A simple, high-throughput stabilization assay to test HIV-1 uncoating inhibitors

Shortly after entering the cell, HIV-1 copies its genomic RNA into double-stranded DNA in a process known as reverse transcription. This process starts inside a core consisting of an enclosed lattice of capsid proteins that protect the viral RNA from cytosolic sensors and degradation pathways. To accomplish reverse transcription and integrate cDNA into the host cell genome, the capsid shell needs to be disassembled, or uncoated. Premature or delayed uncoating attenuates reverse transcription and blocks HIV-1 infectivity. Small molecules that bind to the capsid lattice of the HIV-1 core and either destabilize or stabilize its structure could thus function as effective HIV-1 inhibitors. To screen for such compounds, we modified our recently developed FAITH assay to allow direct assessment of the stability of in vitro preassembled HIV-1 capsid-nucleocapsid (CANC) tubular particles. This new assay is a high-throughput fluorescence method based on measuring the amount of nucleic acid released from CANC complexes under disassembly conditions. The amount of disassembled CANC particles and released nucleic acid is proportional to the fluorescence signal, from which the relative percentage of CANC stability can be calculated. We consider our assay a potentially powerful tool for in vitro screening for compounds that alter HIV disassembly.


Results
Optimization of HIV-1 CANC disassembly. To establish an assay for quantification of HIV-1 CANC array destabilization, we sought to determine conditions required for transition of compact in vitro assembled CANC-tqON tubes into oligo-or monomeric protein subunits with tqON accessible for Exonuclease I degradation. HIV-1 CANC protein, consisting of full-length CA, SP1 and NC sequence (Fig. 1a) was purified by combination of cation-exchange and gel filtration chromatography (Fig. 1b,c) as described in Material and methods and elsewhere [36][37][38] . The tubular structures from the purified HIV-1 CANC protein were first assembled as previously described 36 in the presence of tqON under the following reaction conditions: 18 µM CANC, 2.19 µM tqON, 50 mM Tris, pH 8, 340 mM NaCl and 1 µM ZnCl 2 . To destabilize these in vitro assembled tubular structures (shown in Fig. 1d, panel B), the composition of the buffer was modified by addition of one or more of the following components: NaCl (340 mM to 1 M), the detergent NP-40 (0.5-5%), and the reducing agents DTT (1-10 mM) or β-ME (0.1-1%). Following overnight incubation at room temperature with gentle shaking (480 rpm), the samples were negatively stained and analyzed by TEM (Fig. 1d, panels C-H). None of these conditions led to significant destabilization of the assembled CANC structures. Next, we attempted to destabilize the CANC arrays either by lowering the pH of the final reaction buffer (in the range from pH 5-7) (Fig. 1d, panels I-K) or by lowering both the pH and ionic strength to a final pH of 7 and NaCl concentration of 170 mM (Fig. 1d, panel L). Both conditions led to destabilization of the CANC arrays and caused partial or nearly complete disassembly of the CANC tubular structures (Fig. 1d, panels J-L).
Next, we analyzed whether the conditions triggering CANC disassembly were also suitable for Exonuclease I activity, which is an essential part of the assay. The lower ionic strength (170 mM NaCl) did not influence tqON degradation, and the activity of the exonuclease was the same at 340 mM NaCl as at 170 mM NaCl (Fig. 2a). However, exonuclease activity was strongly inhibited at pH values under 7 (Fig. 2b). To maintain the Exonuclease I activity, we adjusted the pH of the assembly buffer to 7 and the ionic strength to 340 mM NaCl. Under these conditions, CANC assembled into stable, regularly ordered uniform tubular structures (Fig. 2c). Destabilization was triggered by dilution of these assembled CANC tubes with an equal volume of 50 mM Tris, pH 7, 1 µM ZnCl 2 to yield the following final composition of disassembly buffer: 50 mM Tris, pH 7, 170 mM NaCl, 1 µM ZnCl 2 (Fig. 2d).
