The structure of the human tRNALys3 anticodon bound to the HIV genome is stabilized by modified nucleosides and adjacent mismatch base pairs.

Replication of human immunodeficiency virus (HIV) requires base pairing of the reverse transcriptase primer, human tRNA(Lys3), to the viral RNA. Although the major complementary base pairing occurs between the HIV primer binding sequence (PBS) and the tRNA's 3'-terminus, an important discriminatory, secondary contact occurs between the viral A-rich Loop I, 5'-adjacent to the PBS, and the modified, U-rich anticodon domain of tRNA(Lys3). The importance of individual and combined anticodon modifications to the tRNA/HIV-1 Loop I RNA's interaction was determined. The thermal stabilities of variously modified tRNA anticodon region sequences bound to the Loop I of viral sub(sero)types G and B were analyzed and the structure of one duplex containing two modified nucleosides was determined using NMR spectroscopy and restrained molecular dynamics. The modifications 2-thiouridine, s(2)U(34), and pseudouridine, Psi(39), appreciably stabilized the interaction of the anticodon region with the viral subtype G and B RNAs. The structure of the duplex results in two coaxially stacked A-form RNA stems separated by two mismatched base pairs, U(162)*Psi(39) and G(163)*A(38), that maintained a reasonable A-form helix diameter. The tRNA's s(2)U(34) stabilized the interaction between the A-rich HIV Loop I sequence and the U-rich anticodon, whereas the tRNA's Psi(39) stabilized the adjacent mismatched pairs.

INTRODUCTION

Retrovirions are packaged with an enrichment of a host cell tRNA for the purpose of priming reverse transcription. Human immunodeficiency virus type 1, HIV-1 and all other lentiviruses, evolved to select and recruit the cytoplasmic tRNALys3 and the other tRNALys isoacceptors, tRNALys1 and tRNALys2 (1). While the function of tRNALys1,2 in HIV-1 is not known, the selective packaging of HIV-1 reverse transcriptase primer tRNALys3 is required for optimizing both the annealing of tRNALys3 to viral RNA and the infectivity of the HIV-1 population (2). A considerable number of interactions are believed to occur between the human tRNALys3 and HIV-1, as determined by chemical and enzymatic probes (3). Eighteen nucleosides of the tRNA's 3′-terminal sequence (Figure 1A), part of the amino-acid accepting stem (nucleosides 66–76) and the 3′-side of the TΨC domain (tRNA nucleosides 59–65) (4), are bound to a complementary sequence in the viral RNA, the Primer Binding Sequence (PBS) (viral nucleosides 183–200), while the TΨC domain interacts with the Primer Activation Sequence (PAS) (5) (Figure 1A). The uridine-rich anticodon and adjacent nucleosides are bound to the adenosine-rich Loop I of the viral RNA (Figure 1A and C) (4). Deletion of the A-rich Loop I of HIV-1 that is complementary to the uridine-rich anticodon domain resulted in significantly reduced levels of infectivity and reduced synthesis of viral RNA (6–8). Constitution of a virus that was able to use a host cell tRNA other than tRNALys3 required not only the conversion of the PBS to the complement of the investigator-selected tRNA, but also the mutation of the AAAA sequence of HIV-1 Loop I to complement the anticodon sequence of the new tRNA (9–11). Changing one or the other of these sequences was not sufficient for maintaining the new interaction during prolonged culturing. Binding of the tRNA's anticodon region to the A-rich Loop I region has been associated with transition of viral replication from initiation to the highly processive elongation with a possible requirement for the tRNA's post-transcriptional modifications (12).

Figure 1.

Sequence and secondary structure of the tRNALys3 interaction with Loop I of the HIV genome. (A) Sequence and secondary structure of human tRNALys3. The 3′-terminal, 18 nucleoside sequence of tRNALys3 (in red) is complementary to the HIV-1 Primer Binding Site, PBS. Residues in blue are complementary to the HIV-1 Primer Activation Signal, PAS. In purple are the residues of the anticodon domain fragment, ADF, studied in the article. In order not to have frayed and weak termini to the dodecamer duplex, a C43 was substituted for A43 to constitute a terminal C43•G158 base pair. The native post-transcriptional modifications are indicated according to standard nomenclature (80). (B) Chemical structures of the modified nucleosides in the ADF of human tRNALys3. The natural modification occurring at position 34 is 5-methoxycarbonylmethyl-2-thiouridine (mcm5s2U34) and that at position 37 is 2-methylthio-N6-threonylcarbamoyladenosine (ms2t6A37). The uridine at position 39 is modified to pseudouridine, Ψ39. The modifications studied here were 2-thiouridine (s2U34) and pseudouridine, Ψ39. (C) The sequence and secondary structure of the A-rich Loop I of HIV-1 (residues 157–173). The residues in green (158–169) are those residues studied in this article. Residues A158 and C169 were changed to G158 and G159 in order that the dodecamer duplex have terminal G•C base pairs. (D) Duplexes composed of a fragment of the tRNALys3 anticodon domain, the ADF, and the A-rich Loop I of HIV-1, subtype G (VGL). Constructs included the individual and doubly modified ADFs with 2-thiouridine, s2U34 (purple) and pseudouridine, Ψ39 (blue). An additional construct of the ADF was substituted at positions 38 and 39 to create a full complement to the VGL (orange). The labeled pyrimidine nucleosides, [15N3]-uridine and 4-amino-[15NH2]-cytidine, are designated (green). The HIV-1 subtype B construct had a sequence with two Gs substituting for A166 and A167 (gray).

