Protein-based Nanoparticles for Therapeutic Nucleic Acid Delivery

Abstract

To realize the full potential of emerging nucleic acid therapies, there is a need for effective delivery agents to transport cargo to cells of interest. Protein materials exhibit several unique properties, including biodegradability, biocompatibility, ease of functionalization via recombinant and chemical modifications, among other features, which establish a promising basis for therapeutic nucleic acid delivery systems. In this review, we highlight progress made in the use of non-viral protein-based nanoparticles for nucleic acid delivery in vitro and in vivo, while elaborating on key physicochemical properties that have enabled the use of these materials for nanoparticle formulation and drug delivery. To conclude, we comment on the prospects and unresolved challenges associated with the translation of protein-based nucleic acid delivery systems for therapeutic applications.

Graphical Abstract:

1. Introduction

Nucleic acid therapeutics have transformative potential across numerous areas of medicine. Approvals of the messenger RNA (mRNA) SARS-CoV-2 vaccines tozinameran (BioNTech/Pfizer)¹ and elasomeran (Moderna)² along with small interfering RNA (siRNA) therapies, such as inclisiran³ (Novartis/Alnylam Pharmaceuticals) by the EMA and FDA highlight the substantial impact this modality can have on patient care. To fully unlock the promise of nucleic acid therapies, there is a need for non-immunogenic, efficient delivery agents that overcome several in vivo barriers to transport nucleic acid cargo to cells of interest. While efforts in capsid engineering and envelope protein pseudotyping have yielded notable advances, existing viral vectors remain limited by stringent cargo capacities, immunogenicity, potential for oncogenic gene insertion, and manufacturing scaling challenges.^4,5 Lipid nanoparticles have developed into the leading non-viral delivery vector, but the incorporation of active targeting domains, such as monoclonal antibodies (mAbs) and single chain variable fragments (scFvs) requires chemical conjugation that is challenging to scale, and the human relevance of reported passive targeting strategies remains unclear.^6,7 Existing synthetic polymer and peptide-based (<50 residues) nanoparticle nucleic acid delivery systems often suffer from heterogeneous formulation characteristics, poor biocompatibility, and unfavorable pharmacokinetic performance that have limited the clinical progress of these vectors.^8,9

Protein materials exhibit several unique properties that form a promising basis for the design of clinically effective therapeutic nucleic acid delivery systems. Both recombinant and native protein materials are derived from or inspired by natural sources, and therefore are generally biocompatible and biodegradable. Moreover, protein nanoparticles derived from human proteins, such as albumin, ferritin, and elastin, do not contain non-native sequences. Consequently, they are less likely to be immunogenic as compared to viral and other non-viral delivery vectors containing non-native components. While nanoparticles derived from native proteins may often display some degree of heterogeneous physicochemical properties, those produced from recombinant proteins typically exhibit a high level of homogeneity and tunability. Recombinant proteins also enable precise control of features that influence physicochemical properties, such as size and surface charge, the incorporation of targeting motifs with specific stoichiometry, as well as binding domains to facilitate the complexation and uptake of therapeutic cargo. The ability to include such domains biosynthetically eliminates the need for additional chemical modifications. Collectively, these properties have motivated the exploration of both native and recombinant protein-based nanoparticles for a variety of drug delivery applications, including nucleic acid therapeutics.

In this review, we highlight the progress of engineered protein-based nanoparticles for nucleic acid delivery in vitro and in vivo. We consider protein materials to be those composed of greater than 50 amino acids, with insights regarding peptide-based (<50 residues) nucleic acid delivery systems reviewed elsewhere.^10–12 We specifically focus on the use of nonviral protein material; virus-like particles (VLPs) composed of viral proteins have been reported for nucleic acid delivery and progress in this area has been recently summarized.^13–15 Additionally, computationally designed, atomically precise protein nanoparticles have attracted significant interest, but have not yet been used for nucleic acid delivery.¹⁶ After describing the physicochemical properties and reported nucleic acid delivery applications of notable non-viral protein materials, we describe examples of the use of protein materials as safe and effective drug delivery vehicles for a range of non-nucleic acid cargo in humans. To conclude, we comment on the prospects and unresolved challenges for translating protein-based nucleic acid delivery systems.

2. Unique Barriers to Effective Protein Nanoparticle-mediated Nucleic Acid Delivery

Protein nanoparticles face multiple barriers to reaching their intended target, dependent on the route by which they are administered (Fig. 1). In order to be effective as a nucleic acid complexation agent, protein materials must bind anionic nucleic acids with high affinity, generally accomplished by leveraging electrostatic attraction or covalent conjugation in the case of low molecular weight cargo, such as siRNA or antisense oligonucleotides. After intravenous administration, protein-nucleic acid complexes must resist degradation by plasma proteases and nucleases. Protein-based nanoparticles, specifically those constructed using intrinsically disordered, unfolded protein materials, such as elastin-like polypeptides (ELPs) are vulnerable to protease degradation. While this degradation mechanism is favorable for material clearance from the body, it could be unfavorable for nanoparticle performance. A study of ELP degradation in full plasma suggests that rate of degradation is slow, on the order of days, as compared to distribution and elimination half-lives of the material.¹⁷ Given that the self-assembly of protein nanoparticles is concentration dependent in many cases, these particles are also prone to disassembly following dilution in the bloodstream. To counter this effect, protein particles are often stabilized using amino-reactive crosslinkers like glutaraldehyde or genipin, as well as cysteine-based disulfide bridge crosslinking.

Figure 1. Barriers to successful protein nanoparticle-mediated nucleic acid delivery.

Particles traverse a number of extracellular and intracellular barriers on the path to effective delivery outcomes (e.g. gene inhibition, gene addition, or gene editing). Created with BioRender.

Critically, protein-nucleic acid nanoparticles must also evade sequestration by the mononuclear phagocyte system (MPS).¹⁸ The adsorption of opsonins, including complement proteins and antibodies, among other plasma constituents onto thenanoparticle surface leads to rapid clearance from the bloodstream^19–23. Charge and hydrophobicity-mediated adsorption of plasma proteins with subsequent corona formation can also result in nanoparticle aggregation.^24–27 These effects have been mitigated, in part, by incorporation of poly(ethylene glycol) (PEG) or other hydrophilic polymers, as well as through the inclusion of zwitterionic moieties.^28–31 Recombinant incorporation of albumin binding domains has also been shown to improve the circulating half-life of protein nanoparticles, leveraging the long circulation time of serum albumin enabled by FcRN-mediated recycling.^32,33

After navigating the bloodstream, protein-nucleic acid nanoparticles traverse the capillary endothelium to enter the target tissue prior to reaching cells of interest. While nanoparticles tend to accumulate in the liver due to high blood flow, the presence of fenestrated endothelium, and uptake by endothelial and phagocytic cells^34,35, several strategies can be used to promote extrahepatic nanoparticle localization. Direct administration of nanoparticles to muscle via intramuscular injection, central nervous system (CNS) by intracranial, intranasal, or intrathecal administration, and the retina through intravitreal or subretinal routes can concentrate cargo delivery to these sites. Nonetheless, these approaches are suboptimal for many therapeutic indications due to their invasiveness and limited distribution. CNS nanoparticle localization can be promoted by ultrasound enhanced permeability across the blood-brain barrier (BBB), in addition to strategies that derivatize nanoparticles with ligands for endothelial apical transport receptors, which support transcytosis.^36–39 Nanoparticle accumulation in certain tumors can be driven by an enhanced permeability and retention (EPR) effect attributed to the presence of a fenestrated vasculature and poor lymphatic drainage^40–42. However, recent evidence suggests that both endothelial cell transcytosis along with significant nanoparticle uptake in perivascular tumor associated macrophages may be critical mediators of tumoral nanoparticle accumulation.^43–47 Liver-, spleen-, and lung-specific accumulation of lipid nanoparticles in mice has been reported; this differential uptake is thought to be promoted by variations in protein corona composition dependent on the surface charge of the particles.^48,49 Nonetheless, tissue selective extrahepatic nanoparticle delivery remains a significant challenge.

Once within the target tissue, protein-nucleic acid complexes are confronted by a dense fibrous network of extracellular matrix (ECM) proteins present in the interstitial space. The mesh-like organization of the ECM imposes a physical restriction on the diffusion of particles to the cell population of interest.^50,51 Abundant proteoglycans in the ECM contain highly anionic glycosaminoglycans (GAG) that can compete with negatively charged nucleic acid cargo for binding to cationic groups present on the protein carrier, and repel anionic nanoparticles by electrostatic repulsion.^52,53 Particle size and surface charge play an important role in dictating the effectiveness of transport with smaller, neutral particles less susceptible to limiting interactions.^53,54 In addition, the ECM is subject to ongoing remodeling by metalloproteinases, elastases, and cathepsins, among other secreted proteases.⁵⁵ Likewise, as mentioned above, protein-based materials are susceptible to degradation in the absence of shielding strategies. Furthermore, fluid homeostasis in the interstitium also plays an important role in nanoparticle transport. In healthy tissue, colloid osmotic and hydrostatic pressure in the microvasculature and interstitium are balanced, which affords a small net negative interstitial pressure that promotes the flow of fluid and molecules into the tissue.⁵⁶ In contrast, the leaky vasculature and poor lymphatic drainage characteristic of tumors, opposes the influx of nanoparticles.^57,58

Within the target tissue, protein-nucleic acid nanoparticles must bind to target cell membranes and stimulate internalization. Cationic protein materials or materials coformulated with a cationic agent are thought to associate with anionic proteoglycans and phospholipids present on the cell membrane in a non-specific manner.^59–62 To promote particle binding to specific cells, active targeting strategies have been implemented that utilize ligands or binding moieties for receptors present on the surface of target cells. Such targeting domains have included full size monoclonal antibodies^63–65, antibody fragments (e.g. Fabs, scFvs, nanobodies)^66–69, synthetic binding proteins (e.g. DARPins, monobodies)^70–72 and other peptide- and non-peptide ligands.^73–76 The recombinant nature of many protein materials facilitates the direct incorporation of protein-based targeting domains, often with greater efficiency and control than chemical conjugation.^77,78 Notably, protein corona formation may shield active targeting domains from recognition, limiting their ability to induce cell-specific delivery.^47,79–82 The use of anti-fouling surface modifications along with plasma pre-equilibrated nanoparticles have been effective strategies for mitigating this effect.^82–84

Following nanoparticle uptake, the endosome presents another barrier to the functional delivery of nucleic acid cargo. During endosomal maturation, the lumen becomes acidified through the action of V-ATPase proton pumps, and fusion occurs with lysosomes containing a variety of degradative enzymes.⁸⁵ Without a mechanism to enable cytosol escape, conventional nucleic acid cargo is rapidly degraded. Nucleic acid binding to toll-like receptors (TLRs) on the endosomal membrane can also result in innate immune activation and limit cytosolic export.⁸⁶ Chemically modified nucleotide structures such as locked nucleic acids have improved the resilience of RNA cargo to enzymatic degradation and have limited innate immune activation.^87,88 Beyond these cargo-centered modifications, many protein-based nanoparticles incorporate membrane-disruptive cationic endosomolytic peptides to enable cargo delivery^10,89; such peptides are thought to facilitate endosomal membrane disruption to permit cargo escape into the cytosol following cellular internalization. Protein-nucleic acid nanoparticles may also incorporate synthetic cationic polymer materials that can promote endosomal escape of cargo through a membrane disruption mechanism yet to be fully elucidated.^90,91

Altogether, it is clear that protein-based nanoparticles must be designed to overcome several obstacles to effectively mediate intracellular nucleic acid delivery. In the following sections, we will highlight a range of engineered and naturally occurring protein materials used to surpass these barriers and enable effective nucleic acid cargo delivery in vitro and in vivo.

