Cloaking nanoparticles with protein corona shield for targeted drug delivery

Targeted drug delivery using nanoparticles can minimize the side effects of conventional pharmaceutical agents and enhance their efficacy. However, translating nanoparticle-based agents into clinical applications still remains a challenge due to the difficulty in regulating interactions on the interfaces between nanoparticles and biological systems. Here, we present a targeting strategy for nanoparticles incorporated with a supramolecularly pre-coated recombinant fusion protein in which HER2-binding affibody combines with glutathione-S-transferase. Once thermodynamically stabilized in preferred orientations on the nanoparticles, the adsorbed fusion proteins as a corona minimize interactions with serum proteins to prevent the clearance of nanoparticles by macrophages, while ensuring systematic targeting functions in vitro and in vivo. This study provides insight into the use of the supramolecularly built protein corona shield as a targeting agent through regulating the interfaces between nanoparticles and biological systems.

T he application of nanoparticles is promising for the development of imaging and therapeutic agents through improved biodistribution and controlled drug release [1][2][3] . The rationale behind using nanoparticles is that those with diameters of <200 nm extravasate from the leaky tumor blood vessels are retained in the tumor due to the "enhanced permeability and retention" (EPR) effect 4 . Although few nanoparticle formulations (e.g., Abraxane and Doxil) are available in the market, the ubiquitous targeting approach suffers from several limitations including rapid clearance by the mononuclear phagocyte system (MPS) and low uptake into target tumors 5,6 . To improve the targeting ability, along with the EPR effect, an active targeting approach has been attempted by coating the particle surface with antibodies, proteins, or peptides that bind to receptors that are typically overexpressed on cancer cells 7 . However, recent reports have revealed only a modest increase in tumor targeting when this approach is applied; moreover, the addition of targeting ligands increased the clearance of nanoparticles by MPS, indicating that no definitive conclusion has been reached regarding the therapeutic efficacy of this technique 8,9 . In principle, when exposed to physiological environments, the nanoparticle surface is covered by various biomolecules to lower the surface energy by a combination of entropy-driven water molecule displacement, particle surface charge compensation, and screening of hydrophobic parts [10][11][12] . Biomolecule adsorption results in the formation of a layer, called a protein corona, and significantly changes the original molecular identity of the nanoparticle 13 . The formation of a protein corona on the nanoparticle surface can be regulated by modifying the nanoparticle surface with zwitterions, polyethylene glycol (PEG), carbohydrate moieties, and dysopsonic proteins, which can enhance the colloidal stability and prolong the circulation time in blood by enabling escape from MPS clearance [14][15][16][17] . However, these strategies are still limited at conferring targeting specificity since an additional targeting ligand increases the propensity for protein corona formation to mask the targeting ability and inhibit the biological effects of nanoparticles [18][19][20][21][22][23][24] . This can explain why many nanoparticles with active targeting systems have failed in clinical trials 25 . Therefore, to design nanoparticle-based therapeutic agents, there is a considerable need to regulate protein corona formation on nanoparticles [26][27][28] and to obtain a deeper understanding of the molecular mechanism involved in regulating nanoparticle-biological interactions. Here, we present a targeting system in which nanoparticles are supramolecularly precoated with a protein corona shield (PCS) that reduce serum protein absorption while retaining targeting specificity (Fig. 1a).

Results and Discussion
The development of PCS on nanoparticle. To develop PCS on nanoparticles, we use a recombinant fusion protein, GST-HER2-Afb, in which HER2-binding affibody (Afb) is genetically combined with a glutathione-S-transferase (GST) 29 , a well-known fusion tag protein, with an extra linker (GGGLVPRGSGGGCGGGGTGGGSGGG). The preparation of GST-HER2-Afb (molecular weight: 36.3 kDa, >99.0% purity) is confirmed by electrospray ionization time-of-flight mass spectrometry (Fig. 1b) and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) ( Supplementary Fig. 1A). The surface charge of the GST-HER2-Afb at physiological pH is approximately −5.25 mV, which is similar to that of GST (−6.29 mV) (Fig. 1c). The intact binding ability of GST-HER2-Afb is confirmed by monitoring its interactions with the targeting receptor HER2/ErbB2 in real time using a quartz crystal microbalance (Supplementary Fig. 1B) and the cellular uptake of the fluorescein-labeled GST-HER2-Afb to the HER2-receptor-overexpressing cell line (SK-BR3) (Fig. 1e). These results indicate that the recombinant fusion protein with the linker, GST-HER2-Afb, exhibit the ability to bind to the complementary receptor. Furthermore, the toxicity test confirm its biocompatibility, indicating that GST-HER2 itself is not toxic up to 10 μM ( Supplementary Fig. 1C).