The tRNALys3 anticodon domain has post-transcriptional modifications (5-methoxycarbonyl-methyl-2-thiouridine, mcm5s2U34; 2-methylthio-N6-threonylcarbamoyl-adenosine, ms2t6A37; and pseudouridine, Ψ39) that are critical for the decoding of its cognate (AAA) and wobble codons (AAG) on the ribosome (13) and affect its interaction with the HIV Loop I (14). Fully modified, native tRNALys3SUU when bound to the viral RNA protected the HIV-1 Loop-I AAAA sequence (viral nucleosides 169–172) from chemical reactivity and enzymatic cleavage (14). In contrast, the unmodified in vitro transcript of the tRNA was unable to block chemical reactivity of the same sequence (14). In complex with viral RNA, the tRNA anticodon was also protected from chemical and enzymatic probes (3) and the modified nucleoside mcm5s2U34 (Figure 1B) can be crosslinked to the viral RNA (15). When the complementarity between the anticodon of tRNA and the A-rich Loop 1 was disrupted, all four helices of the extended tRNA/HIV interactions were destabilized (16). The importance of the thionyl of mcm5s2U34 to the interaction of the tRNA's anticodon with the viral Loop I was demonstrated when its oxidation was associated with a decrease in the tRNA's protection of Loop I (14). Thus, tRNA's 2-thio modification of U34 may be as important in initiating viral replication (17), as it is in recognizing cognate codon AAA in translation (18).

The chronology of events that lead to the interaction of the U-rich, modified anticodon domain with the A-rich viral Loop I and formation of a duplex could involve an intermediate ‘kissing-loop’ interaction of the two stem and loop hairpins (19,20). However, the 5′-side of the tRNA's anticodon stem, when in complex with the PBS and Loop I, is susceptible to chemical and enzymatic probes (3). In addition, the HIV nucleocapsid protein NCp7 stably denatures tRNALys3 in the presence of the complementary HIV sequences (21–23). Whatever the mechanism that leads to formation of the replication initiation complex, a duplex apparently results from the annealing of the anticodon domain to the HIV Loop I. Yet, the contributions of the modified nucleosides to the stability and structure of the resulting duplex are not known.

To investigate the effect of modified nucleosides on the stability and structure of the duplex that is formed from the anticodon domain and the viral Loop I, an experimental system was designed to mimic the linear, denatured form of the RNAs. Here, we report how two of the more important individual tRNA-modified nucleosides contribute to the stabilization of the tRNA anticodon domain interaction with both the HIV sub(sero)type G, exhibiting the highest complementariness of the Loop I sequences with the anticodon stem and loop domain of tRNALys3 (ASL), and sub(sero)type B, the most prevalent subtype in the Americas, Europe and Oceania. The modified nucleosides s2U34 and Ψ39 were most effective at stabilizing the interaction between the tRNA's anticodon region and the viral A-loop sequence. The NMR-derived structure of one duplex containing the two modified nucleosides, s2U34 and Ψ39 resulted in two coaxially stacked A-form RNA stems separated by two mismatched base pairs, U162•Ψ39 and G163•A38, that maintained a C1′–C1′ distance between 10 and 12.5 Å that is comparable to that of the A-form RNA helix diameter.

MATERIALS AND METHODS

Oligonucleotide synthesis

The tRNALys3 anticodon domain fragments (ADF) and HIV Loop I sequences corresponding to the subtypes G and B (VGL, VBL) were synthesized (Nucleic Acids Facility at North Carolina State University, Raleigh, NC) as dodecamers (Figure 1) with modified nucleosides and the 15N-labeled nucleosides [15N3]-uridine and the 4-amino-[15NH2]-cytidine. The protected, s2U phosphoramidite was synthesized as published (24). The [15NH2]-cytidine, and 15N3-uridine were chemically synthesized (25) and then, derivatized to the protected 3′-phosphoramidite (ChemGenes, Ashland, MA). Oligomers were purified by HPLC (26), desalted (Waters Corporation Sep-pak columns, Milford, MA) and determination of the nucleoside composition of the RNA was performed by enzymatic hydrolysis followed reverse phase HPLC and UV spectroscopy on-the-fly (27). NMR samples were equilibrated to buffer (10 mM sodium phosphate, pH 6.5, 100 mM NaCl, 90% H2O and 10% D2O) using Amicon Centricon 3 concentrators (Millipore, Bedford, MA). For experiments in D2O, the sample solutions were lyophilized and exchanged with D2O (Cambridge Isotopes, Andover, MA). The VGL and VBL dodecamers were titrated with the various ADF oligomers and monitored by 1D NMR. The VGL/ADF duplexes had a final duplex concentration ranging from 0.5 mM to 1.2 mM.

UV and circular dichroism spectroscopy

Thermal denaturations and renaturations were monitored by UV absorbance at 260 nm with a Cary 3 spectrophotometer running WinUV version 3.00. Buffer conditions were identical to those for NMR. Three successive denaturations and renaturations were conducted over a temperature range of 5–55°C and a concentration range of RNA from 1 μM to 1 mM using cuvettes of 1.0 and 0.2 cm pathlengths. The temperature ramp rate was 1.0°C/min or 0.5°C/min with a data sampling interval of 0.2 min or 0.5 min. Circular dichroism spectra were collected (Jasco J600) at 4°C with 1 cm path length using a jacketed, cylindrical sample cell and RNA sample concentrations adjusted to ∼0.2 absorbance units at 260 nm (28,29).

