by the 2′-O-methylation of cap 1 modified mRNAs 16; 17. Mutation of flavivirus MTase residues that interact with the guanosine base of the viral RNA cap is lethal to viral replication 8; 18; 19, indicating that MTase guanosine binding is essential to viral growth. The essential nature of this cap-binding pocket for RNA replication makes it an ideal target for drug development, and a thorough understanding of how the MTase recognizes and binds cap structures is necessary to facilitate rational drug design efforts.
Limited analysis of flavivirus MTase cap binding has been performed 5; 20, but a thorough description of the protein and cap components of binding is lacking. To increase our understanding of how the flavivirus MTase binds RNA cap structures, we have performed a detailed analysis of the guanosine cap binding characteristics of the dengue (DEN) and yellow fever (YF) MTase enzymes. We have solved the three dimensional structure of the YF MTase domain in the apo form at 1.5 Å resolution and complexed with GTP and cap analog at 1.45 Å resolution. We used this high-resolution structural information to guide our assessment of the critical structural components of guanosine binding using solution-based competition assays with GTP analogs and mutational analysis. We show the relative contributions of both known amino acid/cap interactions as well as several novel interactions not previously described. We also evaluated the functional components of GTP necessary for strong interaction with the cap binding site and determined the detailed mechanism for cap guanine recognition by flavivirus MTases. The results of these experiments expand our knowledge of the cap binding site in flavivirus MTase enzymes, and provide a foundation for the rational development of potent and specific antivirals active against flaviviruses. Furthermore, as many of the amino acids involved in MTase activity appear to be conserved across the entire flavivirus genus, antivirals designed to bind to this site may possess significant broad-spectrum activity against multiple flaviviruses.
The YF MTase domain has a compact globular fold where one face of the protein is involved in ligand binding as judged by strong electron density for a bound Ado-Hcy coproduct that co-purified with the protein and the binding site for GTP (Figure 1a). The structure is highly homologous to those of the dengue fever and West Nile virus proteins 5; 8; 11, with structural differences being limited to small changes in the conformation of loop residues, as shown by the maximum-likelihood based multiple structure superpositioning 21 (Figure 1b). Crystal soaking experiments with GTP and GpppA resulted in clear positive difference density for these ligands in a surface exposed pocket located ~15 Å away from the Ado-Hcy coproduct across one face of the protein (Figure 1a). Definitive density was observed for the guanine and ribose groups while phosphate density was generally limited to the α and β phosphate groups, except for a very interesting stacked conformation of the GpppA cap analog as discussed below. Altogether, there are surprisingly few nucleotide specific interactions in the MTase-GTP complex and it is difficult to discern obvious reasons for GTP specificity in the binding pocket by inspection of the structure.
The dominant structural interaction between the bound GTP and the protein is a π–π stacking interaction with Phe24 that positions the nucleotide base moiety on the surface of the YF MTase and locks the ribose into a shallow hydrated binding pocket (Figure 2). This stacking interaction results in a small shift in the position of the α-helix composed of residues 24–34 upon ligand binding (Figure 1b). The exocyclic C2 amine of the guanine forms a pair of hydrogen bonds with the backbone carbonyls of Leu16 and Leu19 in a loop of the protein at the outer edge of the GTP binding pocket. In addition, all the ligand-bound structures show a well-ordered water molecule that is hydrogen bonded between the N1 nitrogen on the guanine ring and the backbone amide of Leu19 (labeled “W” in Figure 2). This water appears to be an integral part of the nucleotide bound complex because there is no electron density evidence whatsoever for it in the apo structure. There is a potential interaction between the exocyclic O6 atom and Lys21 that could provide a guanine specific recognition point, but the lysine sidechain is in fairly weak density and thus this does not appear to be a particularly strong interaction, consistent with our binding data from mutagenesis experiments.
