Tamás
Vermes
ac,
Mark
Kielpinski
b,
Thomas
Henkel
b,
Miquel A.
Pericàs
a,
Esther
Alza
a,
Angelica
Corcuera
c,
Helmut
Buschmann
c,
Thomas
Goldner
c and
Andreas
Urban
*c
aInstitute of Chemical Research of Catalonia (ICIQ), Av. Països Catalans 16, 43007, Tarragona, Spain
bLeibniz-Institut für Photonische Technologien e.V. Albert-Einstein-Str. 9, 07745 Jena, Germany
cAiCuris Anti-infective Cures AG, Friedrich-Ebert-Str. 475, 42117, Wuppertal, Germany. E-mail: andreas.urban@aicuris.com
First published on 17th December 2021
To date, hepatitis B virus (HBV) capsid assembly modulators (CAMs), which target the viral core protein and induce the formation of non-functional viral capsids, have been identified and characterized in microtiter plate-based biochemical or cell-based in vitro assays. In this work, we developed an automated microfluidic screening assay, which uses convection-dominated Taylor–Aris dispersion to generate high-resolution dose–response curves, enabling the measurements of compound EC50 values at very short incubation times. The measurement of early kinetics down to 7.7 seconds in the microfluidic format was utilized to discriminate between the two different classes of CAMs known so far. The CAM (-N), leading to the formation of morphologically normal capsids and the CAM (-A), leading to aberrant HBV capsid structures. CAM-A compounds like BAY 41-4109 and GLS4 showed rapid kinetics, with assembly rates above 80% of the core protein after only a 7 second exposure to the compound, whereas CAM-N compounds like ABI-H0731 and JNJ-56136379 showed significantly slower kinetics. Using our microfluidic system, we characterized two of our in-house screening compounds. Interestingly, one compound showed a CAM-N/A intermediate behavior, which was verified with two standard methods for CAM classification, size exclusion chromatography, and anti-HBc immunofluorescence microscopy. With this proof-of-concept study, we believe that this microfluidic system is a robust primary screening tool for HBV CAM drug discovery, especially for the hit finding and hit-to-lead optimization phases. In addition to EC50 values, this system gives valuable first information about the mode of action of novel CAM screening compounds.
One of the major trends in improving drug discovery processes, especially in the hit-to-lead process, is the utilization of microfluidic systems. This assay format is performed in small enclosed platforms, providing the decisive benefits of miniaturization, such as reduced sample consumption and higher throughput.4 Furthermore, the resulting increased surface-to-volume ratios allows more control over physical variables, such as temperature, pressure, mass transfer, and evaporation, leading to readouts with higher selectivity, increased data resolution, and better reproducibility. These advantages result in higher analytical reliability, and thus, better prediction of drug profiles. Furthermore, compared to microtiter plate-based formats, the continuous operation of microfluidics promotes facile automation, improving reproducibility and reducing the necessity of redundant lab work.5 Finally, the lower detection volume allows improved fluorescence-based readouts, resulting in high volume-to-background ratios similar to confocal setups.6,7
These listed advantages led to the integration of microfluidics into all the different drug discovery stages, from target discovery to drug screening and synthesis.8 In particular, in the hit-to-lead development, multiple enzymatic assays have been successfully integrated.4 Nonetheless, there is still much room for improvement, especially in integrating other druggable viral or microbial non-enzymatic targets.
In line with this, we have developed a microfluidic format for the screening and characterization of novel non-nucleoside hepatitis B virus (HBV) inhibitors. HBV remains a top health priority worldwide with currently over 250 million people chronically infected, and accounting for more than 887000 deaths per year.9 Therefore, new HBV antivirals, which provide a novel mode of action with the potential to cure the infection, are urgently needed.