NMR spectroscopy

NMR spectra were collected on a Bruker DMX 500 MHz instrument equipped with a triple-resonance 1H, 13C, and 15N probe and three-axis pulsed field gradient capabilities, on a Varian 600 MHz instrument equipped with a cryoprobe at NCSU, and on a Varian 800 MHz spectrometer at Duke University (Durham, NC). All NMR data were processed using NMRPIPE (30). Spectra were displayed and analyzed using NMRVIEW 5.0 (31) or SPARKY software (32). NOESY and TOCSY spectra using the MLEV-17 mixing sequence were collected at varying temperatures, and mixing times to aid in the assignment process. T1 relaxation experiments were conducted at various temperatures on the VGL/ADF-s2U34;Ψ39 duplex to identify H2 resonance. A DQF-COSY experiment was conducted at 20°C in D2O with the VGL/ADF-s2U34 and the VGL/ADF-s2U34;Ψ39 duplexes. Two-dimensional heteronuclear spectra aided in the assignment process. Natural abundance 1H-13C HSQC experiments were performed on VGL/ADF-s2U34 in H2O at 4°C and VGL/ADF-s2U34;Ψ39 in D2O at 20°C. 1H-31P HETCOR and a 1H-31P HeteroTOCSY were collected on VGL/ADF-s2U34;Ψ39 in D2O at 20°C. For samples with site-specific 15N labels, 1H–15N HSQC spectra were obtained in H2O at seven temperatures from 2°C to 35°C and at the concentrations used for structural analyses.

NMR restraints generation and structure calculation

Measurement of individual peak intensities, when possible, was performed using the box integration method in the SPARKY software. The cross-peaks were qualitatively classified as strong (0.0–3.5 Å), medium (0.0–4.5 Å), or weak (0.0–5.5 Å). For peaks that overlapped, standard A-form RNA distances were used when an A-form RNA was deduced from the NMR spectra. No restraints were used for the H5′ and H5′′ protons. Intervals with wide boundaries, 0.0–5.0 or 6.0 Å, were used for restraints involving the imino protons. A total of 292 restraints were used to generate the structure (Table 1). Consistent with NMR data, the nine G•C and U•A base pairs were subjected to six and four hydrogen-bond restraints, respectively (1.8–2.5 Å between the hydrogen and acceptor, and 2.7–3.5 Å between the heavy atoms). The G•U base pair was restrained using four hydrogen-bond distance restraints between the GO6 and the UN3H and N3 atoms, and between the UO2 and the GN1 and GN1H atoms. No hydrogen-bonds restraints were used for Ψ39•U162 and G163•A38 base pairs, but an NMR-derived, imino distance restraint between the U162N3H and Ψ39N3H was introduced with an interval of 0.0–6.0 Å. A total of 47 hydrogen bond distance restraints were finally used. Planarity restraints (5–50 kcal/mol Å2) were applied to Watson–Crick and G•U base pairing. Because of severe overlapping in DQF-COSY and HETCOR spectra, torsion angle restraints were set with wide ranges of ±30° and not determined from the direct measurement of coupling constants. β (P-O5′-C5′-C4′), γ (O5′-C5′-C4′-C3′) and ϵ (C4′-C3′-O3′-P) dihedral angles were set respectively to 178 ± 30°, 54 ± 30° and −153 ± 30°. When the presence of the regular A-form RNA structure could be determined, using both the 31P chemical shift and the expected connectivity patterns between the aromatic and sugar resonances, the α (O3′-P-O5′-C5′: −68 ± 30°) and ζ (C3′-O3′-P-O5′: −71 ± 30°) torsion angles were constrained to exclude the trans conformation. For all residues, except G163, the glycosidic torsion angles, χ [O4′-C1′-(N9/N1)-(C4/C2)], were restrained to the anti conformation (−158 ± 30°); the intensities of all the H1′-H6 to H8 connectivities exhibited smaller volumes than the H5 to H6 connectivity. For all residues except G163, the δ angles were restrained to be in C3′-endo conformation. A total number of 156 torsion angles restraints were used for the structure calculation (Table 1).

Table 1.

Restraints and structural statistics

NOEs distances restraints	292
Internucleotide NOEs	139
Intranucleotide NOEs	115
Involving iminos	38
Hydrogen bonds restraints	47
Torsion angles restraints	156
Planarity restraints	11
NOEs per residue	12.2
Total restraints per residue	18.7
Distance violations (Å)	None > 0.5
Dihedral violations (°)	None > 5.0
Backbone RMSD for all atoms
Compared to the average structure (Å)	2.55 ± 0.89
Stems RMSD:
(G158–G161/C40–C43) (Å)	0.80 ± 0.29
(A166–G169/C32–U35) (Å)	0.86 ± 0.39
Loop and adjacent residues (G161–U164/A37–C40) (Å)	1.01 ± 0.34

Molecular modeling of the VGL/ADF-s2U34;Ψ39 structure was achieved using CNS 1.1 (33). One hundred structures were calculated using standard NMR restraint annealing protocol (34). Structures with low total energy, without or with a low number of dihedral or distance restraint violations were chosen for further analysis. The helical parameters of the structure were analyzed using X3DNA software (35). To validate our structures, back calculations of the H5, H6 and H8 protons resonance chemicals shifts were conducted using MOLMOL (Johnson-Bovey model), or NUCHEMICS software (36). The output structures were visualized with MOLMOL (37) and PYMOL (38).

RESULTS

Thermodynamic stability and affinities of HIV-1 A-loop interactions with variously modified fragments of the tRNALys3 anticodon domain

