Plasmids
All plasmids were based on the subviral plasmid pcHIV42 that encodes all proteins of HIV-1NL4-3 except for Nef. The pcHIV variants MA–SP1, NC–p6, MA–SP1:NC–p6, MA–NC and PR− have been described before8,39,43. MA–SP2 was created by introducing alterations at the NC–SP2 cleavage site32 into pcHIV(MA–NC) by overlap PCR using oligonucleotides 5′-GAGAGACAGGCTTCTTTTTTAGGGAAGACCTGGCCTTCCCACAAGGG-3′ and 5′-CCCTTGTGGGAAGGCCAGGTCTTCCCTAAAAAAGAAGCCTGTCTCTC-3′.
Cell lines and virus particle production
HEK293T cells (Research Resource Identifier CVCL_0063) were grown in Dulbecco’s modified Eagle’s medium, 100 U ml−1 penicillin, 100 μg ml−1 streptomycin and 10% fetal calf serum. Genetic characteristics were confirmed by PCR-single-locus-technology, and cells were regularly tested negative for mycoplasma contamination. At 80% confluency, cells were split 1:3 into T175 flasks (CELLSTAR, Greiner BIO-ONE) the day before transfection. Cells were transfected with pcHIV (70 µg per T175 flask) in three T175 flasks per variant using a standard calcium phosphate transfection procedure. At 48 h post transfection, tissue culture supernatant was collected and cleared through a 0.45-μm-pore filter. The filtered supernatant was layered on top of a 20% (w/v) sucrose cushion and subjected to ultracentrifugation at 107,000g for 1.5 h at 4 C. The pellet was resuspended in PBS and stored in aliquots at −80 °C. For quantification, particle-associated reverse transcriptase activity was determined using the Sybr Green Product Enhanced Reverse Transcription assay44.
Immunoblotting
Particles were separated by SDS–polyacrylamide gel electrophoresis (20%; acrylamide: bisacrylamide 30:1). Proteins were transferred to a nitrocellulose membrane (Millipore) by semidry blotting and stained with the indicated antisera in PBS with Intercept (PBS) Blocking Buffer (LICORBIO) (sheep anti-CA, polyclonal, 1:5,000 (in-house); rabbit anti-NC, polyclonal, 1:400 (in-house)), followed by corresponding IRDye secondary antibodies in PBS with Intercept (PBS) Blocking Buffer (LICORBIO) (donkey anti-sheep, 1:10,000 (LiCOR Biosciences); and donkey anti-rabbit, 1:10,000 (Rockland)). Detection was performed using a LiCOR Odyssey CLx infrared scanner (LiCOR Biosciences) according to the manufacturer’s instructions. Blots are shown in Extended Data Fig. 2a.
Cryo-EM
All cryo-EM samples of purified HIV-1 particles were prepared and imaged similarly. Purified virus was diluted in PBS buffer in 1:3 (v/v) ratio. A 3 ml volume of the diluted virus sample was applied on a glow-discharged Quantifoil 2/2 holey carbon grid, Cu 300 mesh (Quantifoil Micro Tools), and plunge-frozen into an ethane/propane 1:1 mixture using the Leica EM GP2 (100% humidity, blot time 3.5 s, 20 °C). Grids were loaded into a Titan Krios G4 transmission electron microscope operated at 300 kV, equipped with a CFEG electron source, a Falcon4i direct detector camera and a Selectris X energy filter (ThermoFisher Scientific). Images were collected in electron-event representation (EER) format45 using EPU (v3) (ThermoFisher Scientific). All datasets were collected at a magnification of ×130,000, resulting in a pixel size of 0.95 Å, with acquisition times ranging from 3.75 to 4.15 s and with a total dose of 40 e− Å−2. Detailed data acquisition parameters and the number of micrographs for all datasets are given in Extended Data Table 1.
