Revisiting Viral RNA-Dependent RNA Polymerases: Insights from Recent Structural Studies

RNA-dependent RNA polymerases (RdRPs) represent a distinctive yet versatile class of nucleic acid polymerases encoded by RNA viruses for the replication and transcription of their genome. The structure of the RdRP is comparable to that of a cupped right hand consisting of fingers, palm, and thumb subdomains. Despite the presence of a common structural core, the RdRPs differ significantly in the mechanistic details of RNA binding and polymerization. The present review aims at exploring these incongruities in light of recent structural studies of RdRP complexes with diverse cofactors, RNA moieties, analogs, and inhibitors.


Introduction
RdRPs (EC 2.7.7.48.) are ancient enzymes that catalyze the formation of phosphodiester bonds between ribonucleotides in an RNA template-dependent fashion using divalent metal ions. The reaction involves the nucleophilic attack of the 3'-hydroxyl group of the nascent RNA on the α-phosphate of the incoming nucleoside triphosphate (NTP), resulting in the formation of a new phosphodiester bond and the release of a pyrophosphate moiety ( Figure 1A). Together with other viral and host proteins, the RdRPs participate in RNA synthesis, discriminating replication and transcription events, and incorporating the 5' cap and the 3' poly A tail [1]. RdRP is the sole enzyme that exhibits significant similarity to ancestral t-RNA sequences, suggesting a descendance from them [2]. Most of the viral RdRPs were identified based on comparative sequence analysis and are highly conserved in the viral genome [3]. The RdRPs have strikingly similar domain and motif arrangements, demonstrating how evolutionary conservation manifests itself at the level of structure [4].
RdRPs are classified as alpha and beta proteins and fold majorly into three subdomains, namely the fingers, thumb, and palm ( Figure 1B) [3,[5][6][7][8]. The majority of the structurally conserved components are located in the palm subdomain, which is sandwiched between the fingers and the thumb subdomains. The palm harbors the catalytic aspartates and the three-stranded RNA-recognizing motif (RRM). The variable thumb subdomain interacts with the primer while the fingers play a crucial role in template binding [1,3]. The RdRPs exhibit "close-hand" conformation as opposed to DNA polymerases owing to the projecting loops ("fingertips") that make the fingers and the thumb subdomains appear interconnected [7,9]. The structural features of the RdRP result in the formation of two channels that converge at the active site ( Figure 1B). The RNA template gains access through the primary channel while the incoming NTPs enter through the secondary channel. To further their ability to service their needs, several RdRPs possess other channels, including the RNA exit channel [6,7,10,11]. Seven conserved motifs, A through G, were discovered by comparing the structures and sequences of RdRPs that are implicated in the binding of metal ions, NTPs, and RNA ( Figure 1B) [12,13]. Motifs A through E reside in the fingers subdomain, while motifs G and F are nested in the palm subdomain. For the group III and IV viruses, a further motif H was found in the thumb subdomain. Motif A houses the first aspartate of the catalytic motif DX2-4D. The motif B of the palm subdomain aids in the binding of the template RNA and in substrate discrimination. The conserved GDD motif, which is used to bind metal ions, is located in the loop of motif C. Due to the presence of a conserved glycine, motif D functions as an anchor for conformational changes related to correct NTP binding. Motif E, commonly known as "the primer grip", aids in the proper positioning of the primer's 3 hydroxyl group during catalysis. Motif F interacts with the phosphate group of the incoming NTP while the loop of motif G forms a part of the template entrance channel in group IV viruses [4,6,9,14].
The sizes of viral RdRPs vary greatly due to the presence of auxiliary domains that aid in catalysis. Those from the family Reoviridae and phage ϕ6 are 148 and 76 kDa, respectively, and possess extended N-and C-terminal domains in addition to the core [10,15]. The picornaviral 3D pol is only ∼50 kDa [9,16], while the flaviviral NS5b is 100 kDa in size as it includes a fused methyltransferase (MTase) domain [11,17]. The nidoviral NiRAN (nidovirus RdRPassociated nucleotidyl transferase) domain spans over 350 residues and is linked to the C-terminus of the RdRP [18]. The RdRPs of the non-segmented single-stranded (ss) (-) RNA viruses, such as the Vesicular stomatitis virus (VSV) (240 kDa), have polyribonucleotidyl transferase (PRNTase) and MTase domains along with a ring-like core [19]. The segmented ss (-) RNA influenza virus replicase complex is~260 kDa, comprising the polymerase acidic (PA), polymerase basic (PB), and RdRP subunits [20]. The structural conservation and variations in the RdRPs both within and across families, have been extensively analyzed [3,4,9,18,21,22]. In this article, we have reviewed the recent structural studies of RdRP complexes from different viral families in terms of modes of replication and inhibition.

The Business of Copying
The structural snapshots of RdRPs in association with nucleic acid moieties and analogs have greatly aided the understanding of their modus operandi. The exact geometry of the active site and the disposition of the template/product, different domains, motifs, and cofactors during the pre-and post-stages of catalysis are now better understood. The following subsections summarize these findings (May 2018 to present) ( Table 1).

