Bacteriophage phi 6 RNA-dependent RNA polymerase: molecular details of initiating nucleic acid synthesis without primer.

Like most RNA polymerases, the polymerase of double-strand RNA bacteriophage phi6 (phi6pol) is capable of primer-independent initiation. Based on the recently solved phi6pol initiation complex structure, a four-amino acid-long loop (amino acids 630-633) has been suggested to stabilize the first two incoming NTPs through stacking interactions with tyrosine, Tyr(630). A similar loop is also present in the hepatitis C virus polymerase, another enzyme capable of de novo initiation. Here, we use a series of phi6pol mutants to address the role of this element. As predicted, mutants at the Tyr(630) position are inefficient in initiation de novo. Unexpectedly, when the loop is disordered by changing Tyr(630)-Lys(631)-Trp(632) to GSG, phi6pol becomes a primer-dependent enzyme, either extending complementary oligonucleotide or, when the template 3' terminus can adopt a hairpin-like conformation, utilizing a "copy-back" initiation mechanism. In contrast to the wild-type phi6pol, the GSG mutant does not require high GTP concentration for its optimal activity. These findings suggest a general model for the initiation of de novo RNA synthesis.

Enzymatic synthesis of nucleic acids can be initiated using two distinct mechanisms. All DNA and some RNA polymerases are strictly primer-dependent. These enzymes add nucleotides to the free hydroxyl group of an appropriate polynucleotide or protein primer. In contrast, most RNA polymerases initiate RNA synthesis de novo, that is without a primer (1,2). In this case, the 3Ј-OH group of the first NTP molecule acts as an acceptor for the second nucleotide. Nucleotidyl transfer is then repeated with the subsequent NTPs giving rise to an extensive RNA product. The de novo initiation may thus require specific molecular interactions to stabilize the initiation complex, because Watson-Crick interactions between the template and individual NTPs may be insufficient to keep them correctly positioned.
Purified 6, 8, and 13 polymerases have been shown to act as replicases and transcriptases in vitro utilizing singlestranded RNA (ssRNA) and dsRNA substrates, respectively (3,4,13). As documented for 6pol, cystoviral polymerases initiate RNA synthesis at the very 3Ј-end of the template employing a primer-independent initiation mechanism (3). HCVpol as well as the related bovine viral diarrhea virus polymerase are also capable of initiation de novo (10, 14 -17). However, under in vitro conditions, these enzymes preferentially utilize a "backpriming" or "copy-back" initiation mode. In this case, the 3Ј-end of the template loops back to form a hairpin structure, which is subsequently extended with the polymerase (5,8,15,18). This type of initiation is obviously deleterious for the virus replication in vivo, because the newly produced daughter strand remains covalently bound to the template strand (2).
High resolution structures of HCVpol and 6pol have been recently determined (19 -22). The structures of the two enzymes are considerably similar (418 of the C␣ atoms of 6pol (665 aa total) can be superimposed on HCVpol with a root mean squared deviation of 3.5 Å). The structural similarity is not expected from the amino acid sequence comparison. This suggests an evolutionary link between the polymerases of the dsRNA viruses infecting bacteria and the positive-sense ssRNA viruses of animals (22). In contrast to many polymerases that have the "open right hand" architecture, with fingers, thumb, and palm subdomains (23)(24)(25)(26), HCVpol and 6pol appear as a cupped right hand with the fingers and thumb strongly interconnected (21,22). Overall, both enzymes appear as compact spherical molecules with internally located active sites, and two positively charged tunnels allowing the access of the RNA template and NTP substrates to the polymerase interior (Refs. 19 -22 and Fig. 1A).
