2'- and 3'-Ribose Modifications of Nucleotide Analogues Establish the Structural Basis to Inhibit the Viral Replication of SARS-CoV-2.

Inhibition of RNA-dependent RNA polymerase (RdRp) by nucleotide analogues with ribose modification provides a promising antiviral strategy for the treatment of SARS-CoV-2. Previous works have shown that remdesivir carrying 1'-substitution can act as a "delayed chain terminator", while nucleotide analogues with 2'-methyl group substitution could immediately terminate the chain extension. However, how the inhibition can be established by the 3'-ribose modification as well as other 2'-ribose modifications is not fully understood. Herein, we have evaluated the potential of several adenosine analogues with 2'- and/or 3'-modifications as obligate chain terminators by comprehensive structural analysis based on extensive molecular dynamics simulations. Our results suggest that 2'-modification couples with the protein environment to affect the structural stability, while 3'-hydrogen substitution inherently exerts "immediate termination" without compromising the structural stability in the active site. Our study provides an alternative promising modification scheme to orientate the further optimization of obligate terminators for SARS-CoV-2 RdRp.

Coronavirus disease 2019 (COVID-19) has caused a global pandemic and has led to over 5.4 million deaths and more than 287 million infections in 196 countries by December 2021.[1] Antiviral agents are under extensive investigation to curb the health crisis caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2).[2−23] RNA-dependent RNA polymerase (RdRp) is one promising drug target due to its necessity for the life cycle of SARS-CoV-2.[7,9,10,24−31] It buries deeply inside the viral capsule and is responsible for viral transcription and replication.[7,13] To perform its functions, RdRp needs to go through the nucleotide addition cycle, which is composed of multiple functional states involving conformational changes of both protein and nucleotides.[32] Inhibitors can be developed to target the functional states or the dynamic transitions in the cycle to impair the processivity of nucleotide addition. As nucleoside triphosphate is the substrate for RdRp to start the extension of the nascent strand in each cycle (Figure A), nucleotide analogues that have similar chemical structures to the cognate substrate but with chemical modifications on either the base or the ribose hold great potential as antiviral drugs that target SARS-CoV-2 RdRp.[5,9,33−40]

Figure 1

(A) Diagram of SARS-CoV-2 RdRp, where ATP binding at the active site is enlarged. (B–H) Chemical structures of six nucleotide analogues in comparison with adenosine, where ribose modifications are highlighted in red, and base modifications are highlighted in blue.

Chain termination is one common inhibitory mechanism exerted by nucleotide analogues to counteract the viral replication and transcription of SARS-CoV-2 RdRp.[9,16,22,32−35,38,41] In this scenario, the inhibitor acts as a chain terminator by hindering the essential check points during the nucleotide addition cycle to terminate the nucleotide addition. For instance, remdesivir is an adenosine analogue with a cyano group attached at the 1′-position of the ribose. Both the experimental and computational studies[7,9,21,27,29,35,37,41−45] have demonstrated that remdesivir could inhibit the translocation through a “delayed” mechanism when it appears at the upstream site of the nascent strand. The efficiency of such termination can be gradually reduced by increasing the NTP concentration, and the full-length RNA product can be yielded.[41,46] In this scenario, the RNA strand with embedded remdesivir was then utilized as a template for synthesizing the second RNA strand, and biochemical experiments have further shown that remdesivir can terminate the nucleotide addition when it is present in the template strand.[46] These studies have suggested that the 1′-cyano substitution on remdesivir plays a deterministic role in chain termination. Sofosbuvir, carbovir, and abacavir are also analogue inhibitors that have 2′- or 3′-ribose modifications and have been proposed as obligate chain terminators for SARS-CoV-2 RdRp in vitro.[33,34,47] They can immediately inhibit the next nucleotide addition once it is incorporated into the nascent strand, and previous work has shown the steric effect caused by the 2′-methyl group substitution could facilitate the immediate termination on SARS-CoV-2 RdRp.[48] The above examples have demonstrated that nucleotide analogues with modification on the ribose could serve as promising inhibitors to terminate the nucleotide addition in SARS-CoV-2 RdRp.

