Highly effective viral RdRP translocation hindrance by the remdesivir nucleotide Authors: Rui Li, Qiang Shao, Meihua Wang, Chunyan Feng, Qin Xu, Qisheng Wang, Luqing Shang, Yechun Xu, Peng Gong Published: 2026 ABSTRACT AbstractNucleoside/nucleotide analogs (NAs) typically interfere with substrate selection and catalysis steps of polymerase nucleotide addition cycle (NAC), thereby affecting polymerase catalytic efficiency. In contrast, their interference with post-chemistry translocation is rarely considered in rational design, partly due to a lack of direct and quantitative evidence for such interference. Remdesivir (RDV) is an adenosine analog with broad-spectrum antiviral activity against diverse RNA virus families and has been approved for COVID-19 treatment. RDV is capable of inducing a characteristic “delayed” intervention, wherein the interference occurs several NACs after its incorporation. By combining X-ray crystallography, quench-flow kinetics, and molecular dynamics simulations, we show that the composite rate of RDV translocation from product strand position −3 to −4 and subsequent catalysis is at least 100-fold slower than that of adenosine in an enterovirus RdRP. However, mutating the corresponding roadblock residue serine417 to glycine largely abolishes this difference. Nonetheless, the incorporated RDV nucleotide can still alter the free-energy landscape even in the S417G mutant complex, enabling the capture of an over-translocated state. Our work demonstrates translocation-related intervention as an effective and broadly applicable polymerase intervention strategy in NA rational design. Introduction Nucleoside/nucleotide analogs (NAs) are an important class of compounds exhibiting both high effectiveness and great broad-spectrum potential in antiviral and anti-cancer studies [1, 2]. Traditional NA intervention modes include immediate polymerase chain termination and virus error catastrophe through NA-mediated mutagenic effect [3, 4]. In anti-RNA virus NAs, the first mode is represented by sofosbuvir, a uridine analog approved for hepatitis C treatment with characteristic modifications at ribose 2′-position [4], while the second mode is represented by favipiravir, a pyrazine carboxamide derivative that is capable of base-pairing with both cytidine and uracil [5]. The base-pairing tolerance of favipiravir and NAs with similar feature primarily accounts for their mutagenic effect [5, 6]. In the early phase of the coronavirus disease 2019 (COVID-19) pandemic, remdesivir (RDV), an adenosine analog known for its effectiveness against Ebola virus and coronaviruses [7–9], and later also for inhibition of picornaviruses including enterovirus 71 (EV71) [10–12], was found to cause an “i + 3” delayed intervention on coronavirus RdRPs including that of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [13, 14]. For clarity, here we use RDV-TP and RDV-MP to represent the substrate form (nucleoside 5′-triphosphate) and incorporated form (nucleoside 5′-monophosphate) of RDV, respectively. This intervention mode is featured by the hindrance of the fourth nucleotide addition following RDV-MP incorporation, leading to the accumulation of the “i + 3” product, where “i” denotes the product formed on RDV-MP incorporation. Such a unique mechanism was proved to be related to the interaction between the RDV nucleotide and an RdRP thumb domain serine residue (S861 of SARS-CoV-2 nsp12), as mutating this residue to the less bulky alanine reduced the level of the specific intervention [15]. Both single-molecule- and enzymology-based studies further suggest that SARS-CoV-2 RdRP is able to escape such an intervention under certain circumstances [16, 17]. Therefore, the RDV nucleotide is neither an immediate nor a complete chain terminator. However, how effective is RDV nucleotide at the delayed intervention site and what is the precise mechanism of its primary intervention both remain to be clarified. Hypothetically, residue S861 may both hinder the translocation of RDV-MP from product strand position −3 to −4 and destabilize RdRP complex with an RDV-MP at position −4. However, structural capture of RDV-MP at product position −4 or in between positions −3 and −4 has not been successful [15, 18]. In a comparative study of RdRPs from SARS-CoV-2 and EV71, a very similar “i + 3” delayed intervention was observed in EV71 RdRP, and a structurally equivalent serine residue (S417) was also found to be responsible for the intervention, as mutating this residue similarly affected the intervention level [16]. Hence, EV71 RdRP can also serve as a model system for dissecting the “i + 3” intervention mechanism of the RDV nucleotide. In this work, we show that EV71 RdRP S417G mutant enables crystallographic capture of RDV-MP at product position −4 of an RdRP elongation complex (EC). EC crystal soaking further allows the observation of a pre-translocation state with an RDV-MP at position −3 and an over-translocated state with an RDV-MP poised between positions −4 and −5, both unfavorable in ECs derived from regular nucleotides, suggesting that RDV nucleotide can reshape the free-energy landscape even in the absence of the roadblock serine side chain. Enzymology data reveal that RDV-MP incorporation rate is only several-fold slower than that of adenosine monophosphate (AMP), whereas the composite rate of RDV-MP translocation from product position −3 to −4 and subsequent catalysis is at least 100-fold slower. Molecular dynamics (MD) simulations suggest that RDV-MP translocation is impeded at product position −3, while introducing the S417G mutation allows the translocation toward position −4, indicating that RDV nucleotide translocation from product position −3 to −4 