Permeability of the HIV-1 capsid to metabolites modulates viral DNA synthesis

Reverse transcription, an essential event in the HIV-1 lifecycle, requires deoxynucleotide triphosphates (dNTPs) to fuel DNA synthesis, thus requiring penetration of dNTPs into the viral core. The central cavity of the capsid protein (CA) hexamer reveals itself as a plausible channel that allows the passage of dNTPs into assembled capsids. Nevertheless, the molecular mechanism of nucleotide import into the capsid remains unknown. Employing all-atom molecular dynamics simulations, we established that cooperative binding between nucleotides inside a CA hexamer cavity results in energetically-favorable conditions for passive translocation of dNTPs into the HIV-1 capsid. Furthermore, binding of the host cell metabolite inositol hexakisphosphate (IP6) enhances dNTP import, while binding of synthesized molecules like benzenehexacarboxylic acid (BHC) inhibits it. The enhancing effect on reverse transcription by IP6 and the consequences of interactions between CA and nucleotides were corroborated using atomic force microscopy, transmission electron microscopy, and virological assays. Collectively, our results provide an atomistic description of the permeability of the HIV-1 capsid to small molecules and reveal a novel mechanism for the involvement of metabolites in HIV-1 capsid stabilization, nucleotide import and reverse transcription.


Introduction
Fusion between HIV-1 virions and target cells engenders the release of the viral capsid into the host cell cytoplasm. The capsid, a cone-shaped protein assembly composed of ~250 capsid protein (CA) hexamers and 12 CA pentamers, encapsulates two copies of the viral single-stranded RNA (ssRNA) genome and viral proteins, including reverse transcriptase (RT) and integrase (IN) 1-3 . Successful infection requires reverse transcription of the ssRNA into double-stranded viral DNA (vDNA) and integration of vDNA into the host cell genome. Based on steric hindrance of the intact capsid by the nuclear pore complex 4 , major structural rearrangement or capsid shedding has been postulated to occur prior to or during nuclear import 5,6 , a process commonly referred to as uncoating. While reverse transcription likely initiates within the capsid and is affected by uncoating, the mechanisms connecting the two viral processes remain a long-standing question in HIV-1 biology [7][8][9][10] . We and others have proposed that reverse transcription mechanically induces changes in capsid morphology and triggers uncoating 11,12 . To support vDNA synthesis, RT requires adequate concentrations of deoxynucleotide triphosphates (dNTPs) within the lumen of the capsid. The HIV-1 capsid has been described as semipermeable and proposed to regulate the passage of ions including dNTPs from the cytoplasm to its interior 3,4,8 .
Inositol phosphates (IPs) are abundant cellular metabolites that play crucial roles in fundamental cellular processes 13 . Interestingly, ions and small molecules like inositol hexakisphosphate (IP6) and benzenehexacarboxylic acid (BHC, or mellitic acid) have been reported to bind to HIV-1 Gag and CA hexamers and impact virion assembly and reverse transcription 4,[14][15][16][17][18] . Here, using a multi-pronged approach combining all-atom molecular dynamics (MD) simulations, atomic force microscopy (AFM), transmission electron microscopy (TEM), confocal microscopy, and virus infectivity assays, we have analyzed the effect of IP6 on HIV-1 capsid stability and reverse transcription. Furthermore, we determined the molecular mechanism that regulates translocation of dNTPs through the capsid.
The CA hexameric cavity is surrounded by six copies of a β-hairpin, helix 1 and a short loop connecting them (Fig. 1a, b). Its radial profile, illustrated in Fig. 1a for the native CA hexamer 19 , contains an exterior tubular channel and an interior conical volume separated by a 1.4 Å radial constriction near Arg18 residues (Fig. 1a). The inner surface of the outward facing cavity is surrounded by residues Asn5, Gln7, and Gln13 that lie within or nearby the β-hairpin. Helix 1 sequences are nearly identical across primate lentiviruses, including invariant charged residues Arg18, Lys25, Glu28, Glu29, and Lys30 (Fig. 1b). These residues, which account for the electrostatic potential of the hexamer cavity, contrast to the negative charge found in the exposed surfaces of the hexamer (Fig. 1c). Overall, based exclusively on the structural features of CA hexamers, the volume occupied by the ring of six Arg18 side chains represents a steric barrier for translocation of small molecules through the central cavity.
