Disassembly of Single Virus Capsids Monitored in Real Time with Multi-Cycle Resistive-Pulse Sensing

Abstract

Virus assembly and disassembly are critical steps in the virus lifecycle; however, virus disassembly is much less well understood than assembly. For hepatitis B virus (HBV) capsids, disassembly of the virus capsid in the presence of guanidine hydrochloride (GuHCl) exhibits strong hysteresis that requires additional chemical energy to initiate disassembly and disrupt the capsid structure. To study disassembly of HBV capsids, we mixed T = 4 HBV capsids with 1.0 to 3.0 M GuHCl, monitored the reaction over time by randomly selecting particles, and measured their size with resistive-pulse sensing. Particles were cycled forward and backward multiple times to increase the observation time and likelihood of observing a disassembly event. The four-pore device used for resistive-pulse sensing produces four current pulses for each particle during translocation that improves tracking and identification of single particles and increases the precision of the particle-size measurements when the pulses are averaged. We studied disassembly at GuHCl concentrations below and above denaturing conditions of the dimer, the fundamental unit of HBV capsid assembly. As expected, capsids showed little disassembly at low GuHCl concentrations (e.g., 1.0 M GuHCl), whereas at higher GuHCl concentrations (≥ 1.5 M), capsids exhibited disassembly, sometimes as a complex series of events. In all cases, disassembly was an accelerating process, where capsids catastrophically disassembled within a few 100 ms of reaching critical stability; disassembly rates reached tens of dimers per second just before capsids fell apart. Some disassembly events exhibited metastable intermediates that appeared to lose one or more trimers of dimers in a stepwise fashion.

Graphical Abstract

Self-assembly and disassembly of the virus protein capsid are critical steps in the virus lifecycle.¹ Understanding these steps provides a general approach to developing antiviral drugs.^2–3 Virus capsid disassembly⁴ is the less understood reaction, but for many viruses is biologically critical to the release of the viral genome.⁵ HBV capsid disassembly in the presence of chaotropes, e.g., guanidine hydrochloride (GuHCl) and urea, has varied outcomes.⁴ HBV in GuHCl shows pronounced hysteresis where assembly is observed at lower GuHCl concentrations, and disassembly requires higher GuHCl concentrations.⁴ Experiments conducted in urea show similar behavior.⁴ Of note, a disulfide crosslink within the dimer appears to slow down assembly and destabilize capsids, suggesting dimer flexibility facilitates assembly and is consistent with the highly dynamic state of the capsid.^6–7

We propose two explanations for hysteresis: allosteric regulation of the assembly reaction and post-assembly maturation. Allostery can lead to hysteresis because protein subunits (dimers for HBV) adopt an assembly favorable conformation, which lowers the net energy needed for association, whereas disassembly of those same subunits from a capsid requires breaking multiple interactions simultaneously to be removed. Evidence of conformation change between free dimer and dimer in the context of capsid has been seen in the closely related Woodchuck Hepatitis Virus.⁸ Consequently, disassembly requires more energy to occur. In a virus where there is a post-assembly maturation transition, observation of assembly is based on assembly of the immature form whereas disassembly, unless the immature form is trapped, would be based on the mature form. Post-assembly maturation is common in bacteriophages.⁹ These two thermodynamic mechanisms are not mutually exclusive. Hysteresis to disassembly may also point to a kinetic barrier that must be overcome and results from free dimers reassociating with incomplete capsids.⁴ The hysteresis in HBV is consistent with the observed paucity of subunit exchange, compared to the extensive exchange predicted from a capsid at thermodynamic equilibrium.^10–11

The impact of chaotropes on protein structure and association among proteins varies with type of chaotrope and its concentration. Chaotropes disrupt hydrogen bonding, thereby weakening the hydrophobic effect, and unfold native protein structure at high concentrations.¹² GuHCl is believed to affect interaction of solutes with water and interact with peptide bonds to modulate protein stability.¹³ Several protein folding and unfolding models which employ a denaturant under sub-denaturing concentrations reveal changes in the local chemistry of the protein.^14–17 How those changes impact the global structure of the supramolecular complex, e.g., the virus capsid, are unknown.