The 21 kDa HBV core protein (HBc) is an attractive target for novel antiviral therapies as it is essential for the HBV life cycle, participating in the viral DNA release at the cell nucleus, transcription of the viral RNA, encapsidation of the viral polymerase, and the reverse transcription of the RNA to relaxed circular DNA (rcDNA) inside the HBc capsid.10,11 HBc assembles into capsids with a T = 4 or T = 3 symmetry, composed of 120 or 90 HBc dimers, respectively.12 The assembly reaction follows sigmoidal kinetics. It starts with a slow nucleation step, followed by an elongation step of fast addition of further dimers upon reaching a pseudocritical concentration of HBc dimers, and finalized by capsid completion.13,14 The reaction is mainly entropy-driven by weak hydrophobic interactions of −3 to −4 kcal, due to the burial of hydrophobic surface area at the HBc interdimer contacts (approximately 1500 Å), leading to a controlled assembly.15 Furthermore, the assembly is dependent on the ionic strength and pH of the solution, temperature, and HBc concentration. The presence of HBV inhibitors targeting capsid assembly (capsid assembly modulators, CAMs) have also been shown to influence the reaction via allosteric activation of the HBc protein.15,16
CAMs can be categorized into two distinguishable classes, (CAM-A) and (CAM-N).17 CAM-N compounds accelerate the spontaneous self-assembly of nucleocapsids, thereby preventing the encapsidation of the viral genome and resulting in empty, non-infectious capsids.18–20 On the other hand, CAM-A compounds misdirect the assembly. The resulting formation of aberrant structures prone to aggregation and proteasomal degradation then lead to the intracellular clearance of HBc.21–23 There is a high interest in the discovery of novel CAM scaffolds with CAM-A mode of action, since so far, only a CAM-A class compound showed sustained virological response in an in vivo HBV mouse model, with a reduction of both serum HBV DNA levels and HBV antigens (HBeAg and HBsAg).24
A microtiter plate-based homogeneous biochemical assay that measures HBV capsid assembly based on fluorescence quenching of dye-labeled truncated version of the core protein (CP150-BO (BODIPY, boron dipyrromethene)) was previously described for the screening of potential CAMs.10,25 However, the experimental identification of the mode of action of different CAMs was mainly performed using thermodynamic assembly data or by analyzing the morphology of assembly products after long incubation times.26 Elaborate techniques have been utilized to provide this information, such as dynamic light scattering,27 size exclusion chromatography and immunofluorescence assays,18,28 small-angle X-ray scattering,29 time-of-flight mass spectrometry30 and fluorescence correlation spectroscopy.31 Several measurements using microfluidics based on resistive-pulse analysis were evaluated to analyze this reaction using different salt concentrations and several CAMs. Nonetheless, since the mixing was performed off-platform, the first results were obtained earliest after 10 minutes and latest after overnight incubation.32–35
In this study, we describe the development of an automated and versatile microfluidic system with a fluorescence quenching readout for biochemical screening and characterization of up to 96 test compounds. Optimal capillary parameters were evaluated mathematically and experimentally to perform convection-dominated Taylor–Aris dispersion (TAD)-mediated dilution of the respective test compounds. Using TAD, we were able to create, for each screening compound, a theoretically infinite number of test concentrations over 3 logarithmic scales, which gave us the possibility to calculate robust half-maximal effective concentrations (EC50s) by fitting the fluorescence data into high resolution dose–response curves. Furthermore, assembly kinetics were measured for several CAMs at various time points from 400 seconds down to 5 seconds post-mixing, which was not possible using the microtiter plate-based capsid assembly assay. At this early time point of the assembly kinetics, significant differences were observed between the assembly properties of different CAMs, facilitating the differentiation between the two classes of CAM scaffolds described so far. In our microfluidic system, CAM-A compounds like BAY 41-410923 and GLS4
51 showed very fast kinetics, with approximately 80% of the core protein assembled after only a seven-second exposure, whereas CAM-N compounds like ABI-H0731,49 NVR 3-778,50 compound 917,47,48 and JNJ-56136379 showed significantly slower kinetics. From our experiments, the differences in assembly kinetics were not determined by the activity of the compounds (EC50 values) but by the mode of action of the compounds. As a proof-of-concept study, this microfluidic assay system was used to analyze two of our in-house screening compounds, compound A and compound B with hitherto unknown mode of action.