Cooperative Cofactors
A repertoire of self and co-opted host-encoded proteins assist the RdRP during genomic replication and transcription. These include the self-encoded helicases; capping enzymes; endonucleases; VPg; phosphoproteins; and the subverted proteins of the host, such as the translation factors, protein chaperones, RNA-modifying enzymes, and cellular proteins. Altogether, the assembly of the replication transcription complex (RTC) is a rather complex process driven by many factors for optimal replication in the infected cells.
In families such as Potyviridae, Picornaviridae, and Caliciviridae, the viral protein genomelinked (VPg) is covalently linked at the 5 end of the genome and functions as a primer during replication [23][24][25][26]. The co-crystal structure of Murine norovirus (MNV) RdRP with VPg highlighted the role of the latter in both priming and multimerization during replication (PDB 5Y3D) [27]. In the presence of RNA, the MNV RdRP-VPg complex  clustered and displayed a tightly packed, ball-like configuration reinforced by the interactions of the RdRP with the disordered C-terminus of the VPg (Figure 2). The resulting ensemble not only offered a sizable interior area to trap ss RNA but also improved cooperative RNA binding for effective replication [27]. The structural investigation of the influenza polymerases from the human A/NT/60/1968 (H3N2) and avian A/duck/Fujian/01/2002 (H5N1) viruses also highlights the necessity of dimerization of the polymerase during genome replication (PDB 6QNW, 6QPF, 6QPG, 6QWL, 6QX3, 6QX8, 6QXE, 6RR7) [28].  Diverse cofactors are required by the coronaviral polymerases for effective transcription and replication of the large and complex coronaviral genomes. Hence, coronaviruses possess a highly processive polymerase (RdRP core with nsp8/nsp7), a system for proofreading (nsp14 with the 3'-5' exoribonuclease activity), a helicase (nsp13) for unwinding the template-product duplex, and the MTases for capping (nsp14, nsp16) [29]. The structure of the replication transcription complex (RTC) of SARS-CoV-2 features two copies of the nsp8 flanking the nsp12 (PDB 7BW4, 6YYT, 7CYQ, 7C2K, 7BZF, 7BTF, 6NUR, 6NUS) [30][31][32][33][34]. The N-terminal helical extensions of the nsp8 act as "positively charged sliding poles", facilitating the process of elongation and engaging with the exiting RNA duplex ( Figure 3A) [30,31,33]. During replication, the two nsp8 copies show varying patterns of association with the nsp12 and nsp7. The pre-and the post-translocated states of the RTC reveal a 45 • difference in the orientation of the N-terminal helical extensions of the two nsp8 copies (PDB 7C2K, 7BZF) [32]. The nsp7 and nsp8 also act as a "non-canonical primase", mirroring the SARS-CoV-1 hexadecameric primase ring complex in the mode of RNA binding [32,35]. During the early stages of replication, the RTC binds to the double-stranded (ds) RNA helix in a sequence-independent manner via the 2'OH of the ribose, demonstrating the basis for the RNA selectivity ( Figure 3A) (PDB 7BV1, 7BV2) [36]. Earlier studies of the RTC with nsp13 proposed the role of the latter in backtracking, thereby facilitating template switching, proofreading, or both during subgenomic RNA transcription (PDB 6XEZ, 7KRN, 7KRO, 7KRP) [37,38]. The association of RTC with two nsp13 moieties aided in transitioning between backtracking and accelerated RNA synthesis in response to cues at the RdRP's active site via an allosterically controlled mechanism (PDB 7RDX, 7RDY, 7RDZ, 7RE0, 7RE1, 7RE2, 7RE3). The nsp13 achieved this extraordinary feat by means of four different conformational states [39]. Recently, Park et al. proposed a carefully orchestrated relay involving the NiRAN domain, nsp9, nsp13, and the MTases to cap the 5 end of the genome ( Figure 3B) (PDB 7THM) [40]. In the first step, the NiRAN domain transfers the 5 -product RNA to the amino terminus of nsp9. The RNAylated nsp9 is then attacked by the GDP, which is produced by the nsp13 via GTP hydrolysis, releasing the capped RNA and regenerating the unmodified nsp9. Finally, nsp14 and nsp16 generate a fully functional cap via RNA methylation. The STAT2 protein, a key player in inducing IFN responses during viral infections, was found to interact in a two-way manner with the nsp5 of the Dengue (DENV) and Zika viruses ( Figure 3C) (PDB 6UX2, 6WCZ) [41]. The RdRP is associated with the NTD of STAT2 while its coiled-coil domain is nestled in a conserved inter-domain crevice between the methyltransferase and the RdRP domains. As both nsp5 and the interferon regulatory factor 9 compete for the same interaction surface of STAT2, the binding of the former resulted in the suppression of the IFN-induced expression of antiviral genes [41].  