EER videos were rendered as an 8,000 × 8,000 grid and further Fourier-cropped into a 4,000 × 4,000 grid using RELION-4.0 (ref. 46). The videos were motion-corrected, dose-weighted and averaged using RELION-4.0 MotionCorr2 algorithm47. Frames were dose-fractionated into groups resulting in a dose of 0.8 e− Å−2 per fraction. Contrast transfer function (CTF) estimation was performed using the patchCTF algorithm in cryoSPARC v3.3 or v4.4 (ref. 48). Particle picking was performed using crYOLO (v1.7.6)49. For each dataset, a new model was trained in crYOLO using a training dataset annotated in a randomly selected set of 50–100 micrographs. Picking models were trained to distinguish virus from background by manually and indiscriminately covering the complete surface of the virus with picks using the crYOLO boxmanager GUI (Extended Data Figs. 3–6).
Cryo-EM data processing for PR− mutant
The data processing pipeline for PR− MA is summarized in Extended Data Fig. 3. A total of 9,942 motion-corrected micrographs and 3,568,755 particle positions were imported into cryoSPARC v4.4 (ref. 48). Particles, which were picked from large numbers of arbitrary positions over the virus surface, were extracted with a box size of 480 × 480 pixels and Fourier-cropped to 180 × 180 pixels. The first step was to perform a high-resolution reconstruction of the well-ordered immature CA layer to generate CA positions and orientations for use as priors to initialize the refinement of the MA layer. Two rounds of 2D classification were performed in which classes showing side and top views of the immature CA layer were selected (1,630,811 particles) and subjected to heterogeneous refinement with 8 classes, C6 symmetry imposed, and using a previously solved immature structure of an in vitro-assembled HIV capsid (Electron Microscopy Data Bank (EMDB) accession code EMD-3782) as a starting reference 50. Classes showing resolved secondary structures in the CA layer with visible densities representing MA and membrane layers were selected (805,005 particles), re-extracted with a box size of 480 × 480 pixels, and Fourier-cropped to 416 × 416 pixels. Duplicate particles were removed on the basis of a spatial separation, and the accepted particles (676,036 particles) were subjected to three further rounds of 3D refinement with local spatial and angular searches (local refinement), C6 symmetry imposed, and a mask comprising the immature CA layer. Local and global CTF refinements combined with Ewald sphere correction as implemented in cryoSPARC v4.4 were performed in between the 3D refinements. Afterwards, the particles were imported to RELION-4.0, re-extracted with a box size of 480 × 480 and Fourier-cropped to 416 × 416 pixels. The particles were then subjected to Bayesian polishing in RELION-4.0 (ref. 51). The polished particles were imported back to cryoSPARC v4.4 and subjected to a final local refinement resulting in a final high-resolution focused map of the immature CA layer.
The coordinates of the CA layer were then used to predict initial coordinates for the MA layer. To do this, the particles from the CA reconstruction (which is centred on the six-fold symmetry axes) were symmetry-expanded using six-fold symmetry and the 3D coordinates of the centre of the box were shifted to define the positions of the six surrounding three-fold axes of the MA layer. The shifted particles (3,920,498 particles) were then extracted using the new box centre with a box size of 512 × 512 pixels and Fourier-cropped to 256 × 256 pixels. Duplicate particles were removed, and the accepted particles were reconstructed with C3 symmetry. The resulting map was then used to subtract densities corresponding to the immature CA layer from the particles. Subtraction was necessary for the immature MA because the immature CA layer otherwise dominated the reconstruction. The subtracted particles were subjected to three rounds of heterogeneous refinement in cryoSPARC v4.4, with C3 symmetry imposed and eight classes using a cryo-ET-derived reconstruction of the immature MA lattice (EMD accession code EMD-13087) low-pass-filtered to 10 Å as a starting reference 1. Particles belonging to classes that showed aligned MA and membrane layers were selected in each iteration. Selected particles (120,104 particles) were then subjected to non-uniform refinement52 with C3 symmetry and a mask comprising both the MA and membrane layers followed by a local 3D refinement with a mask comprising only the MA layer.
Next, the dataset was expanded by using the refined positions of MA trimers to predict the positions of neighbouring MA trimers (lattice expansion). To do this, the particles were symmetry-expanded with C3 symmetry (generating the two additional symmetry-related copies of each particle), the centre of the box was shifted to a neighbouring MA trimer, and the particles were re-extracted with a box size of 512 × 512 pixels and Fourier-cropped to 256 × 256 pixels. Duplicate particles were then removed. Lattice expansion increased the size of the dataset, improving the resolution of our reconstruction and the quality of the map. Two iterations of the lattice expansion were performed. To identify particles containing well-ordered MA lattice, the accepted particles (819,672 particles) were subjected to two rounds of 3D classification without angular search, using 10 classes. Particles were lattice-expanded as described above between the two 3D classifications. Classes showing a well-resolved MA layer were selected (174,245 particles) and subjected to a final round of local refinement, imposing C3 symmetry and a mask comprising the MA layer resulting in a final map of MA with a resolution of 5.8 Å.
