ABSTRACT
Encapsulins, self-assembling icosahedral protein nanocages derived from prokaryotes, represent a versatile set of tools for nanobiotechnology. However, a comprehensive understanding of the mechanisms underlying encapsulin self-assembly, disassembly, and reassembly is lacking. Here, we characterise the disassembly/reassembly properties of three encapsulin nanocages that possess different structural architectures: T = 1 (24 nm), T = 3 (32 nm), and T = 4 (42 nm). Using spectroscopic techniques and electron microscopy, encapsulin architectures were found to exhibit varying sensitivities to the denaturant guanidine hydrochloride (GuHCl), extreme pH, and elevated temperature. While all encapsulins showed the capacity to reassemble following GuHCl-induced disassembly (within 75 min), only the smallest T = 1 nanocage reassembled after disassembly in basic pH (within 15 min). Furthermore, atomic force microscopy revealed that all encapsulins showed a significant loss of structural integrity after undergoing sequential disassembly/reassembly steps. These findings provide insights into encapsulins’ disassembly/reassembly dynamics, thus informing their future design, modification, and application.
INTRODUCTION
Protein nanocages (e.g., virus-like particles (VLPs), ferritins, heat-shock proteins) self-assemble from multiple protein subunits into highly-organised macromolecular structures, that exhibit well-defined inner cavities, outer surfaces, and interfaces between subunits. Their capacity to encapsulate cargo, coupled with the ability to genetically and/or chemically modify their structures, has enabled protein nanocages to be custom-engineered for a multitude of applications, including biocatalysis, materials synthesis, sensing, vaccines, and drug delivery.1, 2
Encapsulins are an emerging class of protein nanocages found inside many archaea and bacteria. They self-assemble from identical protein subunits into hollow icosahedral nanocages that structurally resemble the major capsid protein gp5 of the HK97 virus.3, 4 Based on their triangulation number (T), all encapsulins exhibit one of the following three symmetrical icosahedral architectures: T = 1 (24 nm, 60-mer, 12 pentameric units); T = 3 (32 nm, 180-mer, 12 pentameric and 20 hexameric units); and T = 4 (42 nm, 240-mer, 12 pentameric and 30 hexameric units). 5–8 In nature, encapsulins house cargo enzymes that mediate oxidative stress resistance, iron storage, anaerobic ammonium oxidation, or sulfur metabolism.9–11 Uniquely, encapsulins selectively self-assemble around cargo enzymes tagged with a small encapsulation signal peptide (ESig), packaging them.12 This mechanism has been adapted to load foreign cargo into encapsulins, reprogramming their functionality for different practical applications. 13–16
Encapsulin subunits autonomously assemble, with extraordinary fidelity, into macromolecular nanocages. Such self-assembly is not only driven by folding of the individual polypeptide chains, but also by dynamic noncovalent interactions between the different polypeptide chains both within subunits and at the interfaces between subunits in the assembled supramolecular structure.17 Unravelling the self-assembly mechanisms of protein nanocages is complicated, especially if they exhibit highly symmetric homooligomeric structures, like encapsulins.18 Nevertheless, multiple analytical techniques now allow the molecular mechanisms underlying protein nanocage assembly (e.g., protein folding) to be characterised and subsequently exploited.
For instance, the disassembly/reassembly of protein nanocages belonging to the ferritin family have been studied via a combination of: intrinsic tryptophan fluorescence (ITF), circular dichroism (CD), UV/VIS spectroscopy and synchrotron small-angle X-ray scattering (SAXS) measurements to assess protein conformation19–22, transmission electron microscopy (TEM) and dynamic light scattering (DLS) to evaluate structural integrity, shape and size distribution, and laser light scattering to monitor assembly kinetics.23 One study revealed that ferritin disassembles at extremely acidic pH 1.5, then shows a rapid reassembly upon return to neutral pH 7.0, accompanied by folding, followed by a slow phase in which the final 24-mer nanocage is formed.23 Importantly, this fundamental work led to the rational re-design of ferritin subunit interfaces, resulting in engineered nanocages capable of disassembly at a more amenable pH 4.0.24, 25 Such modification now permits labile compounds (e.g. small-molecule drugs) to be controllably loaded into ferritin nanocages in a facile and non-destructive manner, enabling downstream applications (e.g., drug delivery).24, 25
In contrast, experimental data pertaining to encapsulins’ ability to disassemble/reassemble and the mechanisms that underpin this natural phenomenon are sparse. The most characterised system is the T = 1 encapsulin from Thermotoga maritima (Tm-Enc), whose disassembly/reassembly has been primarily inspected via CD, polyacrylamide gel electrophoresis (PAGE), and TEM.14 Specifically, Tm-Enc has been found to disassociate in strong acidic and alkaline conditions, or high concentrations of denaturing agents (e.g., guanidine hydrochloride, GuHCl). Furthermore, Tm-Enc was shown to spontaneously reassemble upon returning to the initial conditions (i.e., neutral pH or absence of denaturant).27 Interestingly, Tm-Enc can be reassembled in the presence of ESig-tagged cargo (e.g., proteins, nanomaterials), resulting in their selective encapsulation in vitro, and thus further expanding encapsulins’ utility.8, 27, 28,29 Despite these promising findings, key questions concerning the biophysical mechanisms and physicochemical factors that underlie encapsulin disassembly/reassembly, and how they might be controlled, remain unanswered, especially for the T = 3 and T = 4 nanocages.
Motivated by this absence of information, we selected encapsulins with structures representing each of the three known architectures, and then interrogated their disassembly/reassembly. These nanocages included Tm-Enc (T = 1), and the larger and more structurally complex encapsulins from Myxococcus xanthus (Mx-Enc, T = 3) and Quasibacillus thermotolerans (Qt-Enc, T = 4). We combined intrinsic tryptophan fluorescence (ITF) spectroscopy, DLS, PAGE and TEM, to accurately monitor the assembly states of all three encapsulins under varying physicochemical conditions, including exposure to extreme pH, strong denaturants (GuHCl), and elevated temperatures. Furthermore, the effect disassembly/reassembly had on the nanocages’ structural integrity was evaluated by atomic force microscopy (AFM). Together, this work provides critical insights into the dynamic mechanisms that govern the disassembly/reassembly of differing encapsulin structures, which will help to expedite and broaden their future design, modification, and practical application.
