Self-Assembly of Switchable Protein Nanocages via Allosteric Effect

Abstract

Open AccessCCS ChemistryCOMMUNICATION1 Aug 2021Self-Assembly of Switchable Protein Nanocages via Allosteric Effect Miaomiao Xu, Rongjin Zeng, Jun Xiang and Qiang Yan Miaomiao Xu Department of Macromolecular Science, Fudan University, Shanghai 200433 Google Scholar More articles by this author , Rongjin Zeng Department of Macromolecular Science, Fudan University, Shanghai 200433 Google Scholar More articles by this author , Jun Xiang Department of Biomass Science and Engineering, Sichuan University, Chengdu 610065 Google Scholar More articles by this author and Qiang Yan *Corresponding author: E-mail Address: [email protected] Department of Macromolecular Science, Fudan University, Shanghai 200433 Google Scholar More articles by this author https://doi.org/10.31635/ccschem.020.202000437 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail Protein cages have promising applications in highly efficient gene transportation. Many methods of engineering protein–interface interactions have been developed to build such viral mimics. Missing in current approaches is how to turn off the interactions and unpack the protein shell for controlled cargo delivery. Here, we present a concise protein allosteric strategy to create protein nanocages with dismountable coats. This strategy highlights that the precise self-assembly of coat protein only relies on allosterically driven protein–ligand interactions rather than conventional oligomeric protein fusion. Utilizing common calmodulin (CaM) as an allosteric protein element, its Ca2+-mediated conformational interconversion between folded and unfolded states can activate and deactivate its recognition of a specific ligand, thus triggering switchable assembling and uncoating of the protein–ligand supramolecular cages. Moreover, the size, geometry, and architecture of these protein cages can be tailored by modulating the ligand chemistry. Since this new strategy for bottom-up construction of protein cages requires neither protein engineering nor protein interfacial design, we envision that it has potential applications in other allosteric protein families and will offer new vistas on building dynamic and smart virus mimics. Download figure Download PowerPoint Introduction Protein-based nanocontainers continue to attract the interest of chemists since they are considered as ideal self-assembled mimics of natural viral structures and functions.1 As a class of emerging bionanovectors, they offer tantalizing opportunities for targeted cargo delivery,2,3 nanoenzyme intensification,4,5 and biotechnological applications.6,7 To date, intensive efforts have been undertaken in the bottom-up fabrication of such virus-like particles. One classic method is to use directly or reorganize naturally occurring viral capsids (e.g., cowpea chlorotic mosaic virus and hepatitis B virus)8–11 and cage-like protein entities (e.g., ferritins, chaperones, and synthases).12–15 This method is simple and straightforward, but achieving flexible protein functionalization is limited. Another emerging method is to use nonviral or nonnative coat proteins as basic building blocks to de novo assemble protein cages through their protein–protein interactions and oligomeric protein fusion.16–22 Besides protein interactions, other supramolecular forces such as metal–protein binding can also drive the controlled assembly and disassembly of protein cages. Most recently, the Heddle and Tezcan groups23–25 have used different metal–protein interactions to assemble metal-directed ultrastable protein cages. This method allows us to on-demand tailor the size, symmetry, architecture, and even surface chemistry of protein cages; however, it necessitates arduous work on computation-assisted protein interface engineering and precise protein symmetry-oriented design.26,27 Moreover, natural viruses feature the programmed capsidation and decapsidation to protect and transport internal genetic material under the biological signal control, but thus far this process has not been fully explored and replicated yet. To address these two unmet issues, not only building protein cages with self-assembly tunability and dynamics but also circumventing complicated protein engineering and postmodification, a possible thought is to use the conformational adaptability of native protein itself. Almost all artificial virus particles adopt structurally conserved proteins as the capsid-forming units, but this restricts the flexibility of the protein shell and involves multistep protein engineering. For this dilemma, we imagine that a particular family of proteins, called allosteric proteins, may solve