HIV-1 with HBV-associated Q151M substitution in RT becomes highly susceptible to entecavir: structural insights into HBV-RT inhibition by entecavir

Hepatitis B virus (HBV) reverse transcriptase (RT) is essential for viral replication and is an important drug target. Nonetheless, the notorious insolubility of HBV RT has hindered experimental structural studies and structure-based drug design. Here, we demonstrate that a Q151M substitution alone at the nucleotide-binding site (N-site) of human immunodeficiency virus type-1 (HIV-1) RT renders HIV-1 highly sensitive to entecavir (ETV), a potent nucleoside analogue RT inhibitor (NRTI) against HBV. The results suggest that Met151 forms a transient hydrophobic interaction with the cyclopentyl methylene of ETV, a characteristic hydrophobic moiety of ETV. We thus solved the crystal structures of HIV-1 RTQ151M:DNA complex with bound dGTP or ETV-triphosphate (ETV-TP). The structures revealed that ETV-TP is accommodated at the N-site slightly apart from the ribose ring of the 3′-end nucleotide, compared to the position of bound dGTP and previously reported NRTI/dNTP. In addition, the protruding methylene group of bound ETV-TP directly pushes the side-chain of Met184 backward. Met184 is a key residue that confers ETV resistance upon substitution with smaller Ile/Val. These results provide novel insights into NRTI binding to the N-site and further provide important clues for the development of novel anti-HBV/HIV-1 RT inhibitors to overcome critical drug resistance.

. Design of HIV-1 RT mutants mimicking the N-site of HBV RT in this study. (a) The residues lying within 7 Å of the deoxyribonucleoside moiety of the bound NRTI/dNTP determined by previously reported ternary complexes of HIV-1 RT. This diagram was drawn using the structure of HIV-1 RT WT in complex with DNA and dATP (PDB code, 5TXL) 11 . The deoxyadenosine moiety of the bound dATP (DAP) is shown in stick model. The unconserved 8 residues selected for mutational analysis are colored in blue (Ile63 and Leu74), green (Gly112, Asp113, Tyr115 and Phe116), yellow (Gln151) and magenta (Phe160). (b) The 20 selected residues in HIV-1 RT and the corresponding residues in HBV RT. The 8 residues selected for mutational analysis are marked with the same color scheme as in (a). Coincidences with NRTI resistance of HIV-1 RT are also indicated. (c) Amino acid sequence alignment between HIV-1 RT and HBV RT. HIV-1 and HBV RT domains were extracted and used for the alignment. The identical, strongly similar, and weakly similar residues are colored in red, green, and blue, respectively. The residues selected for mutational analysis in this work are highlighted in yellow. The conserved motifs (β2-β3 and motifs A, B, C, D, and E) are indicated. The color bars above the sequence indicate the 20 selected residues with the same color scheme as in (a). The residues involving ETV resistance in HBV RT are boxed. Residue stacking at the ribose ring of primer DNA at the N + 3 position shown in Fig. 6 is highlighted in green.
SCIENTIfIC RepoRts | (2018) 8:1624 | DOI: 10.1038/s41598-018-19602-9 combination therapies against chronic HBV infections to hamper ETV-resistant HBV mutant selection 17,18 . It is important to elucidate the interatomic interactions between ETV-triphosphate (ETV-TP) and HBV RT by X-ray crystallography to understand the structural mechanism of HBV RT inhibition and the reported ETV resistance in detail, which may also be useful for novel HBV RT inhibitor design. Until now, docking simulation studies with homology models based on the HIV-1 RT structure have been conducted to explain the putative interactions between ETV-TP and HBV RT 18,19 .
In this study, we showed that the Q151M mutation of HIV-1 RT confers high ETV and EFdA sensitivity upon HIV-1. The HIV-1 RT Q151M further enabled crystal structure analysis of HIV-1 RT in complex with ETV-TP. Based on the structures of RT Q151M together with the results of the antiviral assay, possible mechanisms of ETV-TP action on HIV-1/HBV RT and of the reported ETV resistance are discussed.