In this work, we have systematically evaluated the inhibitory potential of adenosine analogues that contain 2′- and 3′-ribose modifications (clofarabine, didanosine, fludarabine, vidarabine, 2′-NH2-dA, and 2′,3′-didehydro-2′,3′-dideoxyadenosine, Figure B–H) in SARS-CoV-2 RdRp by extensive molecular dynamics (MD) simulations. Most of them have been approved or investigated for therapeutic applications and inhibitory potential. For example, didanosine is an approved drug that binds to reverse transcriptase and acts as a chain terminator for HIV replication.[49] 2′,3′-didehydro-2′,3′-dideoxyadenosine is an adenosine analogue with the same ribose modifications as stavudine, which has also been approved in antiviral therapy for HIV.[50] Clofarabine and fludarabine have been suggested for the treatment of leukemia[51,52] while vidarabine shows inhibitory activity against herpes simplex virus types 1 (HSV-1), 2 (HSV-2), and varicella-zoster virus.[53] We first performed comprehensive analysis to elucidate how the 2′- and 3′-ribose modifications impact their binding stability in the active site and their incorporation efficiency into the nascent strand. Moreover, we investigated their capability to inhibit the next nucleotide addition when they are present at the 3′-terminal of the nascent strand. On the basis of these findings, we have further proposed that cordycepin may serve as a promising inhibitor. Our study provides valuable insights to orientate further optimization and design of nucleotide analogues to target SARS-CoV-2 RdRp.

An effective chain terminator requires that the nucleotide analogue could be efficiently incorporated into the nascent strand. Therefore, it is essential for the analogue inhibitor to maintain a catalytically active conformation for the efficient incorporation. In this regard, we first calculated the distance between the Pα atom of nucleotide analogues and the O3′ atom of the nucleotide at the 3′-terminal of the nascent strand. Such a distance is required to be below 4 Å for efficient phosphodiester bond formation during catalysis.[54] The population of the Pα–O3′ distance within 4 Å was computed based on the extensive MD conformations with nucleotide analogues adopting the triphosphate (TP) form in the closed active site of SARS-CoV-2 RdRp, and they were compared with that when ATP is occupying the closed active site (Figure A). For each nucleotide analogue, 20 replicas of 50 ns MD simulations were performed (see SI Section 1 for details), and the conformations after first 20 ns were collected for the subsequent analysis (see SI Section 2 for details). Our results demonstrate that the population of the Pα–O3′ distance within 4 Å for vidarabine-TP and 2′-NH2-dATP resembles that for ATP (Figure B), suggesting that vidarabine and 2′-NH2-dATP exhibit a considerable incorporation capability as the cognate substrate. Didanosine-TP and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP also possess a notable population (49.0 ± 7.0% and 42.7 ± 7.9%, respectively) of catalytically active conformations, although the population is marginally reduced compared to that of ATP (66.9 ± 6.3%). This suggests that didanosine-TP and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP are also prone to be incorporated into the nascent strand and satisfy the prerequisite as a chain terminator. This observation could have suggested that the removal of 3′-hydroxyl group as shown in didanosine and 2′,3′-didehydro-2′,3′-dideoxyadenosine seldom disrupt the active conformation in the precatalytic state. In sharp contrast, clofarabine-TP and fludarabine-TP have an obviously lower population of catalytically active configurations (Figure B), indicating that they could be difficult to be incorporated into the nascent strand. It is worthy to note that we have extended the simulations to 100 ns for SARS-CoV-2 RdRp with clofarabine-TP, didanosine-TP, fludarabine-TP, or vidarabine-TP in the active site, and the results from 60, 70, 80, 90, and 100 ns have also been obtained, which show a similar tendency as observed in the 20 × 50 ns simulations and thus validate the convergence of our simulations (Figure S1).