To explore the effects of the dynamic behavior of CA hexamers on the permeability of molecules through assembled HIV-1 capsids, we employed free-energy MD simulations to determine the free energy landscapes of dNTPs, IP6, and BHC for interacting with the central cavity ( Fig. 1d, S1, simulations #1-10). To monitor the displacement of the nucleotide in the cavity, a progress variable (PV) was introduced as the location of the center of mass of each small molecule of interest on the pore axis of the hexamer (Fig. 1c). The resulting free energy landscapes displayed energy basins in proximity to Arg18 for dNTPs, rNTPs, NMPs, IP6, and BHC (Fig. 1d, Fig. S1, Table S1). Unexpectedly, the depth of the energy basin differed between nucleotide types, as a minimum free energy of -7 to -8 kcal/mol was observed for dATP and dGTP, while a minimum of -10 to -12 kcal/mol was observed for dCTP and dTTP. We infer that steric effects of the larger purine nucleotides incur lower binding affinity compared to the more compact pyrimidines. These results reveal nucleoside-specific binding of dNTPs to CA hexamers. However, once bound, release of the dNTP would require overcoming a barrier of at least 6 kcal/mol; thus, dNTPs are highly unlikely to translocate through CA hexamers on their own.
In our simulations, a single nucleotide becomes trapped within the cavity, as the energy well is too deep to enable its release. Our MD simulations revealed that the central hexamer cavity could accommodate two small molecules simultaneously interacting with the ring of Arg18 residues (movie 1; molecular simulation #25, 26). Therefore, we increased the stoichiometric ratio between small molecules and CA hexamer to 2:1 (simulation #11, #12, #13; of CA hexamers to translocate dNTPs in the presence of BHC, IP6 or another dNTP. Specifically, a positive free-energy gradient, which promoted dNTP motion inwards, was observed for dNTPs and IP6 (Fig. 2a, 2c, S2b, S2c), while a negative gradient was observed in the presence of BHC (Fig. 2e,   S2d). The negative gradient induced by BHC implies that it inhibits nucleotide import. Additionally, the difference in the free energy gradient induced by IP6 compared to dNTPs shows that the former is a stronger activator of nucleotide import. Our molecular simulations revealed that IP6 will increase dNTP transport, whereas BHC will block or reduce it.
Alterations of Lys25 were previously found to alter capsid stability but still produce virus-like particles 20,21 . Results from our simulations showed that Lys25 and Glu28 substitutions modified dNTP import (simulations #14-17, Fig. S1d). In particular, K25A was predicted to enhance import while K25E/N were predicted to inhibit translocation, indicating that Lys25 charge is not the sole determinant of nucleotide import. These data indicated that Lys25 plays a role in coordinating solvent molecules around the nucleotide after dewetting of the dNTP through its binding to Arg18 (Fig. S3).
Based on these predictions, the effect of K25N on HIV-1 capsid stability and reverse transcription was examined. In vitro assembly reactions assessed by TEM (Fig. S4) and magic angle spinning (MAS) NMR (Fig. S5) revealed that wild-type (WT) and K25N constructs behaved similarly, while K25A constructs were unable to form mature-like conical capsids and were not pursued for further experiments (Fig. S4). Consistent with these observations, K25N HIV-1 was morphologically similar to the WT virus (Fig. 3a). However, K25N capsids were less stable than WT capsids based on CA retention in permeabilized viruses, which was assessed by antibody staining and confocal microscopy ( Fig. 3b, S6).