The thermal stabilities and the circular dichroism spectra of the duplexes formed between the U-rich anticodon domain dodecamer (ADF) of tRNALys3 and the A-rich HIV Loop I indicated that the s2U34 and Ψ39 facilitated the interaction of the two RNAs. The dodecamer sequences, based on previous studies of the complex formed between tRNALys3 and the HIV-1 RNA (3), were constituted with substitution of C43 for A43 to reduce fraying of the duplex by forming a terminal C43•G158 base pair. This construct allowed us to focus our studies on the contributions made by the two internal base modification chemistries, s2U34 and Ψ39, and the possible mismatch base pairs. As it would be expected for the duplexes, the melting temperatures, Tm, of these bimolecular interactions were concentration dependent (Figure 2A) allowing the binding affinities to be extracted from the results of the UV-monitored thermal experiments (Table 2). The contribution of modified nucleosides to the stability of the VGL/ADF complex was dependent on the tRNALys3 modification introduced. Relative to the unmodified ADF (Tm = 30.0°C at 100 µM; Kd = 320 µM), the individual incorporations of Ψ39 and s2U34 raised the melting temperature of the VGL/ADF duplexes by +1.2°C and +3.9°C, and enhanced the affinity of the ADF for the VGL by some 8- and 72-fold, respectively. Incorporation of both the Ψ39 and s2U34 modifications significantly increased the Tm (+7.2°C) and the affinity of the ADF for the viral sequence (Table 2). The affinity of the doubly modified ADF-s2U34;Ψ39 (Kd = 14.2 µM) for the VGL was significantly stronger than that of the unmodified ADF (Kd = 324 µM) (Table 2). Interestingly, the affinity of the fully modified anticodon stem and loop domain for its cognate codon AAA in the A-site of the Escherichia coli ribosome was comparable (Kd = 3.1 µM) to that of the ADF-s2U34;Ψ39 for the VGL, while ribosomal codon binding of the unmodified stem and loop domain was too low to be determined (E. Gustilo and P.F. Agris, personal communication). The standard free energy values, ΔG°, from duplex formation of the VGL with the ADF-Ψ39, ADF-s2U34, and ADF-s2U34;Ψ39 were lower than that formed from the VGL and the unmodified ADF (Table 2). The ΔG° value for the doubly modified duplex was between that of the two singularly modified duplexes. We expect that the free energy of the VGL/ADF-s2U34;Ψ39 duplex is a composite of local enthalpic and entropic contributions, influenced by the effect of modified nucleosides on local structure and stability. Thus, it is not surprising that the ΔG° of the doubly modified duplex was not simply the sum of the two duplexes containing only one modification each (s2U34 or Ψ39).

Figure 2.

Thermal stability and base stacking of the VGL/ADF duplex. (A) UV-monitored, thermal stability assay of the VGL/ADF-s2U34 duplex. The duplex was repeatedly denatured and renaturated and monitored by UV-absorbance. The Tm of a duplex is concentration dependent. Results from monitoring three concentrations of the VGL/ADF-s2U34 duplex are shown: 1, 14 and 28 µM. (B) Circular dichroism spectra of the VGL/ADF duplexes. The circular dichroism spectrum of each duplex was collected with the RNA concentration of the duplex or the VGL alone adjusted to give an absorbance of 0.2 units (A260) at 25°C. Spectra were collected at 4°C and were not corrected for the background (buffer).

Table 2.

Thermodynamic parameters of VGL and VBL duplex formation with various ADF constructs

VGL or VBL	ADF of tRNALys3	Kd (µM)	ΔG°298	ΔΔG°298	Tm (100 µM)
VGL	ADF	324.0	−4.8	–	30.0
	ADF-Ψ39	43.0	−5.9	1.1	31.2
	ADF-s2U34	4.5	−7.3	2.5	33.9
	ADF-s2U34;Ψ39	14.2	−6.6	1.8	37.2
VBL	ADF	270.0	− 4.9	0.1	30.6
	ADF-s2U34	5.7	−7.2	2.4	33.0
VGL	ADF of tRNALys2	>1900	–	–

Gibbs standard free energy in kcal/mol was calculated at 25°C from: ΔG° = –RTlnK.

A fourth duplex (VGL/ADF-C), formed from a fully complementary ADF sequence in which A39C38 substituted for U39A38, exhibited a Tm (42.2°C) that was 8.3°C higher than the VGL/ADF-s2U34;Ψ39, and 12°C higher than the duplex formed with the unmodified ADF and two mismatched pairs. In addition to increasing thermal stability, the s2U34 and Ψ39 enhanced the degree of base stacking within the VGL/ADF duplexes (Figure 2B), as monitored by circular dichroism spectroscopy (39). The spectral ellipticity of the VGL/ADF was increased with the addition of the single and combined modifications. The duplex with ADF-s2U34 exhibited the highest degree of ellipticity and presumably the greatest degree of base stacking. Surprisingly, the ellipticity of the VGL/ADF-C was comparable to that of the VGL/ADF-Ψ39, and lower than that of the VGL/ADF-s2U34 (Figure 2B). Here, we see again that the ellipticity of the doubly modified complex, VGL/ADF-s2U34;Ψ39, is not simply the sum of that of the two singularly modified complexes VGL/ADF-s2U34 and VGL/ADF-Ψ39.

The ADFs had four contiguous uridines (U33–U36) complementary to the four contiguous adenosines of the dodecamer of the HIV-1 Loop I subtype G (VGL), but were still capable of binding to the Loop I substitutions that are found in other subtypes, such as the two guanosines of the B-subtype (VBL) (Figure 1D). The affinities of the VBL for the unmodified ADF and the ADF-s2U34 were comparable to the VGL (Table 2). There are three human tRNALys species. Notably, an unmodified ADF with the sequence of the human tRNALys1,2CUU having a C34 was hardly bound by the VGL (Table 2).

NMR studies of the VGL/ADF interaction: exchangeable proton resonances identification and assignment

A doubly modified (s2U34;Ψ39) ADF bound to the VGL, VGL/ADF-s2U34;Ψ39, was studied by NMR. Three other duplexes were investigated by NMR: a fully base-paired complement (VGL/ADF-C); a duplex containing only the pseudouridine modification, ADF-Ψ39 (VGL/ADF-Ψ39); and a duplex containing only the 2-thiouridine modification, ADF-s2U34 (VGL/ADF-s2U34) (Figure 1D). These additional constructs aided in the assigning of NMR signals and in elucidating the effect of the different modified nucleosides of tRNALys3 on the interaction of its anticodon domain with the HIV-1 A-rich Loop I (Figure 1D). All four duplexes were studied using classical procedures (40,41). One dimensional spectra of the various duplexes exhibited exchangeable protons resonances between 12.5 and 15 ppm that were attributed to seven hydrogen-bonded imino protons involved in standard Watson–Crick base pairs for the VGL/ADF-Ψ39, VGL/ADF-s2U34 and VGL/ADF-s2U34;Ψ39 duplexes (Figure 3 and Supplementary Data). The imino proton resonance of s2U34 was shifted downfield (14.23 ppm) compared to the corresponding resonance of the similar unmodified U34 (13.85 ppm). The downfield shift of the base paired imino proton of s2U34 has been previously observed (42) and could indicate that the sulfur is contributing to a stronger hydrogen bond through a change in ring current or stacking interactions.

