HIV-1 uses dynamic capsid pores to import nucleotides and fuel encapsidated DNA synthesis

Abstract

During the early stages of infection, the HIV-1 capsid protects viral components from cytosolic sensors and nucleases such as cGAS and TREX, respectively, while allowing access to nucleotides for efficient reverse transcription. Here we show that each capsid hexamer has a size-selective pore bound by a ring of six arginine residues and a 'molecular iris' formed by the amino-terminal β-hairpin. The arginine ring creates a strongly positively charged channel that recruits the four nucleotides with on-rates that approach diffusion limits. Progressive removal of pore arginines results in a dose-dependent and concomitant decrease in nucleotide affinity, reverse transcription and infectivity. This positively charged channel is universally conserved in lentiviral capsids despite the fact that it is strongly destabilizing without nucleotides to counteract charge repulsion. We also describe a channel inhibitor, hexacarboxybenzene, which competes for nucleotide binding and efficiently blocks encapsidated reverse transcription, demonstrating the tractability of the pore as a novel drug target.

Extended Data Figure 1. dATP binds to the R18 pore at the centre of the capsid hexamer.

2Fo-Fc density (grey mesh) contoured at 1.0σ about R18 for the unbound (a) and dATP-bound (b) CA_Hexamer structures. Fo-Fc omit density (green mesh) contoured at 3.0σ is shown for the dATP-bound structure. c, dATP lies on the crystallographic 6-fold axis and significant rotationally-averaged density is observed only for the triphosphate group.

Extended Data Figure 2. Controls for dNTP-binding experiment.

a, Titration of CA_Hexamer into 2 nM fluorescein-labelled dTTP in the presence of 1 mM (physiological) or 5 mM inorganic phosphate. Under the 1 mM conditions, there is no significant effect on hexamer binding to dTTP. At 5 mM apparent affinity is decreased to 851 nM, demonstrating that inorganic phosphate can compete for the pore. However, given that the intracellular [dNTP] is approximately 100 μM, under intracellular conditions dNTP binding would dominate. b, Titration of CA_Hexamer into BODIPY-labelled rGTP-γ-S and fluorescein-labelled dTTP. Each binds with the same affinity suggesting that the R18 pore is unable to discriminate between ribose and deoxyribose nucleoside triphosphates. The difference in the magnitude of the fluorescence anisotropy signals is due to differences in fluorophore excited state lifetimes. K_D values are indicated by a dotted line.

Extended Data Figure 3. DSF melt curves.

The left-hand panels report the ratio of tryptophan fluorescence emission at 350nm and 330nm as a function of temperature. The right-hand panels report the first derivative of the same data, the peak of which is used to determine the T_m value. a, b, Effect of dATP and DTT on WT CA_Hexamer. c, d, Effect of dATP and DTT on R18G CA_Hexamer. e, f, Effect of each dNTP on WT CA_Hexamer. g, h, Comparison of the effects of carboxybenzene compounds on WT CA_Hexamer. i, j, Comparison of the effects of hexacarboxybenzene on WT and R18G CA_Hexamer.

Extended Data Figure 4. Alignment of selected retrovirus capsid sequences bordering the electropositive pore.

The position equivalent to R18 in HIV-1 is marked with an arrow.

Extended Data Figure 5. Confirmation of CA_Hexamer Chimera Assemblies.

a, Non-reducing SDS-PAGE of CA_Hexamer WT:R18G chimera samples demonstrates that the recombinant proteins had reassembled into hexamers. Molecular weight standards (kDa) are presented in the first lane. For gel source data, see Supplementary Fig. 1. b, Comparison of 1:5 homohexamer mix and the equivalent chimera. The 1:5 WT:R18G mix experiences a six-fold loss of apparent K_d, as expected for a 6-fold dilution of WT with a non-binding mutant. In contrast, the 1:5 chimera chimera has a 58-fold decrease in K_d, demonstrating that chimeric hexamers had indeed formed.

Extended Data Figure 6. Effects of HIV-1 CA R18G on viral infectivity.

a, R18G is capable of abrogating TRIM5α-mediated restriction. Rhesus TRIM5α provides a potent block to infection of HIV in FRhK-4 cells. Titration of a non-GFP-expressing virus can compete for TRIM5α-binding and relieve the restriction of a GFP-expressing virus only if it delivers an assembled capsid into the cytoplasm. R18G abrogates restriction but W184A/M185A, which is incapable of forming assembled capsids due to loss of the CTD-CTD dimerization interface, does not. b, Binomial distribution model for the relative proportion of capsid hexamers carrying a discrete number of glycines at position 18 at defined bulk ratios of WT:R18G. c, Six models (dotted lines) predicting the effect of replacing arginine 18 with glycines. Each model assumes a different number of glycines is required to render the pore defective. The data from WT:R18G chimeric virus measurements (solid line) is consistent with a model in which four or more arginines (i.e. 2 or fewer glycines, green) are required to maintain a functional pore.