In the order Mononegavirales, the nucleocapsid-associated template is copied by the L protein with the aid of its cofactor, the tetrameric phosphoprotein (P), during genome replication and transcription [42,43]. The RdRP, cap-specific MTases, and GDP PRNTase domains of the ring-shaped L protein make it multifunctional. The RdRP domain of Human metapneumovirus (HMPV) comprises the canonical fingers-palm-thumb subdomains [44,45]. The L-P complexes were structurally characterized for the segmented HMPV (PDB 6U5O) [44], the Respiratory syncytial virus (RSV) (PDB 6PZK) [46], and the non-segmented Parainfluenza virus 5 (PIV5) (PDB 6V85, 6V86, 6VAG) [47]. In the strikingly similar complexes, the helices and their extensions from the tetramerization domain of the P interact with the L and clasp it in a tentacle-like manner ( Figure 4a). The P monomers associate with the L and the nucleocapsid through overlapping regions exhibiting distinct conformations and interactions [19,44,46]. The L-P complex structure of PIV5 revealed that the central oligomerization domain and the C-terminal X domain of the P attach to two distinct binding surfaces on the L protein. The active site of the MTase was positioned directly above the PRNTase domain, thereby reducing the distance between the former and the conserved HR motif of the latter by~27 Å, representing the complex in a transcriptioncompetent form [47]. The unique conformations of the MTase and the CTD relative to the RdRP domain in PIV5 might have implications for switching between genome replication and transcription ( Figure 4B). The cryo-EM structures of the Influenza C virus (FluC) polymerase in complex with human and chicken acidic nuclear phosphoprotein 32 (ANP32) revealed that the protein served as a replication platform for the viral RNA (vRNA) (Figure 4C) (PDB 6XZD, 6XZG, 6XZP, 6XZQ, 6XZR, 6Y0C). In both complexes, two polymerase molecules were bridged by the amino-terminal leucine-rich repeat domain of ANP32 [48].  The arenaviral zinc-binding regulatory matrix protein (Z) synchronizes the assembly and budding of progeny viruses, negatively regulates RNA synthesis, and blocks host immunity, among other tasks, through interactions with viral and cellular proteins [49]. The structures of the arenaviral L-Z complexes aided in the understanding of the molecular mechanism of the on/off switching of vRNA production. The complexes of the Old World arenavirus, Lassa virus (PDB 7OCH, 7OE3, 7OE7, 7OEA, 7OEB, 7OJJ, 7OJK, 7OJL, 7OJN), and the New World arenavirus, Machupo virus (PDB 7CKM, 7ELC), revealed a single copy of the Z bound to the L polymerase at the interface of the PA subunit carboxy (C)-terminal-like (PA-C-like) regions and the RdRP ( Figure 5) [50,51]. The central L-binding domain of the Z was shown to regulate the polymerase activity by associating with the motifs D and E, whereas the flexible extremities were proposed to interact with diverse proteins during the assembly for genome encapsidation. In the New World Junin virus (JUNV), the Z was found to interact with both the core-lobe and the palm domains of the L, inhibiting its activity during late infection ( Figure 5) (PDB 7EJU) [52]. However, in the early stages, the lower concentration of Z in the cell and a conformational shift of the 680-loop due to interactions with host proteins such as eIF4E38 caused the Z to detach from the L, activating it [52].

Amino-Terminal Augmentations
The RdRPs of many viruses have additional domains at their N-terminus, such as MTases, endonucleases, cap binding domains (CBDs), and PRNTases, which play critical roles during RNA synthesis.
The multifunctional polymerases of the family Peribunyaviridae are involved in both genome replication and transcription. Following replication, either vRNA or copies of the complementary RNA (cRNA) are generated. The 5 caps from the host mRNAs are pirated with the help of the CBD and the endonuclease domains [22,[53][54][55][56]. The previous crystal structure of the L protein of VSV suggested the threading of a~20nt stretch of the template RNA through the polymerase domain during transcription [19]. Further, the disposition of the priming loop of VSV-L on the capping domain led to speculations on the coupling of capping and initiation during polymerization [19]. Nevertheless, the recent structures of the LACV polymerase in the pre-initiation and the elongation-mimicking states reveal the synchronized movement of the priming loop, the lid domain, and the mid-thumb ring linker to accommodate a ten-base-pair template-product duplex during elongation ( Figure 6A) (PDB 7ORI, 7ORJ, 7ORK, 7ORL, 7ORM, 7ORN, 7ORO, 6Z6B, 6Z6G, 6Z8K) [55,56]. The authors suggested a properly orchestrated sequence of events, starting with the modification of a distal duplex promoter, followed by the generation of tension in the nucleotides 6-8 of the 3' template, and finishing with the extension of the prime-andrealign loop for the polymerase to pursue active replication [55].