Figure 2

Examinations of the incorporation capability of nucleotide analogues. (A) Diagram showing the distance between the Pα atom of analogues and the O3′ atom of the 3′-terminal of the nascent strand. (B) Population of MD conformations showing the Pα–O3′ distance within 4 Å for adenosine (ATP), clofarabine (COP), didanosine (DIP), fludarabine (FLP), vidarabine (VDP), 2′-NH2-dA (BNP) and 2′,3′-didehydro-2′,3′-dideoxyadenosine (STP) in the triphosphate form. (C) Diagram of base pairing between ATP/nucleotide analogue and template nucleotide at the active site. (D) Hydrogen bonding probability for base pairing at the active site. (E) Diagram of root-mean-square fluctuation (RMSF) of ATP and nucleotide analogues at the active site. Heavy atoms were included for the RMSF calculations. (F) RMSF of ATP and nucleotide analogues at the active site. (G) The binding free energies of nucleotide analogues relative to that of ATP at the active site of SARS-CoV-2 RdRp calculated by the free energy perturbation method. In (B), (D), and (F), conformations from 20 × 50 ns simulations were used for the analysis.

The stability of nucleotide analogues in the active site is another vital indicator anchoring its inhibitory ability during catalysis and thereby plays a key role in evaluating its incorporation efficiency. In this regard, we examined the base pairing stability between the analogues and the template nucleotide (Figure C,D) as well as the structural flexibility of nucleotide analogues at the active site (Figure E,F). We found that didanosine-TP, 2′-NH2-dATP, and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP can maintain stable base pairing with the template uridine (hydrogen bond probability = 87.4 ± 3.0%, 93.4 ± 1.0%, and 88.3 ± 1.4%, respectively), which is similar to the base pairing stability with ATP at the active site (hydrogen bond probability = 87.1 ± 3.8%). Consistently, the structural fluctuations of didanosine-TP, 2′-NH2-dATP, and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP (root-mean-square fluctuations (RMSF) = 0.919 ± 0.040 Å, 0.795 ± 0.032 Å, and 0.862 ± 0.039 Å, respectively) are as small as that of ATP (RMSF = 0.997 ± 0.052 Å) (Figure E,F). These observations suggest that these three analogues adopt a comparable structural stability with ATP at the active site of SARS-CoV-2 RdRp, which further consolidates an efficient incorporation of them into the nascent strand. On the contrary, vidarabine-TP, which shows a similar population of MD conformations with the Pα–O3′ distance less than 4 Å as ATP (Figure B), has demonstrated reduced base pairing stability and higher structural flexibility (Figure D,F). Clofarabine-TP and fludarabine-TP both demonstrated obviously weakened stability at the active site compared to ATP (Figure D,F), which reduced their potential as effective inhibitors for SARS-CoV-2 RdRp. Moreover, we examined the binding affinity of nucleotide analogues by computing the ΔΔGbinding relative to the cognate substrate ATP (Figure G) using the free energy perturbation method (Figure S2, see SI Section 3 for details). We found that all the six analogues have a comparable binding affinity with ATP, although clofarabine (ΔΔGbinding = 2.4 kcal/mol) and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP (ΔΔGbinding = 2.6 kcal/mol) are less competitive than the other four analogues (ΔΔGbinding < 2 kcal/mol). Altogether, our above analysis for the structural stability and the binding affinity at the active site demonstrates that didanosine-TP, 2′-NH2-dATP, and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP possess the most pronounced conformational stability in the active site.