is hindered by the S417 side chain. Taken together, our work suggests that translocation-related intervention is an underdeveloped but important option in rational design of ribose-modified NA drugs. Materials and methods Plasmid construction and protein production A pET26b-Ub vector-based plasmid containing the wild-type (WT) EV71 3Dpol (RdRP) (strain-SK-EV006-LPS1 (strain-SK) or strain BrCr-ts) gene was used for production of WT RdRP. Note that a C291M mutation was introduced into the strain-SK plasmid for optimal growth of EC crystals, and this mutant is referred to as “WT” for simplicity [19]. The strain-SK plasmid was used as the cloning template to construct mutant plasmids following previously described methods [20–22]. Cell growth, isopropyl-β-d-thiogalactopyranoside (IPTG) induction, cell harvesting, cell lysis, protein purification, and protein storage were performed as described previously [19, 20]. Briefly, the protein was expressed in Escherichia coli strain BL21(DE3)pCG1, induced by 0.5 mM IPTG, purified by sequentially passage through nickel-affinity, anion exchange, and gel filtration columns. RNA preparation and EV71 RdRP EC assembly The 34-mer (T34-1, T34-2) and 33-mer (T33) template RNAs were prepared by T7 RNA polymerase-glmS ribozyme-based approaches described previously [23, 24]. The 10-mer RNA primer (P10) was chemically synthesized (Integrated DNA Technologies). RNA construct assembly and EC assembly, purification by anion-exchange chromatography, and storage were performed as previously described [19]. EC crystallization and EC crystal soaking trials EC crystallization trials were performed as described previously [20]. Briefly, crystals were grown by sitting-drop vapor diffusion at 12 °C or 16 °C using 10 mg/ml EC samples. Strain-SK S417G mutant RdRP EC crystals grew in a precipitant solution containing 0.17 M ammonium sulfate, 0.213 M MES (pH 6.5), 25.5% (wt./vol.) PEG5000 monomethyl ether, and 15% (vol./vol.) glycerol. Strain-BrCr-ts WT RdRP EC crystals grew in a precipitant solution containing 1.6 M ammonium sulfate, 0.08 M sodium acetate (pH 5.0), and 20% (vol./vol.) glycerol. Crystal soaking trials were performed using the precipitant solution supplemented with 5 mM of each nucleotide [NTP(s) or RDV-TP] and 10 mM divalent metal ions (MgCl2 or MnCl2). Crystals were directly cooled and stored in liquid nitrogen prior to X-ray diffraction data collection. Crystallographic data collection, processing, and structure determination X-ray diffraction data were collected at Shanghai Synchrotron Radiation Facility beamlines BL17U1 and BL02U1 (wavelength 0.9792 Å) at 100 K. At least 180° of data were collected in 0.2° oscillation steps. Reflections were integrated, merged, and scaled using HKL2000 (Supplementary Table S1) [25]. The initial structure solution was obtained using the molecular replacement program PHASER [26] with coordinates derived from an EV71 EC structure (Protein Data Bank (PDB) entries 5F8G, chains A–C) as the search model [20]. Manual model building and structure refinement were carried out using Coot and PHENIX, respectively [27, 28]. The 3500-K composite simulated-annealing omit 2Fo − Fc electron density maps were generated using CNS software. Unless otherwise indicated, protein structure superimpositions were performed using the maximum likelihood-based structure superpositioning program THESEUS [29]. The single-nucleotide elongation assay using a quench-flow instrument To measure single-nucleotide incorporation rate of RDV-TP, a rapidly mixing/quenching experiment was performed using an RQF-3 chemical quench-flow instrument (KinTek Corp.). T33 was annealed with P10 at a 1.1:1 molar ratio to generate the T33/P10 construct. EC was assembled using T33/P10 and EV71 RdRP (strain-SK) on incorporation of a CUCU tetra-nucleotide with CTP and UTP as the only NTP substrates (20 μM each, 2 h), yielding a 14-mer (P14)-containing EC (EC14). Unincorporated NTPs were then removed by multiple rounds of concentration and resuspension using an ultrafiltration tube. In a three-syringe quench-flow experimental setting, the reaction was initiated by the mixing of EC14 and ATP/RDV-TP solutions at 22.5 °C. The reaction was quenched by a stop solution containing 95% (vol./vol.) formamide, 20 mM EDTA (pH 8.0), and 0.02% (wt./vol.) xylene cyanol, at various time points (in the range of 5 ms to 10 s). The final reaction mixture contained 50 mM HEPES (pH 7.0), 30 mM NaCl, 50 mM KCl, 5 mM MgCl2, 5 mM TCEP, 4 μM RNA (corresponding to the concentration of T33), 300 μM ATP or RDV-TP, and 6 μM RdRP. The RNA species were analyzed by denaturing (7 M urea) 20% (wt./vol.) polyacrylamide gel electrophoresis (PAGE), followed by Stains-All staining (Sigma-Aldrich), and band intensity quantification was performed using ImageJ (https://imagej.net/ij) as previously described [24]. To estimate the single-nucleotide elongation rate (r) corresponding to P14-to-P15 conversion, the values representing the fraction of 15-mer intensity (f) at all time points (t) were fitted to a single-exponential rise equation: f = amplitude[1 − exp(−r × t)]. The two-nucleotide elongation assay using a quench-flow instrument The experimental settings were identical to those of the single-nucleotide incorporation assay except that T34-2/P10 was used for EC14 assembly using GTP and ATP/RDV-TP (300 μM each), the quench-flow-based reaction duration range was from 5 ms to 120 s and CTP, ATP (80 μM each) were used for the elongation reaction. The medium concentration (80 μM) was chosen to ensure sufficiently rapid catalytic rates while minimizing perturbation by regular NTP-induced misincorporation events. Fitting routines to generate the r1 and r2 values were carried