The RdRPs of the family Orthomyxoviridae are associated with the 3 end of the genome where they are involved in initiating transcription soon after infection. The recent structures of the polymerases from this group have illustrated significant details on this phenomenon and the changes associated with cap-binding and polymerization. The structure of the FluA and Influenza B virus (FluB) polymerases revealed the ordering of the CBD (PB2, 320-489) and the "midlink" domain (251-319 of PB2 with the 490-536 linker) that mimicked the priming state of the polymerase when bound to their physiological substrate, the capped RNA ( Figure 6B) (PDB 6EVK, 6EVJ, 6EUY, 6EUX, 6EUV, 6EVW) [57]. The CBD plays a pivotal role in directing the capped RNA to the endonuclease for cleavage and positioning the severed primer into the polymerase active site via a rotation of about 60 • to commence transcription [20,[58][59][60][61].
The Phenuiviridae RdRPs harbor an endonuclease domain at the N-terminal region followed by the motifs G-F-A-H-B-C-D-E that are located centrally. The large globular core includes the polymerase, the PA-C-like domain, the linker region, and the PB2-N-like domain. The recent cryo-EM reconstruction of the RdRP of Severe fever with thrombocytopenia syndrome virus (SFTSV) captured the structural details of the endonuclease domain of the enzyme in an early pre-initiation state of transcription ( Figure 6C) (PDB 7ALP) [62]. The rotatable endonuclease domain of SFTSV L was housed back against the CBD (1695-1810), while in the LACV L it was flipped by~180 • along the long axis of the endonuclease facing the CBD [21,54]. This mandated substantial changes in the conformations of the SFTSV-L to initiate cap snatching. The crystal structure of the isolated putative CBD in complex with an m 7 GTP cap analog showed similarity to that of the Rift Valley fever virus (RVFV) CBD, comprising a long α-helix packed against a seven-stranded mixed β-sheet (PDB 6XYA, 6Y6K) [62]. The endonuclease domain and the CBD did not have a proper density in the cryo-EM structure of the RVFV L protein (PDB 7EEI). However, an integrated model constructed using SAXS and a low-resolution structure of the full-length L demonstrated that the endonuclease domain and the CBD were adjacent to one another, which is consistent with their functions. Similar to other bunyaviruses, the transcription of RVFV RNA required a primer through a cap-snatching process, although this was not the case with SFTSV [63]. The disposition of the priming loop of RVFV was also quite distinct from that of the SFTSV L and underwent large conformational shifts during transcription ( Figure 6C). A unique feature of the RVFV RdRP was the ability to initiate de novo synthesis of RNA in the absence of 5 vRNA, which was unseen in other bunyaviral RdRPs.
The apo structures of the 3Dpols of the genera Kobuvirus (PDB 6R1I) and Sicinivirus (PDB 6QWT) reveal similarity to the picornaviral polymerases comprising the regular thumb, palm, and fingers subdomains and channels ( Figure 6E) [66]. Using the crystal structures, subgenomic replicon assays, and phylogenetic analysis, it was established that the N-terminus of 3Dpol, liberated upon the proteolytic processing of the 3CD precursor, stabilizes the α10 (234-240) helix and consequently activates the picornaviral polymerases.

C-Terminal Bracelets
The λ3 polymerases of Reoviridae are enormous cage-like structures that are encircled by their long N-and C-terminal extensions. While the N-terminal region wraps the continuous surface between the fingers and thumb subdomains of the polymerase, the C-terminal domain is an annular structure, popularly referred to as the "bracelet domain", with a large opening capable of admitting ds RNA.
The cryo-EM study using Cytoplasmic polyhedrosis virus (CPV) RdPR elucidated the crucial conformational changes associated with the thumb and the bracelet domains during three main stages of de novo transcription: (1) the transition from the quiescent to the initiation state brought on by the unwinding of the capped-terminal RNA; (2) the progress from the initiation to the early-elongation state via the insertion of the non-template RNA into its binding cleft; and (3) the switch from the early-elongation to the elongation state as a result of the template and non-template RNA annealing and the release of the nascent transcript ( Figure 7A) (PDB 6TY8, 6TY9, 6TZ0, 6TZ1, 6TZ2) [67]. The conformational changes in the bracelet domain of the RdRP are dramatic and involve module A (912-1010) and module B (1067-1140). The former exhibits "relaxed" and "stiffened" states owing to changes in the turn number of an α-helix (955-1007), while the latter either "covers" or "uncovers" the transcript exit tunnel during the progress from quiescent to the elongation stages ( Figure 7A) [67].
Pronounced conformational changes were also observable in the bracelet domain (779-1088) of rotaviral RdRP during transcription ( Figure 7B) (PDB 6OGY, 6OGZ) [68]. The multifunctional bracelet domain aided the formation of the transcription bubble and directed the product RNA strands to exit during elongation. Additionally, the investigation emphasized the role of the N-terminal helix-loop-helix subdomain  in modulating RdRP activity through a switchable helix (349-360) of the capsid shell protein [68].