Verification of the DITH principle. The CA domain within the CANC protein is ordered in a mature-like arrangement and therefore can serve as a surrogate for the HIV-1 core, despite the presence of the NC domain 34,35 . As our DITH assay is dependent on the accessibility of tqON to Exonuclease I, we analyzed whether the change in ionic strength used to trigger CANC disassembly affects the interaction of the oligonucleotide with www.nature.com/scientificreports www.nature.com/scientificreports/ stronger (Kd = 0.7 µM) at the low ionic strength of the disassembly buffer (50 mM Tris, pH 7, 170 mM NaCl, 1 µM ZnCl 2 ) than in the assembly buffer (50 mM Tris, pH 7, 340 mM NaCl, 1 µM ZnCl 2 ) (Kd = 4.5 µM). This result confirmed that the disassembly conditions do not facilitate release or dissociation of tqON from NC. On the contrary, the binding of tqON to NC was even stronger under low ionic strength conditions, and thus disassembly was achieved mainly by affecting the CA-CA interactions. This destabilization of the CA lattice subsequently led to exposure of tqON, making it available for exonuclease cleavage. Despite the different binding affinities at different ionic strengths, the result of tqON exonuclease degradation remained unchanged. This confirms that merely binding to non-assembled CANC does not protect tqON from degradation and suggests the robustness of our method to measure the proportion of assembled versus non-assembled CANC.
To quantify the HIV-1 CANC tube disassembly, we compared the fluorescence of the fluorophore (FAM) released from the tqON-CANC tubes in disassembly buffer with that released from the tubes in assembly buffer upon addition of Exonuclease I. As the destabilization is triggered by dilution of the assembly buffer, we wanted to rule out the possibility that disassembly resulted from merely lowering the CANC protein concentration. Following the in vitro assembly of tqON-CANC tubular structures in the pH 7 buffer, the sample was divided in two aliquots that were diluted into different buffers to yield the same final protein and tqON concentrations. One aliquot was diluted into the assembly buffer, while the second one was diluted into 50 mM Tris, pH 7, 1 µM ZnCl 2 . Both samples were then incubated overnight with gentle agitation at laboratory temperature. The next day, Exonuclease I was added to both samples and an increase in fluorescence was monitored for 120 min (Fig. 3a). The samples were negatively stained and analyzed by TEM (Fig. 3b,c). The fluorescence increase confirmed that the preassembled CANC particles incubated at the lower ionic strength of the disassembly buffer were more prone to tqON degradation than those incubated in the assembly buffer (Fig. 3a, red and black lines, respectively). In accord with this, we observed typical tubular structures in the sample diluted into assembly buffer (Fig. 3b), while fully or partially disassembled tubes were observed in the sample diluted into disassembly buffer (Fig. 3c). The relative stabilization of wt CANC (Δ1) was then calculated from the difference between the fluorescence of released and degraded tqON at 90 min in the disassembly and assembly reactions (Fig. 3a). www.nature.com/scientificreports www.nature.com/scientificreports/ Validation of DITH using HIV-1 CA mutants with altered core stability. To validate DITH, we used two HIV-1 CA mutants with distinct core stability: the unstable A42D mutant and the hyperstable E45A mutant 30,39 (Figs. 1a, 4a). Using EMILI mutagenesis 40 we introduced the point mutations in CANC pET22b vector, and following their expression (Fig. 4b), we optimized purification of both CANC mutant proteins (Fig. 4c,d) using the same work-flow described for wild type. Next, we tested the assembly efficiency of these mutants using TEM and the FAITH assay (Fig. 4e,f). We confirmed that the wild-type (wt) and E45A CANC proteins assembled efficiently into long tubular structures. In contrast, the ability of the A42D mutant to assemble was impaired. Therefore, the DITH assay was performed only with the hyperstable E45A CANC mutant. Preassembled wt and E45A CANC particles were incubated overnight in the assembly and disassembly buffers, and fluorescence was measured after Exonuclease I addition (Fig. 4g). Fluorescence released from the wt or E45A CANC in assembly buffer is shown in black, while fluorescence released from the wt and E45A in disassembly buffer is shown in red and yellow (Fig. 4g), respectively. The relative stabilization for wt and E45A mutant tubes was calculated as the difference between the fluorescence of degraded tqON at 90 min in the disassembly and assembly reactions for the wt (Δ1 Fig. 4g) and E45A