3. Polymeric Protein Materials

Analogous to chemically produced synthetic polymers, protein polymers are protein sequences containing multiple repeats of a monomeric unit. In contrast to synthetic polymers, the monomeric unit of a protein polymer is defined by a short amino acid motif, as opposed to a chemical monomer. Polymeric proteins like elastin-like polypeptides, silk, and gelatin do not exhibit defined tertiary structures, self-assemble through interactions between defined secondary structures, and are isolated from unique sources, in bulk from naturally occurring solid structures with chemical and mechanical post-processing or through recombinant expression. In this section, we review the use of polymeric protein materials as nucleic acid delivery vectors.

3.1. Elastin-like Polypeptides (ELPs)

Elastin-like polypeptides (ELPs) are engineered protein polymers that have been explored for a variety of biomedical applications.⁹² Elastin is a fibrous extracellular matrix protein that underlies the extensibility and elasticity of many mammalian tissues, including arteries, heart valves, lung, skin, and ligaments.⁹³ The design of ELPs is inspired by the structure of tropoelastin, a soluble, non-crosslinked elastin precursor composed of alternating hydrophobic and hydrophilic domains.⁹⁴ ELPs are comprised of a short repeating peptide motif of Val-Pro-Gly-X-Gly derived from the hydrophobic domain of tropoelastin, where X represents a guest residue that can be any amino acid except proline. ELPs exhibit lower critical solution temperature (LCST) phase behavior; below a specific transition temperature (T_t), ELPs are soluble in aqueous solution, while above T_t, ELPs reversibly phase separate to form insoluble coacervates.⁹⁵ The transition temperature of ELPs decreases with increased polypeptide length, concentration, and guest residue hydrophobicity^96,97 Additionally, the T_t is modulated by the pH and ionic strength of the aqueous environment.^98,99 ELPs are produced recombinantly and can be purified by affinity purification or by inverse transition cycling (ITC)^100,101, which leverages ELP temperature-dependent solubility to facilitate separation from contaminants.^66,102 Collectively, these properties enable the design of ELP nanoparticles, including 1) multiblock micellar designs containing a phase separated hydrophobic block forming the particle core and a soluble hydrophilic block forming the corona (Fig. 2a), and 2) uniblock designs entirely based on phase separated ELP sequences. The hydrophobic core enables encapsulation and intracellular delivery of apolar small molecule drugs; in addition, the recombinant nature of the material facilitates direct incorporation of peptide drugs for extracellular presentation.^106,107 Furthermore, ELPs can be modified to enable encapsulation and delivery of nucleic acid cargo, as described below (Table 1).

Figure 2. Elastin-like polypeptides (ELPs) for nucleic acid delivery.

a) Temperature-induced ELP self-assembly for micelle formation. T_t = transition temperature for ELP block. Created with BioRender. b) Structure and thermal responsiveness of VGK₈G-ELP₆₀ used for plasmid DNA delivery. From Ref¹⁰³. c) ELP-mediated intratumoral CpG retention following intratumoral administration. d) Quantification of 4T1-luc tumor bioluminescence intensities following daily administration of 0.25 mg/kg luciferase siRNA complexed with Tat-modified ELPs. From Ref¹⁰⁴. e) Serum L-asparaginase (ASP) activity in unmodified ASP, PEGylated ASP, and ELP-modified ASP after intraperitoneal injections on day 0 (1^st), day 14, and day 28 (3^rd). No reduction in ASP activity was observed between the first and third injections in the ASP-ELP₉₀ group. f) Anti-polymer (PEG or ELP) IgG titers after 3 ASP-polymer fusion administrations over 42 days. From Ref¹⁰⁵

Table 1.

Recombinant protein polymer nanoparticles used for nucleic acid delivery.

Material	Cargo	Cell line/Model	In vitro Efficacy	RoA	In vivo Efficacy	Ref
Elastin-like polypeptides (ELPs)
VGK8G-ELP60 Guest residues: V, A, G in 5:2:3 ratio	pEGFP	MCF-7 (human breast cancer)	Limited EGFP expression compared to BPEI control using 1 ⎧g DNA	N/A	N/A	103
p[Asp(DET)]-ELP90 Guest residues: V, A, G in 5:2:3 ratio	pGL4	COS-7 (African green monkey kidney cells)	Effective, less effective than BPEI control	N/A	N/A	116
LAEL-ELP120-Penetratin Guest residues: K, I in 1:4 ratio	pLuc	C6 (Rat glioma cells)	Outperformed luciferase pDNA alone, less effective than PEI	N/A	N/A	113
5TR1/s2.2-ELP72 Guest residues: K 5TR1, s2.2: MUC-1 targeting aptamers	pLuciferase, pPAP-S, pRicin	MCF-7 (xenograft in female BALB/c mice), SKRB (human breast cancer), HepG2 (human liver hepatocellular carcinoma), HFF-1 (human fibroblasts)	Aptamer functionalization improved efficiency. More effective than controls (Turbofect, Lipofectamine) in MUC-1+ cell lines	PT	Repeated administration (MCF-7 xenograft) of ELP-pRicin (d 0, 2, 4, 6, 9, 1 mg/kg) led to complete tumor growth suppression (evaluated d 23)	108,109
Tat-(AP1)4-ELP48 Guest residues: V AP1: IL-4 targeting ligand	siLuciferase, pEGFP	MDA-MB231 (human breast cancer), 4T1-luc (murine breast cancer, xenograft in female BALB/c mice), A549 (human lung adenocarcinoma)	siRNA: luciferase suppression comparable to Lipofectamine control pDNA: Enabled higher EGFP expression than Lipofectamine control in IL-4+ cell lines	IV	siRNA: Enhanced intratumoral siRNA accumulation, suppressed luciferase expression activity with repeated dosing (0.2 mg/kg daily for 6 d)	104,115
ELP12 or 24-K8 Guest residues: A, Y in 4:1 ratio	Yamanaka factor pDNA	HEK293T, Murine fibroblast-derived iPSCs	0.004% fibroblast to iPSC reprogramming efficiency (0.0001–0.001% in other transfection systems)	N/A	N/A	117
ELP60-K12 Guest residues: V	CpG	4T1 (xenograft in female BALB/c mice), murine macrophages	2x increase in NO production in macrophages (signals TLR activation)	IT	Improved intratumoral retention of CpG follow intratumoral injection. 4 mg/kg CpG combined with 131I-ELP brachytherapy reduced tumor size	111
Silk-like peptides (SLPs)
Silk6mer-K30	pEGFP	HEK293FT	Maximum efficiency ~15% (required lipofectamine co-transfection)	N/A	N/A	118
[RGD]11-Silk6mer-K30	pLuciferase	HeLa, HEK293FT	Incorporation of RGD repeats greatly improved efficiency. Less effective than Lipofectamine.	N/A	N/A	119
Silk6mer-K30-[ppTG1]2	pLuciferase	HEK293FT, MDA-MB-435	ppTG1 dimer enabled transfection efficiencies comparable to Lipofectamine.	N/A	N/A	120
Silk6mer-K30-F3	pLuciferase	MBA-MB-435,231, MCF10A	F3-functionalized Silk6mer greatly outperformed unfunctionalized construct. Less efficient than Lipofectamine.	IV	Complexes enabled luciferase expression in MDA-MB-231 tumors in mice	121,122

^aSubscript indicates number of ELP pentapeptide repeats. EGFP, enhanced green fluorescent protein; BPEI, branched polyethyleneimine; GL4, luciferase reporter; iPSC, induced pluripotent stem cell; IT, intratumoral; IV, intravenous; Luc, luciferase; PAP-S, type I Pokeweed antiviral protein; PT, peritumoral; RoA, route of administration.

The first use of ELPs for nucleic acid delivery was reported in 2008.¹⁰³ An ELP consisting of 60 pentapeptide units ([VPGXG]₆₀; where X is Val, Ala, and Gly in a 5:2:3 ratio] was recombinantly modified with an oligolysine-containing cationic block (VGK₈G) to facilitate electrostatic complexation and delivery of enhanced GFP (eGFP) plasmid DNA (pDNA) (Fig. 2b). The construct complexed eGFP pDNA into nanoparticles < 200 nm in diameter for all N/P ratios tested and enabled eGFP expression in vitro. Interestingly, complexation with pDNA greatly reduced the transition temperature of the ELP in an N/P-dependent manner, where lower N/P ratios resulted in lower T_t. This observation suggests that charge neutralization lowers the effective LCST, presumably by reducing the interaction of charged residues with water, consistent with the known effect of guest residue charge on ELP phase transition behavior.⁹⁶ Similar effects have been noted in the studies described below.