To investigate whether capsid disassembly is a piecewise process or an instantaneous, complete dissociation, we coupled resistive-pulse sensing^18–19 with cycling the particle forward and backward (i.e., particle ping pong)²⁰ to measure particle size over an extended period of time and to increase the probability of observing a disassembly event. Resistive-pulse sensing with multiple pores in series yields a unique signature of multiple current pulses that can be easily identified in real time, improves identification of a single particle, and increases the precision of the particle-size measurements when the pulses are averaged.^20–21 With single nanopores, similar ping-pong techniques have been used in conjunction with resistive-pulse sensing to control and analyze DNA molecules^22–24 and nanoparticles.^25–26

Here, we mixed T = 4 HBV capsids with 1.0 to 3.0 M GuHCl and monitored the reaction over time by randomly selecting particles and measuring their size. While trapped in the nanopore region, a fraction of the particles disassembled. Different concentrations of GuHCl gave very different disassembly behaviors, likely due to the types of interactions between the GuHCl and dimer. We show that disassembly was an accelerating process, in which the disassembly rate increased to around tens of dimers per second just before capsids dissociated. Capsids fell apart catastrophically after reaching a threshold for stability, presumably a percolation threshold,²⁷ which appeared to be ~80 dimers. In some cases, metastable intermediates were observed during disassembly that lost one or more hexamers (trimers of dimers) during the initial stages of disassembly.

Results and Discussion

Device Design and Operation.

Figure 1a shows a schematic of the nanofluidic device used for these disassembly experiments. The inset shows a scanning electron microscope (SEM) image of the four nanopores in series. The four nanopores are 60 nm wide, 60 nm deep, and 300 nm long and are labeled p1, p2, p3, and p4. The nanochannels between adjacent nanopores are 300 nm wide, 130 nm deep, and 500 nm long and constitute the pore-to-pore regions. For resistive-pulse measurements, the device is filled with an electrolyte, e.g., 50 mM HEPES buffer with 1.0 M NaCl and 1.0 to 3.0 M GuHCl, and the T = 4 HBV capsids in the same buffer are placed in the sample reservoir. When a capsid passes through a pore, electrolyte is displaced, the resistance increases, and a current pulse is generated. The amplitude of the current pulse (Δi) is proportional to the particle volume (or, because particles are porous, the volume of dimer in the capsid), and the time to migrate electrophoretically between pores (i.e., pore-to-pore time) is inversely proportional to the particle electrophoretic mobility.

Figure 1.

(a) Schematic of the nanofluidic device and scanning electron microscope image of the nanopore region (inset) for the single particle disassembly experiments. The four nanopores are 60 nm wide, 60 nm deep, and 300 nm long and are labeled p1, p2, p3, and p4. The nanochannels between adjacent nanopores are 300 nm wide, 130 nm deep, and 500 nm long. (b) Schematic of device operation. The potential is applied along the nanochannel and controlled by a LabVIEW program, and potential switching is triggered after four current pulses are detected that correspond to a particle translocating through pores p1, p2, p3, and p4. The particle is cycled forward and backward (i.e., ping pong) until a specified time of 60 s (8004 current pulses; 1000.5 cycles) is achieved, the particle disassembles, or the particle is lost.

With the four-pore design, each particle passes sequentially through four pores and generates a set of four current pulses. To cycle the particles forward and backward (Figure 1b), the polarity of the potential is switched after the set of four pulses is detected to drive the same capsid in the opposite direction. Figure 2a shows the final four cycles (32 current pulses) of a T = 4 capsid prior to disassembly. The particle translocates sequentially through pores p1, p2, p3, and p4 in the forward direction (Figure 2b) and through pores p4, p3, p2, and p1 in reverse direction (Figure 2c). Figure 2d shows the four-pulse sequence of the T = 4 capsid just before disassembling completely. Each device was calibrated with a standard mixture of T = 3 and T = 4 particles. To convert the particle size to dimers, the pulse amplitude of the measured particle (Δi) was divided by the average pulse amplitude of the T = 4 capsids (Δi_T=4). The threshold for a minimum detectable particle size was set to 7σ of the baseline current. To ensure each set of four pulses corresponded to the same particle, the pore-to-pore times for that particle had to be within the expected distribution of pore-to-pore times (average ± 2σ).