out using an irreversible two-step consecutive reaction model and the Global Kinetic Explorer software [30]. Molecular simulation system preparation of EV71 RdRP-RNA complexes The atomic coordinates of EV71 WT RdRP_AMP, WT RdRP_RDV-MP, and the S417G mutant complexes were derived from experimentally determined crystal structures described above. For EV71 RdRP, missing residues were reconstructed and refined using the cyclic coordinate descent and kinematic closure protocols in Rosetta v.3.10 [31, 32]. The protonation states of all titratable residues were evaluated at pH 7.5 using Schrödinger suite. All evaluated residues were found in their standard protonation states. A detailed inspection of the environment surrounding each histidine residue revealed that all histidines were neutral, among which H31, H193, H199, H271, H273, H274, and H390 were protonated at the Nδ position and the remaining histidines were protonated at the Nε. For RNA, both template and product strands were truncated to retain the region from −7 to +1. The +2 template nucleotide was excluded because it forms interactions with the RdRP index finger [33], and could bring complexities in the simulations. Each simulated system was solvated in a cubic box filled with a total of ∼23 000 water molecules, with multiple Na+/Cl− ions added to neutralize the protein and RNA charges. AMBER 22 suite [34] was employed for simulations using FF14SB [35], χOL [36], and TIP3P [37] force fields for protein, RNA, and water molecules, respectively. RDV was modeled using the generalized AMBER force field (GAFF) [38] with restrained electrostatic potential partial charges [39] and covalent linkages at the +1 or −3 position of the RNA product strand according to the AMBER tutorial (https://ambermd.org/tutorials/basic/tutorial5/index.php). Gaussian accelerated MD simulations Each aforementioned system was initially minimized for 50 000 steps and heated to 300 K, with protein and RNA heavy atoms being fixed using a harmonic restraint with a force constant of 10.0 kcal·mol−1·Å−2. Subsequently, the protein was relaxed by two short-time steps of equilibrium at constant temperature and constant pressure of 1 atm (NPT ensemble): 2 ns for relaxing the protein side chain and 2 ns for the protein main chain. Finally, a total of four production runs implemented with Gaussian accelerated MD (GaMD) [40] were performed without constrains on any system atoms using the graphical processing unit version of AMBER 22, with randomized initial atomic velocities. The SHAKE algorithm was used to fix all covalent bonds involving hydrogen atoms, and periodic boundary conditions were applied to avoid edge effects [41]. The Particle Mesh Ewald method was applied to treat long-range electrostatic interactions and the cutoff distance for long-range terms (electrostatic and van der Waals energies) was set at 10.0 Å [42]. The Langevin dynamics with a collision frequency of 3.0 ps−1 was adopted to control the temperature. All GaMD production runs were performed at the “dual-boost” level by setting the reference energy to the lower bound, applying one boost potential to the total potential and the other to the dihedral energetic term. The average and standard deviation (SD) of the system potential energies were calculated every 300 000 steps (0.6 ns). The upper limit of the boost potential SD was set to 6.0 kcal· mol–1 for both the dihedral and total potential energetic terms. The coordinates were saved every 10 000 steps for subsequent data analysis. The free-energy landscape on specific collective variables was calculated with the accompanying reweighting algorithm of GaMD. Results Crystallographic capture of RDV-MP at and adjacent to product position −4 With an aim to structurally observe RDV-MP at the product strand position −4 and to understand its impact around this position, we designed an RNA construct comprising a 34-nucleotide (nt) template (T34-1) and a 10-nt primer (P10). This construct allows EV71 RdRP to incorporate a GRGRG pentanucleotide, yielding a 15-nt product (P15)-containing EC (Fig. 1a, R denotes RDV-MP). If the RdRP reaches the post-translocation state, as observed in most previously determined enterovirus RdRP native EC structures [33], the first RDV-MP incorporated will reach position −4, thus allowing structural observation of its interaction with residue 417. When WT RdRP (S417) was used for assembly, a very limited amount of EC was obtained at the high-resolution anion-exchange chromatography step (Supplementary Fig. S1a), not allowing subsequent EC crystallization and structural determination. This observation suggests that either the complex assembly is inefficient, or the assembled RdRP-RNA complex is not as stable as those assembled through incorporation of regular nucleotides. We previously showed that mutating S417 to threonine, alanine, or glycine (S417T/S417A/S417G) can reduce the “i + 3” intervention level of RDV-TP [16]. Hence, using these mutants for complex assembly may increase the yield of the complex. However, only S417G mutant yielded adequate amount of complex for structural studies, while S417T and S417A mutants did not (Supplementary Fig. S1b–d). Using S417G-derived EC crystals and NTP soaking strategy, we obtained a set of four RDV-MP-containing EC crystal structures in the same crystal form previously determined using the same P10 primer and a 35-nt template with only one nucleotide difference from T34-1 [20] (Fig. 1b, Supplementary Figs S2 and S3, and Supplementary Table S1, crystal form 1). The S417G mutant allows RDV-MP to reach and pass through the −4 product position in EV71 RdRP EC. (A) The sequences of the T34-1/P10 construct used in the EC assembly, crystallization, and