Ribonucleoside-5′-diphosphates (NDPs) support RNA polymerase transcription, suggesting NDPs may have been substrates for primordial nucleic acid biosynthesis

A better understanding of the structural basis for the preferences of RNA and DNA polymerases for nucleoside-5′-triphosphates (NTPs) could help define the catalytic mechanisms for nucleotidyl transfer during RNA and DNA synthesis and the origin of primordial nucleic acid biosynthesis. We show here that ribonucleoside-5′-diphosphates (NDPs) can be utilized as substrates by RNA polymerase (RNAP). We found that NDP incorporation is template-specific and that noncognate NDPs are not incorporated. Compared with the natural RNAP substrates, NTPs, the Km of RNAP for NDPs was increased ∼4-fold, whereas the Vmax was decreased ∼200-fold. These properties could be accounted for by molecular modeling of NTP/RNAP co-crystal structures. This finding suggested that the terminal phosphate residue in NTP (not present in NDP) is important for positioning the nucleotide for nucleolytic attack in the nucleotidyl transfer reaction. Strikingly, a mutational substitution of the active-center βR1106 side chain involved in NTP positioning also strongly inhibited NDP-directed synthesis, even though this residue does not contact NDP. Substitutions in the structurally analogous side chain in RB69 DNA polymerase (Arg-482) and HIV reverse transcriptase (Lys-65) were previously observed to inhibit dNDP incorporation. The unexpected involvement of these residues suggests that they affect a step in catalysis common for nucleic acid polymerases. The substrate activity of NDPs with RNAP along with those reported for DNA polymerases reinforces the hypothesis that NDPs may have been used for nucleic acid biosynthesis by primordial enzymes, whose evolution then led to the use of the more complex triphosphate derivatives.

Both NTPs and NDPs possess phosphoanhydride bonds that provide energy to direct biochemical reactions through coupling mechanisms. NTP and their deoxy derivatives are also utilized for nucleic acid biosynthesis by RNA and DNA polymerases, respectively (1). From an energetics standpoint, the choice for NTP substrates over NDPs for enzymatic polymerization reactions by these enzymes is not clear, because the free energy balance for the nucleotidyl transfer reaction for both of these compounds is similar. Thus, hydrolysis of the phosphoanhydride bonds in both NTP and NDP is driven by nearly the same Gibbs energy (2).
Understanding the structural basis for NTP preference by RNA and DNA polymerases helps to define the catalytic mechanisms for nucleotidyl transfer during RNA and DNA synthesis. In our previous study (3), we discovered a new activity of RNAP, phosphorolytic cleavage of nascent RNA transcript, which released 3Ј terminal RNA residues in the form of NDPs. This reaction is functionally analogous to pyrophosphorolysis, which yields NTPs as reaction products (4). Because pyrophosphorolysis is a reversal of RNA synthesis based on nucleotidyl transfer from NTP substrates, we reasoned that NDPs might also serve as substrates for RNA synthesis. Previous reports testing labeled NDPs (5,6) in crude transcription assays revealed some residual incorporation of radioactive nucleotide material into acid-insoluble products in the presence of RNAP and DNA templates. However, it was not clear if this incorporation was because of traces of NTPs in the NDP preparations. In the present study, we used highly purified NDPs to examine their substrate activity with RNAP in an arsenal of biochemical transcription assays and mutational analyses. We find that NDPs can be utilized as substrates for transcription by RNAP, albeit less efficiently than NTP substrates. Previous studies with DNA polymerases (7,8) and reverse transcriptase (9) detected substrate activity of NDPs with these enzymes. Our data, along with the results of these studies, is consistent with the hypothesis that the simpler NDPs may have emerged earlier than NTPs and were utilized as substrates by nucleic acid polymerases at early steps in the origin of life. This is supported by the ability of a currently existing enzyme, polynucleotide phosphorylase (PNP) to utilize efficiently NDP for non-template-encoded RNA synthesis.

