Structural insights into the recognition of nucleoside reverse transcriptase inhibitors by HIV‐1 reverse transcriptase: First crystal structures with reverse transcriptase and the active triphosphate forms of lamivudine and emtricitabine

ABSTRACT

The retrovirus HIV‐1 has been a major health issue since its discovery in the early 80s. In 2017, over 37 million people were infected with HIV‐1, of which 1.8 million were new infections that year. Currently, the most successful treatment regimen is the highly active antiretroviral therapy (HAART), which consists of a combination of three to four of the current 26 FDA‐approved HIV‐1 drugs. Half of these drugs target the reverse transcriptase (RT) enzyme that is essential for viral replication. One class of RT inhibitors is nucleoside reverse transcriptase inhibitors (NRTIs), a crucial component of the HAART. Once incorporated into DNA, NRTIs function as a chain terminator to stop viral DNA replication. Unfortunately, treatment with NRTIs is sometimes linked to toxicity caused by off‐target side effects. NRTIs may also target the replicative human mitochondrial DNA polymerase (Pol γ), causing long‐term severe drug toxicity. The goal of this work is to understand the discrimination mechanism of different NRTI analogues by RT. Crystal structures and kinetic experiments are essential for the rational design of new molecules that are able to bind selectively to RT and not Pol γ. Structural comparison of NRTI‐binding modes with both RT and Pol γ enzymes highlights key amino acids that are responsible for the difference in affinity of these drugs to their targets. Therefore, the long‐term goal of this research is to develop safer, next generation therapeutics that can overcome off‐target toxicity.

Short abstract

PDB Code(s): 6OR7, 6OTZ, 6OUN, 6P2G, 6P1X, 6P1I

Abbreviations

(−)3TC

lamivudine

(−)FTC

emtricitabine

(+)FTC

dextro isomer of emtricitabine

d4T

2′,3′‐didehydro‐2′,3′‐dideoxythymidine or stavudine

dCTP

2′‐deoxycytidine‐5′‐triphosphate

D‐ddCTP

dextro 2′,3′‐dideoxycytidine‐5′‐triphosphate (natural isomer)

dsDNA

double‐stranded DNA

FDA

Food and Drug Administration

HAART

highly active antiretroviral therapy

L‐ddCTP

levo 2′,3′‐dideoxycytidine‐5′‐triphosphate (unnatural isomer)

PEG

polyethylene glycol

Polβ

DNA polymerases β

Pol γ

DNA polymerases γ

Polλ

DNA polymerases λ

RT

reverse transcriptase

TP

triphosphate

TTP

thymidine triphosphate

1. INTRODUCTION

Human immunodeficiency virus (HIV) has been a major health issue since its discovery in the early 80s. The highly active antiretroviral therapy (HAART) has been proven to be a major breakthrough in improving the life expectancy for patients with acquired immune deficiency syndrome (AIDS). Several FDA‐approved drugs for the treatment and protection of HIV infection are nucleoside reverse transcriptase inhibitors (NRTIs). NRTIs are essential components of HAART and play a pivotal role in HIV treatment. NRTIs target the active site of the reverse transcriptase (RT) protein, a critical enzyme for the replication cycle of HIV, and they function as chain terminators of the viral DNA replication upon their incorporation in the DNA primer.1 However, several studies have linked the long‐term usage of the NRTIs to mitochondrial dysfunction2, 3 and, additionally, several other studies have reported the correlation between mitochondrial toxicity, genomic instability, and the off‐target effect of NRTIs toward several human DNA polymerases (Pols) such as Pol γ.4, 5, 6, 7 The design of highly selective NRTIs is challenging as human DNA Pols and RT share a similar mechanism of nucleotide incorporation and discrimination.8 Previous studies have shown that some Pols are more susceptible to NRTIs toxicity than others and have suggested that particular attention needs to be paid in developing and understanding of how natural nucleoside and NRTIs in their 5′‐triphosphate form discriminate between the host Pols, especially Pol γ.5, 7, 9, 10