it.28,29 Protein allostery, as a ubiquitous effect in biology, refers to a remote interaction whereby a ligand binding at one site on a protein can largely change its conformation at another site. It plays a decisive role in biological recognition, cell signaling transduction, and transmembrane transport, and has been applied in various fields, including drug design, enzyme evolution, and biomedicines.30–33 However, harnessing this allosteric effect to direct and modulate protein self-assembly is still in its infancy.34–36 With the aid of this conformational conversion, there is an opportunity to create a dynamic protein container by flexible protein–ligand interactions and totally avoid the protein surface design and engineering process. In this study, we present a minimalist strategy that only exploits native allosteric proteins to construct a switchable protein nanocage. Control of the protein conformational transition by an allosteric signal (allosteric regulator) to turn the protein–ligand interactions on and off can enable a reversible assembly and detachment of the nanocages for internal cargo release. Moreover, we expect that the structural parameter of self-assembly of these protein cages can be precisely regulated by simple ligand chemistries instead of the sophisticated protein symmetry and interface design. To achieve this proof-of-concept, a common allosteric protein, calmodulin (CaM), is selected as the protein building block, and a phenothiazine (PT)-capped collagen-like peptide (CLP) carrying repeated (Proline-Hydroxyproline-Glycine)n tripeptide, PT(POG)n (n = 6, 10, and 14), is designed as the ligand. CaM is a calcium-dependent protein that can shape the orientation of its two binding pockets (N-/C-terminal domains, Figure 1a).37 By folding these two domains with Ca2+, we postulated that CaM could constrict its conformation to “clamp” the specific PT moiety,38 inducing the formation of protein–ligand complexes, CaM-PT(POG)n. Such a complex is structurally similar to a supramolecular amphiphile,39 but a protein-based macromolecular version. This CaM-PT(POG)n supramolecular amphiphile can further self-organize into a CaM-coated protein cage in aqueous media. Upon removal of Ca2+ allosteric signal, a reversible CaM conformational unfolding would release the ligand and elicit the cage dissociation, which resembles the responsive uncoating behavior of natural viruses (Figure 1b). Since these protein nanocages are mainly framed by protein–ligand interactions, their size, architecture, and geometry can be tailor-made by facile regulation of the ligand chemistry. Figure 1 | (a) Allosteric recognition of CaM in open (PDB ID: 1CFD) and closed (PDB ID: 1CTR) conformations to ligand. (b) Ca2+-driven self-assembly and disassembly of CaM-coated protein nanocages through allosterically activated protein–ligand interactions. The chemical structure of specific ligands, PT(POG)n (n = 6, 10, and 14). CaM, calmodulin; PT, phenothiazine. Download figure Download PowerPoint Results and Discussion The designed CLPs, (POG)n (n = 6, 10, and 14), were prepared by standard Fmoc-l-phenylalanine chemistry and (Pro-Hyp-Gly) tripeptide as the repeating unit using solid-phase synthesis on an automated multipeptide synthesizer at a scale of 0.1 mM ( Supporting Information Figure S15). Carboxylation of PT, followed by a condensation reaction with N-hydroxysuccinimide, afforded the PT-modified active ester (PT-NHS). The highly efficient coupling between the NHS end-group and amino residue of peptides yielded the target ligand molecules, PT(POG)n (Scheme 1 and Supporting Information Figures S16–S18). Scheme 1 | Routes for the syntheses of a group of PT-derived CLPs, PT(POG)n (n = 6, 10, and 14), as the ligand molecules to mediate protein allosterism. PT, phenothiazine; CLP, collagen-like peptide. Download figure Download PowerPoint The protein cage self-assembly was studied in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer solution (pH = 7.4). Mixing Ca2+-free CaM with PT(POG)n (equimolar ratio, 0.20 μM) afforded no aggregates. However, adding 0.8 μM of Ca2+ (due to 4∶1 binding number between Ca2+ and CaM)37 into the binary mixture, the solution turned turbid (λtest = 550 nm), implying the appearance of CaM-PT(POG)n protein aggregates ( Supporting Information Figure S1). Negative-stained transmission electron microscopy (TEM) images showed that all three systems can self-assemble into spherical nanostructures (Figures 2a–2c). Close inspection found that the particles in each system had uniform sizes ( Supporting Information Figure S2), and the contrast between gray edge and dark core indicated that these spheres were cage-like, possessing protein