To evaluate infectivity, cell lines, peripheral blood mononuclear cells (PBMCs) and monocyte-derived macrophages were infected with normalized levels of WT or K25N HIV-1NL4-3-based luciferase reporter viruses. While WT HIV-1 infected all cell types tested, K25N was noninfectious (Fig. 3c,d).
Similar results were obtained with WT and K25N HIV-1LAI reporter viruses (data not shown). To determine the step of the virus life cycle at which K25N HIV-1 infection was impaired, early and late reverse transcripts and 2-long terminal repeat (LTR)-containing circles were measured by quantitative PCR (qPCR). Levels of K25N early and late reverse transcripts were reduced 540-fold and 385-fold, respectively, relative to WT HIV-1 (Fig. 3e). The number of 2-LTR circles, a surrogate marker of vDNA nuclear entry, was likewise 500-fold lower for K25N compared to WT HIV-1 (Fig.   3e). These data indicate that K25N HIV-1 is defective for reverse transcription.
Changes in capsid stiffness and morphologies during reverse transcription were monitored in realtime by AFM (Fig. 4) 11,24,25 . The effect of IP6 binding on the stiffness of intact HIV-1 capsids was analyzed using AFM operated in the nanoindentation mode. In agreement with our previously reported findings, isolated WT capsid had an average stiffness value of 0.132 ± 0.007 N/m (n=30) (Fig, 4a). Remarkably, binding of IP6 to isolated HIV-1 capsids increased their stiffness by nearly 2fold to an average value of 0.245 ± 0.021 N/m (n=36). Initially, untreated WT capsids exhibited an average stiffness value of 0.131 ± 0.012 N/m; n=11. Upon IP6 addition, the value increased to 0.258 ± 0.035 N/m (n=15, time = 0 h), but decreased immediately after reverse transcription was initiated, reaching a minimum of 0.078 ± 0.013 N/m (n=7) at 7 h; the minimum stiffness value was reached earlier for capsids treated with IP6 than for capsids without IP6 (5 h compared to ~20 h) 24  showing that IP6 stabilizes HIV-1 capsids and promotes viral DNA synthesis 14 and our computational experiments that revealed BHC blocked nucleotide translocation through the CA hexameric cavity.
To further characterize the effects of IP6 on reverse transcription, we analyzed the morphologies of IP6-treated WT capsids during reverse transcription by AFM operated in the quantitative imaging mode. Prior to reverse transcription, capsids had a well-defined conical appearance. A representative capsid (out of a total of 20 that were imaged) is shown in Fig. 4c. After 5 h, openings at various sizes in the capsid appeared ( Fig. 4d-f). Similar to our previous findings 26,27 , the openings were localized exclusively at or near the narrow end of the capsids. Untreated and IP6-treated HIV-1 capsids underwent complete disassembly during the time course of the experiments. However, complete disassembly was 3.4 times faster in the presence of IP6 (7 h vs. to 24 h). Analysis of the reactions beyond 7 h of reverse transcription revealed mostly fragments of various sizes that lacked a defined morphology (from a total of 52 capsids, 11 remained intact). Overall, we find that IP6 accelerates capsid disassembly during reverse transcription.
The stiffness of the K25N/E45A isolated capsids was also measured during reverse transcription.
While the stability of these capsids was not sufficient to withstand the full duration of the experiment due to spontaneous disassembly, their stiffness remained unchanged over 5 h. Although addition of IP6 did not affect CA retention on K25N/E45A capsids (Fig. S7), their stability was increased, which enabled us to measure their stiffness values during 24 h of reverse transcription. In agreement with our virus infectivity and reverse transcription analyses, the stiffness of the double mutant capsids remained unchanged during 24 hours of reverse transcription (Fig 4c), consistent with the observed reverse transcription defect of K25N/E45A HIV-1.