Several FDA‐approved NRTIs, including ddC and d4T (Figure 1), are now rarely prescribed or discontinued due to the toxicity caused from off‐target inhibition.11 In order to rationally design novel inhibitors with a satisfactory profile of discrimination, investigation of the differences in the mechanism of recognition of the nucleoside analogues by RT is needed. Successful examples of NRTIs designed to bind more selectively to RT are the two pyrimidine analogues containing an 1,3‐oxathiolane ring: emtricitabine, (−)FTC,12 and lamivudine, (−)3TC,13 (Figure 1) which are components of most of the FDA‐approved combination therapies. Interestingly, the triphosphate form (‐TP) of (−)FTC and (−)3TC has a lower mitochondrial toxicity and a much wider therapeutic index than d4T and ddC. In fact, (−)FTC‐TP and (−)3TC‐TP are similarly effective against RT compared to their chiral pair (+)FTC‐TP/(+)3TC‐TP, but they are incorporated much less efficiently into Pol γ.4, 5, 14, 15, 16 Similar stereochemical preferences between D‐ and L‐stereoisomers of nucleic acids were also reported for Polβ and Polλ.7, 9, 10 To date, there are no structural data available for (−)/(+)FTC‐TP and (−)3TC‐TP in complex with RT. The recent success in structure‐based design of novel nonnucleoside reverse transcriptase inhibitors (NNRTIs) with improved pharmacokinetics and less susceptibility toward drug resistance underscore the value of structural data.17, 18, 19, 20, 21, 22 Similarly, in order to understand the molecular basis of recognition for NRTIs by RT and host Pols important in toxicity, obtaining structural data of RT in complex with (−)FTC‐TP and (−)3TC‐TP is imperative. Together with kinetic data, knowledge of the active site architecture for RT with these NRTIs bound is essential for the rational design of new molecules that bind highly selective toward RT.

Figure 1.

Chemical structures of 2′‐deoxycytidine‐5′‐triphosphate (dCTP); thymidine‐5′‐triphosphate (TTP); D‐2′,3′‐dideoxycytidine‐5′‐triphosphate (D‐ddCTP); L‐2′,3′‐dideoxycytidine triphosphate (L‐ddCTP); d4T‐TP, (−)FTC‐TP, and (−)3TC‐TP are the triphosphate forms of stavudine d4T, emtricitabine, (−)FTC, and lamivudine, (−)3TC, respectively. (+)FTC‐TP is the enantiomer of (−)FTC‐TP

In this study, we have determined the crystal structures of RT in complex with dsDNA (primer/template), and six different cytidine analogues (dCTP, D‐ddCTP, L‐ddCTP (−)FTC‐TP, (+)FTC‐TP, and (−)3TC‐TP), and elucidated different binding modes of these six compounds to RT. Our structural comparison of (−)FTC‐TP and (−)3TC‐TP in complex with RT and different Pols, in particular, Pol γ, highlights the key differences in their binding modes, which might be responsible for their differential interaction and binding affinity to their targets. These results can contribute to the development of safer, next generation therapeutics which can overcome the off‐target effects of currently available NRTIs.

2. RESULTS

2.1. Determination of RT crystal structures and binding modes of the cytidine analogues

The engineered RT protein (Q258C and C280S mutations) cross‐linked to the dsDNA primer/template (20‐mer/27‐mer) containing the N²‐cystamine 2′‐deoxyguanosine six bases upstream from the priming site (P site) was used for the crystallographic studies. Structural data are obtained from crystals of RT protein in complex with dCTP, D‐ddCTP, L‐ddCTP, (−)FTC‐TP, (+)FTC‐TP, and (−)3TC‐TP. (PDB ID: 6P1I, 6P2G, 6P1X, 6OR7, 6OTZ, and 6OUN) with resolution of 2.74, 2.99, 2.55, 2.53, 2.86, and 2.66 å, respectively. Data collection, processing, and refinement statistics can be found in Tables 1 and 2.

Table 1.