The in situ structure of the Simian rotavirus RdRP with triple-layer particles (TLPs) and double-layer particles (DLPs) in transcribing and non-transcribing modes revealed that localized conformational changes in the bracelet domain and the capsid shell protein facilitated the exit of the RNA transcript from the polymerase and the viral particle (PDB 6OJ3, 6OJ4, 6OJ5, 6OJ6) [69]. The cryo-EM and subparticle reconstruction of Blue tongue virus (BTV) RdRP in the TLPs and the DLPs provides insights into the role of the bracelet domain (936 to 1302) in the regulation and activation of the RdRP at the time of infection ( Figure 7C) (PDB 6PNS, 6PO2) [70]. During entry, the movement of the VP3 protein, an inner layer protein of the viral subcore, disrupts the interactions of the protein's N-termini with the RdRP at the five-fold axis. This forces a change of conformation of the bracelet domain, providing the requisite space and energy for the initiation of transcription [70].

Motifs, Pockets, and More
The non-canonical RdRPs of the family Permutotetraviridae display a unique cyclically permuted C-A-B-D connectivity of the four palm motifs [71,72]. The structures of the replicative complexes of Thosea asigna virus (TAV) RdRPs bound to the RNA templateprimers were determined in three distinct replication states that provided high-resolution mechanistic details of the nucleotide addition cycle ( Figure 8A) (PDB 7OM2, 7OM6, 7OM7, 7OM9, 7OMA) [73]. It was observed that the active site closed before NTP entry and the catalytic aspartate of motif A repositioned itself upon NTP and metal ion binding. The motif F underwent a major rearrangement as well, relocating by 20 Å in the apo state of the polymerase in relation to the RNA-bound form. The action mimics the conformational changes of motif F in flaviviral RdRPs and orients the ceiling of the NTP channel towards the catalytic cavity [73,74].
The structures of the intermediates during the forward and reverse translocation events of Enterovirus 71 (EV71) 3Dpol underpin the role of motif G in governing the rate-limiting step ( Figure 8B) (PDB 6LSE, 6LSF, 6LSG, 6LSH) [75]. The motif G residues strongly resist the reverse translocation events by restricting the mobility of the RNA template and intermediates during translocation, thereby favoring the power stroke model of catalysis. Furthermore, a pocket in the fingers subdomain was found to be involved in a critical base-stacking interaction with the 5' guanosine of the downstream RNA (PDB 6KWQ, 6KWR) [76]. The pocket comprises three distinct regions and includes the interacting residues His44, Arg277, Asn18, and Ser45, which are conserved in the RdRPs of Picornaviridae.
Interestingly, the role of another pocket was highlighted in the high-resolution cryo-EM structures of the Influenza A virus (FluA) polymerase in association with the RNA ( Figure 8C) [77]. The RNA adopted a loop-like conformation as it re-bound to a partially buried secondary site on the surface of the polymerase between PB1 and the PA-C following its exit. As a result, the 3 end of the template remained tethered to the RdRP throughout the transcription process. This not only allowed the 3 end to stay protected but also aided transcription restart by promoter reformation (PDB 6SZU, 6SZV, 6T0N, 6T0R, 6T0S, 6T0U, 6T0V, 6T0W, 6T2C, 6TU5, 6TW1) [77]. The same was reiterated with the polymerases of Influenza D virus (FluD) and A/H3N2 [28], where the 3 end of the cRNA was bound to a secondary site on the polymerase [78,79]. The FluD polymerase structures in apo and promoter-bound form suggested the presence of two conformations (modes A and B) pertaining to the 3 configuration of the promoter ( Figure 8D) (PDB 6KUJ, 6KUK, 6KUP, 6KUR, 6KUT, 6KUV, 6KV5, 6KUU). While the mode A conformation was shown to promote the synthesis of cRNA, the mode B conformation mimicking the pre-initiation site was seen to be pivotal for vRNA synthesis [79]. Recently, the dynamics of the active site of the FluB polymerase during elongation were investigated using a modified template ( Figure 8E) (PDB 6QCS, 6QCT, 6QCV, 6QCX, 6QCW) [60]. The study captured snapshots of the progressively extruding priming loop and widening active site cavity during the transition from the initiation to elongation steps. The co-crystal structure of the FluC polymerase with a Ser5-phosphorylated CTD peptide revealed the transcription-competent conformation of the polymerase in association with the CTD that facilitated the capture of the capped host mRNAs (PDB 6F5O, 6F5P, 6T0W) [80].
The crystal structure of the RdRPs of both the Tick-borne encephalitis virus (TBEV) (PDB 7D6N) [81] and the Yellow fever virus (YFV) (PDB 6QSN) ( Figure 8F) [82] in association with their MTases showed similar structural features to the known RdRPs of other flaviviruses. Through mutational analysis, the authors identified a highly variable and solvent-accessible region in the nsp5 between the catalytic motifs B and C as the source of host-related diversity in the members of Flaviviridae [81].

