Internal RNA 2′-O-methylation on the HIV-1 genome impairs reverse transcription

Abstract

Viral RNA genomes are modified by epitranscriptomic marks, including 2′-O-methylation that is added by cellular or viral methyltransferases. 2′-O-Methylation modulates RNA structure, function and discrimination between self- and non-self-RNA by innate immune sensors such as RIG-I-like receptors. This is illustrated by human immunodeficiency virus type-1 (HIV-1) that decorates its RNA genome through hijacking the cellular FTSJ3 2′-O-methyltransferase, thereby limiting immune sensing and interferon production. However, the impact of such an RNA modification during viral genome replication is poorly understood. Here we show by performing endogenous reverse transcription on methylated or hypomethylated HIV-1 particles, that 2′-O-methylation negatively affects HIV-1 reverse transcriptase activity. Biochemical assays confirm that RNA 2′-O-methylation impedes reverse transcriptase activity, especially at low dNTP concentrations reflecting those in quiescent cells, by reducing nucleotide incorporation efficiency and impairing translocation. Mutagenesis highlights K70 as a critical amino acid for the reverse transcriptase to bypass 2′-O-methylation. Hence, the observed antiviral effect due to viral RNA 2′-O-methylation antagonizes the FTSJ3-mediated proviral effects, suggesting the fine-tuning of RNA methylation during viral replication.

Graphical Abstract

Graphical Abstract.

Introduction

More than one hundred RNA post-transcriptional modifications have been described so far (1). Among them, N⁶-methyladenosine (m⁶A) is the most widespread. The N⁶-methylation status is notorious for being finely regulated by cellular writer, eraser and reader proteins, and is involved in many cellular physiological and pathological processes [see review (2)]. In contrast, 2′-O-methylation (N_m) is a stable epitranscriptomic mark of the ribose moiety. These modifications are ubiquitous and can be found on both cap and internal sites depending on the RNA template, thereby modulating the structure and function of tRNA, rRNA and mRNA (3). N_m is involved in important intra- and intermolecular interactions for secondary structure (4–6), and can improve stability when found on the RNA cap structure (7) or at the 3′ end of the template to prevent degradation (8–10). When found on mRNA or rRNA internal positions, N_m can disrupt (11,12) or tune translation, respectively [reviewed in (13)]. Interestingly, cap N_m allows immune receptors such as RIG-I and MDA5 to discriminate self- from non-self-RNA (14,15).

In the nucleus of eukaryotic cells, N_m addition is mediated by 2′-O-methyltransferases (MTases) such as FTSJ3 (16) and CMTR-1 (17,18) for rRNA and mRNA as substrates, respectively. Another methylation event involves small nucleolar RNAs (snoRNAs) guiding a ribonucleoprotein complex, allowing a defined methyl addition at base precision (19). N_m is detected and mapped with increasing precision and frequency using techniques such as RiboMethSeq (20), N_m-seq (21), liquid chromatography–tandem mass spectrometry (21) and recently nanopore sequencing (22), thereby outlining the importance of variations of N_m on cellular RNA during human diseases (23,24) and viral infections (25). Remarkably, methods such as RiboMethSeq require low amounts of starting material and are sensitive enough to detect N_m on low-abundance RNA species, such as viral RNA.

RNA viruses are not spared from the post-transcriptional modification process during their replication cycle [for a review, see (26)]. Human immunodeficiency virus type-1 (HIV-1) exhibits an original way to methylate its own RNA genome (gRNA) by hijacking the nucleolar FTSJ3 MTase that is recruited to the TAR RNA-binding protein (TRBP) (27). Hence, the HIV gRNA is decorated with at least 17 internal N_m sites, thereby limiting sensing by MDA5 and subsequent interferon production at the next infectious cycle. Conversely, virions produced in ΔFTSJ3 cells trigger the production of MDA5-mediated interferon. 2′-O-Methylation also protects the viral genome from degradation mediated by the cellular endonuclease ISG20 (28). The HIV-1 methylated RNA genome escapes the innate immune defences by displaying self-like marks and counteracting restriction factors such as ISG20.

N_m has also been shown to impede the reverse transcription activity of some retroviral polymerases, as is the case for the avian myeloblastosis virus (AMV) enzyme at low dNTP concentrations (29,30). The AMV reverse transcriptase pauses during the polymerization of the cDNA strand just before adding a cognate dNTP opposite an N_m base on the RNA template (31). This finding suggests that 2′-O-methylation may have antiviral effects on the efficiency of HIV reverse transcriptase activity during reverse transcription, in opposition to the proviral effects of N_m-modified HIV-1 viral genomes mentioned above (27,28).

In this work, we provide evidence for a significant role played by N_m during the retroviral life cycle. We show that reverse transcription of HIV-1 genomes produced in wild-type (WT) cells is markedly less efficient than that of viral genomes produced in methylation-deficient ΔFTSJ3 cells. We then characterized the capacity of HIV-1 reverse transcriptase to elongate a DNA primer annealed to 2′-O-methylated RNA templates in vitro, and relate that efficiency to the dNTP concentration. Our data show that the presence of 2′-O-modified nucleotides (such as G_m > A_m> U_m, but not C_m) induces a pause during elongation at the −1 site relative to the methylated site, especially at low dNTP concentrations, without increasing mutation rates. Data point to a decrease in the affinity of a dTTP substrate opposite an A_m template site as the limiting factor resulting in stalling of HIV reverse transcriptase. The mechanistic basis of the latter was investigated using reverse transcriptase mutants, showing the importance of K70 in the reverse transcriptase fingers domain for helping the polymerase bypass N_m sites. On one hand, our results suggest differential replication efficiencies in quiescent cells known to contain low dNTP concentrations compared with activated lymphocytes which harbour higher nucleotide pools. On the other hand, they demonstrate for the first time that 2′-O-methylation is not only involved in proviral mechanisms (3,27,28), but also critically impacts the viral polymerase activity at low dNTP concentrations, pointing to a fine-tuning of such epitranscriptomic modifications during viral replication.

Materials and methods

Synthetic RNA templates and DNA primers

All RNA templates were purchased from Biomers with high-performance liquid chromatography (HPLC) purification and resuspended in 10 mM HEPES pH 7.0, 10 μM EDTA. 5′-Cy5 or unmodified DNA primers were purchased from Biomers or Eurofins and resuspended in water. Annealing occurred by mixing templates and primers at a ratio of 1.5:1 with KCl added to 66 mM, and heating at 70°C for 10 min before slowly cooling down to room temperature.

Plasmid construction

The WT pol gene of the HIV reverse transcriptase p66 subunit from pNL4-3 strain (GenBank accession no. MN685337.1) with a C-terminal 6× His tag was cloned into a pTrc99a vector under the control of the trc promoter with an ampicillin cassette, as described previously (32,33). The construction of HIV protease plasmid was reported in (33). Mutants were designed by polymerase chain reaction (PCR) with Phusion® High-Fidelity DNA Polymerase (NEB) and specific primers shown in Supplementary Table S1.