A second lysine-containing ELP construct ([VPGKG]₇₂) for targeted delivery of luciferase pDNA to breast cancer epithelial cells has also been reported.¹⁰⁸ The ELP was functionalized with aptamers with high affinity for MUC1, a glycoprotein overexpressed in epithelial tumors, via electrostatic complexation. Incorporation of MUC1-targeted aptamer 5T1 led to a 5-fold improvement in transfection efficiency with delivery of PAP-S toxin pDNA as a suicide gene therapy. The same platform was also investigated for both in vitro and in vivo delivery of ricin pDNA, with repeated peritumoral injection suppressing tumor growth.¹⁰⁹ In addition, oligolysine-containing ELPs have been explored for the delivery of plasmid-encoded Yamanaka factors for pluripotent stem cell induction¹¹⁰ and for intratumoral delivery of CpG DNA to promote antitumor immunity in a murine model of breast cancer (Fig. 2c).¹¹¹ These early studies demonstrated the potential of ELPs to serve as a nucleic acid delivery vector.

To enhance the potency of ELPs for nucleic acid delivery, the incorporation of cell penetrating peptides (CPPs) and endosomolytic peptides have been explored. Piña et al constructed a lysine-containing ELP ({[VPGIG]₂[VPGKG][VPGIG]₂}₂₄) containing a CPP, penetratin,¹¹² and a fusogenic peptide, LAEL, which transitions from a random coil to an α-helix upon exposure to the low pH environment of the endosome, resulting in membrane association and disruption.^113,114 More recently, ELPs modified with the cationic CPP Tat have been used for the delivery of siRNA and pDNA to cancer cells.^104,115 To promote cell selectivity, ELPs were recombinantly functionalized with AP1, a targeting peptide for the IL-4 receptor which is highly expressed on many human tumors. The most effective construct (Tat-[AP1]₄-[VPGVG]₄₈) protected siRNA and DNA cargo from RNAse and DNAse mediated degradation in vitro. siRNA-mediated luciferase knockdown was observed, comparable to that of a lipid transfection reagent. ELPs without Tat were ineffective in both siRNA complexation and delivery. In luciferase+ tumor-bearing mice, Tat-AP1-[VPGVG]₄₈-siRNA nanoparticles enhanced siRNA accumulation and suppressed luciferase expression following daily intravenous administration and enabled delivery of eGFP pDNA to multiple cancer cell lines (Fig 2d).^104,115

Positively charged ELPs can be challenging to express and purify in high yield due to toxicity and high transition temperatures that complicates ITC purification. As an alternative, ELPs with neutrally charged guest residues that can be covalently modified with cationic moieties have been investigated. Chen et al generated a ELP diblock containing a chemically synthesized cationic poly[diethylenetriamine] (p[DET]) block.¹¹⁶ The ELP, consisting of a 90 pentapeptide repeat with Val, Ala, and Gly guest residues in a 5:2:3 ratio, was functionalized with a poly(aspartate) block via ring-opening polymerization of β-benzyl-L-aspartate N-carboxy-anhydride (BLA-NCA), followed by DET conjugation. The p[Asp(DET)]₅₃-modified ELP complexed luciferase pDNA into 90 nm diameter nanoparticles with modest in vitro efficacy. Similarly, ELPs have been created with methionine guest residues ([VPGXG]₄₀, X = Val, Met at a 3:1 ratio) that were thioalkyted to enable introduction of primary or secondary amine groups.¹²³ Incorporation of the amine groups enabled pDNA complexation at low N/P ratios with efficacy observed at an N/P > 0.6. In a subsequent study, methionine-containing ELPs were covalently modified with cationic TMAEMA oligomers synthesized by reversible addition fragmentation chain transfer (RAFT) polymerization.¹²⁴ ELP-g-p(TMAEMA)₁₅ enabled internalization of FITC-labeled pDNA in HEK 293 cells. In the absence of covalent modification, hollow ELP spheres encapsulating cationic polymer-pDNA polyplexes have also been reported, demonstrating improved efficacy and safety in vitro and in vivo as compared to the polyplexes alone.^125,126

Another key feature of ELP-based protein materials is their minimal immunogenicity, likely related to the high level of self-tolerance developed toward the native VPGXG motif in elastin. For example, recent research demonstrates that the fusion of an ELP with L-Asparaginase (ASP), as a treatment for hematologic malignancies, prolongs the half-life of this enzyme and diminishes its immunogenicity with a reduction in anti-ASP IgG and IgM levels by two orders of magnitude (Fig 2e).¹⁰⁵ Significantly, in comparison to an FDA-approved PEGylated ASP, the IgG and IgM titers generated against the ELP protein polymer are an order of magnitude lower than those against PEG (Fig 2f), which is widely employed in the formulation of lipid nanoparticles. ELPs have demonstrated favorable safety and tolerability profiles as recombinant fusions with peptide therapeutics repeatedly administered to patients, substantiating the benign and redosable nature of the material (Table 4).

Table 4.

Assessment of protein-based drug delivery systems in humans.

Drug	Description	ID	Phase	Indication	# pts	Outcome	Ref
Elastin-like polypeptides
PB1046	Recombinant fusion of vasoactive intestinal peptide (VIP) and ELP	NCT03315507	1	Pulmonary arterial hypertension	3	Dosed SC q1w for > 8 weeks, well tolerated. Improved disease in 1 pt	164
		NCT02808585	2	Heart failure w/reduced ejection fraction	29	Dosed SC q1w for 4 weeks up to 1.2 mg/kg, well tolerated. Significant reductions in diastolic BP vs placebo	221
PB0139	Recombinant fusion of native insulin and ELP	NCT01835730	1	T2D	37	Dose dependent decrease in FBG (range of −2 to −34 mg/dL). Well tolerated	222
PB1023	Recombinant fusion of glucagon-like peptide 1 (GLP-1) and ELP	NCT01236404	1/2a	T2D	80	Dosed SC q1w for 4 weeks. Induced significant reductions in FBG in a dose-dependent manner	223
		NCT01658501	2b	T2D	593	Dosed SC q1w for 4 weeks up to 1.35 mg/kg. Significant reductions in FBG at lowest tested dose (0.3 mg/kg)	2241/ 2/2024 5:42:00 PM
Gelatin
Drug-loaded gelatin MPs	Cisplatin-conjugated gelatin microspheres bFGF-loaded gelatin microspheres	N/A	1	HCC	19	Transarterial chemoembolization using microspheres, ~48% nodule response rate	225
	bFGF-loaded gelatin microspheres	N/A	1	Critical limb ischemia	10	IM injection of 200 μg bFGF-incorporated gelatin MPs enabled significant improvement in tissue oxygenation 24 weeks post treatment	226
Albumin
Nab-paclitaxel	Paclitaxel loaded onto human albumin	NCT00046527	3	Metastatic breast cancer	460	IV injections in 3-week treatment cycle. Albumin-loaded paclitaxel induced higher response rates than paclitaxel alone (33% vs 19%)	227
		NCT00844649	3	Pancreatic cancer	861	Gemcitabine + nab-paclitaxel improved survival rates vs gemcitabine alone	228
Silk
VX-103	Antigen-loaded silk fibroin microneedle array	NCT06125717	1	H1N1 influenza	45	Seroprotection rate of 92% at highest tested dose. Well tolerated.	229
Ferritin
H2HA-Ferittin	Ferritin nanoparticle presenting HA domain from H2N2 influenza virus	NCT03186781	1	H2N2 influenza	50	Neutralizing antibody responses induced following IM administration	2301/2/2024 5:42:00 PM

bFGF, bovine fibroblast growth factor; FBG, fasting blood glucose; IM, intramuscular; IV, intravenous; q1w, once weekly; MP, microparticles; SC, subcutaneous.

3.2. Silk and Silk-like peptides (SLPs)

Silk and silk-like peptides are versatile biopolymers that have been widely investigated for drug delivery.^127–131 Silk is produced by several arthropod species, including bees, wasps, silkworms, spiders, among others. The silk of the Bombyx mori (B. mori) domestic silkworm is commonly used in biomedical applications due to its abundance, ease of production, and clinical precedence.^130,132 B. mori cocoon silk is predominantly composed of fibroin, which forms the core of silk fibers and provides robust mechanical properties, and sericin, which generates an outer adhesive coating. Silk fibroin (SF) is isolated from raw silk during processing and used for downstream applications. SF is comprised of a light chain (~28 kDa) and heavy chain (~391 kDa) linked via a disulfide bond. The fibroin heavy chain is arranged as an amphiphilic block copolymer containing alternating hydrophobic and short hydrophilic regions. The hydrophobic regions of SF are composed of Gly-X repeats, where X is Ala, Ser, Tyr, or Val, while the hydrophilic domains are nonrepetitive and generally anionic. Hydrogen bonding and van der Waals interactions between GAGAGS repeats in the hydrophobic block result in the formation of crystalline anti-parallel β-sheet structures that impart high strength, chemical resistance, and rigidity (Fig 3a).^133–135 Self-assembling behavior enables the production of silk particles, in which SF is dissolved and reformed into particles by treatment with kosmotropic salts or organic solvents to induce β-sheet formation.^136–138 Alternatively, to avoid the use of denaturating agents, mechanical methods are used to produce micro- to nano-scale particles.^139,140 β-sheet formation within the hydrophobic regions of SF facilitates packing of nanoparticles and efficient loading of hydrophobic small molecules; positively charged small molecules can also be effectively encapsulated via electrostatic interactions with the anionic hydrophilic domains of SF. Additional modifications, as described below, enable effective nucleic acid complexation and delivery. Moreover, the biocompatibility of silk is supported by its extensive clinical history as a suture material, as well as recent applications in tissue engineering and drug delivery. Studies have indicated that purified Bombyx mori silk fibroin (BMSF) does not elicit allergic reactions, activate the adaptive immune response, or induce fibrous capsule formation, underscoring its low immunogenicity.^143–145

Figure 3. Silk fibroin as a building block for nucleic acid delivery vectors.

a) Hydrophobicity pattern and self-assembly of B. mori silk fibroin (BMSF) heavy chain. Adapted from Ref¹⁴¹. b) Scheme depicting preparation of BMSF/PEI/Doxorubicin (DOX)/survivin siRNA nanoparticles. Adapted from Ref¹⁴².

