Recent trends in protein and peptide-based biomaterials for advanced drug delivery

pmc Adv Drug Deliv Rev Adv Drug Deliv Rev 3815 pheelsevier Advanced Drug Delivery Reviews 0169-409X 1872-8294 pmc-is-collection-domain yes pmc-collection-title Elsevier - PMC COVID-19 Collection PMC7456198 PMC7456198.1 7456198 7456198 32871201 10.1016/j.addr.2020.08.008 S0169-409X(20)30119-8 1 Article Recent trends in protein and peptide-based biomaterials for advanced drug delivery Varanko Anastasia 1 Saha Soumen 1 Chilkoti Ashutosh ⁎ Department of Biomedical Engineering, Duke University, Durham, NC 27708, USA ⁎ Corresponding author. 1 These authors contributed equally to this work. 2020 29 8 2020 156 364177 133 187 30 6 2020 14 8 2020 14 8 2020 29 08 2020 31 08 2020 03 03 2023 © 2020 Published by Elsevier B.V. 2020 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Engineering protein and peptide-based materials for drug delivery applications has gained momentum due to their biochemical and biophysical properties over synthetic materials, including biocompatibility, ease of synthesis and purification, tunability, scalability, and lack of toxicity. These biomolecules have been used to develop a host of drug delivery platforms, such as peptide- and protein-drug conjugates, injectable particles, and drug depots to deliver small molecule drugs, therapeutic proteins, and nucleic acids. In this review, we discuss progress in engineering the architecture and biological functions of peptide-based biomaterials —naturally derived, chemically synthesized and recombinant— with a focus on the molecular features that modulate their structure-function relationships for drug delivery. Keywords Drug delivery Polypeptides Recombinant proteins Bioinspired materials Hierarchical self-assembly pmc-status-qastatus 0 pmc-status-live yes pmc-status-embargo no pmc-status-released yes pmc-prop-open-access yes pmc-prop-olf no pmc-prop-manuscript no pmc-prop-legally-suppressed no pmc-prop-has-pdf yes pmc-prop-has-supplement no pmc-prop-pdf-only no pmc-prop-suppress-copyright no pmc-prop-is-real-version no pmc-prop-is-scanned-article no pmc-prop-preprint no pmc-prop-in-epmc yes 1 Introduction Research in the past several decades has made it evident that the overall therapeutic benefit of a drug is not proportional to its in vitro potency. Under physiological conditions, drugs encounter biological barriers, such as insolubility, aggregation, degradation, the impermeability of vascular endothelial cell layers, clearance by the kidney and the reticulo-endothelial system that contribute to a drug's short in vivo half-life, non-specific tissue distribution and poor tissue penetration, inefficient cellular internalization, undesired immunogenicity, and off-target toxicities [ 1 , 2 ]. In addition, a drug's potency can be severely compromised due to environmental changes such as pressure, temperature, humidity, and pH, which can occur during storage, administration, or systemic circulation. These factors create a narrow therapeutic window and can result in dismal in vivo performance, thus making the clinical translation of an otherwise potent drug an uphill task. To overcome these challenges, controlled drug delivery systems have been developed to improve the stability, efficacy, and tolerability of existing drugs while mitigating their off-target toxicity and promoting patient compliance. An ideal drug delivery system should be non-toxic, non-immunogenic, and biodegradable, and the architecture, chemical functionality, biological interactions, and mode of administration should all be tailored to optimize the drug's pharmacokinetic (PK) and pharmacodynamic (PD) properties [ 1 , 2 ]. Three major drug delivery strategies have been widely exploited to favorably alter the PK and PD properties of a drug: (i) using a prodrug in which a small moiety is covalently conjugated to the drug, masking its bioactivity until it is activated by a disease-specific stimulus at the desired site; (ii) encapsulating the drug in a delivery vehicle that dictates the PK and PD through its physicochemical properties; and (iii) using an implantable drug-eluting depot or device [ 3 ]. It is important to note that various hybrid approaches —combining more than one delivery strategy discussed above— are used to create a bespoke delivery system. Researchers have explored many synthetic and natural carrier molecules to create drug delivery systems [ 4 ]. Of these, proteins and peptides have garnered significant attention due to their structural diversity, biocompatibility, ability to form hierarchical self-assembly ranging from the nano- to meso- scale, exquisite tunability, non-immunogenicity, ease of synthesis, and scalability [ [5] , [6] , [7] , [8] ]. Because of these diverse traits, peptide materials have been the focus of many innovative drug delivery systems in the past several decades. In this review, we will focus on drug delivery systems that have been engineered from protein materials, including silk, albumin, keratin, collagen, gelatin, elastin, and resilin (

Fig. 1 ). We will discuss the guiding principles for the design of peptide- and protein-based delivery systems and customization of their physicochemical properties for specific applications in drug delivery. Fig. 1 List of protein-based materials discussed in this review. Created with Biorender.com Fig. 1 2 Early advances in protein biomaterials Controlled drug delivery was first conceptualized at the turn of the 20th century by Paul Ehrlich, a Nobel laureate who envisioned a “magic bullet” therapy that could deliver drugs to a specific target while avoiding off-target toxicity [ 9 ]. At that time, drugs were typically administered in pill formulations that instantly released the drug upon contact with water. There was no control over the release kinetics, which led to unsteady drug concentrations [ 10 ]. Small molecule drugs were rapidly cleared from circulation, thus necessitating multiple doses to maintain the drug's minimum effective concentration. Not only are such dosing regimens inconvenient to the patient, they also increase the occurrence of dose-dependent side effects. The lack of target specificity made it impossible to deliver a therapeutic to the organ of interest without impacting other cells, which caused off-target toxicity. In the 1950s, scientists began to focus on developing drug delivery systems ( Fig. 2

). In the next thirty years, they established the principles of drug diffusion, dissolution, and pharmacokinetics [ 11 , 12 ]. Research then shifted toward prolonging drug release and increasing the drug's retention time in circulation [ 11 ]. Scientists also focused on how to specifically deliver drugs to a disease site by engineering systems for local drug release, passive drug accumulation in the diseased tissue, and active targeting [ 12 ]. This research led to the first FDA approval of a drug delivery system —liposomal amphotericin B— in 1990 for treatment of fungal infections [ 13 ]. Since then, controlled delivery has been leveraged to improve the bioavailability and efficacy of numerous therapeutics [ 14 ]. Fig. 2 A brief timeline of advances made in engineering protein-based biomaterials for controlled drug delivery. Fig. 2 Biomaterials have been critical to the success of drug delivery systems. Many drug formulations employ biomaterials to extend the therapeutic window by both sustaining its release from the formulation and slowing its elimination from the body. Beginning in the late 1960s, scientists used synthetic materials, especially polymers, as drug carriers [ 11 , 12 ]. These systems successfully extended the drug's circulation time by increasing its molecular weight to slow renal excretion. Biodegradable systems were also engineered to prolong drug release, thus reducing the frequency of administration [ 15 ]. Despite these accomplishments, synthetic materials can have high immunogenicity or toxic degradation products [ 16 ]. Furthermore, there is limited control over the stereochemistry, structure, and molecular weight of synthetic polymers, which impacts the drug's biodistribution and pharmacokinetics [ 17 , 18 ]. The production of synthetic polymer drug carriers can be difficult and expensive to scale up [ 19 ]. These challenges associated with synthetic polymers led to an interest in the use of natural materials, such as polysaccharides, lipids, nucleic acids, and proteins, as drug carriers. Herein, we will focus on the progress made in the design of protein-based drug delivery systems [ 20 ]. Using protein-based materials in a biomedical setting is not a new concept, as such materials had been used to treat injury or illness for centuries [ 21 ]. The renewed interest in protein-based materials for drug delivery was driven, in part, by the development of better methods to extract proteins from their natural sources, and techniques to characterize them [ 6 ]. Protein-based materials are relatively biocompatible, and their degradation products —amino acids— are nontoxic [ 5 ]. The next major advance that transformed this field was the advent of recombinant DNA technology, which has made it possible to design polypeptides de novo as drug delivery carriers and to customize native proteins for drug delivery applications by manipulating their amino acid sequences. The first class of polypeptides used for biomedical applications are biopolymers based on consensus sequences from naturally derived proteins [ 10 ]. These polypeptides likewise benefit from sequence-level control over their structures and bioactivities, low monodispersity, and lack of toxicity. A second class are de novo designed polypeptides [ 22 ]. For both classes of polypeptide carriers, these methods provide near-absolute control over the carrier's sequence, self-assembly, stimuli-responsiveness, and dispersity that cannot be matched by synthetic polymers [ 23 ]. Further, active functional groups can be readily introduced into the sequence for chemical modification and drug conjugation [ 24 ]. Recombinant techniques are also used to optimize native proteins for a given application by creating fusions of protein and peptide drugs with carriers or targeting proteins. Because these fusions are genetically encoded, this is accomplished with greater precision than is possible synthetic carriers by site-specific introduction of new functional groups in recombinant proteins by unnatural amino acids or post-translational modification [ 5 , 25 , 26 ]. Furthermore, molecular simulations have enabled de novo design of proteins for a given application [ 27 ]. As the field shifts its focus toward “smart” drug delivery systems, the unique properties and exquisite tunability of protein-based materials continue to be attractive. They have been designed to respond to a variety of stimuli, including temperature, pH, oxidative conditions, or the presence of specific biomolecules [ 6 , 28 ]. Furthermore, they can be engineered to self-assemble into a variety of architectures ranging from nanomaterials to hydrogels to porous scaffolds [ 29 ]. These architectures provide numerous opportunities to create precisely engineered drug delivery systems (