MATERIALS AND METHODS
Materials
All chemicals and reagents used in this study were purchased from Sigma-Aldrich, unless stated otherwise.
Molecular cloning of constructs
All inserts were codon optimised for expression in Escherichia coli and custom synthesised as gBlock Gene Fragments (Integrated DNA Technologies). Encapsulins from Thermotoga maritima (Tm) (UniProt: TM_0785), Myxococcus xanthus (Mx) (UniProt: MXAN_3556) and Quasibacillus thermotolerans (Qt) (UniProt: QY95_01592) were each synthesised with flanking restriction sites (NcoI/BamHI). For gene expression in E. coli, Tm-Enc was cloned into pETDuet-1 (Novagen, Merck), and Mx-Enc and Qt-Enc were cloned into pACYC-Duet-1 (Novagen, Merck), summarised in Supplementary Table S1. E. coli α-Select (Bioline, UK) was used for general plasmid storage and propagation. Gene insertion was confirmed by PCR using primer pairs pETUpstream/DuetDOWN (Merk). E. coli BL21 (DE3) cells (New England Biolabs) were used for recombinant protein expression. Herein, cells were transformed with the appropriate plasmids, and the resulting transformants were selected on Luria–Bertani (LB) agar supplemented with either 100 mg/mL of carbenicillin or 50 mg/mL of chloramphenicol (see Supplementary Table S1).
Recombinant Protein Production
Protein expression experiments were performed in LB medium supplemented with carbenicillin (100 mg/mL) or chloramphenicol (50 mg/mL). Briefly, strains were streaked out on LB agar plates and grown overnight at 37°C. A starter culture (1 colony in 5 mL LB) was grown for 16 h at 37°C and used to inoculate 500 mL of LB media. Cultures were incubated at 37°C/200–250 rpm until an optical density at 600 nm (OD600) of 0.5-0.6 was reached. Protein synthesis was then induced by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 0.1 mM. Induced cultures were incubated at 37°C/200–250 rpm for 4 h and then cells were harvested via centrifugation (8,000 × g, 4°C, 15 min). The resulting cell pellets were stored at −30°C until further use.
Protein purification
Cell pellets from 500 mL encapsulin-producing cultures were thawed and resuspended in 30 mL of lysis buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer pH 7.4 (Chem-Supply Pty) and Benzonase® nuclease 10 U/mL). Cells were lysed by three rounds of passage through a French pressure cell at 1000 psi and centrifuged at 8,000 × g, 4°C for 15 min. Supernatant containing soluble protein was heat treated in a water bath at 65°C for 15 min before centrifugation (10,000 × g, 4°C, 10 min). Protein precipitation was initiated by adding 10 % (w/v) PEG8000 and 2 % (w/v) NaCl to the supernatant, followed by incubation on ice for 30 min. Next, the sample was spun down at 10,000 × g for 10 min at 4°C. The precipitated protein was resuspended in 2.5 mL of HEPES buffer (50 mM, pH 7.4) and filtered through a 0.22 mm syringe filter.
All purifications were carried out on an Äkta™ start chromatography system (GE Healthcare). The three encapsulins used in this study were purified via size exclusion chromatography (SEC) using a HiPrep 26/60 Sephacryl S-500 HR column (GE Healthcare) equilibrated with 50 mM HEPES pH 7.4. Fractions showing the corresponding encapsulin band on SDS-PAGE were pooled and subjected to further purification via anion-exchange chromatography using a HiPrep Q 16/10 column (GE Healthcare) equilibrated with 50 mM HEPES pH 7.4. Encapsulin proteins were eluted with linear gradient of 0–0.3 M NaCl and 0.3– 1M NaCl (Supplementary Figure S1). Fractions containing encapsulin, identified via sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE), were pooled, concentrated and buffer exchanged into 50 mM HEPES buffer pH 7.4 using Amicon Ultra-15 centrifugal filter units with a 100 KDa cut-off. Lastly, purified encapsulin Enc proteins were filtered through a 0.22 mm syringe filter and stored at −30°C until further use.
PAGE analysis and protein quantification
Protein samples were denatured, separated, and visualised using SDS-PAGE, with molecular weights compared with a commercial protein ladder (Precision Plus Protein, BioRad). The Bio-Rad mini-protean system (Bio-Rad laboratories) was used for SDS-PAGE analysis. Protein samples were mixed in a 1:1 ratio with 2X Laemmli sample buffer with 50 mM 1,4-dithiothreitol and heated at 99°C for 10 min at 300 rpm. Electrophoresis was performed at 200 V for 30 min on a 4–20% polyacrylamide gel (Mini-PROTEAN® TGX™, BioRad) in SDS running buffer (25 mM Tris, 192 mM glycine, 1% (w/v) SDS, pH 8.3). Gels were stained following the Coomassie G-250 safe stain protocol. Encapsulin assembly was visualised via non-denaturing Blue Native-PAGE (BN-PAGE). The XCell SureLock Mini-Cell Electrophoresis System (ThermoFisher Scientific) was used for BN-PAGE analysis. Protein samples were mixed in a 1:4 ratio with 4X Native-PAGE sample buffer (ThermoFisher Scientific) and loaded into NativePAGE™ 3–12% Bis-Tris protein gels (ThermoFisher Scientific). BN-PAGE was performed using two different running buffers: 1X anode buffer (NativePAGE™ running buffer 20X, ThermoFisher Scientific) in the outer buffer chamber and 1X dark blue cathode buffer (1X anode buffer, 0.02% (w/v) Coomassie G-250) in the inner buffer chamber. Lastly, the samples were run on ice at 150 V for 90 min follow by a second run at 250 V for 30 min. Protein concentration was determined by measuring the absorbance at 280 nm on a NanoDrop 2000 Spectrophotometer instrument (ThermoFisher Scientific).