As an anionic material, B. mori silk fibroin (BMSF) is often functionalized with cationic materials to enable effective nucleic acid complexation. For example, polyethyleneimine (PEI) has been widely leveraged in BMSF-based delivery systems as a cationic co-complexation agent. BMSF nanoparticles coated with 25 kDa branched PEI (bPEI25) effectively complexed and delivered eGFP pDNA in vitro, with lower toxicity than bPEI25 alone.^146,147 Leveraging the presence of a hydrophobic core in BMSF nanoparticles, Norouzi et al utilized bPEI modified nanoparticles to co-complex doxorubicin and survivin siRNA, which in a co-delivery format were more effective in limiting in vivo tumor growth than either monotherapy alone (Fig 3b).¹⁴² In addition to PEI, oligochitosan has also been incorporated into BMSF nanoparticles which enhanced in vitro delivery of luciferase siRNA.¹⁴⁸

BMSF does not inherently promote particle uptake due to its negative charge. In contrast, silk fibroin derived from Antheraea pernyi (A. pernyi) is rich in arginine-glycine-aspartate (RGD) motifs, which supports adhesion to multiple integrin receptors.^149–152 A. pernyi fibroin (ASF)/PEI/pEGFP complexes enabled GFP protein expression in vitro, with transfection efficiency and reduced particle toxicity correlating with the ASF:PEI weight ratio^153,154,155 These particles have also been used for delivery of pDNA encoding pro-angiogenic factors VEGF165 and Ang-1.¹⁵⁵

SF nanoparticles derived from silk obtained from natural sources may contain impurities, display batch-to-batch variability in physicochemical properties, and are not readily amenable to the inclusion of additional functional domains. Alternatively, recombinant silk-like peptides (SLPs) have been used to form nanoparticles for nucleic acid delivery (Table 1). Numata et al created an SLP based on a consensus repeat sequence contained in the Nephila clavipes spider dragline protein MaSp1, which contains a β-sheet-forming poly(alanine) domain. SLPs integrating polylysine domains of various length enabled complexation of pDNA in an N/P ratio-dependent manner. A construct containing a 30-mer polylysine block facilitated optimal GFP pDNA delivery in vitro, though transfection efficiency was low (~15%).¹¹⁸ To enhance transfection, polylysine-containing SLPs have been functionalized with a poly(RGD) domain to improve nanoparticle internalization via receptor-mediated endocytosis (Fig 4a).¹¹⁹ To enable tumor cell targeting, SLP-polylysine constructs have been designed with the tumor homing peptide, F3, that binds nucleolin expressed on tumor associated endothelial cells, and CGKRK, which displays high affinity for heparan sulfates. F3-modified SLPs complexed luciferase pDNA and enabled luciferase expression in orthotopic MDA-MB-231 tumors following repeated intravenous injection.^121,122 A modified SLP-polylysine construct, incorporating the cationic, membrane-disruptive α-helical peptide ppTG1 led to further improvement in luciferase pDNA transfection efficiency, comparable to Lipofectamine (Fig 4b). Notably, SLPs containing a ppTG1 dimer demonstrated 25-fold higher efficiency than those containing a monomeric version.¹²⁰ SLPs modified with Tat have also been used for pDNA delivery in vitro.¹⁵⁶

Figure 4. Recombinant silk-like peptides (SLPs) as a modular platform for nucleic acid delivery.

a) Schematic of procedure for complexation and delivery of pDNA with RGD- and polylysine (K₃₀)-functionalized SLP. From Ref¹¹⁹. b) SLP construct functionalized with polylysine and cell penetrating peptide ppTG1. From Ref¹⁶³.

3.3. Silk-elastin-like proteins (SELPs)

While the materials described above have been the most widely investigated as nucleic acid delivery vectors, several others have been explored due to their compelling physicochemical properties. For example, silk-elastin-like proteins (SELPs) have been utilized to form injectable hydrogels for the controlled release of gene delivery vectors. SELPs are recombinant block copolymers composed of repeating tandem units of silk fibroin-derived (GAGAGS) and ELP-derived (VPGXG) monomers. While the silk units spontaneously self-assemble into tightly packed β-sheets, which act as physical crosslinks, the ELP units display reversible stimulus-responsive assembly. Due to this hybrid composition, SELPs demonstrate an ability to transition from an aqueous solution to a hydrogel (sol-gel transition) at a temperature dictated by molecular weight, silk-elastin ratio, and elastin unit guest residue identities.^131,157,158 Intratumorally administered SELPs designed to gel at a physiological temperature enabled sustained delivery of adenoviral vectors containing a thymidine kinase (Tk-1) transgene, enhancing Tk-1 expression and prolonging antitumor efficacy in mice bearing a head and neck squamous cell carcinoma.^159,160 A similar approach has been used to tailor the release profile and improve the efficacy of adenoviral vectors loaded with a short hairpin RNA (shRNA) targeting the oncogene, C-met.¹⁶¹ To accelerate the vector release profile and hydrogel degradation rate, sequences susceptible to matrix metalloproteinases (MMP) were inserted, with those placed within closely packed silk units displaying slower rates of degradation as compared to those positioned in alternate locations.¹⁶²

Elastin- and silk-based materials represent a promising basis for the development of cell-targeted nucleic acid delivery systems. In vitro and in vivo efficacy has been demonstrated for diverse nucleic acid cargo and clinical studies suggest that native silk and ELPs can be administered repeatedly and display minimal immunogenicity.^143,164 However, to fully realize their potential, novel strategies will be needed to further improve in vivo stability and the potency of these nanoparticles.

3.4. Gelatin

Gelatin, as a hydrolysis product of collagen, is considered GRAS (Generally Regarded as Safe) by the FDA and has a long history of use in food products, cosmetics, and clinical settings. Gelatin is a low-cost, readily available, biocompatible, and biodegradable material. Unlike collagen, which may display antigenicity due to its animal origin, denatured gelatin exhibits reduced immunogenicity. This decrease is linked to the reduction of antigen epitopes during the hydrolysis process. Additionally, gelatin does not produce harmful byproducts during enzymatic degradation, as collagen, being the most abundant protein in animals, undergoes well-tolerated degradation.¹⁶⁵ Other merits of gelatin includes its versatile chemical and biological properties. Gelatin can be either positively charged (type A) or negatively charged (type B) depending on the pH of the hydrolysis conditions in which it is produced.¹⁶⁶ It is also an amphiphilic polymer carrying charged hydrophilic and hydrophobic amino acids^165,167, providing numerous chemical handles for bioconjugation. Furthermore, gelatin carries RGD motifs, which can be recognized by cell surface integrin receptors and promote cell attachment and internalization.¹⁶⁵

Gelatin is characterized by repetitive Gly-X-Y tripeptide motifs, where the X and Y positions are occupied predominantly by proline and hydroxyproline. This specific motif plays a pivotal role in the formation of the triple-helical structure of gelatin, consisting of three polyproline-II type helices, which are stabilized by intermolecular hydrogen bonding and responsible for its sol-gel transition. Numerous methodologies have been developed to manipulate the bulk gelation behavior of gelatin, enabling the creation of nanoparticles. These techniques include desolvation¹⁶⁸, nanoprecipitation¹⁶⁹, and coacervation¹⁷⁰, among others. It should be noted that nanoparticles generated by these approaches typically exceed 200 nanometers. Addressing this limitation, Gwon et al. recently introduced a freezing and thawing step following a two-step desolvation method, resulting in the production of nanoparticles < 100 nanometers.¹⁸³ Gelatin may also be modified in a manner that promotes self-assembly into nanoparticles, including chemical modification with hydrophobic polymers¹⁸⁴ or small molecules.^185,186 Regardless of the manufacturing approach, gelatin nanoparticles typically require immediate crosslinking using agents such as glutaraldehyde and genipin to preserve their morphology and ensure sufficient stability for in vitro and in vivo applications.¹⁶⁵

Both type A and type B gelatin have been used for nucleic acid delivery (Table 2). In contrast to positively charged type A gelatin, which forms electrostatic interactions with anionic nucleic acids, negatively charged type B is used to encapsulate nucleic acids through physical entrapment. Exploiting this approach, Amiji and colleagues have pioneered the development of a multi-compartmental nanoparticle system tailored for oral gene delivery. In this system, a core composed of type B gelatin serves as the carrier for nucleic acid cargo, enveloped within a protective shell constructed from poly-ε-caprolactone or a polymethacrylate-based material. The outer shell shields the core from degradation and facilitates intracellular release of cargo (Fig. 5a). Notably, these vehicles have proven effective for delivery of plasmids encoding IL-10¹⁷¹ and siRNA targeting TNF-α¹⁷² in murine models of dextran sulfate sodium (DSS)- and 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced colitis. Additionally, these nanoparticles have been applied for the delivery of siRNA targeting divalent metal transporter 1 (DMT1) to mitigate pathological iron absorption.¹⁷³

Table 2.

Gelatin-based nanoparticles for nucleic acid delivery.

Material	Cargo	Cell line/Model	In vitro Efficacy	RoA	In vivo Efficacy	Ref
Gelatin
Type-B gelatin nanoparticles embedded in PCL microparticles	IL-10 pDNA	TNBS-induced acute colitis model (Balb/c)	N/A	PO	~10-fold increase in IL-10 protein production in intestinal lining vs empty controls via oral administration, resulting in rescue of acute colitis mouse model	171
	siTNF-α	DSS-induced acute colitis model (Balb/c)		PO	Significant reduction of TNF-〈 expression at d10 post treatment (oral administration), resulting in partial rescue of colitis	172
Gelatin + Eudragit L100–55	siDMT1	CD-1 mice	N/A	PO	Oral administration of siDMT1-loaded NPs resulted in reduced DMT1 expression in duodenum, reduced Fe uptake	173
Type-A gelatin	IGF1, eGFP pDNA	Adult canine chondrocytes	Expression of both cargos observed following transfection	N/A	N/A	174
PEI-gelatin	siASS1	HeLa	Protein expression reduction exceeded that enabled by Lipofectamine with optimized formulation, with less cytotoxicity	N/A	N/A	175
Type A Gelatin functionalized w/chondroitin or dextran sulfate	MUC5AC pDNA	HCE (human corneal epithelial cells), IOBA-NHC (epithelial cells)	Expression of MUC5AC ~20–50% that of positive control (JetPEI-RGD)	IO	Minimal expression of cargo observed in rabbit conjunctiva following ocular administration	176
Chitosan-gelatin-EGCG (CGE)	siTMEM44-AS1	HGC-27, MKN-45 (5-FU resistant); xenograft tumor mouse model	~80% knockdown of tumorigenic lncRNA target in both cell lines	IV	siRNA-loaded CGE nanoparticles + 5-FU prevented tumor growth, more effective than all controls	177
Gelatin-Lipofectamine coformulation	miR-506 mimic	MDA-MB-231, orthotopic TNBC mouse model	Nanoparticle-mediated miR-506 delivery slowed cell migration	IT	Intratumoral administration of NPs resulted in partially slowing of tumor growth (2 injections over 4 week model)	178
	miR-4458 mimic, siCOL11 A1	MCF-7 (human adenocarcinom a), xenograft mouse model	miR delivery slowed cell migration	IT	Single intratumoral administration of both miR-4458 and siCOL11A1-loaded NPs partially slowed tumor growth	179
Thiolated gelatin	Polymeri zed siRFP	RFP+ B16-F10	~80% knockdown of RFP in RFP+ B16-F10,	IV	~80% knockdown of RFP gene expression in B16-F10 xenograft following IV administration on d0, d, 2, d4 (measured on d5)	180
Thiolated PEG-gelatin	p53 pDNA	Panc-1 (human pancreatic adenocarcinom a) xenograft mouse model	N/A	IV	IV administration of p53 pDNA-loaded gelatin NPs resulted in 1000-fold increase in tumoral p53 expression, slowed growth	181
Anti-EGFR mAb functionalized gelatin	siKRAS-G12C	KRAS mutant NSCLC cell line	Target knockdown observed, resulting in ~10–15% reduction in downstream protein expression	N/A	N/A	182