Fig. 3 ) . Fig. 3 Recent advances in the engineering of the peptide-based biomaterials as delivery vehicles. Created with Biorender.com . Fig. 3 3 Pathophysiological and translational challenges in drug delivery Drug delivery systems are designed by optimizing pharmacokinetic–pharmacodynamic behavior, which involves modifying components of the delivery system to overcome the pathophysiological and translational challenges discussed below. Solubility : Drugs must be soluble in blood to achieve prolonged circulation following systemic administration. Hydrophobic drugs with low aqueous solubility can be administered systemically by combining them with surfactants; however, these surfaces often pose health risks [ 30 ]. An alternative is to sequester hydrophobic drugs in peptide-based or polymeric delivery systems (which are surfactant-free) to improve drug safety and efficacy. Degradation and clearance : Enzymes in blood can degrade or deactivate drugs, shortening their half-lives. Drug carriers shield drugs from enzymes and other degradative factors. The size, charge, and hydrophobicity of a drug dictate its clearance rate. If a drug is below the glomerular filtration cutoff of ~60 kDa or 6 nm diameter, it will be cleared rapidly from systemic circulation via renal filtration. Drug carriers can increase the half-life of small-molecule drugs in plasma by increasing their effective size to greater than the renal filtration cutoff [ 31 ]. Small, positively-charged drugs are cleared preferentially from circulation via negatively-charged capillary walls in the kidney's glomerulus [ 32 ]. A drug's charge and hydrophobicity also affect its clearance via opsonization, a process in which blood proteins adsorb to a drug and trigger degradation by the mononuclear phagocytic system [ 33 ]. Shielding the solvent-accessible interface of the drug with a “stealth” carrier can prevent protein adsorption and minimize opsonin-mediated uptake by macrophages. Accumulation : Drug accumulation in the target tissue remains the major challenge for optimizing therapeutic efficacy. Drug carriers have been developed to improve accumulation in diseased tissue by passive targeting (e.g. in tumor tissue via the enhanced permeation and retention (EPR) effect [ 34 ] and by active targeting, wherein drug carriers are decorated with a ligand that binds a receptor that is overexpressed in the diseased tissue [ 35 , 36 ]. Tissue penetration : Once a drug extravasates to the target tissue, it often faces an environment rich in endothelial cells and extracellular matrix that prohibits further transport deep into the tissue [ [37] , [38] , [39] ]. This reduced drug transport reduces drug efficacy and can lead to drug resistance. Drug delivery systems can be loaded with penetrating moieties or modified to tune their physicochemical properties to enhance delivery into the target tissue [ 38 ]. Cell uptake and subcellular trafficking : After reaching the desired tissue, the drug must traverse cell membranes to reach its intracellular target. Cell membranes are permeable to small hydrophobic drugs, but for large and/or hydrophilic drugs, the cell membrane is an impermeable barrier [ [39] , [40] , [41] ]. Once a drug is internalized by a cell, additional barriers within the cell may separate the drug from its therapeutic target. Directing a drug to its subcellular site of molecular action is the penultimate challenge in drug delivery [ 42 ]. Release kinetics : Once a drug carrier permeates the cell membrane, its drug cargo can be released by passive mechanisms such as diffusion or by active mechanisms such as stimuli-responsive release [ 43 ]. Drug release can be triggered by the acidic and enzyme-rich environment of the endosome and lysosome, or by the reducing environment of the cytosol. Most drug carriers are internalized by endocytosis and thus must be designed to escape the endosome to prevent degradation of the drug before it can reach its intracellular therapeutic target. The delivery system must be stable while also allowing spatiotemporal control of drug release. The simplest method to load a drug into a carrier is by physical entrapment, in which the physicochemical properties of the drug are matched to the properties of the carrier to drive encapsulation. Alternatively, drugs can be covalently conjugated to a functionalized carrier. This approach requires reactive groups that do not interfere with the therapeutic function of the drug or the self-assembly of the carrier [ 43 ]. Translational challenges : Once a drug formulation is optimized, its potency is evaluated in preclinical in vitro and in vivo experiments. Preclinical testing is required by regulatory agencies prior to clinical studies. The low correlation of preclinical results with clinical efficacy and the limited relevance of in vitro and non-human in vivo models used in preclinical evaluation are challenges in translating novel protein-based drug carriers to the clinic. The high cost of drug development—up to $10 million for Phase I, $20 million for Phase II, and $50–100 million for Phase III clinical trials [ 44 ]—creates a prerequisite that intellectual property cover a new drug delivery system to ensure exclusivity so as to recover the cost of drug delivery development. 4 Design of peptide-based drug delivery systems to overcome pathophysiological and translational challenges The physicochemical properties of peptide-based delivery systems can be engineered with near absolute precision, which is impossible with synthetic polymers. The vast repertoire of natural, recombinant, and artificial proteins and the ability to customize their amino acid sequences allow construction of on-demand delivery systems. In this section we discuss intrinsic design modules ( Fig. 4

) that can be systematically engineered to create a drug delivery system with tunable physicochemical properties, to overcome the physiological and translational challenges discussed above. The relationship between the structure of these modules and their function as a therapeutic delivery system is described below for each major component of the modules. Fig. 4 Intrinsic design modules of a peptide amenable to precision engineering for drug delivery application. Created with Biorender.com . Fig. 4 Targeting ligand and linker : Like traditional drug formulations, peptide-based drug carriers accumulate in diseased tissue either by passive targeting, which includes diffusion and, in the case of tumor tissue, the EPR effect, or by active targeting via receptor-ligand interactions [ [34] , [35] , [36] ]. Tissue penetration can be enhanced by using cell-penetrating peptides [ 45 ]. The valency and surface density of targeting ligands impacts the affinity of the delivery system for its target receptors and impacts tissue accumulation [ 46 ]. Suboptimal display of targeting peptides on the vehicle surface may inhibit targeting. To address this issue, Wang et al. used a heuristic approach to determine the optimal presentation of a targeting peptide on a delivery vehicle [ 47 ]. By using 98 combinations of 15 tumor-homing peptides presented via 8 peptide linkers (differing in length and charge) on the surface of a delivery vehicle, they showed that two factors—nanoparticle charge and surface hydrophilicity—are critical in determining peptide presentation, and that an intervening peptide linker consisting of hydrophilic and charged residues (e.g. lysine and aspartic acid) prevents undesirable insertion of hydrophobic ligands into the micelle corona or micelle core. These charged linker residues can also be used to counteract extra charged groups within the micelle to create a neutral nanoparticle surface that minimally interferes with electrostatic interactions between ligand and receptor. Peptide backbone and self-assembling block : The peptide backbone and self-assembling block are the heart of the peptide-based drug delivery system. The choice of peptide backbone and self-assembling block dictates the physicochemical properties of the system, which include its (soluble vs. gel), stability, size (nanoscale to mesoscale), morphology (spherical vs. elongated), charge, solubility, payload encapsulation and release, and response to external stimuli. Collagen-based peptides form a stable gel with a slow rate of degradation; gelatin is used when more rapid degradation is required [ 48 ]. A crosslinkable peptide can be used to increase the in vivo stability of a gel [ 49 ]. Albumin is widely used as a soluble drug carrier [ 50 ]. Gliadin (from gluten) facilitates penetration of drug carriers into the gastric mucosa and is used to deliver drugs to the stomach [ 51 ]. The choice of peptide building block can also impact the encapsulation and release of a drug. Due to their negative charge, keratin [ 52 ] and silks [ 53 ] are used to entrap positively-charged molecules and are hence rarely used to deliver negatively charged nucleic acid-based therapeutics. A hydrophilic peptide is desirable to deliver a hydrophobic drug. For example, a paclitaxel-loaded nanoparticle composed of hydrophilic zwitterionic polypeptide showed a wider therapeutic window —the range of drug dose that could treat solid tumors without toxic side-effects— than a neutral elastin-like polypeptide with the same molecular weight [ 54 ]. The stimuli-responsiveness of the delivery system is also governed by the choice of peptide building blocks. Elastin and collagen are widely used to synthesize temperature-responsive drug delivery vehicles, whereas silk is often used as a pH-responsive building block. The hydrophobic / self-assembling block impacts the molecular architecture of the self-assembled peptide-based delivery system. For example, increased resilin content in the self-assembling block in an elastin-resilin fusion peptide shifts the self-assembled morphology from spherical micelles to elongated “worm-like” micelles, and can increase the avidity of a micelle with exposed peptide ligands that bind the α v β 3 integrin receptor by 1000-fold compared to a monomeric ligand [ 55 ]. Lipids and hydrophobic peptides like resilin are used not only to drive self-assembly but also to entrap hydrophobic drugs. Similarly, incorporation of a β-sheet-forming peptide into an elastin-lipid fusion drives self-assembly of morphologies ranging from worm-like micelles to bundled fibers [ 56 ]. Drug-binding domain : Lysine and cysteine are the two most widely used amino acids for covalent conjugation of a drug to a peptide carrier; a drug can be functionalized with N-hydroxy succinimide to target the amino group of a lysine, or with a maleimide to target the thiol group of a cysteine. Selective release of a drug from a drug-binding domain at the target tissue can be achieved by using pH-, redox-, or enzyme-labile bonds between the drug and carrier [ 57 ]. However, chemical conjugation of a therapeutic payload should not interfere with the activity of the drug and disrupt the self-assembly of the carrier in an obstructive manner. For this reason, recombinant strategies are emerging for site specific conjugation of small-molecule drugs to therapeutic or targeting proteins. One such approach exploits sortase mediated ligation that relies on the specificity of the transpeptidase Sortase A (SrtA) for short peptide sequences. SrtA retains its specificity while accepting a wide range of potential substrates [ [58] , [59] , [60] , [61] ]. 5 Applications in drug delivery Protein-based biomaterials have revolutionized drug delivery by providing many unique structural and physico-chemical properties to delivery systems. Numerous techniques have been implemented to engineer protein materials with exceptional release profiles, pharmacokinetics, targeting capacity, and safety. In this section, we will discuss how proteins have been modified and leveraged to improve the delivery of a wide range of therapeutic agents. 5.1 Silk Silk is a water insoluble, fibrous protein produced by silkworms like Bombyx mori and spiders such as Araneus diadematus and Nephila clavipes [ 62 ]. Silk is composed of two main proteins - sericin and silk fibroin (SF). Sericin is a hydrophilic, amorphous protein composed of 18 nonrepetitive amino acids and forms approximately 25% of the total weight of raw silk. It acts as an adhesive to join fibroin filaments. Sericin may cause immunogenic reactions and is thus separated from the SF for biomedical applications [ 63 ]. SF offers a repertoire of materials systems for biomedical applications, such as injectable particles, bioadhesives, hydrogels, implantable scaffolds, and recombinant or chemical conjugates. 5.1.1 Structure and properties of silk fibroin SF is a high MW protein complex composed of a light chain (MW ~26 kDa) and a heavy chain (MW ~390 kDa) covalently held together with a single disulfide bond while non-covalently encapsulating a 25 kDa glycoprotein, P25. The fibroin heavy chain is an amphiphilic block copolymer consisting of alternating hydrophobic and hydrophilic blocks. The hydrophobic, crystallizable blocks, responsible for forming the β-sheet structure, are composed of a highly repetitive dipeptide motif of Gly-X, where X can be alanine, serine, tyrosine, or valine in decreasing frequency. The shorter, hydrophilic, amorphous blocks are composed of nonrepetitive sequences [ 64 , 65 ]. Its unique structure endows SF with highly adaptable properties: (i) its high thermal stability and mechanical malleability are suitable for further processing, such as chemical modification, material fabrication, and sterilization [ 4 , 64 , [66] , [67] , [68] , [69] ]; (ii) in response to external stimuli, SF can self-assemble into various structures ranging from nanoparticles to hydrogels by modulating its β-sheet content [ 64 , 67 ]; (iii) the side chains of SF contain an abundance of active functional groups that enable chemical modification, and new functional groups can be incorporated to tune self-assembly, biodegradation, and payload release [ 69 ]; (iv) its anionic charge can be exploited to deliver a positively charged payload [ 53 ]; (v) recombinant DNA technology provides a modular platform to further engineer SF and create fusions with bioactive peptides [ [70] , [71] , [72] , [73] ]; and (iv) SF is completely biodegradable and biocompatible and has high immunogenic tolerance [ 67 , 68 ]. 