Expression and purification of wild-type and mutant reverse transcriptase

The co-expression of WT or mutant reverse transcriptase p66 with the HIV protease was performed in Escherichia coli XL1 blue cells. In this way, part of p66 is cleaved by the protease, leading to the assembly of competent asymmetric dimers of p66 and p51 subunits. Cultures were grown at 37°C in 2YT medium, supplemented with ampicillin (100 μg/ml) and kanamycin (50 μg/ml) to OD₆₀₀ = 0.6–0.8. They were then induced with 1 mM isopropyl-thio-β-d-galactoside (IPTG) and harvested after an overnight incubation at 37°C. Lysates were resuspended in buffer A [40 mM Tris pH 7.5, 25 mM EDTA, 10% sucrose, 1 mg/ml lysozyme, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM benzamidine, 5 mg/ml DNase] under agitation for 15 min, then diluted in an equivalent volume of buffer B [0.8% IGEPAL CA 630, 20 mM dithiothreitol (DTT), 1 mM PMSF, 2 mM benzamidine]. After two cycles of sonication, centrifuged supernatants were filtered through a 22 μm cut-off membrane and then loaded on a HiTrap® Heparin 5 ml column (Cytiva) previously equilibrated inBinding Buffer (20 mM Tris–HCl pH 8.0, 1 mM EDTA, 1 mM DTT, 2% glycerol). Fractions were eluted with a linear gradient of Elution Buffer (20 mM Tris–HCl pH 8.0, 1 mM EDTA, 1 mM DTT, 2% glycerol, 1 M NaCl) at a flow rate of 0.8 ml/min. Pooled fractions were diluted 5-fold in N1 buffer (5 mM imidazole, 40 mM KPO₄ pH 7.4, 10% glycerol, 0.1% IGEPAL CA 630, 0.32 M KCl), and adsorbed onto a 2 ml chelating Sepharose column loaded with nickel. The affinity column was washed with 50 ml of N1 buffer, then 50 ml of N2 buffer (45 mM imidazole, 40 mM KPO₄ pH 7.4, 10% glycerol, 0.1% IGEPAL CA 630, 0.32 M KCl), prior to elution with 10 ml of N3 buffer (240 mM imidazole, 40 mM KPO₄ pH 7.4, 10% glycerol, 0.1% IGEPAL CA 630, 0.32 M KCl). The N3 buffer was exchanged against Storage Buffer (40 mM KPO4 pH 7.4, 10% glycerol, 0.5 mM EDTA, 1 mM TCEP, 20 mM KCl) by dialysis or with a desalting column (PD10, GE Healthcare), and samples were concentrated and stored at −20°C.

Primer elongation assays and quantification

All reverse transcriptase enzymes, primer/template (P/T) and dNTP pre-mixes were prepared separately on ice at the desired concentrations by diluting the respective biomolecules in water and reaction buffer (50 mM Tris pH 8.0, 50 mM KCl, 0.05% Triton X-100, 10 mM MgCl₂ final concentrations). Reverse transcriptase and P/T were incubated at 37°C for 5 min prior to the addition of dNTP mix to start the elongation. At several time points (i.e. 0, 0.5, 1, 5, 10, 20 and 50 min), a fraction of the reaction mix was quenched in formamide buffer with bromophenol blue and 10 mM EDTA at a ratio of 1:5, then heated at 70°C for 10 min, cooled on ice and stored at −20°C. Samples of 4–7 μl were sujected to denaturing polyacrylamide gel electrophoresis (PAGE; 17–20% acrylamide, 7  M urea, TBE buffer) and scanned by a fluorescence imager (Amersham Typhoon). The gel analysis was done using ImageQuantTL software (Cytiva) by picking out three bands on every lane (referred to as 10-mer, 26-mer and 40-mer products). The percentage of each band was used to build plots on GraphPad Prism version 5.00 for Windows (GraphPad Software, San Diego, CA, USA, www.graphpad.com).

Determination of K_m and V_max for dTTP

Primer elongation assays were performed as described above with WT or mutant reverse transcriptase at 0.8 μM and P/T at 0.2 μM, but with no pre-incubation time, and with increasing concentrations of dTTP (0.01, 0.1, 0.2, 0.5, 1, 5, 10 and 20 μM for WT reverse transcriptase and mutants K70E, L74V and K82A, and 0.01, 0.1, 0.2, 0.5, 1, 5, 10, 20, 50, 100, 200 and 500 μM for mutant K70A). A duplex of 24-mer RNA template annealed to a 17-mer Cy5-labeled DNA primer was used. The resulting products were separated by denaturing PAGE (20% acrylamide, 7  M urea, TBE buffer). Quantification of the 17-mer (lowest band) and elongated DNA (all the bands detected above the 17-mer) was carried out on ImageQuantTL. About 0.8 pmol was loaded on every lane, the total of which is shared between the volume of the 17-mer and the volume of the elongated bands. The P/T quantity (pmol) of elongated products is then plotted against time (min) for each experiment and the linear part of the curve (usually between 0 and 5 min) is fitted by linear regression. The slope coefficients were designated as initial rates (V_i, pmol/min), and plotted against dTTP concentrations, resulting in curves that were fitted with the Michaelis–Menten analysis using GraphPad Prism to determine K_m and V_max values.

Reverse transcriptase excision activity

The WT reverse transcriptase of HIV-1 was produced and purified as mentioned above, except as a last step after the Ni affinity column it was buffer exchanged into 75 mM NaCl and 10 mM Tris–HCl pH 8 by gel filtration. The excision reaction was triggered at 37°C by mixing the reaction buffer and the reverse transcriptase (0.8 μM) with either a control or an N_m-modified RNA template (0.2 μM) paired with a 17-mer DNA primer, and addition of a range of PPi concentrations (0, 1, 5, 10, 20, 50 and 100 μM). As in primer elongation assays, samples were quenched in formamide, bromophenol blue and 10 mM EDTA buffer, at several time points (i.e. 0, 0.5, 1, 5, 10, 20 and 50 min), at a ratio of 1:5, prior to being heated at 70°C for 10 min, cooled on ice and stored at −20°C. Samples of 4 μl were loaded on a denaturing PAGE gel (20% acrylamide, 7  M urea, TBE buffer) and scanned by a fluorescence imager (Cytiva Typhoon). The data showed enhanced excision from the methylated template. The sum of intensity measurements for all the excision product bands was numerically fit to a burst kinetics model to yield an amplitude for the initial rapid exponential burst phase followed by a steady-state linear increase in product. The rates of the burst phase are too rapid to be measured in these manual quench experiments. The observed burst amplitudes were fit to a standard concentration dependence model, Y_obs = Y_max × [PPi]/(K_d,app + [PPi]) to determine the concentration of burst-competent complexes in the sample.

Fluorescence polarization

RNA templates and Cy-5-labeled DNA primers were annealed as described above. Duplexes of RNA/DNA were incubated with increasing protein concentrations in reaction buffer (50  mM Tris pH 8, 50  mM KCl, 0.05% Triton X-100, 10 mM MgCl₂ final concentrations) in black non-binding Greiner microplates. Fluorescence polarization (FP) measurements were performed on a microplate reader (PHERAstar FS; BMG Labtech) with an optical module equipped with polarizers and using excitation and emission wavelengths of 590 and 675 nm, respectively. Dissociation constants (K_d) were determined using a quadratic binding equation curve fitting, with a = 1, b = ProbeConc + K_d + X, c = ProbeConc × X , and Y = {[b– sqrt(b × b – 4 × a × c)]/(2 × a)] × (Bmax/ProbeConc) + Baseline}.

High-throughput sequencing

A first in vitro reverse transcription step was performed with the purified WT reverse transcriptase, with an unmodified control or a 2′-O-methylated RNA template (Supplementary Figure S1), both annealed to the same DNA primer, and with either 4, 40 or 250 μM dNTPs. After a 2 h incubation at 37°C, a 10 min denaturation at 70°C and a brief centrifugation, 100 ng of the resulting DNA strands were submitted to 25 cycles of PCR amplification using cognate primers (FW: AGTCATTGAGCTCGATGTGCACCTCCC, REV: GCAGTCAGCAGCTCGCC) and the Phusion® High-Fidelity DNA Polymerase (NEB), following the manufacturer's instructions. After DNA quantification using the Qubit® dsDNA HS Assay Kit and Qubit 2.0 fluorometer (ThermoFisher Scientific), libraries for sequencing were built by adding barcodes for sample identification, and primers to DNA using AB Library Builder System (ThermoFisher Scientific). To pool the barcoded samples equimolarly, a quantification step by real-time PCR using the Ion Library TaqMan™ Quantitation Kit (ThermoFisher Scientific) was carried out. An emulsion PCR of the pools and loading on a 520 chip were performed using the automated Ion Chef instrument (ThermoFisher Scientific). Sequencing was processed using the S5 Ion torrent technology (ThermoFisher Scientific) following the manufacturer′s instructions. Consensus sequences were obtained after trimming of reads (reads with quality score < 0.99) and mapping the reads on a reference (the DNA copy of RNA 783A, Supplementary Figure S1) using CLC genomics workbench software v.21 (Qiagen). Parameters for reference-based assembly consisted of match score = 1, mismatch cost = 2, length fraction = 0.5, similarity fraction = 0.8, insertion cost = 3 and deletion cost = 3. Variants with frequency >0.5% were studied, meaning that our analysis aimed at identifying single mutations repeatedly found at the same sites on at least 0.5% of the reads.