In addition to the apoenzyme, the 6pol initiation complex structure has been solved. This provides a detailed view of the enzyme associated with an oligonucleotide template and two NTPs complementary to the template 3Ј-end (22). This information is not available for the HCVpol. One intriguing feature of the 6pol initiation complex is a chain of stacking interactions encompassing the bases of the two initiatory NTPs, Tyr 630 and perhaps Trp 632 . Both residues are located in the C-terminal loop 630 -633 that has been referred to as the "initiation platform" (Ref. 22 and Fig. 1B). This type of stacking is likely to be preserved in the 8 and 13 polymerases, because both proteins have aromatic residues at the equivalent positions (4,27,28). It has been suggested that Tyr 630 of 6pol could stabilize the NTPs in the process of initiation. Following the initia-tion step, the C-terminal domain containing this loop is believed to move, allowing the exit of the newly synthesized dsRNA product. Interestingly, an analogous structural element containing a tyrosine residue is also present in HCVpol (19 -21) but not in the RdRp subunit of poliovirus (26). The latter enzyme is known to utilize a protein primer to initiate RNA synthesis (29). Furthermore, it has been observed (2) that the critical tyrosine is conserved at least across flaviviral and pestiviral polymerases. The proposed initiation platform (␤-hairpin aa 443-454) of HCVpol has been shortened from LDC-QIYGACYSI to LGGI (30), leading to an increased propensity of the polymerase to initiate from an internally annealed primer, as compared with the wild-type enzyme that could only utilize short primers complementary to the template 3Ј-end. It was concluded that the ␤-hairpin acts as a gate preventing the 3Ј terminus of the template RNA from slipping through the polymerase active site and ensuring terminal initiation of replication. However, the primer-independent initiation mode, crucial for the in vivo initiation, has not been studied in the HCVpol using mutated polymerases.
Here, we use a series of 6pol mutants to address the role of the 630 -633 loop (the tentative initiation platform) in the de novo initiation of RNA replication. Mutations affecting conserved aromatic residues implicated in stacking interactions significantly decrease de novo initiation on 6-specific (ϩ) sense ssRNAs. Nevertheless, the mutants replicate rather efficiently some 3Ј-modified (ϩ) sense RNAs using "back-priming" initiation mechanism, reported earlier for other viral RdRps (5,8,11,15,18,(31)(32)(33). When the initiation platform of 6pol is disordered by the mutation YKW (630 -632) to GSG, back-priming becomes the major mode of initiation. Overall, our results extend the HCVpol data and suggest a model where the de novo initiation is assured by a specialized element of the polymerase polypeptide chain and is further controlled by the secondary structure on the 3Ј-end of viral RNA.

EXPERIMENTAL PROCEDURES
Plasmids-Plasmid pLM659 (34) was used to produce (ϩ) sense ssRNA copies of the small 6 genomic segment (s ϩ ). Plasmid pEM15 containing the 6 s ϩ segment with an internal deletion (13) was used to prepare s⌬ ϩ RNA and its 3Ј-terminally modified variants (s⌬ ϩ 13 and s⌬ ϩ HP ). Plasmid pEM19 was derived from pEM15 by inserting two duplexes of annealed phosphorylated oligonucleotides TL1/TL2 and TL3/TL4 at the XbaI-SacI sites (see Table I for oligonucleotide sequences). The plasmids encoding for the 6pol mutants were derived from the wild-type 6pol expression plasmid pEM2 (3). First, a short fragment of 6pol gene was PCR-amplified using Pfu polymerase (Stratagene) and oligonucleotide seq3_p2 as an upstream primer. The downstream primers p2_Y630F, p2_Y630A, and p2_GSG were designed to introduce corresponding mutations into the initiation platform loop. The PCR products were digested with NruI-NsiI and ligated with the large fragment of the similarly cut pEM2. The resultant plasmids pSJ4 (encoding for the Y630F mutant), pSJ5 (Y630A mutant), and pEM28 (YKW (630 -632) to GSG mutant) were partially sequenced to verify the mutations. Preparation of ssRNA Substrates-Synthetic ssRNAs were produced by run-off transcription in vitro with T7 RNA polymerase (3). Templates for the T7 transcription were prepared by either cutting plasmid DNA with restriction endonucleases or by PCR amplification. RNAs s⌬ ϩ 13 and s⌬ ϩ HP were transcribed from the SmaI cut plasmids pEM15 and pEM19, respectively. The s⌬ ϩ fragment was PCR-amplified from pEM15 using Pfu DNA polymerase and oligonucleotides T7_1 and 3Јend, as upstream and downstream primers, respectively. RNA s ϩ was produced as described by Ref. 3. All ssRNAs were dissolved in sterile water, and the RNA concentration was measured (A 260 ). The quality of each preparation was checked by electrophoresis in 1% agarose gels. 6 Polymerase Assay-Both wild-type and mutated 6 polymerases were expressed in Escherichia coli BL21(DE3) containing the appropriate expression plasmid at 20°C for 15 h and purified to homogeneity as described previously (3). The replication activity of wild-type 6pol and 6pol mutants were typically assayed in 10-l reaction mixtures containing 50 mM HEPES-KOH, pH 7.8, 20 mM ammonium acetate, 6% (w/v) polyethylene glycol 4000, 5 mM MgCl 2 , 1 mM MnCl 2 , 0.1 mM EDTA, 0.1% Triton X-100, 1 mM each NTP (Amersham Biosciences, Inc.), and 0.8 unit/l RNasin. The final concentration of the RNA substrates was 50 g/ml. Unless indicated otherwise, the mixture was supplemented with 0.1 mCi/ml [␣-32 P]UTP (Amersham Biosciences, Inc., 3000 Ci/mmol). Reactions were initiated by adding 6pol protein to a final concentration of 27-270 nM. In the control reactions, a corresponding volume of 6pol control buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.1 mM EDTA) was used instead of polymerase. The mixtures were usually incubated at 30°C for 1 h and treated for further analysis as described below.