outer jackets, and hollow lumens, analogous to the morphology of virus capsids. Notably, the cage dimension seemed to be ligand-length-dependent. With the increase of CLP repeating unit, dynamic light scattering (DLS; Supporting Information Figure S3) revealed that their hydrodynamic diameters (Dh) elevated from 19.2 nm (n = 6) to 29.4 nm (n = 10) to 40.5 nm (n = 14). By further solution small-angle X-ray scattering (SAXS) tests, the scattering patterns for the three protein assemblies exhibited regular oscillating profiles on a q−2 decay (Figure 2d), which corresponded to a geometric model of concentric spherical shell.40 According to this model, the varying maxima/minima ratios of q vector and the shift of valley values indicated their different shell thickness (dwall) of 5.5, 6.7, and 8.0 nm and their exterior diameter (dSAXS) of 18.8, 28.1, and 38.5 nm ( Supporting Information Figure S4). These results accorded well with the DLS and TEM data, confirming that the ligand length can dictate the size and shape of such protein–ligand nanocages. Figure 2 | Ca2+-driven self-assembly of protein cages and ligand-length-dependent size and morphological shifts of CaM-PT(POG)n cage. (a–c) TEM micrographs of the protein–ligand self-assembled structures in the presence of Ca2+: (a) CaM-PT(POG)6, (b) CaM-PT(POG)10, and (c) CaM-PT(POG)14 (scale bars = 100 nm). (d) SAXS profiles of the three kinds of protein cages. CaM, calmodulin; PT, phenothiazine; TEM, transmission electron microscopy; SAXS, small-angle X-ray scattering. Download figure Download PowerPoint We next wanted to explore what impetus drives CaM to undergo a Ca2+-mediated self-assembly. A rational speculation is that there could exist two major interactions to underpin the protein cages: protein–ligand interaction and ligand–ligand association. On the one hand, Ca2+-induced protein–ligand complexation was identified by 15N heteronuclear single quantum coherence (15N-HSQC) NMR spectroscopy. This is a sensitive NMR technique that offers insight into protein–ligand binding and the interacting sites.41 Using PT(POG)10 as a ligand model, when mixed with equimolar 15N-labeled Ca2+-free CaM, all its 148 residues only gave slight resonance perturbations (Δδ < 0.04 ppm; Figure 3a, bottom panel), indicating no specific interactions. However, upon addition of Ca2+, some of these signals changed significantly, among which a particular group of residues (Glu11 in N-domain; Phe92, Met124, Phe141, and Met144 in C-domain) responded intensely (Δδ > 0.2 ppm; Figure 3a, top panel and Supporting Information Figure S5). This points out the specific binding sites of PT(POG)10 ligand within CaM protein, which is in accord with the crystallography of small-molecule CaM-PT complex.42 Their binding affinity (Ka), determined by isothermal titration calorimetry (ITC), grew by seven orders of magnitude from 7.9 × 10−1 M−1 to 4.92 × 106 M−1 before and after treatment with Ca2+ ( Supporting Information Figure S6). This means that the introduction of Ca2+ can fold CaM conformation to trap PT-bearing molecule for protein–ligand pairing. Figure 3 | (a) 15N-HSQC NMR-monitored bar graphs showing the chemical shift of per residue in 15N-labeled CaM-PT(POG)10 mixture with and without Ca2+. (b) CD spectral variation of the CaM-PT(POG)10 solution under different conditions. 15N-HSQC, 15N heteronuclear single quantum coherence; CaM, calmodulin; PT, phenothiazine; CD, circular dichroism. Download figure Download PowerPoint On the other hand, it is known that CLP is prone to form a triple helix.43,44 Hence, we wondered if this ligand could trimerize in the presence of Ca2+. Probing by circular dichroism (CD) spectroscopy, PT(POG)10 and CaM without Ca2+ only showed typical α-helical CD signals ascribed to CaM (Figure 3b, black line)45; but Ca2+ addition led to a signature CD spectral change, yielding a minimum at 200 nm and a maximum at 225 nm, indicative of the presence of polyproline-type triple helices (Figure 3b, red line).46 The CD thermal analysis further reflected that the association constant of the triple helix is 1.45 × 104 M−1 ( Supporting Information Figure S7 and Table S1), inferior to that of protein–ligand affinity. Moreover, this trimerization is reversible, as attested to by the recovery of the CD curve upon removal of Ca2+ (Figure 3b, blue line). These facts together demonstrate that the Ca2+-driven allosteric effect of CaM can activate the multiple protein–ligand and ligand–ligand interactions, which collectively facilitate protein self-assembly. From the terms of free energy (ΔG), the individual contribution of the protein–ligand partner and ligand trimer accounts for 62% and 38% of the total energy ( Supporting