Measuring multiple compounds was achieved using RotAxys360 positioning system from CETONI (Korbussen, Germany) equipped with a neMESYS low pressure 1 syringe and a needle as an interface allowing the transfer of liquids from a 96-well plate to the respective 5 mL pressure glass reservoirs (Fluigent, Jena, Germany) sealed with septa. Initiating the flow was conducted using Flow EZ™ 1000 mbar pressure pumps (Fluigent, Jena, Germany), and the flow rates were controlled using the respective flow rate sensor L, M, or S. The modules were evaluated with respect to their accuracy and reproducibility and, if necessary, adjusted by measuring the pressure flow through characteristics of different biochemically relevant solutions and oils and comparing the measured data with the displayed data (ESI, Fig. S1†). The logarithmic dilution of the compounds was conducted by generating a sharp compound margin using the L-Switch, a 6-port bidirectional valve (Fluigent, Jena, Germany) followed by the injection into a PFA capillary (length 2.5 cm, ID 1 mm, JR-T-4007-M3; ANALYTICS SHOP, Munich, Germany) to create convection-dominated TAD dilution curves. Following dispersion in the capillary, each smoothed pulse passes to the rotate-split-and-recombine-based microfluidic mixer (ROSAR GF-T75) from IPHT (Jena, Germany),36 where it is combined with the target protein and the required buffers. The glass ROSAR GF-T75 passive micromixer had an etching depth, mesh size and thickness of 75, 75, and 250 μM, respectively. The mixing is based on a 9-step sequential multi-lamination, allowing optimal mixing conditions even at the low Reynold numbers of the system (Fig. S2†).36 Bubble traps were installed on pump 1 and 2 before the flow measure unit to prevent air in the system (Fluigent, Jena, Germany). Finally, the solutions were passed onto a long glass chip WC1AB (IPHT, Jena, Germany), in which the fluorescence and reflection was measured using the fluorescence spectrometer OptoReader, equipped with an FITC filter set (excitation of 480 nm and an emission of 520 nm) (Elveflow, Paris, France) (Fig. S2†). With the provided alignment platform and through the miniaturization-mediated reduction of sample volume, a strong 3D focus was achieved in the microreactor. Furthermore, variations in gain and diode power of the OptoReader allowed optimal measurements of dyes from 3.2 nM concentrations up to 32 μM with regards to linear detection range (Fig. S3†), signal to noise (S/N), and signal to background (S/B). The limit of detection and quantification resulted in an S/B (protein signal/50 mM HEPES) of 100 for 1.5 μM CP150C (target protein), and an S/B of 11500 for 32 μM sodium fluorescein (maximal concentration of control dye).
We have written a workflow in the open-source software KNIME (version 4.9) (KNIME AG, Zurich, Switzerland) to allow the evaluation of the fluorescent data obtained from the OptoReader. The workflow operated with text and Microsoft Excel files and cut the complete measurement into separate compounds and replicates. By superimposing the respective three technical replicates, the arithmetic mean values and the respective standard deviations were calculated. Furthermore, the limit of detection and quantification were determined according to Currie's definition, and the Z′-factor was recorded for each measurement individually.37,38 The processed data, containing the flow rates, pump pressures, and fluorescence and reflection measurements, were saved in the form of Microsoft Excel sheets, allowing a general overview and troubleshooting.