Data collection and refinement statistics

PDB ID code	6OR7 ((−)FTC‐TP complex)	6OTZ ((+)FTC‐TP complex)	6OUN ((−)3TC‐TP complex)
(A) Data collection and processing
Wavelength (å)	0.979180	0.979314	0.979303
Space group	C 2 2 21	C 2 2 21	C 2 2 21
Unit cell parameters a, b, c, (å); α, β,γ, (°)	166.5, 169.8, 103.3; 90, 90, 90	167.1, 170.4, 103.2; 90, 90, 90	167.0, 171.8, 105.9; 90, 90, 90
Matthews coefficientb (å3/Da)	2.8	2.8	2.9
Solvent contentb (%)	55.7	56.0	57.4
(B) Diffraction data
Resolution range (å)	50–2.53 (2.68–2.53)	30–2.86 (3.03–2.86)	30–2.66 (2.82–2.66)
Unique reflections	48,991 (7,702)	34,440 (5,442)	44,155 (6,977)
R(I)sym (%)	9.3 (96.7)	16.3 (85.4)	13.1 (148)
Wilson B factor (å2)	57.7	56.4	76.1
Completeness (%)	99.7 (98.4)	99.8 (99.6)	99.8 (99.5)
Redundancy	7.3 (7.9)	12.8 (12.9)	26.3 (28.6)
<I/σ(I)>	16.6 (1.9)	12.5 (3.1)	19.5 (1.9)
CC(1/2)	99.8 (77.8)	99.6 (88.3)	99.9 (92.0)
(C) Refinement
Resolution range (å)	47.57–2.53	29.82–2.86	29.57–2.66
Reflections used in refinement (work/free)	48,969 (46,522/2,447)	34,423 (32,701/1,722)	44,086 (41,881/2,205)
Final R value for all reflections (work/free) (%)	21.2/24.9	19.8/24.1	20.7/25.9
Protein residues (chain A/chain B)	545/ 388	554/403	546/397
dsDNA (primer/template)	18/23	18/23	18/22
Water molecules	48	2	6
RMSD from ideality: bond lengths (å)	0.003	0.003	0.005
RMSD from ideality: bond angles (°)	0.569	0.625	0.763
Mean B factor protein (å2)c	62.0	60.2	85.9
Mean B factor ligand (å2)c	78.6	44.4	116.4
Mean B factor water molecules (å2)c	42.8	50.6	66.9
Ramachandran plot d:
Residues in most favored regions (%)	93.1	92.4	90.6
Residues in additionally allowed regions (%)	6.8	7.6	9.2
Residues in generously allowed regions (%)	0.1	0.0	0.1
Residues in disallowed regions (%)	0.0	0.0	0.1

Note: Values in parenthesis describe the highest resolution shell.

^bCalculated with Matthews_coef program from CCP4 suite version 7.0.053.36, 37

^cMean B‐factors were calculated with MOLEMAN.50

^dCalculated with PROCHECK.51

Table 2.

Data collection and refinement statistics

PDB ID code	6P1I (dCTP complex)	6P1X (L‐ddCTP complex)	6P2G (D‐ddCTP complex)
(A) Data collection and processing
Wavelength (å)	0.979314	0.979314	0.979314
Space group	C 2 2 21	C 2 2 21	C 2 2 21
Unit cell parameters a, b, c, (å); α, β,γ, (°)	166.4, 169.5, 103.1; 90, 90, 90	169.2, 171.5, 107.3; 90, 90, 90	168.2, 171.2, 105.7; 90, 90, 90
Matthews coefficientb (å3/Da)	2.8	3.0	2.9
Solvent contentb (%)	55.5	58.5	57.8
(B) Diffraction data
Resolution range (å)	30–2.74 (2.90–2.74)	30–2.55 (2.71–2.55)	30–2.99 (3.16–2.99)
Unique reflections	38,484 (5,873)	97,351 (15,225)	393,959 (31,320)
R(I)sym (%)	16.0 (90.0)	11.5 (71.5)	25.2 (67.8)
Wilson B factor (å2)	56.8	53.9	53.9
Completeness (%)	99.1 (94.9)	99.4 (96.8)	99.1 (95.0)
Redundancy	13.5 (	7.1 (6.8)	12.6 (12.8)
<I/σ(I)>	12.4 (2.3)	12.1 (2.2)	10.2 (2.9)
CC(1/2)	99.7 (86.5)	99.7 (83.7)	99.1 (93.6)
(C) Refinement
Resolution range (å)	28.93–2.74	29.75–2.55	29.59–2.99
Reflections used in refinement (work/free)	38,466 (36,541/1,925)	97,318 (92,465/4,853)	31,276 (29,711/1,565)
Final R value for all reflections (work/free) (%)	18.3/23.6	18.8/23.0	21.5/26.7
Protein residues (chain A/chain B)	548/392	557/406	545/388
dsDNA (primer/template)	18/23	19/24	18/22
Water molecules	17	33	–
RMSD from ideality: bond lengths (å)	0.007	0.007	0.006
RMSD from ideality: bond angles (°)	0.904	0.951	0.872
Mean B factor protein (å2)c	55.0	55.7	50.2
Mean B factor ligand (å2)c	30.4	60.9	41.6
Mean B factor water molecules (å2)c	43.9	43.8	–
Ramachandran plot d:
Residues in most favored regions (%)	91.0	91.2	91.4
Residues in additionally allowed regions (%)	9.0	8.7	8.4
Residues in generously allowed regions (%)	0.0	0.0	0.2
Residues in disallowed regions (%)	0.0	0.1	0.0