Cryo-EM data processing for MA–SP1
The data processing pipeline for MA of MA–SP1 is summarized in Extended Data Fig. 4. A total of 14,222 motion-corrected micrographs and 5,704,512 initial particle positions were imported into cryoSPARC v3.3 (ref. 48). Particles, which were picked from large numbers of arbitrary positions over the virus surface, were extracted with a box size of 512 × 512 pixels and Fourier-cropped to 192 × 192 pixels. The first step was to perform a high-resolution reconstruction of the well-ordered immature CA layer to generate CA positions and orientations for use as priors to initialize the refinement of the MA layer. Particles were subjected to two rounds of 2D classification, and classes showing side and top views of the immature CA layer were selected (2,969,039 particles; Extended Data Fig. 4b). To accelerate computation, the selected particles were divided into 4 approximately equally sized subsets (each containing about 750,000 particles). Each subset underwent heterogeneous refinement with six classes, enforcing C6 symmetry, using a previously determined structure of an in vitro-assembled HIV-1 capsid (EMDB accession code EMD-3782) as a starting reference 50 (Extended Data Fig. 4c). The highest-quality classes from each batch were then pooled for a series of refinement steps. They were subjected to non-uniform 3D refinement before particle re-extraction with a box size of 512 × 512 pixels, and Fourier-cropped to 384 × 384 pixels. Duplicate particles were removed on the basis of a spatial separation distance, and the accepted particles (885,809 particles) were subjected to 3D refinement with local angular and spatial searches (local refinement), imposed C6 symmetry and a mask comprising the immature CA layer. The particles were then subjected to local CTF refinement, followed again by 3D refinement. Particles were re-extracted with a box size of 512 × 512 pixels, and Fourier-cropped to 450 × 450 pixels, and again subjected to local CTF refinement and local 3D refinement. Heterogeneous refinement was then performed to remove any remaining low-quality particles, using three classes, from which the highest-quality class was selected. Global CTF refinement53 and further local refinement were then performed. Afterwards, the particles were imported to RELION-4.0 and subjected to Bayesian polishing51. The polished particles were imported back to cryoSPARC v3.3 and subjected to further 3D refinement and local CTF refinement to generate a final high-resolution immature CA reconstruction (Extended Data Fig. 4d).
The coordinates of the CA layer were then used to define initial coordinates and orientations for the MA-layer-focused reconstruction. The 3D coordinates of the centre of the box were shifted to the centre of the three-fold symmetry axis of the MA layer, which could be seen in the high-resolution CA reconstruction. Local refinement, without provision of a new reference, was then performed with imposed C3 symmetry and a mask comprising only the MA layer (Extended Data Fig. 4e). Next, the dataset was expanded by using the refined positions of MA trimers to predict the positions of neighbouring MA trimers (lattice expansion). To do this, the particles were symmetry-expanded with C3 symmetry (generating the two additional symmetry-related copies of each particle), and the centre of the box was shifted to the neighbouring six-fold symmetry axis, before duplicate particles were removed on the basis of a spatial separation distance. The accepted particles (1,398,844 particles) were subjected to a further round of local refinement with C6 symmetry enforced. The particles were once again lattice-expanded, with the centre of the box shifted to the three-fold axis of neighbouring MA timers. Duplicate particles were removed, and the accepted particles (5,486,693 particles) were reconstructed with C3 symmetry imposed to generate the final MA reconstruction (Extended Data Fig. 4e).
Cryo-EM data processing for MA–NC
The processing strategy for MA–NC was almost identical to that of MA–SP1, including for the initial CA reconstruction and for the subsequent MA refinement steps. The data processing pipeline is summarized in Extended Data Fig. 5.