Figure 3.

Imino proton resonances and NOESY sequential assignment NMR spectra of the VGL/ADF. (A) A 1D NMR spectrum of VGL/ADF-s2U34;Ψ39 (90% H2O/10% D2O and at 10°C). The region of the spectrum in which imino protons resonate (9.5–15.0 ppm) is shown. (B) Sequential assignment of the aromatics to H1′ resonances of the VGL/ADF-s2U34;Ψ39 duplex. The NOESY spectrum was recorded at 10°C (90% H2O/10% D2O) with a mixing time of 300 ms. For clarity, only a portion of the ‘NOE walk’ is displayed. The residues belonging to HIV-1 Loop I and to the tRNA are in green and purple, respectively. The H5-H6 connectivities are shown in rust.

Imino proton resonances between 10 and 12 ppm corresponded to those protons involved in non-canonical base pairs. The characteristic chemical shifts (43) and observation of an intense NOE connectivity between the N1H of G160 and N3H of U41, together with the imino–imino NOE connectivity between N3H of U41 and N1H of G161 allowed the unambiguous assignment of the imino protons of G160 and U41 in the VGL/ADF-C NOESY spectrum. The formation of this base pair was observed in all four duplexes including the VGL/ADF-s2U34;Ψ39 (Supplementary Data). The imino protons resonances of Ψ39-N1H, Ψ39-N3H and U162-N3H were identified and assigned on the basis of the VGL/ADF-Ψ39 and VGL/ADF-s2U34;Ψ39 NOESY spectra at low temperatures (2°C and 4°C, respectively) and of the 15N-HSQC spectra. The chemical shifts of the Ψ39 and U162 imino protons resonances agreed with previously published studies (44). Moreover, a weak imino–imino NOE connectivity between the Ψ39-N3H and the U162-N3H was observed on the 200 ms NOESY spectra of the VGL/ADF-Ψ39 duplex (Supplementary Data). In addition, NOE connectivities were observed between Ψ39-N1H and A38H2′ and A38H3′ on the NOESY spectra (data not shown). A broad imino resonance at 11.50 ppm was assigned to the imino proton of G163. Despite this proton being protected from exchange with solvent, no NOE connectivity between this proton and other exchangeable or non-exchangeable protons could be observed.

The thermal denaturations of the VGL/ADF-C, VGL/ADF-Ψ39 and VGL/ADF-s2U34;Ψ39 duplexes were followed by NMR (Supplementary Data) and were consistent with the results of the UV-monitored melting studies. The VGL/ADF-C exhibited a Tm between 40°C and 45°C, whereas the Tm values of the VGL/ADF-Ψ39 and VGL/ADF-s2U34;Ψ39 were between 25°C and 30°C.

Non-exchangeable protons resonances identification and assignment

The non-exchangeable resonances of the VGL/ADF-s2U34;Ψ39 duplex were identified and assigned with a variety of homo- and heteronuclear NMR experiments using standard procedures (41) and the help of the three other duplexes. Using the different duplexes, all of the aromatics and H1′ resonances were assigned, excepted that of G163 in VGL/ADF-s2U34, VGL/ADF-Ψ39 and the VGL/ADF-s2U34;Ψ39 (Figure 3B and Supplementary Data). Although assignment of the G163 H8 and H1′ resonances remains tentative because of severe overlapping of signals, an NOE connectivity was observed between the U164H6 and the G163H1′ in the VGL/ADF-Ψ39 duplex, as well as a strong NOE between the U164H6 and G163H2′ in the NOESY spectra of the VGL/ADF-Ψ39 and VGL/ADF-s2U34;Ψ39 duplexes at low mixing times. The chemical shifts of the H5 and H6 protons are known to be most affected by the ring current from the 5′ adjacent base (36). The H5 and H6 chemical shifts of U164 were consistent with protons experiencing a ring current effect from its 5′-adjacent G163 (36), indicating that the G163 and U164 were stacked, and consequently that G163 is inside the helix. The U162H6 and H5 chemical shifts were also consistent with a uridine experiencing the ring current of G161 in spite of a more important displacement toward the high-field for the U162H6 chemical shift as compared to U164H6. Moreover, the volume of the U162H5 to H6 NOE connectivity was less than all of the other H5-H6 NOE connectivities indicating that the U162 was more dynamic than the other pyrimidines. An inter-strand NOE connectivity between the A165H2 and the A37H2 was observed for the VGL/ADF-s2U34;Ψ39.

The Ψ39H6 proton resonated at 6.8 ppm, characteristic of a pseudouridine (45). An aromatic to aromatic NOE connectivity was observed between Ψ39H6 and A38H8, indicating that Ψ39 and A38 were stacked. Because of overlapping peaks, we were not able to unambiguously identify the connectivities of A38 to the A37 H2. The A168H2 proton exhibited an unusual NOE connectivity with the s2U34H1′, indicating an enhanced stacking that we attributed to the 2-thio modification. The chemical shifts of A38H8 and H1′ were consistent with an adenosine involved in a typical A-form RNA. The A38H2 proton, identified and assigned by T1 relaxation experiments and confirmed through the natural abundance 1H-13C-HSQC spectra, relaxed slowly, as did other H2 resonances for the VGL/ADF-s2U34;Ψ39 duplex. Taken together along with strong sequential NOEs between the aromatic to H2′ protons, this information showed that the entire duplex adopted a conformation similar to a canonical A-form RNA. Moreover, the data indicated that all the bases were stacked and inserted into the helix.