The GSH-MSN was then coated with GST-HER2-Afb to give the PCSN through the supramolecular interaction at the GSHbinding site. The attachment of GST-HER2 on GSH-MSN was confirmed by measuring the surface charge, which changed from −40 mV (GSH-MSN) to −5.3 mV (PCSN) (Fig. 1c). The DLS experiment indicated an increase in the hydrodynamic radius of PCSN (~270 ± 20 nm) compared with that of GSH-MSN (~140 ± 20 nm) ( Supplementary Fig. 4A), and the TEM image analysis further confirmed the protein-coating layer (Fig. 1f). The maximum number of GST-HER2-Afb attached on GSH-MSN was 270 μg/mg, which was confirmed by BCA assay (Supplementary Fig. 4B). No significant aggregation of PCSNs was observed in the DLS analysis up to 2 weeks at 4°C (Fig. 1d).
The interaction between PCSN and serum proteins. Next, we investigated interactions occurring at the interfaces between serum proteins and PCSNs. Control particles, GSH-MSN, and PEG-modified MSN (PEG-MSN) 31 were used to study the effects of GST-HER2-Afb pre-coating on particles. We first incubated PCSN, PEG-MSN, and GSH-MSN with 55% fetal bovine serum (FBS) for 1, 2, and 4 h and isolated the serum proteins that adsorbed to them by centrifugation to completely remove unbound proteins. The molecular composition of the adsorbed serum proteins was measured by denaturing SDS-PAGE and the protein density was plotted (Fig. 2a, b). The protein profiles observed for PEG-MSN and GSH-MSN were fairly similar, but the intensities of the bands for PCSN were significantly reduced (~15-fold lower intensity than for GSH-MSN for the case of 1 h of incubation), indicating that the GST-HER2-Afb pre-coating on PCSN could reduce the interactions with serum proteins. To further confirm the particle-protein interactions, the physical characterizations were also investigated by using DSL, zeta, and PDI ( Supplementary Fig. 5). The treatment of 55% serum on GSH-MSN and PEG-MSN increased its hydrodynamic radius and changed its surface charge while PCSN showed no significant changes, confirming the decreased PCSN-serum protein interactions.
We then investigated the composition of serum proteins adsorbed on each particle using shotgun proteomics [liquid   Fig. 2d, middle). The proteins adsorbed on PCSN were composed of lower-molecular-weight proteins (20-60 kDa) (Fig. 2d, left) and included an abundance of proteins with pI of approximately 7-8 compared with the proteins adsorbed on GSH-MSN and PEG-MSN (Fig. 2d, right). Moreover, the amount of immune response involving proteins, complement, and coagulation proteins adsorbed on PCSN was significantly lower than those on GSH-MSN and PEG-MSN, suggesting that PCSN can increase their circulation time in the blood by decreasing the uptake from the immune system 20 . It is also noted that the relative quantity of apolipoproteins among serum proteins absorbed on PCSN increased with respect to GSH-MSN and to PEG-MSN ( Fig. 2d, middle). Considering that apolipoproteins on nanoparticle play an important role in association with cellular uptake to target cells [32][33][34] , targeting capability of PCSN partly could be linked with some serum proteins recruited at the surface of PCSN.