Inhibitor Complexes of RdRP
RdRP has emerged as the most versatile protein for developing drugs against RNA viruses [4,14,83]. These drugs are broadly categorized into nucleoside and non-nucleoside inhibitors since they act by binding to the RdRP's active site and allosteric regions, respectively. In the present study, the crystal structures of recently solved RdRP-inhibitor complexes ( Table 2) have been analyzed for a greater understanding of their mode of binding and inhibition.

Picornaviral Inhibitors
Previously, Picornaviridae inhibitor complexes were shown to target antivirals at the interface of the fingers and the palm subdomains [4]. The EVD68 RdRP crystal structures in the apo state (PDB 6L4R) and in complex with the non-nucleoside-like inhibitor NADPH (PDB 5ZIT, Ligand ID NDP) revealed a binding pocket at the fingers subdomain close to the RNA template channel that correlated to a similar one in the Coxsackie virus B3 RdRP [84,85]. The binding of NADPH showed no apparent conformational variations with respect to the apo form and involved interactions with the residues Glu104-Asp107, Val206, Gly207, Ser284-Ser290, and Tyr322 ( Figure 9A). The IC 50 values of the NADPH against the RdRPs of EVD68, EVA71, and PV were determined to be 232.9 µM, 746.6 µM, and 342.8 µM, respectively.

Coronaviral Inhibitors
Repurposing existing RdRP inhibitors for SARS-CoV-2 remains a viable option given the similarities in the essential drug-binding sites between the Coronaviridae members. A recent study of 22 SARS-CoV-2 RdRP inhibitors offered insights into the mechanics of viral RNA replication and a logical model for antiviral design [86]. The structures of the SARS-CoV-2 RdRP in complex with suramin and its derivatives (PDB 7D4F, Ligand ID H3U) revealed the presence of two distinct binding sites ( Figure 10A) [36]. The first site (site 1) is located at the RdRP's template binding region and involves the conserved motif G and the amino terminus of motif B. The naphthalene-trisulfonic acid moiety of suramin interacts with the residues Asn497, Lys500, Arg569-Lys577, Asn496, and Gly590 in site 1 and prevents the binding of the template ( Figure 9B). Site 2 is positioned near the active site and is formed by the conserved motifs A, C, E, and F between the fingers and the thumb subdomains where the residues Arg550-Arg555, Arg836, and Asp865 participate in hydrogen bonding ( Figure 9C). All the interacting residues are found to be conserved in the RdRPs of the family Coronaviridae. Suramin (IC 50 0.26 µM) and its derivatives are at least 20 times more potent than remdesivir against SARS-CoV-2 RdRP, but their use is limited because of poor cellular absorption caused by a strongly negative surface charge and off-target effects on helicases and polymerases [86].
The crystal structures of SARS-CoV-2 RTC with inhibitors favipiravir (PDB ID: 7DFG, Ligand ID 1RP) and ribavirin (PDB ID: 7DFH, Ligand ID RVN) are available from an unpublished work and reveal the binding of the inhibitors inside the RNA template channel. Favipiravir interacts with the residues Ser682, Asn691, and Lys545, while ribavirin interacts with the residues Asn691 and Ser682 via their carbamoyl and ribofuranosyl groups, and both show similar efficacy with IC 50 of~0.26 µM ( Figure 9D,E, respectively). The cryo-EM reconstruction of the favipiravir ribonucleoside triphosphate (favipiravir-RTP) with the RdRP and the template-primer RNA duplex shows the binding of the inhibitor at the catalytic site of the RdRP in a non-productive conformation (PDB 7AAP, Ligand ID GE6) [87].
Recently, the mode of action of the highly potent drug remdesivir (Ligand ID GS-5734) was investigated by independent groups using cryo-EM ( Figures 9F and 10B) (PDB 7BV1, 7BV2) [86,88,89].  showed that the binding of remdesivir to the RTC along with the template-primer RNA (50 bp long) revealed significant conformational changes involving: (a) the movement of nsp7 with respect to nsp12; (b) the loop connecting the first two helices of the thumb subdomain; and (c) the residues of motif G. The inhibitor was covalently incorporated into the primer strand at the first replicated base pair, terminating the chain elongation [36]. Another study along the same line by Cramer et al. captured the structure of the remdesivir-stalled RdRP (PDB 7B3B, 7B3C, 7B3D). According to the authors, the cyano group of the inhibitor is sterically hindered in its transit past the ser861 side chain of the nsp12, resulting in a translocation barrier during elongation [88]. On the contrary, the structure of the elongation complex of SARS-CoV-2 with remdesivir by Gong and group revealed "i+3" delayed intervention of the inhibitor in meta-stable complexes, resulting in only a temporary pause of elongation (PDB 7DTE) [89]. The molecular mechanism underlying the inhibitory activity of molnupiravir ( Figure 9G) (PDB 7OZU, Ligand ID 7OK, and PDB 7OZV Ligand ID 16B), the oral prodrug of beta-D-N4-hydroxycytidine (NHC) against SARS-CoV-2 RdRP, was recently elucidated [90]. When used in its active form as NHC triphosphates, molnupiravir is incorporated instead of CTP or UTP during the RNA synthesis. The structure of the RdRP-inhibitor complex demonstrated that the NHC can establish stable base pairs in the RdRP active center with either G or A, therefore explaining how the polymerase circumvents proofreading and creates altered RNA.