Results and Discussion
Viral replication and antiviral assay. Based on the previously reported ternary complex structures of HIV-1 RT, a total of 20 amino acid residues lying within 7 Å of the deoxyribonucleoside moiety of the bound NRTI/ dNTP were extracted (Fig. 1). Among the selected residues, 8 were not conserved in HBV RT; 6 were located at the N-site of HIV-1 RT, whereas the other two were located at the β2-β3 region of the finger domain and stabilize the base moiety of the template DNA for base-pairing with incoming NRTI. These 8 unconserved residues were finally selected for the HBV RT-mimicking mutational strategy ( Fig. 1 and Supplementary Fig. S1). Of these 8 mutation candidates, we reasoned that substitution of the hydrophilic Gln151 for the bulky and hydrophobic methionine (Q151M) would be the most significant, while the remaining 7 candidates were categorized as mutations with similar side-chain properties. It should also be noted that Q151M is widely known as a multi-NRTI-resistant mutation in HIV-1 RT, as described later. We created 5 HBV RT-mimicking HIV-1 RT mutants in this study: RT Q151M , RT Q151M/F160L , RT Q151M/G112S/D113A , RT Q151M/Y115F/F116Y , and RT Q151M/I63V/L74V (Table 1). Wild-type HIV-1 (HIV-1 WT ) and the 5 HIV-1 variants carrying RT N-site mutations (e.g., HIV-1 Q151M ) were employed and their replication examined. Their replication kinetics suggested that although HIV-1 Q151M/F160L completely failed to replicate when cultured up to 9 days, the other 4 variants successfully propagated ( Supplementary Fig. S2). HIV-1 Q151M , HIV-1 Q151M/Y115F/F116Y , and HIV-1 Q151M/I63V/L74V propagated slower than did HIV-1 WT , but HIV-1 Q151M/Y115F/F116Y reached similar or slightly greater supernatant p24 levels on day 9 than did HIV-1 WT . These HIV-1 variants were considered able to propagate, but their replication was somewhat slower than that of HIV-1 WT . Notably, the magnitude of enzymatic activity of RT Q151M was almost identical to that of RT WT (Supplementary Table S1), and further investigation will be needed to explain the slight difference between HIV-1 RT activity and the propagation profile of HIV-1 with the Q151M mutation. In contrast, the propagation of HIV-1 Q151M/G112S/D113A was relatively slow, with a day 9 viral level approximately 40% that of HIV-1 WT . Taken together, the results implied that the F160L mutation was exceptionally harmful to viral replication.
Next, we examined the antiviral activity of these NRTIs against HIV-1 Q151M , HIV-1 Q151M/Y115F/F116Y , and HIV-1 Q151M/I63V/L74V . As Shirasaka et al. and Harada et al. reported, HIV-1 RT Q151M, known as a multi-NRTI-resistant mutation, usually occurs in combination with accessory mutations, including A62V, V75I, F77L, and F116Y (Q151M-complex). The Q151M-complex causes high-level resistance to AZT and certain NRTIs with the dideoxy configuration, such as ddC and ddI 10,22 . Notably, Q151M is also an important AZT-resistant mutation to HIV-2 RT 23,24 . We confirmed that HIV-1 Q151M also caused moderate-level resistance to AZT (~5-fold); the combination of Q151M and F116Y decreased AZT sensitivity further, as expected (IC 50 : 0.2 μM) ( Table 2). Notably, mutations including Q151M and some accessory mutations are also reportedly cause intermediate-level resistance to 3TC and TDF 25 ; however, such resistance was not caused by any variants tested in this study (Table 2). Critically, EFdA activity against HIV-1 Q151M and HIV-1 Q151M/Y115F/F116Y increased (IC 50 : 30 pM for both variants) when compared to that against HIV-1 WT (IC 50 : 0.4 nM). Furthermore, the anti-HIV-1 activity of ETV drastically improved in the presence of Q151M and additional mutations Y115F and F116Y; the IC 50 numbers of ETV against HIV-1 Q151M and HIV-1 Q151M/Y115F/F116Y decreased to 42 nM and 13 nM, respectively (26-fold and 85-fold less than that of HIV-1 WT ). This suggested that the Q151M mutation was strongly associated with ETV-TP binding to the N-site of HIV-1 RT, with additional Y115F/F116Y mutations further enhancing the ETV sensitivity of HIV-1 Q151M . Considering the results of the viral replication and antiviral assays, we reasoned that HIV-1 RT Q151M would be a suitable candidate for crystallographic study to elucidate the HIV-1 RT N-site structure in the ETV-TP binding state.