The ribose moiety is involved in several hydrogen bonds to protein residues and bound waters (Figure 2). The outside of this pocket is composed of three consecutive serine residues (150–152) that are all involved in ribose-triphosphate interactions. Ser150 forms a hydrogen bond with the β phosphate group, the backbone carbonyl of Ser151 helps to position Lys13 for interactions with the 2′ and 3′ ribose hydroxyls, and the backbone carbonyl of Ser152 is hydrogen bonded to the sidechain nitrogen of Asn17, positioning the Asn17 sidechain oxygen for a hydrogen bond with the 2′ hydroxyl. The GTP phosphate groups extend out over the surface of the protein toward the Ado-Hcy binding site and the α and β phosphate groups are generally well ordered while little density is observed for the terminal γ phosphate.
The structure of the GpppA cap analog in the citrate crystal form is intriguing in that there is clear electron density for the entire GpppA molecule in an internally stacked conformation where the adenosine is packed onto the solvent exposed face of the bound guanosine (Figure 2C). A similar conformation was recently reported in DEN MTase crystals soaked with various cap analogs 10. In contrast, the low ionic strength MPD crystal form only shows density for the guanine ring, ribose, and α-β phosphates while the adenosine is disordered (Figure 2F) similar to what was observed with MVEV MTase complexed with m7GpppA 12. In both cases the stacked orientation was not observed when the ionic strength of the crystallization buffer was low (Figure 2F and reference 12) whereas the stacked orientation is only observed when the ionic strength is very high (Figure 2C and reference 20). The stacked interaction is likely stabilized by the high ionic conditions, reducing electrostatic repulsion within the triphosphate moiety that is now folded back on itself, although we do not see any well-defined electron density for ordered sodium counterions. Because this stacked conformation has not been observed in less than 1.6 molar sulfate or citrate, the biological significance of the stacked conformation is unclear.
Last, we see crystallographic evidence for a second triphosphate binding site in the MPD crystal form complexes with GTP and GpppA. In these structures there is significant density for a di- and tri-phosphate moiety in a new site near Arg27 and Arg87 that is located above the bound Ado-Hcy coproduct (Figure 1a). This is the same site in which a sulfate ion was previously observed in the DEN MTase structure 5; 11 as well as in our DEN structure, and has been suggested act as a channel for RNA entry into the MTase 13; 20. In examining the electron density maps, there is no evidence of a specific orientation of the attached but unresolved ribose and guanine base in this binding site.
GTP binding has been used previously as a surrogate for guanosine cap binding 7 because the guanosine cap structure on cellular mRNAs and flavivirus RNA resembles an inverted GTP. Previous analysis of GTP/cap binding has utilized a UV crosslinking assay with radiolabeled GTP to study binding, followed by PAGE and autoradiography. To study the binding of GTP, and by extension the guanosine cap structure, to the MTase enzyme, we have established an in vitro fluorescence polarization (FP) based binding assay. This assay allows for sensitive, homogenous, and real-time determination of GTP binding to purified MTase protein. FP assays have been successfully used to analyze GTP binding to a number of proteins 22; 23. We first tested the binding affinity of the DEN, WNV, and YF MTase proteins for GTP-Bodipy using binding saturation curves. Using this approach, we obtained similar Kd values for GTP-Bodipy bound to each protein (DEN = 72 ± 2 nM; YF = 124 ± 6 nM; WNV = 115 ± 6 nM (Figure 3A)). To confirm that GTP-Bodipy has a similar Kd to that of GTP we measured the change of FP as a function of DEN protein concentration in the presence of various concentrations of GTP (GTP-Bodipy Kd = 126 ± 15 nM; GTP Ki = 119 ± 25 nM (Figure 3B)). These data demonstrate that GTP-Bodipy and GTP bind with similar affinity, thus validating our assay. DEN, WNV, and YF MTase proteins also have very similar Kd values for GTP-Bodipy binding (data not shown). To establish the specificity of guanine cap binding, we attempted to determine Kd values for MTase ATP-Bodipy binding. DEN MTase bound ATP-Bodipy very poorly as compared to GTP-Bodipy (Kd > 13 μM (Figure 3C)), but we were not able to determine an exact Kd for ATP-Bodipy because we could not get the protein concentration in our experiments high enough to saturate binding. Finally, we determined the Ki of GTP and ATP for binding to the DEN MTase using a GTP-Bodipy displacement assay (Figure 3D). 10 nM GTP-Bodipy was complexed with 500 nM DEN MTase protein (~75% fractional saturation), increasing amounts of competitor was added, FP signal was detected, and the resulting Ki value calculated. We determined the Ki of GTP to be 40 ± 10 nM, and the Ki of ATP to be 350 ± 39 μM. Because the MTases appear to be nearly identical in terms of structural and GTP-Bodipy binding characteristics and DEN MTase was the most soluble during purification, we chose to perform the remaining biochemical characterization of guanosine cap binding only with the DEN MTase.