Hepacivirial Inhibitors
The allosteric inhibitors of hepaciviruses and flaviviruses restrict interactions between the thumb and fingers subdomains, limiting catalysis and elongation [4]. In vivo studies of HCV RdRP in association with GSK5852 (PDB 4KHM, Ligand ID 1PV) is the first experimentally validated palm II NS5B non-nucleoside inhibitor that has broad-spectrum antiviral activity owing to its N-benzyl boronic acid moiety (IC 50~5 0nM). Although strong in action, this inhibitor requires a large daily dose for effectiveness due to its short plasma half-life (5 h) in human subjects [91]. An earlier work by Chong et al. (2019) sought to find second-generation palm II NS5B inhibitors that prevent the inhibitor's benzylic oxidation, preserving its efficacy while increasing its plasma life [92]. These include compounds 25 (PDB 6MVK, Ligand ID K4J), 31 (PDB 6MVQ, Ligand ID K4M), 45 (PDB 6MVP, Ligand ID K4S), and 49 (GSK8175, PDB 6MVO, Ligand ID K4P). Many of these inhibitors form extensive water-mediated hydrogen bonds with the polymerase except compound 25, which forms direct hydrogen bonds with Arg200 and Tyr448 via its oxazolidinone group and with Asn316 and Ser365 through the methyl carbamoyl moiety ( Figure 11A). The water-mediated hydrogen bonds are formed by the N-phenyl group of compound 45 with residues Arg200, Tyr448, Asn316, Arg386, and Tyr415 ( Figure 11B) and the triazole moiety of compound 31 with residues Arg200 and Tyr448 ( Figure 11C). The oxaborole group of compound 49 replaces the boronic acid of compound 31 and contributes to two of the six water-mediated hydrogen bonds with residues Gly449, Arg386, Asn411, Ser365, and the conserved Arg200 ( Figure 11D). Compound 49, a sulfonamide-Nbenzoxaborole analog, was the most potent inhibitor amongst all the screened derivatives; it showed broad-spectrum activity with a long plasma half-life of about 60-63 h and IC 50 of 16 µM in humans.
The co-crystal structures of HCV RdRP with N-ethylmethylsulfonamide analogs, compound 5 (PDB 5QJ0, Ligand ID J6D) and compound 13 (PDB 5QJ1, Ligand ID J6J), are available from a recent drug screening study that was aimed at solving the P-glycoprotein and BCRP transporter-mediated efflux problems of benzofuran-derived palm site allosteric inhibitors [93]. Both derivatives exhibit comparable binding patterns analogous to compound 49, but the former showed enhanced pharmacokinetic properties and oral bioavailability in studies with rat models. In the crystal structures of the complexes, compound 13 orients towards Gly449, Arg386, and Asn411 due to the presence of an oxaborole moiety, while compound 49 tilts in the opposite direction towards Arg200 and Tyr195 due to interactions with the carbamoyl group in the binding pocket ( Figure 11E,F). Compound 49 was potent against genotype (GT) 1, while compound 13 showed efficacy against GT 2a owing to a hydrogen bond between the ethylamino moiety and the Gln414 of NSP5 [93].

Flaviviral Inhibitors
Recently, the crystal structures of DENV3 RdRP with two non-nucleoside inhibitors, NITD-434 (PDB 6XD0, Ligand ID 6V0M) and NITD-640 (PDB 6XD1, Ligand ID 6V0J), with moderate activity against the polymerase were reported [94]. The IC 50 values ranged from 6 to 17 µM in DENV4 and 0.9 to 13 µM in DENV1 RdRPs. The NITD-434 binds inside the RNA channel, displacing Val603 in motif B, whereas the NITD-640 orders the loops of the fingers domain preceding motif F near the tunnel entrance upon binding ( Figure 12A,B). Most of the interacting residues in the binding pockets of the two compounds are conserved in other flaviviruses such as Japanese encephalitis virus, YFV, West Nile virus, and Zika virus [94].
The derivatives of the NIT29 inhibitor (Ligand ID KY3) that showed preferential binding to an allosteric pocket between the thumb and palm subdomains (PDB 6LD1, 6LD2, 6LD3, 6LD4, 6LD5) were used to screen for potential non-nucleoside inhibitors of Zika viral RdRP [96]. Fluorescence-based alkaline phosphatase-coupled polymerase assays (FAPAs) for initiation and elongation revealed inhibitory concentrations of 44.8 µM and 51.9 µM, respectively. In a cell-based experiment, compounds 13, 14, and 15 possessing the scaffold and aryl sulfonamide moiety displayed modest activity against RdRP and interacted with the residues Arg739, Trp797, Arg731, and Thr796 ( Figure 12E-H). Compound 7, comprising the thiophene group and the acid moiety from the phenyl ring, showed no activity [96].