Structure determination of RT Q151M binary and ternary complex. Recombinant HIV-1 RT WT and RT Q151M were produced by Escherichia coli and purified by Ni-affinity and ion-exchanging chromatography (see Methods). The RT assay for the purified sample by enzyme-linked immune-sorbent analysis (ELISA) demonstrated that RT Q151M and RT WT showed similar activity levels (Supplementary Table S1). Quantitation by polymerase chain reaction (PCR) using viral supernatant of 293 T cells showed that RT Q151M activity was approximately 77% that of RT WT (Supplementary Table S1). These results indicated that the Q151M mutation does not significantly affect the enzymatic activity of HIV-1 RT. For crystallization of the RT Q151M :DNA binary complex, the previously developed 38-mer DNA aptamer was applied, albeit with 3 base substitutions to trap guanosine analogue ETV-TP at the N-site (Fig. 3a) 26,27 . After mixing RT Q151M and the DNA aptamer, gel-filtration chromatography was applied to separate unbound DNA and the RT Q151M :DNA complex. Native polyacrylamide gel electrophoresis (PAGE) analysis showed that the peak fractions contained no DNA-free RT Q151M , indicating that the RT Q151M :DNA was stable and well equilibrated ( Supplementary  Fig. S3). The structure of the RT Q151M :DNA binary complex was determined by molecular replacement with the previously reported RT WT :DNA binary complex (PDB code, 5D3G) 27 as a search model and refined to 2.6 Å resolution ( Fig. 3b and c). There are two RT Q151M heterodimers (p66/p51) in the asymmetric unit of the rhombohedral R3 lattice. The final refined model contains amino acid residues 3-553 for chains A/C (p66 subunit), 4-427 for chains B/D (p51 subunit) except 214-230 owing to poor electron density, and two DNA aptamers (nucleotides from −1 to 33 for chain E and from −4 to 33 for chain F). In the previously reported RT WT :DNA structure, the side-chain of Gln151 interacted with the adenine base of the dA0 via ordered solvent molecules, stabilizing base stacking with  the cytosine base of dC1 27 . In contrast, the present structure showed that the hydrophobic side-chain of Met151 keeps nearby residues, the base moiety of dC0, and solvent away, and that the cytosine base of dC0 is deviated by approximately 35° relative to the guanine base of dG1. Consequently, the present structure of the RT Q151M :DNA binary complex exhibits a more open conformation than that of the previously reported RT WT :DNA structure ( Supplementary Fig. S4). These results indicated that the bulky, hydrophobic Met151 side-chain increases the flexibility of the N-site and forms a distinct conformational state in the absence of dNTP/NRTI. RT Q151M :DNA:dGTP and RT Q151M :DNA:ETV-TP ternary complex crystals were prepared by soaking the crystals of the binary complex in the crystallization mother liquor supplemented with dGTP/ETV-TP. The structures of the RT Q151M :DNA:dGTP and RT Q151M :DNA:ETV-TP ternary complexes were determined at 2.38 Å and 2.45 Å resolution, respectively. The obvious electron density was observed for the bound dGTP/ETV-TP and Mg 2+ (Fig. 4). The two heterodimers in the asymmetric unit are well superimposed on one another with main-chain root-mean-square deviation (RMSD) below 1.0 Å. In contrast, apparent local conformational changes were observed in the β2-β3 region of the finger domain between the binary and ternary complexes with maximum atomic displacement of approximately 8.5 Å (Fig. 3c). The Arg72 and Lys65 side-chains, derived from the β2-β3 region, interact with the bound ETV-TP to stabilize the ternary complex during catalytic reactions (Fig. 5). In RT WT structures, a hydrogen bond is formed between the side-chain of Gln151 and Arg72, whereas the thioether group of Met151 in the RT Q151M structures occupies roughly the same position as Gln151 but keeps the distances of van der Waals contacts with nearby Arg72, Phe115, and C2′ of dGTP/ETV-TP (3.6−4.0 Å). The same observations have also been made in the recently reported structures of HIV-1 RT Q151M and RT Q151M-complex ternary complexes 11 . The structures reported in this work have been determined in the rhombohedral lattice system for the first time, whereas the structures of RT Q151M :DNA:dGTP/ETV-TP basically exhibit a typical closed conformation of the HIV-1 RT ternary complex, and thus superimpose well on previously reported structures of the dNTP/ NRTI ternary complex with main-chain RMSD of ~1.2 Å 11,28,29 . These results suggest that the lattice forces and the inter-molecular crystal contacts do not affect the overall structural conformation of HIV-1 RT ternary complex.