Information Table S1). In other words, the protein–ligand interaction is a decisive driving force for the protein assembly while the ligand–ligand association plays an auxiliary role. For the other two protein cages, their assembly also depends on the same forms of interactions and responds to Ca2+. By contrast, their interligand affinity varied consistently with the CLP number ( Supporting Information Figure S7 and Table S1, KPT(POG)6 = 3.12 × 103 M−1 and KPT(POG)14 = 2.85 × 104 M−1), illustrating that the stability of protein cages is enhanced as the length of ligand triple helices extends. To further map the three-dimensional (3D) self-assembly structure of the protein cage, we used cryo-electron microscopy (cryo-EM) to unravel the protein packing fashion. EM imagery revealed that the CaM-PT(POG)10 cage embodied quasivirus morphology with protruded protein superstructure (Figure 4a). Single-particle micrographs affirmed that the protrusions are protein-constituted ring nanoobjects, and the particles overall manifest polyhedral geometry (Figure 4b). 3D reconstruction further showed that they have 282 Å diameter (close to SAXS data, 28.1 nm) with polyhedral geometry. Clearly, the particle shell can be divided into two layers: The outer layer is the protein coat, in which every three CaMs adopt tripod arrangement to form a CaM trimer and further self-organize into a hexon (corresponding to the ring-like entities observed in EM image) to cover the entire surface (Figure 4c, left half). The inner layer is composed of the ligands, which are distributed evenly inside the protein coat (Figure 4c, right half). To understand the self-assembled structural basis, we also combined the 3D reconstruction and computational modeling to analyze one single protein cage. Figure 4d gives a closeup view of one individual protein-hexon structure. It is comprised of six CaM trimers and together forms a near hexagonal topology, where two adjacent CaM belonging to two different trimers are noncovalently connected and produce interfacial contacts. We deduced that there possibly exist CaM–CaM protein–protein interactions. The zoom-in image (Figure 4e) revealed that the linking domains from two CaMs overlapped each other and constituted an interlocked, X-shaped CaM dimer with a certain dihedral angle, which provided impetus for a spherical assembly rather than a discrete trimer. Figure 4 | The atomic-level structural basis of protein cage self-assembly. (a and b) Cryo-EM and single-particle EM micrographs of CaM-PT(POG)10 cage. (c) 3D reconstruction shows the icosahedral topology and double-layered architecture of CaM-PT(POG)10 cage (left half: CaM-coated outer layer; right half: ligand-forming inner layer as indicated by the green distribution). (d) Protein-hexon structure that is assembled by six identical CaM trimers via CaM–CaM protein–protein interactions. (e) The structural basis of CaM–CaM dimer through an interlocked X-shaped overlap of linking domains from two CaMs (Ca2+ is indicated as red sphere). (f) The structure basis of one CaM-PT(POG)10 trimer in top (left) and side (right) view. The purple ribbon represents CaM protein and colored sticks represent PT(POG)10 ligand (Ca2+ ions are omitted for clarity). (g) Atomic structure of protein–ligand binding sites between CaM residues (orange series) and PT moiety (green series). The interactions of Glu11, Phe92, Met124, Phe141, and Met144 with PT group are indicated by dashed lines. (h) Schematic illustration of cone-angle of CaM-PT(POG)n trimer dictating the size difference of protein cages. CaM, calmodulin; cryo-EM, cryo-electron microscope; PT, phenothiazine. Download figure Download PowerPoint On the other hand, we also investigated the formation of CaM trimer. Typically, it follows a triangle protein arrangement, each CaM seizing one ligand (Figure 4f). The PT moiety is buried into the hydrophobic pockets of CaM, tightly binding with Glu11 (3.81 Å) in N-domain, and Phe92 (4.29 Å), Met124 (3.79 Å), Phe141 (4.60 Å), and Met144 (3.31 Å) in C-domain (Figure 4g). These data are identical to the molecular spacing of CaM-PT complex,42 corroborating the NMR result. The three CLP chains of ligands approach each other to intertwine into a triple helix, whose overall shape is like a rigid rod. This is not only conducive to the CaM trimerization but also can structurally support the exterior protein framework (Figure 4f, right panel). For the other two cages, there are only minor differences in the protein–ligand contacts ( Supporting Information Figure S8). In addition, the difference of ligand helicates in length can interpret the size distinction of protein cages. Because the shape of the protein–ligand