Note: Values in parenthesis describe the highest resolution shell.

^bCalculated with Matthews_coef program from CCP4 suite version 7.0.053.36, 37

^cMean B‐factors were calculated with MOLEMAN.50

^dCalculated with PROCHECK.51

The superimposition of the six crystal structures did not show any substantial protein conformational changes induced by binding of the inhibitors, displaying a mean RMSD of 0.71 ± 0.29 å, between the alignment of the Cα atoms of the structures, as calculated with COOT27 (Figure S1). The electron density for the amino acids between P133 and G141 of the fingers domain was poor, and it allowed the modeling of this loop only for the complexes with (+)FTC‐TP and L‐ddCTP. Differences in the fingers domain are due to the high flexibility of this region and are commonly observed.28 Overall, the structure of the RT protein in complex with the dsDNA and the NRTIs are similar to previously reported protein‐DNA structures.29, 30 The electron density for all six triphosphate cytidine analogues are clear and well‐defined in the active site of RT (Figure 2). The cytosine base moieties of the compounds make a Watson–Crick base pair with the guanine base (5‐dG) of the template DNA oligo, and it forms a π–π stacking with the guanine base 21‐ddG of the primer (Figure 3). The cytosine base moieties of the six compounds display very similar H‐bond distances to the 5‐dG despite the difference in stereochemistry, the missing 3′‐OH, or the lack of the 5‐F (Figure S2). For all the six complexes, only one Mg⁺² ion could be detected coordinated with amino acid side chains D110 and D185 and the carbonyl of the backbone of V111. The Mg⁺² ion, corresponding to metal B in the polymerization mechanism, is additionally coordinated with the α and γ phosphate groups of dCTP, D‐ddCTP, and (+)FTC‐TP and the β and γ phosphate groups of L‐ddCTP, (−)FTC‐TP, and (−)3TC‐TP. The Mg⁺² ion corresponding to metal A in the polymerization mechanism is not detected in the electron density of the RT structures presented in this work. The two Mg⁺² ions have different roles in the incorporation mechanism of the nucleotide triphosphate: Mg⁺² ion A functions by lowering the pKa of the 3′‐OH of the DNA primer, promoting the attack of the 3′‐OH by the γ phosphate, while Mg⁺² ion B assists in facilitating the leaving of the pyrophosphate.32 Due to the lack of a 3′‐OH on the chain‐terminated DNA primer and the NRTIs, the Mg⁺² ion A binding is weakened and may not have full occupancy in the crystal structures to be observed. This is not surprising as Mg⁺² ion A is detected in only a few crystal structures of RT in complex with dsDNA and in complex with DNA/RNA hybrid.33, 34, 35 The three phosphate groups are firmly anchored in the active site by an extensive hydrogen bond network involving additionally R72, D110, D113, A114, D185, and the Mg⁺² ion. However, the three phosphate groups of the different enantiomers coordinate in a dissimilar fashion within the protein active site. The α and γ phosphate groups of the (−) or L enantiomers are in H‐bonding distance with the backbone of D113 and A114, while the same interactions are achieved by the β and γ phosphate groups for the (+) or D enantiomers. The γ phosphate groups of dCTP, (−)FTC‐TP, and (+)FTC‐TP are additionally stabilized by H‐bond interactions with K220, whereas in the structure of RT in complex with L‐ddCTP, K220 is too far to achieve a H‐bond interaction with γ phosphate (3.6 å) and, for the D‐ddCTP and (−)3TC‐TP complex structures, the density of this amino acid is disordered (Figure 2). The β phosphate group of dCTP is additionally interacting with a water molecule.