Mutants
Structural description of the RT Q151M N-site occupied by ETV-TP. The overall conformation of the present ternary complex and the position and orientation of bound dGTP/ETV-TP are similar to those in previously reported HIV-1 RT:dNTP/NRTI complexes, whereas slight but significant structural differences at the N-site occupied by ETV-TP were also detected. First, the ribose-analogous cyclopentane ring of the bound ETV-TP lies slightly apart from the 3′-end nucleotide of the primer DNA. Second, relocation of the side-chain of Met184 occurs by pressing with the protruded exocyclic methylene group of the ETV-TP (Fig. 5, Supplementary  Figs S5 and S6). The side-chains of Gln91 and Gln161 also moved in response to the relocation of the side-chain of Met184, retaining appropriate interatomic distances. We selected and listed two interatomic distances in Supplementary Table S2 as representative of the observed differences in the present structure of the ETV-TP ternary complex. In previous docking simulation studies, the exocyclic methylene of ETV-TP was predicted to be fitted with the small hydrophobic pocket created by Ala87, Phe88, Pro177, Leu180, and Met204 of HBV RT 19 . The corresponding hydrophobic pocket is observed in the present structure of the ETV-TP complex formed by Ala114, Tyr115, Pro157, Phe160, and Met184 of HIV-1 RT. However, the methylene did not exploit the pocket, but directly pushed the Met184 side chain (Fig. 5, Supplementary Figs S5 and S6). It is thus likely that the side-chain thioether of the Met184 plays a crucial role in ETV-TP acceptance through hydrophobic interaction with its methylene group, a characteristic moiety of ETV. Notably, Met184 of HIV-1 RT corresponds to Met204 of HBV RT, with M204I/V being a critical amino-acid substitution in acquiring ETV resistance. HIV-1 RT M184V was also detected from HIV-HBV co-infected patients receiving ETV monotherapy, and in vitro experiments further demonstrated that M184V of HIV-1 RT confers resistance to ETV 30 . The substitution of the bulky thioether group with a smaller isopropyl group could diminish contacts with methylene, leading to decreased binding affinity of ETV-TP to the N-site. Based on these results, we propose that modification of the ETV methylene moiety may represent a strategic candidate to develop new anti-viral agents for overcoming ETV resistance.