trimer can be approximately viewed as a conical building block, the longer the helicate is, the smaller its cone angle is, and accordingly, the larger the resulting cage is (Figure 4h). To fully elucidate the proposed protein self-assembly mechanism, we also designed several control experiments to rule out other possibilities. First, we studied whether the ligand trimerization was necessary for this cage formation. To this end, we synthesized a PT-derived peptide counterpart, which comprises ten glycine–alanine–glycine (GAG) repeating units, PT(GAG)10, to perform the assembly experiments in the same conditions. It is known that the GAG tripeptide unit is a β-sheet-forming sequence found in silk protein,47 but does not lead to helical structure. Using the GAG tripeptide instead of PT(POG)10, even though protein nanospheres can form in solution, we observed that size and morphology were not identical and they produced interparticle connections and fusions ( Supporting Information Figure S9a). Their Dh was determined to be 41 nm with a broad polydispersity (PDI) of 0.24, as measured by DLS ( Supporting Information Figure S9b). These results indicate that the architectural accuracy of protein spherical assembly is related to the precision of peptide chain aggregation. Because the β-sheet formed by (GAG)10 peptides lacks a precise aggregation number and secondary structure, we could only obtain the protein nanospheres of uneven size rather than well-defined nanocages. Second, we wanted to know if ligand trimerization requires protein synergy or an independent process. To confirm this point, we used PT(POG)10 as a typical model. By CD detection, PT(POG)10 (0.2 μM), regardless of the presence or absence of Ca2+, only showed weak CD signals, indicating no helicate yield ( Supporting Information Figure S10, blue solid and dash lines). However, upon addition of equal amounts of CaM protein, the CD profile gave two characteristic CD signals in the presence of Ca2+, indicating the formation of collagen-like triple helix. On the contrary, upon removal of Ca2+, only α-helical signals of CaM protein could be seen ( Supporting Information Figure S10, red solid and dash lines). This result proved that ligand trimerization requires the assistance of CaM. A rational reason is that the trimeric association of any CLP has a critical concentration. PT(POG)10 alone in solution is lower than this critical concentration; thus it exists in a unimer form. Once it binds with CaM to form CaM-PT(POG)10 complex, its total volume and hydrophobicity can increase, which greatly lowers the critical concentration and leads to the helical structure. Third, we wondered whether we could use a simple PT molecular triad as the ligand to replace complex peptide derivative. For this, we designed a three-arm PT compound with a glycerol core (PT3). However, under similar conditions, mixing PT3 with Ca2+-saturated CaM can only obtain irregular, small protein aggregates. TEM imagery revealed that their size is smaller on average than 10 nm ( Supporting Information Figure S11), close to that of CaM oligomer, indicating that their further assembly from the CaM-PT3 complex to a larger cage is inhibited. One main reason is the insufficient hydrophobicity of protein–ligand complex. If a supramolecular complex can form large-scale assemblies, the suitable molecular amphiphilicity is necessary. In the case of CaM-PT3 complex, PT3 is a small-molecule ligand and lack of long peptide chain in contrast to PT(POG)n, it thus does not have enough hydrophobicity to induce further spherical assembly. Fourth, to elucidate the crucial role of CaM–CaM dimerization in driving spherical assembly, we carried out the protein mutation experiments. It is known when the CaM recognizes the PT ligand, it will change to a close conformation, inducing the formation of CaM–CaM dimer through X-shaped protein interfacial contacts. We found that the protein contacts lie in the linking domains of two CaMs. In its open conformation, the linking domain exhibits an entire α-helical motif ( Supporting Information Figure S12a); however, in its closed conformation, the linking domain folds into an upper and lower helical section (helix I and II) with a certain angle ( Supporting Information Figure S12b). These two linking domains form a reverse intersecting dimer, where the helix I of linking domain A is located above the helix II of linking domain B, whereas the helix II of linking domain A is located below the helix I of linking domain B ( Supporting Information Figure S12c). This structure is reminiscent of a twisted knot ( Supporting Information Figure S12d), which greatly stabilizes the protein assembly. The residues of