![]() | ||
Fig. 2 Chemical structures of the CAMs along with their known mechanism of action (CAM-A and CAM-N). |
The purification of CP150C by serial affinity and size-exclusion chromatography steps and the labeling with BODIPY-FL maleimide was performed as described previously.18,25
The dilution of CAMs for the high-resolution dose–response curves was conducted using a 2.5 cm dilution capillary with an ID of 0.5 mm using the MAT and the system setup described in Data evaluation and automation. Further dilution of the compounds occurred in the connection capillaries between the mixing unit and the WC1AB chip with a total length of 4 cm and 0.75 mm ID. In the final assay, two pumps were used with a flow rate of 12 μl min−1. One pump was dedicated to the CP150-BO (3 μM in 50 mM HEPES), and the other one was dedicated to the compound (64 μM diluted in 50 mM HEPES with 600 mM NaCl), resulting in the final assay conditions of 1.5 μM CP150-BO, 300 mM NaCl and 32 μM compound. Each experiment consisted of 5 minutes for equilibration, 10 minutes for measurement, and 15 minutes for cleaning, resulting in a total run time of 30 minutes (Fig. S4†). Percentage assembly values are normalized to CP150-BO (1.5 μM) dimers in 50 mM HEPES as the negative control, and 2 M NaCl-assembled CP150-BO (1.5 μM) capsids at a residence time of 120 seconds as the positive control (1).
(1) Calculation of CAM mediated percentual CP150-BO assembly based on P150-BO mixed with 50 mM HEPES as negative, and BO mixed with 2 M NaCl as the positive control at 2 minutes post mixing.
![]() | (1) |
As an internal standard, the dilution curves of sodium fluorescein were measured under assay conditions in three replicates. Using three independent measurements with different laser and gain settings of the fluorescence reader, the encoder's high-resolution dilution curve was generated, covering compound concentrations from 3.2 nM up to 32 μM (Fig. S3†). The dilution curve was fitted against the capsid assembly curves of the reference compounds, generating the high-resolution dose–response curves.
Initially, the adaptation to droplet flow seemed a suitable technique to conveniently choose the appropriate time point for measuring different compounds. However, during the evaluation of this technique, we found that CP150-BO assembly was impaired, most likely due to the hydrophobic interface between continuous and disperse phase. Fluorinated PEG-based surfactants in fluorinated oils could not resolve this issue, making the adaptation of this assay to droplets challenging (Fig. S5†).
To determine the effect of NVR 3-778, BAY 41-4109, compound A, and compound B on the localization of HBV capsids and core protein (HBcAg), 3 × 104 HepG2.117 cells were seeded onto sterile cover slips in 12-well plates. For 6 days, cells were cultured without doxycycline and treated with 50-fold EC50 of the compounds. Afterwards, the cells were fixed with 4% paraformaldehyde in phosphate buffered saline (PBS), and permeabilized with 0.2% Triton X-100 in PBS. The cells were then blocked with 5% FBS in PBS. HBcAg was detected with an anti-HBc primary antibody (Abcam ab115992) and an anti-rabbit-Alexa 488 secondary antibody (Invitrogen A-11034). The nuclei were then counterstained with DAPI (Sigma D9542). Samples were mounted onto glass slides with VECTASHIELD antifade medium (Vector Laboratories) and viewed under an inverted fluorescence microscope (Leitz DMRB). Images were taken with the pco.panda 4.2 camera (Visitron Systems) and the VisiView software (Visitron Systems, ver. 4.5.0.5). Images were analyzed using ImageJ (National Institutes of Health, ver. 1.53 g).
While TAD has been utilized in previous flow formats for the generation of high-resolution dilution curves, it was often used in thin capillaries resulting in a complicated evaluation dependent on the radial diffusion of compounds.41–44 In this paper, volume normalized capillaries with different inner diameters were used to experimentally determine the optimal conditions for convection-dominated TAD with high resolution and a short measurement time.
These tasks were achieved at a high Peclet number compared to the length ratio of the capillary, which is accomplished by increasing the ID of volume normalized capillaries (117 μl) (0.5 mm, 0.75 mm, 1 mm) (Fig. 3). The effect of convection-dominated forces were further increased by reducing the length ratio of the system to a PFA tubing with 1/16 × 1.0 mm ID and a length of 2.5 cm (Table S1†). The capillary for the unilateral convection-dominated TAD of the compounds was evaluated both experimentally and mathematically. Seven BODIPY derivatives with molar masses ranging from 291–475 g mol−1 were diluted in isopropanol and dispersed in the capillary at three different flow rates, resulting in the perfect alignment of the normalized dispersion curves (Fig. 5 and Table S1†). Even the dilution curve of compounds with high molar masses, such as fluorescein-dextran (20000 g mol−1) and albumin-FITC (∼620
000 g mol−1) showed a strong alignment with FITC, especially at higher flow rates (Fig. 6).