Figure 2.

(a) Current trace of a T = 4 hepatitis B virus (HBV) capsid being cycled forward and backward during the final four cycles before disassembly in 2.0 M guanidine hydrochloride (GuHCl). The blue and red rectangles highlight current pulses of typical amplitude for a T = 4 HBV capsid in the forward and reverse directions, respectively, and the green rectangle highlights current pulses of lower amplitude as the particle disassembles. (b)-(d) Enlarged regions of panel (a) that show four current pulses associated with the T = 4 capsid translocating sequentially through (b) pores p1, p2, p3, and p4 in the forward direction and (d) pores p4, p3, p2, and p1 in the reverse direction. (d) Final four current pulses measured for the T = 4 capsid before complete disassembly. Scale bars for panels (b)-(d) are the same.

The four-pore design has several benefits. With four pores in series, four independent measurements of each capsid are made and averaged, which improves the precision of the particle-size measurement by a factor of 2 compared to a single measurement.²¹ Consequently, incomplete capsids that differed in size by three dimers were easily distinguished. Additional pores in series (e.g., eight pores) would further narrow the particle-size distribution, but the overall resistance of the nanopore region would increase and lead to lower pulse amplitudes (Δi) and poorer limits of detection. For these experiments, four pores in series were a reasonable compromise between the 2-fold improvement in particle-size measurement precision and a minimum detectable particle size of ~40 dimers. Moreover, the four-pulse sequence provides a unique current pulse pattern during a translocation event for each particle that is easily recognized by the data acquisition software, which then triggers the potential to switch.

In these experiments, particles were cycled forward and backward for up to 60 s (8004 pulses or 1000.5 cycles), or the capsid disappeared, either by disassembling or being lost by the control program. There are two primary reasons the capsid might be lost by the control program. Reversing the direction of a capsid usually takes several milliseconds, during which the total distance that each capsid travels differs due to electrophoretic migration and diffusive transport. The capsid may travel too far from the pore entrance and may not return to the pore in the designated time window. Or, if the capsid is too close to the pore entrance, the four current pulses on the return trip might not be detected, because the electronics need a relaxation time of ~2 ms before the baseline stabilizes after switching the potential. Another reason is the time needed to collect the current signal, analyze it, and send a command to switch the potential. To minimize this time overhead, the program used to control the experiment was optimized. Although a single particle can easily be trapped for over 60 s with the optimized program, we set the upper limit to 60 s (or 8004 measurements) because trapped particles typically disassembled, did not react, or were lost within this time window (Figure 3).

Figure 3.

Variation of particle size in dimers with trapping time of single T = 4 HBV capsids during disassembly experiments in 2.0 M GuHCl. (a) Trapping of six T = 4 capsids that do not change in size and do not disassemble before they are lost by the cycling program or the 60-s monitoring period ends (8004 current pulses; 1000.5 cycles). (b) Trapping of six T = 4 capsids that exhibit a rapid downturn in size and disassemble (marked with arrows). Full-size T = 4 HBV capsids are 120 dimers (dashed gray lines), and the detection limit is ~40 dimers.

Disassembly Reactions in GuHCl.

The nanofluidic devices were equilibrated with 50 mM HEPES buffer with 1.0 M NaCl and 1.0 to 3.0 M GuHCl. Purified T = 4 HBV capsids were prepared in the same buffer and immediately loaded into the sample reservoir of the device (Figure 1a). Mixing and loading of the sample took ~90 s. After the sample was loaded onto the device, resistive-pulse measurements commenced. Table 1 summarizes the GuHCl concentration, number of experiments conducted, total analysis time of those experiments, total number of measured particles, number of disassembly events, and number of multi-level disassembly events.

Table 1.