In vitro disassembly/reassembly of encapsulin
To characterise the in vitro disassembly of encapsulin, the presence of a denaturing agent (GuHCl), changes in pH, and thermally induced disassembly methods were used. GuHCl was added to the encapsulin sample to final concentrations between 1–7 M. For pH disassembly, dilution of the purified encapsulins in 50 mM HEPES at different pH (3–13) was performed. Final pH was confirmed with pH strips. Once in their respective conditions, encapsulin samples were incubated for 1 h at room temperature. For thermally induced disassembly, the encapsulin sample was subjected to a temperature ramp from 20–90°C at a rate of 2°C/min. In all disassembly methods, the final encapsulin concentration was 5 μM, dithiothreitol (DTT) was added to the encapsulin sample to a final concentration of 1 mM and 50 mM HEPES pH 7.4 was used to reach the desired final volume.
For stability experiments, 5 μM of encapsulin was incubated for 1 h at room temperature under reducing (10 mM DTT), oxidising (10 mM H2O2), and high ionic strength (1 M NaCl) conditions.
The subsequent reassembly of encapsulins was initiated by returning the sample to original conditions. Briefly, the samples were dialysed against 50 mM HEPES buffer pH 7.4 and 1 mM DTT at room temperature overnight, using 3.5K MWCO SnakeSkin dialysis tubing (ThermoFisher Scientific). To measure the reassembly rate, samples were removed from dialysis every 15 min for 75 min and centrifuged for 5 min at 10,000 × g to remove any aggregated proteins.
Intrinsic tryptophan fluorescence
ITF measurements of encapsulins in their varying states of assembly were performed with an FP-8500 Spectrofluorometer (JASCO) using a 3 mm pathlength micro-volume quartz cuvette. Samples (60 μL) were prepared in triplicate with a final encapsulin concentration of 5 μM. Samples were excited at 290 nm and emission spectra were collected from 300–450 nm. The measurement parameters were: excitation and emission bandwidths of 5 nm, response of 0.2 s, medium sensitivity, data interval of 0.1 nm, scan speed of 100 nm/min, and 4 measurement accumulations were averaged. To investigate the effect of temperature on the different encapsulins, spectra were collected from 20–90°C with a temperature ramp of 2°C/min. The obtained spectra were further processed by buffer spectra subtraction using Spectra Manager™ software (JASCO), and the ratio between the fluorescence intensity at 360 and 330 nm (360/330) was calculated and plotted in Microsoft Excel.
Transmission electron microscopy (TEM)
To visualise morphology, size, and state of encapsulin assembly, TEM was performed using a Philips CM10 microscope (100 kV accelerating voltage). Briefly, encapsulin samples (0.2 mg/mL) were adsorbed onto pioloform-coated 200 mesh copper grids (ProSciTech) for 2 min. Prior to imaging, samples were negatively stained for 1 h using uranyl acetate replacement (UAR-EMS, Electron Microscopy Sciences), washed with ultrapure water, and allowed to dry for at least 15 min.
Dynamic light scattering
To measure the diameter of encapsulins, DLS was performed with a Malvern Zetasizer ZSP instrument equipped with a 633 nm laser. Samples with a final encapsulin concentration of 5 μM were prepared as described above. Three measurements were performed at 25°C, using a plastic microcuvette (ZEN0040, Malvern), 173° backscatter and automatic attenuator selection. Data analysis was performed in the Zetasizer Nano software. All DLS sizes reported, herein, are size by number values calculated using distribution analysis. A 1 cm quartz cuvette was used for temperature ramp experiments. Heat maps were made using GraphPad Prism 9.
Atomic Force Microscopy (AFM)
Images and force curve measurements were obtained using a Cypher ES Atomic Force Microscope (Oxford Instrument, Asylum Research, Santa Barbara, CA, USA) at room temperature (25°C). Protocols were adapted from Collett et al 26, 27. For imaging, the instrument was operated in amplitude modulated-AFM (AM-AFM), while force measurements were obtained in contact mode. Biolever BL-AC40TS cantilevers (Oxford Instruments, Asylum Research, Santa Barbara, CA, USA, nominal spring constant kc = 0.09 N/m) were used for all measurements. All experiments were completed within a droplet of media (~100 μL) with a concentration of ~100 ng of encapsulin deposited onto freshly cleaved muscovite surface (the supporting substrate). Prior to experimentation, each cantilever was calibrated via the thermal spectrum method and the lever sensitivity was determined using force spectroscopy. Processing of AFM data involved using a combination of the Asylum Research software, custom MATLAB codes, and the Gwyddion software package 28.
Determining elasticity of intact encapsulins
Force versus distance curves (FD curves) were obtained from the central region of encapsulins. Following image location, FD curves were first obtained on an area of bare mica next to the nanocage to ensure linear (non-elastic) FD curves were observed, which provided a reference for determining elasticity of the particles. The spring constant (kc) of the cantilever was determined for each cantilever used as described above (values in the range of 0.05–0.1 N/m were obtained for all cantilevers). The tip was then directed to the central region of the particle to obtain accurate indentation data. Upward of 50 FD curves were recorded across several particles for each sample.
The tip-sample contact point between the AFM cantilever and the encapsulin was performed independently for each FD curve analysed using methods previously described 29. Specifically, the contact point between the AFM cantilever tip and the encapsulin is defined as the point at which the cantilever first makes physical contact with the surface adsorbed particle. Following this point, the cantilever bends in response to interactions with the encapsulin, while the nanocage itself is indented. In a raw FD curve (z-displacement (Z) vs. cantilever deflection (d)), this point of first contact is mathematically defined as Z0 and d0, respectively. This is observed in the AFM force curve as an increase in force (y-axis), above the zeroed baseline, as a function of distance (x-axis prior to zeroing the indentation).
