Structure-Based Nanocarrier Design for Protein ... - ACS Publications

Letter pubs.acs.org/macroletters

Structure-Based Nanocarrier Design for Protein Delivery Xu Wang,† Changying Shi,† Li Zhang,†,∥ Mei Yun Lin,† Dandan Guo,† Lili Wang,† Yan Yang,†,⊥ Thomas M. Duncan,‡ and Juntao Luo*,†,§ †

Department of Pharmacology, ‡Department of Biochemistry and Molecular Biology, and §Upstate Cancer Center, State University of New York Upstate Medical University, Syracuse, New York 13210, United States ∥ Department of Applied Chemistry, China Agricultural University, Beijing 100193, P. R. China ⊥ College of Marine Life Sciences, Ocean University of China, Qingdao 266003, P. R. China S Supporting Information *

ABSTRACT: Structure-based nanocarrier design for protein delivery remains challenging and has rarely been documented in the literature. We herein present a facile computer-aided approach for rational and customized design of a unique linear−dendritic telodendrimer that self-assembles into a nanocarrier for therapeutic protein delivery, e.g., insulin. Virtual screening of a small-molecule library was performed to identify optimal protein binding moieties, which were conjugated precisely in the telodendrimer backbone preinstalled with charged moieties. We systematically tested our hypothesis and obtained significant correlations between the computational predictions and experimental results. The D-α-tocopherol (vitamin E)-containing nanocarrier showed strong binding affinity for insulin in both computational prediction and experiments, which led to improved blood glucose control. This study affirms the concept and validates the approach of structure-based nanocarrier design for protein delivery. ffinity-controlled release systems,1 e.g., hydrogels decorated with protein-binding ligands, have shown great promise for protein loading and delivery in a “green” manner to maintain protein bioactivity.2,3 However, the application of such a strategy in nanoparticle systems for protein delivery is sometimes limited by the lack of specific ligands for therapeutic proteins, e.g., insulin. In addition, the accessibility of such specific binding sites can be compromised by random deposition of the tethered ligands in the self-assembled polymer nanoparticles. Alternatively, the decoration of nanocarriers with selective building blocks that interact with different residues on protein surfaces has demonstrated effectiveness for protein encapsulation and delivery via synergistic multivalent interactions.4−7 Salt bridges and hydrophobic interactions play important roles in stabilizing protein structures, which inspire the construction of nanocarriers to target protein surfaces globally. The introduction of charged groups in nanoparticles is straightforward to target the net charges on protein surfaces.4,5,8 In contrast, the hydrophobic moieties are largely variable, which are known to be important for both protein loading5 and intracellular delivery of therapeutics.9 Computational chemistries, such as molecular modeling and simulation, have been applied routinely in studying drug−protein interactions. It is straightforward to apply such computational tools to identify the hydrophobic protein binding moieties (PBMs) for the rational construction of affinity-based nano-

A

© XXXX American Chemical Society

carriers. However, the random introduction of charged and hydrophobic pendant groups into polymer backbones during polymerization or by postmodification hinders the capability for rational design and optimization of nanocarriers. A novel strategy and versatile platforms are desired for rational nanocarrier design for protein delivery. Previously, we presented a well-defined linear−dendritic telodendrimer platform synthesized by peptide chemistry.10,11 The synergistic combination of charged and hydrophobic moieties in telodendrimers has been demonstrated to be crucial for stable protein encapsulation via multivalent hybrid interactions.5 However, the selection of charged and hydrophobic functionalities based on the empirical approach significantly limits nanocarrier optimization. The unique telodendrimer architecture allows for precise conjugation of various building blocks, and it serves as a blueprint to leverage the synergism between computational prediction and combinatorial chemistry for structure-based nanocarrier design, which may dramatically accelerate nanocarrier development for protein delivery. We hypothesize that the stronger PBMs identified by virtual screening will lead to nanocarriers with stronger protein binding affinity and loading stability. To test our hypothesis (Figure 1), we synthesized a library of Received: January 3, 2017 Accepted: February 27, 2017