Reverse transcription and the consequent disassembly of the HIV-1 capsid are fueled by the ability of the capsid to import dNTPs from the cytoplasm. Free energy calculations reveal a molecular dynamic view of the interaction between a freely diffusing nucleotide and the HIV-1 CA hexamer ( Fig  5). First, a nucleotide freely diffuses from the exterior solvent to the beta-hairpin region.
Subsequently, the loss of entropy by binding of the nucleotide to Arg18 is compensated by the strength of the electrostatic interactions between Arg18 and dNTP phosphate residues. Hydrogen bonds between the nucleotide base and polar residues in helix 1 confer distinct binding affinities for individual nucleotide species, which acts as a selectivity filter. However, to be released into the capsid lumen, a single nucleotide needs to overcome a high energetic penalty (free energy barrier of over 6 kcal/mol). Engagement of a second charged molecule such as dNTP or IP6 within the channel provides the required thermodynamic effector to release the initial dNTP within the capsid lumen.
Ion channels and transporters have evolved to employ a cooperative translocation mechanism to displace cargos between distinct environments 28,29 . Our free energy calculations for multiple small molecules suggest that binding of two nucleotides within the CA cavity promotes dNTP import.
Interestingly, similar to water molecules in aquaporins 29 , dNTPs flip as they pass through hexamers.
Furthermore, pure electrostatic interactions between CA and dNTPs are insufficient to describe the molecular mechanism of dNTP import, as the loss and gain of entropy associated with dewetting dNTPs and Arg18 is a key contributor to the free energy of binding. As a result, the cooperative mechanism for nucleotide translocation is not universal for any set of negatively charged molecules, as IP6 promotes dNTP import while BHC inhibits it.
We conclude that HIV-1 has evolved to import dNTPs from the cytoplasm into the capsid and to discern nucleoside type. Molecular determinants of this nucleoside-specific importing mechanism are encoded in highly conserved residues such as Lys25. In addition, we conclude that IP6 enhances import of dNTPs for efficient reverse transcription. The discovery of metabolite-dependent nucleoside-specific import provides a unique target for development of new therapies against HIV-1.

CA hexamer model building
The initial coordinates of the HIV-1 CA hexamer (Fig. S8a, b) were generated by applying a six-fold symmetry operation onto a native full-length HIV-1 CA (PDB accession code 4XFX). The two loops between residues 5 to 9 and residues 222 to 231, missing in the original structure, were built using Modeler 26 . Once the hexamer was built, the protonation states of titratable residues, namely histidine, asparagine, lysine and cysteine, were assigned using PDB2PQR 30 at pH 7.4.

Molecular mechanics parameterization of NTPs, IP6 and BHC
Parameters for the negatively-charged small molecules (Fig. S9), except ATP, which has parameters available in the CHARMM general force field, were derived by analogy following the CGENFF protocol 31 . The parameter penalties and charge penalties in each generated parameter files were less than ten indicating good analogy with the available atom types present in CGENFF 32 . A magnesium ion was added to the triphosphate group present in dNTPs. The coordination number of the magnesium ion with the phosphate group of the dNTP was constrained using coordnum in Colvars 33 .

Simulation setup
Small charged molecules such as NTP/IP6/BHC were placed in the central cavity of the CA hexamer model. The models were then solvated using the TIP3P water model 34 . Additionally, excess of TIP3P water molecules were deleted to transform the cubic water box into a hexagonal orthorhombic cell of dimension 92.328 Å in the x̂-ŷ plane and 90 Å in the ẑ direction. The length of the system in the ẑ direction provided sufficient solvent padding, greater than 24 Å, to avoid interactions between periodic images. Na and Cl ions were then added to neutralize the system and the bulk salt concentration was set to the physiological concentration of 150 mM. The total number of atoms of the resulting CA hexamer models was 60,000 (Fig. S8c, d). Solvated WT and K25A, K25N, K25E and K25E/E28K CAs were derived using the mutator plugin in VMD.