crystal soaking trials. A 15-mer-containing EC was assembled using GTP and RDV-TP and then used in crystallization and crystal soaking trials. (B) EC structures captured in different NAC states and with RDV-MP at various positions along the product chain. From left to right: the crystallized EC in state 1 (C1S1); a post-catalysis pre-translocation state 4 complex after incorporation of two nucleotides (C2S4); a post-catalysis over-translocated complex after incorporation two nucleotides (C3S1o); a translocation intermediate after incorporation of three nucleotides (C3S6). Color scheme: template, cyan (+1 nucleotide, orange); product, green (first and second incorporated RDV-MP, purple and slate, respectively); palm, gray (YGDD sequence, magenta); thumb, slate (S417G, red). Capital letters with gray oval-shaped background indicate corresponding RdRP motifs. The horizontal bar indicates the reference location of the product strand position −4. 2Fo − Fc electron density maps (contoured at 1.0 σ) of the RDV nucleotides are overlaid onto the structural models. “.x” in position labels denotes an intermediate location between two consecutive reference positions. (C) Schematic free-energy diagrams for working models depicting NMP (denoting regular nucleotide) and RDV-MP translocations. In NMP-containing EC, the pre-translocation state (S5) could establish fast equilibrium with the intermediate states (S6), and the subsequent transition to the post-translocation state (S1) is rate-limiting. The over-translocated state (S1o) is energetically unfavorable (high-energy). However, in the RDV-MP-containing EC, both the translocation intermediate state and the post-translocation state are less stable (corresponding to higher free-energy levels), while the S417G mutation stabilizes the post-translocation state and readily allows the inter-conversion between S1 and S1o. For all four structures, the manually superimposed versions with the 3500-K composite simulated-annealing omit 2Fo − Fc electron density maps overlaid are presented in Supplementary Fig. S3. For a better illustration of the structural details of EC, we adopt a nomenclature used for viral RdRP nucleotide addition cycle (NAC) structural states: a post-translocation state 1 (S1) represents the start of NAC; an NTP bound state 2 (S2) with an open active site; a catalytically competent state 3 (S3) with the active site in the closed conformation; a post-chemistry pre-translocation state 4 (S4) with the active site still closed; a pre-translocation state 5 (S5) with a re-opened active site; structurally captured translocation intermediate state(s) (S6) with RdRP poised between the pre- and post-translocation positions [43, 44]. As this crystal form allows incorporation of up to three nucleotides in NTP soaking trials, NAC cycle number was assigned along with the NAC state for each structure (e.g. C1S1 denotes state 1 of the first in-crystal NAC). As expected, the S417G mutant/RDV-MP-derived EC structure is a post-translocation C1S1 complex with the two RDV-MP nucleotides poised at product positions −2 and −4. The conformation of this complex is highly consistent with the reference C1S1 complex obtained using regular nucleotides and WT enzyme (PDB entry 5F8G) (Fig. 1b and Supplementary Fig. S2). The nitrogen atom of the −4 position RDV-MP 1′-cyano has a 3.2-Å close contact with the Cα atom of the engineered G417 (Supplementary Fig. S2). Modeling a serine at this site leads to a steric clash of its side chain and the cyano group (shortest non-hydrogen atom distance: 1.3 Å, Supplementary Fig. S2). CTP soaking of crystal obtained using WT RdRP and regular nucleotides typically yields a C3S1 complex with the RdRP entering the third NAC after two CMP incorporations and two translocation events [20]. Interestingly, we obtained two different states with the S417G mutant/RDV-MP EC. The first is a C2S4 state that has the RdRP retained at the pre-translocation position, and the second is a previously unidentified C3S1o state that we define as “over-translocated” (denoted by the lowercase subscript “o”). The RNA conformation of C3S1o is very similar to previously determined C3S6 reference complex (PDB entry 5F8M), except that the 3′-nucleotide is poised between positions −1 and −2 (Fig. 1b and Supplementary Fig. S2). Hence, C3S1o is not a conventional translocation intermediate, but an off-pathway state. Its 3′-nucleotide of the product RNA needs to move toward the downstream direction (opposite direction of regular translocation) to reach position −1 for NTP binding and catalysis. The capture of both the C2S4 and C3S1o complexes suggests that the free-energy landscape is altered by the RDV nucleotide even when the side chain of the roadblock serine is replaced by a hydrogen of the engineered glycine. Hence, RDV-MP at position −4 (post-translocated, C1S1), position −3 (pre-translocated, C2S4), or between positions −4 and −5 (over-translocated, C3S1o) are likely energetically comparable in the S417G-derived ECs (Fig. 1 b and c and Supplementary Video S1). Interactions of the first RDV-MP 1′-cyano in the C2S4 complex and the second RDV-MP 1′-cyano in the C3S1o complex both with residue D413 likely contribute the stability of these structurally captured complexes (Supplementary Fig. S2, compare with reference structures in panel a). Although we have not observed the second RDV-MP at position −4, the structural capturing of a C3S6 complex in CTP, UTP soaking trials demonstrates that the third nucleotide incorporation can indeed occur, indicating that the second RDV-MP can move from position −3 to the intermediate position between −4 and −5, and during this process the RdRP does incorporate one nucleotide