TNBS, 2,4,6-Trinitrobenzene sulfonic acid; DSS, dextran sulfate sodium; DMT1, Divalent metal transporter 1; IGF1, Insulin-like growth factor 1; ASS1, arginosuccinate synthase 1; EGCG, epigallocatechin gallate; lncRNA, long noncoding RNA; TNBC, triple negative breast cancer.

Figure 5. Gelatin as a basis for nucleic acid delivery vectors.

a) Scanning electron microscopy images of nanoparticles-in-microsphere oral system (NiMOS) development by Amiji et al. DNA cargo is encapsulated in gelatin nanoparticles that are subsequently loaded in poly(ε-caprolactone) microparticles for protection during gastric transit and to enable intestinal release. From Ref¹⁷¹ b) Images of HGC-27/R (human gastric cell) subcutaneous xenograft tumors in mice following treatment with 5-FU (chemotherapy) and chitosan-gelatin-epigallocatechin (CGE) nanoparticles formulated with siRNA against a resistance-promoting long noncoding RNA. From Ref¹⁷⁷. c) Assembly and siRNA delivery mechanism of thiolated siRNA and gelatin coformulated into nanoparticles. From Ref¹⁸⁰. d) Volumes of human adenocarcinoma xenograft tumors in mice following treatment with gemcitabine and EGFR-targeted, PEGylated gelatin nanoparticles loaded with pDNA-encoded tumor suppressor p53. From Ref¹⁸¹.

Similar to other protein polymer systems, gelatin can harness the advantages of an augmented positive charge through functionalization with cationic polymers, such as ethylenediamine¹⁷⁴, PEI¹⁷⁵, cholamine¹⁸⁷, and spermine¹⁸⁸. Zorzi et al. created a gelatin nanoparticle system comprised of positively charged type A gelatin that was further cationized with spermine, which exhibited efficient plasmid incorporation and transfection of ocular epithelial cells. Topical delivery of plasmid-encoded MUC5AC has also been demonstrated with potential for treatment of dry eye disease.¹⁷⁶ In addition to N-hydroxysuccinimide (NHS) mediated coupling to produce branched cationic polymers, nanoparticles have been created by combining gelatin, chitosan, and epigallocatechin gallate in the presence of sodium tripolyphosphate. The resulting cationic particles were used for systemic delivery of siRNA targeting long non-coding RNA (lncRNA) TMEM44-AS1 with a pronounced tumor-suppressive effect (Fig. 5b).¹⁷⁷ Liu et al. have recently reported complexing pDNA, siRNA, and miRNA with lipofectamine followed by loading within gelatin nanoparticles. Intratumoral injection led to a therapeutic response in mice bearing triple-negative and estrogen receptor-positive breast cancers.^178,179

In addition to addressing charge and delivery efficiency, considerable attention has been directed toward enhancing particle stability, extending circulating half-life, and refining targeting specificity in nanoparticle systems. Traditional crosslinking methods, such as those involving glutaraldehyde, have raised concerns regarding toxicity and the elicitation of immune responses. Consequently, crosslinking strategies based on disulfide bond formation have emerged that lead to particle disassembly within the reducing environment of the intracellular space. Lee et al. observed a reduction in tumor expression of red fluorescent protein after intravenous administration of thiolated type B gelatin nanoparticles containing siRNA (Fig. 5c).¹⁸⁰ Another well-established approach for further augmenting nanoparticle half-life and improving biodistribution is through PEGylation. Kaul and Amiji demonstrated that PEGylated gelatin particles exhibited higher levels of beta-galactosidase expression within tumors¹⁸⁹.

While native RGD motifs and positive surface charge confers gelatin nanoparticles with a propensity for non-specific cell uptake, enhanced specificity can be achieved through surface modification with distinct targeting ligands. An array of targets and targeting moieties have been investigated. For example, Xu et al. have designed a comprehensive system in which thiolated and PEGylated gelatin nanoparticles are decorated with a peptide targeting the epidermal growth factor receptor (EGFR) for tumor treatment. Intravenous administration of particles loaded with plasmids encoding p53 effectively suppressed the growth of human pancreatic adenocarcinoma in a SCID mouse model (Fig. 5d).¹⁸¹ Moreover, EGFR antibodies have been affixed to gelatin nanoparticles to facilitate the co-delivery of siRNA targeting KRAS and a tyrosine kinase inhibitor, resulting in the inhibition of H23 non-small cell lung cancer cell growth in vitro.¹⁸² In another innovative approach, Zhao et al. engineered gelatin/silica nanoparticles derivatized with PEG, a fusogenic peptide (HA2), and a tumor-targeting aptamer (ARGO100) with a strong affinity for nucleolin expressed in cancer cells. These modified particles exhibited tumor accumulation following intravenous injection in A549 tumor-bearing mice in vivo.¹⁹⁰ Lastly, Tian et al. have demonstrated the therapeutic potential of Tat-decorated gelatin-siloxane nanoparticles carrying a gene encoding the vasodilator calcitonin gene-related peptide (CGRP). This nanoparticle system exhibited notable therapeutic effects in a rat model of subarachnoid hemorrhage-induced cerebral vasospasm upon intracisternal injection.¹⁹¹

Interestingly, in a comparative study evaluating cationized atelocollagen, albumin, and gelatin for ocular gene delivery, gelatin cationized with spermine emerged as the most promising candidate. This choice struck an optimal balance between reduced toxicity, efficient encapsulation of pDNA and siRNA, and favorable functional outcomes in vitro.¹⁹² Similarly, casein, bovine serum albumin, and gelatin have each been evaluated for non-targeted and antibody-targeted siRNA delivery. As anticipated, antibody conjugated gelatin demonstrated the highest efficiency for intracellular siRNA release.¹⁹³

In summary, gelatin nanoparticles have demonstrated commendable biocompatibility, stability, and efficiency for the delivery of therapeutic nucleic acids. Established methodologies have been devised for their production, crosslinking, PEGylation, and conjugation to targeting moieties. Notably, this system has exhibited promising therapeutic effects when delivering plasmid DNA and short RNAs in various in vivo animal models through systemic, local, and topical administration. Nonetheless, the limitations of gelatin nanoparticles include an inherent difficulty in controlling the molecular weight and composition of this biopolymer. Moreover, the introduction of various chemical modifications complicates the manufacturing process, introducing additional costs and variability through reaction and purification cycles.

4. Globular proteins

In contrast to polymeric protein materials, globular proteins do not contain repetitive sequences, exhibit defined tertiary structures, self-assemble through hydrophobic interactions enabled by their tertiary structures, and must be expressed recombinantly or isolated from human fluids (i.e. plasma). In this section, we review the use of globular proteins as nucleic acid delivery vectors.

4.1. Albumin

Albumin nanoparticles have been utilized as carriers for a wide variety of therapeutic agents, including nucleic acids. As the most abundant protein in human plasma, albumin is a small (MW ~67 kDa), water-soluble, anionic (pI 4.7) protein that is responsible for maintaining intravascular osmotic pressure and acts as a carrier for several classes of endogenous hormones, vitamins, and fatty acids, and exogenous small molecule drugs.¹⁹⁴ Due to its lack of immunogenicity, human serum albumin (HSA) is commonly used in clinical therapeutics, though bovine serum albumin (BSA) shares significant (~80%) sequence homology with HSA and typically employed for preclinical formulation studies due to low cost and ease of sourcing.¹⁹⁵ Albumin has been clinically approved as an excipient for multiple hydrophobic drugs, most notably Abraxane, which consists of paclitaxel co-formulated with albumin nanoparticles to improve tumor specificity and limit toxicity.²⁰⁹ Albumin nanocarriers traffic to tumors due to high affinity interactions with gp60 and SPARC (secreted protein acidic rich in cysteine) that are expressed on tumor endothelial cells.^195,196 This tropism, along with an extended half-life (14–16 hours), biodegradability, and amenability to conjugation have motivated the investigation of albumin as a nucleic acid delivery vector (Table 3).

Table 3.

Globular protein-based delivery systems for nucleic acids.