Agarose Gel Electrophoresis-Standard agarose gel electrophoresis was used to achieve separation of the positive-sense ssRNA and the corresponding dsRNA segments (3,35). It was carried out in 1.2% agarose or 3% MetaPhor-agarose gels containing 0.25 g/ml ethidium bromide and buffered with 1 ϫ TBE (50 mM Tris borate, pH 8.3, 1 mM EDTA, and 0.25 g/ml ethidium bromide). The 6pol replication reaction was stopped with an equal volume of U2 buffer (10 mM EDTA, 0.2% SDS, 0.05% bromphenol blue, 0.05% xylene cyanol FF, 6% (v/v) glycerol, and 7-8 M urea). For strand-separation gels, the samples were boiled for 2 min and then incubated on ice for 3 min before loading into the gel. After RNA separation (5 V/cm), gels were photographed under UV light exposure, dried on Whatman 3 filter paper or Hybond-Nϩ membrane (Amersham Biosciences, Inc.) followed by autoradiography and/or phosphorimaging (Fuji BAS1500) analysis of the product bands.
RNase I Digestion-6pol replication reactions were assayed as described above but in a 20-l reaction volume. The reactions were stopped after 1 h at ϩ30°C by adding EDTA to a final concentration of 10 mM. NH 4 Ac was added to a final concentration of 0.2 M, and the RNase digestion of the reaction products was initiated by adding 0.2 unit of RNase I (RNase ONE reaction buffer, Promega). An equivalent volume of 1ϫ reaction buffer (RNase ONE) instead of the RNase I was added to the reactions without RNase digestion. After incubation of 1 h at 30°C, the RNase I reactions were stopped with 0.1% SDS and purified by phenol extraction and gel filtration (Amersham G-50) according to the manufacturer's instructions. The reaction products were processed further as described above for strand separation gel electrophoresis.
Primer Extension Assay-The oligonucleotide anti s-117 used in primer extension assays was designed to be complementary (in the 3Ј35Ј direction) to nt 117-136 of the s ϩ segment (36). Primer extension reactions were done according to a previous study (37) with some modifications: 2 l of s ϩ or s⌬ ϩ RNA (1.4 g) was combined with 2 l of the hybridization buffer (0.25 M K-HEPES, pH 7.0, 0.5 M KCl), 0.1 mM EDTA, and 4 pmol of ␥-32 P-labeled primer (T4 polynucleotide kinase, Promega, [␥-32 P]ATP, Amersham Biosciences, Inc., 3000 Ci/mmol) in a final volume of 9 l. Tubes were incubated at 65°C for 1 min and slowly cooled to 30°C over a period of 30 min. The primed RNA templates were mixed with the rest of the 6 polymerase assay components (total volume 20 l) and incubated at 30°C for 1 h. For the control reaction 6pol control buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.1 mM EDTA) was added instead of polymerase. The reactions were stopped by adding EDTA to a final concentration of 10 mM, purified with gel filtration (Amersham Biosciences, Inc., G-50), vacuum-dried, and dissolved into the sample buffer (95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF). Sequencing lanes (A, C, G, and T) were produced with T7 Sequenase 2.0 (Amersham Biosciences, Inc.) from cloned cDNA of the s ϩ segment using the same primer, anti s-117, as in the primer-extension reactions. The primer extension mixtures were incubated at 100°C and sequencing reactions at 80°C for 2 min and analyzed in a 6% polyacrylamide gel containing 7.5 M urea. After electrophoresis, the gels were dried and exposed to phosphorimaging and autoradiography analysis. If the primer anti s-117 was not labeled, [␣-32 P]UTP (Amersham Biosciences, Inc., 3000 Ci/mmol) was added to the reaction in a final concentration of 0.1 mCi/ml. The reactions were stopped by adding EDTA to a final concentration of 10 mM, purified with gel filtration (Amersham Biosciences, Inc., G-50), and U2 buffer was added for analysis in standard or strand separation agarose gel electrophoresis as described above.