presumably when this RDV nucleotide is poised at position −4 (Fig. 1b, Supplementary Fig. S2, and Video S1). RDV-MP incorporation rates are similar for EV71 WT RdRP and the S417G mutant and are 5-to-6-fold slower than AMP incorporation Although the observation of RDV-MP around product strand position −4 in EV71 RdRP strongly suggests that its translocation from position −3 to −4 is hindered primarily due to the steric effect between its characteristic 1′-cyano group and the side chain of residue S417, direct evidence of such translocation intervention is still needed for a proof. We therefore carried out a set of enzymology characterizations to collect further evidence. Since efficient RDV-MP incorporation is a prerequisite for subsequent “i + 3” intervention, we first determined the pre-steady-state single-nucleotide incorporation rate under relatively saturating substrate concentration for both AMP and RDV-MP using a T33/P10 RNA construct and a quench-flow instrument. Such a construct allows a 14-nt product-containing EC (EC14) assembly using the CTP, UTP combination followed by AMP/RDV-MP incorporation to produce a 15-nt product (Fig. 2a, P14 to P15; Supplementary Fig. S4a). For WT EV71 RdRP, the incorporation rate (r) for AMP is 7.5 s−1, while the rate for RDV-MP is 1.4 s−1 (Fig. 2b). It turned out that the S417G mutation only brings moderate impact on AMP and RDV-MP incorporation with rate values of 6.1 and 1.1 s−1, respectively (Fig. 2b). The only 5-to-6-fold rate reduction of RDV-MP for both WT and S417G mutant RdRPs suggests that RDV-MP incorporation is still quite efficient. To reveal the structural details of RDV-MP incorporation, the same T33/P10 construct was used for EC14 assembly and crystallization. The EC14 crystals were then subjected to RDV-TP soaking for determination of an RDV-MP-containing EC structure at 2.7-Å resolution (Supplementary Table S1, crystal form 2, Fig. 2c, and Supplementary Fig. S4b). This structure has one RDV-MP incorporated but not yet translocated (i.e., at position + 1). The backbone conformation of the active site motifs is consistent with the S4 state but no divalent metal ions are observed at the active site. We therefore used a previously determined C1S4 structure (PDB entry 5F8J) as a reference for structural comparison [20]. The RDV-MP base forms a regular Watson–Crick base pair with the +1 template uridine, but the RDV-MP ribose ring moves about 1.4 Å toward the direction of the motif B helix (Fig. 2c and Supplementary Fig. S4b and c). As a result, the hydrogen bond between the ribose 3′-hydroxyl and the backbone imino group of motif A D238 is lost, the hydrogen bond between the ribose 2′-hydroxyl and motif B S289 hydroxyl is maintained but weakened (bonding distance: 3.5 Å vs. 2.6 Å), and newly established close contacts between the RDV-MP 1′-cyano nitrogen and residues T294 and S295 are observed (nitrogen-threonine Cγ distance: 3.3 Å; nitrogen-serine Oγ distance: 3.5 Å). Taken together, the structural data suggest that both gains and losses of interactions occur upon recognition of the incorporated RDV-MP, and the ribose position difference between the RDV-MP and the reference structure +1 nucleotide likely explains the moderate reduction in incorporation rate, as the 2′- and 3′-hydroxyls play key roles in pre-catalysis active site closure [20, 45]. Kinetics and structural characterization of RDV-MP incorporation. (A) Sequences of the T33/P10 construct used for EC assembly and a subsequent single-nucleotide incorporation assay. A P14-containing EC was assembled using CTP and UTP and then used in quench-flow trials by addition of ATP or RDV-TP for real-time monitoring of P14-to-P15 conversion. (B) Determination of incorporation rates of AMP or RDV-MP in WT and S417G mutant ECs. The fractional values of 15-mer intensity measured in three independent experiments were used for curve fitting to generate the rates and fitting errors. Fitting error values below 0.1 were set to 0.1, indicated by underlining. (C) Stereo-pair images comparing the RDV-MP-containing structure (color, RDV-MP at the +1 position) and the pre-translocation reference state (gray, PDB entry 5F8J). Color scheme is the same as in Fig. 1b. The side chains of several RDV-MP interacting residues are shown as stick representations. Green dashed lines indicate hydrogen bonding interactions between RDV-MP and side chains of residues T294 (3.3 Å), S295 (3.5 Å) and S289 (3.5 Å), and gray dashed lines indicate hydrogen bonding interactions between CMP and residues D238 (3.0 Å) and S289 (2.6 Å). The composite rate of RDV-MP translocation from position −3 to −4 and subsequent catalysis is at least 100-fold slower than that of AMP, and the S417G mutation largely eliminates this difference To precisely probe RDV-MP translocation from product position −3 to −4, we designed another quench-flow-based enzymology assay. Another 34-nt template (T34-2) was used in combination with P10 to first assemble WT or S417G mutant-derived EC14 complexes with either a GAGA or a GRGR tetra-nucleotide incorporated (Fig. 3a, P10 to P14). After removal of the NTPs used for EC14 assembly, CTP and ATP were provided for extension of the P14 to 16-nt products (P16) using the quench-flow device (Fig. 3a, P14 to P16; Supplementary Fig. S5). In such a setting, the in-gel detection of the 14-to-15-nt conversion precisely probes the catalysis step to yield the 15-nt intermediate products (P15), while detection of the 15-to-16-nt conversion includes one translocation step (corresponding to the −3 to −4 translocation of the first incorporated AMP or RDV-MP in EC14 assembly) and the catalysis step to yield P16 (Fig. 3b). Here, we use an irreversible two-step consecutive reaction