Number of experiments, total analysis time of those experiments, total number of measured particles, number of disassembly events, and number of multi-level disassembly events for T = 4 HBV capsids during disassembly experiments in 1.0 to 3.0 M guanidine hydrochloride (GuHCl).

[GuHCl] (M)	number of experiments	total time (min)	total measured particles	dis-assembly events	multi-level events
1.0	1	171	2596	5	0
1.5	2	141	1184	393	28
2.0	3	220	924	264	56
2.2	4	205	562	69	29
2.5	11	296	629	25	12
2.8	7	102	289	11	3
3.0	2	13	11	0	0

While being trapped, some capsids disassembled, whereas others did not disassemble (Figure 3). The timescale and degree of disassembly depended on the GuHCl concentration (Figure 4 and Table 1). At 1.0 M Gu, the GuHCl concentration was too low to effect disassembly. From 1.5 to 2.8 M Gu, disassembly events were observed at each GuHCl concentration with the fraction of disassembly events decreasing with increasing GuHCl concentration. At 3.0 M, no disassembly events were observed because disassembly is too rapid, and only 11 particles were successfully trapped over 13 min.

Figure 4.

(a) Probabilities of all disassembly events and multi-level disassembly events of T = 4 HBV capsids in 1.0 to 3.0 M GuHCl. (b) Variation of fraction of counts with start time of particle trapping and time to disassemble for 2.0 and 2.5 M GuHCl. (c)-(f) Scatter plots of particle starting and ending sizes in dimers for HBV capsids trapped in 1.0, 1.5, 2.0, and 2.5 M GuHCl. Blue lines set the cutoff (90% of the T = 4 capsid size) below which particles are counted as disassembled and reported in panel (a). Vertical and horizontal red lines show the T = 4 capsid size (120 dimers), and the 45° red lines indicate where the starting and ending sizes are the same.

Distributions for the start time of particle trapping and time to disassemble for experiments in 2.0 and 2.5 M GuHCl are shown in Figure 4b. For ≤ 2.0 M GuHCl, capsids were easily trapped and observed for up to 2 h after initiating the reaction. Because minimal disassembly was observed at 1.0 M GuHCl, the experiments were stopped after 2596 events were detected (~171 min). However, at 2.5 M GuHCl, each experiment lasted < 50 min, and no particles were detected after 50 min (Figure 4b). At > 2.5 M GuHCl, the reaction window was even shorter.

Figure 2a shows the last four cycles (32 current pulses) of a capsid in 2.0 M GuHCl before the capsid disassembled. The pulse amplitude (Δi) is directly proportional to the capsid volume and decreased quickly within a few 100 ms as the capsid disassembled. From each of these groups of four peaks, the particle size at a given time was extracted over time and tabulated. The variation of capsid size with time for ten capsids is plotted in Figure 3, and the majority of these traces did not show changes in the size over time. At any given time, thousands of capsids reside in the microchannels, but only a portion of them might undergo the disassembly process. During the experiment, the system randomly selected a particle and monitored it for up to 60 s. The probability of seeing a particle disassemble while trapped was ~30% at 1.5 M GuHCl and lower for other GuHCl concentrations (Figure 4a). Figure 3a shows five capsids that did not change in size and did not exhibit disassembly. The monitoring time for each trace varied because the time at which a capsid was lost by the program is random. The fact that most particle traces showed no size change over time suggested that the multi-cycle resistive-pulse system itself did not contribute to the disassembly of capsids. Figure 3b shows examples of disassembly monitored in real time. In all cases, the trapped HBV capsids tended to change size very slowly at the beginning, but when disassembly occurred, capsids catastrophically fell apart within a few 100 ms (see arrows in Figure 3b).