After comprehensive structural investigations into the interaction between the nucleotide analogues and the protein environment (Figures S3–S5, see SI Section 2 for details), we found that the interplay between the ribose modification and Asp623 could dominate the structural stability of nucleotide analogues. When ATP occupies the active site, its 2′-hydroxyl group can form hydrogen bonds with the side chain of Asp623 with a probability of 55.8 ± 9.1% (Figure S3B). 2′-NH2-dATP that substitutes the 2′-hydroxyl group with 2′-amino group can maintain the hydrogen with Asp623 (hydrogen bonding probability of 39.2 ± 9.9%) as ATP. However, when clofarabine-TP binds at the active site, its 2′-ribose fluorine substitution is electrostatically repelled by the negatively charged side chain of Asp623. This leads to distortion of the ribose ring in clofarabine-TP (Figure S6A) as shown by the variations of the dihedral angle between the ribose and backbone of Asp623 (Figure A). Such a dihedral angle for clofarabine-TP is concentrated at ∼53 deg (Figure B), obviously larger than that of ATP (∼25 deg). Both fludarabine-TP and vidarabine-TP have a 2′-hydroxyl group at the opposite side of the ribose, which can form stronger hydrogen bonding interactions with Asp623 (Figure S3B) and thereby induces the conformational change of the ribose (Figure D,E and Figure S6C,D). There is no specific interaction between the side chain of Asp623 and the ribose of didanosine-TP as well as 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP due to the absence of hydroxyl groups at both 2′- and 3′-ribose positions (Figure D,H). Under such a circumstance, the side chain of Asp623 mainly interacts with Lys621, which is also observed when ATP is located (Figure S7). Therefore, the ribose orientation in didanosine-TP and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP relative to Asp623 is similar to that in ATP (Figure C,G Figure S6B,F). It is worthy to note that the ribose configuration could also be affected by base modifications. For example, the fluorine substitution in fludarabine could introduce extra electrostatic repulsion with the base of the template. This unfavorable interaction would push the base of fludarabine away from the template nucleotide (Figure S8D) and propagate to further affect the ribose configuration and the overall structural stability (Figures and 3D). This also rationalizes why vidarabine-TP demonstrates more stable binding than fludarabine-TP in the precatalytic state (Figure ).

Figure 3

Investigations on the relative orientation between the backbone of Asp623 and the ribose of nucleotide analogues at the active site. (A) Cartoon model showing the dihedral angles between the ribose plane of nucleotide analogues and the plane of Asp623 backbone. The ribose plane was defined by O4′, C4′, and C1′ atoms, and the plane of the Asp623 backbone was defined by C, CA, and N atoms. (B–G) Distributions of the dihedral angle as described in (A) for the six nucleotide analogues (clofarabine, didanosine, fludarabine, vidarabine, 2′-NH2-dA, and 2′,3′-didehydro-2′,3′-dideoxyadenosine) in the triphosphate form. In (B–G), conformations from 20 × 50 ns simulations were used for the analysis.

The above examinations suggest that didanosine-TP, 2′-NH2-dATP, and 2′,3′-didehydro-2′,3′-dideoxyadenosine-TP demonstrate higher structural stability at the active site than other analogues, while all three of them together with vidarabine-TP can maintain the catalytically active conformations at the active site of SARS-CoV-2 RdRp for efficient incorporation into the nascent strand. After incorporation into the nascent strand at the i site, the nucleotide analogue will move upstream by one base position to reach the i+1 site (Figure ). The effective obligate terminator should be able to immediately inhibit the next nucleotide addition after its incorporation. Therefore, it is necessary to further examine the substrate incorporation efficiency when the analogue is present at the i+1 site of the nascent strand. As the O3′ atom of the 3′-terminal of the nascent strand is required to form a phosphodiester bond with the Pα atom of the substrate during the catalysis for nucleotide addition, didanosine and 2′,3′-didehydro-2′,3′-dideoxyadenosine in which the 3′-hydroxyl group is substituted by one hydrogen atom (Figure D,H) would inherently abolish the next nucleotide addition when it appears at the 3′-terminal of the nascent strand. For the remaining four nucleotide analogues including clofarabine, fludarabine, vidarabine, and 2′-NH2-dA with a hydroxyl group attached at 3′-ribose as adenosine (Figure C,E–G), we modeled each nucleotide analogue at the 3′-terminal of the nascent strand of SARS-CoV-2 RdRp and performed twenty 50–100 ns MD simulations with different random seeds (see SI Section 1 for details). To assess their inhibitory effect on the nucleotide addition at the active site, we computed the distance between the Pα atom of NTP and the O3′ atom of the nucleotide analogue at the 3′-terminal of the nascent strand (Figure A). We found that the catalytically active conformation is well maintained, as the population of MD conformations with the Pα–O3′ distance within 4 Å is 94.7% ± 2.1%, 92.8% ± 4.7%, 98.5% ± 0.8%, and 75.1 ± 5.6% for clofarabine, fludarabine, vidarabine, and 2′-NH2-dA at the 3′-terminal of the nascent strand, respectively (Figure D). We also found that the stability of NTP and the base pairing with the template nucleotide at the active site is well maintained except for clofarabine (Figure E–F), in which the fluorine substitution in the ribose may repel the O4′ in the ribose of NTP and weaken its stability (Figure S9). Overall, these results suggest that NTP incorporation is not effectively inhibited when fludarabine, vidarabine, or 2′-NH2-dA is present at the 3′-terminal of the nascent strand. Although clofarabine’s incorporation may partially weaken the next NTP binding in the active site, its unfavorable binding stability at the active site (Figure ) has reduced its potential as an effective immediate terminator in the first place.