Cryo-EM data processing for WT
The data processing pipeline for WT MA is summarized in Extended Data Fig. 6. A total of 19,530 motion-corrected micrographs and 5,801,053 initial particle positions were imported into cryoSPARC v3.3 (ref. 48). In contrast to the case for the PR−, MA–SP1 and MA–NC samples described above, there is no immature CA layer present in WT virions, so MA was reconstructed directly. Particles, which were picked from large numbers of arbitrary positions over the virus surface, were initially extracted with a box size of 480 × 480 pixels, Fourier-cropped to 240 × 240 pixels and subjected to 2 rounds of 2D classification to identify and select classes showing top and side views of the MA layer (Extended Data Fig. 6b). They were then subjected to two rounds of heterogeneous refinement, each with four classes and imposed C6 symmetry, using a cryo-ET-derived mature MA lattice reconstruction as a starting reference (EMDB accession code EMD-13088; Extended Data Fig. 6c). The resulting highest-quality class was selected (61,672 particles) and subjected to non-uniform 3D refinement with C6 symmetry imposed52. Duplicate particles were removed on the basis of spatial separation distance, and the accepted particles (57,260 particles) were then subjected to 3D refinement with local spatial and angular searches (local refinement), using a refinement mask comprised of the MA layer. The dataset was expanded by using the refined positions of MA trimers to predict the positions of neighbouring MA trimers (lattice expansion). To do this, the particles were symmetry-expanded with C3 symmetry (generating the two additional symmetry-related copies of each particle), the centre of the box was shifted to the neighbouring six-fold symmetry axis, and the particles were re-extracted with a box size of 480 × 480 pixels and Fourier-cropped to 356 × 356 pixels (276,547 particles). The particles were subjected to local CTF refinement, followed by a further round of local refinement. The particles were once again lattice-expanded, with the centre of the box shifted to the neighbouring MA timers. Duplicates were again removed, and the accepted particles (898,502 particles) were finally reconstructed with imposed C3 symmetry (Extended Data Fig. 6d).
Automated lipid fitting
Automated docking of lipid candidates into the MA–SP1 ligand density (cholesterol, PtdIns(4,5)P2, phosphatidylserine, phosphatidylcholine and phosphatidylethanolamine) was performed using RosettaEmerald, using the protocol described in ref. 54. All resulting fits were visually inspected and Q-scores of all ligand fits were determined using MapQ in USCF Chimera 1.15 (ref. 55). The top ten fits for each ligand, according to Q-score, are provided (Extended Data Fig. 7a).
Fourier analysis of 2D class averages of cleavage mutants
Cryo-EM data preprocessing and particle picking were performed as described in the cryo-EM subsection. Images were extracted with a box size of 480 × 480 pixels and downsampled to 240 × 240 pixels resulting in a final pixel size of 1.9 Å. 2D class averages of mutants that displayed side views of membranes were selected manually in cryoSPARC v4.4. Afterwards, the 2D classes were reoriented according to a reference class in which the membrane bilayer was oriented perpendicularly to the y axis of the 2D class using cross-correlation (Extended Data Fig. 8). Then, the pixel values along the MA layer (section parallel to the inner leaflet of the viral membrane) were interpolated with 1,024 points and exported as 1D vectors. The vector was filtered using a Hann function and zero-padded to a total of 4,096 sampling points. Fast Fourier transformation of the zero-padded signal was plotted in MATLAB v2022a (MathWorks) and analysed for peaks corresponding to MA lattice spacing frequencies. All of the signal processing steps were performed in MATLAB v2022a (MathWorks).
Atomic model building and refinement of SP2 bound to MA
The solution structure of myristoylated HIV-1 MA (myrMA; Protein Data Bank accession code 2H3I)12 was used as an initial MA structure for building into the MA–SP1 density. Initial coordinates for SP211–16 were generated using AlphaFold (v2.2.0)56, which were fitted roughly into the SP2 density in USCF Chimera55. The initial model was then flexibly fitted, with manual adjustments, into the density with ISOLDE 1.3 (in ChimeraX 1.3)57,58. A single round of real-space refinement was then performed in Phenix-1.21. Final model validation statistics and the map-to-model Fourier shell correlation were calculated in Phenix-1.21 (ref. 59) and are given in Extended Data Table 2.