Next, we conducted a mathematical derivation of the convection-dominated TAD in order to gain an estimation for the consensus between theoretical and experimental values. Mathematically, the convection-mediated dilution can be described in circular tubing with Reynold numbers below 10. Under laminar flow conditions, the pressure drop in the tubing is expressed with the Hagen Poiseuille equation, where Δp equals the pressure difference between inlet and outlet, L the length of the tubing, Q the volumetric flow rate, μ the dynamic viscosity, and R the radius of the tubing (Fig. 4, eqn (2)).
![]() | ||
Fig. 5 Normalized fluorescence spectra (y-axis) of the convection-dominated TAD in isopropanol of the BODIPY derivatives B-01 to B-07 (legend) in a 2.5 cm tubing with an inner diameter of 1 mm followed by a 1.5 cm tubing with an inner diameter 0.75 mm as a function of time in s (x-axis). The dilution curves of all seven compounds aligned perfectly at the three different flow rates evaluated: 9 μl min−1 (A), 13 μl min−1, and (C) 17 μl min−1. TAD without diffusion was calculated using eqn (5.1) for the two short capillaries (ID 1 mm, L 2.5 cm; ID 0.75 mm, L 1.5 cm) and the microfluidic channel of the chip (ID 0.04 mm, L 3 cm). Measurements for each compound were repeated in three technical replicates. |
Eqn (2) is converted toeqn (2.1) by substituting Q with eqn (3).
![]() | (2) |
![]() | (2.1) |
Q = vmeanπR2 | (3) |
The radial velocity distribution vz under laminar flow in the cylindric channel is described by eqn (4), with r as the radial distance from the tube center orthogonally to the flow direction and p as the respective pressures (Fig. 4).
![]() | (4) |
Solving eqn (2.1) and substituting it into eqn (4) results in eqn (4.1) describing the radial velocity distribution dependent on the difference to the center of the tubing.
![]() | (4.1) |
By the variation of time, a stepwise transition between a fluid phase S (sample) and phase B (pure buffer) is created. Regarding the composition of the mixture, this means that the fluid fraction αs varies between a percental ratio between 0 and 1 (eqn (5)).
![]() | (5) |
For the derivation of convection-dominated TAD, we start with the velocity distribution, which yields a particular velocity Z of the fluid at a distance r from the channel center. Please note the differences between flowrates dependent on the radial distance of the center vz(r) and the mean flowrates vmean. Since vz(r) is uniform, vz(r)/dz = 0. Therefore, we can substitute vz(r) by vz(r) = Z/t and solve the equation for r2:
![]() | (4.2) |
![]() | (4.3) |
![]() | (5.1) |
After simplification, the final model for the pure Taylor dispersion without diffusion terms is obtained.
![]() | (6) |
During the simplification, the geometrical parameters are completely lost in the equation. Hence, the model is highly generic and depends only on the average fluid velocity, which can be easily calculated for a given channel from the flow rate and the channel cross-sectional area. The model yields negative values for 2 × t × vmean < Z, which can be substituted by zero.
In most microfluidic setups, the capillary of interest is connected to other capillaries or microreactors with varying inner diameters, leading to a further dispersion of compounds. Eqn (6) can be modified by substituting vmean for Q/A (eqn (3)), to account for the total TAD-mediated dispersion. The final equation was simplified to eqn. (6.1) and the hereby calculated values were compared to the measured BODIPY's dilution curves. As shown in Fig. 4 the experimental values were not only in high consensus with each other, but also showed a strong approximation to the theoretical value calculated using eqn (6.1) (Fig. 4, brown graph).