We performed experiments with GuHCl concentrations ranging from 1.0 to 3.0 M. At low concentrations, e.g., 1.0 M GuHCl, only five capsids disassembled out of 2596 capsids measured over 171 min (Table 1). Conversely, at very high GuHCl concentrations, e.g., ≥ 3.0 M GuHCl, capsids fell apart very quickly, and only 11 capsids and no disassembly events were detected over two experiments for 13 min. For each GuHCl concentration, we extracted the starting and ending sizes and generated scatter plots (Figure 4c–f). The vertical and horizontal red lines demark full-sized T = 4 capsids (120 dimers), and the diagonal red line is where the starting and ending sizes are the same. The blue line designates the threshold at which the ending size is 90% of the starting size. All particles that fell below the blue line were considered disassembly events. The probability of disassembly events is summarized in Figure 4a. Almost all events at 1.0 M GuHCl (Figure 4c) started as a full capsid and ended as a full capsid, whereas at 2.5 M GuHCl (Figure 4f), many events started with incomplete capsids (< 120 dimers) and ended with incomplete capsids, indicating that these incomplete capsids were metastable for the duration of the experiment (diagonal line in Figure 4f). These data indicated that low GuHCl concentrations did not lead to capsid disassembly, in agreement with previous experiments.⁴ More importantly, these data show that our multi-cycle resistive-pulse system, which drives the same particle forward and backward up to a thousand times in an electric field, did not induce capsid disassembly, even in the presence of low GuHCl concentrations (1.0 M).

Although bulk measurements show extensive disassembly events in 1.5 M GuHCl,⁴ the timing for disassembly events was very different compared to reactions in higher GuHCl concentrations (Figure 4b). At low GuHCl concentrations, the capsids were stable up to several hours, and very few disassembly events were detected for the first hour of the experiment. After ~1 h of incubation, a high percentage of events collected were due to disassembly. This incubation period suggests that a threshold event, or events, must take place before initiating disassembly. At high GuHCl concentrations (> 2.0 M), capsids were stable for less than an hour, and disassembly events started almost immediately after the experiment started and without an incubation period (Figure 4b). The transition between necessitating an incubation period for disassembly at ≤ 2.0 M GuHCl and almost immediate disassembly at > 2.0 GuHCl correlates well with the transition from nondenaturing to denaturing conditions for the dimer, which occurs between 2.0 and 2.5 M GuHCl.²⁸

Hysteresis may arise from a conformational transition prior to assembly, i.e., allosteric transition of dimers regulate assembly, or from a post-assembly capsid maturation transition. Disassembly at 1.5 and 2.0 M GuHCl showed that most capsids required an hour incubation before disassembly was initiated. This observation suggests that the capsid must first undergo a transition into a conformation that can dissociate. In GuHCl > 2.0 M, this lag time was much shorter or altogether nonexistent, and dissociation more frequently involved metastable intermediates (see below), which suggests a difference in capsid stability and mechanism of disassembly. Again, this abrupt transition between 2.0 and 2.5 M GuHCl notably corresponds to the transition from nondenaturing to denaturing conditions for the dimer.²⁸

Acceleration of the Disassembly Process.

For disassembly events, exemplified in Figure 3b, the slope of the particle size over time was fitted to determine the disassembly rate (dimers/s; Figure 5). Specifically, when the particle size was > 90% of a T= 4 capsid (120 dimers), the slope fitting included 200 adjacent points (~1.5 s), because most events showed particle size changing slowly at the beginning. When the particle size fell below 90%, the slope fitting only included 50 adjacent points (~400 ms), because the particle size changed rapidly in this size range. From the fittings, the relation between disassembly rate and particle size was extracted for each event.

Figure 5.

(a) Variation of average disassembly rate with particle size for HBV capsids disassembled in 1.5, 2.0, 2.2, and 2.5 M GuHCl. The average disassembly rate was calculated from at least five particles having the same size (dimers). (b) Slope analysis data for all disassembly events at 2.0 M GuHCl and average disassembly rate for those data. Heat map of slope data ranges from most frequent (red) to least frequent (purple), and the yellow line with black outline designates the average disassembly rate.