Figure 2.

Crystal structures of RT in complex dsDNA primer/template and NRTIs. Fo‐Fc difference electron densities are shown as gray mesh at a contour level of 3σ. The triphosphate cytidine analogues and K220 are shown as stick model. The carbon atoms for dCTP are colored brown (a), for D‐ddCTP are colored light green (b), for L‐ddCTP are colored light gray (c), for (+)FTC‐TP are colored purple (d), for (−)FTC‐TP are colored cyan (e), and for (−)3TC‐TP are colored magenta (f). All structural representations were prepared with PyMOL.31 NRTI, nucleoside reverse transcriptase inhibitor; RT, reverse transcriptase

Figure 3.

Crystal structures of RT in complex with dsDNA primer/template and pyridine analogues. Close‐up view of the RT active site bound to dCTP (in brown, a), D‐ddCTP (in light green, b), L‐ddCTP (in light gray, c), (−)FTC‐TP (in cyan, d), (−)3TC‐TP (in magenta, e), and (+)FTC‐TP (in purple, f). The inhibitors are shown as stick models. The amino acids and nucleic acids, within a distance of 5 å, are shown as stick models in gray. H‐bonds interaction distances within 3.5 å to the compounds are depicted as black dotted lines. The distance between 3′‐C of the ribose and the α P is given in å and is depicted as orange dashed lines. Mg⁺² ions are shown as green spheres. Water molecule is shown as a red sphere. Full list of H‐bond distances between the compound and RT can be found in Figure S2. RT, reverse transcriptase

3. DISCUSSION

3.1. Comparison between the binding modes of different pyrimidine analogues

The pyridine moiety is binding into the active site in a nearly identical fashion for all six compounds; however, a large conformational change is observed for the phosphate groups of these compounds (Figure 4). Overall, the binding mode between (−)FTC‐TP and (−)3TC‐TP is very similar with the minor differences in the position of the β and γ phosphate groups (Figure 4a). Those differences result in additional interactions of (−)FTC with RT especially with K220. The more extensive hydrogen bond network between (−)FTC‐TP and RT observed in the crystal structure is in agreement with the kinetic data that show a slightly higher rate of incorporation for (−)FTC‐TP compared to (−)3TC‐TP (Table 3).4, 15 The phosphate groups of unnatural (L or (−)) enantiomer (−)FTC‐TP, (−)3TC‐TP, and L‐ddCTP are all shifted downward with the α P located the same region of the corresponding β P of dCTP, D‐ddCTP, and (+)FTC‐TP resulting in an increased distance of ~3 å between the 3′‐C of the dideoxyribose of the 21‐ddG of the DNA primer and the α P (Figures 3 and S2). Comparison of the NRTIs binding mode and the natural isomers (D or (+)) dCTP, D‐ddCTP, and TTP (PDB ID: 1RTD) provide evidence that the binding mode of (+)FTC‐TP is consistent with the binding mode of the natural deoxynucleotide triphosphate but L‐ddCTP, (−)FTC‐TP, and (−)3TC‐TP exhibit significant changes (Figures 4 and S2). The observed rearrangement of the phosphate groups of the (−) or L compounds are striking as it has never been reported for the RT protein. The longer distance between the DNA primer and the unnatural enantiomer of the compounds is also in agreement with the slower incorporation rates in contrast with natural substrates, dCTP and TTP, by RT (Table 3). Previous pre‐steady‐kinetic analysis showed the lack of a burst for (−)3TC‐TP and (−)FTC‐TP when a dsDNA primer template was used as a substrate.4, 38 The lack of a burst indicates that the k _pol rate is similar or equal to the steady‐state rate, k _ss. These data imply that either a conformational change or a chemistry rearrangement may now represent the slowest step in the reaction pathway. Other studies examining (−)3TC‐TP using a fluorescently tagged form of RT have also suggested more complex kinetics.39 The structural findings for the two unnatural NRTIs, (−)3TC‐TP and (−)FTC‐TP, together with the previous kinetic data strongly suggest a more complex mechanism of incorporation by RT for the unnatural NRTIs compared to the one suggested for the natural analogues. In considering the structure of RT in complex with natural (+)FTC‐TP, the structural basis for the lower incorporation rate of (+)FTC‐TP compared to dCTP, D‐ddCTP, and TTP is less clear as they bind the protein in a nearly identical way. As the crystal structures provide only a snapshot of the binding events between the NRTIs and the protein, it is possible that additional dynamic changes are not captured that may offer further explanation of the kinetic studies.