model for data fitting to yield two distinct rates to aid our analysis. The first rate (r1) corresponds to the incorporation step for the 14-to-15-nt conversion, and the second rate (r2) is a composite parameter including both the −3 to −4 translocation and the 15-to-16-nt conversion immediately following the translocation. It turned out that both rates are relatively fast (r1: 20.1 s−1; r2: 6.4 s−1) and comparable to the single-nucleotide incorporation rate of AMP or RDV-MP for the ATP-derived WT EC14, suggesting that the translocation rate contributing to r2 is also fast and at least comparable to the catalytic rate (Fig. 3c, top left; Fig. 3d). Strikingly, the r2 value is about 100-fold lower for the RDV-TP-derived WT EC14 (0.042 s−1 vs. 6.4 s−1), while the r1 value remains high (12.8 s−1). These data demonstrate that RDV-MP significantly impairs the process comprising translocation from position −3 to −4 and subsequent catalysis (Fig. 3c, top right; Fig. 3d). For the S417G mutant RdRP, ATP-derived EC14 yielded rates comparable to those of its WT counterpart (r1: 14.0 s−1 vs. 20.1 s−1; r2: 6.2 s−1 vs. 6.4 s−1), consistent with the aforementioned moderate impact on catalysis for the S417G mutation (Fig. 3c, bottom left; Fig. 3d). For RDV-TP-derived S417G mutant EC14, the r2 value was fully restored to a level comparable to its r1 value (16.7 s−1 vs. 18.8 s−1), indicating that the S417G mutation almost eliminates the RDV-mediated site-specific translocation-related intervention. Note that such an intervention could take place when NA either moves in between two translocational registers or distorts a post-translocated state. Detection of RDV-MP translocation from product position −3 to −4 by quench-flow kinetics. (A) Sequences of the T34-2/P10 construct used for EC assembly and a subsequent two-nucleotide incorporation assay. A P14-containing EC was assembled using GTP and ATP/RDV-TP and then used in quench-flow trials by addition of CTP and ATP for real-time monitoring of P14-to-P16 conversion. The incorporation event of the first RDV-MP (yielding the P12 product) is denoted as i, and therefore the P15 product corresponds to the i + 3 event. The “i + 3” RDV-MP intervention is expected to occur after P15 formation. (B) The kinetic model used for curve fitting to generate the rates corresponding to the in-gel conversions of P14-to-P15 (r1) and P15-to-P16 (r2). (C and D) Determination of r1 and r2 values for AMP-/RDV-MP-containing ECs assembled using WT and S417G mutant RdRPs. The fractional values of 14-/15-/16-mer intensities measured in three independent experiments were used for curve fitting shown in (C) to generate the rates and fitting errors listed in (D). MD simulation data support RDV-induced site-specific translocation hindrance and the compensatory effect brought by S417G mutation The structural and enzymology data together highlight the hallmark site-specific translocation-related intervention mechanism of RDV in EV71 RdRP, and a similar mechanism can presumably be applied to the SARS-CoV-2 RdRP system. In principle, this mode of NA intervention/regulation can be used in NA core structure rational design for any nucleic acid polymerase system. To explore computation-based approaches for probing such intervention, and for proof-of-concept efforts to establish an effective and efficient methodology of translocation-intervention-based NA evaluation, we use EV71 RdRP EC structure-derived models to perform MD simulations to monitor RdRP translocation events and how their behavior is modulated by RDV-MP and the S417G mutation. Accordingly, the systems of WT RdRP_AMP, WT RdRP_RDV-MP (RDV nucleotide at the +1 or −3 position in the product strand), and S417G mutant RdRP_RDV-MP (RDV nucleotide at the +1 or −3 position) complexes were comparatively studied with a sophisticated enhanced sampling technique called GaMD [40]. The use of GaMD allows for the repeated observation of RdRP translocation events and thus the quantitative investigation of the translocation path. The accumulated simulation time was 4.8 µs per system. All simulation parameters are depicted in Supplementary Table S2. The hallmark distances between atom pairs of RdRP residue S417 and AMP at the −5 position of product strand (CV1) and RdRP H199 and UMP at −5 position of template strand (CV2) were used to track the relative motion of the two corresponding RNA strands toward RdRP (Fig. 4a). As shown in a representative trajectory of WT RdRP_AMP (Fig. 4b), a translocation event occurs with the product strand moving ahead of the template strand. Moreover, the free-energy landscape calculated from the accumulated simulation data of this system further shows that the translocation path exhibits an “L” shape (Fig. 4c). Accordingly, in the intermediate state, the product strand moves almost one register forward whereas the template strand is apparently lagging behind (Supplementary Fig. S6a). These simulation data suggest that the motion of product strand precedes that of the template strand, consistent with crystallographic observations of the asymmetric movement of the two RNA strands during translocation [19, 20]. When the +1 position AMP is replaced by RDV-MP, translocation occurs with a similar asymmetric feature (product strand preceding template strand, as shown in Fig. 4d, middle plot). By contrast, translocation is hardly seen in the same simulation timescale when the −3 AMP is replaced by RDV-MP (Fig. 4e, middle plot and Supplementary Fig. S6b), consistent with the very slow composite rate (r2) observed in the enzymology characterization. For the S417G mutant, translocation is observed regardless of whether RDV-MP is at position +1 or −3 (Fig. 4d and e, right plots). The