System minimization and equilibration
The solvated systems were then subjected to minimization in two stages, both using the conjugated gradient algorithm 35 with linear searching 36 . Each stage consisted of 10,000 steps of energy minimization. During the first stage, only water molecules and ions were free to move, while the CA protein and NTP/IP6/BHC molecules were fixed. In the second stage, the backbone atoms of the CA protein were restrained with a force constant of 10.0 kcal mol -1 Å -2 . Convergence of the minimization procedure was confirmed once the variance of the gradient was below 0.1 kcal mol -1 Å -1 . Following minimization, the systems were tempered from 50 K to 310 K in increments of 20 K over 1 ns.
Subsequently, the systems were equilibrated at 310 K for 100,000 steps, while the protein backbone atoms were restrained. Then positional restraints were gradually released at a rate of 1.0 Kcal mol -1 Å -2 per 400 ps from 10.0 Kcal mol -1 Å -2 to 0.0 Kcal mol -1 Å -2 . All MD simulations were performed using NAMD 2.12 37 with the CHARMM36m force field 38 . An internal time step of 2 fs was employed in the multi-step r-RESPA integrator as implemented in NAMD, bonded interactions were evaluated every 2 fs, and electrostatics were updated every 4 fs. Temperature was held constant at 310 K using a Langevin thermostat with a coupling constant of 0.1 ps -1 . Pressure was controlled at 1 bar using a Nose-Hoover Langevin piston barostat with period and decay of 40 ps and 10 ps, respectively. The Shake algorithm was employed to constraint vibrations of all hydrogen atoms. Long range electrostatics was calculated using the particle-mesh-Ewald summation with a grid size of 1 Å and a cutoff for short-range electrostatics interactions of 12 Å.

Gibbs free energy calculations
Progress variables (PVs), akin to reaction coordinates, for all free-energy molecular dynamics calculations were chosen as the location on the Z axis of the center of mass of the small-molecule(s) of interest. The origin of the progress variable was set to the center of mass of Cα atoms in the Nterminal domain of the CA hexamer (Fig. 1c). One-and two-dimensional potentials of mean force (PMFs) along the PVs were calculated using the Hamiltonian Replica-exchange/Umbrella Sampling (HREX/US) method 35 implemented in NAMD 2.12 31,39 . The initial coordinates for the HREX/US windows were derived from constant-velocity steered MD (SMD) simulations in which molecules were pulled along the PV at a rate of 0.1 nm/ns. The center of mass of the small-molecules were positionally resta rained in the US windows with a harmonic force constant of 2.5 Kcal mol -1 Å -2 using Colvars 40 . The width of all US windows was set to 0.75 Å; except for single nucleotide HREX simulations (simulation #1-8) in which the window width was set to 1.0 Å. The number of US windows in the present HREX simulations were chosen so that the nucleotide translocation from capsid exterior to interior through the central cavity was uniformly sampled (Fig. 1c).
Two dimensional (2D) HREX/US simulations were employed to study the cooperativity between small molecules for translocation. For this purpose, two progress variables were employed: PV1 that determined the location of dATP, and PV2 that determined the location of dATP, IP6 or BHC in the CA cavity. The initial configurations for the 2D simulations were derived by pulling dATP using constant velocity SMD at 0.1 nm/ns along PV1, while the center of mass of dATP/IP6/BHC was restrained with a harmonic force constant of 2.5 Kcal mol -1 Å -2 using Colvars (cycle 1, Fig. S10a-c).