The indentation (δ) is then calculated via: and transformed into indentation vs. force curves using Hooke’s law; The elastic response of the samples is then fit to the Hertz/Sneddon equation (3) to obtain a Young’s elastic modulus (E) for each of the fitted curves. Using custom MATLAB scripts, curves were fitted to the equations: and; where F is loading force, δ is indentation depth, α is the cone opening angle, E* is the apparent elastic modulus, and v is the Poisson’s ratio. A conal tip angle of 34.4° and a Poisson’s ratio of 0.5 was used for processing of all force curves. All elasticity values were obtained from curves fitted with an R2 value of 0.9 or above. Data falling below this quality was rejected for further analysis.
RESULTS AND DISCUSSION
Monitoring encapsulin assembly/disassembly using intrinsic tryptophan fluorescence
Unloaded Tm-Enc, Mx-Enc, and Qt-Enc were produced in E. coli and purified by SEC and anion exchange chromatography prior to biophysical characterisation. Purification and correct self-assembly were confirmed using SDS-PAGE, blue native-PAGE, DLS, and TEM (Figure 2g, Supplementary Figure S1). TEM images of self-assembled Tm-Enc, Mx-Enc, and Qt-Enc displayed structures with consistent shape and size (Figure 2g) and DLS analysis indicated a diameter of 23.7 ± 4.9 nm for Tm-Enc, consistent with its crystal structure data12, and a diameter of 32.0 ± 6.0 nm for Mx-Enc and 38.2 ± 7.3 nm for Qt-Enc, consistent with their Cryo-EM structures.6, 8 Mx-Enc was produced primarily in its T = 3 structure; however, populations of smaller, T = 1 like, structures were evident in Native-PAGE and TEM results (~18 nm) (Supplementary Figure S1, Figure 2g). This variation in size of Mx-Enc has been previously observed, where recombinantly produced Mx-Enc without the presence of ESig-tagged cargo will form heterogenous T = 1 and T = 3 populations.8
Tryptophan fluorescence is extremely sensitive to local environment. As such, protein unfolding, disassembly or conformational transitions often result in a change in emission spectra of the Trp(s) within a protein; a lower maximum wavelength (blue-shifted) when the Trp(s) are buried, and a higher maximum wavelength (red-shifted) when solvent exposed.30
Encapsulin subunits adopt a HK97-fold and have three conserved structural regions, a peripheral (P)-domain, axial (A)-domain, and an elongated (E)-loop region (Supplementary Figure S2).3, 4 Conveniently, each subunit of Tm-Enc, Mx-Enc, and Qt-Enc contains five, three, and two Trp residues, respectively. For each encapsulin, at least one of these Trps is located within the interface between subunits, and one is located within the hydrophobic core of a single subunit (Figure 1a, b & c). Therefore, these intrinsic Trp residues are likely to be suitable reporters for both assembly/disassembly of the encapsulin macrostructure and folding of the individual subunit polypeptide chains. Indeed, intrinsic Trp fluorescence (ITF) has previously been used to monitor the re-folding of T = 1 encapsulin from Rhodococcus erythropolis N771 when desorbed from a zeolite substrate.31 ITF is also an appealing technique to monitor the process of encapsulin assembly/disassembly owing to its relative simplicity, ability to report on a dynamic ensemble of structures in solution and non-destructive nature, which enables measurements to be performed in real-time. Additionally, unlike fluorescence resonance energy transfer (FRET)-based techniques, ITF does not require modification of the protein with any extrinsic labels that may alter the assembly/disassembly dynamics. Although circular dichroism (CD) may provide more specific secondary structural information, unlike ITF, CD is not compatible with the high concentration of GuHCl and extreme pH likely required for encapsulin disassembly.32
Figure 1 shows the ITF of assembled Tm-Enc, Mx-Enc and Qt-Enc overlayed with the spectra obtained after 1 h incubation with 7 M GuHCl. For all three assembled encapsulins, blue-shifted Trp fluorescence spectra are observed (~334 nm maximum emission wavelength for Tm-Enc and Qt-Enc; and ~338 nm for Mx-Enc), while in the presence of 7 M GuHCl, a dramatic red-shift of the fluorescence spectra is seen (~354 nm maximum emission wavelength). A maximum emission wavelength over 350 nm is observed where the Trps within a protein are completely solvent exposed, as in an unfolded polypeptide chain.33 Therefore, it appears that 7 M GuHCl fully disassembles Tm-Enc, as has previously been demonstrated14, as well as both Mx-Enc and Qt-Enc, by completely unfolding the protein.
It is also interesting to note that the overall fluorescence emission spectral shape represents an average of all Trp environments present in the protein.34 Where a protein contains multiple Trps, these may exist in different environments and therefore emit at different wavelengths. While we cannot draw conclusions about the foldedness/assembly of Tm-Enc, Mx-Enc and Qt-Enc relative to one another (as they all contain a different number of Trps), changes in the spectral shape for the same encapsulin may discern encapsulin assembly (i.e., interaction between subunits) as opposed to the folding of individual subunit polypeptide chains. This is particularly true for Qt-Enc. As can be seen in Figure 1f, the peak-shape for assembled Qt-Enc is flattened. This is consistent with its two Trp residues being located in different solvation environments such that the spectra two Trps overlap, but are not resolved (i.e., Trp95 being more buried in a helix at the pentamer/hexamer interface than Trp154, which is located on a loop in the A-domain, nestled within the hydrophobic core of a single Qt-Enc subunit) (Figure 1c, Supplementary Figure S2).
Instead of reporting the maximum fluorescence wavelength, herein, we have presented much of our ITF data as a fluorescence intensity ratio of emission at 360 nm and 320 nm (360/320). This ratio is used as a proxy for the shift in overall Trp fluorescence emission peak, and avoids any bias arising from the spectral peak shape.35 This approach has allowed us to expand the repertoire of methods used to monitor encapsulin assembly/disassembly across a range of conditions.
GuHCl induced encapsulin disassembly
GuHCl is a denaturant that affects protein structure by disrupting non-covalent interactions, including hydrogen bonding and hydrophobic effects.37 Our initial ITF results indicated that 7 M GuHCl completely unfolds the Tm-Enc, Qt-Enc and Mx-Enc protein polypeptide chains (Figure 1), and thus, disassembles the encapsulin macrostructure. To further examine the effect of GuHCl on encapsulin folding and assembly, all encapsulins were incubated in varying concentrations of GuHCl (0–7 M) for 1 h before performing ITF and DLS.