To test the contribution of conserved residues suggested to be involved in cap binding (Figure 4), we mutated individual residues and assessed the effect the mutations had on the affinity of the protein for GTP-Bodipy (Table 2). Of the mutants tested, the most significant effects were observed with the Phe24Ala, Lys28Ala, and Ser215Ala mutants. Interestingly, mutation of Phe24 to Tyr had almost no effect on GTP-Bodipy binding. This result was surprising because the Phe24 is invariant in all known flaviviral MTase enzymes, suggesting that the conservation of Phe24 is due to factors other than guanine binding. Mutation of Lys28 or Ser215 to Ala also resulted in decreased binding (Table 2), indicating that two contacts, one strong (Lys28) and one weak (Ser215), are made with the α-phosphate as evidenced by the structure. Ser150 was previously suggested to interact strongly with the α-phosphate 5;13, but our structural data indicate that an interaction with the β-phosphate is more structurally favorable (Figure 2B). Mutation of Ser150 to Ala resulted in a very modest decrease in binding, suggesting that this interaction makes only a minor contribution to binding. Interestingly, we observed that Arg213 was in a position to potentially interact with the GTP γ-phosphate, though the 2Fo − Fc density maps showed weak density for the residue and the γ-phosphate. Mutating Arg213 to Ala resulted in a modest 2.5 fold reduction in binding affinity, indicating that an interaction between this residue and the γ-phosphate is possible, but not very strong. The γ-phosphate was disordered in the YF GTP and GpppA 2Fo − Fc density maps and the β-phosphate was in weak density, indicating significant 1 flexibility of these phosphates and weaker interactions with the protein as compared to the α-phosphate.
Lys21 is a charge-conserved residue located near α-helix 24–34 in a position near the cap guanosine. The 2Fo − Fc density map shows minimal density for the sidechain of Lys21. However, one of the sidechain rotamers can be positioned within 3 Å of the C6 carbonyl group on the guanosine ring structure. To test if the Lys21 sidechain contributes to binding of the guanine ring, we mutated Lys21 from a basic residue to acidic glutamic acid. The Lys21Glu mutation reduced the affinity by 3.2 fold. However, when we changed Lys21 to Ala, we observed only a minor increase in GTP affinity as compared to wildtype protein, suggesting that the positively charged lysine may be slightly inhibitory to cap binding. Therefore, the Lys21 sidechain does not seem to contribute to guanosine binding, and the effect seen with the Lys21Glu mutant is likely due to repulsion between the C6 amine and the Lys21 amine groups or alteration of the peptide backbone position.
We also attempted to test the contribution of residues that appear to interact with the ribose 2′ and 3′ hydroxyl groups, Lys13 and Asn17. Neither of these mutants expressed in BL21 E. coli despite repeated attempts. DEN Asn17 was mutated and expressed previously 5, so our inability to purify the proteins may be due to the strain of DEN virus MTase used (16681 vs New Guinea) or the difference in the length of the C-terminal domain between the two constructs. We also attempted to purify a Lys28Ala/Ser215Ala double mutant to determine if binding was affected more than with the individual mutants, but the protein was completely insoluble and could not be purified.