Material	Cargo	Cell line/Model	In vitro Efficacy	RoA	In vivo Efficacy	Ref
Albumin
PDMA-BSA	eGFP, luciferase pDNA	HEK293T, NIH/3T3, D1 (murine bone marrow cells)	Efficacy comparable to PEI at optimized BSA:DNA mass ratios	N/A	N/A	196
Ethylenediamine-functionalized BSA	siBcl2	B16 (IV administered B16 cells for lung metastasis model in C57BL/6)	~50% target knockdown in optimized conditions, comparable to Lipofectamine	IV	Particle accumulation in mouse lungs related to particle size (D ~5 μM). cBSA+siBcl2 particles enabled significant reduction in # of lung metastases vs controls	197
PEI-HSA	eGFP, luciferase pDNA	A549	PEI-HSA nanoparticles performed comparably to Lipofectamine with less cytotoxicity	N/A	N/A	198
PEI-HSA	Luciferase pDNA	Mouse cerebellar granule cells	~80% of treated wells showed luminescence following transfection	N/A	N/A	199
bPEI-HSA	siTurboGFP	HPMEC, MDA-MB-231	~90% knockdown in both cell types, comparable to Lipofectamine	N/A	N/A	200
PEI-BSA	Cas9 pDNA	MDA-MB-231	Nanoparticles enabled efficient uptake of pDNA; editing efficiency not evaluated	N/A	N/A	201
BSA	siKRAS (G12S)	A549	Significant target knockdown resulting in reduced cell viability	N/A	N/A	202
MSA (in situ)	Maleimide-anti-miRNA-21 ASO	U87	Significant changes in miRNA-21 target gene expression	IV	In situ binding to endogenous circulating albumin resulted in prolonged circulating half life, increased tumor accumulation, and ~50% slowing in tumor growth	203
Heat shock proteins
HSP-Tat-RGD RGD: integrin targeting	siTERT	CT26 (murine colorectal carcinoma cells/xenograft in BALB/c nude mice)	RGD addition significantly reduced TERT expression; apoptosis rate ~1.5-fold higher than that of the Lipofectamine/si TERT control.	IV	Intravenous injection of HSP-Tat-RGD/siTERT (d 0, 2, 4, 6, 8) resulted in markedly reduced tumor volume (evaluated d 20)	204
Ferritin
Human heavy chain apoferritin	siInsR	Caco-2, Human PBMCs	Enabled ~85% knockdown of insulin receptor in Caco-2 cells, ~50% knockdown in activated human PBMCs	N/A	N/A	205
Pas-HumAfFT (humanized archaea ferritin); HumAfFT point mutation (M54C); PA (cationic polyamine)	siGAPDH	HeLa, HepG2 (human hepatoma), MCF-7(human breast cancer)	> 20% GAPDH silencing observed in HeLa, HepG, and MCF-7 cells.	N/A	N/A	206
Humanized A. flugidus ferritin	miRNA-145–5p	NB4 (human acute promyelocytic leukemia)	Granulocytic differentiation observed in cell morphology analysis, although not sufficient to achieve full differentiation	N/A	N/A	207
Poly(L-lysine) modified heavy chain apoferritin (4L-HFn)	siEGFR	HeLa, 4T1 (murine breast cancer cells/tumor model in female BALB/c mice)	Successfully downregulation of EGFR protein (>70%), but not as effective as Lipofectamine.	IV	Significant accumulation of 4L-HFn@siRNA in tumors from 2 to 10 h; tumor growth suppression effect following treatment every 2 d for 10 d	208

ASO, antisense oligonucleotide; bPEI, branched poly(ethylenimine); BSA, bovine serum albumin; HPMEC, human pulmonary microvascular endothelial cells; HSP, heat shock protein; HSA, human serum albumin; lnsR, insulin receptor; MSA, Mouse serum albumin; PAMAM, polyamidoamine; PDMA, poly(N,N-dimethylacrylamide); PEI, poly(ethylenimine).

As with other non-cationic proteins, albumin has been functionalized with cationic polymers to facilitate electrostatic complexation with nucleic acids. Residues on the surface of albumin provide several chemical handles for the direct conjugation of polymers. Cationic bovine serum albumin (BSA)-based nanoparticles for plasmid DNA delivery have been synthesized through in situ polymerization of (dimethylamino)ethyl methacrylate (PDMA) on bromide-derivatized free amines (Figure 6a).¹⁹⁶ Han et al and Chen et al pursued a similar approach, in which carboxylic acid groups were modified with ethylenediamine and branched PEI, respectively.^197,198 Ethylenediamine-BSA particles were used to deliver siRNA targeting Bcl-2, reducing lung metastasis in a B16 melanoma model.¹⁹⁷

Figure 6. Albumin-based nanoparticles as a non-immunogenic material for nucleic acid delivery vectors.

a) Synthesis of 2-(dimethylamino)ethyl methacrylate (DMA)-modified cationic BSA via in situ polymerization on bromide-functionalized free amines in BSA. Figure from Ref²¹⁴. b) Schematic depicting co-formulation of BSA and PEI with Cas9 RNP. Figure from Ref²⁰¹. c) Synthesis of albumin nanoparticles via desolvation and glutaraldehyde crosslinking. Nucleic acid (NA) cargo is combined in aqueous albumin solution for albumin-NA nanoparticle formation. Figure from Ref²¹⁰. d) (Left) Plasma concentration and (right) organ distribution of an anti-miRNA-21 oligonucleotide following intravenous administration in tumor bearing mice. AM-Anti-21: unmodified oligonucleotide, MI-Anti-21: maleimide-functionalized oligonucleotide to facilitate in situ conjugation to endogenous circulating albumin via reaction with Cys34, MSA-Anti-21: maleimide-functionalized oligonucleotide pre-incubated with mouse serum albumin prior to administration.

Alternatively, albumin has been used in combination with cationic polymers for co-complexation and delivery of nucleic acids. Langiu et al combined HSA with linear PEI to encapsulate luciferase pDNA, enabling transfection of HeLa and mouse cerebellar granule cells.¹⁹⁹ A similar formulation combining branched PEI with HSA demonstrated improved siRNA uptake and protein knockdown compared to bPEI alone.²⁰⁰ More recently, BSA-PEI nanoparticles effectively facilitated cellular uptake of plasmid and ribonucleoprotein (RNP) Cas9 (Figure 6b).²⁰¹

Albumin nanoparticles have also been employed for nucleic acid delivery through co-desolvation and direct conjugation. Mehta et al prepared BSA-nucleic acid nanoparticles by desolvation of aqueous BSA and nucleic acids, followed by glutaraldehyde crosslinking.²¹⁰ An optimized preparation method yielded an 85% encapsulation efficiency of pDNA with delivery of both GFP pDNA and KRAS-targeted siRNA to A549 lung cancer cells in vitro (Figure 6c). While direct conjugation of cargo to albumin results in markedly lower carrying capacity, Cys34 on the surface of HSA has been used as a chemical handle for conjugation to heavily-modified oligonucleotides, leveraging a monobromomaleimide linker to improve stability in serum.²¹¹ Cys34 can also be leveraged as an in situ conjugation handle. Kwak et al designed a maleimide-functionalized antisense oligonucleotide targeting (ASO) miRNA-21, a microRNA that suppresses tumor progression.²⁰³ When this ASO was administered intravenously in tumor bearing mice, covalent bonding to albumin facilitated a marked improvement in ASO circulating half-life and tumor growth suppression in vivo (Fig. 6d)

4.2. Protein Nanocages

Protein nanocages (NC) have recently gained interest as a delivery vehicle due to their unique material properties, as their cage-like structure enables monodisperse size, robust stability, innate biocompatibility and biodegradability, and ready functionalization by genetic or chemical means.^204,212,213 NCs are defined by the self-assembly of a defined number of protein subunits into a hollow core-shell structure. Importantly, this precise organization facilitates both surface and internal presentation of distinct functional elements, such as receptor targeting ligands, antigenic proteins, and cargo binding motifs in specific amounts and geometries. Additionally, advances in recombinant protein production render NCs amenable to production and purification at scale. Altogether, these advantages have encouraged efforts to harness protein nanocages for nucleic acid delivery (Table 3). While existing efforts to engineer NCs as drug nanocarriers have included both viral and nonviral proteins, viral protein nanocages have been reviewed elsewhere.²¹⁵ This review will focus on nonviral systems that avoid some of the key challenges of virus-derived carriers.

Among nonviral protein nanocages, human ferritin or apoferritin (demineralized ferritin) has been exploited for the delivery of nucleic acids, particularly siRNAs (Fig. 7a) Ferritin is a 24-mer heteropolymer that plays a pivotal role in the storage, transport, and detoxification of iron. In mammalian organisms, each of these subunits is encoded by a heavy (H) or light (L) chain, which integrate during a reversible self-assembly process to create a hollow, spherical structure with an inner diameter of ~8 nm and an outer diameter of ~12 nm.^216,217 The resulting inner cavity provides ferritin with a unique storage ability. Due to an ability to be endocytosed through H-ferritin and TfR1 (transferrin 1 receptor) interactions, combined with tunable recombinant expression and intrinsic biocompatibility, ferritin nanocages represent an interesting opportunity for intracellular drug delivery. In addition, as a recombinantly expressed human protein, ferritin is expected to have minimal immunogenicity. Li et al demonstrated that unmodified human apoferritin can mediate delivery of a variety of siRNAs resulting in gene silencing in both tumor cells and human peripheral mononuclear blood cells (PBMCs) in vitro.²⁰⁵ However, the ferritin cavity naturally contains many negatively charged residues, inhibiting efficient encapsulation of siRNA molecules.²⁰⁶ To overcome this limitation, Pediconi et al functionalized human ferritin with positively charged cationic piperazine-based compounds (Pas) to physically entrap siRNA duplexes.²⁰⁶ PA-functionalized ferritin compounds were able to deliver GAPDH-targeted siRNA to HeLa, HepG2, and MCF-7 cancer cells with superior silencing as compared to traditional transfection methodologies (Fig 7b).²⁰⁶ Similarly, Huang et al synthesized a poly(L-lysine)-modified apoferritin nanocage that exhibited gene silencing in vitro and inhibited 4T1 tumor progression in vivo.²⁰⁸

Figure 7. Nonviral Protein Nanocages as an siRNA delivery platform.

a) Wild-type human ferritin (PDB ID: 3AJO) with 24 monomers displayed in different colors (left), charge distribution of each subunit (middle), and the predicted structure of each subunit (right). From Ref²¹⁸. b) Schematic of recombinant human ferritin (HFn) nanocages functionalized with cationic domains such as piperazine-based compounds (PAs) or poly(L-lysine) (4L) for improved siRNA encapsulation. Adapted from Ref²⁰⁶. c) Schematic of heat shock protein (HSP)-derived protein nanocages decorated with transactivator of transcription (Tat) (green) and arginine-glycine-aspartate (RGD) motifs (yellow), abbreviated HSP-Tat-RGD. d) In vivo antitumor efficacy of HSP-Tat-RGD-siTERT complexes delivered in CT26 xenograft mice, as demonstrated by reduced tumor volumes compared to saline and free siTERT2 controls. c) and d) adapted from Ref²⁰².