To understand the interactions between PCS and external biological components, we investigated the molecular mechanisms of the interactions associated with the conformational change of GST-HER2-Afb. When non-covalently adsorbed on nanoparticles with precisely designed supramolecular interaction, specific domains of proteins can interact with the surface of nanoparticles in thermodynamically favorable manners, leading to the colloidal stabilization and dysopsonization of particles in a physiological environment 35,36 . On the other hand, covalent attachment of their counterparts on particles significantly changes their conformations and particle surface properties (e.g., hydrophobicity and charge density), inducing potential detrimental interactions with external biomolecules, rather than stabilizing them 13,19,37 . Therefore, assuming that the orientation and conformational change of proteins adsorbed on particles are subject to interaction with serum proteins, the effects of chemical modification of GST-HER2-Afb on biological consequences were investigated. First, PCSNs having randomly orientated GST-HER2-Afb [PCSN(R)s] were prepared by chemically conjugating GST-HER2-Afb to succinimidyl-modified MSN via amide formation with amine groups on the protein surface (Fig. 2e, right). When assessed by SDS-PAGE (Fig. 2e, left), the protein absorption for PCSN(R) was enhanced five-fold compared with that for PCSN, indicating that the randomly orientated conjugation of GST-HER2-Afb may change surface properties and increase protein-nanoparticle interactions. Next, the succinic anhydride-modified PCSN [PCSN(−)] were prepared to generate additional carboxylate groups on GST-HER2-Afb, which has a negatively charged function but still binds on GSH-MSN in the same orientation as GST-HER2-Afb (Fig. 2e, middle). The SDS-PAGE analysis confirmed that the protein absorption observed for PCSN(−) was similar to that of PCSN, implying that once thermodynamically favorably orientated on nanoparticles, the supramolecularly adsorbed proteins can stabilize the colloid and significantly reduce the interactions with external serum proteins. We further employed coarse-grained molecular dynamics method to observe the interactions between coated silica nanoparticles (i.e. PEG-MSN, PCSN(R), PCSN (−), and PCSN) and serum protein (i.e. albumin) in the vicinity of their interface at the molecular level (simulation method details in the Supplementary Figs. [6][7][8][9][10][11][12], which suggests that the orientation of GST-HER2-Afb attached to the surface of silica nanoparticles as well as the electrostatic interactions can be an important factor. Stealth and targeting effect of PCSN. We studied the cellular uptake of PCSN by macrophages, one of the most important components of the immune defense system, which act by clearing foreign molecules from the blood. Being able to evade internalization by phagocytic cells would provide a drug carrier that can accomplish long-term blood circulation and enhanced arrival at the target tumor 38,39 . PCSNs or PEG-MSNs loaded with a fluorescent dye, DiI, were pretreated with 55% FBS for 1 h at 37°C and then incubated with a murine macrophage-like cell line, RAW264.7, for 6 h in culture medium supplemented with 10% serum (Fig. 3a). Confocal microscopy imaging revealed that the internalization of the DiI-loaded PEG-MSN was significant, but was rarely observed for the DiI-loaded PCSN (Fig. 3b), which was further confirmed by flow cytometry (Fig. 3c). Cell viability analysis also confirmed reduced cellular uptake of PCSN when the camptothecin-loaded PCSN was applied (Fig. 3d). These results indicate that PCS on PCSN significantly reduced internalization into macrophages. A recent study reported that the biomolecular corona formation with a specific composition provides a stealth effect on the macrophage recognition 40 . Similarly, our results suggest that the scarce corona formation on PCSN caused by the supramolecularly pre-coated proteins could confer stealth properties to evade immune cells and subsequent elimination by the MPS.