Figure 4.

Superimposition of the binding mode of the compounds bound to the RT binding site. The inhibitors are shown as stick models. The carbon atoms of (−)FTC‐TP are shown in cyan (a and b), carbon atoms of (+)FTC‐TP are shown in purple (b and d), carbon atoms of (−)3TC‐TP are shown in magenta (a and f), carbon atoms of dCTP are shown in brown (c), carbon atoms of D‐ddCTP are shown in light green (c, d, and e), and carbon atoms of L‐ddCTP are shown in light gray (e and f). RT, reverse transcriptase

Table 3.

Nucleotide and analog incorporation by RT

DNA/DNA	Pyrimidine analogues	k pol (s−1)	K d (μM)	k pol (s−1)/k d (μM−1)
D23/D45	(−)FTC‐TP	0.039 ± 0.003a	12 ± 2a	0.0033a
D23/D45	(+)FTC‐TP	0.031 ± 0.003a	14 ± 2a	0.0022a
D23/D45	(−)3TC‐TP	0.019 ± 0.001b	15 ± 3b	0.0013b
D23/D45	dCTP	2.9 ± 0.2c	56 ± 9.8c	0.052c
D23/D45	D‐ddCTP	0.35 ± 0.02b	69 ± 11b	0.0051b
D23/D45	L‐ddCTP	0.013 ± 0.001d	27 ± 5d	0.00049d
D22/D45	dTTP	1.0 ± 0.1e	1.5 ± 0.5e	0.67e

Abbreviation: RT, reverse transcriptase.

^aFeng et al.4

^bFeng et al.38

^cRay et al.52

^dRay et al.53

^eVaccaro et al.14

The comparison of the binding mode of L‐ddCTP, (−)FTC‐TP, (+)FTC‐TP, and (−)3TC‐TP with the crystal structure of dCTP, D‐ddCTP, and d4T‐TP in complex with RT (PDB ID: 6AMO 40) also leads to the conclusion that the peculiar binding mode of L‐ddCTP, (−)FTC‐TP, and (−)3TC‐TP is not caused from the missing second Mg⁺², from the missing 3′‐hydroxyl of the compounds, nor from the difference in the ribose‐like moiety. In fact, dCTP, (+)FTC‐TP, D‐ddCTP, and d4T‐TP exhibit the same binding mode as the natural substrates, and TTP (PDB ID: 1RTD 35) despite the nonobservable second Mg⁺² ion and the 3′‐hydroxyl presence or lacking on the ribose‐like heterocycle (Figures 4 and 5). In addition, L‐ddCTP, (−)FTC‐TP, and (−)3TC‐TP bind in the same atypical fashion suggesting that is the stereochemistry and not the presence of the sulfur in the oxathiolane ring of (−)FTC‐TP and (−)3TC‐TP the trigger for the different conformation adopted in the active site by these three compounds.

Figure 5.

Close‐up view of the RT active site bound to (−)FTC‐TP (in cyan, b), (−)3TC‐TP (in magenta, c), TTP (in yellow, a), and d4T‐TP (in green, a–c). Protein and primer/template are shown as ribbon model. The inhibitors are shown as stick models. The amino acids and nucleic acids, within a distance of 5 å, are shown as stick models in gray. Mg⁺² ions are shown as spheres in the same color as the compound of the complex structure. RT, reverse transcriptase; TTP, thymidine triphosphate