NDPs are substrates for Escherichia coli RNA polymerase
We first asked if NDPs could extend RNA primers in ternary elongation complexes (TEC) 2 assembled from RNAP and synthetic oligonucleotides. We constructed TEC containing a 9-nucleotide RNA and then elongated it with the next cognate substrate, [␣-32 P]CTP. The resulting radiolabeled TEC10 was further extended to TEC12, which was used to probe the substrate activity of NDPs. As seen in Fig. 1, incubation of TEC12 with the mixture of four NDPs reduced the amount of the original 12-nucleotide RNA and led to the appearance of longer RNAs. Increasing NDP concentrations yielded progressive increases in RNA lengths (Fig. 1, lanes 2-4). These results demonstrate that NDPs are active RNAP substrates. To examine the template specificity of this RNA synthesis, we challenged TEC12 with various NDPs (Fig. 2, lanes 4 -7). Only the cognate GDP promoted RNA extension, indicating that NDP incorporation is template-specific.

The K m values of NDPs and NTPs are comparable, but the NDP V max is significantly reduced.
As seen from Fig. 1, for both NDP and NTP, the rate of RNA synthesis progressively increased up to a nucleotide concentration 100 M, whereupon a plateau was reached (compare lanes 2-4 with 8 -10 of Fig. 1). This result indicates that the K m values for NDPs and NTPs are comparable. For a more precise determination of the catalytic parameters of these substrates, we assayed them in a single-nucleotide RNA extension assay in TEC10. RNA elongation of TEC10 by ATP proceeds unusually slow, thus obviating the need in fast kinetics devices for K m and V max determination. As seen in Table 1 and Fig. 3, A and D, consistent with the experiment described above (Fig. 1) the K m values of ATP and ADP were roughly equivalent (20 M and 80 M, respectively). Remarkably, a similar correspondence in dissociation constants (K d ) for UDP and UTP from the RNAP active center (5-fold) has been reported (10), confirming that interactions of the NTP ␥-phosphate do not contribute significantly to active center binding. However, as shown in Table 1 and Fig. 3D, the V max was about 200-fold lower for ADP compared with ATP, reflecting a significant contribution of the NTP terminal phosphate to the catalysis. This difference in V max corresponds to a ϳ3 kcal/mol decrease in activation energy for ATP compared with ADP. To provide independent evidence for this calculation, we determined the activation energy for ADP and ATP incorporation by measuring the incorporation rate at various temperatures (Fig. 4). Activation energies for ADP and ATP according to the Arrhenius equation were 21.6 and 17.6 kcal/mol respectively. The difference ϳ4 kcal/mol is in good agreement to that suggested from the difference in V max determined in the previous assay. Notably, the activation energy for ADP in RNAP nucleotidyl transfer (ϳ22 kcal/mol) is close to that for dADP (25 kcal/mol) determined in DNA polymerization with Taq enzyme (8).

Effect of Mn 2؉ ions on the rate of single-step NDP and NTP incorporation
Mn 2ϩ ions increase the rate of RNA synthesis of modified NTPs with poor substrate capacity (11) as well as with natural NTP substrates. Stimulation is more pronounced with a mixture of Mn 2ϩ and Mg 2ϩ ions (11). Moreover, Mn 2ϩ can suppress inhibition by active RNAP center mutations (12)(13)(14)(15). This effect of Mn 2ϩ is attributed to stronger substrate coordination at the active center by the metal ion, which facilitates formation of the reaction transition state. We therefore asked whether Mn 2ϩ can stimulate RNA synthesis with NDPs in lieu of Mg 2ϩ . We examined the effect of these two metal ions alone and together on nucleotidyl transfer with ATP and ADP in TEC10. As shown in Fig. 5, Mn 2ϩ alone, or mixed with Mg 2ϩ , significantly (ϳ2-fold) increased the rate of ATP incorporation. Importantly, we saw no stimulation of ADP incorporation by Mn 2ϩ . This result is explained by our molecular models (see below).

