Nature offers a vast array of biological building blocks that can be combined with synthetic materials to generate a variety of hierarchical architectures. Viruses are particularly interesting in this respect because of their structure and the possibility of them functioning as scaffolds for the preparation of new biohybrid materials. We report here that cowpea chlorotic mottle virus particles can be assembled into well-defined micrometre-sized objects and then reconverted into individual viruses by application of a short optical stimulus. Assembly is achieved using photosensitive dendrons that bind on the virus surface through multivalent interactions and then act as a molecular glue between the virus particles. Optical triggering induces the controlled decomposition and charge switching of dendrons, which results in the loss of multivalent interactions and the release of virus particles. We demonstrate that the method is not limited to the virus particles alone, but can also be applied to other functional protein cages such as magnetoferritin.
Multivalent interactions are ubiquitous throughout nature and are used to create materials that are unparalleled with regard to properties and function1,2. Synthetic materials that interact with biological target molecules are of paramount importance for the development of various biotechnological applications3,4,5. Materials that have multiple binding ligands organized on a single molecular unit are particularly useful, because they may show enhanced binding with target elements due to the effect of multivalency6,7. Branched molecules, such as dendrimers and dendrons, offer an inherently versatile synthetic scaffold with which to construct surfaces with a high density of binding ligands, allowing strong binding affinities for biomolecules8. For example, the interaction of a fourth-generation polylysine dendrimer that is functionalized on its surface with multiple copies of naphthalene disulphonate surface groups (VivaGel®) has been shown in preclinical trials to prevent human immunodeficiency virus (HIV) infections9,10,11.
We are interested in developing temporary binding platforms12,13,14 that are able to adhere on the surface of virus and other protein cage-based particles. This approach is interesting from a medical point of view15,16,17,18, and also from a materials research perspective19,20,21,22,23,24,25,26,27,28,29,30. Controllable high-order structures based on virus capsids have significant potential in nanofabrication and fundamental studies of metamaterials with three-dimensional periodic nanoscale architecture and also the electromagnetic properties originating from such periodicity, interactions and collective behaviour31,32,33,34,35. Furthermore, cationic polymers have been shown recently to enhance diffusion-limited transduction of retroviral particles by charge shielding and virus aggregation25. Such aggregates may prove useful in promoting the viral delivery of therapeutic agents such as chemotherapeutic taxol36 or nucleic acids for gene therapy37,38,39.
Here, we present a self-assembly and photo-triggered disassembly method based on electrostatic interactions to assemble viruses into hierarchical supramolecular complexes. This method may enable the controlled release of virus particles that hold compounds or materials within the capsid interior. We use the cowpea chlorotic mottle virus (CCMV), which is an icosahedral (Caspar-Klug triangulation number T = 3) assembly of 180 identical coat protein subunits that encapsulate the viral RNA. The outer and inner diameters of the virus are ∼28 and ∼18 nm, respectively40. An interesting feature of the CCMV capsid is that it shows reversible pH-dependent assembly/disassembly behaviour. This assembly mechanism can be used to encapsulate compounds and materials into the spatially confined interior of the virus23.
Our approach is schematically presented in Fig. 1. The dendrons are composed of a Newkome-type scaffold functionalized with a polyamine (spermine) surface that is positively charged at physiological conditions (pKa amines > 8)13,41. The isoelectric point (pI) of the CCMV is ∼3.8 and therefore the outer surface of CCMV is negatively charged42,43 at pH ≥ 5. Attractive electrostatic interactions between these two oppositely charged weak polyelectrolytes are therefore expected in the pH range 4–8, leading to the formation of a hierarchically ordered self-assembled complex29,30. Disassembly of the complex is based on the photochemical degradation of the dendron, and the consequent release of virus particles can take place as a result of two factors. First, the cationic polyamine groups, which are attached to the dendritic scaffold through an o-nitrobenzyl linker, undergo photocleavage when exposed to long-wavelength (λ ≈ 365 nm) ultraviolet (UV) light. This optically triggered release of the cationic surface groups destroys the multivalent supramolecular interactions between the virus and the dendron. Second, the cleavage reaction creates negatively charged carboxylate functions on the surface of the dendron that repel the negatively charged virus particles. Use of light as an external stimulus has special advantages, such as being highly orthogonal, quick and easy to apply, but, most importantly, it allows precise spatiotemporal control over the release event.