Sequence alignments

HIV-1 and AMV reverse transcriptase amino acid sequences of their respective catalytic subunit, i.e. p66 and p95, were aligned using the Multalin webserver (http://multalin.toulouse.inra.fr/multalin/) (34), while the figure was rendered using ESPript 3.0 (ESPript; https://espript.ibcp.fr) (35) and centred on the fingers domain of HIV reverse transcriptase. The alignment of 6802 sequences from all complete sequences of the HIV-1/SIVcpz pol gene from the year 2021 was carried out on the online database https://www.hiv.lanl.gov/ and the conservation was established on WebLogo 3 (36).

Cells and viruses

HEK293T cells knocked-out for FTSJ3 (HEK-FTSJ3 KO) were kindly provided by Y. Bennasser (IGH, Montpellier, France) (27). HEK-FTSJ3 KO and parental HEK293T (HEK WT) cells were cultured at 37°C and 5% CO₂ in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum (FCS), supplemented with 1% penicillin–streptomycin (ThermoFisher Scientific). Jurkat cells were cultured at 37°C and 5% CO₂ in RPMI 1640 medium containing 10% FCS, supplemented with 1% penicillin–streptomycin (ThermoFisher Scientific). Luciferase-encoding HIV-1 plasmid (pNL4-3.Luc.R-E-) was kindly provided by N. Landau (Aaron Diamond AIDS Research Center, The Rockefeller University, New York, USA) (37). The HIV-1 reverse transcriptase mutants K70E, L74V and K82A were generated by site-directed mutagenesis on the pNL4-3.Luc.R-E- proviral plasmid, using the QuikChange II site-directed mutagenesis kit (Agilent) and cognate primers (Supplementary Table S1). VSV-G pseudotyped luciferase-encoding HIV-1 particles were produced by transient transfection of HEK293T (WT or FTSJ3 KO) with pNL4-3.Luc.R-E- and pVSV-G using calcium phosphate, and were concentrated by ultracentrifugation for 1 h at 22 000 g (Beckman Coulter) at 4°C (26). Replication-competent, infectious HIV-1 particles were produced by transient transfection of HEK293T cells (WT or FTSJ3 KO) with pNL4-3 and were concentrated using a Lenti-X Concentrator (Takara). Virus yields were measured by p24 enzyme-linked immunosorbent assay following the manufacturer’s instructions (Lenti-X p24 Rapid Titer Kit, Takara). All pseudotyped and infectious HIV-1 particles were treated with 250 U/ml benzonase (Merck) at 37°C for 20 min before being used in the experiments in order to remove residual DNA.

Endogenous reverse transcription (ERT) assays

Viral supernatants containing 75 ng of p24 were mixed with 2× ERT buffer (100 mM Tris–HCl, pH 8.0, 0.5 mM dNTPs, 6 mM MgCl₂, 6 mM EDTA and 0.016% NP-40) and incubated at 37°C for 5 or 10 h. Reactions were terminated by the addition of an equal volume of stop buffer (2% SDS, 20 mM EDTA, 200 μg/ml proteinase K) followed by a 3 h incubation at 56°C. Following ERT, viral DNA was phenol–chloroform extracted, ethanol precipitated and resuspended in 34 μl of H₂O. ERT products were digested with DpnI to eliminate residual transfection plasmid. Quantitative PCR (qPCR) was performed in triplicate using Takyon ROX SYBR MasterMix blue dTTP (Eurogentec) on an Applied Biosystems QuantStudio 5 (ThermoFisher Scientific), using the following program: 3 min at 95°C followed by 40 cycles of 15 s at 95°C, 20 s at 60°C and 20 s at 72°C. Rreverse transcriptase products were quantified using M667 and M661 primers, which detect the presence of complete minus-strand DNA (38), and results were normalized to expression levels of glyceraldehyde phosphate dehydrogenase (GAPDH) transcripts. Primer sequences were as follows: M667, GGCTAACTAGGGAACCCACTG; M661, CCTGCGTCGAGAGAGCTCCTCTGG; GAPDH Forward, ACTTCAACAGCGACACCCACT; GAPDH Reverse, GTGGTCCAGGGGTCTTACTCC.

Reduction of the intracellular dNTP pool

HEK293T or Jurkat cells were pre-treated with 1 mM hydroxyurea (Merck) for 16 h. Cells were washed in propidium iodide (PI) staining buffer (100 mM Tris pH 7.4, 150 mM NaCl, 1 mM CaCl₂, 0.5 mM MgCl₂, 0.1% NP-40) supplemented with 50 μg/ml RNase A and 10 μg/ml PI. They were incubated at room temperature in the dark for 15 min and analyzed by flow cytometry on a NovoCyte flow cytometer (Agilent). Collected data were analyzed by using NovoExpress software (Agilent).

Quantitative reverse transcription–PCR (RT–qPCR) analyses

Total RNA was extracted using the RNeasy Mini kit and submitted to DNase treatment (Qiagen), following the manufacturer's instructions. RNA concentration and purity were evaluated by spectrophotometry (NanoDrop 2000c, ThermoFisher Scientific), and 500 ng of RNA were reverse transcribed with both oligo(dT) and random primers, using the PrimeScript RT Reagent Kit (Perfect Real Time, Takara Bio Inc.) in a 10 μl reaction. Real-time PCRs were performed in duplicate using Takyon ROX SYBR MasterMix blue dTTP (Eurogentec) on an Applied Biosystems QuantStudio 5 (Thermo Fisher Scientific). Transcripts were quantified using the following program: 3 min at 95°C followed by 35 cycles of 15 s at 95°C, 20 s at 60°C and 20 s at 72°C. Values for each transcript were normalized to expression levels of RPL13A (60S ribosomal protein L13a), using the 2^−ΔΔCt method. Primer sequences were as follows: RPL13A forward, AACAGCTCATGAGGCTACGG; RPL13A reverse, TGGGTCTTGAGGACCTCTGT; FTSJ3 forward, CATCCGGGGTCACCAGTTAT; FTSJ3 reverse, TCACCGTCGTCCTCAACATC.

Results

2′-O-Methylation inhibits reverse transcription by HIV reverse transcriptase

Endogenous reverse transcription of HIV-1 full-length genome is inhibited by 2′-O-methylation

Among retroviral polymerases, AMV reverse transcriptase is commonly used to map 2′-O-methylation sites in RNA templates (31). It has been reported that at low dNTP concentrations, AMV reverse transcriptase pauses DNA elongation at the −1 site relative to the 2′-O-methyl group. As HIV genomic RNA was previously proposed to carry internal 2′-O-methylation sites added by the cellular methyltransferase FTSJ3 (27,39), we addressed the possibility that these RNA modifications could impair the reverse transcription of the viral genome. After verification of the significantly lower level of FTSJ3 expression in FTSJ3 KO compared with WT HEK293T cells [Figure 1A, also described in (27,28)], purified HIV-1 virions, produced in either cell type, were permeabilized and submitted to ERT assays. The efficiency of HIV-1 genome reverse transcription was then assessed by qPCR (Figure 1B). Virions from FTSJ3 KO cells yielded significantly more cDNA copies than the viral particles from WT cells, thus suggesting an inhibition of ERT caused by internal 2′-O-methylation.