RESULTS
The Initiation Platform Mutants Fail to Replicate Efficiently 6-specific (ϩ) Sense ssRNAs-To examine the role of the Cterminal loop 630 -633, three 6pol mutants were constructed and purified ( Fig. 2A). In two of the mutants, Tyr 630 was substituted with either alanine (Y630A) or phenylalanine (Y630F). In the third case, three bulky amino acids YKW (aa 630 -632) were changed to considerably smaller residues GSG (6pol(GSG)). All mutants were expressed and purified accord-ing to the protocol described for the wild-type 6pol (3). RNAsynthesizing activities of the purified enzymes were initially assayed using the 710 nucleotide (nt)-long ssRNA template The RNA concentration was 50 g/ml in all reactions. The dsRNA products (0.71 kb), labeled with [␣-32 P]UTP, were detected by autoradiography. C, the 3Ј-terminal secondary structures of the ssRNA molecules used in this study: s ϩ (single-stranded positive-sense s-segment of 6 phage), s⌬ ϩ (as s ϩ but with an internal deletion), s⌬ ϩ 13 (as s⌬ ϩ but with a 13-nt long addition at the 3Ј-end), and s⌬ ϩ HP (as s⌬ ϩ but with a stable tetra-loop added to the 3Ј-end). The sequences in the boxes are conserved between the three 6 segments (39). s⌬ ϩ 13 allows hairpin formation whereas the s ϩ and s⌬ ϩ are left with a short single-stranded 3Ј-end.  TL1  5Ј-CTAGGGGTTCGC  TL2  5Ј-CGAACCC  TL3  5Ј-CCCGGGTACCGAGCT  TL4  5Ј-CGGTACCCGGGG  T7_1 5Ј 5Ј-GCCATGCATCAGTACCTCGTGGATATTCGCCGAGACATCGGCCTCGGTCCATTTAAACTGGAGTT s⌬ ϩ , which is a plus (ϩ) sense copy of the 6 small genomic segment (s ϩ ) containing an extensive internal deletion (38).
The assays were carried out as specified under "Experimental Procedures" and subsequently analyzed by gel electrophoresis followed by autoradiography. The results revealed significant differences between the wild-type and the mutated 6 polymerases (Fig. 2B). None of the tested mutants could utilize the s⌬ ϩ substrate efficiently, whereas the wild-type 6pol control contained a readily detectable amount of the full-length dsRNA product.
The Initiation Platform Mutants Employ a Back-priming Initiation Mechanism-The platform mutants were also assayed using chimeric ssRNAs templates. One of these templates, s⌬ ϩ 13 , was similar to s⌬ ϩ RNA but contained a 13-nt extension . . . CUAGAGGAUCCCC-3Ј originating from the plasmid polylinker. Both the wild-type and mutated polymerases accepted this template producing a full-length dsRNA product (Fig. 2B). Mutants Y630F and Y630A demonstrated relatively low activity, but 6pol(GSG) mutant replicated s⌬ ϩ 13 more efficiently than the wild-type enzyme. There are two principal explanations for this difference: (i) The mutated polymerases might initiate de novo on the 3Ј terminus of s⌬ ϩ 13 but not s⌬ ϩ . Indeed, it has been shown that the addition of one or several cytosines to the template 3Ј terminus stimulates initiation by the wild-type 6pol (13). (ii) The 6pol mutants might use an alternative initiation mechanism on s⌬ ϩ 13 , which differs from the de novo mechanism of the wild-type enzyme. The 3Ј terminal regions of these two ssRNAs have been suggested to form a tRNA-like secondary structure. In s⌬ ϩ , this tRNA like element contains a 5-nt single-stranded 3Ј tail, which apparently does not form any stable intramolecular base pairs (39). Conversely, the 13-nt longer tail of s⌬ ϩ 13 has a potential to form a transient hairpin structure, which might be used by the polymerase mutants to prime RNA synthesis (Fig. 2C).