ModelAngelo predictions of SP2
The 3.1-Å-resolution map of the mature MA of the MA–SP1 cleavage mutant was used for automated ModelAngelo v1.0 (ref. 60) atomic model predictions. The box size was cropped to 128 × 128 pixels and only sequences built into the central trimer were considered. The ModelAngelo job was run with default parameters in RELION-5.0. The sequence input was either the HIV-1 Gag sequence or no sequence. Afterwards, sequences built into the side pocket cryo-EM densities from the central MA trimer were extracted and analysed by the Clustal Omega multiple sequence analysis tool61. The ModelAngelo automated model building and sequence prediction results are shown in Extended Data Fig. 7b–d.
Expression and purification of HIV-1 MA
The expression plasmid pET11b-MA encodes the HIV-1 pNL4–3 MA domain with a six-residue C-terminal His tag62. BL21(DE3) Escherichia coli competent cells for protein expression were co-transformed with the pET11b-MA plasmid and a plasmid encoding the yeast N-terminal myristoyltransferase. The protein was expressed and purified as described previously17,63 with modifications. Cells were grown at 37 °C at 180 rpm in a lysogeny broth medium containing 100 mg l−1 of ampicillin and 50 mg l−1 of kanamycin. When absorbance at 600 nm (A600nm) reached about 0.6, 15 mg l−1 myristic acid (Sigma-Aldrich) was added to the lysogeny broth medium. After 30 min, cells were induced with 0.5 mM isopropyl β-d-1-thiogalactopyranoside and grown for another 5 h at 37 °C at 180 rpm. Afterwards, the cells were spun down at 3,500g, and the pellets were stored at −80 °C until further use. For purification, 6 g of the cell pellet was diluted in 60 ml of lysis buffer (25 mM Tris pH 8, 500 mM NaCl, 2 mM TCEP and 2 mM phenylmethylsulfonyl fluoride) and sonicated for 2 min. Then, 20 µl of benzonase was added to the lysed cells. The lysate was incubated on ice for 10 min and then spun down at 50,000 rpm at 10 °C for 45 min (Beckman Coulter, 50.2 TI rotor). The supernatant was collected and treated with polyethyleneimine to a final concentration of 0.03%, incubated on ice for 5 min and subsequently centrifuged at 10,000 rpm at 4 °C for 10 min (Beckman Coulter, JA-25.50 rotor). The supernatant was collected, and powdered ammonium sulfate (approximately 15 g) was added to the supernatant on ice with constant stirring until protein precipitate was observed, followed by centrifugation at 10,000 rpm at 4 °C for 10 min (Beckman Coulter, JA-25.50 rotor). The pellet was resuspended in 4 ml of binding buffer (20 mM Tris-HCl, pH 8, 100 mM NaCl, 2 mM TCEP) and loaded onto a HisTrap column (Cytiva) equilibrated with the binding buffer. The column was then washed with a wash buffer (20 mM Tris-HCl, pH 8, 100 mM NaCl, 2 mM TCEP and 20 mM imidazole), and the protein was subsequently eluted with elution buffer (20 mM Tris-HCl, pH 8, 100 mM NaCl, 2 mM TCEP and 250 mM imidazole). Fractions containing myrMA were collected and further purified by gel filtration using Superdex 75 16/600 (Cytiva) equilibrated in buffer containing (20 mM Tris-HCl, pH 8, 500 mM NaCl, 1 mM TCEP). Fractions containing myrMA were collected and stored at −80 °C. The presence of myristoylation modification was confirmed by mass spectrometry.