Figure 1.

2′-O-Methylated HIV-1 genomes are less prone to replication in cells. (A) Total RNA was extracted from HEK293T WT or FSTJ3 KO cells and the relative expression of FTSJ3 transcripts was quantified by RT–qPCR. Data represent the mean ± standard deviation (SD) of duplicates. (B) WT HIV-1 particles produced in HEK293T WT or FSTJ3 KO cells were submitted to ERT for 5 or 10 h, as indicated. The amount of reverse transcriptase products was estimated by qPCR. The cDNA copy numbers per ng of virus are shown (n = 12 from four independent experiments), with the bars showing the mean ± SD. **P < 0.01, ****P < 0.0001, as determined by Student's t-test. (C) HEK293T cells treated or not with 1 mM hydroxyurea (HU) for 24 h, before the cell cycle profiles were assessed by PI labeling. (D) HEK293T cells treated or not with HU were transduced with VSV-G pseudotyped HIV-1 particles produced in WT or FSTJ3 KO HEK293T cells, as indicated. At 6 h post-infection, total cellular DNA was extracted, and HIV-1 reverse transcriptase products were quantified by qPCR. Data represent the mean ± SD of three independent experiments performed in duplicate. ****P < 0.0001, as determined by Student's t-test. (E) Jurkat cells treated or not with 1 mM HU for 24 h, before the cell cycle profiles were assessed by PI labeling. (F) Jurkat cells treated or not with HU were transduced with HIV-1 particles produced in WT or FSTJ3 KO HEK293T cells, as indicated. At 6 h post-infection, total cellular RNA was extracted, and HIV-1 reverse transcriptase products were quantified by qPCR. Data represent the mean ± SD of three independent experiments performed in duplicate. ****P < 0.0001, as determined by Student's t-test.

2′-O-Methylation causes reverse transcriptase to pause at low dNTP concentration

To investigate the impact of 2′-O-methylation on HIV-1 reverse transcription in target cells and to evaluate the influence of dNTP concentration, we treated HEK293T cells with hydroxyurea to deplete the intracellular dNTP pool (Figure 1C) prior to their transduction with VSV-G pseudotyped HIV-1 particles harvested from WT or FTSJ3 KO HEK293T cells. Subsequent qPCR of newly reverse-transcribed HIV genomes shows a lower amount of reverse transcriptase products in both cell types after treatment by hydroxyurea, and the effect is exacerbated for virions harboring 2′-O-methylated genomes produced in WT cells (Figure 1D). The inhibition of reverse transcription on methylated genomes at low dNTP concentration is confirmed in Jurkat cells, a human CD4+-T-cell line, upon infection with WT HIV-1 (Figure 1E, F). These data suggest that 2′-O-methylation of the HIV-1 genome hinders the viral life cycle at the step of reverse transcription.

To unravel the mechanism underlying how RNA methylation could interfere with reverse transcription, we set up an experimental system to determine whether HIV-1 reverse transcriptase is able to bypass 2′-O-methylation. WT HIV-1 reverse transcriptase was produced and purified (Supplementary Figure S2) and tested for its ability to extend a 5′-Cy5-labeled 10-mer DNA primer annealed to a 40-mer RNA template carrying a single N_m site at position 27 (RNA-A_m27) (Figure 2A). The elongation profile of RNA-A_m27 at low dNTP concentrations (0.5 μM) shows a well-defined 26-mer intermediate band product that is missing when we used the non-methylated template RNA control (Figure 2B). We conclude that WT reverse transcriptase pauses at the −1 site relative to the 2′-O-methyl group, similarly to what has been observed for AMV polymerase (31). We further quantified three bands corresponding to the initial primer (10-mer), the intermediate reverse transcriptase pause product (26-mer) and the full-length 40-mer in a time course assay (Figure 2C, D). We observe that elongation on the unmethylated RNA template efficiently proceeds to the full-length 40-mer product with no major intermediate species. In comparison, reverse transcriptase extension on the RNA-A_m27 template shows a 26-mer intermediate band product (>30% after 5 min), and a limited amount of the full-length product. Nevertheless, full-length products appear upon longer incubation periods and reach >30% at the end of the reaction (Figure 2D). A range of dNTP concentrations (0.5–50 μM) was next tested in additional primer extension assays with the RNA-A_m27 template. Data in Figure 3A and B show that the pause induced by the 2′-O-methylated nucleotide is overcome in a dNTP concentration-dependent manner, suggesting that reverse transcriptase impairment due to N_m is not an issue in activated T cells. Indeed, the latter are known to contain higher dNTP levels (∼6 μM) than those found in quiescent cells such as macrophages (∼0.03 μM) (40). To further understand the N_m bypass requirements, we conducted experiments with a low concentration of all four dNTPs (1 μM) to which we added an excess of each dNTP individually (up to 20 μM total). We observe that the percentage of the 26-mer intermediate band (%Pause) decreases only in the presence of a high dTTP concentration (Figure 3C). Hence, the availability of the cognate dNTP to be incorporated opposite an N_m site seems to be critical to reduce the pause during reverse transcription.

Figure 2.

Primer extension assays on a 2′-O-methylated RNA template reveal a pause during reverse transcription at low dNTP concentration. (A) Sequences of the two DNA/RNA primer/template (P/T) duplexes used for primer extension assays. The DNA primers are both 10-mer and Cy5-labeled. The RNA templates have identical sequences, with RNA-A_m27 being 2′-O-methylated on the 27th nucleotide defined from the 3′ end. (B) Primer extension assays were carried out on both templates shown in (A) with [RT_WT] = 0.08 μM, [P/T] = 0.02 μM, [dNTP] = 0.5 μM. Reverse transcriptase (RT) is incubated for 5 min at 37°C with either P/T duplex. The reaction starts by the addition of dNTP, and is quenched at several time points (0, 0.5, 1, 5, 10, 20 and 50 min, as indicated at the bottom of the gel) in a 10 mM formamide–EDTA buffer. Samples are loaded on a denaturing PAGE gel (20% acrylamide, 7 M urea) and detected by a fluorescence imager (Amersham Typhoon). A highly contrasted lane along with the elongated DNA sequence, corresponding to the primer extension assay on RNA-A_m27 after 50 min, juxtaposes the gels. (C and D) Quantification of initial DNA (10-mer), intermediate DNA products (26-mer) and full-length DNA (40-mer) resulting from primer extension assays for RNA control and RNA-A_m27, respectively. Each data point represents the mean ± SD of three independent experiments.

Figure 3.

Reverse transcriptase (RT) pause depends on the dNTP concentration and the 2′-O-methylated nucleic acid base. (A–C) Primer extension assays were carried out on primer/template RNA-A_m27 shown in Figure 2A. (A) Primer extension assays were performed in the following conditions: [RT_WT] = 0.8 μM, [P/T] = 0.2 μM, four dNTP concentrations as indicated (0.5, 1, 5 and 50 μM), and quenched at different times (0, 0.5, 1, 5, 10, 20 and 50 min). The samples were separated on a denaturing PAGE gel (20% acrylamide, 7 M urea), and visualized using a fluorescence imager (Amersham Typhoon). (B) Quantification of the 26-mer intermediate band of the primer extension assays shown in (A). Each data point represents the mean ± SD of three independent experiments. (C) Quantification of the 26-mer intermediate band of primer extension assays performed with [RT_WT] = 0.08 μM, [P/T] = 0.02 μM and [dNTP] = 1 μM, plus an excess concentration of another dNTP as indicated, resulting in a 20 μM concentration of the respective dNTP. Each data point represents the mean ± SD of three independent experiments. (D) Quantification of the 26-mer intermediate band of primer extension assays performed with [RT_WT] = 0.08 μM, [P/T] = 0.02 μM and [dNTP] = 1 μM on the previously used DNA primer (A) annealed to four different RNA templates (see Supplementary Table S2) where the 27th position site is either a G_m, A_m, U_m or C_m. Each data point is the mean ± SD of three independent experiments.