A seeding method similar to that reported 41 was employed to generate new simulation windows along PV2 as follows. For each conformation resulting from the SMD simulation, two new seeds were generated by displacing by ±0.75 Å the harmonic restraints for dATP/IP6/BHC along PV2 (cycle 2, Figure S10a-c). Subsequently, 5 ns of HREX/US MD simulation were performed. The last configuration from each replica was then used as the initial seed for a subsequent cycle where the harmonic restraints were displaced by another ±0.75 Å along PV2 (cycle 3, Figure S10a-c). The seeding process was repeated one last time and production 2D HREX/US simulations (production runs, Figure S10a Potentials of mean force were derived from the resulting sampling in each of the US windows using the weighed histogram analysis method (WHAM) 43 . In WHAM, PMF bins are obtained from US/HREX simulation windows. Convergence of the 1D US/HREX calculations was characterized by the changes in the resulting PMF in trajectory increments of 10 ns (Fig. S11). That is, simulations were considered to have converged once the maximum change in one of the PMF bins resulting from adding more simulation data, was less than 1 Kcal mol -1 .
Convergence of the 2D calculations was examined by means of the fluctuations of the average root mean squared error for each PMF bin (RMSE) (Fig. S12a). Each 2D HREX/US simulation was divided into 60 time-slices of width 0.5 ns. Then the PMF was calculated for each slice using WHAM.
Where the sum runs over the total number of PMF bins (Nbins), Gi, k and Gi, reference are the free energies at the i-th PMF bin for the -th and last slice, respectively. Convergence of 2D PMF was established once the RMSE was not larger than 3.0 kcal/mol. The bins before convergence were discarded.
Subsequently, utilizing the converged slices, the 2D PMF surfaces (Fig. 2) and their standard deviations (Fig. S12b, c and d) were computed using WHAM. The statistics of the exchange ratios between neighboring US replicas in HREX simulations are shown in Fig. S11. Overall, the exchange ratios are greater than 20% in 2D HREX simulations, indicating enough sampling efficiencies (Fig.   S13). Capsids were isolated from concentrated virus-containing material using a previously described protocol 46 with modifications. Briefly, ∼40 μL of purified HIV-1 pseudovirions was mixed with an equal amount of 1% Triton-X diluted in 100 mM 3-(N-morpholino) propane sulfonic acid MOPS buffer (pH 7.0) and incubated for 2 min at 4 °C. The mixture was centrifuged at 13,800 × g for 8 min at 4°C.
After removing the supernatant, the pellet was washed twice by adding ∼80 μl of MOPS buffer and centrifuging at 13,800 × g for 8 min at 4°C. The pellet was resuspended in 10 μl of MOPS buffer.

AFM measurements and analysis
AFM measurements and analysis were performed as previously described 11 Capsid stiffness was obtained by the nanoindentation method as previously described 11,24,46 . Briefly, the stiffness value of each capsid was determined by acquiring ~400 force-distance (F-D) curves. To determine the stiffness value of capsids, 20 F-D curves at rate of 20 Hz at each of 24 different points on the capsid surface were determined. To confirm that the capsid remained stable during the entire indentation experiment, we monitored individual measured point stiffness as a histogram (Fig. S14a) and as a function of the measurement count (Fig. S14b). Samples whose point stiffness values decreased consistently during experimentation were discarded, since they underwent irreversible deformation.
The maximum indentation of the sample was 4 nm, which corresponds to a maximum loading force of 0.2 to 1.5 nN. Prior to analysis, each curve was shifted to set the deflection in the noncontact section to zero. The set of force distance curves was then averaged (Fig. S14c). From the slope of the averaged F-D curve, measured stiffness was derived mathematically. The stiffness of the capsid was computed using Hooke's law on the assumption that the experimental system may be modeled as two springs (the capsid and the cantilever) arranged in series. The spring constant of the cantilever was determined during experiment by measuring thermal fluctuation 47 . To reduce the error in the calculated point stiffness, we chose cantilevers such that the measured point stiffness was <70% of the cantilever spring constant. Data analysis was carried out using MATLAB software (The Math Works, Natick, MA).