5.1.2 Silk nanoparticles Silk fibroin nanoparticles (SFNPs) have been extensively studied as injectable drug carriers to control the release of bioactive substances both in vivo and in vitro [ 74 ]. Their wide therapeutic application stems from the fact that their properties, including size, shape, zeta potential, and secondary structure, can be modified during self-assembly by external stimuli, such as pH [ 53 ], salt concentration [ 53 ], or the amount of co-solvent [ 75 , 76 ]. Lammel and coworkers prepared SFNPs in which the pH of the solution could control the secondary structures and zeta potential of the nanoparticles by salting out a silk fibroin aqueous solution with potassium phosphate [ 53 ]. At pH 6, the SFNPs predominantly resembled a silk II (crystalline) structure, whereas at pH 9, particles were composed of silk I (less crystalline). The authors also proposed a model to predict the effect of pH and kosmotropic salts on particle formation. Model small molecule drugs, such as alcian blue, rhodamine B, and crystal violet, were loaded into SFNPs by absorption, and their release was governed by SF crystallinity; more crystalline structures demonstrated a greater release rate. Shi et al. synthesized a SFNP for loading and release of hydrophobic small molecules and protein therapeutics [ 77 ]. Over 50 days, 23% FITC-BSA and 34% rhodamine B were released from the SFNPs and internalized by cells, as seen by microscopy and flow cytometry. Crivelli et al. synthesized a SFNP using a desolvation technique to encapsulate the anti-inflammatory drugs celecoxib (CXB) or curcumin for osteoarthritis (OA) treatment [ 78 ]. The release of the drug was controlled by varying the drug loading into the SFNP. In vitro release experiments indicated that the release reached equilibrium after 24 h, which was much faster than the release of the same drugs from silk based hydrogel systems. Covalent functionalization can be used to tune self-assembly and interactions with therapeutics and the biological environment. The active amino acid residues on SF, such as serine, threonine, aspartic acid, glutamic acid, and tyrosine, make it amenable to chemical functionalization to modulate its properties for a given application. For example, SFNPs 40–120 nm in diameter were synthesized by an acetone extraction method and conjugated to insulin with glutaraldehyde as a crosslinker [ 79 ]. Insulin-conjugated SFNPs were resistant to trypsin digestion and had a half-life 2.5 times higher in human serum compared to bare insulin, which demonstrates the potential of SF nanoconjugates for peptide or enzyme delivery. Recombinant silk-like peptides (SLPs) have also been used to create nanoparticles with reproducible sizes for drug and gene delivery. The negatively charged SF cannot complex with nucleic acids through electrostatic interaction, thus limiting its applications in gene therapy. To address this limitation, Numata et al. recombinantly synthesized silk-based block copolymers with poly( l -lysine) domains for gene delivery. The pDNA complexes of silk-polylysine prepared at a polymer:nucleotide ratio of 10:1 showed the highest transfection efficiency. The pDNA complexes were also immobilized on silk films and could directly transfect cells from these surfaces [ 80 ]. Recombinant SLP sequences derived from the native sequence of the dragline protein MaSp1 sequence from the spider Nephila clavipes were combined with a polylysine domain to produce hybrid systems that formed nanocomplexes of varying sizes based on the polymer to pDNA ratio or the molecular weight of the polylysine domain [ 80 , 81 ]. Transfection efficiency was also significantly enhanced by introducing cell-specific targeting groups like the arginine-glycine-aspartic acid (RGD) tripeptide [ 82 ]. Fusion of a cell-penetrating peptide with an Sp1-based SLP produced a delivery vehicle that was 45-fold more efficient at transfection than poly(ethyleneimine) at low pDNA concentrations. Hybrid SFNPs have also been reported for injectable drug delivery. Yang et al. synthesized a multifunctional SF@MnO 2 nanoparticle-based platform using SF as a reductant and template via a one-step biomineralization-inspired crystallization process ( Fig. 5

) [ 83 ]. The authors took advantage of the mesoporous structure and carboxyl residues of the SF@MnO 2 nanoparticles to conjugate the photodynamic agent indocyanine green (ICG) and the chemotherapeutic drug doxorubicin (DOX) to form a SF@MnO 2 /ICG/DOX nanocomplex (SMID). TEM images suggested that the SMID nanocomplex possesses a well-defined spheroid structure and an average diameter of 60 nm. The presence of MnO 2 made it highly reactive with endogenous hydrogen peroxide, which decomposed into O 2 to enhance tumor-specific photodynamic therapy (PDT). In addition, the SMID nanocomplex demonstrated a stable photothermal effect upon near-infrared (NIR) irradiation for photothermal therapy (PTT) due to the photothermal response of SF@MnO 2 and conjugated ICG. In vivo NIR fluorescence and magnetic resonance (MR) imaging indicated significant accumulation of the SMID nanocomplex in the tumor, which consequently improved tumor regression efficacy after a combination of PTT, PDT, and DOX chemotherapy [ 83 ]. Mao et al. conjugated the cyclic pentapeptide cRGDfk that targets the α v β 3 integrin and the photodynamic agent Chlorin e6 (Ce6) to SF polypeptides using a simple acid-amine coupling reaction and genipin peptide as a crosslinker to formulate a 5-fluorouracil (5-FU) -doped SFNP ( Fig. 6

) [ 84 ]. The authors investigated the active targeting properties and photodynamic effects of the nanoparticles in vitro. Results revealed that treatment with multifunctional SFNP and infrared radiation produced high levels of reactive oxygen species (ROS) and induced cell death in MGC-803 gastric cancer cells. The multifunctional SFNP, which combined targeted 5-FU chemotherapy with PDT, induced a remarkable antitumor effect in a xenograft mouse model for gastric cancer. However, the low colloidal stability of SFNPs under physiological conditions restricts widespread use of these system in vivo. Shao and colleagues addressed this issue by fabricating SFNP composite materials with a core–shell structure (CS-SFNPs). The authors electrostatically coated the negatively charged SFNPs with four different cationic polymers —glycol chitosan, N , N , N -trimethyl chitosan, polyethyleneimine, and PEGylated polyethyleneimine. Dynamic light scattering and nanoparticle tracking analysis revealed that CS-SFNPs had much greater colloidal stability than bare SFNPs in biological media. Fig. 5 Synthesis procedure of SF@MnO2/ICG/DOX (SMID) nanoparticles. SMID nanoparticle acts as a multifunctional drug delivery platform for in vivo MR/fluorescence imaging-assisted tri-modal therapy of cancer. Adapted with permission from [ 83 ]. Fig. 5 Fig. 6 Conjugation of the cyclic pentapeptide cRGDfk and the photodynamic agent Chlorin e6 (Ce6) to SF was achieved using a simple acid-amine coupling reaction and the resulting conjugate was doped with 5-fluorouracil (5-FU) using genipin peptide as a crosslinker. Adapted with permission from [ 84 ]. Fig. 6 5.1.3 Silk hydrogels and depots The adaptable mechanical properties and thermal stability of SF make it an ideal candidate material for forming hydrogels and scaffolds. Its degradation can be tuned by controlling the self-assembly of SF which can stably encapsulate the therapeutics and release them on demand. Kaplan and colleagues have extensively studied the abilities of SF and its hybrids to form hydrogels. They developed a method to synthesize a thixotropic silk nanofiber hydrogel from aqueous solution, which otherwise requires an organic co-solvent [ 85 , 86 ]. The injectable nanofiber hydrogel stably entraps DOX, solidifies in situ, and demonstrates pH-triggered sustained release of DOX [ 86 ]. Kaplan et al. also used an injectable SF-hydrogel system to sustain the delivery of anti-vascular endothelial growth factor (anti-VEGF) therapeutics [ 87 ], and small molecule drugs [ 88 ]. However, the prolong gelation time (weeks-months) of SF hydrogel acts as a major roadblock for their practical application. To decrease the gelation time Gong et al. synthesized another class of thixotropic hydrogels by blending regenerated SF and hydroxyl propyl cellulose (HPC), a natural cellulose ether approved by FDA [ 89 ]. HPC has lower critical solution temperature (LCST) of about 40°C above which it precipitates in water. The HPC-SF blend gelled at 37°C within 1 h. Results from confocal laser scanning microscopy (CLSM), Raman spectroscopy, and 13 C NMR spectroscopy suggested that the conformational transition of SF from random coil to β-sheet during phase separation resulted in gel formation through β-sheet crosslinking and immobilization of the molecules of SF and HPC in the dispersed phase. The blended hydrogel encapsulated mice fibroblasts and protected them against high shear force during injection, which suggests that the hydrogel can be used for cell delivery [ 89 ]. Germershaus et al. developed heuristics to decipher protein interactions with SF using protamine and polylysine as model proteins [ 90 ]. The author concluded that the interaction between protein and SF arises primarily from entropy-driven complex coacervation, which depends on the ionic strength of the solution and presence of kosmotropic and chaotropic salts. SF-hydrogels have also been fabricated with various nanoassemblies including SFNPs, carbon nanotubes and hybrid nanoparticles of different materials. Mao and colleagues encapsulated curcumin-loaded cationic nanoparticles of RRR-α-tocopheryl succinate-grafted-ε-polylysine conjugate into a SF-hydrogel, which promoted the penetration of curcumin into the thickening corneum of psoriatic mice and thus inhibited skin inflammation. Compared to 49% of curcumin released from the nanoparticle, only 30% was released from the mixed hydrogel-nanoparticle system. The author concluded that this slow release of curcumin might be due to adherence of the nanoparticles into the SF hydrogel, making it difficult for the embedded curcumin to diffuse out of the gel. This delayed release profile resulted into significant accumulation of curcumin in the stratum corneum at the 48 h [ 91 ]. Wu et al. incorporated salinomycin and paclitaxel-loaded SF-nanoparticles into a SF hydrogel to inhibit cancer stem cell and tumor growth. The dual drug-loaded hydrogel had homogeneous drug distribution and exhibited increased tumor inhibition compared to the single drug-loaded hydrogel, as evidenced by fewer CD44 + CD133+ tumor cells in vivo. Because paclitaxel and salinomycin interacted differently with SF, their release profiles were different, with paclitaxel showing sustained release and salinomycin an initial burst release [ 92 ]. Gangrade et al. introduced carbon nanotubes into SF hydrogels to construct an on-demand, tumor-targeting system [ 93 ]. They synthesized folic acid functionalized, DOX-loaded, single-walled carbon nanotubes (SWCNT) and incorporated them into an SF hydrogel matrix. Only 7% of DOX was released from the composite material over a period of five days under physiological conditions. However, intermittent exposure to near-infrared light stimulated on-demand DOX release (~15%) due to the photothermal property of SWCNT. He et al. developed an injectable silk fibroin nanofiber hydrogel system complexed with upconversion nanoparticles and nano-graphene oxide (SF/UCNP@NGO) for upconversion luminescence imaging and photothermal therapy [ 94 ]. The NaLuF 4 :Er 3+ ,Yb 3+ upconversion nanoparticles were complexed with the nano-graphene oxide and then doped into an aqueous SF solution to form a hybrid hydrogel system. SF/UCNP@NGO hydrogels efficiently ablated 4T1 breast cancer cells via the photothermal effect both in vitro and in vivo. Recombinant techniques can also be utilized to modulate the properties of silk-based scaffolds and hydrogels. Anderson et al. recombinantly synthesized SLP segments conjugated to a human fibronectin segment and cell attachment domain [ 95 ]. Crystal structure characterization of (GAGAGS)-based SLPs revealed the hydrophobic domains of silk collapsed to form anti-parallel β-sheets that assembled into crystallite whiskers at the nanometer scale. This self-assembled material is mechanically tough, can undergo processing in the presence of bioactive molecules, and can be processed for into thin films, hydrogels, and three-dimensional scaffolds. Schacht et al. fabricated highly porous foams made from recombinant spider silk protein eADF4(C16) and a variant containing an RGD motif using a salt-leaching technique [ 96 ]. In contrast to other salt-leached silk scaffolds, the swelling behaviors of these scaffolds as measured by Discovery V20 stereomicroscope were low, and the mechanical properties were suitable for soft tissue engineering. The compressive moduli of the foam in a hydrated state was 3.24 ± 1.03 kP at a protein concentration of 8% ( w / v ). The pore size and porosity of the foams were optimized by altering the salt crystal size, thus rendering them suitable to adhere and culture fibroblasts. As silk fibroin can self-assemble into hydrogels from an aqueous solution, it has been widely used for ocular delivery of various drugs ranging from small molecules to antibodies and other therapeutic proteins. In one such example, silk hydrogel formulations of bevacizumab, a clinically used antibody as angiogenesis inhibitor showed sustained release of the antibody over a three months' period in an intravitreal injection model in Dutch-belted rabbits [ 87 ]. The bevacizumab concentration in the vitreous humor on day 90 using hydrogel formulation at both standard (1.25 mg /50 μl injection) and high dose (5.0 mg /50 μl injection) was equivalent to the levels achieved with positive control on day 30 (1.25 mg bevacizumab/50 μl injection). [ 87 ]. This concentration is estimated to be the therapeutic threshold based on the current dosage of 1 injection/month. These gels also got degraded after 3 months, indicating a repetitive dosing may be possible. Additionally, the propensity of SF to bind positively charged molecules through electrostatic interaction can be exploited for topical drug delivery on the eye surface. Dong et al. electrostatically coated an ibuprofen-encapsulated liposome of cationic lipids with SF for ocular drug delivery [ 97 ]. The SF being a potential mucoadhesive biopolymer aided in retention of the drug on eye surface and facilitated its sustained release. Recently a phase II clinical trial ( NCT03889886 ) has been undertaken to evaluate the ocular and systemic safety and efficacy of SDP-4, a naturally occurring silk-based ophthalmic solution in subjects with moderate to severe dry eye disease over a 12-week treatment period. 5.2 Keratin Keratin is a fibrous structural protein abundant in epithelial cells of human and animal skin, hair, nails, scales, feathers, and other epidermal appendages. It forms the body's protective barrier by facilitating cell-to inter-cellular adhesions. Keratins are highly tunable and responsive, which makes them ideal materials for drug delivery applications. Like other protein carriers, keratins benefit from a high MW and avoid rapid renal clearance, thus increasing the circulation half-life of their cargo. They are very durable and stable due to their high level of intramolecular bond formation. Furthermore, the unique amino acid composition of keratins allows them to stably interact with a variety of therapeutics and respond to numerous biological stimuli [ 98 ]. 5.2.1 Structure and properties of keratin Unlike other proteins reviewed herein, keratins have a high cysteine content that provides the protein with mechanical, chemical, and thermal stability. Typically, a greater cysteine content, which results in more disulfide bonds within the protein, creates harder keratins such as those found in hair and nails. Soft keratin, which is found in skin, has fewer disulfide bonds. In addition to being biocompatible, biodegradable, non-toxic and tunable, keratins can also contain cell adhesion motifs, a useful characteristic for facilitating drug delivery [ 99 ]. Keratins can be subdivided into three molecular configurations. The first, α-keratin, has a relatively low sulfur content and an α-helical structure with four intertwined, right-handed helices [ 100 ]. These α-helices form the fibers that are abundantly found in soft tissues. The second, β-keratin, is rich in glycine, lysine, histidine, alanine, serine, and tryptophan. It forms a β-sheet structure stabilized by hydrogen bonds and provides rigidity to skin, scales, and nails. The third configuration, γ-keratin, has high levels of cysteine, glycine, and tyrosine. It has an amorphous structure that forms many intermolecular and intramolecular disulfide bonds. It is a key matrix protein that primarily holds α-keratin fibers together and provides mechanical strength to hair. Keratins are negatively charged; thus, positively charged therapeutics can easily adhere to the surface of keratin hydrogels or nanoparticles. Carboxyl, amine, and carbonyl groups in the protein also provide useful drug attachment sites, either through hydrogen bond formation or chemical conjugation. Keratins have cell-targeting capacity due to the presence of cell attachment sites, including RGD sequences and leucine-aspartic acid-valine (LDV) sequences, within the protein. Additionally, disulfide bonding, electrostatic interactions, and hydrogen bonding provide keratin with mucoadhesive properties, which make it a useful material for delivery of drugs to the gastrointestinal tract [ 101 ]. Furthermore, keratins have been employed as “smart” materials for stimulus responsive drug delivery. The large number of carboxyl groups within their sequence makes them pH-sensitive so that in response to an increase in pH, release their cargo in a controlled manner as these groups become deprotonated [ 102 ]. Additionally, their rich cysteine content renders them redox-responsive. Tuning the number or level of crosslinking of disulfide bonds within the protein alters the duration of drug release from a keratin-based material. Similarly, keratin is responsive to changes in glutathione (GSH) concentration. This is particularly useful for delivery of chemotherapeutics to metastatic cancer cells, which have significantly higher concentrations of GSH than healthy cells. Moreover, the high lysine and arginine content in some keratins enables their cleavage by high concentrations of trypsin [ 103 ], which is frequently overexpressed in inflamed tissue [ 104 ]. Thus, keratin could be a useful material when targeting injured or tumorous tissues. 5.2.2 Keratin nanoparticles Due to its unique material properties, keratin has been used to create nanoparticles that encapsulate and sequester a therapeutic cargo before releasing it in response to biological stimuli. These nanoparticles can stably carry therapeutics via electrostatic interactions, hydrogen bonding, disulfide bond formation, or chemical conjugation. Keratins are durable and remain stable in the bloodstream, which increases the half-life of its cargo [ 100 ]. Positively charged drugs electrostatically adsorb to the surface of negatively charged keratin nanoparticles; this interaction provides long-term, sustained release of the drug from its carrier. Zhi et al. first explored this strategy by complexing the model drug chlorhexidine (CHX) with keratin nanoparticles generated by ionic gelation [ 102 ]. Carboxylate groups on the nanoparticle surface stabilized the polyanion complex with CHX. This interaction provided a remarkable CHX encapsulation efficiency of 91.2% and a loading content of 9.2%. Zhi et al. demonstrated that for 140 h, the drug was released in a pH-dependent manner with greater release observed at neutral and slightly acidic pH, which indicates the utility of this system to deliver chemotherapeutics to the acidic tumor microenvironment [ 102 ]. Similarly, Li et al. demonstrated that keratin-based drug-loaded nanoparticles (KDNPs) can be loaded with DOX via electrostatic interactions [ 105 ]. These nanoparticles were designed to exploit keratin's responsiveness to pH and glutathione concentration to release the drug, which is a useful strategy to treat solid tumors, which have a pH of 6.2–6.9 and GSH concentration of 0.5–10 mM, which is approximately 10-fold greater than the GSH concentration of healthy tissue [ 105 ]. KDNPs were created with a desolvation method that destabilizes keratin with ethanol and causes it to aggregate into nanoparticles that are crosslinked with glutaraldehyde. In an environment with a low pH or high concentrations of glutathione, KDNPs experienced a stark increase in zeta potential and underwent a negative-to-positive charge conversion. The positive charge facilitated internalization by cells and electrostatically repelled the drug, thus accelerating its release ( Fig. 7

). Moreover, the charge conversion destabilized the nanoparticles, causing them to aggregate and accumulate at the tumor due to the enhanced permeation and retention (EPR) effect, which further enhanced their anti-cancer potency [ 105 ].The stimulus-responsive properties of KDNPs were expanded by Li et al., whoe engineered triple stimuli-responsive KDNPs via drug-induced ionic gelation. In addition to releasing DOX in response to changes in pH and glutathione concentration, these nanoparticles broke down in the presence of high concentrations of trypsin, which digested the peptide bonds within the keratin [ 106 ]. Similarly, keratin graft poly(ethylene glycol) nanoparticles have also been explored as glutathione-responsive vehicles; DOX-HCl entrapped in the disulfide-crosslinked keratin core was released in response to increased intracellular glutathione concentrations [ 107 ]. Fig. 7 Dual stimulus responsiveness of keratin-based drug loaded nanoparticles (KDNPs). (A) Schematic of nanoparticle fabrication and GSH- or pH- stimulated drug release (B) An acidic environment shifts the size and zeta potential of KDNPs and accelerates the release of entrapped DOX (C) The presence of GSH shifts the size and zeta potential of KDNPs and accelerates the release of entrapped DOX. Adapted with permission from [ 105 ]. Fig. 7 Due to its exquisite and versatile loading capacity, keratin has also been harnessed to create bimodal nanoformulations that combine chemotherapeutics and photodynamic therapy. Martella et al. functionalized high molecular weight keratin with the photosensitizer Chlorin-e6 (Ce6) and induced spontaneous nanoparticle formation by mixing the protein with paclitaxel, a chemotherapeutic that aggregates with the hydrophobic residues in keratin [ 108 ]. This is an attractive bottom-up nanoparticle fabrication strategy, as it did not require any toxic crosslinkers or downstream purification steps for nanoparticle fabrication. When administered to an osteosarcoma cell line, the nanoparticles localized to the cell lysosome. Although Ce6 typically experiences a drop in fluorescence in acidic conditions, the keratin protected the photosensitizer and no decrease in fluorescence was observed. Further, the nanoparticle formulation transported the paclitaxel in a three-dimensional tumor model system without reducing the drug's potency [ 108 ]. Keratin nanoparticles have also been engineered for mucoadhesive drug delivery. Kerateine (KTN) and keratose (KOS) are two forms of keratin that have been extracted and processed in its reduced and oxidized forms, respectively. Cheng et al. demonstrated that, by altering the KTN:KOS ratio, the mucoadhesive properties and thus drug release, gastric retention time, and bioavailability of keratin nanoparticles could be tuned [ 109 ]. An evaluation of the hydrophobicity, surface charge, and terminal groups of the nanoparticles revealed that the mucoadhesive properties of KTN were dominated by electrostatic interactions, whereas those of KOS were primarily due to hydrogen bonding with gastric mucin. Moreover, Cheng et al. discovered that gastric retention time decreased with an increase in KOS, and release of the model drug, amoxicillin, increased with a larger proportion of KTN due to its pH responsiveness [ 109 ]. 