3.2. Comparison between the NRTIs binding modes in RT and host DNA polymerases (Pols)

In a previous study describing the binding mode of (−)FTC‐TP and (+)FTC‐TP with Pol γ (PDB ID: 5C52 5 and 5C53 5, respectively), the differences in binding between the two stereoisomers were primarily due to changes in the 5‐F cytosine base pairing with the dG of the template.5 The unnatural (−)FTC‐TP exhibited a distorted Watson–Crick base pair with extended hydrogen bonding pairs by 0.8–1.0 å, which may explain the large differences observed in incorporation rate (k _pol(s⁻¹) 0.84 for (+)FTC‐TP and 8.6 × 10⁻³ for (−)FTC‐TP4). However, it is evident, comparing the superimposition of the two Pol γ structures, that there are no significant differences between the orientation of the phosphate groups of the natural (+) and the unnatural (−) enantiomer of the compounds (Figures 6 and S3) and that the binding mode of (+)/(−)FTC‐TP in Pol γ was consistent with the binding mode of TTP in RT (Figure 6b). On the other hand, the binding mode of (−)FTC‐TP and (−)3TC‐TP in RT is distinct from the one of (−)FTC‐TP in Pol γ (Figure 6c,d). As the different enantiomer in complex with Pol γ did not result in substantial conformational changes in the binding mode of the compounds, it was predicted that the binding mode of (−)FTC‐TP and (−)3TC‐TP in RT also should have been similar to the one of (+)FTC‐TP/TTP.5 An additional insight into underlying reasons for such shifted binding modes of (−)FTC‐TP and (−)3TC‐TP in RT can be obtained by considering structural comparisons with the DNA damage repair enzymes, Pol λ7 and Pol β.9, 10 In those cases, multiple (−)FTC‐TP and (−)3TC‐TP binding modes are observed for these DNA Pols in contrast to the natural dCTP substrate. The two NRTIs are captured in several conformational states in transition to a productive catalytic complex the DNA polymerase active site.

Figure 6.

Superimposition of the binding mode of the compound bound to the RT binding site and DNA polymerase γ (Pol γ). The ligands are shown as stick models. The carbon atoms of (−)FTC‐TP bound to Pol γ are shown in pink (a, c, and d), carbon atoms of (+)FTC‐TP bound to Pol γ are shown in teal (a and b), carbon atoms of TTP bound to RT are shown in yellow (b), carbon atoms of (−)FTC‐TP bound to RT are shown in cyan (c), and carbon atoms of (−)3TC‐TP bound to RT are shown in magenta (d). RT, reverse transcriptase; TTP, thymidine triphosphate

4. CONCLUSIONS

We elucidated the crystal structures of RT protein in complex with a dsDNA primer template and six different triphosphate pyridine nucleotides, dCTP, D‐ddCTP, L‐ddCTP, (−)FTC‐TP, (+)FTC‐TP, and (−)3TC‐TP, and highlighted novel binding modes for the unnatural (−) or L enantiomer compounds. This knowledge not only provides additional understanding of the kinetic data but also gives invaluable structural insight into the binding mechanism and lays the path for further rational optimization and development of more selective therapeutic against HIV with lower off‐target‐related side effects.