Sugar conformations and backbone geometry

The DQF COSY spectra showed that all the residues of the VGL/ADF-s2U34;Ψ39 duplex adopted the C3′-endo sugar pucker (JH1′–H2′ < 3 Hz) characteristic of an A-form RNA helix. A narrow spread of the 31P chemical shifts (2.2 ± 0.9 ppm) around that attributed to A-RNA helix-like values indicated that on average the phosphate backbone conformed to a regular A-RNA conformation without significant deformation (Supplementary Data) (41). Each of the aromatic to H1′ connectivities of the VGL/ADF-s2U34;Ψ39 had an intensity significantly lower than that for the H5–H6 indicating that all the bases, including Ψ39, adopted the anti form of the glycosidic torsional angle (46). Nevertheless, the resonance overlaps and tentative assignment for G163 did not allow us to determine its glycosidic torsional angle.

Structure of VGL/ADF-s2U34;Ψ39

The spectral data indicated that the structure of the VGL/ADF-s2U34;Ψ39 duplex was similar to a standard double helical A-form RNA, that all the bases were inserted into the helix and that the 2 × 2 internal loop 5′-U162G163-3′/3′-Ψ39A38-5′ was structurally organized. One hundred structures were calculated using a simulated annealing protocol (34). Among the 100 calculated structures, 41 converged to satisfy all of the NOE distance and dihedral angle restraints within 0.5 Å and 5°, respectively. The U162•Ψ39 and G163•A38 mismatch base pairs were observed with great attention. Among the 41 accepted structures, 23 presented an imino G•A base pair and the lowest total energy. The other 18 structures presented different kinds of G•A base pairs. Of the 23 structures exhibiting an imino G•A base pair, the 10 structures with the lowest total energies were selected for further analysis. In order to validate the selected structures, a chemical shift back-calculation was conducted for the H5 and H6 chemical shifts of C159 (standard), U162 and U164, and for the Ψ39H6 and G163H8. Indeed, the chemical shifts of the H5 and H6 were most influenced by the 5′ bases and thus, could be used as probes to validate our structure. The back-calculated chemical shifts for the U164H5 and the H6 were in excellent agreement with the observed chemical shifts for U164 (Supplementary Data). This confirmed the stacking of the U164 and G163 and moreover, tended to prove that the imino G•A base pair observed in the 10 selected structures was consistent with the NMR data. The predicted Ψ39H6 chemical shift was consistent with the observed chemical shift. The back-calculated and experimental chemical shifts for U162H5 were almost identical (Supplementary Data). This contrasted with U162H6 for which these chemical shift values displayed a difference of ∼0.7 ppm. In this case, the position of U162 in relation to G161 seemed to be different from that observed in our structures.

The overall structures are well defined with a mean backbone RMSD for all residues of 2.55 ± 0.89 Å. The structures have two canonical, A-form RNA stem regions of four and six Watson–Crick base pairs including the wobble G•U, respectively, separated by a structured 2 × 2 internal loop. Relative to the mean structure, the backbone RMSDs were 0.80 ± 0.29 Å and 0.86 ± 0.39 Å, respectively, for the G158-G161/C40-C43 (Figure 4A) and the A166-G169/C32-U35 (Figure 4B) A-form RNA stems. However, the backbone RMSD for the loop and the immediately adjacent residues G161-U164/A37-C40 was greater than that of the two A-form RNA stems with a value of 1.01 ± 0.34 Å (Figure 4C).

Figure 4.

Structure of the VGL/ADF-s2U34;Ψ39 duplex. The structure of the VGL/ADF-s2U34; Ψ39 duplex was determined from NMR-restrained, molecular dynamics. Members of a family of 10 structures without distance or torsion angle violations were superimposed. Both the U162•Ψ39 and imino G163•A38 base pairings are within the structures. (A) Best fit for the stem residues G158-G161/C40-C43. (B) Best fit for the stem residues A166-G169/U35-C32. (C) Best fit for internal residues G161-U164/A37-C40. (D) Superposition of the ten best structures (left) and the mean structure (right). (E) The G163•A38 and U162•Ψ39 base stacking (top) and the geometry of the two base pairings (bottom). (F) The mean G163•A38 and U162•Ψ39 base stacking (top) and mean geometry of the two base pairings (bottom). (G) The imino and sheared G•A base pairs. Top: The imino G•A base pair employs the GN1H as a donor for the AN1 acceptor, whereas the carbonyl oxygen of G is the acceptor for an amino proton of A. Bottom: The sheared G•A pair employs an amino proton of G as a donor with the AN7 as the acceptor, and an amino proton of A as a donor for the N3 of G.

Both G163, being loosely restrained, and U162 exhibited conformational dynamics (Figure 4E and F). The average C1′–C1′ distance between U162 and Ψ39 was 11.4 ± 0.5 Å, more than the conventional U•U base pair that averages a C1′–C1′ distance of ∼9 Å (47–49). The average hydrogen bond distance observed between the U162N3H and the Ψ39N3H was 3.8 ± 0.6 Å. This distance is consistent with the moderately weak NOE observed between the two imino protons. In comparison, distances between the U162N3 and Ψ39O4 and between the U162O4 and Ψ39N3 were on average 5.4 ± 0.6 Å and 4.4 ± 0.9 Å, respectively. The former distance excluded the formation of a direct hydrogen bond, and possibly the latter as well, but did not exclude the formation of a water-mediated hydrogen bond for either. The aforementioned distance between the U162 and Ψ39 agreed with the formation of an imino base pair between G163 and A37. Indeed, the C1′–C1′ distance between G163 and A37 was nearly 12.4 ± 0.2 Å in the 10 selected structures. The average distance between Ψ39N1 and the Ψ39OP1 was 4.3 ± 0.6 Å. Once again, this distance did not exclude the possible formation of a hydrogen bond or a water-mediated hydrogen bond between the N1H imino proton of Ψ39 and a phosphate atom, such as that observed in crystallographic structures (50).