mutant (Δ2 Fig. 4g). The relative percent of stabilization was then determined using simple calculation: relative % of stability = 100 * Δ2/Δ1. The calculated relative percentage values of stability were then plotted   (Fig. S2a), HIV-1 CANC proteins in chemiluminescent mode (Fig. S2b). (e) TEM analysis of negatively stained wt HIV-1 CANC and the A42D and E45A CA mutants assembled in the presence of tqON in assembly buffer. (f) FAITH quantification of wt HIV-1 CANC and CA mutant assembly in the presence of tqON. (g) Fluorescence emission curves demonstrating the kinetics of tqON release from preassembled wt and E45A CANC particles incubated in assembly buffer (black curve) and disassembly buffer (red and yellow curves, respectively). The relative stabilization was calculated as the difference between the relative fluorescence of tqON at 90 min in the disassembly and assembly reactions according the calculation: relative percent of stabilization = 100 * Δ2/Δ1. (h) DITH quantification of the relative stability of preassembled wt and E45A CANC particles incubated in the disassembly buffer measured and calculated as described in (g). Relative stability of wild type in disassembly buffer was considered as 0%. (2019) 9:17076 | https://doi.org/10.1038/s41598-019-53483-w www.nature.com/scientificreports www.nature.com/scientificreports/ and compared with the wild type in the disassembly buffer whose stability was considered as 0% (Fig. 4h). As expected, E45A CANC mutation increased stability of CANC tubes in the disassembly buffer by about 50% compared to the wild type (Fig. 4h).
Evaluation of DITH with the capsid inhibitor PF74. As the main purpose of our in vitro disassembly method is to screen for compounds and small molecules that alter the stability of the hexameric array of CANC tubular structures, we used the inhibitor PF74 for further DITH validation. PF74 is a small molecule that binds to the capsid core and inhibits HIV-1 at an early stage of infection 41 . PF74 preferentially binds to the hexameric lattice of the capsid core, affecting both intra-and inter-CA hexamer interactions 14,42,43 . This alteration subsequently results in enhanced stability of the HIV-1 capsid core, which in turn affects reverse transcription and viral infectivity.
To evaluate whether optimized conditions for destabilization of the CA hexameric lattice of preassembled CANC tubes would allow measurement of the stabilizing effect of bound small molecules, we used DITH to quantify the stabilization effect of PF74 binding to CANC. Various concentrations (4 to 24 µM) of PF74 were added to the assembled CANC tubes following 1 h incubation. The samples were then divided into two aliquots and diluted into either assembly or disassembly buffer. Following overnight incubation, Exonuclease I was added to the samples and the kinetics of fluorescence release were measured (Figs. 5a, S3). Fluorescence released from the wt assemblies in assembly buffer is shown in black (Fig. 5a), while fluorescence released from the wt in disassembly buffer is shown in red (Fig. 5b). Fluorescence released from the PF74 treated wt in disassembly buffer is shown in yellow (4 µM PF74), pink (6 µM PF74), violet (8 µM PF74), salmon (10 µM PF74), brown (16 µM PF74), green (18 µM PF74) and blue (24 µM PF74). The proportion of assembled and degraded wt tubes (Δ1) was calculated as the difference between the fluorescence of tqON at 90 min in the disassembly and assembly reactions (Fig. 5a). This fluorescence range was then used to calculate the relative stabilization of CANC assemblies by PF74 as the difference between the fluorescence of released tqON from PF74 treated tubes and the wt upon 90 min incubation in the disassembly buffer (e.g. the difference for 10 µM PF74 is shown as Δ2 in Fig. 5a). The relative percent of stabilization was then calculated as follow: relative percentage of stability = 100 * Δ2/Δ1. The relative percentage of stability of wt CANC assemblies for each tested PF74 concentration was then plotted and compared with the wild type in the disassembly buffer whose stability was considered as 0% (Fig. 5b). Based on the DITH results, we observed a PF74 induced increase of CANC stability in comparison to untreated sample in the whole range of tested PF74 concentrations (Fig. 5b). Lower stabilization was observed at PF74 concentrations ranging from 16 to 18 µM, where we consistently observed a slight decrease in stabilization capacity of PF74, which was overcome at higher PF74 concentration (24 µM) (Fig. 5b). Stabilization effects of E45A mutation (Fig. 5b, grey) was similar to that of PF74 at concentrations ranging from 8-10 µM (Fig. 5b violet and salmon, respectively).