The observed shift in Trp emission peak for Tm-Enc, Mx-Enc and Qt-Enc in the presence of GuHCl are shown in Figure 2a. As expected from our previous results, all three encapsulins show a blue to red-shift from the assembled (0 M GuHCl, Δ360/320 = 0) to unfolded state (7 M GuHCl, Δ360/320 = 1). However, the concentration of GuHCl at which this transition occurs varies significantly between the different encapsulins. Tm-Enc Trp solvation began from approximately 3 M GuHCl, with complete Trp solvation from 6 M GuHCl, while Mx-Enc and Qt-Enc Trp solvation began at a lower concentration of 1 M GuHCl, with complete Trp solvation from 4 M and 5 M GuHCl, respectively. This variation may reflect structural complexity, with the smaller and pentameric Tm-Enc being the most robust to GuHCl compared to the larger pentameric and hexameric Mx-Enc and Qt-Enc.
To complement these ITF results, DLS analysis was performed to characterise the size distribution of encapsulin in solution for the same samples (Figure 2b, Supplementary Table S2). For Mx-Enc, no intact encapsulin macrostructure is detectable by DLS at concentrations ≥ 3 M GuHCl, which also correlates with the significant red-shifted Trp emission at 3 M GuHCl (Δ360/320 = 0.88). This result suggests that Mx-Enc is both disassembled and largely unfolded at GuHCl concentrations ≥ 3 M. However, for Tm-Enc and Qt-Enc, absence of the intact encapsulin macrostructure occurs at lower GuHCl concentrations than the major red-shift in Trp emission. For example, at 4 M GuHCl, the Tm-Enc macrostructure is not detectable by DLS, but the Trp emission remains significantly blue-shifted (Δ360/320 = 0.20), corresponding to a native-like fold. A similar result is seen for Qt-Enc at 3 M GuHCl (Δ360/320 = 0.27). TEM and native-PAGE images of each encapsulin after incubation in the lowest GuHCl concentration required for significant Trp solvation also show the absence of nanocage macrostructures, confirming their disassembly. Upon dialysis in buffer (50 mM HEPES, 1 mM DTT) overnight, all encapsulins subsequently reassembled into their original macrostructure (Figure 2h, Supplementary Figure S3).
Taken together, these results point to the existence of an intermediate state(s) in Tm-Enc and Qt-Enc GuHCl-induced disassembly, in which the nanocage breaks down into smaller (<0.5 nm) entities with a native-like fold. There is insufficient evidence from ITF or DLS to define the stoichiometry or structure of this intermediate state(s). However, a previous native mass spectrometry study of the T = 1 encapsulin from B. linens suggested reassembly occurs via formation of stable dimers prior to the final nanocage.38 It is possible that disassembly of Tm-Enc and Qt-Enc proceeds via a similar mechanism. We also cannot rule out that Mx-Enc proceeds via an equivalent intermediate state upon GuHCl-induced disassembly. However, if this is the case, our data suggests the chemical stability of the nanocage and the intermediate for Mx-Enc are more closely matched, and less chemically stable, than those of Tm-Enc or Qt-Enc, such that both disassembly and complete unfolding occur at close to 3 M GuHCl. This potential model of encapsulin disassembly warrants further investigation, as disassembly of the nanocage utilising lower GuHCl concentrations, without completely unfolding the encapsulin protein, may provide an effective scheme for loading and/or releasing more chemically sensitive cargo.
Overall, our results for GuHCl-induced encapsulin disassembly suggest that folding plays a critical role for encapsulin nanocage assembly/disassembly. The smaller, less complex Tm-Enc structure requires more GuHCl for unfolding than the larger, more complex Qt-Enc and Mx-Enc. This agrees with the difference in the high stability of ferritin, which is a small protein nanocage comprised of 24 subunits, compared to the high sensitivity of larger and more structurally complex VLPs in denaturants. Ferritin is stable up to 6 M GuHCl39, and requires 8 M of denaturant urea for disassembly.40 A concentration of 2 M GuHCl is not strong enough to disassemble ferritin, but is able to unfold the α-helices surrounding ferritin channels, thereby enlarging its pores to allow entry of cargo without the need for complete disassembly.41 Contrastingly, the larger P22 VLP, which assembles from 420 subunits into a T = 7 nanoparticle-like structure that is 56 nm in diameter, dissociates at just 3 M GuHCl.42 However, the chemical stability of the encapsulins tested is not directly related to their size, as Qt-Enc requires more GuHCl for unfolding than the smaller Mx-Enc. Hence, other factors, such as the symmetry and stability of intermediate state(s), also needs to be taken into consideration.
pH induced encapsulin disassembly
To investigate the role of electrostatic interactions in maintaining encapsulin macrostructure, ITF and DLS were also performed for the three encapsulins at varying pH (Figure 2c & d). The theoretical pI of Tm-Enc, Mx-Enc and Qt-Enc are 4.90, 5.45 and 5.02, respectively.43 If electrostatic interactions play a major role in the mechanism of encapsulin assembly, we would expect to see disassembly of the encapsulin nanocage at pH < pI (where the overall charge on the protein would be positive rather negative, as at neutral pH). However, this is not the case. ITF indicates that all encapsulins remain relatively stable across a broad pH range, with a significant change in Trp exposure only observed in extreme alkaline conditions (pH 12-13) (Figure 2c). DLS demonstrates the presence of assembled encapsulin at all pHs, with the exception of pH 13, and smaller structures (~13 nm) at pH 12 for Mx-Enc (Figure 2d, Supplementary Table S3). Disassembly of the nanocage structures at pH 13 was also confirmed by TEM (Figure 2i). Therefore, loss of electrostatic interactions between and/or within encapsulin subunits is unlikely to be a major driving force for disassembly. Disassembly more likely arises at extreme alkaline pH due to deprotonation of the protein side chains, resulting in a loss of the hydrogen bonding that holds the protein scaffold together.44
Thermal encapsulin disassembly
To assess encapsulin thermostability, assembled samples were heated to 90°C with ITF and DLS measurements taken every 10°C up to 90°C, as well as upon return to 20°C. Additionally, encapsulins were heated to 90°C with TEM samples prepared both immediately after heating, and also after being cooled at 20°C for 1 h. ITF results showed Tm-Enc encapsulin remained stable throughout heating, with only a very slight observed red-shift in emission wavelength at temperatures above 80°C, which was recovered when cooled back to 20°C (Figure 2e). No significant change in size of the Tm-Enc nanocages was observed by DLS, and TEM images showed assembled cages after heating to 90°C, suggesting that the Tm-Enc nanocage is resistant to thermal disassembly (Figure 2f & j, Supplementary Table S4).