Reverse transcriptase shows differential sensitivity to template 2′-O-methylation depending on the nucleobase

For the HIV-1 RNA genome, it has been previously described that among the 17 high-confidence N_m sites, 15 involve adenosines and 2 involve uridines (27). This prompted us to address if the type of 2′-O-methylated nucleotide could affect the efficiency of reverse transcription on viral 2′-O-methylated templates. We designed templates similar to RNA-A_m27 except we changed the N_m nucleic base at position 27 to generate RNA-G_m27, RNA-U_m27 and RNA-C_m27 templates. We also changed a few nucleotides to minimize potential RNA secondary structures (Supplementary Table S2). Comparing reverse transcriptase elongation on these templates in the presence of 1 μM dNTPs shows a very strong pause for a G_m, a 3-fold weaker one for A_m, an even weaker pause for U_m and finally no detectable pause for C_m (13, 5, 1 and 0.2% respectively, at 10 min, Figure 3D). These data illustrate that 2′-O-methylated purines impair elongation more efficiently than 2′-O-methylated pyrimidines.

2′-O-Methylation does not impact reverse transcriptase fidelity

It was previously shown that 2′-O-methylation was found on highly conserved sites among major HIV-1 subtypes, with conservation frequencies >90% for 14/17 sites and no consensus of the nucleotide sequence in the 40 nucleotide vicinity surrounding the methylated sites (27). The error rate of HIV-1 polymerase is ∼0.6–2 × 10⁻⁴ regardless of the type of template (RNA or DNA) (41), and we next explored whether a 2′-O-methylated site could favor mutagenesis during the synthesis of the cDNA strand by the reverse transcriptase. Primer extension assays were performed with RNA templates containing nucleotides 734–799 of the HIV-1 genome (pNL4-3 strain) and including ± A_m versions of the 783A site that is known to be 2′-O-methylated (27) (here at position 67 of the template, Supplementary Figure S1A), in the presence of different dNTP concentrations (4, 40 and 250 μM). The single-stranded DNA products (Supplementary Figure S1B) served as templates for PCR amplification, and the resulting double-stranded DNA was sequenced by Ion Torrent next-generation sequencing (NGS). Any mutation found in >0.5% of the reads is reported in Supplementary Table S3. Compared with the reference sequence of the native RNA template, sequencing yielded only a few minor variants (0.5–4%), affecting both templates (783A and 783A_m, 40 or 250 μM dNTP), that do not affect A at position 67, or nucleotides in the close vicinity. Consequently, an A_m site does not lead to any hotspot of mutation, an observation that is consistent with the high conservation of the sequence in the vicinity of the HIV-1 methylation site (27).

2′-O-Methylation impedes the chemistry step of nucleotide incorporation and translocation

The reverse transcriptase catalytic cycle encompasses five major reaction steps: (i) binding of the primer–template duplex; (ii) initial binding of the dNTP; (iii) a rate-limiting transition to a closed and catalytically competent conformation; (iv) the chemical step of dNTP incorporation; and (v) PPi release followed by translocation of the newly incorporated base from the dNTP-binding site (N-site) to the primer site (P-site) [reviewed in (42)]. We suspected that since reverse transcriptase forms hydrogen bonds with several RNA 2′-OH ribose moieties (43), methyl groups could interfere with critical interactions during the catalytic cycle, leading to the observed pause. In order to identify the N_m-limited step during the catalytic cycle, we examined binding, elongation and translocation on methylated and unmethylated templates.

2′-O-Methylation does not affect reverse transcriptase:RNA/DNA binding

We used FP to determine whether template binding (step i) is impaired for a duplex of DNA/2′-O-methylated-RNA. As FP experiments require a fluorescent partner with a relatively low molecular weight compared with the binding protein, we designed a smaller nucleic acid component with a sequence similar to the inner part of the previous RNA-A_m27 template. It is composed of a 24-mer RNA template, with or without A_m18 (named midRNA-control and midRNA-A_m27, respectively) and a 17-mer DNA primer matching the first 17 nt of the RNA template. The annealed DNA–RNA hybrid mimics the intermediate band position where elongation stopped at the −1 site regarding the N_m position, with a 5′-overhang end of the RNA template (Figure 4A), and the complex is sufficient for a tight binding of the reverse transcriptase that usually spans 17–18 nt between the polymerase and RNase H sites. We first performed active site titration of the reverse transcriptase enzyme by manual rapid quench flow and we determined that ∼100% of our enzyme carries reverse transcriptase activity (Supplementary Figure S3). Both methylated and non-methylated RNA templates show a similar K_D (16 ± 3 nM and 13 ± 3 nM, respectively, Figure 4B) that are in the same nanomolar range as what was previously described for poly(rA)/oligo(dT) with other methods (44,45). Overall, we demonstrate that reverse transcriptase interacts similarly with both RNA templates, leading us to conclude that template binding is not the limiting step of elongation that could explain the pause observed in the context of a single N_m, in agreement with a previous study showing that binding of reverse transcriptase does not rely on RNA 2′-hydroxyl groups (46).

Figure 4.

The chemistry step opposite a 2′-O-methylated A site represents a limiting step in reverse transcription. (A) Two duplexes of DNA/RNA primer/templates (P/T) used for FP and primer extension assays. DNA primers are 17-mer Cy5-labeled oligonucleotides. RNA templates have the same sequence, except that midRNA-A_m27 is 2′-O-methylated on the 18th nucleotide, as defined from the 3′ end. (B) Fluorescence polarization with 0.0125 μM of both P/T pairs and a 0.00061–10 μM range of reverse transcriptase concentrations. The data were fitted with a quadratic form binding equation curve, yielding similar K_d values of 16 ± 3 nM for the midRNA-A_m27 template and 13 ± 3 nM for the midRNA-control template. Each data point represents the mean ± SD of three independent experiments. (C) The K_m of dTTP for the P/T·RT complex was assessed by performing primer extension assay, mixing the reverse transcriptase, either of the P/T duplexes shown in (A) and a range of dTTP concentrations (0.01, 0.1, 0.2, 0.5, 1, 5, 10, 15 and 20 μM), prior to stopping the reaction at different times (0, 0.5, 1, 2, 5, 10 and 20 min) in 10 mM formamide–EDTA buffer. The samples are analyzed using denaturing PAGE (20% acrylamide, 7 M urea), and scanned using a fluorescence imager (Amersham Typhoon). The elongated DNA products were quantified and plotted against time. The slope coefficient allowed determination of the initial speed (V_i, pmol/min) by linear regression in the linear fraction of the curve. The V_i values were plotted against the dTTP concentration prior to a Michaelis–Menten fit, allowing the determination of K_m and V_max. This figure shows the typical results of one replicate, but this experiment was done as a duplicate, yielding similar results (see Table 1 for the data of the duplicate). (D and E) The K_m (D) or the V_max (E) of dTTP were extrapolated from two independent experiments performed on each template. ‘midRNA-control₁’ and ‘midRNA-Am27₁’ are the K_m or V_max ± standard error (SE) of the fit of replicate 1, and ‘midRNA-control₂’ and ‘midRNA-Am27₂’ are the K_m or V_max ± SE of the fit of replicate 2. Both K_m values or V_max values were also pooled together to obtain the mean ± SEM of the K_m or V_max for the duplicate (referred to as ‘midRNA-control₁₊₂’ and ‘midRNA-Am27₁₊₂’). Data are differentiated by colors, as indicated next to (E).