5. MATERIALS AND METHODS

5.1. Expression, purification, RT‐DNA cross‐linking, and crystallization of the recombinant RT protein

Recombinant RT enzyme containing the mutations C280S and Q258C was expressed in Escherichia coli and purified as following protocols previously described.28, 30, 41, 42 For the cross‐linking experiments, a 27‐mer DNA template (5′‐ATGGGCGGCGCCCGAACAGGGACTGTG‐3′), standard desalted, was purchase from IDT (Integrated DNA Technologies, Coralville, Iowa), and a 20‐mer DNA primer (5′‐ACAGTCCCTGTTCGGXCGCC‐3′) was purchased from TriLink BioTechnologies (San Diego, California). The cross‐linking N²‐cystamine 2′‐deoxyguanosine in the DNA primer is indicated by an X. The cross‐linking reaction between RT and 27‐mer/20‐mer primer template and the purification of the protein–dsDNA complex were carried out as described by Sarafianos et al.30, 41 SDS PAGE was used to confirm the purity of the dsDNA–protein complex (results not shown). dCTP and ddCTP were purchased from Sigma. The (+) and (−)FTC‐TP were synthesized as previously described.15 The purity of (−)FTC‐TP and (+)FTC‐TP compounds (~99%) was verified by high‐pressure liquid chromatography analysis as well as liquid chromatography‐electrospray ionization mass spectrometry. (−)3TC‐TP was purchased from Toronto Research Chemicals and was purified before use as previously described.5 For the crystallography studies, 1 μL of RT enzyme cross‐linked to the DNA/DNA primer template (~90 μM) was combined with an equal volume of mother liquor of 2–4% (w/v) PEG 8000, 15 mM magnesium sulfate, 50 mM MES adjusted at pH 6.0, and 0.1 mM of either (−)/(+)FTC‐TP or (−)3TC‐TP. For the structures in complex with dCTP, D‐ddCTP, and L‐ddCTP, no compounds were added to the mother liquor. After growing for 2 weeks at 22°C, crystals were exposed for 1 hr to the mother liquor solution with the addition of 0.3 mM of one of the triphosphate cytidine analogues and a final concentration of 15% (w/v) PEG 8000. The PEG 8000 concentration was increased by 5% each hour to a final concentration of 35%, and crystals were incubated overnight in the resulted solution. Crystals were then harvested and exposed to a solution containing 35% (w/v) PEG 8000, 15 mM magnesium sulfate, 50 mM MES pH 6.0, 20% (v/v) glycerol, and 0.1 mM of the triphosphate cytidine analogues for 10 s and then flash frozen in liquid nitrogen.

5.2. Data collection, processing, and structure determination and refinement

Crystals of RT protein in complex with (−)FTC‐TP were collected at APS on beam line 24‐ID‐E through NE‐CAT. Crystals of RT protein in complex with dCTP, D‐ddCTP, L‐ddCTP, (+)FTC‐TP, and (−)3TC‐TP were collected at NSLS‐II FMX beamline. The data collection of the complexes were achieved at a wavelength of 0.98 å and a temperature of 100 K on a CCD HF‐4M detector (PDB ID: 6OR7) and on a silicon Dectris Eiger 16M detector (PDB ID: 6OTZ, 6OUN, 6P2G, 6P1X, and 6P1I). Data sets for the best diffracting crystals were indexed, processed, and scaled with XDS.43 Phases were solved by molecular replacement with the program PHASER MR44 from the CCP4 suit45 using the structure 3KJV as initial search model. A subset corresponding to 5% of all reflections were omitted during refinement and used for the calculation of R _free. Model building was performed in COOT27, 46 and refinement using PHENIX.refine version 1.14‐3260.47 dCTP, D‐ddCTP, L‐ddCTP, (−)FTC‐TP, (+)FTC‐TP, and (−)3TC‐TP SMILE codes were created with Molinspiration v2018.10,48 built with the grade web server.49 Compounds restraints were generated with the GWS and eLBOW.23 Cartesian simulated annealing, applying default parameters, was used as a first refinement step. Subsequently, refinement of XYZ coordinates and individual B‐factors were alternated with structural adaption until the model was readily built and gave the best possible explanation of the electron density, acceptable R‐factors, geometry statistics, and Ramachandran statistics were achieved. TLS refinement was performed for all structures except for 6P2G with TLS groups selected from the TLSMD web server.24, 25 Crystallography programs were compiled by SBGrid.26

AUTHOR CONTRIBUTIONS

N.B. and K.S.A. designed research; N.B. and A.H.C. performed research; N.B., Y.W.Y., and K.S.A. analyzed data; R.F.S. contributed new reagents and analytic tools; R.F.S and Y.W.Y. reviewed the manuscript; and N.B. and K.S.A. wrote the manuscript.

Supporting information

Figure S1. Overall superimposition of the six novel crystal structure complexes.

Figure S2. Crystal structures of RT in complex with DNA/DNA primer/template and pyridine analogues with H‐bond distances.

Figure S3. Crystal structure of (−)FTC‐TP and (+)FTC‐TP in complex with Pol γ.

Bertoletti N, Chan AH, Schinazi RF, Yin YW, Anderson KS. Structural insights into the recognition of nucleoside reverse transcriptase inhibitors by HIV‐1 reverse transcriptase: First crystal structures with reverse transcriptase and the active triphosphate forms of lamivudine and emtricitabine. Protein Science. 2019;28:1664–1675. 10.1002/pro.3681