Multi-step mechanism of RT inhibition by ETV-TP. Although single Q151M substitution leads to a drastic increase in ETV sensitivity, the current structure of the ternary complex revealed no direct interactions between the Met151 side-chain and the exocyclic methylene of ETV-TP. It is likely that Gln151 does not impede the binding of ETV-TP to the N-site. These structural aspects as well as the results of antiviral assays strongly suggest that Q151M contributes not to the tight binding of ETV-TP to the N-site as observed in the crystal structure analysis, but to the process of ETV-TP entry into the N-site. Gln/Met151 is located at the entrance of the N-site and likely acts as a lid (Supplementary Fig. S4). Notably, Q151M also intensified sensitivity against EFdA by approximately 13-fold. EFdA has a characteristic ethynyl group at the C4′ position. The previously reported structure of the HIV-1 RT:DNA:EFdA-TP ternary complex revealed that the 4′-ethynyl is bound at a hydrophobic pocket created by the Ala114, Tyr115, Phe160, and Met184 side-chains 28 , yet does not directly interact with Gln151. ETV and EFdA share an increased hydrophobicity conferred by a methylene and ethynyl group, respectively. Therefore, we presumed that Met151 provides the primary binding site for ETV-TP/EFdA-TP through transient hydrophobic interactions, prior to the stable ternary complex formation observed in the crystal structures. The primary binding at Met151 might also stimulate local conformational change of the finger domain, thereby decreasing the energy barrier for entry of ETV-TP into the N-site. Collectively, we propose a three-step mechanism of ETV-TP action on HIV-1 RT: (i) primary binding to HIV-1 RT at Met151 via transient hydrophobic interactions; (ii) tight binding to N-site accepted by relocated Met184 side-chain, allowing incorporation of ETV-MP into primer DNA; and (iii) delayed chain termination at the N + 3 position as reported previously (Fig. 6) 16 . There might be a similar multi-step mechanism of ETV-TP action on HBV RT. In recent structural studies of the HIV-1 RT Q151M and RT Q151M-complex , a multi-NRTI resistance mechanism was proposed: the Q151M mutation increases the perturbation of the N-site by removal of the hydrogen bond between Gln151 and Arg72, leading to decreased incorporation of 3′-exo-forming dideoxy NRTIs 11 . We further propose the additional possibility that Gln/Met151 plays a role as a selector in response to the hydrophobicity of incoming NRTI. For example, Q151M confers resistance against AZT by 5-fold 31 . AZT has a highly polarized azide group at the C3′ position, which might not conform to the primary binding site provided by hydrophobic Met151, thus hindering the entry of AZT to the N-site and consequently decreasing binding affinity. Because the amino acid residues forming the N-site are identical between HIV-1 and HIV-2 RT, similar mechanisms, i.e., N-site perturbation and increased hydrophobicity caused by the Q151M mutation, likely underlie the AZT resistance of HIV-2 RT Q151M . Crystal structure analysis of HIV-2 RT Q151M :DNA:NRTI would be needed to discuss the NRTI-resistance mechanism of HIV-2 RT more precisely.

Insights into the NRTI sensitivity and N-site structure of HBV RT. The structural information on
HIV-1 RT Q151M together with the results of the antiviral assays enable us to further discuss the individual role of each amino acid residue at or near the N-site in affecting NRTI sensitivity and resistance. The antiviral assay revealed that the Y115F and F116Y mutations increased the ETV sensitivity of HIV-1 Q151M by 3-fold. A recent structural study of HIV-1 RT Q151M-complex suggested that the F116Y mutation improves polymerization fitness, which might be caused by N-site stabilization through hydrogen bond formation between the side-chain hydroxyl of Tyr116 and the main-chain O of Lys73 11 . The same stabilization effect is expected when ETV-TP is bound to the N-site of HIV-1 RT Q151M/Y115F/F116Y , thus leading to the increase of ETV sensitivity. In addition, Phe115 might be more suitable for stacking interaction with the cyclopentyl ring of ETV-TP than is Tyr115, which would also increases ETV-TP binding affinity. In contrast to Y115F/F116Y, the I63V/L74V mutations markedly decreased ETV sensitivity. These two residues are located near the cytosine base of dC0, and their side chains play a role in stabilizing base-pair formation with the guanine base of ETV-TP. I63V/L74V thus weaken the stabilization of bound ETV-TP at the N-site, which might drastically decrease ETV-MP incorporation into primer DNA. The results also suggested that local structures, including Ile63 and Leu74 (β2-β3 region), might differ considerably between HBV and HIV-1 RTs.
Definitive structural differences between HIV-1 RT Q151M and HBV RT likely exist, because HIV-1 RT Q151M exhibits high sensitivity to both ETV and EFdA. EFdA is not potent against HBV RT 32 . We speculate that the hydrophobic pocket of HIV-1 RT accommodating the unique 4′-ethynyl group of EFdA may be distorted or destroyed by F160L and neighboring G112S/D113A mutations, leading to the preclusion of EFdA-TP binding to the N-site of HBV RT. Unfortunately, HIV-1 could not be propagated upon introduction of the F160L mutation. Phe160 is located at the depth of the hydrophobic pocket adapted by EFdA 4′-ethynyl, and forms the hydrophobic core of the thumb domain with the side-chains of Met164 and Phe214, which are not conserved in HBV RT. Therefore, the local structure around Phe160 is likely an important factor yielding NRTI sensitivity differences between HIV-1 and HBV RT.