E31 and S38 in linking domain A can strongly interact with T109 and E103 in linking domain B through hydrogen bonding, and vice versa. Their distance of E31-T109 and S38-E103 are 2.79 and 3.01 Å, respectively. With these facts in hand, we have designed single-site (E31A) and dual-site (E31A and E103A) CaM mutants to confirm CaM dimerization. For the case of single-site CaM mutation (E31A, glutamic acid to alanine residue), the same assembly process can only yield shape-irregular protein nanosheets ( Supporting Information Figure S12e). This is understandable since the E31A mutation can break two strong H-bonding interactions (E31-T109), which leads to the instability of X-shaped CaM dimer and loses the precise dihedral angle. This will cause the failure of cage-like assembly, and the two remaining weak H-bonding interactions (S38-E103) can only underpin an irregular lamellar assembly without structural periodicity. For the case of dual-site mutation, owing to the removal of all H bonding on the protein contacting interfaces, no available protein assemblies can be observed except for several CaM oligomers ( Supporting Information Figure S12f). These findings verify that the CaM–CaM interactions play a crucial role in the support of virus-like assembly. Based on the above facts, it can be deduced that the allosteric protein CaM, allosteric ligand unit of PT, and triple helix of (POG)n peptide chain are indispensable prerequisites to realize this allosteric-effect-driven protein cage self-assembly. Finally, given that the two conformers of CaM can interconvert in response to the Ca2+ allosteric signal, it is possible to manipulate the opening and closing of protein cages by regulating the Ca2+ level. As expected, applying ethylene glycol tetraacetic acid (EGTA) to chelate Ca2+, the protein shells can detach within 30 min, owing to CaM conformational relaxation (Figures 5a and 5b). Conversely, once CaM reverts to its contractive configuration by readdition of Ca2+, the cages can be rebuilt (Figure 5c). This coating/uncoating process is reversible, which is reminiscent of the capsidation and decapsidation effect of real viruses ( Supporting Information Figure S13). With this feature in mind, the ability of the cage to deliver nucleic acid in the cell was studied ( Supporting Information Figure S14). Using calcein amidite-modified DNA (c-DNA) as a cargo, 50 nM of the c-DNA loaded CaM-PT(POG)10 cage was incubated with HepG2 cells. If c-DNA is taken up by cells, nonfluorescent c-DNA can be cleaved at the amidite bond by cytoplasma esterase and expose the strong fluorescent calcein label. Confocal microscopy found that cells treated with cage-loaded c-DNA only emitted faint fluorescence (Figures 5d and 5e). Interestingly, as adding EGTA into the cell, bright intracellular fluorescence appeared (Figure 5f). The results indicate that cells can endocytose DNA-packaged protein nanocages, and the Ca2+-switchable cage walls, resembling the natural virus capsids, can achieve the defense and control release of internal genetic material, presaging their potential in nanomedicines and biomaterials. Figure 5 | Ca2+-switchable protein cages for controllable DNA package and release: TEM images showing the CaM-PT(POG)10 protein cage uncoating (a and b) upon EGTA (0.8 μM) and (c) rebuilding upon addition of Ca2+ (1.0 μM). Confocal laser microscopy of HepG2 cells treated with (d) buffer, (e) c-DNA-packaged cage, and (f) c-DNA-loaded cage, then adding EGTA. Nucleus stained by Hoechst 33342 and green fluorescence indicating the calcein emission from uncoating protein cages. TEM, transmission electron microscopy; CaM, calmodulin; EGTA, ethylene glycol tetraacetic acid. Download figure Download PowerPoint Conclusion To confer dynamic assembling ability and adaptable behavior of protein nanocages is crucial to constructing “living” virus mimics. This work demonstrates an unprecedented protein cage system that is predicated on protein–ligand interactions, and suggests a new strategy that harnesses the allosteric effect of protein itself to dynamically switch the self-assembly and disassembly of protein cages. This strategy allows us to obtain tailor-made protein cages with tunable size, structure, and function by simple allosterically regulated protein–ligand chemistry that replaces complicated protein interface chemistry, which promises a new direction for building smart virus mimics. Supporting Information Supporting Information is available. Conflict of Interest There is no conflict of interest to report. Funding Information The authors are grateful for the support of the National Natural Science Foundation of China (nos. 21674022 and 51703034), the National Defe