Following DITH, sample treated with 10 µM PF74 and untreated CANC sample were negatively stained and analyzed by using TEM (Fig. 5c,d). The TEM analysis clearly confirmed that the addition of PF74 stabilized the CANC tubular structures and prevented their degradation during incubation in disassembly buffer (Fig. 5d). In contrast, in the absence of PF74, the capsid lattice of CANC tubular structures disintegrated (Fig. 5c).

Effect of inositol hexakisphosphate (IP6) on HIV-1 CANC stability.
In addition to synthetic organic compounds such as PF74, small cellular polyanions such as IP6 can be incorporated into assembling immature HIV-1 particles and enhance core stability 26 . We employed the DITH assay to test the stabilization effect of IP6 on assembled mature-like CANC tubular structures. The tubular particles were preassembled from HIV-1 CANC protein in assembly buffer in the presence of tqON and various concentrations of IP6: 0.7 µM (green). Following 3 h incubation, the samples were divided into two aliquots and diluted into either assembly buffer or disassembly buffer. After overnight incubation, Exonuclease I was added to all the samples and the kinetics of fluorescence release were measured (Fig. 6a). Fluorescence released from the wt assemblies in assembly and disassembly buffers is shown in black and red, respectively (Fig. 6a). Fluorescence released from the wt in disassembly buffer at various IP6 concentrations is shown in green (0.7 µM), yellow (1.8 µM), violet (2.3 µM) and blue (3 µM) (Fig. 6a).
Relative percentage of stabilization effect of IP6 was calculated as described above for PF74, plotted and compared with the wild type whose stability in the disassembly buffer was considered as 0% (Fig. 6b). The greatest stabilizing effect (by 1.5-fold) was observed for CANC:IP6 ratios ranging from 1:10 to 1:8, corresponding to IP6 concentrations of 1.8 µM to 2.3 µM, respectively. A comparison of relative stability of E45A mutant (Fig. 6b, grey) and wt CANC in the presence of 10 µM PF74 (Fig. 6b, salmon) in the disassembly buffer revealed similar level of stabilization as that observed for 1.8 µM IP6 (Fig. 6b, yellow).

Discussion
Here, we describe development of a new high-throughput fluorescence method that enables direct measurement of the kinetics of disassembly of HIV-1 CANC tubular particles formed in vitro. The method is based on measurement of fluorescence generated upon degradation of an oligonucleotide that becomes accessible upon disassembly of CANC tubular structures. The oligonucleotide is a dually labelled, TaqMan-based, 40-base-pair, single-stranded DNA called tqON. When tqON is incorporated inside the preassembled CANC tubular www.nature.com/scientificreports www.nature.com/scientificreports/ structures, it is protected from degradation by Exonuclease I. Uncoating or disassembly of the hexameric CA lattice of CANC complexes is triggered by dilution of the preassembled CANC tubular structures into disassembly buffer. The disintegration of the tubular structures renders the released tqON vulnerable to degradation by Exonuclease I. Exonuclease cleavage leads to separation of the fluorescence label (FAM) from the quencher (BHQ). The amount of disassembled CANC particles is therefore proportional to the FAM-derived fluorescence signal, from which the relative percentage of CANC destabilization or stabilization can be calculated.
HIV-1 CANC can assemble in vitro into narrow tubular structures 38,44 . Despite the presence of the NC domain, the CA domain of CANC adopts a mature-like arrangement 35 . Therefore, this in vitro system can be used experimentally as a surrogate for the HIV-1 core 34 , as demonstrated with the capsid stabilization assay 33,34 . In this assay, CANC protein is assembled in vitro in the presence of a 25-nucleotide DNA strand (TG 25 ). The in vitro preassembled CANC tubular structures are incubated in a destabilization buffer to instigate disassembly, and the material is ultracentrifuged through a 70% sucrose cushion. The level of CANC stability is determined by comparing the CA content in the pelleted fraction with the input CA amount using Western blot 33 . Our new DITH method uses the same HIV-1 CANC protein, but differs in its usage of dually labelled FAM-BHQ oligonucleotide. This arrangement is advantageous because the amount of disassembled CANC particles is proportional to the FAM-derived fluorescence signal. Thus, DITH allows direct real-time determination of stabilization or destabilization without the need for ultracentrifugation and Western blotting steps. As DITH can be carried out entirely in a 96-well plate, it is optimal for small molecule screening or testing of CA-binding factors or proteins.