Molecular modeling of the NDP polymerization reaction
Notably, there are no high-resolution structures of RNAP with an NTP substrate bound to the active center in a catalytically competent mode consistent with the S n 2 reaction mechanism (12, 16 -21). Thus in the 1R9S structure of the yeast pol II enzyme (16), the NTP ␤-phosphate does not coordinate with either of the Me 2ϩ ions of the active center, whereas this coordination is seen in all relevant structures of DNAP-NTP com-  NDPs are substrates for nucleic acid polymerases plexes ( Fig. 6, C and D). In the structure of Thermus thermophilus RNAP (22), the coordination of the NTP by active center Mg 2ϩ ions is severely distorted, possibly reflecting a catalytically inactive intermediate state. To explain our results, therefore, we used our previously constructed model (18), in which the NTP triphosphate chain position was adjusted to that seen in the DNAP complexes ( Fig. 6), on the premise that both enzymes operate by a universal S n 2 catalytic mechanism. In this model, the triphosphate chain of the bound NTP makes a set of contacts with two active center Me 2ϩ ions and with side chains of the basic ␤ subunit Lys-1073, Arg-1106 and Arg-678 residues of RNAP (Fig. 6A). This configuration of the triphosphate chain is stabilized by ␣-phosphate coordination with Mg-I, and by ␤and ␥-phosphate coordination with Mg-II, as well as by electrostatic/hydrogen-bonding interaction of ␥-phosphate with ␤Lys-1073, ␤Arg-678, and ␤Arg-1106. In this arrangement the ␣-phosphate group is poised for nucleophilic attack by the 3Ј hydroxyl of the terminal RNA residue (shown as a red sphere in Fig. 6, A and B), so that the directions of the attack and the release of the leaving pyrophosphate group are aligned in the transition state. Next, we built a model for active center-bound NDP by eliminating the ␥-phosphate residue from the previous structure (Fig. 6B). In the absence of the constraints imposed by interactions of the terminal ␥-phosphate residue in the NTP-TEC complex, the ␤-phosphate can readily change its conformation to acquire additional energetically favorable contacts. Thus it can simultaneously coordinate to Mg-II through two oxygens and engage the amino group of ␤Lys-1073 side chain, which would change the side chain conformation seen in the NTP complex to approach the NDP terminal phosphate. This acquisition of new contacts is consistent with the subtle change in K m value for NDP compared with NTP. However, this NDP diphosphate chain rearrangement would misalign the reactant groups (Fig. 6B), and thus adversely affect catalysis. This distortion can explain the dramatic decrease in V max for NDP compared with NTP (see Table 1). The NDP binding model can also account for the different effects of Mn 2ϩ on the NDP and NTP nucleotidyl transfer reactions. Because the orientation of the reactant groups in nucleotidyl transfer is suboptimal with NDP due to the change in ␤-phosphate coordination, it is expected that the reaction proceeds more efficiently in the correct NDP coordination state

NDPs are substrates for nucleic acid polymerases
observed with NTP (Fig. 6A), which can exist in equilibrium with suboptimal orientations. The enhancement of NTP incorporation by Mn 2ϩ , because of the stronger coordination of the attacking 3Ј RNA hydroxyl group and ␣-phosphate group compared with Mg 2ϩ , might be negated by ␤-phosphate coordination with the Mn-II ion. The stronger coordination by Mn 2ϩ would reduce NDP incorporation by shifting the coordination equilibrium toward the substrate suboptimal orientation, thus nullifying enhancement by Mn-I.

Effects of active center mutations on NDP/NTP incorporation
As shown in Fig. 6, ␤Lys-1073, ␤Arg-1106 are the residues closest to the active center-bound NTP/NDP substrates and are likely to participate in catalysis. We therefore investigated the effect of alanine substitutions of these residues on nucleotidyl transfer catalysis. The ␤K1073A mutation increased the K m 2.5-fold for NDP in single-step elongation of TEC10 (Fig. 3, B and D, and Table 1), in accord with the proposed involvement of this residue in NDP retention at the active center (Fig. 6B). The substitution also caused a ϳ100-fold reduction in the maximal reaction rate. Surprisingly, the ␤Lys-1073 mutation enhanced NTP binding ϳ10-fold, while decreasing V max ϳ500fold (Fig. 3, B and D, and Table 1). The reduced K m by the substitution is puzzling and was not previously observed in testing of this mutant with promoter templates (23). We suggest that the reduction in K m might be specific to TEC10. In our previous study (3) we demonstrated that TEC10 is mostly pretranslocated. Thus, the substitution might influence the NTP K m indirectly by affecting RNAP translocation equilibrium. The ␤R1106A substitution moderately increased K m for both ADP and ATP (2.5-and 5-fold, respectively) but reduced the maximal incorporation rate ϳ800-fold for both substrates (Fig.  3, C and D, and Table 1). Altered substrate retention in the mutant enzyme for ATP can be explained by elimination of the

NDPs are substrates for nucleic acid polymerases
salt bridge between the Arg guanidine group and ATP ␥-phosphate. The effect of the substitution on ADP K m could reflect long-range electrostatic interactions between the substrate and the ␤Arg-1106 side chain. Because this side chain does not contact ADP (Fig. 6B), why ␤Arg-1106 reduces the maximal reaction rate of ADP cannot be easily explained. We return to this point below.