5.2.3 Keratin films Keratin-based films have been widely explored for biomedical applications due to the presence of cell attachment sites, their biocompatibility, and their large surface area. These mechanically and chemically stable materials have been useful for delivering a variety of drugs and peptides. In an early study of keratin films, Fuji et al. extracted keratin from human hair in the absence of surfactant, thus creating a water-soluble film comprised primarily of α-keratins [ 110 ]. Alkaline phosphatase, a model enzyme, was incorporated into the film by mixing it with the keratin prior to gelation. The biochemical properties and bioactivity of alkaline phosphatase were maintained for two weeks after loading [ 110 ]. While keratin films have excellent biocompatibility, they lack mechanical strength. Thus, many studies combine keratin with other materials or treat the protein with a crosslinking agent to improve its rigidity, stiffness, and stability. In one example, keratin was blended with SF to create a composite film for the delivery of Bowman-Birk inhibitors (BBI), synthetic peptides designed to inhibit elastase in wound healing applications [ 111 ]. SF provided structural stability to the film while keratin governed the degradation and BBI release rates. Using FITC -tagged bovine serum albumin (BSA) as a model protein, Vasconcelos et al. demonstrated that, although the SF was compact and rigid, increasing the percentage of hydrolytic keratin could increase the rate of FITC-BSA release by film degradation and diffusion [ 111 ]. Another study used transglutaminase (TGase) to crosslink keratin films to improve the mechanical strength and chemical stability of the material [ 112 ]. Treatment with TGase resulted in more compact network formation and increased the mechanical strength from 5.18 MPa to 6.22 MPa. This reduced the solubility of the film and thus delayed the release of the model drug, diclofenac [ 112 ]. These studies illustrate how keratin materials can be modified to improve drug elution for wound healing and tissue engineering applications. 5.2.4 Keratin hydrogels Keratin hydrogels have been widely studied for biomedical applications, due to their stability, durability, and large number and variety of attachment sites for drugs or cells. The latter makes keratin hydrogels attractive for tissue engineering and regenerative medicine applications [ 100 , 101 , 113 ]. Additionally, keratin hydrogels have been explored for local, controlled delivery of small molecule drugs and macromolecules. Drug release is mediated by keratin degradation, the rate of which can be systematically tuned by controlling the number and type of bonds within the hydrogel [ 111 ]. The simplest keratin hydrogels rely on electrostatic interactions to load the drug of interest, which is typically positively charged, onto a keratose-based hydrogel. This strategy was used by Saul et al. to locally deliver and sustain the release of ciprofloxacin, a broad-range antibiotic [ 114 ]. Ciprofloxacin was loaded into the hydrogel by mixing it with the keratose solution before gelation. Because keratose is processed with sulfonic acid, it lacks disulfide bonds, and its gelation relies on hydrophobic interactions and physical chain entanglement. Saul et al. observed that ciprofloxacin release strongly correlated with keratose degradation. Approximately 40% of the drug was released within the first 24 h, followed by a period of linear release for the next six days, with release detectable for up to three weeks. Drug bioactivity was not impacted by its incorporation into the hydrogel [ 114 ]. Halofuginone, a type I collagen synthesis inhibitor, was loaded into a keratose hydrogel by the same technique [ 115 ]. One day after administration, the drug was released at a steady rate from the hydrogel for four days. After seven days, 60% of the halofuginone had been released, consistent with the degradation profile of the keratose hydrogel. As with the ciprofloxacin study, halofuginone remained bioactive [ 115 ]. These studies highlight the applications for keratin hydrogels in local drug release. The rate of drug release from keratin hydrogels can be modified by altering the crosslinking density through alkylation. Han et al. alkylated KTN by treating keratin with iodoacetamide to “cap” the cysteine thiol groups and thus modulate the number of disulfide bonds within the protein [ 116 ]. Three therapeutics – ciprofloxacin, recombinant human insulin-like growth factor 1 (rhIGF-I), and recombinant human bone morphogenic protein (rhBMP-2) – were loaded into the hydrogel. Han et al. found that the increased rate of hydrogel degradation correlated with the amount of iodoacetamide used, though it was not directly proportional due to the binding affinity of the model drug to the keratin [ 116 ]. A disulfide shuffling strategy was used by Cao et al. to explore how disulfide bond formation impacts drug release. They cleaved intramolecular disulfide bonds with a reductive agent (e.g., cysteine) to free thiol groups that could then form intermolecular disulfide bonds (

Fig. 8 A) [ 117 ]. This tactic increased the mechanical strength of the hydrogel, reduced its gelation time, and required a lower amount of keratin for gelation. The release rates of ciprofloxacin and DOX were inversely proportional to the level of cysteine in the hydrogels. Hydrogels that entrapped ciprofloxacin also demonstrated zero-order release kinetics in PBS ( Fig. 8 B & C). Moreover, when these hydrogels were exposed to increasing concentrations of glutathione, the degradation rate increased in response to the more rapid degradation of disulfide bonds ( Fig. 8 D & E) [ 117 ]. These studies indicate that the degradation rate of keratin hydrogels can be altered for specific disease states and drugs. Fig. 8 Disulfide shuffling to modulate drug release from keratin hydrogels. (A) Schematic of gelation by the disulfide shuffling strategy. Crosslinking density and thus microstructure, mechanical strength, and degradation can be tuned with this method (B,C) Hydrogel degradation and ciprofloxacin release in PBS demonstrate that greater crosslinking, represented by a greater number of Cys residues, slows degradation and release. (D,E) Hydrogel degradation and DOX release in the presence of GSH demonstrates the redox responsiveness of these hydrogels. Adapted with permission from [ 117 ] . Fig. 8 pH-responsive keratin hydrogels have also been designed to reversibly swell in response to their environment to modulate the release of their cargo. In many of these examples, keratin has been combined with itaconic acid, N-isopropyl acrylamide, or methacrylic acid to achieve pH responsiveness [ [118] , [119] , [120] ]. More recently, Peralta Ramos et al. reported a stimulus-responsive hydrogel that did not depend on chemical grafting to achieve its mechanical properties and pH-responsivity [ 121 ]. This was credited to a novel synthesis strategy during which disulfide bridges were broken and the α-keratin reorganized. At an acidic pH, the carboxyl groups were protonated and formed hydrogen bonds, which created a higher fraction of β-sheet conformations and kept the gel in a collapsed state. At a basic pH, hydrogen bonds broke to create more sites for water adsorption, thus forcing chain reorganization and allowing the material to swell [ 121 ]. Villanueva et al. expanded upon this work by incorporating antimicrobial ZnO nanoplates into the pH-responsive hydrogel [ 122 ]. Upon administration to a chronic wound, the hydrogel swelled in response to the basic environment and locally released the biocidal agent. As the wound healed and the presence of microbes decreased, the gel collapsed and inhibited the release of ZnO [ 122 ]. This study demonstrates how keratin hydrogels can be used to adjust the delivery of a therapeutic agent in response to alterations in pH. 5.3 Albumin Albumin is the most abundant protein in human plasma and has a set of properties that make it a unique molecular carrier for drugs: (i) it is a natural physiological carrier of native ligands and nutrients; (ii) it bypasses systemic clearance and degradation by the body's own innate mechanisms, so that it has an exceptionally long half-life of 19 days in humans, and similarly long half-lives in most animal species [ [123] , [124] , [125] , [126] ]; (iii) it preferentially accumulates at sites of vascular leakiness; (iv) it is highly internalized and metabolized by rapidly growing, nutrient-starved cancer cells; and (v) it is biodegradable and has no known systemic toxicity. Because of these properties especially its extraordinarily long plasma circulation, albumin has attracted a lot interest as a carrier for diverse drugs. 5.3.1 Types of albumin Three naturally derived albumin molecules have been used to address the delivery challenges associated with small molecule drugs. Ovalbumin (OVA), a highly functional food protein with a molecular weight of 47 kDa and isoelectric point (pI) of 4.8, is a monomeric phospho-glycoprotein consisting of 386 amino acid residues [ 127 ]. Each OVA molecule has one internal disulfide bond and four free sulfhydryl groups. OVA was originally chosen as a carrier for drug delivery because of its availability, low cost, ability to form gel networks and stabilize emulsions and foams, and pH- and temperature-sensitive properties [ 128 ]. Bovine serum albumin (BSA), which has a molecular weight of 69 kDa and a pI of 4.7 in water (at 25°C), has also been widely used for drug delivery because of its abundance, low cost, ease of purification, unusual ligand-binding properties, and wide acceptance in the pharmaceutical industry [ [129] , [130] , [131] , [132] , [133] , [134] ]. However, due to possible human immunologic response to OVA and BSA in vivo, human serum albumin (HSA) is now exclusively used for drug delivery [ 50 ]. HSA, which has a MW of 66.5 kDa, is the most abundant serum protein with a concentration of 30–50 g/l in human serum. It is long-circulating with an exceptionally long in vivo half-life of ~19 days [ [123] , [124] , [125] , 135 ]. About 10–15 g of albumin is synthesized by liver hepatocytes daily and released into circulation [ 136 ]. When albumin extravasates into tissue, it is naturally recycled and returned to the vascular space via the lymphatic system. The same approximate mass of 10–15 g of albumin entering the intravascular space is also catabolized daily. HSA is a carrier of a wide variety of endogenous and exogenous compounds and facilitates the colloidal solubilization and transport of hydrophobic molecules, such as long chain fatty acids, and variety of other ligands, including bilirubin, hormones, amino acids, metal ions, and drugs [ 136 , 137 ]. 