RNA having 2′-O-methylated adenosines decreases reverse transcriptase affinity for dTTP

The reverse transcriptase pause at a −1 position relative to an A_m modification suggests that the polymerase is slow to incorporate a dNTP opposite the A_m site, causing an accumulation of intermediate products. To probe the kinetics of the chemistry step in this context, we assessed the dTTP concentration dependence of the polymerization in steady-state primer elongation assays (Supplementary Figure S4), obtaining affinity constants (K_m) of 0.09 ± 0.01 and 1.4 ± 0.5 μM for dTTP opposite A and A_m template sites, respectively (WT reverse transcriptase in Figure 4C, D and Table 1). These data thus show an ∼16-fold increase in the dTTP K_m on the methylated versus the non-methylated template, suggesting that the pause at −1 of the A_m site could result from a lower apparent affinity for dTTP. The 0.09 μM observed K_m on the unmethylated template is similar to the 0.1 μM established previously by isothermal titration calorimetry for dTTP binding to reverse transcriptase–DNA/DNA primer–template (47). Regarding V_max values, we observed that they fall within a similar range (∼0.6–0.7 pmol/min), regardless of the template (Figure 4E; Table 1), indicating that in our experimental system the processivity of the reverse transcriptase is not affected by the template methylation.

Table 1.

K _m of dTTP and V_max of reverse transcriptase of WT and mutants

	K m dTTP (μM)		V max (pmol/min)
	midRNA -control	midRN-A m 27	midRNA -control	midRN-A m 27
WT	0.09 ± 0.01	1.4 ± 0.5	0.71 ± 0.03	0.66 ± 0.04
K70E	0.13 ± 0.05	8.9 ± 6.4	0.52 ± 0.10	0.40 ± 0.13
K70A	2.5 ± 0.9	480 ± 200	0.03 ± 0.01	0.04 ± 0.02
L74V	0.14 ± 0.03	1.27 ± 0.02	0.66 ± 0.04	0.5 ± 0.2
K82A	0.4 ± 0.2	6.5 ± 0.7	0.3 ± 0.10	0.23 ± 0.16

The table shows the mean ± SEM of two independent experiments.

2′-O-Methylated adenosines enhance pyrophosphorolysis of the primer

It has been previously demonstrated that HIV-1 reverse transcriptase can excise the last incorporated nucleotide of the DNA strand during elongation through phosphorolysis mediated by PPi or ATP, which is an enhanced feature for some reverse transcriptase mutants exerting resistance against nucleosidic reverse transcriptase inhibitors (NRTIs) (48). Excision occurs when the last incorporated dNTP is found in the N-site, rather than the P-site where it sits after the translocation step. It can then interact with a reactive phosphate group present in the active site and be removed by a reversal of the canonical nucleotide addition reaction. Thus, if the translocation step does not occur or, alternatively, if there is backtracking of the enzyme on the nucleic acid duplex, then the 3′-end dNTP stalls in the N-site and the probability of excision should increase. We used this reverse transcriptase feature to test whether N_m could impair the translocation step. WT reverse transcriptase was incubated with midRNA-control or midRNA-A_m27 annealed to the cognate DNA 17-mer primer (Figure 4A) in addition to a range of ATP or PPi concentrations, and in the absence of dNTPs. The phosphorolysis reaction was efficient with both ATP (Supplementary Figure S5) and PPi (Figure 5). However, in our experimental conditions, the primer excision is observed upon a long incubation period (several hours) with ATP, while it happened faster with PPi (Figure 5). Therefore, we further studied phosphorolysis mediated by PPi. The DNA primer is degraded with both templates, but excision of the first base to yield the −1 product is more efficient with midRNA-A_m27 than with midRNA-control (Supplementary Figure S6; Figure 5). Production of the −1 band shows burst kinetics with both RNA templates, indicating that a portion of the complexes in solution have their P/T duplex poised for rapid and immediate excision, while the remaining complexes are degraded more slowly (Figure 5A, B). We cannot determine the rates of the initial burst phase from these manual quench data, but their amplitudes could be reliably determined. Numerical curve fitting of the PPi concentration dependence of the burst amplitudes indicates that the concentrations of rapidly excised primer–template complex in the reactions increased from 67 ± 6 nM for the midRNA-control template to 97 ± 8 nM for the midRNA-A_m27 template, but the two templates have equivalent PPi K_d,app values of 5.4 ± 2 μM (Figure 5C). The subsequent linear phases reflecting excision of the remaining material are also ∼3-fold faster for the methylated template than for the unmodified template (Figure 5A, B).

Figure 5.

2′-O-Methylation promotes excision of the DNA primer. (A and B) Time courses of the build-up of excision products at 0, 1, 5, 10, 20, 50 and 100 μM PPi (light to dark blue or red) for the control (A) and A_m27 (B) template RNA strands. The data show burst kinetics that suggest that a portion of the P/T pre-exists in the backtracked state and is poised for rapid excision, while the remainder of the P/T is excised on a slower time scale limited by rebinding or backtracking rates. The A_m27 RNA results in a larger burst amplitude and an ∼3-fold faster continued excision rate that are both indicative of more efficient backtracking with the 2′-O-methylated RNA base. (C) Concentration dependence curves show different maximum burst amplitudes of 67 ± 6 nM for control RNA (blue) and 97 ± 8 nM for A_m27 RNA (red), but no change in the midpoint PPi concentrations which is 5.4 ± 2 μM.

The burst phase kinetics indicate that a significant portion of the primer–template is pre-positioned for rapid excision, while the remaining material is more slowly excised in a steady-state phase that is presumably rate limited by P/T release and rebinding. It has been shown that the P/T pair is bound to HIV reverse transcriptase in a dynamic equilibrium, with the 3′ end of the DNA primer being found in either the P-site as expected after translocation or in the N-site due to lack of translocation or backtracking (49). In the context of such a positional equilibrium, the greater burst phase amplitude observed with A_m27 RNA indicates that template strand methylation alters the equilibrium to increase the amount of backtracked material whose 3′ end is in the N-site and susceptible to rapid PPi-mediated excision. Considering the 0.8 μM reverse transcriptase and 0.2 μM P/T concentrations used in our experiments and their 13–16 nM affinities (Figure 4B), the binding equilibrium is 97.5% saturated and 195 ± 2 nM reverse transcriptase·P/T complex is formed. If we then consider the 67 and 97 nM maximal amplitudes of the burst phases as representing the concentration of complexes present in the backtracked state, we can estimate the equilibrium constant (K_eq) for the partitioning of the P/T complex between the two states. This calculation indicates that the native RNA template has a K_eq value of 1.9 ± 0.2 in favor of the P-state that is reduced to 1.0 ± 0.1 with the 2′-O-methylated template. Thus, both RNAs favor the P-state binding register for the P/T duplex, but the partitioning is reduced by 1.9 ± 0.3-fold in the presence of a 2′-O-methyl-modified ribose at the templating base.