To validate the initial assembly of the CCMV–dendron complex, we first conducted dynamic light scattering (DLS) measurements. Figure 2a shows typical DLS profiles recorded during titration of G1 (2) into a buffered (pH 5, 10 mM NaCl) CCMV solution. The measured hydrodynamic radius (Dh) of CCMV is 31.1 nm, which corresponds well with the outer diameter of the virus (28 nm). A slight increase in the observed value is to be expected because of the presence of a water layer surrounding the virus. Adding increasing amounts of G1 (2) to the virus solution led to the gradual disappearance of the free virus and the formation of a larger secondary virus assembly, which increased in size with an increase in dendron concentration. The size of the secondary assembly increased to result in a Dh value of 717 nm at a G1 (2) concentration of 7 mg l−1. The same data are plotted in a different way in Fig. 2b, that is, as a percentage of the peak corresponding to free virus, which decreases with increasing dendron concentration. The hydrodynamic radius of the corresponding secondary assembly is plotted on the secondary axis, which, as expected, increases when the dendron concentration becomes higher. A similar trend is observed for dendrons G0 (1) and G2 (3), and the combined results obtained with the three different dendrons clearly point to a multivalency effect (Fig. 2b). In line with this, dendron G0 (1) is unable to fully complex the virus particles even at very high concentrations (100 mg l−1). It is of interest to note that the amount of G2 (3) required to fully complex the free virus is significantly lower than the amount needed in the case of G1 (2). Furthermore, micrometre-sized assemblies are observed with the former dendron at a concentration of only 1 mg l−1. At NaCl salt concentrations of 150 mM, G0 (1) completely loses its affinity for the virus and the affinity of G1 (2) is significantly decreased (Fig. 2c). In contrast, the multivalent G2 (3) can effectively compete with Na+ ions and induce the formation of dendron–virus complexes (size 600–800 nm) even at a low concentration of 6–9 mg l−1. This emphasizes the dendritic effect of binding—multiple binding units organized on a single subunit clearly enhance the interaction with the virus. Similar titrations were also conducted in the absence of salt and at 300 mM NaCl concentration. These experiments also showed that increasing salt concentration decreases the binding affinity of dendrons, in line with the experiments above (Supplementary Fig. S1 and Table S1). The assembly of the dendron–virus complex occurs immediately after adding the dendron. The mean count rate detected with DLS increases within seconds after dendron addition, whereas the size of the secondary assembly continues to grow in time and reaches a plateau within ∼5 min (Supplementary Fig. S2).
Variable-temperature circular dichroism (CD) experiments were performed to determine the effects of dendron binding on the virus surface structure. Figure 2d presents the CD spectra of CCMV (pH 5, 150 mM NaCl) obtained between 20 and 90 °C. At 20 °C the CD spectrum shows a distinct negative absorbance at 208 nm, which is characteristic for the virus. When the temperature is increased to 90 °C, the virus is almost fully denatured and the CD signal is lost. In a subsequent series of experiments the three different dendrons were separately added to a solution of CCMV, and the loss of the CD signal was monitored at 208 nm as a function of temperature between 20 and 90 °C (Fig. 2e). Melting temperatures (Tm) were obtained from the maximum of the first derivative of the melting curves (data not shown). Blank CCMV showed a Tm of 77 °C. As expected, the addition of G0 (1) (5 mg l−1) had no effect on the melting curve due to the weak interaction with the virus surface and, as a result, Tm remained unchanged at 77 °C. The addition of G1 (2) (5 mg l−1) had a minor effect on the Tm value, which decreased to 74 °C. The thermal stability of the virus, however, turned out to be significantly affected by the presence of G2 (3). The addition of G2 (3) (5 mg l−1) decreased the Tm value by 15 °C to 62 °C, indicating strong interactions with the non-native conformation of the coat proteins. As the thermal denaturation of CCMV is not fully reversible and the binding of dendrons leads to the formation of larger assemblies, the melting curves also reflect the aggregation behaviour and solubility of the non-native form of the virus. Nonetheless, the observed differences are dependent on the generation (Fig. 2e) and concentration of the dendrons, as well as salt concentration (Supplementary Table S2). These results are in good agreement with the DLS measurements.
Gel electrophoresis and ζ-potential measurements were used to further demonstrate CCMV–dendron complex formation and disassembly (Fig. 2f). Surface-charge neutralization of the virus and formation of larger assemblies could be observed as a reduction of the mobility in the gel electrophoresis experiment. Free CCMV (150 mg l−1) is shown in lane 1. The addition of G0 (1) (65 mg l−1) to the virus did not lead to any changes in the electrophoretic mobility (lane 2). However, when CCMV was complexed with G1 (2; 65 mg l−1), no clear band corresponding to the free virus was visible (lanes 3 and 4). Instead, smearing towards the anode was observed, indicating the formation of complexes that are still able to partially penetrate the gel. Compound G2 (3) (65 mg l−1) formed a strong positively charged complex with the virus particles and hence also efficiently prevented migration of the free CCMV into the gel (lane 4).