Discussion
We find that NDPs can be used, although less efficiently than NTPs, as substrates by RNA polymerase. Together with previous observations (7-9) this finding demonstrates that NDP utilization is a common property of nucleic acid biosynthetic enzymes. Molecular modeling performed in this study and analysis of X-ray structures provide a rationale for our data. Specifically, our model explains why NDP incorporation into RNA is less efficient than NTP incorporation. The terminal phosphate of an NTP substrate makes important interactions with RNAP active center side chains that contribute to substrate retention and pose the substrate for catalysis. This explains the increase in K m and the drastic decrease in V max for NDP relative to NTP. Notably, compared with RNAP, the dNDP/dNTP K m ratio increases more dramatically (Table 1) for DNA polymerases (about 20-fold with RB69 and TaqDNAP and about 500-fold with HIV RT). The striking increase in K m in

NDPs are substrates for nucleic acid polymerases
the case of HIV RT can be explained by extensive interactions of the dNTP ␥-phosphate with HIV RT active center residues Lys-65, Lys-70, and Lys-219 (Fig. 6C). These are lost with an NDP substrate. Molecular modeling of other polymerizing enzymes suggests that some of the active center residues interacting with the NTP ␥-phosphate can also salt bridge to the NDP ␤-phosphate-␤Lys-1073 in RNAP and likely Lys-560 in RB69 DNAP. These contacts could account for the more moderate increase in NDP K m .
The maximal reaction rate with NDP for all polymerases is also slower than that with NTP, which reflects the role of the NTP terminal phosphate interactions in lowering the reaction activation barrier. Thus the strongest V max drop was observed with RB69 DNAP (400-fold) followed by RNAP (140-fold), HIV RT (115-fold), and TaqDNAP (ϳ8-fold). The smaller difference between the dNTP-and dNDP-directed reactions with the Taq enzyme could be because of the high reaction temperature (72°C) used in the assay. In fact, incorporation of dNDP is more temperature-dependent than dNTP incorporation. This is because of the higher activation energy of nucleotidyl transfer from dNDPs compared with that from natural substrates. Thus, the difference between dNTP and dNDP incorporation rates is smaller at higher than at lower temperatures.

NDPs are substrates for nucleic acid polymerases
The relative efficiency of NDP substrates can be expressed as the ratio of the maximal reaction rate to K m . As seen in Table  1, the drop in V max /K m ratio for NDP relative to NTP was about 550-fold for RNAP. This value is comparable with that for TaqDNAP (168-fold) (9), but much less than that for RB69 DNAP (6900-fold) (7), and HIV RT (5.7 ϫ 10 4 -fold) (10).
The effects of substitution of the positively charged active center residues bracketing the nucleotide substrates on the NDP versus NTP incorporation rates provide insight into the mechanism of polymerization catalysis. Thus, alanine substitutions of E. coli RNAP ␤Lys-1073 and ␤Arg-1106 residues strongly reduced catalysis with both NTP and NDP substrates. These residues interact with the ␥-phosphate of NTP. The significant decrease in catalytic efficiency of NTP with the ␤R1106A substitution is less obvious with NDP; this residue does not contact NDP. A reduced catalytic efficiency was observed when the structurally analogous Lys-560 and Arg-482 residues in RB69 DNAP were mutated (see Fig. 6D and Table 1). Similarly, mutations in HIV RT Lys-65, which forms a salt bridge with NTP ␥-phosphate, but is not in a contact with the bound NDP ␤-phosphate (see Fig. 6C), marginally reduces NTP catalysis (9) but completely abolishes NDP incorporation. Notably, the mutant enzyme retained the ability to bind NDP. These findings define a common catalytic factor involved in phosphodiester bond formation by nucleic acid polymerases that is not mediated through immediate contacts of the active center residues with NDP and, possibly, NTP. These residues might play a role in active center dynamics accompanying catalysis that escapes the resolution of crystallographic analysis, or in an intermediate state of substrate binding. Thus, despite the loss of the catalytic performance with NDP compared with natural NTP substrates for all of the tested enzymes, the activity of NDPs is sufficient to support nucleic acid biosynthesis.
Structurally, NDPs are simpler compounds than NTPs and may have been biosynthesized prior to NTPs at earlier steps of evolution. Like NTPs, NDPs are energy-rich phosphoanhydrides. Importantly, the hydrolysis of a phosphoanhydride bond in both di-and triphosphate compounds provides a comparable amount of energy, suggesting that NDPs could efficiently substitute for NTPs in virtually all biochemical reactions that rely on a supply of high energy, including phosphoryl and nucleotidyl transfer.
NDPs substrate activity with RNAP reported here is consistent with the idea that during the development of life on Earth NDPs might have played significant roles in support of many biochemical reactions, including nucleic acid biosynthesis. Notably, NDPs are more resistant to hydrolytic cleavage at high temperature than NTPs, because they contain only one phosphoanhydride bond. This might have been a significant factor in maintaining the pool of these molecules at the elevated temperatures in which first living systems evolved.
The substrate activity of NDPs with single-subunit RNAP of phage and mitochondrial origin has yet to be explored. The possibility of NDP to support RNA synthesis by these enzymes comes from high structural homology of their active center with that of DNAP.
That NDPs could support efficient polymerization stems from the precedent of polynucleotide phosphorylase (24), which releases P i as its reaction product. Unlike PNP, however, nucleotide polymerization with NTPs or dNTPs by RNAP or DNAP is not readily reversible because the pyrophosphate reaction product is rapidly cleaved by inorganic pyrophosphatase, which maintains a low micromolar level of intracellular pyrophosphate concentration (25). In contrast, NDP polymerization yields P i , whose concentration in the cell is high, about 0.01-0.1 M (26,27). Therefore, in the cell PNP readily catalyzes RNA phosphorolysis, the reverse reaction to NDP polymerization, which releases NDP products. These considerations support the idea that the introduction of NTPs in the cellular metabolome might allow more efficient synthesis of nucleic acids by rendering polymerization irreversible.
Polyuridylic acid was from Sigma. Radioactive products were resolved by electrophoresis in 20% PAAG in the presence of 7 M urea and quantified by phosphoimagery using a Molecular Dynamics device (GE Healthcare) and Storm 60 software. Molecular modeling was performed using WebLab Viewer-Light 4.0 (Molecular Simulations Inc.). The experiments described below were repeated three times, unless specified otherwise.