These partitioning values are consistent with footprinting studies showing that modified 3′-end nucleotides can similarly alter the P-site to N-site occupancy ratio (49), but differ mechanistically in that the equilibrium is now being perturbed by interactions with the template RNA strand instead of a modified 3′-terminal nucleotide bound into the active site. To further explain this at a molecular interaction level, we examined the 2.5 Å resolution structure of HIV reverse transcriptase with RNA/DNA and a dATP substrate (PDB: 4pqu) (42,50). Several residues grip the RNA template by interacting mainly with the RNA backbone, notably V75, R78, N81, G152, W153 and K154 in the fingers domain and E89, Q91, G93 and P157 in the palm domain (43), as diagrammed in the NUCPLOT (51) analysis shown in Figure 6A. Residues E89, Q91, D76 and G152 interact directly with RNA 2′-OH moieties through hydrogen bonding. We then modeled 2′-O-methyl groups onto the RNA ribose groups of the templating ‘+1’ nucleotide (U705) and the downstream ‘+2’ nucleotide (C706) to approximate the locations of the N_m modification in either a templating or a backtracked P/T binding register (Figure 6B). Considering that we observe the elongation pause one nucleotide before the N_m position, having the N_m ribose in the +1 site means that the DNA primer is in the non-excisable P-site register (i.e. the +1/P-site), but having the N_m base in the +2 site means the P/T is backtracked and the DNA primer is in the excisable N-site register (i.e. that +2/N-site). Inspection of the structure shows that an O-methyl group in the +1/P-site register has highly unfavorable interactions; the carbonyl oxygen of Gly152 forms a hydrogen bond with the 2′-OH group of the +1 base, and adding a 2′-O-methyl group creates a significant steric clash with a very close 1.4 Å carbon–oxygen distance (Figure 6C). There is also a generally unfavorable interaction with a polar pocket formed by the backbone of residues 75–77 (Figure 6C). These interactions will clearly alter the positioning of the +1 ribose and in turn will alter the positioning of the templating base during catalysis, providing a structural explanation for the higher K_m observed for dTTP incorporation. In contrast, placing the N_m into the +2/N-site puts the 2′-O-methyl group in a significantly more favorable environment where the entire nucleotide is on the reverse transcriptase surface and positionally unconstrained, giving it the flexibility to pack the methyl up against the surface-exposed hydrophobic Phe61 side chain and away from the negatively charged Asp76 residue. Thus, from a structural interactions perspective, we predict that it is energetically more favorable to have the 2′-O-methyl group in the +2 site than in the +1 site, and this will drive the P/T binding equilibrium toward the backtracked register with an N_m template, increasing the amount of DNA primer in the N-site where it is poised for excision.

Figure 6.

Model of the structural features of a 2′-O-methylation addition on a DNA/RNA–reverse transcriptase competent complex. (A) A NUCPLOT analysis was performed on PDB 4pqu, a structure of a reverse transcriptase bound to a DNA/RNA primer/template duplex, with an incoming dATP in the polymerase active site. This analysis shows the interactions between the p66 subunit (residue number and atom name, chain A) and nucleic acids (chains T and P for RNA and DNA, respectively). Interactions with water molecules are not shown, −1 elongated site, +1 and +2 sites for elongation are identified. (B–D) Structure of the HIV reverse transcriptase–P/T complex (PDB 4pqu) with 2′-O-methyl modeled in the +1 and +2 template strand positions. Reverse transcriptase (gray), DNA (tan), RNA (yellow), incoming dATP (green), 2′-O-methyl (magenta, half-sized spheres). (B) Overview of the active site with +1 position (N-site) base shown in cyan and base-paired to a dNTP (green) bound in the active site. 2′-O-Methyl was modeled onto the RNA at the +1 and +2 position based on superpositioning of ribose groups from PDB 310d. (C) Detailed view showing that a 2′-O-methyl in the +1 position (cyan) will be placed in a polar environment with a significant steric clash with G152 carbonyl and unfavorable backbone interactions with V75–F77. In contrast, a 2′-O-methyl in the +2 position (dark cyan) will be in a more hydrophobic environment involving C_β of D76 and the aromatic rings of F61 and W24. (D) Amino acids targeted for mutagenesis are shown in cyan and colored by atom in the vicinity of the reverse transcriptase active site.

Identification of critical amino acids for reverse transcriptase sensitivity to A_m sites

To better understand the underlying molecular mechanisms of reverse transcriptase sensitivity to the presence of N_m sites, we looked for specific amino acid substitutions that could alter the efficiency of the reverse transcriptase bypass of N_m sites. We established three categories of potentially relevant reverse transcriptase mutants. The first category encompasses mutations related to NRTI resistance and includes K65R, K70E, K70R, L74V, V75T and M184V that are generally found in the polymerase active site, opposite the dNTP-binding site (except for M184V), and close to the RNA template. These NRTI resistance mutations are known to yield active enzymes and confer different capacities compared with the WT regarding substrate discrimination [M184V (52), K70E(53)], excision [K70R (54)] or accommodation, template alignment and processivity [reviewed in (55–57)]. The second category of mutations is based on a sequence alignment of the HIV-1 and AMV catalytic subunits that showed several positions close to the template where residue charges and sizes differ greatly (Supplementary Figure S7). As AMV reverse transcriptase is known to pause at N_m sites, we wanted to determine if some residues that are not found in HIV-1 reverse transcriptase were helping the AMV ‘pause phenotype’. We highlighted three interesting differences in the fingers domain (K70S) and in the α-helix in close vicinity to the 5′ extremity of the template (K82A and E79A). The third category clusters mutations in the vicinity of the −1 site that are also involved in the template grip, and we designed D76A, D76V, D76S, N81A, N81V and N81T mutations to assess the effect of residues that are less bulky and less prone to interact with the template. The mutated sites are illustrated in Figure 6D.

The mutants from the third category were not well expressed in bacteria (we were unable to purify any reverse transcriptase N81V) and were weakly active, and therefore deemed unsuitable for further study. We note that D76 and N81 are highly conserved (>99%) (Supplementary Figure S8), suggesting that these residues are critical for either reverse transcriptase activity or folding. In contrast, the other mutants were produced and purified (Supplementary Figure S2), and all the mutants in categories 1 and 2 were found to be active in preliminary primer extension assays. Among them, we were looking for mutants showing a critically different ‘bypass’ or ‘stop’ behavior compared with the WT reverse transcriptase on a methylated template. Our preliminary tests identified K70 mutants (K70E, K70R and K70S) in the fingers domain. For these mutants, the pause seems considerably stronger during primer elongation assays on a methylated template. The role of K70E reverse transcriptase was characterized in a previous study (53) and, as this mutant was expressed and purified well, we tested it in primer extension assays to measure the K_m and V_max by addition of dTTP to incorporate opposite a native adenosine or a 2′-O-methyladenosine. The K70E strong ‘pause’ phenotype is confirmed, as the K_m of dTTP is ∼70-fold higher with the midRNA-A_m27 template than with midRNA-control (Table 1; Figure 4D). To further test the role of residue 70, it was mutated to an alanine (K70A), and this was less active with a lower V_max than the WT, and a 200-fold higher K_m for dTTP on the methylated template compared with the control templae. L74 is known to participate in template positioning through its interaction with the templating base (58), but the K_m and V_max values of the L74V mutant are not significantly different from the WT reverse transcriptase (Table 1; Figure 4D). Finally, the K82A mutant, found within a helix close to the modified +1 position (Figure 6D), appeared slightly more able to bypass methylated sites in early tests. Eventually, thorough primer extension assays allowed us to determine that K82A shows a phenotype that is comparable with the WT enzyme (Table 1; Figure 4D, E). Overall, V_max values are rather alike for both templates, independent of the mutation (Table 1; Figure 4E), meaning that N_m sites do not greatly affect the processive elongation by the enzyme.

To confirm our biochemical observations using the purified reverse transcriptase, we next conducted ERT experiments on WT or hypomethylated HIV-1 particles (produced respectively in WT or FTSJ3 KO cells), containing the K70E reverse transcriptase mutant, previously identified as a potential ‘pause phenotype’. The ERT assays show that the efficacy of reverse transcription of the reverse transcriptase K70E mutant is dramatically decreased for WT virions compared with hypomethylated viruses (Figure 7A). This is particularly striking when comparing the ratio of reverse transcriptase product relative expression from FTSJ3 KO cells over WT cells (Figure 7B), where the K70E mutant seems again to be poorly able to reverse transcribe methylated templates, and is significantly less efficient than the WT. Consequently, these results confirm that the K70 residue plays a critical role in the bypass of 2′-O-methylation. In contrast, mutation L74V did not significantly change the ratio of reverse transcriptase product compared with the WT (Figure 7B), while mutation K82A yielded less efficient replication, but equivalent reverse transcriptase product relative expression regardless of FTSJ3 expression (Figure 7A, B).