Disassembly of the hierarchical CCMV–dendron complex was studied by exposing the CCMV–dendron complexes to UV light and following the release of the free virus particles using gel electrophoresis, DLS and transmission electron microscopy (TEM). Gel electrophoresis demonstrated that after UV exposure the CCMV complex with G1 (2) and G2 (3) clearly released virus particles, allowing them to freely migrate into the gel with a similar electrophoretic mobility and surface charge as the native CCMV particles (Fig. 2f, lanes 6 and 7). UV exposure did not affect the mobility of the G0 (1)–virus complex (lane 5). Figure 3a,b shows the DLS profiles of CCMV–dendron complexes (G1 (2) or G2 (3)) before and after exposure to 60 s of UV light. Before UV exposure the CCMV–dendron assemblies were over 600 nm in size and relatively monodisperse (polydispersity index, PDI < 0.25). One minute of UV exposure seem to be enough to break the assemblies, which allowed the peak corresponding to free CCMV at ∼33 nm to be observed again. The assembly and disassembly cycle can be repeated several times by the addition of fresh dendron following UV exposure (Supplementary Fig. S3). Importantly, virus assemblies formed with compound A as a control (a previously reported first-generation dendron without the photolabile linkers41) were inert to UV exposure, and the free virus was not observed afterwards (Fig. 3c). TEM was used to image the free virus before and after UV exposure (Fig. 3d,f) and to characterize the morphology of the CCMV–dendron complexes (Fig. 3e). Negatively stained CCMV before complexation was visible as monodisperse spherical objects with an average diameter of 27.9 ± 1.0 nm. Following the addition of G1 (2) to the CCMV solution, large assemblies with diameters of several hundred nanometres were observed (Fig. 3e, left inset). A larger magnification image revealed well-ordered arrays of CCMV with apparent hexagonal packing, as shown in Fig. 3e (see Supplementary Fig. S4 for the G2–virus complex). The size of the individual viruses in the complex (27.1 ± 1.2 nm) corresponded well with the size of the native CCMV. Similar complexes in which the viruses appeared as dark spheres are also readily observed on the same grid (Supplementary Fig. S4). After UV exposure, virus assemblies were no longer observed, indicating that the photolytic reaction had taken place, thus allowing free virus with a diameter of 27.6 ± 1.0 nm (Fig. 3f) to be observed again.
The successful assembly and disassembly of the CCMV–dendron complex also suggest that other negatively charged protein cages could be formed and tuned in the same manner. To demonstrate the generality of the concept, the assembly of magnetoferritin (MF) was studied44,45. MF has an apo-ferritin cage, which consists of 24 protein subunits that form a globular and hollow protein cage to host synthetic superparamagnetic magnetite/maghemite (Fe3O4–γ-Fe2O3) particles. The outer diameter of the shell is ∼12 nm and the inner cavity size is 8 nm. The pI of ferritin is in the range 4.3–4.6 and its assembly with the dendron relies on the same electrostatic principle as assembly with CCMV. MF is an interesting building block because of its magnetic properties34, which may find useful applications in magnetic sensory devices such as magnetic relaxation switches46.
Assembly of MF was studied with G1 (2). Figure 4 presents the volume-averaged DLS and TEM results for free MF and the MF–G1 complex before and after UV exposure. DLS clearly shows that the individual MF units can be assembled into large complexes (Dh: 870 nm), which can be subsequently disassembled back into individual particles by UV exposure (Fig. 4a). The elevated PDI values for free MF are to be expected, because in a low-ionic-strength medium the ferritin cages are known to interact with themselves, which is observed as partial spatial ordering47. TEM images complement and confirm the DLS results and show that MF is also hexagonally packed in the assembly (Fig. 4b–d). Overall, the interaction of MF with the dendrons is in good agreement with that of CCMV, demonstrating the versatility of the approach. The reversible self-assembly of similar superparamagnetic iron oxide nanoparticles into stable complexes has previously been demonstrated to enhance the spin–spin relaxation times of water protons, which can be used, for example, to improve contrast in magnetic resonance imaging or to detect molecular interactions with magnetic relaxation measurements46,48.
In conclusion, we have presented a straightforward and previously unused method for the optically reversible self-assembly of protein cages into large ordered architectures by using cationic dendrons that bind to the negatively charged cage surface. Facile control of the assembly size can be achieved by adjusting the generation and concentration of the dendron as well as the salt concentration. Such oppositely charged polyelectrolyte complexes are generally considered to be kinetically frozen systems in which the mobility of individual polyelectrolyte components is limited. In our approach, however, we use optically triggered destruction of the multivalent binding interactions as a tool to release the protein cages from their complexes. Development of assembly–disassembly procedures for such three-dimensional structures is an important prerequisite for the preparation of reversible colloidal architectures, which may find applications as virus-based optical metamaterials27 or bioengineered magnetic crystals49. Furthermore, the focal point of the dendron offers a site that can be easily modified with functional molecules, which allows these materials to be organized on the surface of protein cages through supramolecular interactions. Finally, this procedure allows compounds and materials that are encapsulated inside the protein cage to be assembled together with the protein shell50. These complexes can be disassembled by using external stimuli and released at selected locations, thereby opening new routes for the controlled assembly, delivery and release of compounds.