We performed a series of GTP-Bodipy displacement assays similar to Figure 3D with GTP analogs to determine which components of GTP are important in binding to the MTase. Ki values for each analog are shown in Table 3. We observed that GTP was the most effective competitor of GTP-Bodipy binding with a Ki of 40 nM, as compared to the Kd of GTP-Bodipy (72 nM). Our finding that the Ki of GTP is comparable to the Kd for GTP-Bodipy confirms that GTP-Bodipy is binding to the MTase in a similar manner to GTP. The slight difference between the Kd of GTP-Bodipy and the Ki of GTP is likely due to weak steric hindrance caused by the Bodipy group as compared to no Bodipy group with GTP. Once we had a baseline for binding, we tested the contribution of each of the phosphates to binding. Removing the γphosphate (GDP) or γ– and β–phosphates (GMP) increased the Ki by 3 and 29 fold, respectively. Removing all three phosphates (guanosine) almost completely abolished the ability to compete with GTP-Bodipy binding (6400 fold reduced). Therefore, while the γ- and β-phosphates do provide some energy to guanosine cap binding, the α-phosphate interaction with Lys28 and Ser215 appear to be critical to tight cap binding. Interestingly, GpppG and GpppA had slightly higher Kis that GTP, which may be due to a slight steric clash with the second guanine or adenine with the MTase protein that would not be present in with GTP.
We then tested the contribution of the ribose 2′ and 3′ hydroxyl groups to binding. dGTP (lacking the 2′ OH) displayed a Ki of 610 nM, which represents a 15 fold decrease in binding as compared to GTP. Removing both hydroxyl groups (ddGTP) resulted in a 770 nM Ki, indicating only a weak contribution from the 3′ hydroxyl.
The purine ring structure of the guanosine cap is a major component of cap binding. To understand what aspects of the guanine ring are important for guanine binding and specificity, we performed competition experiments with guanosine triphosphate analogs modified at different positions on the guanine ring (Table 3). Adenosine triphosphate (ATP), which is identical to GTP except for the purine N1, C2, and C6 prosthetic groups, was unable to effectively compete with GTP-Bodipy (Ki = 350 μM). Therefore the groups at N1, C2, and/or C6 positions are primary determinants for to specificity of binding guanine as compared to adenine. We first examined the contribution of the C2 position by using GTP analogs with substitutions at the C2 position, which was previously suggested to be the primary determinant for guanine specificity 5; 11. Inosine triphosphate (ITP) removes the C2 amine group from GTP, which represents an intermediate between ATP and GTP. ITP was able to compete well for binding (3 fold reduction), indicating a minor contribution of the C2 amine to binding. We then tested xanthine triphosphate (XTP), which replaces the GTP C2 amine with a carbonyl group. XTP was a slightly weaker competitor than ITP (12 fold reduced compared to 3 fold reduced). Therefore, alteration of the C2 amine on the guanine ring structure appears to weakly contribute individually to guanine recognition by the flavivirus MTase but is not the only determinant for binding. We then tested the ability of guanine N1 to act as a hydrogen bond donor to the water that resolved in the YF structures. We used 6-methylthio-GTP (6-MT GTP), which replaces the C6 amine with a thiomethyl group and results in a C6-N1 double bond that removes the N1 hydrogen atom, to test the contribution of the N1 position to binding. 6-MT GTP competed only slightly worse than ITP (230 nM vs 120 nM), but was much better that ATP (350 μM), indicating that the N1 position is a necessary but not sufficient component of guanine recognition.