Figure 4

Investigation of the inhibitory effect of four analogues (clofarabine (CO3), fludarabine (FL3), vidarabine (VD3), and 2′-NH2-dA (BN3)) at the i+1 site (filled in gray in (A–C)) on the nucleotide addition at the active site (i site). (A) Diagram showing the distance between the Pα atom of ATP at the active site and the O3′ atom of the analogue in the 3′-terminal of the nascent strand. (B) Diagram of base pairing between ATP and template nucleotide at the active site when the analogue has been incorporated into the nascent strand. (C) Diagram of RMSF of ATP at the active site when the analogue has been incorporated into the nascent strand. Heavy atoms were included for the RMSF calculations. (D) Population of MD conformation with a Pα–O3′ distance within 4 Å. (E) Hydrogen bonding probability for base pairing at the active site. (F) RMSF of ATP at the active site when the analogue has been incorporated into the 3′-terminal of the nascent strand. In (D–F), the conformations from the 20 × 100 ns simulations were used for AD3, CO3, FL3, and VD3, while those from 20 × 50 ns simulations were used for BN3.

According to the above observations, didanosine and 2′,3′-didehydro-2′,3′-dideoxyadenosine surpass the other four nucleotide analogues to become the most promising immediate terminators due to their merits in maintaining a stable and catalytically active conformation in the active site for the efficient incorporation into the nascent strand, as well as its inherent capability to terminate the next nucleotide addition. On the contrary, vidarabine and 2′-NH2-dA show binding stability in the active site; however, they could not efficiently inhibit the incorporation of the next substrate. Clofarabine and fludarabine could not form stable binding at the active site, while they only have a minor effect on the next nucleotide incorporation into the nascent strand. Therefore, didanosine and 2′,3′-didehydro-2′,3′-dideoxyadenosine are the most possible obligate terminators among the six nucleotide analogues under investigation. These findings are consistent with previous experimental results that have suggested the uridine analogue of 2′-NH2-dATP has no termination effect, while the thymidine analogue of 2′,3′-didehydro-2′,3′-dideoxyadenosine can terminate RNA synthesis in SARS-CoV-2 RdRp.[33] Moreover, previous work developing bioinformatics pipeline based on single-cell RNA sequencing data has suggested that didanosine can effectively inhibit nucleotide addition in SARS-CoV-2 RdRp.[55]