Conclusion and perspectives
We generated 5 HIV-1 RT mutants to mimic the HBV RT N-site, constructing recombinant HIV-1 containing those HIV-1 RT mutants for viral replication and anti-viral assays, and found that the Q151M mutation of HIV-1 RT alone is critical for conferring high ETV sensitivity on HIV-1. The structure of ETV-TP is identical to that of dGTP, except that ETV-TP has a cyclopentyl methylene instead of a ribose ring oxygen. Therefore, the results of the antiviral assays suggest that the side chain of Met151 forms a transient hydrophobic interaction with the cyclopentyl methylene of ETV-TP; this interaction is apparently required for smooth ETV-TP entry into the N-site. We also determined the crystal structure of HIV-1 RT Q151M in complex with DNA, DNA:dGTP, and DNA:ETV-TP, which revealed that ETV-TP is bound at the N-site, directly pushing the Met184 side chain. The corresponding Met204 substitution by the smaller Val/Ile, known as a major ETV-resistance mutation in HBV RT, could decrease the binding affinity of ETV-TP to the N-site. Among the total of 8 amino-acid substitutions introduced in this study (Table 1), Q151M, Y115F, F116Y, and L74V have been reported as mutations associated with NRTI resistance in HIV-1 RT (Fig. 1b). Therefore, exhaustive structural studies of these NRTI-resistant HIV-1 RT mutants in complex with various NRTIs would be helpful for further understanding of the relationship and differences in NRTI sensitivities between HIV-1 and HBV RT.

Methods
Viral replication kinetics and antiviral assays. The genes encoding HIV-1 RT p66 and p51 used in this study originated from the HIV-1 clone NL4-3 (GenBank: M19921.2). All site-specific mutations were introduced by inverse PCR 33 using pCDF_p66 as a template DNA 34 . HIV-1 NL4-3 -based infectious clones with RT mutations were constructed using the In-Fusion HD Cloning Kit (Clontech). The In-Fusion enzyme fuses the PCR-generated DNA sequence (RT region with mutations) and a linearized vector (HIV-1 NL4-3 -based plasmid) by recognizing a 15 bp overlap at their ends. Viral population replication kinetics were measured as previously described 35 . In brief, each virus (wild-type or carrying RT mutations) was harvested by the transfection of HIV-1 plasmids to 293 T cells. MT-4 cells (10 4 , in a 96-well plate) were exposed to 200 TCID 50 of each HIV-1 and cultured without antiretroviral agents for 9 days. No fresh MT-4 cells were added during the culture period. The p24 values in supernatants were determined on days 0, 3, 6 and 9. Antiviral assays (p24 assay) using wild-type HIV-1 (HIV-1 WT ) and replicable HIV-1 variants (HIV-1 Q151M , HIV-1 Q151M/Y115F/F116Y , and HIV-1 Q151M/I63V/L74V ) were also conducted as previously described 35 . In brief, MT-4 cells were exposed to a virus at 100 50% tissue culture infectious doses (TCID 50 s). After exposure, the cell suspension (5 × 10 3 cells in 100 µL) was plated into each well of a 96-well culture plate containing various concentrations of drugs. After incubation for 7 days, the amounts of p24 antigen in supernatants were determined using a Lumipulse G1200 immunoassay system (Fujirebio) as previously described 36 with minor modifications and the drug concentrations suppressing p24 Gag protein production by 50% (50% inhibitory concentration [IC 50 ]) were determined by comparison with p24 production levels in drug-free control cultures. All assays were performed in duplicate. Supplementary Fig. S7.