Exploring NADPH as a Broad-Spectrum Inhibitor
The in vitro titration of the polymerases of EVD68, EV71, and Poliovirus with NADPH resulted in decreased activity where the ligand acted as an allosteric inhibitor rather than a chain terminator during replication [84]. Furthermore, the crystal structure of the EVD68-NADPH complex showed that the interacting residues were conserved in all enteroviruses. This led us to speculate that the ligand could have similar binding and inhibitory properties in the RdRPs of other viruses as well. Hence, we performed blind docking of NADPH (Ligand ID NDP) against RdRPs of select viruses in the apo and template-bound forms using AutoDock Vina 1.1.2. These include HCV (PDB 1NB4, 1NB7), SARS-CoV-2 (PDB 7BTF, 7C2K), Poliovirus (PDB 4K4W, 4R0E), and EV71 (PDB 6KWQ, 3N6L). After protein preparation and removal of water molecules, the ligand-based docking was carried out with default parameters using a genetic algorithm. The NADPH showed a striking similarity in the binding poses that depended solely on the occupancy of the template channel of the RdRP. In the apo structures, the NADPH was bound inside the template channel closer to the pinky of the fingers subdomain, whereas in the template-bound elongation complexes, the ligand was displaced by more than 5 Å away from its original position and towards the thumb subdomain owing to the binding of the RNA (Figure 13). The interaction energies of the NADPH with the RdRP in the apo states were consistently lower than the corresponding elongation complexes in all viruses owing to additional interactions of the ligand with the RNA (Table 3). Intriguingly, the ability of NADPH to effectively engage with the polymerase at different phases of replication is demonstrated by the changes in binding sites. As suggested in the earlier structural and in vivo studies with EVD68 [84], the NADPH binding in other viral RdRPs could exert comparable effects, impairing translocation and catalysis, thus proving its suitability to be developed as a broad-spectrum allosteric inhibitor.

Conclusions and Future Directions
Recent investigations, largely cryo-EM-based, have revealed structural details of the RdRPs bound with cofactors, nucleic acid moieties, analogs, and inhibitors, enhancing our understanding of their mechanism of action. Although the fundamental structure of RdRP has stayed constant over time, there are variations at the individual level that remain unexplored. The association of RdRPs with cofactors that are either self-encoded or derived from the host facilitates the timing of genome replication and transcription. For instance, the in situ structures of reoviral RdRPs demonstrated that the disruption of capsid protein interactions with RdRP during disassembly provided the essential room and energy for transcription to start [69,70]. Likewise, the proteolytic processing of the 3CD of picornaviruses [66] and the association of VPg in caliciviruses [27] aided in the activation of the respective RdRPs. The binding of the P with the L proteins in mononegaviruses [46,47], and the nsp7-9, nsp13, nsp14, and nsp16 with nsp12 in SARS-CoV-2 [38,97], promoted active replication, while the Z protein of the arenaviruses [52] inhibited the synthesis of vRNA. The RdRP complexes with RNA or its analogs offer previously unknown details of structure-function relationships. For example, the re-association of the 3 template RNA at the polymerase surface of FluA following exit protected it against nucleases [77], and the interactions of the NTDs of the pestiviral RdRPs promoted the unwinding of the RNA duplex [65]. The varied nature of the inhibitor binding pockets of RdRPs from distinct viruses necessitates the search for antivirals with unique pharmacophoric properties [14,98]. For example, the benzofuran-, thiazolidinone-, or anilinobenzothiazole-derived inhibitors were effective against HCV, while pyrazole or bibenzylsubstituted piperazine-containing compounds performed well against Zika and the octahydroquinazoline derivatives strongly inhibited DENV RdRP. In addition, our in silico investigations using NADPH demonstrated the possibility of using it as a broad-spectrum antiviral after experimental verifications. The comparison of the catalytic complexes of representative viral RdRPs, including SARS-CoV-2, revealed the broad-spectrum effectiveness of nucleotide analogs with 1' modifications as potential antivirals [89]. Curiously, the SARS-CoV-2 oral drug molnupiravir induced widespread mutagenesis in the viral genome, presenting an unusual antiviral strategy [90]. Thus, the effective modifications of the current antivirals and the discovery of new ones should eventually result from a comprehensive understanding of the dynamics of polymerization, mandating additional research. The end is just the beginning.
Author Contributions: S.V. was involved in the conceptualization, analysis, and writing of the manuscript. K.R. reviewed and described the inhibitor-bound complexes of RdRP. M.R. helped in drafting the section on RdRP complexes with cofactors and nucleic acid and overall editing. S.R. and T.J. were involved in the revision of the manuscript. All authors have read and agreed to the published version of the manuscript.