On the basis of the above analysis and the molecular mechanisms that underline the mutual interaction between ribose and Asp623, we have proposed nucleotide analogue, which removes the 3′-hydroxyl group while possesses a 2′-ribose modification that would not perturb the Asp623’s side chain, may hold great potential as an effective obligate terminator. In this regard, we screened through the purine nucleotide analogues and expect that cordycepin, which is cytotoxicologically inactive and has high level of biosafety by the analysis using data from the Prediction of Toxicity of Chemicals (ProTox-II) Virtual Laboratory,[56,57] may serve the purpose (Figure A). On one hand, cordycepin replaces the 3′-hydroxyl group with a hydrogen atom and would naturally inhibit the next nucleotide addition after its incorporation into the nascent strand. On the other hand, cordycepin keeps the 2′-hydroxyl group and thereby would maintain the interaction with Asp623. To evaluate our expectation, we performed twenty 100 ns MD simulations for SARS-CoV-2 RdRp with cordycepin-TP binding at the active site. Our calculations indicate that cordycepin-TP has sampled ∼54% catalytically active conformation and suggests a reasonable incorporation efficiency similar to that of ATP (Figure B). Calculations of ΔΔGbinding of cordycepin relative to ATP indicates that cordycepin-TP has a comparable binding affinity with ATP (ΔΔGbinding = −0.767 ± 0.437 kcal/mol). Furthermore, we observed that the relative orientation between the ribose of cordycepin-TP and Asp623 resembles that when ATP is at the active site (Figure C). The base pairing stability (83.4 ± 3.5%) and RMSF (0.99 ± 0.05 Å) of cordycepin-TP is also comparable with that of ATP (base pairing stability = 90.5 ± 3.2% and RMSF = 0.97 ± 0.05 Å). These observations further consolidate that maintaining the relative orientation between ribose and Asp623 is important for ensuring a normal ribose configuration and structural stability in the active site. Moreover, the results of cordycepin again pinpoint that removal of the 3′-hydroxyl group can assign the nucleotide analogue the inherent inhibitory capability, while it does not perturb the structural stability of substrate at the active site of SARS-CoV-2 RdRp. Interestingly, after we submited this manuscript, a recent work was been published,[56] which has supported our proposal and indicated that cordycepin can indeed inhibit the viral replication in SARS-CoV-2.

Figure 5

Investigations on the stability and the incorporation efficiency of cordycepin-TP at the active site of SARS-CoV-2 RdRp. (A) Chemical structure of cordycepin with the ribose modification highlighted in red. (B) Histogram of the Pα–O3′ distance for cordycepin-TP (in orange) at the active site compared with that for ATP (in gray). (C) Histogram of the dihedral angle between the plane of the Asp623 backbone and the ribose plane of cordycepin-TP (in orange), which is compared with the scenario when ATP is at the active site (in gray). In (B) and (C), conformations from 20 × 100 ns simulations were used.

To further validate our computational model and protocol adopted in the current study, we have also performed analysis for the well-characterized adenosine analogue Remdesivir based on our previous simulations[42] as a control (see SI Section 4 for details). Our analysis has well reproduced the previous experimental and computational results that remdesivir-TP can be incorporated into the nascent strand,[26,35,41,58] suggesting the robustness of our current protocol to estimate the incorporation capability of nucleotide analogues. The similar performance of remdesivir-TP with that of cordycepin-TP and didanosine-TP has further consolidated our proposal that cordycepin-TP and didanosine-TP could be added into the nascent strand. It is worthy to note that the selectivity of the nucleotide analogue in the host polymerase is also important for evaluating the efficacy and the development of nucleotide analogues to counteract the viral infection. In this regard, further investigations about the performance of nucleotide analogues in host polymerase are required to achieve a more comprehensive evaluation of the efficacy of nucleotide analogues.

In this study, we have systematically investigated the inhibitory effect of nucleotide analogues possessing 2′- or 3′-ribose modifications targeting at SARS-CoV-2 RdRp. We have elucidated that the 2′-ribose modification has an obvious response from the environment and accordingly affects the overall stability as well as the incorporation efficiency. Moreover, substitution of the 3′-hydroxyl group by one hydrogen atom would render the nucleotide analogue’s inherent capability to inhibit the next nucleotide addition while maintaining a stable precatalytic conformation. On the basis of the above extensive analysis by estimating the incorporation probability and capability to inhibit the next nucleotide addition, we have proposed that cordycepin and didanosine could hold great potential to hamper the nucleotide addition in the active site. Although a limited number of analogues investigated herein may not be sufficient to reach general conclusions, our study has provided valuable molecular insights into how the 2′- and 3′-ribose modifications can alter the inhibition effect, which could orientate the rational design of nucleotide analogues targeting SARS-CoV-2 RdRp.