During optimization of disassembly of the CANC tubular structure, we assessed various conditions that led to the decomposition of CANC complexes and exposure of tqON to Exonuclease I cleavage. The most efficient disassembly was achieved by lowering the pH below pH 6. However, this non-physiological acidity could affect the overall or local charge of the CA and NC domains or the compounds tested, and thus alter their interactions. Moreover, we found that low pH conditions were not suitable for the assay, as the exonuclease was inhibited. However, we found that CANC tubular structures assembled at pH 7 were more prone to disassembly that those assembled at pH 8. Therefore, we altered conditions for the in vitro assembly of CANC tubular structures and preassembled the particles at pH 7. Disassembly was then triggered by simply diluting the preassembled CANC particles into the same pH 7 buffer with a lower ionic strength to yield a final NaCl concentration of 170 mM. The change in ionic strength had no effect on Exonuclease I activity.
We used several means to validate and evaluate the DITH method. First, we prepared CANC protein with E45A mutation in the CA domain, which has been shown to form a hyperstable core 30,39 . In accord with previous data, our results showed that the E45A mutant exhibited 50% higher stability than the wild-type. We also used DITH to assess the core stabilization effect of PF74, a small molecule that binds the CA hexameric lattice 14,[41][42][43] . Results of previous studies on effects of PF74 are conflicting. In vitro experiments demonstrated completely converse impact of PF74 on HIV-1 core: it either accelerated HIV-1 core disassembly by reducing its stability [41][42][43] or it increased the stability of HIV-1 CA cores thus slowing down the disassembly 42,45 . Recently, researchers using atomic force microscopy (AFM) measured stiffness of capsids assembled in vitro and HIV-1 isolated cores. They showed that binding of PF74 increased the stability of both structures in a concentration-dependent manner 46 . We obtained similar results by using DITH, finding that the CANC stability was enhanced in the presence of PF74. However, even though we observed a concentration-dependent effect of PF74 on HIV-1 CANC stabilization in the inhibitor concentration ranging from 4-10 µM, we consistently observed a slight drop in PF74 stabilizing effect at concentrations within 16-18 µM. This effect disappeared at PF74 concentration 24 µM, and the same level of stabilization, observed for 10 µM was re-established. Due to DMSO concentration limitation (maximum in the reaction is 1%), we could not increase the PF74 concentration above 24 µM, which prevented to determine further trend of stabilization. However, based on the data published by Rankovic et al., who showed the CA assemblies stability increases in the PF74 concentration-dependent manner, but not above 20 µM PF 74 46 , we can hypothesize that there would not be any significant increment in CANC stability at PF74 concentration higher than 24 µM. The reason for the observed drop in PF74 stabilization at 16-18 µM however, remains unclear. Nevertheless, our data on the comparison of the stabilization effect of PF74 and E45A mutation are in concert with those showed by Rankovic et al. 46 . Using AFM they showed that the stiffness, observed upon binding of PF74 at 10 µM concentration, of in vitro assembled capsids achieved the level similar to that of the E45A mutant 46 . In our assay, similar level of stabilization observed for E45A mutant was achieved at PF74 concentration of 8 µM.
Our previous work indicated that PF74 does not affect the in vitro assembly of either mature-like or immature-like structures 36 , in accord with a published crystal structure that reveals a binding pocket between helices 3 and 4 of HIV-1 NTD-CA 41 . Other in vitro studies have suggested that PF74 binding significantly increases the rate of CA assembly 41,42 . We observed previously that PF74 affects the structure of immature-like particles, which form mature-like short rods instead of spherical particles 36 , suggesting that the binding of PF74 may shift the equilibrium towards mature-like assembly 36 . This is in agreement with the results on core stabilization described in the present work.