Additional efforts to engineer nonviral protein nanocages for nucleic acid delivery have focused on heat shock proteins (HSPs). HSPs are a large family of protein chaperones involved in the first line of defense against a variety of cellular stressors, including heat shock, ionizing or ultraviolet radiation, and reactive oxygen species.²¹⁹ These ubiquitously expressed proteins protect cell integrity and survival by various mechanisms, including genomic repair and the prevention of protein misfolding and aggregation.²¹⁹ Much like ferritin, Hsp16.5, derived from Methanococcus jannaschii, is composed of 24 subunits, which self-assemble into a spherical nanocage. Guan et al engineered Hsp16.5 by introducing a cationic polyarginine motif at the C-terminus to enable condensation of siRNA within the nanocage.²¹³ This delivery system protected condensed siRNA from degradation and enabled delivery to HeLa-EGFP cells with downregulation of GFP expression.²¹³ Furthermore, HSP NCs decorated with Arg-Gly-Asp (RGD) and Tat peptide motifs were used to deliver telomerase reverse transcriptase (TERT)-targeting siRNA, inducing tumor cell apoptosis in vitro, and inhibiting the growth of tumor CT26 xenografts in vivo (Fig 7c–d).²⁰⁴ Together, these data demonstrate potential for Hsp-derived protein nanocages to serve as an efficient and biocompatible siRNA delivery system. Nonetheless, the carrying capacity of these natural protein scaffolds remains near the effective diameter of siRNAs (8–9 nm), precluding the application of nonviral nanocages to larger, protein-encoding nucleic acid cargo. Moreover, as bacterial-derived proteins, the immunogenicity of these nanoparticles requires further investigation.

In summary, diverse protein nanoparticles exhibit varying levels of in vitro and in vivo efficacy in delivering nucleic acid cargo. In numerous instances, their delivery efficiency rivals or even surpasses that of traditional lipid- or cationic polymer-based transfection agents (Tables 1 and 2). It is important to highlight that, in certain scenarios, these effects result from the modification of the protein material with cationic polymers, raising questions about the distinctive contribution of the protein component itself. The primary contribution of the protein component oftentimes lies in its ability to mitigate the cytotoxicity associated with cationic polymers while simultaneously enhancing delivery efficiency. This phenomenon is consistently observed in systems such as cationic ELPs^103,220, silk^{139,140,151–153}, gelatin¹⁹¹, and albumin¹⁹⁹. This effect is believed to stem from a combination of reduction of surface positive charge, improvement of nanoparticle stability, and the enhancement of tissue targeting and cellular uptake, which is mediated via diverse interactions between the protein motifs and their native receptors.

5. Clinical Development

Preclinical data from the studies detailed above is part of a growing body of evidence supporting the clinical application of non-viral protein materials for nucleic acid delivery. At the time of publication of this review, none of the materials described have been used for nucleic acid delivery in humans. However, a proof-of-concept has been established for the use of many of these materials as safe and effective drug delivery vehicles for a range of non-nucleic acid cargo (Table 4). Leveraging the recombinant and thermoresponsive nature of elastin-like polypeptides, fusions of ELPs and therapeutic peptides have been created that form sustained release depots following subcutaneous administration, due to the coacervation of the ELPs. Slow egress of the ELP-peptide fusions from the depots into circulation leads to target engagement and therapeutic effects; the presence of the long ELP chain on the therapeutic peptide improves its circulating half-life and further prolongs its effects. ELP fusions to vasoactive intestinal peptide (VIP), insulin, and glucagon-like peptide 1 (GLP-1) were well tolerated and demonstrated sustained therapeutic benefit in pulmonary arterial hypertension and type 2 diabetes following weekly subcutaneous administration to patients in Phase 1/2 clinical studies (see Table 4).^{164,221–224} Repeated dose ELP fusions to hundreds of patients without adverse effects related to an anti-ELP immune response provides evidence of the minimal immunogenicity of this material, an important property for its use for nucleic acid delivery in humans.

Both gelatin and albumin have also been effectively used for therapeutic delivery in humans. Gelatin microspheres have been effectively used as a sustained release system for both proteins (bovine fibroblast growth hormone, in critical limb ischemia patients)²²⁶ and small molecules (cisplatin, in patients receiving transarterial chemoembolization of hepatocellular carcinoma nodules).²²⁵ Likely the most prominent example of a protein-based delivery vehicle used in humans is Abraxane (generic name nab-paclitaxel), an approved therapy consisting of hydrophobic chemotherapeutic paclitaxel co-formulated with albumin nanoparticles intended to 1) avoid the toxicities associated with alternative solvents such as polyethylated castor oil and 2) increase tumor exposure to the drug. In a Phase 3 trial directly comparing unformulated paclitaxel and nab-paclitaxel in metastatic breast cancer patients, nab-paclitaxel induced higher overall response rates (33% vs 19%), and lower rates of severe neutropenia (9% vs 22%).²²⁷ In a separate trial, nab-paclitaxel combined with gemcitabine enabled improved survival rates in late-stage pancreatic cancer patients compared to gemcitabine alone.²²⁸

Other protein materials have been utilized as part of vaccination delivery system in humans. Due to its mechanical strength, biodegradability, and biocompatibility, B. mori silk fibroin (BMSF)-derived thread has been used as the material basis for surgical repair mesh and sutures.^231,232 More recently, BMSF has been used to create room temperature-stable, protein antigen-loaded microneedle arrays for minimally invasive vaccination. Sustained antigen release from the microneedles following application to the skin better recapitulates the natural course of exposure to an infectious agent; in a Phase 1 trial, 92% of patients who were vaccinated against H1N1 influenza through this approach produced a protective level of antibodies.²²⁹ Leveraging the self-assembly, multivalent structure, and high manufacturability of ferritin nanoparticles compared to traditional attenuated virus vaccines, Houser et al recombinantly produced a H2 hemagglutinin-ferritin nanoparticle vaccine (H2HA-Ferittin) that displayed eight copies of H2N2 influenza HA and produced neutralizing antibody responses in vaccinated patients.²³⁰ Collectively, these data provide some evidence for the clinical safety of the reviewed protein-based materials, setting the stage for the investigation of these materials as nucleic acid delivery vectors in humans.

6. Outlook and Unresolved Challenges

Protein-based nanoparticles have shown great promise as delivery vectors for multiple types of nucleic acid cargo. To realize the full potential of these materials, several unresolved challenges will need to be addressed (Fig. 8). Many of these challenges are ultimately related to the challenge of improving potency – the concentration or dose of nucleic acid cargo required to generate a desired therapeutic effect. As the leading non-viral nucleic acid delivery system, lipid nanoparticles (LNPs) have displayed greater potency when compared to protein-based systems. LNP-siRNA formulations incorporating the highly active ionizable lipid DLin-MC3-DMA achieved transthyretin (TTR) mRNA knockdown over 24 hours at a median effective dose (ED₅₀) of 0.03 mg siRNA/kg in mice. In clinical studies, patisiran, a TTR-targeted siRNA drug formulated in a DLin-MC3-DMA based LNP, has been effectively at 0.3 mg/kg administered every 3 weeks.^233,234 Other LNPs have demonstrated pDNA transfection efficiencies comparable or exceeding Lipofectamine in vitro.^235,236 In contrast, protein nanoparticles evaluated as delivery platforms for siRNA and pDNA, particularly those not coformulated with synthetic polymers such as PEI, have largely failed to outperform Lipofectamine in vitro and require frequent multiple mg/kg dosing in vivo.^{104,109,119,122} While the advantages of protein nanoparticles remain evident, including reduced cytotoxicity, improved biodegradability, genetic programmability, and highly tunable physicochemical properties, those features do not outweigh the importance of achieving therapeutic delivery at scalable nucleic acid doses.

Figure 8. Emerging solutions to address key challenges faced by protein-based nucleic acid delivery vehicles.

If leveraged appropriately, innovations in the areas of endosomal escape, biostability, targeted delivery, and safety and manufacturing can come together as the missing pieces for constructing an ideal protein nanoparticle. Created with BioRender.

6.1. Endosomal escape

Entrapment in the endolysosomal system following cellular internalization is a key limitation that must be surmounted to enhance the potency of protein-based nucleic acid delivery systems. One approach to address this is through the use of cationic amphiphilic drugs (CADs), a class of molecules that accumulates in acidic lysosomal compartments, promotes phospholipidosis, and leads to lysosomal disruption.^237,238 Chloroquine, a commonly used antimalarial medication, is a CAD that has been explored as an endosomolytic agent.²³⁹ Other CADs, such as desloratadine and loperamide, have demonstrated a remarkable ability to potentiate siRNA delivery by several types of nanocarriers.²³⁸ While CADs can be co-formulated into nanoparticles with nucleic acid cargo, more often these drugs have been administered separately, creating additional dosing complexities. Critically, CADs present a risk of toxicity, with chloroquine displaying immunosuppressive effects. Other agents with potentially better toxicity profiles are being investigated as alternatives.^240,241

In contrast to small molecule endosomolytic agents, direct incorporation of membrane-disruptive endosomolytic peptides can improve nucleic acid cargo delivery without the need for a co-administered agent. While the precise mechanism of action is debated, interactions between positively-charged residues and negatively-charged phospholipids, coupled with membrane intercalation by hydrophobic peptide segments, appear to be critical aspects of peptide-mediated endosomolysis. It has been proposed that cationic endosomolytic peptides induce cargo-loaded vesicle budding from the parent endosome and eventual collapse.^89,242,243 An alternative mechanism suggests that endosomolytic peptides may promote the fusion of intraluminal vesicles with the parent endosomal membrane, resulting in cargo release without requiring direct movement across the membrane.^244–246 Recent studies have established that the amphiphilicity of endosomolytic peptides, a property of either the primary peptide sequence or emerging from its secondary structure, correlates with membrane disruptive potential.^247,248 Tat, an arginine-rich, HIV-derived peptide²⁴⁹, penetratin and ppTG1, both cationic, amphiphilic α-helical peptides that are examples of secondary amphiphilic peptides,^112,250 have been used to enable nucleic acid cargo complexation and mediate endosomal escape in protein nanoparticle systems.^{104,113,115,120,156} While significant progress has been made in improving the effectiveness of such peptides for nucleic acid delivery²⁵¹, further exploration of membrane-disruptive peptides may well yield protein-nucleic acid nanoparticle designs with enhanced potency.