In this study we performed a detailed biochemical and structural analysis of the guanosine cap binding characteristics of the DEN and YF virus NS5 MTase domains. We present new high-resolution structures of the YF MTase domain complexed with GTP and cap analog, a higher resolution DEN MTase structure, and define many of the critical protein and GTP components of guanosine cap binding. Based on the combination of our structural and biochemical analysis, we present a detailed scheme for the binding of the guanosine triphosphate portion of the cap structure to the flavivirus MTase enzyme (Figure 5). The relative binding affinities for most of the critical interactions observed in the structures are shown. Our data provides the first highly detailed description of the correlates of guanosine cap binding by the flavivirus MTase enzymes. Our studies have indicated a much higher affinity for guanosine cap binding than had previously been recognized, identified several new protein-cap interactions, and demonstrate a detailed mechanism of guanine recognition.
Using our new fluorescence polarization based assay, we determined that the Kd of the flavivirus MTase for GTP-Bodipy is approximately 100 nM, and that the Ki for both GTP and cap analog is in a similar range. The Kd value of the flavivirus MTase domain for cap analog is somewhat higher that the 2 nM cap-binding affinity of the cellular cap-binding eIF4E protein for cap analog as determined by stopped-flow fluorescence titration experiments 24. However, eIF4E stacks the cap guanine ring between two tryptophane residues, which likely provides more stabilization that the single phenylalanine residue in the flavivirus MTase and accounts for the increased affinity observed with eIF4E. In light of the high affinity of eIF4E for guanosine cap, the 100 nM Kd value determined using our assay seems more reasonable than the 58 μM Kd that was previously determined 5.
We performed a comprehensive mutational analysis of the DEN MTase enzyme to determine the contribution of individual amino acids to guanosine cap binding. Figure 4 shows residues within a 10 Å sphere of the cap binding pocket. Conserved residues within hydrogen bonding distance of GTP in our crystal structures noted and were tested for their contribution to GTP interaction. Phe24 and Lys28 have been previously suggested as major interaction points 5; 20. By far the most dramatic effect was seen with the Phe24Ala mutant, indicating that π–π interaction between the guanosine base and the phenylalanine side chain is essential for binding. However, mutation of Phe24 to Tyr did not significantly alter binding. The Phe residue is invariant across the flavivirus family, even though a single uracil to adenine transversion mutation in the second position of the codon would change the residue from Phe to Tyr. This suggests that the Phe residue is functionally conserved, but there may be an additional role for the residue outside of cap binding that requires the residue to remain phenylalanine rather than a tyrosine. Experiments mutating Phe24 to Tyr in infectious virus assays may shed light on why this residue is highly conserved.
Ser215 appears to coordinate the cap α-phosphate in combination with Lys28. The combination of Lys28 and Ser215 interacting with the α-phosphate appears to provide a great deal of stability to the bound cap, as the α-phosphate is well resolved in our structures whereas the β and γ phosphates are not. Mutational and analog competition analyses support this observation, as the Lys28Ala mutation binds GTP-Bodipy weakly and guanosine is a very poor competitor. We also observe a minor effect on binding when we mutate Arg213, even though there is only very weak electron density for the residue. This residue appears structurally poised to influence γ-phosphate binding. Arg213 is also in a region that could be involved in interaction with the 5′ adenosine ring structure of the first viral nucleotide. The position of the viral RNA or 5′ RNA adenosine has not been observed as of this report, but we have performed preliminary computer modeling of a capped viral RNA into the YF MTase structure and observed that the molecular distances are possible for this interaction. Additionally, if the 5′ adenosine base is stabilized between Arg213 and Arg56, this would position the ribose 2′ hydroxyl near AdoMet, increasing the efficiency and specificity of the 2′-O-methyltransferase reaction.