Synthesis of ETV-TP. ETV-TP was synthesized according to the scheme shown in
First, hydroxyl groups of ETV 1 were protected as triethylsilyl ether. After protection of the N 2 -amino group with an isobutyryl group, the triethylsilyl group was removed to give compound 2. Second, 5′-and 3′-hydroxyl groups were protected with 4,4′-dimethoxytrityl and isobutyryl groups, respectively. Deprotection of the 4,4′-dimethoxytrityl group gave a compound 3. Third, phosphorylation of the 5′-hydroxyl group of compound 3 by the phosphoramidite method followed by deprotection gave ETV 5′-monophosphate 4 (ETV-MP). Finally, ETV-MP was converted to the corresponding 5′-triphosphate 6 (ETV-TP) by the phosphoroimidazolidate method, as previously described 37 . Protein expression and purification. HIV-1 RT expression vectors (pET28_His 6 -p51 and pCDF_p66) were prepared as previously described 34 . To increase the stability and crystallizability of the RT:DNA complex, two Cys residues were substituted with Ser (C162S and C280S) for both p51 and p66, the 4 C-terminal residues were truncated from p66, and the human rhinovirus 3 C protease cleavage sequence was inserted between His 6 and the p51 gene fragment 38 . Recombinant HIV-1 RT WT and RT Q151M were overexpressed by E. coli strain BL21-CodonPlus (DE3)-RIL and purified as described 34 with slight modification as follows. The crude extract was applied to a Ni-affinity column (Sigma-Aldrich) pre-equilibrated with buffer A (50 mM Na phosphate pH 8.0, 600 mM NaCl, 2 mM MgCl 2 , and 10% glycerol). The column was washed with buffer A followed by buffer B (same composition as buffer A but with pH 6.0); bound samples were eluted with a linear gradient of 0-400 mM imidazole in buffer B. The collected eluent was dialyzed against buffer C (50 mM Tris-HCl pH 8.0, 2 mM MgCl 2 , 1 mM dithiothreitol [DTT], and 10% glycerol) and loaded into a DEAE Sepharose Fast Flow column (GE Healthcare). The pass-through was collected, diluted with a 3-fold volume of buffer D (same composition as for buffer C but with pH 7.5), and concentrated 4-fold by Amicon Ultra centrifugal filters with 50 kDa cut-off (Millipore). The samples were further applied to a CM Sepharose CL-6B column (GE Healthcare), and the bound RT was eluted with a linear gradient of 0-400 mM NaCl in buffer D. The fractions containing equal molar of p66 and p51 were collected, dialyzed against a buffer consisting of 10 mM Tris-HCl pH 8.0 and 50 mM NaCl, and concentrated to 14 mg/mL using 50 kDa cut-off Amicon Ultra centrifugal filters.
HIV-1 RT enzyme assays. The RT activity of HIV-1 RT WT and RT Q151M was determined using ELISA and PCR methods. ELISA was conducted using the Reverse Transcriptase Assay kit (Roche) according to the manufacturer's instruction. Purified His-tagged enzymes produced by E. coli were used for the ELISA assay. For the PCR-based RT assay, cell-free viral supernatants from 293 T cells were collected 3 days after transfection with an HIV-1-coding plasmid, and a SYBR Green-based real-time PCR-enhanced RT (SG-PERT) assay was performed as previously described with slight modifications 39 . In brief, viral lysate was prepared by incubating the mixture containing cell-free viral supernatant and virus lysis buffer (0.5% Triton X-100, 1 mM MgCl 2 , 25 mM Tris-HCl and 1 mM DTT final concentrations) for 10 min at room temperature. RT reaction mixture (10 µL) containing the viral lysate, MuLV RT buffer (1×) (New England Biolabs), dNTPs (1 mM), RNase inhibitor (4 U) (TaKaRa Bio), MS2 reverse primer (500 nM), and 1 µL pre-heated (65 °C, 5 min) MS2 RNA (Roche) was prepared, and the following reactions were performed: 60 min RT reaction at 37 °C, 5 min RT inactivation at 95 °C. SYBR Green-based real-time PCR using the following MS2 primers (MS2 fwd: 5′-TCC TGC TCA ACT TCC TGT CGA G-3′, MS2 rev: 5′-CAC AGG TCA AAC CTC CTA GGA ATG-3′). RT activity standard curve values were determined by running a dilution series of commercial recombinant HIV-1 RT (Bio Academia) in parallel.