Additionally, we used DITH to evaluate the effect of small cellular polyanion IP6, which can stabilize immature HIV-1 capsid lattices by binding to two lysine residues in CA and SP1 regions of Gag hexamers 26 . The arginine residue at the N-terminus of CA, which becomes unmasked by proteolytic cleavage of Gag during maturation can serve as an alternative IP6 binding site 26,27 . However, quantitative analysis of stabilizing effect of IP6 on mature CA lattice was not published. Here, using DITH, we tested concentration-dependent effect of IP6 on stabilization of CANC mature-like structures and we identified the optimal ratio to achieve the highest stability. We found that 1.8 µM IP6 provides similar 50% increment of CANC tubular structures stability as that shown for E45A mutant, 8-10 µM PF74.
These data clearly show that DITH is a reliable method that can be used for initial screening of compounds that bind to and stabilize the CA lattice and have the potential to block HIV-1 infectivity. The method might also be applicable to compare the effects of a newly designed series of inhibitors that potentially form more extensive contacts with CA, as recently proposed by molecular docking and structural analysis for molecules GS-CA1 and GS-6207 47 .

Fast assembly inhibitor test for HIV (FAITH). FAITH was used for in vitro testing of inhibitors of
assembly of HIV-1 particles 36 . A 60 µg portion of HIV-1 CANC was pre-incubated with inhibitor for 1 h on ice. Following incubation with 3 µg of dually labelled oligonucleotide (tqON) in assembly buffer (50 mM Tris, pH 8.0, 1 µM ZnCl 2 , 340 mM NaCl) in a 96-well for 3 h at room temperature, Exonuclease I (ExoI) with Mg 2+ ions was added to the mixture. The fluorescence of the fluorophore released from degraded tqON present in the solution was measured using a Tecan M200Pro plate reader.

Stabilization fast assembly inhibitor test for HIV (DITH).
Small molecules were added at a final concentration of 18 µM to HIV-1 CANC tubular structures in a 96-well plate. The mixture was incubated for 3 h at room temperature. Next, 100 µl of disassembly buffer (50 mM Tris, pH 7.0, 1 µM ZnCl 2 ) was added to the each sample in the plate, and incubation under moderate agitation continued overnight at room temperature. ExoI and Mg 2+ ions were added 16 h later, and fluorescence was measured using a Tecan M200Pro plate reader. The relative stabilization for wt and E45A mutant, PF74 or IP6 treated CANC assemblies was calculated as the difference between the fluorescence of degraded tqON at 90 min in disassembly and assembly reactions for the wt (Δ1) and E45A mutant, PF74 and IP6 treated CANC assemblies (Δ2). The relative percent of stabilization was then determined using following formula: relative percentage of stability = 100 * Δ2/Δ1. The calculated relative percentage values of stability were then plotted and compared with the wild type in disassembly buffer whose stability was considered as 0% fluorescence anisotropy binding assay. The fluorescence anisotropy binding assay was carried out as previously described 49 . Briefly, a solution of 10 nM 5′-FAM-labeled ssDNA (5′-GATTAAAAGTGAAAGTAAACTC-3′) was titrated with HIV-1 CANC to an overall protein concentration ranging from 8 to 6,000 nM. Changes in 5′-FAM ON DNA fluorescence anisotropy induced by changes in the CANC protein concentration and the equilibrium dissociation constant (Kd) were determined as previously described 50 .
Synthesis of PF74 derivatives. The PF74 molecule consists of three synthons: 2-(2-methyl-1H-indol-3-yl) acetic acid, L-phenylalanine and N-methylaniline. The compound was purified by flash column chromatography on silica gel and analytically characterized by NMR and mass spectrometry. The thin-layer chromatography (TLC) was performed using (TLC Silica gel 60 F254, Merck). TLC detection (254 nm). The column chromatography was performed using silica gel (100-160 μm, Merck) and glass column with 2 cm in diameter and 12 cm in height. NMR spectra were acquired using device from Agilent Technologies with working frequency 400 MHz, chemical shift in ppm (δ) and J-constants in Hz. The mass spectra were measured using LC-MS TSQ Quantum Access Max (Thermo). The products were identified via 1 H NMR, 13C NMR, COSY, HMQC and MS (Supplementary Table 1).
All data generated or analyzed during this study are included in this published article. Data that are not shown are available from the corresponding author on reasonable request.