![]() | (6.1) |
At maximal compound concentrations of 32 μM, only the CAM-A compounds BAY 41-4109 and GLS4 reached an assembly of 100%, while the other CAMs yielded lower assembly, ranging from 91% (compound B) to 58% (ABI-H0731) (Fig. 8). The resulting EC50 values were higher in comparison to the ones derived from microtiter plates, mainly due to the shorter measurement times and the different assay format (Fig. 7 and 8). Compounds with slower assembly kinetics showed significantly higher EC50 values due to the incomplete assembly reaction at this time point compared to the longer incubation times in the microtiter plate format. This was visualized by comparing the resulting dose–response curves, which show an exponential increase instead of a sigmoidal curve (Fig. 7). Measuring accurate EC50 values for “slower” CAMs could be performed at later time points, but the inhibition of further dilution due to TAD has to be prevented for comparative results.
The assay presented here measured the EC50 values of several CAMs at early timepoints in an automated and reproducible manner. Information about the mode of action of the compounds was provided based on their assembly kinetics and is further elucidated in the following section.
Therefore, to distinguish between the modes of action of CAMs in this study, we opted to use the differences in the CAM-mediated CP150-BO assembly kinetics instead of the morphology of the resulting capsids and aggregates. Microfluidic systems fit these requirements perfectly, since they allow measurements after just a few seconds post-mixing, giving new insights about the reaction.
Using our microfluidic platform, we screened the assembly kinetics of six well-characterized reference inhibitors, two of which were categorized into CAM-A and four into CAM-N, as well as the two in-house compounds with unknown mode of action. The goal was to observe clear differences between the assembly kinetics of the known CAM-As compared to the CAM-Ns, and based on this data, to categorize the two in-house compounds. The reaction was performed in flow with 10-fold EC50 concentration of the compound obtained from the microtiter plate-based assay. Percentage assembly was calculated at several time points, ranging from 5 seconds up to 336 seconds (Fig. 9). Both CAM-A compounds led to a rapid assembly (GLS4: 99% assembly; BAY 41-4109: 82%) completed at 10 seconds post-mixing. On the other hand, CAM-N compounds showed variations in their respective kinetics and were categorized into two groups. “Slow” CAM-N led to an assembly of 6–18% at 7.5 seconds, while “fast” CAM-N compounds, such as compound B, NVR 3-778 and compound 917, were able to achieve assembly values between 44% to 61%. At the endpoint of 336 seconds, all the compounds reached an assembly of at least 88% with the highest values obtained by GLS4 (109%), BAY 41-4109 (100%), compound B (98%), JNJ-56136379 (92%) and compound A (94%). These tendencies showed that compounds with a CAM-A mode of action indeed have faster assembly kinetics, reaching slightly higher endpoint values in flow.
To further study the impact of CAM concentration on capsid assembly kinetics and to facilitate the characterization of the mode of action of the two in-house compounds, measurements were conducted with 1, 10, and 100 μM CAMs. By using 1 μM CAM and a residence time of 2 minutes, CAM-A compounds and compound B reached approximately 40% assembly, while other compounds reached only around 29%. Increasing the concentration to 10 μM, while maintaining the same residence time of 2 minutes, led to a nearly complete assembly of approximately 95% for the CAM-As and 77% for compound B. Contrarily, the other compounds only reached maximal values ranging from 58% to 37%. Interestingly, significant differences in assembly between CAM-A (BAY 41-4109: 75%, GLS4: 89%) and the other compounds (31% to 2%) could be observed by reducing the residence time to 7.7 seconds, while maintaining a 10 μM concentration. Among the other compounds, compound B, NVR 3-778, and compound 917 achieved the highest assembly with approximately 27%. Finally, the CAM concentration was further increased to 100 μM with a residence time of 2 minutes, at which all the compounds reached an assembly of at least 75%. CAM-As again yielded the highest assembly rates at around 100%, followed by compound B at around 91%, and JNJ-56136379 at 87%. Measurement of the effect of 100 μM CAM at a shorter residence time of 7.7 s again displayed the difference between CAM-A (BAY 41-4109: 98%, GLS4: 92%) and CAM-N (around 65%) (Fig. 10). In summary, the reactions with 10 μM (or 32 μM as used in the EC50 measurements) CAMs were most appropriate to elucidate the difference between the two compound classes by eliminating oversaturation and, at the same time, maintaining a good signal to background ratios. Therefore, to accurately distinguish the CAMs'mode of action, we suggest using a 10 μM solution of the respective CAMs with 7.7 and 51 s as residence times, which led to the most apparent differentiation of the CAM's mode of action. At 7.7 s, CAM-A reached assembly values around 50% higher than CAM-N, and at 51 s, the assembly reaction for CAM-A was mostly completed contrary to CAM-N.