compensatory effect on RDV-MP translocation from position −3 to −4 brought about by the S417G mutation is also consistent with the restoration of the r2 value upon S417G mutation (Fig. 3d). MD simulations of AMP/RDV-MP translocation between specific sites in the product strand. (A) Schematic graph showing the collective variables (CVs) selected as the hallmark distances between the AMP C1′ atom at the −5 position of the RNA product strand and the RdRP S417 Cα atom (CV1) and the UMP C1′ atom at the −5 position of the template strand and the RdRP H199 Cα atom (CV2) to describe strand movement, respectively. (B) A representative GaMD trajectory as described by the time series of CV1 or CV2 for AMP translocation at the template or product strand in the WT RdRP_AMP complex. Dashed lines show the time when each strand begins its translocation. (C) Free-energy landscape along CV1 and CV2 for the WT RdRP_AMP complex. Contours in the two-dimensional subspace are spaced at intervals of 1.0 kcal·mol−1. Yellow dashed lines show the translocation paths from i to i + 1 site. (D) Left: the structure of the RdRP_RDV-MP complex as the RDV is at the +1 position of the product strand. Right: free-energy landscapes for the corresponding WT_RDV-MP and S417G_RDV-MP complexes. (E) Left: the structure of the RdRP_RDV-MP complex as the RDV is at the −3 position of the product strand. Right: free-energy landscapes for the corresponding WT_RDV-MP and S417G_RDV-MP complexes, respectively. . Discussion For decades, antiviral NA development has been oriented to direct catalysis intervention of nucleic acid polymerases. Here, we classify direct catalysis intervention according to three specific sites that NA occupies on its intervention. The first is active site (corresponding to product strand position +1) intervention (Fig. 5a, Ia and Ib). If NA binding at the active site does not lead to catalysis, such as the situation recently observed in the Lassa virus (LASV) RdRP on adenosine analog HNC-1664-TP intervention [46], the intervention can be defined as “non-catalytic occupation” (Fig. 5a, Ia). If an NA can be incorporated into the product chain and keeps occupying the active site, the intervention can be described as “post-catalysis occupation” (Fig. 5a, Ib). The second is the priming site (corresponding to product strand position −1) intervention, where the presence of NA at the priming site interferes with NTP binding or incorporation at the active site (Fig. 5a, Ic). The third is the templating site intervention, and the most notable examples are the mutagenic NAs, such as favipiravir and molnupiravir, that lead to high error rates in RNA synthesis due to base-pairing tolerance (Fig. 5a, Id) [5, 47]. As mentioned above, intervention can also occur when NA moves in between two translocational registers or distorts a post-translocated state. If such an intervention occurs during the NA movement from product position +1 to −1 (Fig. 5a, IIa), it can lead to immediate chain termination by partially occupying the active site. Note that situations Ib and Ic in catalysis intervention can also result in immediate chain termination, and all these three situations (highlighted by magenta circles in Fig. 5a) cannot be distinguished by traditional gel-based analysis, as all these intervention events lead to the accumulation of a product corresponding to the incorporation of the NA-MP. More importantly, translocation-related intervention can also take place at upstream positions (Fig. 5a, IIb). A proposal of translocation-related intervention as a generally applicable strategy for NA rational design. (A) A comparison of direct catalysis intervention (I) and translocation-related intervention (Ⅱ). Direct catalysis intervention is further classified by three intervention sites and four specific situations (Ia–Id), and translocation-related intervention is classified by movement either from active site to priming site (IIa) or between further upstream positions (IIb). (B) Stereo-pair images showing ribose-protein close contacts along the minor groove of template-product RNA duplex in three representative EV71 RdRP EC structures. If present, one representative close contact (with a distance of 4.2 Å or shorter between non-hydrogen atoms, yellow dashed line) between the C1′ or C4′ atom and the RdRP is shown for each nucleotide. Top: an on-register pre-translocation complex (PDB entry: 5F8N); middle: an early-stage translocation intermediate (PDB entry: 5F8N); bottom: a late-stage translocation intermediate (PDB entry: 6LSH). Color scheme is the same as in Fig. 1b. Previous studies, particularly in SARS-CoV-2 and EV71, have shown that steric hindrance by a structurally equivalent serine residue in the RdRP thumb domain is responsible for the characteristic “i + 3” intervention by the RDV nucleotide [14–16, 18]. Our work here further provides a quantitative measure of this site-specific intervention by determining the composite rate of translocation from product position −3 to −4 and the subsequent catalysis (Fig. 3). Although the enzymology data cannot discriminate the contribution of two rate components of the composition, the MD simulation data, obtained at the microsecond timescale, nevertheless suggest the contribution of the translocation component in the intervention (Fig. 4e). Very interestingly, the RDV nucleotide seems still to alter the free-energy landscape even with the roadblock serine replaced by a glycine that has the smallest side chain, as evidenced by the structure of an over-translocated state that has not been captured previously (Fig. 1b, C3S1o). In the WT RdRP context, the RDV nucleotide makes both its translocation intermediate state between product positions −3 and −4 and the post-translocation state at position −4 more unstable than a regular nucleotide does (Fig. 1c, compare black traces of RDV-MP and NMP). Presumably, the serine-to-glycine mutation can stabilize both states, making their stability comparable to the over-translocated state (Fig. 1c, indicated by red arrows). The capture of this over-translocated state with the glycine mutant suggests that the progressive movement of the RDV nucleotide along the product binding path can be affected even without an obvious side-chain roadblock, and therefore highlights the general inhibition potential of RDV and NAs with similar ribose modifications, with respect to both roadblock residue type and roadblock position. Indeed, close contacts are observed at multiple product strand positions between ribose and EV71 RdRP along the template-product duplex minor groove, as surveyed in one on-register EC and two translocation intermediates (Fig. 5b). The effectiveness of site-specific translocation-related intervention by RDV establishes a benchmark for developing antiviral NA drugs capable of intervention on translocation essentially at any position. Any steric hindrance of an NA in the product strand by polymerase residues provides an opportunity of translocation-related intervention. In a sense, translocation-related intervention-oriented NA rational design can be taken as a general strategy in NA drug or NA probe development. Considering the fact that duplexed nucleic acids are structurally much more rigid than single-stranded nucleic acids, we surveyed the product strand nucleotides within the template-product duplex region in representative viral RdRP ECs, looking for nucleotides having close backbone (i.e., riboses and phosphates) interactions with RdRP residues (distance cutoff 5 Å). Two take-home messages can be extracted from such an analysis. First, RdRPs from three major groups of RNA viruses, positive-strand (Supplementary Fig. S7a), double-stranded (Supplementary Fig. S7b), and negative-strand (Supplementary Fig. S7c), all have multiple nucleotides involved in such close interactions (nucleotides shown in magenta). Second, the interacting patterns exhibit relatively high diversity with interacting residues either from the palm-fingers-thumb core of RdRP or from other RdRP regions and/or co-factors within the RdRP EC, such as nsp8 of SARS-CoV-2 (Supplementary Fig. S7a), PB2 subunits of influenza A virus, and polyribonucleotidyltransferase (PRNTase) region of Nipah virus (NiV) RdRP (Supplementary Fig. S7c). Hence, a particular NA capable of undergoing one or more translocation events can result in different consequences in different polymerase systems: no intervention, intervention at one specific position, or intervention at multiple positions. For example, RDV “i + 3” intervention was evidenced in Coronaviridae SARS-CoV-2 and Picornaviridae EV71 RdRPs [14, 16], but has not yet been found in RdRPs from other virus families. As discussed above, translocation-related intervention expands the space for NA drug development in intervention sites (Fig. 5a), creating new avenues for NA rational design in particular for ribose and α-phosphate modifications that are responsible for nucleic backbone interactions. Although not investigated in this work, ribose 1′- (e.g. RDV) and 4′-positions are among the most probable modification sites to generate translocation-related intervention compared with 2′- and 3′-positions [48]. Modifications at ribose 2′-positions have the greatest potential to interfere with NTP binding and active site closure, while modifications at ribose 3′-α-position will allow at most one incorporation event. As a consequence, 2′/3′-modified NAs likely have a relatively low probability to reach upstream positions where most candidate sites for translocation-related intervention exist. With respect to NA rational design-based discovery of translocation-related-intervention NA, identifying “delayed intervention”-related intermediate products by traditional gel-based methods and capturing translocation-related intervention in action by structural biology methods are probably the most effective experimental biology approaches. Consistent with crystallography and enzymology characterizations in probing translocation, the MD simulation methodology for translocation monitoring established in this work provides a reliable and powerful tool for such a type of discovery. In principle, both experimentally determined and predicted polymerase EC structures can be used to generate starting models with candidate NA present or modeled at different positions in the product strand, and this approach will greatly expand the number of polymerase systems and NA molecules in the discovery scheme. In summary, by obtaining multiple-approach-based evidence in the EV71 RdRP, we have quantitatively characterized site-specific translocation-related intervention by the RDV nucleotide and found the more general translocation-related impact brought by this ribose-modified nucleotide. By differentiating between general translocation-related intervention and traditional direct catalysis intervention, we propose the former as a widely applicable NA-based intervention strategy in drug or probe development targeting nucleic acid polymerases. Approaches aiding the discovery and validation of translocation-related intervention NAs are discussed with a highlight of the MD simulation-based method that is worth promoting as a powerful and practical tool. By expanding intervention sites beyond the polymerase active site and its vicinity, this translocation-related intervention-based strategy could bring a profound impact on NA drug development. Supplementary Material