2D crystallization of HIV MA
2D crystallizations were performed in a cleaned polytetrafluoroethylene block containing 60-ml side-entry reservoirs, combining previous protocols64. A 58 ml volume of crystallization buffer (12 mM sodium phosphate buffer pH 7.8; 2.5 mM sodium acetate pH 7.6; 150 mM sodium chloride; 10% glycerol)33,65 was added to 6 crystallization reservoirs. Afterwards, 1 µl of a freshly prepared lipid mixture mimicking the inner leaflet of the viral membrane (molar fractions: 31% cholesterol; 6% POPC; 29% POPE; 27% POPS; 7% PtdIns(4,5)P2 (ref. 34)) in 9:1 (v/v) chloroform/methanol solution at a lipid concentration of 0.01 mg ml−1 was carefully added on top of each buffer surface. The polytetrafluoroethylene block was then incubated in a closed Petri dish with a wet filter paper placed underneath the block for 60 min to allow a lipid monolayer to form at the air–water interface. Then a Quantifoil 2/2 holey carbon grid, Au 200 mesh (Quantifoil Micro Tools) was placed on top of each reservoir. A solution containing purified myrMA was then injected into each reservoir from the side entrance. The final concentration of myrMA in the reservoir was 12 mM. After 10 min, HIV-1 SP2 in PBS buffer was added to 3 of the 6 experimental reservoirs to a final concentration of 120 mM. The same amount of PBS buffer was added to the three control reservoirs. All samples were incubated for an additional 60 min, and then the grids were carefully lifted from the surface of the reservoirs and plunge-frozen using the Vitrobot Mark IV (4 s blot time; 3 blot force; 100% humidity).
Grids were loaded into a Talos Glacios cryo-transmission electron microscope operating at 200 kV and equipped with a Falcon4i direct electron detector (ThermoFisher Scientific). For each grid, 5 grid squares were selected manually and in each square 30 holes were randomly selected in the EPU software (v3). A single acquisition position was selected in the centre of each hole resulting in 150 micrographs automatically collected from each grid. The micrographs were collected as videos of 40 frames with a total dose of 40 e− Å−2 at a magnification of ×92,000 resulting in a pixel size of 1.20 Å per pixel. Videos were motion-corrected, dose-weighted and averaged using the RELION-4.0 MotionCorr2 algorithm46,47. CTF estimation was performed using CTFfind466. Particles were picked as a grid of points separated by 128 pixels placed in a 4,000 × 4,000 micrograph, resulting in 139,500 particles for each dataset (Extended Data Fig. 11b). Particles were extracted with a box size of 256 pixels and Fourier-cropped to 128 pixels. The particles were then imported to cryoSPARC v4.4 (ref. 48) and subjected to two rounds of 2D classification (Extended Data Fig. 11). Classes showing a 2D crystal MA lattice were selected, and particles from these classes were considered as particles containing a 2D crystal lattice. All six datasets were processed the same way. To facilitate visual comparison of class averages and simulations in Fig. 4e–h, greyscales were made similar using Fiji (ImageJ v1.54f).
Plasmids, reagents and cell lines for fusion assays
HIV-1 GagPol was expressed by pCMV ΔR8.2 (Addgene plasmid no. 12263). The pCAGGS HIV-1JRFL gp160 expression plasmid was provided by J. Binley. The pN1 CypA–HiBiT plasmid (made by J. Grover) was derived from pEGFP-N1-CypA. To generate this plasmid, human cyclophilin A was cloned from HeLa cDNA and inserted into pEGFP-N1 using the EcoRI and BamHI sites. To generate CypA–HiBiT, a synthetic oligonucleotide, which encoded the linker sequence GSGSSGGGGSGGGGSSG followed by the HiBiT peptide VSGWRLFKKIS, was inserted to replace eGFP at the C terminus of CypA using the BamHI and NotI sites. This construct is based on a similar design by G. Melikyan.
The Gag cleavage mutants were constructed through site-directed mutagenesis and Gibson assembly. The alterations for pCMV ΔR8.2 MA–CA, pCMV ΔR8.2 MA–SP1, pCMV ΔR8.2 MA–NC and pCMV ΔR8.2 MA–p6 were recreated from previous literature6,67,68. Gene blocks of the MA–CA, MA–SP1, MA–NC and MA–p6 GagPol sequences were ordered from Twist Bioscience and combined with fragments of the pCMV ΔR8.2 backbone amplified by PCR for Gibson assembly using the Gibson Assembly Master Mix from New England Biolabs. The alterations for pCMV ΔR8.2 MA–SP2 and pCMV ΔR8.2 NC–p6 were created using site-directed mutagenesis using the pCMV ΔR8.2 MA–p6 and pCMV ΔR8.2 constructs as templates, respectively. These fragments were combined with pCMV ΔR8.2 backbone fragments amplified by PCR and combined with Gibson assembly using the same protocol as above.