The disassembly rates at 1.5, 2.0, 2.2, and 2.5 M GuHCl are shown in Figure 5a, and as an example, the raw slope data in the form of a heat map and average disassembly rate for 2.0 M GuHCl are plotted in Figure 5b. Only data from 1.5 to 2.5 M GuHCl are plotted here because very few measured disassembly events occurred at 1.0 M GuHCl or ≥ 2.8 M GuHCl. One consistent trend is that the disassembly rate increased as the particle size decreased, which indicated that as particles broke apart, disassembly was an accelerating process. The disassembly rate increased from several dimers/s to tens of dimers/s as the particles disassembled. The disassembly rates at 1.5 and 2.0 M GuHCl mapped similarly to each other and were faster than the disassembly rates at 2.2. and 2.5 M GuHCl (Figure 5a). At 1.5 M GuHCl, most of the observed disassembly events began with particles that had more than 80 dimers, and very few particles continued to dissociate below 80 dimers, which suggests smaller particles completed disassembly very rapidly (black line in Figure 5a). At 2.0, 2.2, and 2.5 M GuHCl, a measurable fraction of particles persisted well below 80 dimers in size and their disassembly could be observed at or above 50 dimers in size.

Multi-level Disassembly of Capsids.

Some disassembly events proceeded by loss of blocks of subunits, creating multiple levels of intermediates. Figure 6a shows disassembly trajectories for two such particles. Particle 1 shows steps at 117 and 104 dimers, whereas particle 2 shows steps at 117 and 108 dimers (Figure 6b). Interestingly, disassembly events at higher GuHCl concentrations not only had a slower disassembly rate, but also showed a higher fraction of multi-level events (Figure 4a and Table 1). This stepwise disassembly is direct evidence of metastable intermediates, and each peak in the distribution represents a metastable intermediate. The sizes of these intermediates for disassembly in 2.0 M GuHCl are compiled in Figure 6c. Of interest, the most abundant intermediates have sizes of 111, 114, and 117 dimers, which show differences of three dimers, corresponding to the trimeric lattice of dimers that forms the capsid surface. Similar distributions were observed for 1.5, 2.2, and 2.5 M GuHCl (data not shown). Through visualization of individual stepwise traces, we also found that very few traces exhibit large steps, i.e., a capsid breaking into two large pieces. As seen in Figure 5a, faster disassembly rates at lower GuHCl concentrations may seem counterintuitive. However, one may speculate that breaking interdimer contacts for the loss of several dimers at a time, e.g., multi-level events (Figure 6), is likely to be much slower than releasing one dimer at a time, which suggests a basis for slower average disassembly rate in higher GuHCl concentrations. An alternative explanation is that a given partial particle, especially in high GuHCl, may partially collapse on itself to form a metastable structure.

Figure 6.

(a) Examples of two T = 4 HBV capsids exhibiting multi-level disassembly in 2.0 M GuHCl. Points are the average of 4 pulses (0.5 cycle), and lines are the average of 20 current pulses (2.5 cycles). The dashed red line corresponds to the size of a T = 4 capsid (120 dimers). (b) Variation of particle counts with particle size for the line data in panel (a) for particles 1 and 2 projected onto the y-axis. (c) Variation of particle counts with particle size for all metastable intermediates measured in 2.0 M GuHCl. The peaks in panels (b) and (c) denote common particle sizes observed during stepwise disassembly events.

Conclusion

Multi-cycle resistive-pulse sensing provides a platform to precisely measure particle sizes in solution with high temporal resolution. Application of this system to HBV disassembly in GuHCl demonstrated the ability to monitor disassembly at a single particle level over a range of reaction conditions and long periods of time. By trapping enough particles, we were able to develop a picture of the disassembly reaction with most disassembly events showing a similar pattern. HBV capsids tended to have a lag time prior to disassembly and to change size very slowly at the beginning. At high GuHCl concentrations, some capsids exhibited metastable intermediates during disassembly. Towards the end, all capsids disassembled very rapidly and catastrophically fell apart within a few 100 ms. Notably, at 1.5M GuHCl, the disappearance of partial particles smaller than 80 dimers is consistent with the percolation threshold of a 90-mer for particle stability previously observed for HBV capsids.^27,29 This study provided new insight into the basis of HBV hysteresis to dissociation. This nanofluidic system is not restricted to this particular reaction. Any reaction that results in changes of particle size that can be tracked over seconds to tens of seconds could theoretically be monitored with this system.