Figure 7.

Endogenous reverse transcription in cell-free particles of mutant HIV-1 containing WT or hypomethylated genomes. (A and B) HIV-1 WT or mutant reverse transcriptases were produced in HEK293T WT or FSTJ3 KO cells and submitted to ERT for 10 h. The amount of reverse transcriptase products was estimated by qPCR. Results are represented as mean ± SD and shown as raw values (A) or as ratios of reverse transcriptase products in viruses produced in HEK293T KO versus the WT (B). **P < 0.01 and ****P < 0.0001 as determined by Student's t-tests.

Discussion

In this study, we shed light on the extent to which 2′-O-methylation affects HIV-1 genome replication. We used WT or FTSJ3 KO cells to produce WT and hypomethylated HIV-1 virions, respectively, and performed ERT assays and infection of cells with a depleted pool of dNTP. The limited amount of reverse-transcribed RNA coming from WT virions in dNTP-depleted cells suggested the impairment of reverse transcription caused by 2′-O-methylation in a dNTP concentration-dependent manner. This was further confirmed by biochemical assays, emphasizing the pause of the reverse transcriptase at low dNTP concentration when encountering a G_m, A_m or U_m, but not a C_m, site. Consequently, we evidence an antiviral role for 2′-O-methylation in the context of an HIV-1 infection.

The pause during reverse transcriptase elongation could result from altered interactions with the 2′-O-methylated RNA template, effects on dNTP binding or incorporation rates, or the post-catalysis translocation step, but not with primer/template binding as its affinity was unaffected by A_m. 2′-O-Methylation was observed to increase DNA excision, and we propose that the polymerase is slow to incorporate dTTP and backtracks when encountering an A_m site. To support this, we analyzed the consequence of RNA 2′-O-methylation by modeling a methyl group on the +1 and +2 nucleotides of the RNA template in the structure of the reverse transcriptase–DNA/RNA complex. This showed substantial steric clashes with the methyl group in the +1 site, but a much more favorable binding environment in the +2 site, consistent with the biochemical data indicating a 1.9 ± 0.3-fold shift of the primer/template binding register equilibrium toward the backtracked state. We thus hypothesize that the 2′-O-methylation causes a repositioning of the template, potentially reminiscent of that found in the reverse transcriptase–RNA/RNA complex (59) where initiation of reverse transcription is slow and shows low processivity (60).

Nevertheless, HIV-1 reverse transcriptase efficiently bypasses 2′-O-methylation when the dNTP concentration is high. Interestingly, retroviruses such as AMV and murine leukemia virus (MLV) replicate well in proliferative and cancerous cells that have high concentrations of dNTP required for cell division. Hence, although MLV viral RNA was found to be modified by epitranscriptomic marks such as 2′-O-methylation (61), the fact that the virus replicates mostly in actively dividing cells would limit the pause caused by 2′-O-methylation during reverse transcription. However, HIV-1 belongs to the lentivirus family known to infect cells with low dNTP concentrations, such as quiescent LT CD4+ cells or cells from the monocyte/macrophage lineage [∼6 μM and 0.03 μM dNTP, respectively (40)]. Lentiviruses have thus possibly evolved to minimize the effect of 2′-O-methylation on reverse transcription in an environment depleted of dNTP. Notably, we showed that the reverse transcriptase K70 residue is critical to favor a bypassing phenotype, while AMV reverse transcriptase has an S70 (Supplementary Figure S7) that could impede the bypass. Whether other viral proteins such as the nucleocapsid (p7) help with bypassing 2′-O-methylation was addressed by ERT assays, demonstrating that other viral proteins do not seem to rescue the sensitivity of the reverse transcriptase regarding 2′-O-methylated sites.

From a structural point of view, K70 is located in the flexible β3−β4 loop of the fingers subdomain, the residues of which become part of the binding pocket upon dNTP binding. In this context, the side chain of K70 does not show direct interactions with either the substrate (dNTP), the P/T or other amino acids. Previous studies suggest that residues 65−73 within the fingers contribute to dNTP binding (62) and K70E could be close enough to K65 to disturb its location and possibly destabilize the flexible loop structure, impairing dNTP binding by disrupting how K65 interacts with the γ-phosphate of incoming dNTP (Figure 6D) (53). Hence, we hypothesize that RNA 2′-O-methylation causes a template mispositioning and the dNTP addition steps becomes even less efficient, and the side effects of a K70E mutation on the chemistry step are even more deleterious, increasing the strength of the pausing phenotype. An alternative perspective, which complements the previous interpretation, suggests that Lys70 is located within the cleft leading to the dNTP-binding site and may facilitate the attraction of negatively charged dNTPs. This interaction could lead to the formation of an elongation complex, enabling proper dNTP binding. Some interactions of K70 are indeed possibly critical during the catalytic cycle, explaining why the K70A mutation, which should not displace K65R, drastically limits the methylation bypass. In addition, the fingers bending closer to the catalytic site is the conformational adjustment ensuring a closed and competent complex and is considered as the rate-limiting step during elongation. Thus, the mutant K70E being less able to replicate when the HIV-1 genome (or a synthetic template) is methylated highlights the key role of the fingers and notably K70 in ensuring efficient nucleotide incorporation opposite a 2′-O-methylated site. In contrast, the L74 residues interacts with the DNA templating base through a van der Waals contact (58), but it does not differentially impact elongation compared with the WT reverse transcriptase, in vitro or in cells, if replaced by a valine. Finally, K82A, a mutation found in residues near the primer grip, is found to negatively impact replication in cells, regardless of the presence of FTSJ3-mediated methyl addition, while the purified enzyme shows a phenotype relatively comparable with the WT reverse transcriptase upon in vitro primer extension assays.

Eventually, 2′-O-methylation has been described as a modification allowing cells to discriminate self- from non-self-RNA with immune sensors such as RIG-I and MDA5 (15,27,63). 2′-O-Methylation also counteracts the activity of the restriction factor ISG20 (28). However, 2′-O-methylation seems to have an antiviral role against HIV-1 at low dNTP concentration. Knowing that they impede reverse transcription, and that cellular 2′-O-MTases such as the cap 2′-O-MTase CMTR1 are suspected to be interferon induced (64,65), the potential role of 2′-O-MTases as restriction factors would be worth further investigation. Moreover, whether 2′-O-methylation negatively affects translation of viral RNA, as is the case for mRNA (11,12), or other protein interactions such as with the viral capsid, remains to be determined. Consequently, we suspect that the process of 2′-O-methyl addition to be finely regulated so that viruses succeed at replicating despite potential antiviral effects of RNA modifications. Hence, identification of the targeted RNAs (genomic, unspliced or mature mRNA) and where and when 2′-O-methylation occurs are interesting issues. Ultimately, a better understanding of the trade-off between pro- and antiviral effects of epitranscriptomic modifications on pathogenic viruses could pave the way for the discovery of new antiviral strategies.

Data availability

Sequencing results are available under the BioProject accession no. PRJNA936659 (bio samples SAMN33360799, SAMN33360800, SAMN33360801, SAMN33360802, SAMN33360803, SAMN33360804, SAMN37640539, SAMN37640540, SAMN37640541, SAMN37640542, SAMN37640543 and SAMN37605388).

Supplementary data

Supplementary Data are available at NAR online.

Funding

The Agence Nationale de la Recherche sur le SIDA et les Hépatites virales [ECT280C/U160]; and the Fondation pour la Recherche Médicale [FDT202204014965 grant awarded to A.D.].

Conflict of interest statement. None declared.