Elongation complexes
TEC9A was assembled as described previously (3). Briefly, RNA9 was mixed with equivalent amounts of DNA template strand to final concentration of 1 M in transcription buffer (TB) (20 mM Tris-HCl, pH 8.0, 0.1 M NaCl, 10 mM MgCl 2 ) at 40°C and left to cool in a water bath to room temperature for ϳ30 min. This mixture was added to an equivalent amount of His 6 -tagged RNAP immobilized on nitrilotriacetic acid (NTA) agarose and kept at room temperature in a shaker for 10 min followed by addition of the equivalent amount of DNA nontemplate strand. The mixture was agitated for another 10 min at room temperature and supplemented with TB containing 1 M NaCl. The pellet was separated and washed with regular TB. If required, the TEC was further elongated to TEC11 by subsequent incubation with 1 M CTP and 50 M ATP (each incubation 20 min, 0°C in TB containing 1 mM MgCl 2 ). Labeled TEC10C, and TEC12C, were obtained by elongation of TEC9A NDPs are substrates for nucleic acid polymerases or TEC11A by an equivalent amount of [␣-32 P] CTP (3000 Ci/mmol, 10 mCi/ml; MP Biomedicals) for 2 min at ambient temperature followed by washing with TB (4 ϫ 1 ml).

Elongation of TEC12 in the presence of NDP and NTP
The elongation reactions (20 min, 20°C) were performed in TB using NDP/NTP concentrations indicated in Figs. 1 and 2. Control mixtures lacked NDP/NTP or MgCl 2 . The reactions were initiated by addition of MgCl 2 and stopped by mixing with equal volume of 10 M urea, 50 mM EDTA, pH 8.0.

Single nucleotide elongation of TEC10C by ADP and ATP
This was performed in TB under the conditions indicated in Figs. 3 and 4. The reactions were started and terminated as described above.

Effect of Mn 2؉ on ATP and ADP incorporation in TEC10C
TEC10C in TB lacking Mg 2ϩ was supplied with 100 M ADP or 10 M ATP and the reaction started by addition of Mg 2ϩ (1 mM), Mn 2ϩ (0.2 mM), or Mg 2ϩ /Mn 2ϩ (1 mM, 0.2 mM, respectively). After incubation at the conditions specified in Fig. 5, the reaction was stopped as described above.