Because back-primed synthesis should result in the daughter strand covalently attached to the template, we analyzed heat-denatured RNA polymerization products by agarose gelelectrophoresis. In the case of the de novo initiation, the complementary strands of the duplex RNA molecule would migrate as single-stranded after denaturation. However, if RNA synthesis is initiated by the back-priming mechanism, the hairpinlike dsRNA product should re-anneal immediately after the denaturation step and appear in the gel at the position of dsRNA (Fig. 3A). As expected, wild-type 6pol produced daughter strands, which could be almost completely separated from the s⌬ ϩ 13 template upon heat denaturation (Fig. 3B, compare  lanes 1 and 9). On the contrary, most replication products of 6pol(GSG) mutant could not be converted to the singlestranded form (Fig. 3B, lane 11). In the case of Y630A and Y630F mutants, approximately half of the RNA products appeared as double-stranded after denaturation (Fig. 3C, lanes 25  and 27). This indicates that all three mutants have a substantially increased propensity to generate back-primed RNA products consisting of covalently linked template and daughter strands.
To confirm that the dimer-sized RNA products observed after the denaturation were indeed hairpin-like species, we introduced an RNase digestion control. RNase I of E. coli readily hydrolyzes single-stranded and partially double-stranded RNA but not perfect RNA duplexes (40). The loop at the 3Ј-end of the hairpin product should not be base-paired and is, therefore, RNase-sensitive. As expected, RNase digestion had almost no effect on the wild-type 6pol replication products (Fig. 3B,  lanes 1, 5, 9, and 13). In the case of mutated polymerases, RNase digestion converted heat-resistant double-stranded products into the single-stranded form (Fig. 3, B and C, lanes  15, 29, and 31).
Thus, on the s⌬ ϩ 13 ssRNA template, the GSG mutant utilizes predominantly back-priming initiation mechanism, whereas Y630A and Y630F mutants use de novo and backpriming initiation modes with nearly equal efficiencies.
Kinetic Analysis of de Novo Initiation-Because the wildtype 6pol and the mutants can use s⌬ ϩ 13 template for de novo initiation, it was possible to compare enzymatic constants of the four 6pol variants for the initiatory nucleotide (GTP). For this purpose, polymerization reactions were carried out with an excess of s⌬ ϩ 13 template using variable concentrations of [␥- 32  into RNA chains initiated de novo but not into back-primed products (Fig. 4A). This allowed us to selectively detect the de novo initiated dsRNA products. Accumulation of dsRNA products reaches steady state within ϳ0.5 min and is linear for at least 5 min for all four polymerases (not shown). We therefore measured the amount of dsRNA products produced within the first 5 min as an approximation for the initial velocity of de novo initiation. The velocities, thus determined, were plotted as a function of GTP concentration for all 6pol variants (Fig. 4B). The resultant curves are distinctly S-shaped, which indicates a cooperative binding of the initiatory GTPs. Consequently, the data were fitted to the Hill equation, V 0 ϭ V max [S] H /(K 50 H ϩ [S] H ), where V 0 is the initial velocity; V max , the maximal velocity; [S], initial substrate concentration; H, Hill cooperativity coefficient; and K 50 , the half saturation constant, using nonlinear regression software EZ-Fit (Perrella Scientific, www. JLC.net/ϳfperrel). The apparent K 50 and relative V max values calculated from the regression analysis are presented in Fig.  4C. The Hill coefficient was ϳ2 for all polymerases. It is obvious that, compared with the wild-type, all three mutants have increased K 50 and decreased V max values.
The GSG Mutant Polymerase Accepts ssRNA Substrates with a Preformed Hairpin Loop at the 3Ј Terminus-Because the 6pol platform mutants were capable of back-priming on the 3Ј-end of the template, we further tested whether the mutants could initiate on the 3Ј-end containing a pre-formed hairpin. For this purpose, a very stable hairpin structure . . . CUAGGGGUUCGCCCC-3Ј containing the UUCG-tetraloop was introduced to the s⌬ ϩ ssRNA. The length of this template (s⌬ ϩ HP ) was 725 nt (Fig. 2C). Agarose gel electrophoresis of the 6pol reaction products shows that the wild-type 6pol accepts s⌬ ϩ HP at least one order of magnitude less efficiently than the s⌬ ϩ 13 template. Most of the products migrated as singlestranded RNA after heat denaturation (Fig. 3B, lanes 10 and  10A). Conversely, 6pol(GSG) readily utilizes s⌬ ϩ HP RNA template producing hairpin-like RNA molecules. Y630A and Y630F failed to initiate on the s⌬ ϩ HP template (Fig. 3C, lanes 18 and  20), thus suggesting that the phenotype of these two mutants is intermediate between the wild-type and GSG.