5.3.2 Structure and properties of HSA HSA is a soluble, globular, monomeric protein consisting of 585 amino acid residues. It contains 35 cysteinyl residues that form one sulfhydryl group and 17 disulfide bridges [ 123 , 135 , 138 ]. Fig. 9 A shows the crystal structure of albumin and the sites where ligands can bind [ 139 ]. Three distinct features of albumin make it ideal for use in drug delivery applications: binding, trafficking, and recycling. Fig. 9 Ribbon diagram of the three-dimensional structure of human serum albumin (A). Schematic of the in vivo thiol-maleimide conjugation reaction (B). Chemical structure of DOX-EMCH/Aldoxorubicin (C). Adapted with permission from [ 140 ] (A) and [ 144 ] (B). Fig. 9 Binding : Each class of drugs has a distinct binding affinity to different binding sites of albumin. Dicarboxylic acids and sterically demanding anionic heterocyclic molecules (e.g., Warfarin) bind to Sudlow's site I ( Fig. 9 A), whereas aromatic carboxylic acids with a single negatively charged acid group separated by a hydrophobic center (e.g., diazepam, ibuprofen) bind to Sudlow's site II [ 140 ]. HSA has seven long-chain fatty acid binding sites (FA1–7 in Fig. 9 A, asymmetrically distributed throughout its three domains) that promote binding of up to two moles of fatty acid per mole of HSA under normal physiological condition [ [141] , [142] , [143] ]. Compounds with p-isothiocyanate (p-SCN) or NHS ester (N-hydroxysuccinimide) functional moieties covalently react with albumin's lysine, with Lys199 residue being the most reactive [ 144 , 145 ]. However, due to multiple other (at least ten) surface accessible lysine residues in HSA, reaction with lysine residues leads to poorly defined conjugates with a range of stoichiometries [ 146 , 147 ]. Alternatively, the solvent accessible cysteine34 in HSA can be selectively modified with a prodrug containing a maleimide functionality, as (i) 70% of circulating serum albumin possess one free cysteine; (ii) other major serum proteins lack a solvent-accessible free cysteine; and (iii) the 34th cysteine residue of albumin has the most reactive thiol group (pK a  ~ 7) in human plasma [ 144 , 148 ]. Trafficking : Albumin naturally transcytoses across the vascular endothelium, which normally creates an impermeable barrier to most plasma proteins. This process is attributed to the 60 kDa vascular endothelium receptor sialoglycoprotein (gp60). Albumin binds to gp60 and forms a cluster in association with Cav-1, the main protein critical to caveolae formation. Clustered albumin-gp60 receptors and compounds bound to albumin are then internalized and transported to the basolateral membrane to complete transcytosis [ 149 ]. Interestingly, modified albumins show preferential binding to gp18 and gp30 [ 150 ]. Preferential internalization of albumin in cancer cells has also been correlated with albumin binding to SPARC (secreted protein acidic and rich in cysteine). For example, immunohistochemical staining detected stromal SPARC in ∼70% of non-small cell lung cancer (NSCLC) cases, and increased chemotherapeutic efficacy of albumin-bound paclitaxel was correlated with high stromal SPARC reactivity of NSCLC cells [ 151 ]. Recycling : The long half-life of albumin is attributed to the neonatal Fc receptor (FcRn), a widely distributed intracellular receptor responsible for salvaging albumin from cellular catabolism [ 143 ]. FcRn binds to albumin in the acidic endosome, diverts it from the lysosomal degradation pathway, and the complex is exocytosed. Albumin is released from FcRn at extracellular pH and enters circulation through the lymphatics, thus prolonging its half-life. Albumin also avoids renal clearance by reabsorption through megalin and cubilin receptor-mediated endocytosis in the renal proximal tubule. 5.3.3 Endogenous albumin for drug delivery In situ albumin-binding drugs have been designed to covalently react with or noncovalently bind to endogenous albumin, which capitalizes on the long-circulating property of the protein to enhance the PK and hence PD of the drug. Indeed, exploitation of endogenous serum albumin has several advantages over the use of exogenous albumin: first, although commercial exogenous albumin can be isolated with high yield and purity, it is often contaminated with pathogens. Second, the cost, effort, and time required for manufacturing exogenous albumin is altogether avoided. Third, as no external macromolecular carrier is involved, the quality control associated with endogenous albumin is comparable to small molecule drug candidates. Irreversible covalent bond formation between endogenous albumin molecules and a therapeutic payload can be utilized to enhance the pharmacokinetics of the latter. Kratz et al. pioneered this strategy by synthesizing a prodrug that selectively reacts with the 34th cysteine residue of circulating serum albumin after systemic administration, which improves the drug's plasma half-life and protects it from premature degradation. The therapeutic cargo (e.g., DOX) was anchored with a thiol reactive maleimide group via a pH-sensitive hydrazone bond and a carefully optimized alkyl spacer. Upon intravenous injection, the highly reactive maleimide group formed an irreversible thioether bond with the thiol group of the albumin. The pH-sensitive hydrazone bond facilitated the liberation of the covalently attached DOX from albumin after en endocytosis by cells. Aldoxorubicin, also known as DOX-EMCH ( Fig. 9 B & C) is currently under various phases of clinical trial for the treatment of soft tissue sarcomas ( NCT01673438 and [ 152 ]) and small cell lung cancer ( NCT02235688 ). Inspired by DOX-EMCH, many other albumin-binding prodrugs have been developed. These prodrugs often consist of an anticancer drug, a maleimide group as the thiol-binding moiety, and a cleavable linker [ 144 ]. Native albumin-binding ligands such as fatty acids can be directly conjugated to therapeutics, which results in docking of the drug to endogenous albumin upon systemic administration. A notable example of this class is Semaglutide, an FDA approved drug that is a glucagon-like peptide-1 (GLP-1) analog with an octadecyl fatty diacid conjugated to the epsilon amine of a lysine residue in the peptide via a glutamyl ethylene glycol spacer [ 153 ]. The plasma half-life (t 1/2 , seven days in humans) and GLP-1 receptor (GLP-1R) binding affinity (K d  ~ 0.38 nM) of Semaglutide make it suitable for once-weekly administration for treatment of type 2 diabetes [ 153 , 154 ]. This was an important advancement, as its predecessor Liraglutide —which contains a hexadecyl monoacid conjugated to the same lysine residue of GLP-1 via a γ-Glutamic acid spacer— has a t 1/2 of only 0.5 days in humans, and is hence only approved for once-daily treatment of type 2 diabetes, despite having a three-fold higher affinity for GLP-1R (K d  ~ 0.11 nM) than Semaglutide. This dramatic improvement was attributed to both the spacer and ligand chemistry of the GLP-1 analog in Semaglutide, which influenced in vivo albumin binding [ 153 ]. Other therapeutic payloads are bound to endogenous HSA by conjugation of the drug to fatty acids through various cleavable linkers [ 125 , 144 ]. Synthetic small molecules can also bind endogenous albumin [ 144 ]. Most drugs and small molecules that interact with HSA are anionic, although a few cationic drugs have detectable affinity [ 142 ]. Evans blue (EB), an aromatic dye with four anionic charges, reversibly binds to serum albumin with a micromolar affinity. The affinity of EB for albumin allows 100% of the injected dose to be retained in the blood; thus, it can be used to quantify total plasma volume of a test subject [ 155 ]. Albumin-bound EB can also be used to assess blood brain barrier (BBB) permeability, as the BBB is impermeable to the complex in healthy, non-pathological conditions [ 156 ]. EB derivatives with various affinities to albumin have been used to deliver therapeutic cargo [ 144 , 157 ]. Yamamoto et al. developed an o -toludine EB analogue that chelates gadolinium (III), as a T 1 -weighted MRI contrast agent for imaging of blood vessels [ 158 ]. Following this seminal work, the Chen group developed the o -toludine EB analogue as a positron emission tomography (PET) tracer for in vivo labeling of serum albumin. They conjugated o -toludine EB to 1,4,7- triazacyclononane-N,N′,N″-triacetic acid (NOTA) to prepare a new analog named NEB. The NOTA chelator provides a simple radiolabeling route that allows labeling with different PET isotopes (e.g., 68 Ga, 64 Cu) [ [159] , [160] , [161] ]. Chen et al. evaluated labeled NEB in mice as a blood pool imaging agent under various pathological conditions, including myocardial infarction (MI) and serum leakage from permeable or abnormal blood vessels. Sentinel lymph nodes (SLNs) were visualized in all 24 pathologically diagnosed breast cancer patients within 4.0–10.0 (5.6 ± 1.4) min [ 161 ]. by analyzing 68 Ga-NEB PET/CT images. To identify molecular features required for albumin binding, the Neri group screened a DNA-encoded chemical library comprising >600 oligonucleotide-conjugates of potential albumin-binding molecules coded with unique six-base-pair sequences for identification [ 162 ]. The HSA-binding tags were chosen through Systematic Evolution of Ligands by Exponential Enrichment (SELEX). The selection, amplification, and microarray analysis of the pool suggested that following structural features are important for albumin binding: (i) the presence of a 4-phenylbutanoic acid moiety is essential; (ii) a butanoyl acid moiety increases albumin affinity as propanoyl and pentanoyl acid residue was not enriched in the pool; and (iii) substitution at the para position of the phenyl ring with a hydrophobic functionality increases the affinity toward albumin. Albumin-binding domains (ABDs) are small protein domains that non-covalently associate with serum albumin. ABDs are structurally robust and stable, as evidenced by their ability to withstand denaturation at extremely low pH (2.4) and high temperatures. and have also been explored for drug delivery. ABDs are used for drug delivery by either fusion with a therapeutic protein at the gene level if the protein is recombinantly produced or are chemically synthesized to conjugate small molecule drugs for systemic administration. Sjöbring and colleagues isolated a 14-kDa albumin binding protein fragment from Streptococcal protein G. This 46-amino acid native ABD (ABDN) is a 3-helix bundle and has nanomolar affinity for HSA [ 163 , 164 ]. Using ABDN as a template, Jonsson et al. identified an engineered version of ABDN from a phage library. Specifically, they targeted 15 residues in a combinatorial protein engineering strategy to identify an albumin -binding sequence with improved HSA affinity. The high -affinity ABD (ABDH) has superior thermal stability and binds HSA with femtomolar affinity. Chilkoti and colleagues recombinantly fused ABDN and ABDH to the N-terminus of a hydrophilic, thermally responsive chimeric polypeptide (CP) that contained multiple C-terminal cysteine residues [ 126 ]. The recombinant polypeptides formed highly monodisperse micellar nanoparticles upon covalent conjugation of multiple copies of hydrophobic DOX molecules to the. cysteine residues through the pH-sensitive hydrazone bond (using maleimide-thiol chemistry) ( Fig. 10 A). The DOX-loaded ABDN/ABDH-decorated polypeptide micelles —termed ABD-CP-DOX— bind both human and mouse serum albumin with high affinity, as seen by native PAGE and isothermal titration calorimetry ( Fig. 10 B & D ) . Hence, upon systemic injection, they instantaneously bind endogenous albumin with high affinity and are coated with an albumin corona. The albumin decorated ABD-CP-DOX nanoparticles had significantly superior PK in both murine and canine models compared to a negative control, CP-DOX nanoparticles that do not present ABD domains on the surface of the nanoparticle ( Fig. 10 E & F) The long plasma circulation of the ABD-CP-DOX nanoparticles also resulted in higher accumulation in tumors than the CP-DOX nanoparticles, and these nanoparticles also showed better tumor regression, and a wider therapeutic window than the CP-DOX nanoparticles. These results clearly show the enhancement in PK and PD conferred by decorating a drug-loaded nanoparticle with endogenous albumin. Fig. 10 Design of doxorubicin(DOX)-conjugated albumin-binding nanoparticles. DOX was conjugated via a pH-sensitive hydrazone linker. (B) Albumin-binding properties of ABDN-CP-DOX and ABDH-CP-DOX were qualitatively demonstrated using native PAGE. (C) Cryo-TEM images of ABDN-CP-DOX (I) and CP-DOX (II) micelles. (D) Isothermal titration calorimetry of ABDN-CP-DOX and ABDH-CP-DOX micelles with MSA. The solid red line represents the best fit of the binding isotherm. (E, F) Pharmacokinetics of ABD-decorated nanoparticles in murine (E) and canine (F) models. Adapted with permission from [ 126 ]. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Fig. 10 Furthermore, the same group took advantage of the native sortase reaction to directly install DOX onto the ABD (ABD-DOX) through a pH-sensitive linker, without forming nanoparticles [ 59 ]. ABD-DOX bound to both human and mouse serum albumin with nanomolar affinity, had a terminal t 1/2 of 29.4 h in mice, and increased tumor accumulation by ~120 fold compared to free DOX ( Fig. 11 A). In multiple mouse xenograft models, ABD-DOX resulted in greater tumor regression than Aldoxorubicin, DOX-prodrug under clinical development that was designed to covalently bind endogenous serum albumin. Other notable examples of albumin-based delivery systems involve the genetic fusion of ABD to various therapeutic proteins including affibodies [ 165 , 166 ], human soluble complement receptor type 1 [ 167 ], single chain antibody-drug conjugates [ 168 ], insulin-like growth factor II [ 169 ], immunotoxins [ 170 ], and respiratory syncytial virus subgroup A (RSV-A) G protein (G2Na) [ 171 ]. Fig. 11 (A) Synthetic scheme of ABD–DOX. Elastin-like polypeptide (ELP) was used as a purification tag and removed following drug conjugation using sortase A. Inclusion of the KEKE peptide at the N-terminus disrupted micellar self-assembly upon DOX conjugation and enabled the subsequent sortase A cleavage of ELP from the ABD–DOX conjugate. (B) Synthetic scheme of albumin−polymer−drug conjugates. Drug-containing monomer (drug: panobinostat, dark blue) was copolymerized with HPMA using RAFT agents, which allow one-step conjugation of albumin. Adapted with permission from [ 59 ] (A) and [ 172 ] (B). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Fig. 11 5.3.4 Exogenous albumin formulations for drug delivery Albumin-bound therapeutics can also be formulated ex vivo with purified albumin molecules prior to administration. In contrast to other physiological proteins, albumin tolerates a wide range of pH (stable in the range of 4–9), temperatures (can be heated at 60°C for up to 10 h), and organic solvents [ 50 ]. These properties have motivated researchers to covalently and non-covalently append various payloads, including radioisotopes (e.g., 18 F, 68 Ga, 111 In) and anticancer drugs (e.g., DOX, curcumin, methotrexate), to albumin for imaging and delivery purposes [ 125 , 144 ]. However, modification of exogenous albumin with drugs is often associated with suboptimal drug loading, as denaturation and albumin molecule crosslinking potentially compromise its physiological behavior. Smith et al. addressed these challenges by synthesizing an albumin-polymer-drug conjugate with high drug loading efficiency [ 172 ]. Using a convergent, reversible addition−fragmentation chain transfer (RAFT) polymerization reaction, they synthesized a polymeric prodrug of panobinostat, an FDA-approved anticancer agent, with HPMA (N-2-hydroxypropylacrylamide). Multiple copies of panobinostat were attached to the polymeric prodrug, which was then conjugated to albumin using NHS chemistry under physiological conditions, thereby minimally affecting the albumin molecule ( Fig. 11 B). The first albumin-drug conjugate that entered phase I/II trial was a methotrexate-albumin (MTX-HSA) conjugate synthesized using covalent coupling between carboxylate of the MTX and lysine residues of HSA. Seventeen patients not amenable to standard care were treated with up to eight injections given in weekly intervals. Tumor responses were seen in three with no sign of toxicity and drug accumulation. The MTD for weekly administration was found to be 4 × 50 mg/m 2 and a injection of 50 mg/m 2 every 2 weeks was recommended for a further study [ 173 ]. However, in a subsequent phase II study no objective responses were seen, although eight patients did not have disease progression (stable disease) for up to 8 months (median 121 days) [ 174 ]. A phase II study showed combination of MTX-HSA with cisplatin as an effective treatment modality against urothelial carcinomas with an acceptable toxicity profile [ 175 ]. Patients were treated with a loading dose of 110 mg/m 2 of MTX-HSA followed by a weekly dose of 40 mg/m 2 starting on day 8. Tumor response was observed in 7 patients. Complete response (CR) and partial response (PR) was observed in 1 patient each (overall response rate: 29%) but no follow up clinical investigation was undertaken. Albumin nanoparticle formulations represent the most widespread utilization of exogenous albumin as a carrier. Fabrication techniques such as desolvation, emulsification, thermal gelation, nano-spray drying, and self-assembly have been used to formulate albumin nanoparticles [ 125 ]. The most notable albumin nanoparticle is Abraxane, also known as nab-paclitaxel, an FDA approved paclitaxel formulation for the treatment of multiple cancer types [ 176 , 177 ]. Abraxane was developed by American Bioscience using the so-called nab-technology in which paclitaxel and human serum albumin are passed through a jet under high pressure to form nanoparticles with a mean diameter of 130 nm. Several albumin nanoparticle of small molecule drugs has been developed using nab technology and are under various phases of clinical trials [ 178 ]. Nab-rapamycin is currently under phase I clinical trial to treat non-muscle invasive bladder cancer ( NCT02009332 ) and under phase II clinical trial to treat progressive high-grade glioma ( NCT03463265 ). Nab-docetaxel proved to be effective against hormone refractory prostate cancer ( NCT00477529 ) and metastatic breast tumors ( NCT00531271 ). Nab-5404, a novel albumin formulation of thiocolchicine dimer exhibits dual inhibition of tubulin polymerization and topoisomerase I activities. It showed antiangiogenic and vascular targeting activities and was found to be effective against solid tumors and lymphomas ( NCT01163071 ). Lin et al. reported a method to simultaneously encapsulate paclitaxel by denaturing albumin with sodium borohydride at high urea concentration, followed by addition of hydrophobic drugs to the denatured protein, and diluting the mixture to spontaneously refold albumin and form nanoparticles [ 179 ]. This method avoids chemical crosslinkers and high-pressure emulsification techniques. By adding a cell-penetrating peptide to the dual drug-loaded albumin nanoparticles, they achieved BBB penetration through interaction with SPARC, which resulted in greater survivability in a U87 orthotopic glioma model [ 179 ]. Albumin has also been derivatized into a diblock format in which a hydrophilic albumin block gets exposed at the surface of the hybrid nanoparticle and a synthetic hydrophobic polymer, such as polycaprolactone, forms the core. To achieve this, Jiang et al. synthesized two different polymers based on ring-opening polymerization: poly(oligo-(ethylene glycol) methyl ether acrylate)–poly(ε-caprolactone) (POEGMA-PCL) and maleimide–functionalized polycaprolactone (MI-PCL). MI-PCL was conjugated to the free cysteine on bovine serum albumin (BSA-PCL) [ 180 ]. The authors formulated hybrid nanoparticles by co-assembling POEGMA-PCL, BSA-PCL, and curcumin with increasing albumin content. Cellular uptake of the nanoparticles in cancer cell lines increased with albumin content [ 120 ]. Additionally, albumin nanoparticles can be decorated with targeting ligands, such as mannose [ 181 ], folic acid [ 182 ], antibodies [ 183 , 184 ], and aptamers [ 185 ], to provide greater receptor specificity. Recombinant albumin is another alternative to animal-derived albumin and is increasingly being used in exogenous formulations. Direct genetic fusion to recombinant albumin enables one-step fabrication of therapeutic proteins. For instance, rlX-FP, a genetically encoded fusion protein linking recombinant human albumin with human coagulation factor IX, has a 5-fold longer plasma half-life than unmodified coagulation factor IX (FIX) products. An advanced phase III clinical trial (PROLONG-9FP) demonstrated the long-term safety, tolerability, and efficacy of rIX-FP for prophylactic and on-demand treatment of bleeding episodes in children [ 186 ]. Li et al. synthesized an albumin-lidamycin conjugate by recombinant DNA technology. Lidamycin, a cyclic enediyne antibiotic is ~1000 fold more potent than DOX against multiple cultured cancer cell line and it consists of an apoprotein (LDP) and an enediyne chromophore (AE) that can be separated and reassembled in vitro [ 187 ]. A DNA fragment encoding HSA-LDP was constructed and the fusion protein was purified from Pichia pastoris using IMAC. The HSA-LDP conjugate was then reconstituted with the AE to form the albumin-lidamycin conjugate that had a sub nanomolar IC 50 value in various cancer cell lines, significant tumor retention, and potent in vivo tumor regression efficacy against mouse hepatoma H22 model [ 188 ]. Other recombinant albumin proteins include genetic fusions with interleukin-2 [ 189 ] and interferons [ 190 , 191 ]. Albuferon® aka Albinterferon-α-2b was developed as a fusion protein of albumin and interferon-α-2b (INFα-2b) and is under phase III for the treatment of hepatitis C infection ( NCT00724776 ). Several other phase III trials are under progress aimed at evaluating the efficacy and safety of Albinterferon-α-2b [ 177 ]. However, high-yield synthesis of pure recombinant albumin and albumin fusions requires a complex procedure and specialized yeast expression system that are not readily available to most researchers. Nguyen et al. made significant progress by purifying recombinant HSA from a bacterial expression system using maltose-binding protein and protein disulfide isomerase [ 192 ]. But an optimized bacterial expression system to purify recombinant albumin is still needed. 5.4 Collagen Collagen is the most abundant protein in mammals, making up nearly one-third of all proteins in the body. A diverse family of structural proteins, collagen is a major component of the extracellular matrix (ECM) and connective tissue. It has a broad spectrum of functions, including cell adhesion, cell migration, tissue scaffolding, and repair [ 193 ]. To date, thirty different classes of collagen have been identified and characterized [ 194 ], though not all are relevant for this discussion. The most abundant class of collagen —fibrillar collagen— represents more than 90% of human collagen and includes types I, II, III, V, and XI. This class is a major component of skin, hair, ligament, tendon, cartilage, bone, and placenta. Other common types of collagens, like IV and VIII, construct the network structure of basement membranes [ 193 , 194 ]. Before discussing their molecular architecture and biomedical applications, it is important to define the term “collagens.” Gelatin, a thermally degraded, collagen-derived product, is also termed “collagen” in the literature despite losing the characteristics of real collagen, and will hence be discussed separately [ 195 ]. 5.4.1 Principles of collagen self-assembly Although the molecular architectures and functions of collagens vary greatly, they all share the same tertiary structure —the collagen triple helix [ 196 ]. Under the electron microscope, native collagen shows a thread-like structure —fibrils— consisting of collagen molecules organized in three interlocking polytripeptide chains that form a triple helix, with each chain having the general sequence GIy-X-Y, where X and Y are occupied by Proline (Pro) and (4 R )-hydroxyproline with the highest statistical frequency. Each collagen polypeptide chain is twisted around a threefold screw axis and exists in a secondary structure analogous to the left-handed polyproline II-helix. The three polyproline II-helices are held together by hydrogen bonding, which repeats periodically between the glycine amide in one helix and the carbonyl group of the amino acid residue of the neighboring helix. The three polyproline II-helices form a right-handed triple helix stabilized by hydrogen bonds [ 194 , 196 , 197 ]. With tunable self-assembly, gel-forming ability, biodegradability, and responsiveness to external stimuli, collagens have been increasingly used for drug delivery and tissue regeneration [ 195 , [198] , [199] , [200] ]. Current research is focused on developing short, bioinspired model collagens, termed as collagen-mimetic peptide (CMPs) or collagen-like peptides (CLPs) [ 201 , 202 ]. CLPs are short, synthetic peptides arranged in the same triple-helical conformation as native collagens. CLPs have been used: (i) to elucidate the triple helix structure and the molecular forces responsible for this architecture; (ii) to target pathological collagen in vivo through triple helix hybridization; and (iii) as a bioactive domain to construct smart biomaterials through hierarchical self-assembly. Each of these applications will be discussed separately as they are relevant for drug delivery. 