In contrast to Tm-Enc, both Mx-Enc and Qt-Enc disassemble at elevated temperatures. Mx-Enc Trp solvation began from 60°C and disassembled from 80°C, as shown by DLS (Figure 2e & f, Supplementary Table S4), and appeared disassembled at 90°C via TEM imaging (Figure 2j). Additionally, in contrast to Tm-Enc, disassembly of Mx-Enc by temperature is not reversible, as when cooled back to 20°C only large aggregates were found (Figure 2f, Supplementary Table S4). Similarly, ITF showed Qt-Enc began Trp solvation from 40°C and DLS indicated the majority of sample disassembled at 80°C and 90°C. However, even at 90°C, a smaller population (approx. 16%) of Qt-Enc subunits remained in intermediate structures of ~13 nm (Figure 2f, Supplementary Table S4). When cooled back to 20°C, DLS and TEM indicated Qt-Enc reassembled to its original size and ITF showed a blue-shift of emission (Figure 2e, f & j, Supplementary Table S4).
The difference in thermal stability and reassembly of Tm-Enc, Mx-Enc and Qt-Enc points to kinetic complexity in the encapsulin assembly pathway. For Mx-Enc, aggregation of the unfolded/disassembled state is likely the main obstacle to achieving proper reassembly. Our GuHCl data also suggest that the Mx-Enc unfolded/disassembled state(s) is less stable than for Tm-Enc or Qt-Enc. Whilst Qt-Enc began Trp solvation at 40°C, DLS showed it did not change in size until 80°C, indicating Qt-Enc undergoes structural change from 40°C, but remains assembled until 80°C. The stability of Tm-Enc and Qt-Enc at high temperatures are consistent with them being derived from thermophilic bacteria.6, 12 Whereas Mx-Enc is derived from bacteria that lives in soil45, which may explain its lower tolerance to temperature.
Alternate disassembly conditions
We have begun to use the ITF technique, developed herein, to investigate other chemical additives that may alter the folding and/or assembly of encapsulins, including redox conditions (10 mM DTT or H2O2) and ionic environments (1 M NaCl). However, we are yet to find another condition to induce disassembly. All Encs remained stable in NaCl, and only Mx-Enc displayed a small blue-shift in emission in H2O2, and an observable increase in diameter after incubation in either DTT or H2O2 (Supplementary Figure S4, Supplementary Table S7). This may be due to reduction/oxidation of disulfide bonds causing a weaker association between subunits, resulting in swelling of the nanocage.46
Encapsulin Reassembly
Based on the above results, high concentrations (6 M) of GuHCl and pH 13 conditions were selected for further analysis of the Tm-Enc, Mx-Enc, and Qt-Enc reassembly mechanism. Notably, these conditions disassemble encapsulin to differing degrees. ITF revealed maximum peak emissions for disassembly with pH 13 to be ~341 nm for Tm-Enc, and ~347 nm for Mx-Enc and Qt-Enc, whereas disassembly with 6 M GuHCl resulted in a maximum peak emission at ~354 nm for all encapsulins (Figure 3a–c). As previously discussed, a Trp emission maxima ≥350 nm (as observed for all encapsulins in 6 M GuHCl) is expected for unfolded protein. However, in pH 13, all encapsulins display a Trp max emission < 350 nm, suggesting that at least some of their subunits’ secondary structure remains intact.
Reassembly of all encapsulins after disassembly in either 6 M GuHCl or pH 13 was initiated via overnight dialysis into reassembly buffer. Following disassembly in 6 M GuHCl, Tm-Enc, Mx-Enc, and Qt-Enc all reassembled back to their original structures, as indicated by ITF maximum emission and DLS size (Figure 3a–c, Supplementary Figure S3). In contrast, after disassembly in pH 13, only Tm-Enc reassembled to its original structure. For Mx-Enc and Qt-Enc, a blue-shift in ITF emission, resembling the Trp maximum emission of the original self-assembled material, was observed. However, native-PAGE results showed no bands (Supplementary Figure S3), and TEM analysis found no structures for Mx-Enc, and only a single nanocage was imaged for Qt-Enc (Figure 2i). Indeed, DLS analysis of pH 13 reassembled Mx-Enc and Qt-Enc showed the presence of either large aggregates or smaller structures (~17 nm), which suggests the subunits may be reassembling into smaller cages and/or intermediate structures at a concentration too low to detect by native-PAGE and TEM.
As previously discussed, at pH 13, hydrolysis of the peptide bonds can occur, leading to protein misfolding and/or aggregation44, which may account for some of this inefficiency in encapsulin reassembly. Interestingly, although the position of the Trp emission peak for the original self-assembled, GuHCl reassembled and pH 13 reassembled samples of Qt-Enc are not significantly different, a difference in spectral shape is noted for the pH 13 reassembled sample. The blue-shifted shoulder of the Qt-Enc Trp emission peak, thought to correspond to Trp95 buried in the pentamer/hexamer interface of the intact nanocage, is lower in intensity for the pH 13 reassembled sample. This result is consistent with reassembly of a different structure with a native-like fold, instead of the original intact T = 4 Qt-Enc nanocage, upon reassembly from pH 13. The reassembly of Mx-Enc and Qt-Enc after pH 13 disassembly into smaller nanocages and/or alternative structures in inefficient quantities suggests this pathway, in its native form, may not be suitable for some applications.