DISCUSSION

Modifications stabilize the interaction of the tRNALys3 anticodon domain with the HIV-1 Loop I

Replication of the human immunodeficiency virus (HIV) requires base pairing of the reverse transcriptase primer, human tRNALys3, to the viral RNA. Though conclusions from our UV-monitored studies of the thermal denaturations and structures of the duplexes are limited by the fact that the duplex sequence was altered to avoid frayed ends, the circular dichroism spectra and the NMR-derived structure underline the contributions of the 2-thio group and the pseudouridine to the stability of the interaction between the Loop I and the ADF. One of the most dramatic affects on nucleotide conformation and thermodynamic stabilization is the substitution of the carbonyl oxygen at C2 of uridine by a sulfur atom (s2U). The van der Waals’ radius of sulfur being some 20% larger than oxygen contributes to the greatly increased stability of the 3′-endo, gauche+ ring pucker and anti conformation of s2U and causes the adjacent nucleotides to favor 3′-endo ring pucker and A-form geometry (14,18,51–54). The extremely stable nucleoside conformation of s2U and that of the adjacent nucleosides stabilizes the canonical base pairings. In the NMR spectra, the downfield chemical shifts for the H5, H6 and H1′ protons relative to that of the unmodified uridine are indicative of an altered ring current that would contribute to the enhanced base pairing, as well as to base stacking. Though the N1 proton of Ψ has long been suspected to be responsible for contributing through direct hydrogen bonding to base pairing, little evidence has been found in investigations of RNAs. However, observation of the N1 proton resonance indicates that it is sufficiently protected from chemical exchange with the solvent water to invoke a stabilizing effect on base pairing through water-mediated hydrogen bonding, enhanced base stacking, or both. As with previous studies of Ψ (55,56), we saw no evidence for hydrogen bonding of the N1 proton of Ψ. Enhanced base stacking is observed when Ψ is in stems or in loops (57–59) and this could contribute to the greater stability of the U-rich anticodon domain interaction with the A-rich HIV Loop I. The modification ms2t6A37 is missing from the studies presented here. This modification excludes the formation of a canonical U•A base pair across the anticodon loop (60). The ms2t6A37 and the t6A37 modifications decreased the overall Tm of the ASL (19,60). However, crystallographic studies of the ASL bound to cognate and wobble codons on the ribosome (61) indicate that the ms2t6A37 modification would act as a hydrophobic platform in stabilizing the base pairing between the tRNA anticodon and the series of As in the Loop I of HIV-1.

The HIV-1 Loop I/tRNALys3 ADF structure

The dodecamer duplex formed between the VGL and the ADF-s2U34;Ψ39 exhibited an A-form RNA conformation with a relatively well-organized internal loop composed of two mismatch base pairs. The NMR data clearly demonstrated that C40, Ψ39, A38 and A37 are stacked and inserted into the helix. Nevertheless, despite a relatively well-organized internal loop, the VGL/ADF-s2U34;Ψ39 duplex exhibited a melting temperature lower than the full complement, VGL/ADF-C. The U162 residue was dynamic, as suggested by the weakness of the H5–H6 connectivity in comparison to other uridines or cytidines. The difference between the observed and the back-calculated chemical shifts indicated that the stacking observed in the structures between G161 and U162 does not reflect the NMR data. An intense NOE connectivity is expected between the uridine imino proton of residues involved in a U•U base pair (49,62). In those base pairs, the average distance between the two imino protons is ∼2.5 Å, significantly lower than the average distance observed in our structures between U162N3H and Ψ39N3H (3.8 Å). Although a classical U•U base pairing cannot be formed, the observed distance between acceptor and donor does not exclude the formation of a water-mediated hydrogen bond stabilizing the Ψ•U interaction.

A G•A base pair can adopt more than nine conformations, but two most often observed are the imino and the sheared (43,63) (Figure 4G). The type of G•A base pair depends on the neighboring base sequences and pH (64). In the imino G•A conformation, an NOE connectivity between the GN1H and the AH2 is observed (65), whereas in the sheared G•A the GN1H proton is observed at very low temperatures (66). Conformational dynamics coupled with the presence of the neighboring weak Ψ39•U162 pair could explain the broad proton resonance attributed to the G163N1H, and could explain why the expected connectivity between G163N1H and A38H2 could not be observed. The chemical shift of the U164H5-H6 cross peak and its back-calculated chemical shift were similar and in good agreement with a uridine experiencing a 5′-guanosine (G163) ring current. This is supportive of an imino G•A base pairing. Moreover, it has been suggested that a single G•A base pairing cannot adopt the sheared G•A conformation observed in a tandem G•A pair (67). Formation of an imino base pair between G163 and A38 would result in an average C1′–C1′ distance of 12.5 Å. Such a base pair could prevent the formation of a Ψ•U base pair with a C1′–C1′ distance of ∼9Å. We observed that the average C1′–C1′ distance for the Ψ39 and U162 pair was 11.4 Å. These results are consistent with the 31P spectrum of the VGL/ADF-s2U34;Ψ39, and previous studies showing that in contrast to a sheared G•A, an imino G•A base pair resulted in little to no deformation of the backbone (68,69). A model of a 2 × 2 internal loop composed of 5′-UG-3′/3′-UA-5′ was proposed from NMR and thermal stability studies of a model rRNA system (70). Despite a general pre-organization of the 2X2 internal loop, no hydrogen-bond was observed between the two U's, and the G was external to the helix. Nevertheless, a G•A base pair was not excluded and it was suggested that an imino G•A base pair was possible. We found that the 5′-UG-3′/3′-ΨA-5′ was relatively well-organized with stacking of the C40, Ψ39 and A38, and that a water-mediated hydrogen-bond between Ψ39 and U162 probably existed. Furthermore, G163 was inside the helix and probably involved in an imino G•A base pair that would be supported by the NMR data.