267

DOI: 10.1021/acsmacrolett.6b00982 ACS Macro Lett. 2017, 6, 267−271

Letter

ACS Macro Letters

Arginine is more efficient for intracellular protein delivery, in comparison with lysine, once introduced in peptides,12 as well as in our telodendrimer system for cytotoxic protein delivery.5 In this study, we choose lysine as a building block to provide positive charges for two reasons: (1) Insulin binds to the receptors on the cell surface and triggers the influx of glucose; therefore, extracellular insulin release is needed, and the arginine-induced cell uptake of nanoparticles5 should be avoided. (2) The amino group on lysine has more flexibility in orientation to interact with carboxylic groups on proteins to form an efficient salt bridge. In comparison, arginine can form strong bivalent hydrogen bonds only with the planar guanidinium,13 which may reduce the chance to form stable salt bridges between telodendrimer protein in the assembly. We envision that the lysine-containing nanocarriers may have stronger binding with insulin than the arginine-containing ones. Therefore, a lysine-containing telodendrimer scaffold, PEG5k(LysLys-L-R)4 (Figure 2b), was applied for telodendrimer nanocarrier synthesis, where PEG5k, LysLys, L, and R represent a polyethylene glycol (PEG, 5 kDa), two lysines with primary amino groups, a flexible linker, and a PBM, respectively. A number of biocompatible small molecules were collected into a model library (Figure 2c) for virtual screening of insulin binding by molecular docking. The docking energies were ranked by the average of 100 docking assays. Aromatic ring structures are often used to endow polymer nanocarriers with hydrophobic features for protein delivery,4,14,15 so we included some aromatic molecules for virtual screening. Simple aromatic structures in our library, such as 2-phenylacetic acid (PA) and phenylalanine (Phe), as well as the aromatic-rich molecules of anthraquinone (AQ) and rhein (Rh), showed relatively high (unfavorable) average docking energies (Ea), implying poor binding affinity with insulin. An obvious approach to improve the binding affinity is to build more PBMs in the telodendrimer nanocarriers. As evidenced by the docking results in Figure 2c, the Ea of the bivalent LysRh2 was decreased to −5.11 kcal· mol−1 from −3.91 kcal·mol−1 for monomeric Rh. A similar result was observed for LysCA2 (Ea: −5.26 kcal·mol−1), although it was much weaker than the Ea sum for two monomeric cholic acids (CAs) (Ea: −4.03 kcal·mol−1). It can be explained by the restriction of the size and shape of the binding sites on protein surfaces for the dimers of CA or Rh to be fully integrated in protein binding. Moreover, the increased amount of hydrophobic PBMs in PEG5k(LysLys-L-R)4 may break the balance of the hydrophobicity and hydrophilicity of a nanocarrier and cause precipitation. Instead, efficient monomeric protein-binding units can be identified with optimal interactions with the target protein at a specific site and globally. Among the monomeric PBMs, vitamin E (VE, D-αtocopherol) has both aliphatic chain and aromatic ring structures and exhibits the strongest docking with insulin (Ea: −4.55 kcal·mol−1), which is nearly comparable to the Ea of bivalent LysRh2 and LysCA2. Once an optimized PBM is identified, the density of PBMs, charges, and PEG chain length can be further optimized precisely and systemically. To validate the above computational predictions, seven PBMs, including PA, Co, CHO, CA, C17, VE, and LysCA2 (full names and chemical structures, Figure 2c), were conjugated to PEG5k(LysLys-L-R)4 backbones for combinatorial synthesis of a telodendrimer library. As determined by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, the molecular weights of the telodendrimers

telodendrimers using representative PBMs with different rankings in virtual screening for systematic evaluation and methodology validation.

Figure 1. Schematic illustration of the rational design and combinatorial synthesis of telodendrimers for systematic evaluation and optimization of nanocarriers for protein delivery.

Insulin surface displays different regions composed of either neutral, positively/negatively charged, or hydrophobic residues (Figure 2a). The continuous hydrophobic regions of insulin provide multiple binding sites for PBMs, and the overall negative net charge of insulin calls for the application of a positively charged telodendrimer for insulin encapsulation.