Timescale of encapsulin reassembly
To advance encapsulins as a cargo-carrying platform, disassembled encapsulin needs to be able to be reassembled on a viable timescale. The encapsulin nanocages were disassembled for 1 h via 6 M GuHCl or pH 13 before reassembly was initiated using dialysis, with samples measured with ITF and DLS every 15 min over a 75 min timeframe. The rate of reassembly appeared to vary between each Enc. Tm-Enc reassembled faster than Mx-Enc and Qt-Enc, with DLS indicating Tm-Enc reached its assembled diameter within 15 min after being disassembled via pH 13, and between 15–30 min when disassembled via 6 M GuHCl (Figure 4a & b, Supplementary Table S5 & S6). The faster reassembly after pH 13 disassembly compared to GuHCl may be explained by the subunits still maintaining some structure after incubation in pH 13 conditions. However, Tm-Enc reassembly after pH 13 disassembly displayed a more gradual Trp burial and a slight decrease in size at 45 mins (Figure 4a & b, Supplementary Table S6). This suggests reassembly after pH 13 may follow a more dynamic pathway, with rapid nanocage formation, followed by restructuring. In contrast, after 6 M GuHCl disassembly, Qt-Enc reassembled into its original diameter between 30–45 min, and Mx-Enc appeared to partially assemble between 45–60 min, but only completely assembled between 60-75 min (Figure 4a & b, Supplementary Table S5). The difference in rates between encapsulins may reflect the overall stability of the complex i.e., complexes that are more stable may reassemble quicker. In a recent report, the in vivo loading of cargo proteins into Tm-Enc during its self-assembly was found to be ~8 times less efficient than with Mx-Enc.47 The authors suggested that this striking difference was partly due to the smaller Tm-Enc self-assembling at a faster rate than Mx-Enc, thus limiting the available contact time between ESig-tagged cargo and the nanocage’s interior surface. This hypothesis is now supported by our observation that Tm-Enc reassembles up to four times quicker than the larger Mx-Enc (Figure 4). This therefore highlights the critical role assembly time-frames play in cargo-loading efficiency, and that modifying such properties can enhance cargo packing, density and/or stoichiometry.
The rapid burial of Trp residues for all encapsulins within 15 min, as shown by ITF, prior to formation of the cages, suggests a reassembly pathway where subunits first fold into an intermediate structure before gradually forming the cage. A previous study on encapsulin from B. linens (T = 1) suggested reassembly occurred with subunits first forming stable dimers prior to the final nanocage formation, with a preference for even numbered stoichiometries as demonstrated via mass spectrometry analysis.38 Additionally, ferritin has been found to have a biphasic reassembly, where an initial fast step occurs with folding of subunits and unknown stable intermediates, followed by the slower restructuring of intermediates into the nanocage.23 Data of pH 13 reassembly rate for Mx-Enc and Qt-Enc was not included, as the presence of aggregation prevented extraction of accurate DLS size values. This suggests that reassembly from the pH 13 disassembled state for the larger and more complex Mx-Enc and Qt-Enc may be prone to protein misfolding pathways and/or partial assembly, and thus, without additional re-engineering, may not be ideal for use in future applications.
Nanomechanical stability of self-assembled vs reassembled encapsulins
To understand any changes in the structural integrity of Tm-Enc, Mx-Enc, and Qt-Enc after reassembly in the conditions tested above, AFM was utilised to compare the morphology, rupture point, and elasticity between self-assembled and reassembled encapsulins. Previous AFM analysis has been done on the rigidity of Tm-Enc and encapsulin from B. linens (T = 1), and suggested that the presence of cargo within encapsulin from B. linens may lead to some destabilisation, as indicated by a lower rupture force in loaded versus unloaded encapsulin.38 However, the effect of reassembly on the nanomechanical stability on Tm-Enc, as well as the larger and more structurally complex Mx-Enc and Qt-Enc, is yet to be elucidated.
Individual encapsulins were imaged with a scan size of ~100 nm × 100 nm to visualise their topographical detail. Self-assembled encapsulins were found to be at their expected size, including both T = 3 and T = 1 conformations of Mx-Enc (Figure 5a). However, clear morphological variation can be seen after reassembly, with all encapsulins demonstrating flattening of their surfaces. Additionally, Qt-Enc appeared more aggregated after reassembly.
Being composed of similar constituents, approximately the same elasticity would be expected between each self-assembled encapsulin, which was found to be the case of Tm-Enc and Mx-Enc, with an average elasticity of 2.77 ± 0.96 and 2.12 ± 0.81 MPa, respectively (Figure 5b, Supplementary Figure S5). However, Qt-Enc displayed a higher average elasticity of 25.60 ± 14.92 MPa, indicating it has greater structural integrity and is therefore more resilient to deformation (Figure 5b, Supplementary Figure S5). Qt-Enc may therefore be more strongly self-assembled than Tm-Enc and Mx-Enc, which could be attributed to its higher structural complexity, as the different symmetries and subunit-subunit contacts may influence the strength of interfacial interactions.48 Upon reassembly following 6 M GuHCl disassembly, the elasticity of all encapsulins decreased significantly, with Tm-Enc average elasticity 1.16 ± 0.93 MPa (58.12% decrease), Mx-Enc 1.39 ± 0.78 MPa (34.43% decrease), and Qt-Enc 0.85 ± 0.62 MPa (96.68% decrease), indicating a significant reduction in structural integrity (Figure 5b, Supplementary Figure S5). Elasticity data of reassembled Tm-Enc after disassembly via pH 13 was not included, as, during testing, the sample became unstructured and did not retain any elasticity or show any elastic response. This suggests that even though Tm-Enc reassembled after disassembly via pH 13, these conditions may have irreversibly altered the protein by hydrolysis, thereby affecting stability.44
The rupture force represents the maximum force a protein cage withstands prior to puncture by the AFM tip.49 The rupture force of all self-assembled encapsulins were between ~0.1-0.2 nN, indicating they are fairly fragile (Figure 5b). This is in contrast to AFM results from a previous study on self-assembled empty Tm-Enc and encapsulin from B. linens, which had rupture forces of 0.63 and 0.64 nN, respectively38, indicating that the encapsulins used in the current study are slightly more fragile. After reassembly, the rupture force decreased further by a factor of ~5 for each encapsulin system tested (Figure 5b).