Experimental Section

Device Fabrication.

The nanofluidic devices consisted of two V-shaped microchannels connected by a series of nanochannels and nanopores (Figure 1a).^21,30 Microchannels were patterned into a D263 glass substrate by standard UV photolithography followed by wet chemical etching of the chromium film with chromium etchant and the glass substrate with buffered oxide etchant.³¹ Access holes at the ends of the microchannels were sandblasted through the substrate with aluminum oxide powder. Nanochannels and nanopores (Figure 1a) were directly milled in the 10-μm gap between the V-shaped microchannels with a focused ion beam instrument (Auriga 60, Carl Zeiss AG). The nanopores were 60 nm wide, 60 nm deep, and 300 nm long, and the pore-to-pore nanochannels were 300 nm wide, 130 nm deep, and 500 nm long. The substrate and cover glass, also D263, were cleaned, hydrolyzed in NH₄OH/H₂0₂/H₂O, rinsed with H₂O, brought into contact with each other, and annealed at 545 °C for 12 h. After bonding, reservoirs were installed over the access holes on the devices with epoxy. The channels were then rinsed with 0.1 M NaOH, rinsed again with H₂O, and coated with 2-[methoxy(polyethyleneoxy)6–9propyl]dimethylmethoxysilane (Gelest, Inc) to reduce electroosmotic flow and minimize particle adsorption.³²

Sample Preparation.

HBV core protein dimers (Cp149 dimer, 34 kDa) were expressed in E. coli and purified by established procedures.³³ T = 3 and T = 4 HBV capsids were assembled from Cp149 dimers by 24 h incubation in 300 mM NaCl and separated after assembly on a 10−40% (w/v) continuous sucrose gradient in 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; pH 7.5) with 300 mM NaCl that was centrifuged for 6 h at 150 000g.³⁴ The lower particle band (T = 4 capsids) was extracted and dialyzed into a 50 mM HEPES buffer with 1 M NaCl. For the disassembly experiments, 0.2 μM solutions of T = 4 capsids were prepared in 50 mM HEPES with 1 M NaCl and 1.0 to 3.0 M guanidine hydrochloride (GuHCl). The solution was then immediately loaded into the sample reservoir on the nanofluidic device to perform the disassembly experiments. The sample mixing and loading onto the device took ~90 s.

Resistive-Pulse Measurements.

For the resistive-pulse measurements, the HBV capsids were trapped and cycled back and forth across the four nanopores in series.²⁰ After the sample was loaded onto the device, the potential was applied with an Axopatch 200B (Molecular Devices, LLC) between the sample and waste reservoirs along the nanochannel (Figure 1a), and the current was monitored with the same Axopatch 200B connected through a data acquisition card (PCIe-6353, National Instruments) to a computer. Current traces were analyzed in real time by a program written in LabVIEW (National Instruments). Current pulses were identified by setting a detection threshold of 7σ, where σ is the standard deviation for current measurements taken at 1.0 M GuHCl. After four current pulses were detected, a timing delay (typically 8 ms) allowed the capsid to migrate electrophoretically away from the pore entrance, and then, the potential was reversed. This delay permitted the current spike due to the capacitance of the sensing electronics to decay to baseline (~2 ms).

After the timing delay, the program resumed collecting pulses, and after observing another series of four pulses, the program reversed the potential again. This process was repeated until the particle was analyzed for 8004 current pulses (1000.5 cycles), disassembled, or was lost. The data were tagged with labels at different phases of pulse identification (e.g., waiting for a particle to enter the nanopores, recognizing four current pulses in series, and delaying before potential switching).

Data Processing.

During the ping-pong experiments, current, potential, pulse threshold, and the labels for different phases of pulse identification were all saved to simplify subsequent data analysis. Data were continuously saved to hard disk in TDMS format and analyzed offline with a MATLAB program. The analysis program parsed the data and extracted pulse sequences by the recorded pulse identification phase label. Each 4-pulse sequence was averaged to determine the particle size, and five 4-pulse sequences were averaged to determine trends, such as multi-level disassembly events.