All reagents were commercially available and used as supplied without further purification. The production and purification of CCMV were carried out according to literature procedures51,52. CCMV was characterized by fast protein liquid chromatography, UV–vis spectroscopy and TEM24. Compounds G0 (1), G1 (2) and G2 (3) were synthesized as previously reported13. MF was prepared, magnetically separated and purified following a literature procedure49.
Dynamic light scattering
DLS analyses were carried out with a Zetasizer Nano S from Malvern Instruments at 25 °C. Results were the average of at least five measurements. All CCMV samples were prepared in filtered acetate buffer (10 mM NaAc, 1 mM EDTA, 1 mM NaN3 with 0, 10, 150 or 300 mM NaCl, pH 5). CCMV solution (1 ml, 40 mg l−1) was titrated with a polyamine solution (G0 (1), G1 (2) or G2 (3)) in aqueous solution (0.001–10 mg ml−1depending on the compound and measurement). The added polyamine solution did not exceed 5% of the total volume, so no corrections were made for sample dilution. After each titrant addition, the samples were thoroughly mixed and allowed to equilibrate for 2 min. If the dendron was added at once to fully complex the sample, the samples were allowed to equilibrate for at least 5 min. For MF measurements, a stock solution of MF was diluted with filtered MilliQ water to a concentration of 60 mg l−1, which was subsequently complexed with G1 (2) (10 mg l−1).
CD spectra of the thermal scans of CCMV were recorded in the UV region (205–300 nm) using a JASCO J-810 spectropolarimeter equipped with a JASCO PFD-425S temperature control unit by the accumulation of 20 spectra. Thermal scans were carried out with 80 mg l−1 CCMV and different dendron concentrations at 208 nm with the following settings: heating rate, 1 °C min−1; resolution, 1 nm; data pitch, 1 °C; integration time, 8 s. A 1 ml quartz cuvette with a 4-mm light path and buffered water (10 mM NaAc, 1 mM EDTA, 1 mM NaN3 with 0, 10 or 150 mM NaCl, pH 5) were used for all measurements. Melting temperatures (Tm) were obtained from the maximum of the first derivative of the melting curves.
Agarose gel electrophoresis and ζ-potential
Agarose gels were prepared from a solution of 1.2% agarose in dilute acetate buffer (10 mM NaAc, 1 mM EDTA, 1 mM NaN3, pH 5) and stained with 10 µl of ethidium bromide solution (10 mg ml−1). Samples contained CCMV (150 mg l−1) and the dendron (65 mg l−1) in 10 mM NaAc, 1 mM EDTA, 1 mM NaN3, 10 mM NaCl, pH 5 buffer. A volume of 1 µl of 5× nucleic acid loading dye (BioRad) was added to each sample. A sample of the solution (12 µl) was run at 100 V for 30 min. The UV light irradiation time was 3 min. The ζ-potential of the same sample solution (∼0.5 ml) was determined with a Nano-ZS Zetasizer from Malvern Instruments at 25 °C. Results were the average of three measurements from duplicate samples.
Transmission electron microscopy
TEM micrographs were recorded on a JEOL JEM-1010 instrument. Samples were prepared on Formvar carbon-coated copper grids (Electron Microscopy Sciences) by placing a 5 µl drop of a solution containing free CCMV (40 mg l−1), free MF (60 mg l−1), CCMV (40 mg l−1)–G1 (8 mg l−1) complex or CCMV (40 mg l−1)–G2 (1 mg l−1) complex before or after UV exposure. The sample drop was left on the grid for 1 min, after which time the excess buffer was blotted away with filter paper. Samples were negatively stained by applying 5 µl of stain (1% uranyl acetate in MilliQ water) onto the grid and removing the excess stain away after 15 s with filter paper. The samples were dried under air flow for 5 min before imaging. Each size distribution histogram was plotted based on more than 80 particles.
UV light irradiation
Samples were irradiated in cuvettes placed over an ice bath for the given time periods using a UVA Spot 400T lamp from Hönle UV Technology.
This work was supported by the Netherlands Organization for Scientific Research (Vidi grant to J.J.L.M.C. and top grant to R.J.M.N.), by the European Research Council (Euryi grant to J.J.L.M.C.) and by the Royal Netherlands Academy of Science (endowed chair to R.J.M.N. and Beijerink award to J.J.L.M.C.). O.K. was supported by the Engineering and Physical Sciences Research Council UK. M.A.K. was supported by the Academy of Finland, Instrumentarium Science Foundation and the Alfred Kordelin Foundation.