The human CD4-expressing vector pcDNA-hCD4 was provided by H. Gottlinger. The pMX-puro PH-PLC∂LgBiT plasmid was provided by Z. Matsuda30. The following reagents were obtained through the National Institutes of Health (NIH) HIV Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases (NIAID), NIH: indinavir sulfate, ARP-8145, and ritonavir, ARP-4622, both contributed by Division of AIDS, NIAID; human CCR5 expression vector (pcCCR5), ARP-3325, contributed by N. Landau.
HEK293 cells (ATCC no. CRL-1573) were grown in the presence of 5% CO2 using RPMI-1640 medium from ThermoFisher Scientific supplemented with 10% fetal bovine serum (Invitrogen), 100 U ml−1 of a penicillin and streptomycin solution and 2 mM l-glutamine. Cells were transfected at 60%–80% confluency, and culture medium was exchanged before transfection. Cells were regularly tested negative for mycoplasma contamination.
Virus-like particle preparation for fusion kinetics
Virus-like particles (VLPs) were produced by transfecting three 10-cm plates of HEK293 cells with 12 µg DNA per 10-cm plate using polycation polyethylenimine (pH 7.0, 1 mg ml−1). Immature VLP preparations were transfected with a final concentration of 1 µM indinavir and 10 µM ritonavir in the cell culture medium. Plasmids were transfected in a 1:1:1 ratio of Gag/Env/CypA-HiBiT. Cell culture medium was collected 2 days after transfection, and then spun down for 5 min to pellet cells. Supernatants were transferred to 38.5-ml ultracentrifuge tubes and underlaid with 5 ml sterile-filtered 15% sucrose in PBS. VLPs were then pelleted by ultracentrifugation at a maximum of 131,453 RCF using a Beckman Coulter SW28 swinging bucket rotor at 27,000 rpm for 1 h at 4 °C. Supernatant was removed and viruses were resuspended in 1:100 volume (300 µl) of serum-free CO2-independent medium (ThermoFisher Scientific). Particles produced in the presence of protease inhibitors were resuspended in serum-free CO2-independent medium with final concentrations of 1 µM indinavir and 10 µM ritonavir. Each VLP preparation was aliquoted into about 20 tubes at 15 µl per tube and stored at −80 °C until use. After collection, each aliquot of VLPs was used once for fusion kinetics assays to minimize particle destruction during freeze–thaw cycles.
VLP normalization for fusion kinetics
VLP volumes were normalized on HiBiT incorporation using the Nano-Glo HiBiT Lytic Detection System from Promega according to the manufacturer’s instructions. Three volumes of the VLPs (2 µl, 4 µl and 6 µl) were measured per sample. Plates were placed in a Promega GloMax Explorer GM3500 Multimode Microplate Reader and read using the manufacturer’s suggested protocol. Readings were then graphed in Excel with a linear trendline. The trendline equation was used to calculate the volume of each sample containing 3 × 107 relative light units of HiBiT.
Fusion kinetics assay
The split nanoluciferase fusion kinetics assay described in ref. 30 was modified to enable real-time live monitoring of fusion events. Briefly, HEK293 cells were transfected with a 1:1:1 ratio of CD4, CCR5 and PH-LgBiT. After 24 h, cells were collected and Endurazine (Promega) and DrkBiT peptide were added to the solution to a final concentration of 1× and 1 µg ml−1, respectively. A white 96-well flat-bottom plate was prepared with 100 µl of the cell solution at a density of about 2 × 104 cells per well. Prepared VLPs were added to the wells and spinoculation was performed at 1,200 RCF and 12 °C for 2 h. The plate was then covered with sterile BREATH-EASY*GAS PERMEABLE film (USA Scientific, no. 9123-6100). The plate was read continuously in a Promega GloMax Explorer GM3500 Multimode Microplate Reader using an automatic protocol for 24 rounds of reading wells every 2 min with a 1.5 s integration time per well at 37 °C. Readings were exported into Excel, in which relative light units were calculated, and then processed into percentage of total fusion. The percentage of total fusion curves were processed using an in-house Mathematica code (Supplementary Data 1; written by A. Lee) to generate T1/2 for each sample. Results were plotted using GraphPad Prism v10.4.1.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