Next, to investigate the cell-specific targeting ability of PCSN, the HER2-receptor-overexpressing cancer cell line SK-BR3 and the HER2-receptor-negative cell line HEK293T were treated with Dil-or camptothecin-loaded PCSN after pre-incubation in 55% FBS for 4 h (Fig. 3e). The significant uptake of Dil-loaded PCSN by SK-BR3 cells rather than HEK293T cells, mediated by receptor-mediated endocytosis (especially macropinocytosis), was confirmed by confocal fluorescence microscopy ( Fig. 3f and Supplementary Fig. 13A) and FACS (Supplementary Fig. 13B). The targeted internalization was further confirmed by measuring cellular viability for the camptothecin-loaded PCSN, which exhibited dose-dependent toxicity on SK-BR3, while exhibiting lower toxicity on HEK293T (Fig. 3g). A control, HER2-binding affibody (Afb) modified MSNs (Afb-MSN) was also evaluated and it confirmed that GST plays an important role in reducing interactions among the serum proteins as well as supramolecularly conjugating on particle (Supplementary Fig. 14). To examine the versatile targeting platform for PCS, EGFR-binding Afb combined with glutathione-S-transferase (GST-EGFR-Afb) was also applied on GSH-MSN and its targeting ability to MDA-MB468 cells (EGFR-positive cancer cells) was confirmed Fig. 1 Protein corona shield nanoparticle (PCSN). a We introduce the protein corona shield (PCS) concept for an efficient target drug delivery system. Generally, nanoparticle drug carriers with a target ligand lose their targeting ability on being coated by blood proteins in a biological environment. However, the PCS system can inhibit blood protein adsorption to maintain the targeting ability and avoid unwanted clearance by the mononuclear phagocyte system. b Mass spectrometry analysis of the GST-HER2-Afb showed a mass of 36.3 kDa. c Zeta-potential analysis of mesoporous silica nanoparticle (MSN) (−23 mV), GSH-MSN (−39 mV), GST-HER2-Afb (−5.25 mV), and PCSN (−5.3 mV). d Size distribution plots of PCSN. e Images of cellular uptake of fluorescein 5 maleimide-modified GST-HER2-Afb by the target cell (SK-BR3) and the negative control (MCF-10A). f Transmission electron microscopic images of GSH-MSN and PCSN (scale bar represents 100 nm). All bar graphs were reported as means ± standard deviations (SDs) for three experimental replicates (n = 3) ( Supplementary Fig. 15). Taken together, these results show that PCSN exhibited cell-specific targeting ability as well as stealth properties.
Effect of PCSN on the targeting and antitumor efficacy. To translate the outcomes of PCSN in an in vivo tumor model, nanoparticles loaded with the far-red fluorescent dye DiD were systemically injected via the tail vein into nude mice (n = 6 mice per group) bearing SK-BR3 cell xenografts, after which the tumor accumulation of PCSN was monitored. Considering the lowest accumulation efficacy of GSH-MSN observed from initial experiments ( Supplementary Fig. 16A), our studies focused on the comparison between PCSN and PEG-MSN. The in vivo live imaging results showed the presence of 1.8-fold enhanced fluorescent signals in tumor sites from the PCSN-treated group compared with the PEG-MSN-treated group at 0.5 h after injection ( Supplementary Fig. 16B), which was maintained until 24 h, while the fluorescent signals for PEG-MSN gradually decreased with time at 8 h ( Supplementary Fig. 16C).
The biodistribution of nanoparticles was then assessed by measuring the fluorescence of DiD in harvested organs and tumors upon necropsy 48 h after injection. The fluorescent signals of tumors from the PCSN group were enhanced 2.5-fold in comparison with those of the PEG-MSN group (Fig. 4a). Moreover, the fluorescent intensity in the tumor for the PCSN group was seven-fold higher than that in reticuloendothelial organs (e.g., liver and spleen), whereas no significant difference in fluorescent intensities between tumors and the reticuloendothelial organs was observed for the PEG-MSN groups (Fig. 4b), indicating that PCS enabled the nanoparticles to evade the immune system and undergo enhanced accumulation in the target tumor. In vivo antitumor efficacy was further evaluated by intravenously administering camptothecin-loaded PCSN (PCSN (CPT)), PEG-MSN (PEG-MSN(CPT)), PCSN, camptothecin (Free-CPT), and phosphate-buffered saline (PBS) to the SK-BR3 tumor-bearing mice. Compared with the PEG-MSN(CPT)treated group, the PCSN(CPT)-treated group exhibited higher therapeutic efficacy, resulting in~90.0% inhibition of tumor growth in terms of volume and 2.5-fold enhancement of inhibitory effects (Fig. 4c). Subsequently, by examining tumor weight reduction and histopathology after necropsy on day 22, the PCSN(CPT)-treated group was shown to have enhanced therapeutic efficacy for tumor growth inhibition ( Supplementary  Fig. 16D, 17). Additionally, hematoxylin and eosin (H&E)-stained lung, liver, spleen, and kidney samples showed no apparent abnormalities or lesions at day 21 after camptothecin-loaded PCSN treatment (Supplementary Fig. 17). These results indicate that PCSN increases the tumor targeting ability, enhancing the efficacy of cancer chemotherapy. Collectively, we found that the supramolecular binding of GST-HER2-Afb on particles enabled the maintenance of conformational stability and further minimized interactions with serum proteins. In vitro and in vivo experiments confirmed that PCSN improved the targeting ability and therapeutic efficacy. These findings indicate that exploiting protein coronas can provide a tool for a targeting platform.