6pol(GSG) Does Not Require Increased GTP Concentration for Optimal Activity-The initiation mechanism of 6pol was further studied by titrating NTP concentrations in the replication reactions and analyzing the reaction products. In contrast to the data presented in Fig. 4, [␣-32 P]UTP label was used in this experiment allowing one to detect products of both de novo initiation and back-priming. Reduction of the GTP concentration from 1 to 0.1 mM in the replication reaction mixture containing s⌬ ϩ 13 ssRNA template significantly decreased the level NTPs, 50 g/ml s⌬ ϩ 13 ssRNA template and 270 nM of either wild-type 6pol, GSG, Y630A, or Y630F were incubated at 30°C for 5 min and analyzed using standard gel electrophoresis. Intensities of dsRNA product bands were quantified with phosphorimaging (Fuji BAS1500). The velocity of the RNA synthesis was estimated, and the graphs were normalized so that the V max of the wild-type 6pol (WT) is set to 1. C, the results from panel B were fitted to Hill equation using non-linear regression software EZ-Fit and the apparent K 50 in mM and relative V max (rV max ) are shown in the table.
of the wild-type 6pol-directed RNA synthesis (Fig. 5A, compare lanes 1 and 5). This is not surprising, because GTP is used here as a priming nucleotide for initiation de novo. Phosphorimaging analysis revealed that the GTP concentration sufficient for a half-maximal level of RNA synthesis was ϳ0.2 mM, as measured with the s⌬ ϩ 13 template (Fig. 5B). The overall RNA synthesis was detected by the non-denaturing agarose gel analysis. In contrast, the 6pol(GSG) mutant was less dependent on the GTP concentration with either the s⌬ ϩ 13 or the s⌬ ϩ HP ssRNA templates (Fig. 5A, compare lane 3 with 7 and lane 4 with 8). In the case of the s⌬ ϩ 13 template, the GTP concentration for a half-maximal synthesis by the GSG mutant was only ϳ0.02 mM (Fig. 5B). The requirements of the two enzymes do not differ in respect to the other NTPs (Table II).
6pol(GSG) Can Utilize Oligonucleotide Primers-The platform mutants were also tested for their ability to extend a complementary oligonucleotide primer. For this purpose, 5Јlabeled primer (anti s-117) complementary to nt 117-136 of the 6 s ϩ segment (2948 nt long) was annealed to the corresponding ssRNA, and this primed template was used in wild-type and mutant 6pol assays. The reaction products were analyzed in a denaturing polyacrylamide gel along with the sequencing lanes produced with the same anti s-117 primer and cDNA of the of s ϩ segment (Fig. 6). The reaction with 6pol(GSG) resulted in ϳ20-fold more efficient initiation from the primer than the wild-type 6pol reaction (Fig. 6, compare lanes 1 and  2). 6pol mutants Y630A and Y630F were able to initiate from the primer but clearly not as efficiently as the GSG mutant. Mutant Y630A was more prone to primer-dependent initiation than Y630F. In the control reaction no RNA synthesis occurred on the primed template incubated with buffer instead of 6pol (Fig. 6, lane 5). The adjacent sequencing lanes show that the primer extension products were full-length, initiated accurately from the primer (Fig. 6, lanes A, C, G, and T).
The following experiment was carried out to estimate the efficiency of the 6pol(GSG)-directed primer extension reaction compared with the wild-type-directed de novo initiation. Unlabeled anti s-117 primer was annealed to the s⌬ ϩ ssRNA template, and this was used in the 6pol reactions containing [␣-32 P]UTP (Fig. 7A). As apparent from the autoradiogram, the primer does not affect the wild-type 6pol-catalyzed reaction, but in contrast, it does change the 6pol(GSG) product pattern dramatically (Fig. 7, lanes 1, 3, 5, and 6). As expected, 6pol(GSG) is unable to utilize the s⌬ ϩ ssRNA template (Fig. 7, lane 4, see also Fig. 2B); however, in the presence of the primer 6pol(GSG) it gives rise to partially double-stranded RNA products. It is evident that the amount of the primer-extended product in the case of the GSG mutant is actually higher than the amount of the full-length dsRNA product initiated de novo by the wild-type 6pol (compare lanes 2 and 3 in Fig. 7). DISCUSSION This report provides direct experimental insights into the mechanism of de novo initiation of RNA-dependent RNA polymerization. The high resolution structure of the 6pol quaternary complex with a template and two NTPs provides an excellent ground for biochemical studies on the primer-independent initiation mechanism of RNA synthesis. Based on the structural data, the C-terminal platform (aa 630 -633) of 6pol forms stacking interactions with the initiatory NTPs, thus suggesting that this structural element might be critical for the de novo initiation (Ref. 22 and Fig. 1). This idea is further supported by the fact that a similar polypeptide loop is present in the HCVpol, which is also capable of initiation de novo (19 -21). Furthermore, this element is absent from the polioviral RNA-dependent RNA polymerase of poliovirus that is strictly primer-dependent (26,29). The recent work by Hong et al. (30) addressed the role of the C-terminal ␤-hairpin (aa 443-454) of HCVpol in the terminal initiation, but its role in the de novo initiation remains elusive.