Infectivity assays and measurement of reverse transcripts and 2-LTR circles
Cells were plated overnight in 24-well or 96-well plates and challenged with virus at equal amounts of p24. Virus infectivity was determined by luciferase production (Promega) after 48h using a

Statistics of infectivity assays, quantitative PCR, and capsid stability assay
Results were analyzed for statistical significance by two-sided student t test (infectivity assays and quantitative PCR) or one-way ANOVA with Tukey's multiple comparisons test (capsid stability assay) with Prism software (GraphPad). A p-value of less than or equal to 0.05 was used to indicate statistical significance.

TEM of viruses
Virus produced from HEK 293T cells was centrifuged to remove cells, filtered through a 0.45 μm Polyethersulfone (PES) syringe filter (Millipore), transferred into 25x89mm polyallomer ultracentrifuge tubes (Beckman) and ultracentrifuged in an Optima XL-100K ultracentrifuge (Beckman Coulter) using an SW28 rotor at 25,500 rpm (117,250 x g) for 2.5 h at 4°C. Following aspiration of supernatant the pellet was fixed in FGP fixative (1.25% formaldehyde, 2.5% glutaraldehyde, and 0.03% picric acid in 0.1 M sodium cacodylate buffer, pH 7.4) for 2 h at room temperature and stored at 4°C. Ultrathin sections (60 nm) were cut on a Reichert Ultracut-S microtome, transferred to copper grids stained with lead citrate, and observed using a JEOL 1200EX microscope with an AMT 2k charge-coupleddevice camera. Images captured at 30,000X magnification were visually inspected to classify viral particles as mature, immature, eccentric, or containing 2 apparent capsids, and over 100 particles were counted per virus preparation. Linear regression analysis of the eccentric particles was performed using GraphPad Prism software.
Preparation of recombinant CA assemblies U-13 C, 15 N labeled CA K25N and K25A mutants were isolated and purified using the protocol reported previously with minor modifications 18 . Pure proteins were assembled as reported previously 11 with minor modifications. First, the proteins were dialyzed overnight against 50 mM MES buffer pH 6.0 containing 300 mM NaCl. The dialyzed proteins were concentrated to 40 mg/mL and diluted to 1:1 volumetric ratio with assembly buffer (50 mM MES pH 6.0, 600 μM NaCl). The final assembly was carried at room temperature by 1:1 volumetric dilution with 1800 μM inositol hexaphosphate (IP6), pH 6.0. CA assemblies were pelleted after the incubation for 1 hour, and stored at 4 °C.
Protein was purified and assembled as previously described by the authors 46 .

Transmission electron microscopy of CA assemblies
After assembly, a small aliquot of each sample was removed and immediately stained with 2% uranyl acetate on copper grids. Images were collected using the TALOS F200C at the Keck Center for        (e) Correlation analysis between loss of solvation molecules and formation of bonds with Arg18 and NTPs. Importantly, solvation of the dNTP/rNTP as it moves toward the interior of the capsid is assisted by several charged or polar residues including K25/30 and E28/29. Figure S5. 2D 13 C-13 C correlation MAS NMR spectra of conical assemblies of CA (CORD, magenta) and CA K25N mutant (RFDR, light blue). The assemblies contain 20 mg/ml of the respective protein in 50 mM MES (pH 6.0), 0.9 μM IP 6 . The similar chemical shifts indicate that K25N mutant is folded and its overall structure is the same as in the WT CA conical assemblies.  The mean fluorescence intensity of CA staining for each imaged virus particle is shown, with the overall mean intensity for the population indicated by the purple bar. Representative results are shown from one experiment. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001.   In each cycle, new US windows were generated to increase the sampling in 2D space. The initial conformations of the first cycles were extracted from SMD simulations pulling dATP through the hexamer central pore. After that, the initial conformations for new US windows (shown as yellow dots) were copied from the last frame of nearest previous US windows (shown as white dots). After 3 cycles, the generated seeds were used in 30 ns production runs. Figure S11. Convergence of the 1D HREX/US simulations. Sequential changes of PMF from every 10 ns HREX/US simulations are shown. After 10 ns, the PMF profiles were used to compute the 1D PMF for NTP translocation.