Endocytic pathway analysis. To check the endocytosis-mediated uptake, SK-BR3 cells were seeded into four-well chambers with cover glass and pretreated with different inhibitors, including sucrose (clathrin-mediated uptake, 400 nM), methylbeta cyclodextrin (caveolae-mediated uptake), and amilorin (macropinocytosis), in a serum-free DMEM for 1 h and replaced with fresh media. Afterward, DiI-loaded PCSNs were added to the medium for another 1 h of incubation. Then, the cells were analyzed with a confocal microscope (Olympus FV1000) connected to a CO 2 incubator.
Animals and tumor models. Female nude mice (18 ± 2 g, 6 weeks of age) were fed under the condition of 22 ± 2°C and 55 ± 5% humidity, with free access to food and water. All animal experiments were approved by the Institutional Animal Care and Use Committee at Ulsan National Institute of Science and Technology and conducted in compliance with the guidelines (UNIST-IACUC- [17][18][19][20][21][22][23][24][25][26][27][28][29]. To set up the tumor xenograft model, mice were subcutaneously inoculated in the right lower leg with 1 × 10 6 human breast cancer SK-BR3 cells. Tumor volume (V) was determined by the following equation: V = L × W 2 /2, where L and W are the length and width of the tumor, respectively. SK-BR3 tumor-bearing mice were used in the experiments when the tumor volumes reached approximately 100 mm 3 .
In vivo and ex vivo fluorescence imaging. To evaluate the biodistribution of GSH-MSN, PEG-MSN, and PCSN loaded with a lipophillic dye, DiD (at 10 wt% loading capacity) was intravenously injected (0.1 mg/mL) into SK-BR3 tumorbearing nude mice (7 mice/group). Then, the mice were anesthetized and imaged using an in vivo imaging system (Bruker Xtreme) with an excitation wavelength of 630 nm and emission wavelength of 700 nm. In vivo images were taken at 0.5, 1, 2, 4, 8, 10, 12, 24, 32, and 48 h. Then, the mice were killed to separate the organs and tumors for ex vivo imaging to determine the biodistribution pattern of DiD-loaded nanocarriers.
In vivo tumor inhibition analysis. Female nude mice (average weight 18.5 ± 2 g) were obtained from Orientbio, Korea, for the in vivo studies. All protocols for the in vivo experiments were approved by the UNIST-IACUC animal ethics approval committee. SK-BR3 tumor-bearing mice were used to carry out experiments when the tumor volumes reached approximately 300 mm 3 . The tumor-bearing nude mice were randomly divided into five groups (6 mice per group). PBS, free-CPT, PCSN, CPT-loaded PEG-MSN, and CPT-loaded PCSN were intravenously injected into mice every 3 days seven times. CPT dosage was 1.5 mg/kg body weight. The first day of treatment was defined as day 0. The body weight and tumor size were recorded every 3 days, and the survival of mice was monitored throughout the . c Growth curve of tumor volume after intravenous injection with various groups of carriers until day 21 (n = 6 mice per group, mean ± 1 day [n = 6 mice per group, mean ± SD, statistical significance was calculated by one-way analysis of variance, *P < 0.05, **P < 0.01] experiment. The concentrations used for the groups treated with PCSN were the same as that of CPT-PCSN. After treatment for 21 days, related mice were killed, and the tumor tissues were removed from the bodies to investigate the morphology and use for further studies.
Histochemical analysis. At day 22, one mouse in each group was sacrificed to separate the tumors, and the mice were then fixed in 4% formaldehyde solution in PBS. For histochemical staining, the fixed tumors were dehydrated by gradient ethanol washing and embedded in paraffin blocks, sectioned, and stained with H&E staining.
Statistical analysis. All data were reported as means ± standard deviations (SDs). Microsoft Excel software was used to calculate P-values for unpaired t-tests. Statistical significance was determined by using Student's t-test and one-way analysis of variance, and P-values of <0.05 were considered to be indicative of statistical significance.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.