Here, the platform was minimized by changing Tyr 630 -Lys 631 -Trp 632 to GSG. Alternatively, point mutations were introduced that only affected Tyr 630 , immediately involved in stacking interaction with the initiatory NTPs (mutants Y630A  and Y630F). As predicted, these platform mutants were found to be inefficient in the initiation on 6-specific (ϩ) sense ssRNA. The impaired de novo initiation of Y630F is somewhat surprising, because this mutant only lacks the phenolic hydroxyl compared with the wild-type enzyme. We notice, however, that this OH group is located close to the side-chain carboxyl of Asp 624 , suggesting that this potential hydrogen bonding is somehow important for an adequate stacking interaction with the incoming nucleotide. This conclusion is further supported by the presence of Asp 624 in the polymerases of 6-related viruses 8 and 13.
However, the mutated enzymes were fully functional with some chimeric ssRNAs (Fig. 2B). We noticed that the 3Ј-end of these ssRNA (s⌬ ϩ 13 ) can fold back and form a hairpin structure, which may allow the polymerase to utilize the backpriming initiation mode. To address this hypothesis, we heatdenatured the replication reaction products before gel electrophoresis and confirmed the results using RNase digestion. Unlike the dsRNA products of the wild-type 6pol, which could be converted to the single-stranded form by heat denaturation, most of the RNA species produced by the 6pol mutants migrated as dsRNA even after extensive boiling. Only after the RNase I pretreatment, the mobility of these species was shifted to that of ssRNA thus confirming the hairpin-like nature of the replication products (Fig. 3, B and C). The 6pol(GSG) mutant clearly prefers back-priming initiation mode to the de novo initiation whereas Y630A and Y630F mutants can use both initiation mechanisms with almost equal efficiencies. Kinetic analysis of the de novo initiation on the s⌬ ϩ 13 ssRNA template corroborates the idea that stacking interaction between the aromatic side chain of Tyr 630 and incoming nucleotide (GTP) stabilizes the initiation complex. Interestingly, when compared with the wild-type, mutations in the platform loop both increase K 50 and decrease V max for the de novo initiation (Fig. 4). K 50 of the Y630F is closer to the wildtype than Y630A and GSG, consistent with the notion that the phenylalanine side chain can still stack against GTP base. All mutants are characterized by substantially reduced relative V max values, GSG having somewhat higher value than the point mutants. Additional studies are clearly needed to rationalize this dramatic decrease in V max . The calculated Hill coefficient is ϳ2 for all four polymerases, which probably reflects cooperative binding of two GTP molecules to the initiation complex. This observation is supported by our previous structural data (22).
Several laboratories have previously reported that HCVpol can initiate RNA synthesis in vitro using back-priming mechanism, in addition to the de novo initiation mode (5,15,18). In fact, studying de novo initiation by HCVpol requires specific measures to be taken to prevent back-primed initiation. Several other viral RdRps that also utilize back-priming initiation in vitro have been described (8,11,(31)(32)(33)41). Because hairpinlike products resulting from this reaction must be deleterious for viral replication, back-priming is unlikely to be a genuine initiation mechanism used in vivo. Our results indicate that the shift from de novo initiation mechanism of wild-type 6pol to the back-priming mechanism is caused by the modifications at the C-terminal initiation platform. In the case of HCVpol, the bias toward back-priming mode might be a consequence of using soluble forms of the enzyme in in vitro experiments, whereas in vivo HCVpol is associated with intracellular membranes using its C terminus as an anchor (9).
In addition to the back-primed RNA synthesis, the 6pol(GSG) mutant can extend a complementary oligonucleotide primer annealed to the template (Fig. 6). At least in the case of anti s-117 primer, the primer-dependent initiation of 6pol(GSG) is even more efficient than de novo initiation of the wild-type enzyme as shown in Fig. 7B. Primer-dependent initiation of the 6pol(GSG) is consistent with the results on the FIG. 8. Schematic model for the de novo and back-priming initiation mode of 6pol. The 6 polymerase is shown as green, the oligonucleotide template coming via template tunnel is blue, and the nucleotides are red. The metal ions are shown in black. The initiation platform (P) of the wild-type 6pol (WT) forms stacking interactions with the incoming nucleotides and stabilizes the initiation complex in de novo RNA synthesis. Importantly, the platform prevents the backpriming of the template shown for the 6 polymerase mutants and does not allow utilization of RNA substrates with preformed loop at the 3Ј-end. The mutated platform allows the 6 polymerase to initiate via back-priming mechanism (GSG).