We observed electron density in the positively charged region across from the cap binding site in structures soaked with GTP (Figure 1A) or cap analog (data not shown). We were able to model a triphosphate from GTP into the resolved density, although density for the ribose and guanine on the GTP was not observed. This density could represent a binding site for the triphosphate between the first and second nucleotides in the genomic RNA, with the adenosine residue bound somewhere on the face of the MTase between the AdoHcy and GTP binding sites. Consistent with this, the Arg213/Arg56 pocket at the back of this face could provide hydrogen bonding interactions with the adenosine base. Alternatively, this second phosphate binding site could be involved in binding the cap guanosine residue for the guanine N7 methylation reaction because we can fairly easily place the guanosine N7 atom within a few ångstroms of the Ado-Hcy sulfur after modeling a triphosphate into the observed density. As the methylation of the flavivirus cap occurs sequentially with the guanine N7 being methylated before the ribose 2′ hydroxyl, the observed triphosphate density could indicate the path for the first methylation event.
The α-phosphate is a major interaction point between the protein and cap structure. One possibility for the strength of this α-phosphate: Lys28: Ser215 interaction is that it may represent the as yet unidentified guanylyltransferase activity. Egloff et al. previously showed that α-32P GTP can be covalently crosslinked to DEN MTase in a Mg2+ independent manner 5; 20, which could represent the protein-GMP adduct. Such a role may explain the strength of interaction that is observed between the α-phosphate and the protein, whereas the β- and γ-phosphates appear more weakly associated with the protein. Additionally, in high ionic strength crystallization conditions (1.6M sodium citrate (Figure 1) or 0.4M ammonium sulfate/1.2 M lithium sulfate 20) the cap analog appears to fold back on itself, leaving the α-phosphate exposed for nucleophilic attack or adduct formation. This orientation has been suggested to be involved in guanylyltransferase activity 20, and Lys28 appears to be well positioned to act as a nucleophile. Non-UV based α-32P-GTP labeling experiments examining the formation of DEN MTase-GMP adducts using established guanylyltransferase assays 25 did not detect adduct formation with either the wildtype or K28A DEN MTase proteins (Aaron Shatkin, personal communication). Additionally, the stacked orientation mentioned above was not observed in our lower strength ionic MPD conditions with the same crystal packing lattice (Figure 2) or that of the MVEV MTase 12, raising questions about the physiological relevance of this stacked conformation. Further studies are needed to determine if the stacked conformation indeed plays a role in flavivirus guanylyltransferase activity.
The tight cap binding by MTase (Kd ~100 nM) may represent a mechanism to sequester the RNA cap structure from cellular eIF4E or ribosomes until the positive-strand genomic RNA has become long enough to prevent polyprotein translation from interfering with continued RNA replication. This model would require the guanosine cap to be released from its binding site either by a directed release mechanism (i.e. conformational change) or by folding of an RNA structure that triggers the release of the cap from the MTase. Flavivirus MTase specifically binds GTP (or cap), but binds the structurally similar ATP very poorly. It was initially postulated that the mechanism for guanine specificity lay in the interaction of the guanine C2 amine group with the backbone carbonyls of Leu16 and Leu19 5. While this interaction does provide a small contribution to binding, our analog studies show that the C2 moiety is not the sole factor in guanine specificity. The observation that ITP, which lacks the C2 amine group but possess a hydrogen bond capable hydrogen at N1, can strongly compete with GTP-Bodipy binding indicates that there is at least a second contact between the guanine ring and the MTase. Were the C2 amine position the only selective interaction point, ITP would bind as weakly as ATP. We originally hypothesized that the Lys21 sidechain could interact with the C6 carbonyl group on guanine. However, our mutational analysis indicated that the Lys21 sidechain does not play a role in the observed specificity. Surprisingly, a structurally conserved water molecule is found between the guanine N1 hydrogen and the Leu19 backbone carbonyl in each of our GTP/cap soaked structures derived from completely different crystallization conditions. The presence of this bridging water molecule, which is within hydrogen bonding distance of the guanine N1 hydrogen and the Leu19 backbone carbonyl oxygen, was not observed in previously published lower resolution MTase-GTP structures 5; 9; 10; 11; 12; 13; 14, but is present in a 1.8 Å Den MTase-cap analog structure (PDB Code 2P41) 10. The presence of the water molecule in three different crystallization conditions with two different flavivirus MTase proteins suggests that the coordinated water is not a crystallization artifact and appears to represent an important interaction. The combination of the N1 water-bridge and C2 amine interactions with MTase backbone interactions appear to contribute almost equal amounts of binding energy based on our competition data, and the combination of the two interactions appear to confer specific recognition of guanine largely independent of side-chain composition. These two interactions increase our understanding of guanine binding by flavivirus MTase proteins. This recognition mode is considerably different from that observed in other RNA cap binding proteins (vaccinia VP39, cellular CBP20 and eIF4E) which bind guanosine caps between two aromatic sidechains 24; 26; 27. Water-mediated nucleoside recognition has been reported previously with calf purine-nucleoside phosphorylase 28, although in that case the water interaction was observed between the C6 carbonyl and an Asn sidechain.