RT Q151M :DNA complex formation. The previously established 38-mer hairpin DNA aptamer 27 was employed for the structural study of the RT Q151M :DNA:dGTP/ETV-TP ternary complex with 3 base substitutions to accommodate dGTP and ETV-TP at the N-site (Fig. 3a). The 38-mer DNA was purchased from Hokkaido System Science, Co., Ltd. The DNA dissolved in TE buffer (10 mM Tris-HCl pH 8.0 and 1 mM EDTA) was heated at 80 °C for 10 min, then cooled slowly to 25 °C. The DNA and purified RT Q151M were mixed and incubated overnight at 4 °C. The mixture was injected onto a HiLoad 16/600 Superdex 200 pg gel-filtration column (GE Healthcare) using a buffer consisting of 10 mM Tris-HCl pH 8.0, and 50 mM NaCl. Native PAGE with a Tris-borate buffer was performed to examine fraction contents and purity. The gel was initially stained by fluorescent dye (ATTO) for detection of DNA, followed by staining with Coomassie Brilliant Blue (CBB) for detection of proteins ( Supplementary Fig. S3). The peak fractions were collected and concentrated to approximately 15 mg/ mL for crystallization.

Crystallization.
All crystallization experiments were conducted by the vapor-diffusion technique at both 4 °C and 20 °C. Initial crystallization screenings using commercially available sparse matrix kits were performed by the sitting-drop vapor-diffusion method in 96-well plates. Drops consisted of equal volumes (100 nL) of sample and reservoir solution and were equilibrated against 70 μL of reservoir solution. Optimization of the initial hit condition was performed using hanging-drop vapor-diffusion with the 24-well plates. The hanging-drops were set up by mixing sample solution (1.0 μL) with reservoir solution (1.0 μL), and equilibrated against 500 μL reservoir solution. Crystals suitable for structure analysis were obtained at 20 °C using a reservoir solution consisting of 20 mM Bis-tris-HCl pH 6.0, 60 mM di-ammonium hydrogen citrate, 20 mM MgCl 2 , 1.5-3% PEG 6000, 2-4% sucrose, and 4-8% glycerol. The crystals were soaked in a cryoprotectant solution consisting of crystallization mother liquor with increasing concentrations of PEG 6000, sucrose, and glycerol up to 10%, 5%, and 25%, respectively. For preparation of the RT Q151M :DNA:dGTP and RT Q151M :DNA:ETV-TP ternary complex, the crystals were further soaked briefly in the same cryoprotectant solution supplemented with 2.4 mM dGTP/ETV-TP. The crystals were flash-cooled and stored in liquid nitrogen for the X-ray diffraction experiments.
Structure determination and refinement. X-ray diffraction data sets were collected using synchrotron radiation at BL-17A and BL-1A beamlines of the Photon Factory (Tsukuba, Japan). X-ray diffraction data were processed using the program XDS 40 . The crystals belonged to rhombohedral space group R3, with unit-cell parameters a = b = 284, c = 96 Å. The structure was solved by molecular replacement using the program MOLREP 41 , with the previously reported HIV-1 RT WT structure in complex with the DNA aptamer as a search model (PDB code, 5D3G) 27 . The atomic model was fitted manually, using the program Coot 42 . Model refinement was performed using the programs REFMAC5 43 and Phenix 44 . An atomic model of ETV-TP was created using the program SKETCHER, and the geometrical restraints files were generated using the LIBCHECK module provided in the CCP4 program package 45 . Data collection and refinement statistics are summarized in Supplementary  Table S3. MolProbity software 46 was used for model validation.
Data availability statements. The atomic coordinates and structure factor amplitudes of the RT Q151M :DNA, RT Q151M :DNA:ETV-TP, and RT Q151M :DNA:dGTP complexes have been deposited in the RCSB Protein Data Bank (http://www.rcsb.org) under accession codes 5XN0, 5XN1, and 5XN2, respectively. Other data generated and/or analyzed during this study are available from the corresponding authors upon reasonable request.