The HIV Loop I/tRNA anticodon interaction: a ‘kissing’ loop complex?

Interactions between the A-rich Loop I and the anticodon domain of tRNALys3 are essential for priming the initiation of HIV-1 reverse transcription (8,9,71). In fact, the A-rich Loop I region or the complex formed by the A-rich Loop I with the tRNALys3 anticodon could represent promising targets for developing new therapies. Polyamides target the A-loop sequence (PNAAL), and by binding the A-loop prevent tRNALys3 from annealing to the viral RNA. Consequently, initiation of reverse transcription in vitro is blocked (71). Therefore, an understanding of the interaction between the HIV-1 A rich Loop I and the tRNALys3 anticodon is imperative.

The annealing of the tRNA primer to the PBS requires unwinding of the acceptor and TΨC stems of the tRNA cloverleaf structure, thus releasing the tRNA sequence complementary to the PBS. Reverse transcriptase (RT) binding facilitates the annealing of the tRNA to the PBS (72). The nucleocapsid protein (NC) is known to have general unwinding and strand renaturation activities and this protein or its precursor Gag has been implicated in the placement of tRNA onto the viral genome (73,74). HIV-1 NC is a short basic protein with two zinc finger domains and functions as a nucleic-acid chaperone (75). Like RT, NC appears to destabilize base pairing in tRNA molecules, without a complete melting of the structure (76). Moreover, it has been suggested that the annealing of tRNA to PBS could open the TΨC stem of the tRNA, allowing residues 48–55 to interact with the PAS portion of HIV-1 (77). Our expanded duplex model was constructed with the purpose of determining the effect of modifications on the binding the HIV Loop I by tRNALys3. The equilibrium binding constant that we observed for the s2U and Ψ-modified 12 mer to the Loop I sequence (Kd = 14.2 µM) was indicative of a higher affinity, but consistent with that found for the kissing loop interactions of the fully modified anticodon stem and loop (Kd = 71 µM) (19). There is evidence that the mechanism of duplex formation may include an initial kissing loop interaction (19,20). In an annealing experiment monitored by NMR, the anticodon stem of a ‘recombinant’ tRNALys3 produced in E. coli remained intact while the loop was reported to have undergone a conformational exchange without a directly observable loop–loop interaction (78). Chemical probing of infected cells and virions indicated that a kissing loop interaction of the anticodon and Loop I could precede and drive the annealing of the two RNAs (79). Therefore, the unwinding of the tRNALys3 anticodon stem and that of the Loop I stem may occur subsequent to a kissing loop interaction. Whatever the chronology of the steps in forming a duplex, the interaction between the U-rich anticodon loop of the primer tRNALys3 and the A-rich Loop I of HIV is stabilized by the tRNA's natural modifications.

HIV subtypes include the A/G forms predominantly from Africa, and the B form predominantly from Europe and North America, representing almost all sequenced isolates. These isolates use tRNALys3 as the primer for reverse transcription, as do all lentiviruses. In addition, they have in common the A-rich Loop I sequence bound by the tRNALys3 anticodon with the one difference of the B form having the G166G167 substitution for the A166A167 in the A/G subtype. The VBL/ADF duplex had a G166•U35 wobble pair adjacent to a modified G167•s2U35 base pair, and exhibited a stability little different from that of the ADF-s2U34 bound to the VGL. The Loop I of the viral A/G subtype was protected from chemical probing in the initiation complex for reverse transcription (79). However, the Loop I of the viral B subtype was not protected. Therefore, the interaction of the tRNALys3 anticodon with the Loop I of the viral B subtype may not be as stable as indicated by equilibrium binding experiments. It is possible that our in vitro system might not adequately reflect the in vivo assembly of the initiation complex for the B subtype of the virus.

The anticodon loop nucleoside sequences of tRNALys1,2 are identical to that of tRNALys3 except for a C34 substituting for mcm5s2U34 and t6A for ms2t6A37. Though the C34 that would pair with G167 of the VBL would be expected to improve the stability of the duplex, C34 would miss pair with A167 of the VGL and the stability of the duplex would decrease. The tRNALys1,2 are not used as the primer in the natural viral isolates. The viral primer binding site, the A-rich Loop I sequence and the recruitment of the host cell tRNALys3 by the HIV proteins all play important roles in maintaining the virus’ dedication to tRNALys3 as the primer of reverse transcriptase. The detailed biochemistry and structural biology of the recruitment and relaxation of the tRNALys3 structure and its annealing to the HIV and other lentivirus genomes are yet to be determined.

Accession number

The coordinates of the 10 lowest energy structures, NMR restraints and VGL/ADF-s2U34;Ψ39 chemical shifts have been deposited in the Protein Data Bank (accession number: 2K7E) and in the Biological Magnetic Resonance Bank (BMRB entry: 15915). This material is available free of charge via the Internet at http://pubs.acs.org.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

National Institutes of Health [R01 GM23037 to P.F.A.]; the National Science Foundation [MCB MCB-0548602 and MCB-9986011 to P.F.A.]; the TRIoH Integrated Project of the European Union [LSHB-CT-2003-503480 to J.V.]; Polish Committee for Scientific Research Grant [to M.A.]. Funding for open access charge: National Science Foundation (MCB-0548602).

Conflict of interest statement. A potential conflict arises in that Dr Andrzej Malkiewicz is a major shareholder in a start-up biotech company (TRANA Discovery, Inc.) interested in the annealing of the HIV primer tRNA to the HIV genome.