Interestingly, while compound A behaved more like a classical “fast” CAM-N compound, compound B showed an in-between kinetic behavior at 1 μM and 10 μM concentrations, making a clear classification for this compound challenging. The difference of compound B to CAM-N was especially eye-catching at a concentration of 10 μM and 51 s at which both the other fast CAM-Ns compound 917 and NRV-33-778 showed a lower CP150-BO assembly rate. Consequently, size exclusion chromatography and immunofluorescence assays were conducted to further elucidate their modes of action.
In addition to the biochemical analysis, the mode of action of compound A and B were analyzed in an antiviral assay using the immunostaining of HBcAg as a readout. In the HepG2.117 cell line, the expression of HBV is regulated by a tetracycline-responsive minimal CMV promoter. As shown in Fig. 12, HBV replication was repressed in cells treated with doxycycline, a second-generation tetracycline, whereas in DMSO-treated cells, HBcAg was detected in both the nucleus and the cytoplasm. As previously described, the HBcAg staining in cells treated with the CAM-N reference NVR 3-778 was almost exclusively found in the cytoplasm, whereas BAY 41-4109 which is known to follow the CAM-A mode of action, induced punctuated aggregates in or near the nucleus.18 In line with the observation from the SEC, treatment with compound A induced a CAM-N phenotype, whereas treatment with compound B had the tendency to induce aggregated dot-like structures exclusively seen after CAM-A treatment. However, the number and size of these aggregates were significantly lower than in BAY 41-4109-treated cells.
From these results, we classify compound A as having a classical CAM-N mode of action. On the other hand, this study showed evidence in all three experiments, namely the kinetics analysis using the microfluidic system, the SEC, and the immunostaining of HBcAg in HepG2.117 cells, that compound B displays a profile which is similar but weaker to that of the CAM-A reference compound.
Hence, the aggressive assembly displayed by CAM-A compounds is a possible identification factor for the mode of action of different CAMs. Evidently, the microfluidics platform offers an excellent opportunity for further study of capsid kinetics, as well as the differentiation of the modes of action of CAMs in a rapid and reliable manner.
Interestingly, the in-house compound B showed faster and higher maximal assemblies compared to other CAM-Ns, particularly at a concentration of 10 μM and residence times of 7.7 and 51 s but was not quite reaching the levels of the other CAM-As. This compound showcased similar in-between behavior in the immunofluorescence assay and SEC as well. Therefore, we conclude that aside from the morphology of the assembled products and the localization in cells, the aggressive assembly displayed by CAM-A compounds is a possible identification factor for the mode of action of different CAMs. In conclusion, the microfluidic system offers a valuable opportunity to rapidly screen for novel CAMs. It provides information about the activity and kinetics of compounds simultaneously, also giving a first hint about the mode of action of novel CAMs. Furthermore, with the evaluation of the kinetic data at three independent time points (7.7, 51, and 122 s), the differentiation of mode of action per compound is possible within 4 min, as opposed to the analytical size exclusion chromatography requiring at least 3 h for sample preparation and evaluation. Finally, the possibility of measuring reactions at a couple of seconds provides an excellent opportunity to study capsid kinetics and further novel viral or bacterial targets.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1ay01227d |
This journal is © The Royal Society of Chemistry 2022 |