5.4.2 Elucidation of the triple helix structure To exploit CLPs to fabricate biomaterials, it is critical to elucidate the intrinsic parameters affecting the sequence-based thermal stability of the triple helix. In contrast to native collagen, CLPs exhibit reversible phase transition behavior. Their slow folding rate upon denaturation and small size allow researchers to thermodynamically characterize the folding and melting processes of CLPs by X-ray crystallography, atomic force microscopy (AFM), light scattering, and circular dichroism (CD) and nuclear magnetic resonance (NMR) spectroscopy. Many researchers have synthesized and studied polypeptides with Gly-X-Y sequences that fold into a triple-helical structure. Persikov et al. investigated the role of guest amino acid residues at the X and Y positions on the thermal stability of the triple helix [ 203 ]. They documented that the guest triplets Gly-X-Hyp and Gly-Pro-Y can be used to quantitate the conformational propensities of all 20 amino acids to form a triple helix at the X and Y positions in a (Gly-Pro-Hyp) 8 host peptide. Persikov proposed that the triple-helical structure of CLPs is the direct result of the propensity of amino acids to adopt a polyproline II-like conformation, which is driven by the high propensity of ionizable residues in the X position for interchain hydrogen bonding and a low propensity of bulky residues in the Y position for effective solvation. Building upon this concept, many studies have deciphered the structure-function relationship of the CLP phase transition [ 201 , 202 , 204 ]. To improve the stability of the triple-helical structure of CLPs, C-terminal covalent crosslinking was performed using an interchain cystine knot derived from collagen type III [ [205] , [206] , [207] ]. Such covalent modification also includes the crosslinking of three α-chains at the C-terminus using various templates such as cis,cis-1,3,5-trimethylcyclohexane-1,3,5-tricarboxylic acid (KTA) and tris(2-aminoethyl)amine-(succinate-OH) 3 (TREN) [ 208 , 209 ]. Comparative analysis using CD and NMR spectroscopy revealed that the flexibility of the TREN scaffold is superior to that of the KTA scaffold for the induction of triple helicity [ 208 ]. However, to avoid cumbersome covalent crosslinking, simple noncovalent crosslinking strategies have been developed based on: (i) interchain metal bridging by functionalization of the terminus of CLPs with a chelating ligand [ 210 , 211 ] and (ii) hydrophobic interactions using a single saturated hydrocarbon or lipid tail [ [212] , [213] , [214] ]. The thermal stability of the triple helix of lipid-modified CLPs increased as the monoalkyl chain length increased from C 6 to C 16 [ 214 ]. Furthermore, introducing amino acids with guest residues that have a Cγ substitution with a highly electronegative atom, such as fluorine and chlorine, in place of hydroxyproline also increased triple helix stability [ [215] , [216] , [217] , [218] ]. It is important to note that the triple-helical structure restricts the sequence space available for synthetic CLPs, and until recently, was thought to be intolerant to substitution of glycine in the Gly-X-Y triplet. Multiple recent reports have shown that modification of glycine with a thioamide, nitrogen atom, or azaglycine (azGly) yielded comparable or better stability of CLPs compared to their non-substituted counterparts. Another recent approach incorporated metal-binding sites at the ends of CLPs to drive thermodynamically stable, nano-micro scale self-assemblies of various shapes [ 219 , 220 ]. Zheng et al. elucidated the role of surface electrostatics and hydrogen bonding on the stability of the triple helix and provided computational tools for de novo design of CLPs beyond the Gly-X-Y triplet [ 221 ]. Taken together, these structure-function studies indicate that nature may not have optimized the stability of the collagen triple helix, and that there is plenty of room for further improvement through precise design of CLPs beyond tandem arrays of the Gly-X-Y triplet sequence. 5.4.3 Single-stranded CLPs in targeting native collagens Although the major impediment in CLP research is deciphering the triple helix conformation and supramolecular assembly formation (discussed in the next section), biomedical applications of monomeric CLPs have gained attention in recent years using denatured collagen. Degradation of the ECM is a crucial element governing the progression of tissue remodeling in many life-threatening diseases, including cancer, cardiovascular diseases, and organ fibrosis, as well as prevalent debilitating conditions, such as arthritis and intervertebral disc degeneration. During tissue remodeling, collagen molecules within the collagen fibers and networks are degraded by proteases, such as MMPs or cathepsin, and denature at body temperature. In a seminal work, Yu and coworkers demonstrated that unfolded, single-stranded CLPs with a sequence (GPO) n (where n =6–10) have a high propensity for binding denatured collagen through a “strand hybridization process”, which is similar to the binding of complementary DNA strands [ 222 , 223 ]. Due to the high serum stability of such collagen -hybridizing peptides [ 224 ] and their affinity for native collagen [ 222 ], they have been used to selectively stain native collagen both in vitro and in vivo by fluorescently labeled CLPs [ [225] , [226] , [227] , [228] , [229] ]. Fluorescently labeled CLPs can also target solid tumors through the triple helix hybridization process, as high MMP-9 activity in tumor tissue exposes denatured collagen [ 230 ]. Because homotrimeric CLPs have little driving force for collagen hybridization, CLPs must be thermally dissociated by heating at 80°C to the monomeric state before binding to a collagen substrate. To avoid the preheating step, the Yu group developed a sequence (GfO) 9 by replacing the hydroxyproline with a fluoroproline (f) [ 231 ]. This new sequence cannot self-trimerize at body temperature but maintains the ability to hybridize with natural collagen chains. By appending an octa-glutamic acid residue at the N-terminus of the (GPO) 9 , the authors attracted vascular endothelial growth factors (VEGFs) to the collagen-binding site of the endothelial cells through charge-charge interactions, which resulted in tubulogenesis [ 232 ]. Triple helix hybridization has also been used to impart collagen-targeting properties to nanoparticles [ 233 , 234 ] and for sustained release of nucleic acids from collagen films and depots [ 235 , 236 ]. Recently, the Hubbell lab recombinantly fused a collagen-binding domain (CBD) to interleukin-12 ( Fig. 12

), a potent cytokine that stimulates the innate and adaptive immune system (CBD–IL-12). Intravenously administered CBD–IL-12 preferably accumulated in the tumor stroma and provided a sustained intratumoral level of interferon-γ, thereby eliciting superior anti-tumor effects and fewer off-target toxicities when compared to naked IL-12 [ 237 ]. Fig. 12 CBD–IL-12 binds to collagen with high affinity without compromising functionality. (A) Schematic of the fusion sites of the von Willebrand factor A3 CBD to the mouse p35 and p40 subunits via a (GGGS)2 linker. (B) Dose–response of phosphorylated STAT4 to IL-12 and CBD–IL-12 in preactivated primary mouse CD8+ T cells. EC 50 , half-maximum effective concentration; MFI, mean fluorescence intensity. (C, D) Binding of CBD–IL-12 to collagen I (C) and collagen III (D) as measured by SPR. The curves represent the specific responses (in resonance units (RU)) to CBD–IL-12. (E, F) Affinity of bare IL-12 (E) or CBD–IL-12 (F) to human melanoma cryosections was imaged using fluorescence microscopy. Scale bars, 100 μm. Adapted with permission from [ 237 ]. Fig. 12 5.4.4 Hierarchical self-assembly of CLPs and their hybrids: biomedical implications In the past few decades, the wealth of knowledge accumulated about the structure of collagen has driven a surge in research focused on the supramolecular assembly of collagen-like peptides. In the previous section, we described how various molecular interactions can stabilize the formation of a triple helix. In this section, we delve deeper into the molecular determinants of higher-order self-assembly of CLP–based biomaterials —beyond the triple helix— and how they can be used in drug delivery. We have organized this section by the specific molecular stimuli that trigger self-assembly. Cysteine knot in CLP self-assembly : In an early example, Raines and coworkers used two cysteine knots to covalently link three CLPs, one of which protruded out to produce a sticky end during triple helix assembly [ 238 ]. The sticky ends forced the trimers to configure themselves in a head-to-tail fashion, which resulted in a long, single collagen triple helix. Such collagen-like fibrils were much longer (400 nm) than the native type I collagen (300 nm). In a similar approach, the Koide group developed a synthetic CLP system in which three CLPs were preorganized in a staggered configuration that was locked in place by two cysteine knots [ 239 ]. CD, ultrafiltration, and laser diffraction analysis indicated that the staggered trimers formed large supramolecular architectures through intermolecular triple helix formation. However, hydrogels made of these collagen fibers did not successfully form due to concentration-dependent aggregation. To address this limitation, Yamazaki et al. appended a hydrophilic arginine residue at the cysteine knotted end of the synthetic CLPs [ 240 ]. The disulfide-linked trimers of the Gly-X-Y triplet repeats formed hydrogels by spontaneous intermolecular triple helix formation upon cooling. The thermal sol-gel transition is reversible, and the design of the peptides can tune the transition temperature. Koide's group also incorporated an integrin-binding sequence GFOGER into one CLP strand of the same knotted trimeric base unit. The supramolecular structure exhibited adhesion to human dermal fibroblasts that was comparable to natural collagens [ 241 ]. To further tune the gel properties, Ichise et al. appended multiple end-to-end disulfides through crosslinking of chemically synthesized triple-helical CLP. Rheology showed that gel stiffness is controlled by the number of cysteine residues up to three, at which point it plateaus. With the incorporation of an integrin α 2 β 1 -binding sequence, the peptide polymer showed receptor-specific cell binding in vitro. Moreover, cell signaling activity and biodegradability of such a system can be tuned by altering the content of integrin binding ligand in the peptide polymer and by varying the weight percentage of CLP [ 242 ]. π–π interaction in CLP self-assembly : The Maryanoff group introduced an aromatic-aromatic recognition motif at either the N- or C-terminus of each peptide chain of the CLP triple helix to drive the supramolecular assembly of CLPs [ 243 , 244 ]. The aromatic π–π interaction facilitated head-to-tail stacking of CLPs into a micrometer-sized fibrillar structure, as evidenced by CD, 1 H NMR, dynamic light scattering (DLS), TEM, AFM, and computational energy dynamics. The fibrillar CLPs induced the aggregation of human blood platelets, an ability lacking in less organized CLPs, with nearly the same potency as native type I collagen [ 245 ]. Kar et al. investigated how the presence of aromatic residues on neither end, one end, or both ends of a collagen model peptide affected the kinetics of self-association. CLPs with aromatic residues at both termini self-assembled and aggregated too quickly to be monitored by turbidimetry at a concentration of 7 mg/ml [ 246 ]. Similar interactions, such as CH-π interaction between amino acids (Proline and Hydroxyproline) and aromatic residues (Phenylalanine and Tyrosine) and cation-π interaction between a positively charged N-terminal Arginine and a C-terminal Phenylalanine, were also introduced to fabricate collagen-like fibrillar structures in a head-to-tail manner [ 246 , 247 ]. Hydrophobic interaction in CLP self-assembly : Hydrophobic interactions also contribute to CLP-hybrid self-assembly and collagen-mimicking properties. The Field and Tirrell research groups independently synthesized CLP-hybrids by conjugating collagen peptides with single or double hydrocarbon tails. Lipidation of CLPs increased the triple helix stability, triggered se