Machine learning has emerged as a powerful tool for accelerating the discovery of membrane active peptide sequences.^252–254 In contrast to in vitro screening based on rational design, in silico screening enables the exploration of much larger spans of physicochemical space encompassing potentially active peptides in a far less resource-intensive manner. Many machine learning models designed to support cell penetrating and endosomolytic peptide discovery have been trained using the CPPsite database, which contains structural information for over 1800 CPPs and associated cell uptake efficiency.^{252,255–257} As opposed to using this non-standardized database for training, some groups have generated datasets derived from peptide library screens labeled with nucleic cargo delivery efficacy as a basis for predictive model generation.^258–260 For example, data from an in vitro delivery screen of 600 CPP-phosphorodiamidate morpholino oligomer (PMO) conjugates, which enable EGFP expression in treated cells was applied to train a convolutional neural network for the prediction of peptide efficacy. This model was leveraged for the identification of de novo sequences with the discovery of peptides displaying two fold greater activity than the initial lead peptide.²⁵⁹ These studies highlight the potential of in silico design of potent endosomolytic peptides, which in turn can be incorporated into protein-nucleic acid nanoparticle systems to achieve enhanced potency.

Beyond endosomolytic peptides, membrane disruptive proteins are emerging as another means to improve endosomal escape of nucleic acid cargo. Goldbach et al created membrane disruptive, pH-responsive helical bundle proteins designed to lyse membranes via direct intercalation within a narrow range of pH (6.0–5.5).²⁶¹ Naturally occurring proteins have also been investigated, including Perfringolysin O (PFO), Listeriolysin O (LFO), and Phospholipase A₂ (PLA₂), which are pathogenic proteins that enable pore formation in lipid membranes (PFO and LFO) or hydrolyze phospholipids found in membranes (PLA₂). These proteins have been used as fusions with receptor targeting and RNA-binding protein or synthetic polymer domains to enable delivery of siRNA and DNA cargo to cell lines in vitro.^{251,262–264} Given that these modules are proteins, they are amenable to direct incorporation into the recombinant protein nanoparticle systems described above, including elastin-like polypeptides, silk-like polypeptides, and protein nanocages and may be worthy of further exploration. However, there are two important limitations that must be addressed for membrane disruptive proteins to be used as endosomolytic agents in humans: 1) immunogenicity, as the variants of these proteins used are typically derived from pathogens or venomous animals, and 2) membrane selectivity, as undesired activity on the cell membrane can result in significant toxicity. The use of human pore-forming proteins and phospholipase homologues may represent less immunogenic alternatives worthy of further exploration.^265,266 Emerging in silico protein design tools may be useful for reengineering these proteins to improve endosomal membrane specificity, either via reducing activity at neutral pH or tuning activity towards anionic lipids specifically enriched in endosomal membrane.^267,268

6.2. Improving biostability

Sustained colloidal stability prior to use, together with resistance to protein aggregation following drug administration ensure that particles can be stored for an extended period and achieve effective biodistribution to target tissues. Studies that assess nanoparticle stability have generally been lacking and many reports have been limited to investigations under reduced serum transfection conditions suggestive of limitations in particle stability or uptake in physiologic conditions. Anti-fouling surface modifications have been explored as a means to prevent nanoparticle aggregation, improve particle biostability, enhance circulating half-life, and thereby improve the potency of in vivo cargo delivery.

Coating nanoparticles with hydrophilic polymer chains, such as PEG, promotes surface hydration, creates a steric barrier that discourages nanoparticle aggregation, and establishes an entropic penalty for protein adsorption to the nanoparticle surface.^22,27,28 Silk fibroin nanoparticles bearing PEG5000 displayed improved stability, limited particle aggregation, and maintained particle size over 24 hours under physiologic conditions.²⁶⁹ Similarly, gelatin nanoparticles modified with a PEG-poly(2-dimethyl amino ethyl methacrylate-co-itaconic acid) copolymer (PEG-P(DMAEA-co-IAc)) have been used to co-encapsulate and deliver doxorobucin and betanin to breast cancer cells.²⁷⁰ Other synthetic and natural-occurring anti-fouling polymers explored have included poly(glycerol)^271,272, poly(acrylamide)^273,274, and hyaluronic acid.²⁷⁵

While feasible, polymer conjugation to recombinant proteins is challenging to implement and scale, and the widespread use of PEG has resulted in the emergence of anti-PEG immunity and reports of an associated accelerated blood clearance (ABC) phenomenon.^28,276 Zwitterionic polypeptides have emerged as an alternative strategy for incorporating anti-fouling domains into recombinant protein nanoparticles with formation of a well-hydrated, compact surface that prevents protein adsorption.^277,278 Poly(glutamate-lysine) (poly(EK)) peptides have demonstrated robust anti-fouling properties in multiple nanoparticle systems, resulting in improved circulation times and biodistribution.^31,279–282 Importantly, zwitterionic polypeptides can be easily incorporated into recombinant protein nanoparticle systems without the need for chemical conjugation, and with precise control of size and valency. Thorough pharmacokinetic studies are needed to establish the benefit of zwitterionic and other anti-fouling modifications on nanoparticle half-life and biodistribution. Improvements in circulating half-life mediated by zwitterion functionalization can ultimately increase target cell transfection efficiency in vivo, while reducing non-specific uptake.

6.3. Extrahepatic and extratumoral delivery

The majority of approved and late-stage nanoparticle drug delivery systems are intended to promote hepatic and tumoral drug accumulation following intravenous administration.^283,284 A majority of protein nanoparticle nucleic acid delivery vectors have also been directed towards liver and oncologic applications. As noted, liver and tumors contain a fenestrated vasculature, a shared characteristic that favors nanoparticle drug delivery.^35,40 However, many indications require delivery to tissues and cells, such as myocytes and neural cells for the treatment of inherited muscular dystrophies and neurodegenerative diseases, respectively, that harbor significant barriers to particle transport.^285,286 Both trafficking to the tissue of interest and internalization by target cells remain critical factors that dictate cell-specific cargo delivery. In turn, enhancing target tissue or cell tropism will reduce the amount of nucleic acid cargo required to achieve a therapeutic effect, thereby improving potency.

Recombinant protein-based nanoparticles can easily incorporate receptor targeting domains, including peptide motifs and antibody fragments that promote internalization via receptor-mediated endocytosis. As described, this strategy has been pursued in ELP and SLP-based nucleic acid delivery systems, leveraging peptides targeting surface receptors that are upregulated in a variety of cancer types.^104,109,119 Likewise, antibody variable fragments derived from camelid single-chain antibodies (nanobodies) have been used to decorate protein nanoparticles and promote targeted internalization.^77,78 Costa et al. designed diblock ELPs modified with a nanobody targeting EGFR overexpressed in many cancers, which led to a significant improvement in uptake in EGFR+ cell lines.⁷⁷ In addition to active targeting, passive targeting can be achieved by altering surface chemistry to modulate the composition of the corona and particle trafficking. This underexplored approach can potentially be utilized to generate protein nucleic acid delivery vectors with unique tropisms.^48,287,288

6.4. Safety and Manufacturing

For protein nanoparticle delivery systems to be clinically viable, there will be a need for scalable production, as well as analytical methods to ensure that these materials meet medical grade quality and safety standards. Immunogenicity is an important feature to consider in assessing the safety profile of protein materials. Many protein materials derived from or inspired by human proteins, such as ELPs, nanocage proteins, and human serum albumin have demonstrated low immunogenicity in settings relevant to their use as nucleic acid delivery vectors. Other proteins with high homology to human proteins, such as bovine serum albumin, gelatin, and collagen, as well as silk are capable of inducing an allergic response in rare cases, but have typically demonstrated low immunogenicity.^292–294

Of greater concern are xenogeneic proteins and other molecules that may be present in material preparations, which may be residual from material source organisms, including E. coli for recombinant proteins, silkworms or spiders for silk fibroin, or bovine or porcine sources for collagen, gelatin, and albumin. Non-mammalian-derived endosomolytic and cell-penetrating peptides incorporated to facilitate cytosolic penetration of cargo may represent another source of potential immune stimulation, though there is some evidence suggesting that certain CPPs are non-immunogenic.²⁹⁵ The purification of protein materials by affinity chromatography or acidic or thermal extraction, along with the removal of inflammatory contaminants like endotoxin should minimize the effect of xenogeneic contaminants. In addition, in silico predictions of peptide antigenicity can be leveraged in the design of cell penetrating and endosomolytic peptides to select for sequences displaying minimal risk of immunogenicity.^296–298

Recombinant protein materials to be used as excipients for nucleic acid delivery can be produced at scale using existing biomanufacturing methods. E. coli expression systems have been favored in the laboratory setting due to low cost, high yields, and well-established protocols.^299–301 While E. coli do not easily generate human-like glycosylation or phosphorylation features, such post-translational modifications are typically not necessary for the assembly of protein nanoparticles, making E. coli an ideal vector for scaled recombinant protein material production if paired with high fidelity purification methods. Recently, continuous manufacturing has been explored as a strategy to improve yields and quality, reduce operational costs, and increase efficiency by reducing the pauses and transfers associated with batch manufacturing.^299,302 Downstream of protein production and purification, the quality of protein materials used for nucleic acid delivery should be assessed using quality control tests focused on purity (e.g. SDS-PAGE, and liquid chromatography methods), homogeneity (e.g. DLS or other light scattering methods, size exclusion chromatography), and identity (e.g. mass spectrometry).^303,304

7. Conclusions

Addressing the longstanding delivery challenges associated with nucleic acid therapies is critical to unlocking the full potential of this modality to transform disease outcomes. As highlighted in this review, protein-based nanoparticles are well-positioned to serve as a vehicle that both combines many desirable features of existing nucleic acid delivery systems and overcomes their limitations. Recombinant protein materials exhibit the genetic programmability of existing viral vector systems with improved robustness to modification, enabling highly selective incorporation of peptide-based cell receptor targeting domains and membrane active domains to mediating cargo endosomal escape. Moreover, in contrast to viral delivery systems, engineered protein materials also demonstrate a level of scalability, resilience to harsh, denaturing purification conditions, and cargo accommodation capacity comparable to existing non-viral delivery systems. More broadly, protein materials are biodegradable and can be designed with limited immunogenic risk, and therefore, suitable for many clinical applications requiring repeated nucleic acid delivery. As issues of potency and biostability are addressed, protein nanoparticles are poised to mature into a promising new class of nucleic acid delivery vectors for the treatment of human disease.