Figure 2. (a) Surface structure of human insulin. (b) Chemical structure of the telodendrimer. (c) Representative protein binding molecules ranked by Ea for insulin binding in molecular docking. *Building blocks selected for telodendrimer synthesis. 268

DOI: 10.1021/acsmacrolett.6b00982 ACS Macro Lett. 2017, 6, 267−271

Letter

ACS Macro Letters are close to their theoretical values (Table S1 and Figure S1). The observed discrepancies are likely attributed to the enhanced molecular aggregation of amphiphilic telodendrimers, which hinders the desorption of telodendrimers during MALDI-TOF analysis, especially for higher molecular weight portions.16 The narrow polydispersity of telodendrimers, as analyzed by MALDI-TOF mass spectrometry (Table S1), indicates that the well-defined dendritic structures have been synthesized on the terminus of PEG chains. The proton nuclear magnetic resonance (1H NMR) spectra for the telodendrimers also confirm their well-defined chemical structures with calculated formula being close to the designed structures (Table S1 and Figures S2−S5). The colloidal behavior of telodendrimers was characterized by critical micelle concentration (CMC) and dynamic light scattering (DLS) measurements (Figures S6 and S7). Insulin forms aggregates (∼40 nm) in phosphate-buffered saline (PBS, pH 7.4) even in the presence of 1 mM ethylenediaminetetraacetic acid (EDTA), as determined by DLS (Figures S8) and transmission electron microscopy (TEM) (Figures 3a). Addition of telodendrimers to the insulin solution significantly improves the dispersity of insulin, leading to decreased particle sizes ranging from 12 to 25 nm (Figure 3a), which affirms the formation of insulin−telodendrimer nanocomplexes. The stability of the insulin−telodendrimer nanocomplexes was evaluated with an agarose gel retention assay in which free or loosely bound protein molecules can be separated from proteins stably loaded in the nanoparticles based on their sizes and net charges.5 Insulin mixed with PA-, Co-, and CHO-containing telodendrimers migrated to a distance similar to free insulin (Figure 3b), indicating poor stability of those nanoformulations under an electric field. In contrast, insulin was partially loaded in CA−telodendrimer nanoparticles, and most of the insulin was tightly loaded in C17−, VE−, and LysCA2−telodendrimer systems, suggesting good stability for insulin loading. This trend indicates that the insulin encapsulation stability in telodendrimer nanoparticles generally increases with decreasing Ea of the PBMs. Notably, the compounds of CHO, C17, and VE have close hydrophobicity with similar partition coefficient (log P) values of 8.23, 7.80, and 8.88, respectively. However, the CHO-− elodendrimer has significantly weaker insulin loading stability when compared to the C17 and VE analogues determined by the electrophoresis study (Figure 3b). It suggests that the hydrophobicity of the PBM is not the only determing factor to target hydrophobic regions on proteins, which likely requires conformational and structural matches. The kinetics of protein−telodendrimer binding was studied by biolayer interferometry (BLI)17 to illustrate the structure− property relationship in insulin encapsulation. Insulin was first covalently immobilized on BLI sensors. Then the sensors were incubated in different telodendrimer solutions with varied concentrations for association followed by dissociation in PBS at 37 °C, and the kinetics of the association/dissociation phases for different telodendrimers are shown in Figures 3c and S9. Significant association of the VE−telodendrimers on insulinimmobilized sensors and slow dissociation in buffers were observed (Figure 3c), indicating strong interactions between VE−telodendrimers and insulin. The association constant (kon) and dissociation constant (koff) were obtained by globally fitting the association−dissociation kinetics to a 1:1 model algorithm (Figures S10a and S10b). Among the tested telodendrimers, the VE−telodendrimer showed the smallest koff (Figure S10b),

Figure 3. (a) TEM images of free insulin and insulin-loaded telodendrimer nanoparticles (insulin/telodendrimer, 1/3, w/w) with negative staining. The corresponding hydrodynamic diameters (Dh) are shown above the TEM images. (b) Loading ability and stability of telodendrimer nanocarriers for FITC insulin determined by an agarose gel retention assay. (c) Representative BLI kinetics for association in telodendrimer solutions at concentrations ranging from 33 to 600 nM (