FIG. 7.
Primer-dependent polymerization activity of 6pol(GSG) compared with the primer-independent wild-type 6pol. A, a schematic drawing to describe the primer-extension with unlabeled primer annealed to the template (gray). Radioactive nucleotides were used in the reaction, and therefore the newly synthesized RNA strands are labeled (black). The reaction products of wild-type and 6pol(GSG) were heat-denatured. B, standard agarose electrophoresis of wild-type 6pol (WT) and GSG mutant (GSG) replication reactions in the presence of anti s-117 primer annealed to s⌬ ϩ ssRNA template (lanes 1, 3, 5, and 6) or template without primer (lanes 2 and 4) carried out as described under "Experimental Procedures." The reaction products in lanes 5 and 6 were heat-denatured and analyzed in 3% Meta-Phor agarose gel electrophoresis. The size markers are shown on the left and right for both gels independently.
HCVpol C-terminal ␤-hairpin mutant (30). However, it is not clear how mutated polymerases with completely encircled active sites access an internally bound primer. It is unlikely that 6pol(GSG) conformation is more open than that of the wildtype enzyme, because mobility of the two proteins on a sizing column (Superdex 200) under native conditions is undistinguishable (data not shown). Two possibilities appear plausible: (i) the template threads through the enzyme and the polymerase pauses at the annealed primer, or (ii) occasional fraying of the fingertips allows the polymerase to bind in the middle of the template.
The model for de novo initiation of RNA-dependent RNA polymerization, which emerges from our data, is summarized in Fig. 8. In this model, the specific C-terminal platform loop (aa 630 -633) protruding into the active site forms stacking interactions with the first two incoming nucleotides, thus stabilizing the initiation complex. Furthermore, the same structural element prevents abnormal back-primed initiation, which otherwise would lead to the accumulation of inactive hairpinlike dsRNA products. This type of negative control is only possible when the platform loop is bulky and rigid as it is in the wild-type 6pol and HCVpol. Hence, mutations introduced to the C-terminal region of the enzyme may promote the backpriming initiation mode.
We further notice that, at least in the case of 6 (ϩ) RNAs, the native secondary structure of the 3Ј-end does not favor the formation of back-primed intermediates. The terminal portions of all three segments are folded in a tRNA-like structure, with the five 3Ј-proximal nucleotides in a single-stranded form. This 5-nt terminus is sufficient to span the template channel of 6pol (22) but is not sufficient to loop back. However, when short 3Ј-terminal extensions are added to the RNA, the backprimed conformation will become plausible, as observed (Fig.  2C). This suggests that the 6-specific 3Ј-end secondary structure might be a result of evolutionary selection to assure accurate de novo initiation. Because many RNA genomes have compact secondary structures at their 3Ј termini, this evolutionary trend might be common for different viral families. Intriguingly, 6 RNA segments lacking the 3Ј-terminal secondary structure can be replicated both in vitro and in vivo. However, in the latter case the native tRNA-like 3Ј-end is soon regained using heterologous recombination (39).
In contrast to the wild-type 6pol, the GSG mutant, deficient in the primer-independent initiation, does not require high concentrations of GTP. This suggests that the elevated GTP concentration is associated with primer-independent but not with primer-dependent initiation (Fig. 5). An elevated concentration of the initiatory purine nucleotides is also required for other RdRp subunits (14,17,42,43). Moreover, both singlesubunit RNA polymerases of DNA phages and complex cellular DNA-dependent RNA polymerases usually initiate RNA synthesis with either GTP or ATP (44), nucleotides known to have high stacking propensities.
In conclusion, the mechanism of primer-independent initiation of RNA polymerization is dissected for the polymerase of bacteriophage 6. We propose that efficient and accurate de novo initiation of RNA-dependent RNA synthesis is controlled by three factors: (i) The polymerase stabilizes the initiation complex using its primer-mimicking loop; the same loop also interferes with the template back-priming. (ii) Viral RNAs possess specific secondary structures at their 3Ј-ends that dis-favor the back-primed conformation. (iii) The de novo initiation complex is further stabilized by high concentration of initiation nucleotides.