Together, these results show that, after undergoing disassembly in either 6 M GuHCl or pH 13 and subsequent reassembly, all three encapsulin architectures significantly lose structural integrity. In terms of encapsulins loaded with cargo under in vitro conditions, the observed increased fragility could be beneficial if wanting a system that can be easily broken down to release cargo, such as intracellular drug delivery or immunotherapy. If in need of a more robust system (e.g., enzyme-loaded nanocages for industrial biocatalysis), however, these results demonstrate the importance to understand and modify encapsulins so they can disassemble in more benign conditions. Furthermore, to enhance understanding of the reassembly pathway of encapsulins, future studies could assess real-time assembly of the subunits. This has been achieved with large bacterial microcompartments, where AFM revealed that the shell facets assemble from preformed oligomers.50
Conclusion
In summary, this study characterised the conditions facilitating the disassembly and reassembly of Tm-Enc, Mx-Enc, and Qt-Enc, with each encapsulin system varying in degree of sensitivity to GuHCl, pH, and temperature. When disassembled in high pH conditions, the subunits appeared to individually remain structurally intact, as opposed to the complete unravelling observed when disassembled in high concentrations of GuHCl. Furthermore, Tm-Enc, Mx-Enc, and Qt-Enc were all able to reassemble within 75 min after disassembly with GuHCl; however, only Tm-Enc was able to reassemble after disassembly at pH 13 (within 15 min). Our results suggest that disassembly by GuHCl and pH 13 affect different interfacial interactions. Self-assembly, disassembly, and reassembly of protein nanocages are highly dependent on interfacial subunit interaction17, and therefore, further understanding of these interactions is a fundamental step required to harness their potential. Furthermore, using AFM, we revealed a previously unknown effect of in vitro disassembly/reassembly on encapsulin stability, finding that encapsulin loses a substantial amount of structural integrity when disassembled and subsequently reassembled. These findings have implications if requiring in vitro loaded encapsulin for a particular application, but may be beneficial when nanocage dissociation and cargo release is favored.
In high concentrations of GuHCl, all encapsulins completely unfold and refold efficiently; however, these harsh conditions pose limitations for applications involving in vitro loading of cargo that may be sensitive to these conditions, as it could be destroyed in the process. Furthermore, the need to use high concentrations of GuHCl or alkaline pH in order to achieve disassembly may limit use for biomedical applications, where biocompatibility is paramount. Therefore, understanding the disassembly/reassembly conditions is an important step to elucidate target regions within encapsulin subunits that can be engineered to allow nanocage disassembly to occur under milder conditions. For example, the elucidation of ferritin disassembly/reassembly conditions has allowed for targeted amino acid engineering to enhance its potential as a nanocarrier for pH-sensitive compounds. As previously discussed, ferritin requires strong acidic conditions (pH ≤ 2) to disassemble, and molecules sensitive to these conditions are unable to be loaded. Therefore, by removing the last 23 amino acids from the carboxyl terminal, researchers found that ferritin was able to disassemble at pH 4 and reassemble again at pH 7, with the ability to load the bioactive agent curcumin and maintain its regular structure.24
Engineering interfaces can also allow enhanced control over size and cargo-loading efficiency of protein nanocages. For instance, due to being disassembled under high acidic conditions, a lot of ferritin is lost during this process and therefore it has a low loading efficiency. Consequently, ferritin variant subunit interfaces were engineered to display His6 motifs, which allowed nanocages to reassemble in the presence of either transition metal ions or pH 10, which increased loading efficiency by 1.6-3.6 times compared with previous acid disassembly.51 Due to ferritin being limited to one size, researchers engineered the subunit interface by first identifying the key subunit interfaces and removing the amino acids that do not participate in the interfacial interactions, resulting in the formation of a 17 nm 48-er nanocage instead of the 12 nm 24-er native ferritin structure.52 With an increased understanding of encapsulin structure, disassembly/reassembly conditions, molecular mechanisms, stability, and stoichiometry of intermediate structures, future genetic engineering efforts of subunits may similarly lead to enhanced control over conditions that trigger disassembly/reassembly, size, structure, and cargo-loading efficiency of encapsulins.
In summary, the findings of this study advance our understanding of encapsulins by providing critical insight into their unique disassembly/reassembly dynamics. This knowledge provides a roadmap towards an encapsulin “tool-kit” comprised of nanocages with varying structural architectures and biochemical/biophysical properties, that can be readily selected and further customised for a specific nanobiotechnological application.
ASSOCIATED CONTENT
Supporting Information
List of genetic constructs, examples of purification chromatographs, gel images of purified encapsulins, molecular graphics of Trp residue positions within encapsulin subunits, gel images of disassembled/reassembled encapsulins, stability study of encapsulins in REDOX and ionic condition, AFM elasticity histograms, and DLS measurement tables.
AUTHOR INFORMATION
Corresponding Author Andrew Care Email: andrew.care{at}uts.edu.au
Author Contributions
I.B co-designed the research, generated all nanocage constructs, conducted all disassembly/reassembly characterisation work, performed data analysis and wrote the manuscript. S.G assisted in experimental design, data interpretation, and wrote the manuscript. D.D assisted in protein production and purification, TEM imaging, protein modelling, and revised the manuscript. A.E performed AFM experiments and data analysis. L.C.P supervised the project and revised the manuscript. A.C. conceptualised and co-designed the study, supervised the project, and wrote the manuscript.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
I.B. is supported by a Dementia Australia Research Foundation PhD scholarship. A.C. is supported by the UTS Chancellor’s Postdoctoral Research Fellowship scheme. This work was supported by grants to A.C and L.C.P by the Dementia Australia Research Foundation, the National Foundation for Medical Research and Innovation, the Brain Foundation and the NeuroSurgical Research Foundation.