The results of the experiments described here further our understanding of the architecture of the MTase guanosine cap-binding site. The flavivirus MTase is a new and potentially valuable antiviral target to develop drugs against flavivirus replication. Drugs that interfere with influenza virus RNA cap binding have been previously described 29, indicating that viral cap binding is a viable drug target. Our increased understanding of how the flavivirus MTase binds RNA cap structures will aid in the rational development of highly potent and specific drugs for the treatment of flavivirus infections.
Recombinant NS5 MTase domains from DEN (strain 16681, AA 1-267), YF (strain 17D, AA 1-268), and WNV (strain NY1999, AA1-268) were cloned into an inducible T7 expression plasmid that contains a carboxy-terminal 6-histidine tag. YF Protein was produced in BL21 (DE3) Codon Plus E. coli cells (Novagen) and DEN and WNV protein was produced in BL21 (DE3) pLysS cells. 750 ml cultures were induced with 400 μM IPTG overnight at 22°C and the bacterial pellets were collect and stored at −20°C in his-tag lysis buffer 30. The frozen pellets were thawed and incubated with additional lysozyme for 1 hr on ice, lysed by sonication, and the lysate was clarified by centrifugation at 18K RPM in a SS-24 rotor. The histidine-tagged protein was purified from clarified lysates using a nickel- loaded chelating sepharose column on an AKTA Purifier FPLC system. The eluted proteins were further purified over a HiTrap–S cation exchange column (Amersham), concentrated using 10K Amicon Ultra concentrators (Millipore), and buffer exchanged into 200 mM NaCl, 20 mM Tris pH 7.5, 0.02% sodium azide and 5mM Tris(2-Carboxyethyl) phosphine Hydrochloride (TCEP-HCl) on a Superdex 200 gel filtration column (Amersham). The isolated proteins were >99% pure as estimated from SDS-PAGE and Coomasie Blue staining. Purified proteins for use in crystallization studies were concentrated using 10K Amicon Ultra concentrators to approximately 400 uM and the concentrations were determined by the absorbance at 280nm using extinction coefficients obtained from the ExPASy web site (http://www.expasy.ch/tools/protparam.html). Purified protein intended for crystallization studies was stored at 4°C and used within 1 week of preparation. Purified protein for use in biochemical studies was supplemented with TCEP-HCl and concentrated to 300 μM, then brought to 20% glycerol and stored at −20°C until use.
Yellow Fever virus NS5 MTase crystals were grown by hanging drop vapor diffusion at 16°C using ~10 mg/ml protein in two different conditions yielding distinct crystal morphologies with identical crystal lattices. The “tapered-rod” morphology grew in ~5 days with a precipitant/well solution containing 25% (v/v) MPD, 0.1 M Tris pH 8.5, 2 mM DTT and 0.02 % (w/v) sodium azide. The “bow-tie” shaped crystals grew in ~10 days with a precipitant/well solution containing 1.6 M sodium citrate pH 6.5. To obtain the ligand-bound structures, crystals were soaked for 2 to 6 hours at 4°C in mother liquor solutions containing ~3 mM GTP or GpppA. The tapered-rod crystals were flash frozen in liquid nitrogen with a final MPD concentration of 25% (v/v), and the bow-tie crystals were briefly dunked in a solution containing 1.6M sodium citrate (pH 6.5) before being frozen. Dengue virus MTase crystals were obtained using a smaller protein construct than that used for the previous structure 5; 11 by deleting the last 31 amino acids that were disordered in that structure (residues 1-267 vs. 1-298). Optimal crystallization conditions were found by screening around the original conditions and resulted in crystals that diffracted to 2.2 Å resolution.