Extended Data Figure 7. ERT assay.

a, HIV-1 cores were prepared by ultracentrifugation through a Triton X-100 layer over a sucrose gradient. Resulting fractions were subjected to ELISA for p24 and fractions 3 – 7 were pooled for further experiments. b, Endogenous RT activity for strong stop in the presence of DNase I using HIV-1 fractions that were prepared with or without the Triton X-100 spin-through layer. Input levels of p24 were normalized between reactions. c, dNTP’s were added to HIV-1 cores prepared by Triton X-100 spin-through in the presence of DNase I. Reactions were stopped at the indicated time point by shifting to -80° C and levels of strong stop were quantified. d, Levels of strong-stop (RU5), first-strand transfer (1ST) and second-strand transfer (2ST) DNA after overnight incubation of HIV-1 cores with or without dNTP’s in the presence of DNase I. e, Levels of naked HIV-1 DNA genomes untreated or incubated overnight with DNase I or Benzonase. f, Effect of carboxybenzene compounds on recombinant reverse transcriptase activity.

Extended Data Figure 8. Comparison of WT and H12Y crystal structures.

The H12Y monomer (in the context of the hexamer, purple) superposes on the WT (green) with RMSD = 0.2471 Å. Residues 4-9 of the H12Y structure have been modeled in two alternate conformations owing to flexibility towards the tip of the hairpin.

Figure 1. HIV-1 capsid hexamers have a pore at the 6-fold symmetry axis.

a, Superposition of N-terminal domains from solved capsid structures. A detailed view of the boxed region shows that the β-hairpin toggles between closed (green) and open (pink) states as a result of the hydrogen-bond network about P1, H12, and D51. b, β-hairpin (coloured) conformations dictate the presence of a pore at the 6-fold axis. Hexamers of CA N-terminal domain (CA-NTD) structures have been assembled using symmetry operators from CA_Hexamer structures. c,d Displacement of Q7 or H12-D51 distance as a function of crystallization pH. e, Correlation of Q7 displacement with H12-D51 distance.

Figure 2. The HIV-1 capsid pore is strongly electropositive and recruits dNTP’s with rapid association and dissociation kinetics.

a, Model of an HIV-1 virion with hexamers in an open conformation reveals that the capsid is porous. Surface electrostatic potential shows that the pores are highly electropositive. b, Cross sections through the closed (β-hairpin green) and open (β-hairpin pink) CA_Hexamer showing a central chamber that is accessible in the open state. R18 (cyan) creates a bottleneck at the base of the chamber underneath the β-hairpin. c, Fluorescence anisotropy measurements of dNTP’s binding to CA_Hexamer. d, Example of pre-steady state association kinetics of dCTP with CA_Hexamer. e, Apparent rate constant (k_app) at increasing CA_Hexamer concentrations. f, Dissociation of unlabeled dCTP:CA_Hexamer by excess fluorescent-dCTP. g, R18 co-ordinates the phosphates in a dATP-bound CA_Hexamer structure.

Figure 3. R18 is crucial for nucleotide recruitment, reverse transcription and infectivity.

a, Superposed monomers of R18G (light-pink) and wild-type (light-green) CA_Hexamer. b, Binding of capsid variants to dCTP as measured by fluorescence anisotropy. c, DSF stability measurements expressed as T_m for WT and R18G ± DTT. d, DSF measurements of the effect of dNTP’s on the stability of WT and R18G expressed as ΔT_m relative to unbound. e, Fluorescence anisotropy titrations of dTTP-binding by chimeric CA_Hexamers with different R:G ratios at position 18. f, Comparison of infectivity and reverse transcription of chimeric viruses. g,h, Correlation between HIV-1 capsid dTTP affinity, viral infectivity g and reverse transcription h.

Figure 4. HIV-1 reverse transcription is inhibited by blockade of the capsid pore.

a, In vitro endogenous reverse transcription measuring strong-stop transcripts. b, Residues surrounding Y12 in the H12Y hexamer structure. c, Cartoon and surface representations of the β-hairpin in the H12Y hexamer. d, WT and H12Y reverse transcription kinetics. e, Competition binding of carboxybenzene compounds to CA_Hexamer. f, Change in wild type and R18G CA_Hexamer T_m as measured by DSF in the presence of carboxybenzene compounds. g, CA_Hexamer crystal structure in complex with hexacarboxybenzene, which is co-ordinated by R18. h, Effect of carboxybenzene compounds on endogenous reverse transcription.