Diffraction data were collected at the MBC beamline 4.2.2 at the Advanced Light Source (Berkeley, CA). Reflections were integrated, merged, and scaled using d*TREK 31, as summarized in Table 1. The initial structure solutions were obtained by using molecular replacement using the program CNS 32 with the Dengue NS5 methyltransferase structure (PDB code 1R6A) as the search model. Manual model rebuilding was performed using O 33 and refined with the CNS package using the MLI target. Refinement parameters for GTP, GpppA, and S-adenosyl-homocysteine were obtained from the HIC-up database 34 and Dundee PRODRG2 server (Schuttelkopf and van Aalten, 2004). Figures were generated with Pymol Molecular Graphics System (www.pymol.org) 35.
Fluorescence polarization (FP) assays were performed with purified MTase protein and GTP-Bodipy γ-phosphate labeled fluorescent analog (Invitrogen Catalog #G22183) as the ligand. All FP experiments were performed in 50 μl volumes in opaque black 384 well microtiter plates (Corning # 3573). Reactions were carried out in 50 mM Tris pH 7.5, 0.01% NP-40, and 10 nM GTP-Bodipy. To perform binding saturation experiments, purified proteins were diluted as 10X concentrates using gel filtration buffer in a 1.5 fold dilution series, and 5 μl diluted protein was added to each well containing 45 μl of the reaction mixture. Plates were incubated at room temperature for 5 minutes and FP and total fluorescence signals were detected using a Victor V multimode plate reader with fluorescence polarization capability (Perkin Elmer). To perform analog competition studies, 11 nM GTP-Bodipy was incubated with 550 nM wildtype DEN MTase protein (~70–80% fractional saturation as determined by binding saturation curves performed with each experiment) in the reaction buffer described above in a volume of 45 μl. 5 μl of diluted competitor was added to the wells (diluting GTP-Bodipy to 10 nM and DEN WT to 500 nM), incubated for 5 minutes at room temperature, and FP and total fluorescence signal was detected. All FP experiments were performed three times in duplicate. Nucleotide analogs used in competition studies were purchased from Jena Bioscience (Jena, Germany).
Coordinates and structure factors have been deposited in the Protein Data Bank with accession numbers 3EVA, 3EVB, 3EVC, 3EVD, 3EVE, 3EVF, and 3EVG.
GTP-Bodipy binding affinity for MTase protein was directly determined by fitting the fluorescence polarization data to equation 1, where f.p.obs is fluorescence polarization observed, f.p.i is fluorescence polarization initial, Δf.p. is change in fluorescence polarization or
fluorescence polarization max minus fluorescence polarization initial and f.b. is fraction bound. Fraction bound (f.b.) is equal to equation 2, where Pt is the total protein
concentration and Kd is the dissociation constant. The affinity of GTP for DEN MTase was determined using a competition based assay and measuring the affinity of GTP-Bodipy in the presence of multiple concentrations GTP. These data were globally fit using Prism (www.graphpad.com) to equation 1 where f.b. is equal to equation 3, where P·GTP is GTP bound protein.
P·GTP is equal to equation equation 4, where GTPt is total GTP added and
KdGTP is the disassociation constant for the GTP. Other inhibition assays were fit using Kalidagraph software package (Synergy Software) and Microsoft Excel with Ki and Kd equations based on 36.
This work was supported by a grant from the Rocky Mountain Regional Center for Excellence (U54 AI-065357) to BG, SK, and OP.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.