Self-assembling protein nanoparticles for cytosolic delivery of nucleic acids and proteins

Abstract

Safe and effective intracellular delivery of biomacromolecules can treat a variety of clinical disorders. Here, we report the development of Elastin-based Nanoparticles for Therapeutic delivERy (ENTER), a recombinant elastin-like polypeptide (ELP)-based delivery system for effective cytosolic delivery of biomacromolecules in vitro and in vivo. Through iterative design, we developed fourth-generation ELPs fused to cationic endosomal escape peptides (EEPs) that self-assemble into pH-responsive micellar nanoparticles and enable cytosolic entry of cargo following endocytic uptake. In silico screening of α-helical peptide libraries led to the discovery of an EEP (EEP13) with 48% improved protein delivery efficiency versus a benchmark peptide. Our lead ELP-EEP13 shows similar or superior performance compared to lipid-based transfection reagents in the delivery of mRNA-encoded, DNA-encoded, protein-form Cre recombinase and CRISPR gene editors, as well as siRNAs to multiple cell lines and primary cell types. Intranasal administration of ELP-EEP13 combined with Cre protein achieved efficient editing of lung epithelial cells in reporter mice.

Introduction

There is growing interest in the use of nucleic acid and protein-based drugs to modulate intracellular targets, enable cytosolic protein production, and edit the genome. The past decade has witnessed the emergence of highly effective therapies for vaccination,^1,2 gene knockdown,³ gene editing,^4,5 protein replacement,⁶ and cell reprogramming^7,8 for the treatment of a variety of infectious diseases, genetic disorders, and cancer. Nonetheless, the full potential of these therapeutic modalities has yet to be realized due to several challenges associated with the delivery of these molecules. Several classes of viral and non-viral delivery vehicles have been explored to facilitate either direct cell membrane penetration or endosomal escape of therapeutic macromolecules.^9,10 Existing viral vectors remain limited by stringent cargo capacities, immunogenicity, potential for oncogenic gene insertion, and manufacturing scaling challenges.^11,12 Lipid nanoparticles are the leading non-viral delivery modality, but the incorporation of active targeting domains requires chemical conjugation that is challenging to scale, and the human relevance of passive targeting strategies reported in animal models remains unclear.^13,14 Cationic membrane disruptive peptides, including cell penetrating peptides such as Tat¹⁵, oligoarginine¹⁶, and penetratin¹⁷ or other endosomolytic peptides, as exemplified by S10¹⁸ have also been examined as vectors in the form of conjugates or as untethered peptide shuttles.^19,20 However, these peptides remain limited by low potency and cytotoxicity.

Elastin-like polypeptides (ELPs) are recombinant protein polymers that exhibit thermally responsive self-assembly, enabling the formation of micelles that have been utilized as drug carriers for various cargo.^21–23 ELPs are composed 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. Specifically, ELPs are soluble below a transition temperature (T_t), inversely correlated to guest residue (X) hydrophobicity and protein size^24,25, but above the transition temperature reversibly phase separate to form insoluble coacervates.²⁶ We have previously designed amphiphilic diblock ELPs that self-assemble into micellular nanoparticles with a hydrophilic surface and hydrophobic core for targeted extracellular drug delivery of proteins and small molecules.^27–29 However, attempts to utilize ELPs for the intracellular delivery of proteins, gene editor ribonucleoproteins (RNPs) and mRNA have not been reported and efforts to deliver siRNA and plasmid DNA (pDNA) have been associated with low potency, presumably due to particle instability related to amorphous polyion complex (PIC) architecture and an inability to mediate effective endosomal escape.^30–32

In this report, we describe the design of Elastin-based Nanoparticles for Therapeutic delivERy (ENTER), a versatile delivery system that achieves efficient cytosolic delivery of therapeutic macromolecule modalities to a variety of cell types both in vitro and in vivo. Through refinements over four generations of designs, we have engineered ELP multiblock protein polymers that self-assemble as crosslinked micellar nanoparticles displaying a hydrophilic corona and a histidine-containing hydrophobic core, which includes cationic endosomal escape peptides (EEPs) (Fig. 1a). This organized, pH-responsive micellar architecture enables effective encapsulation of nucleic acid cargo and drives disassembly of particles upon trafficking to acidic endosomes. Significantly, iterative machine learning-enabled in silico design and in vitro screening led to our discovery of an EEP (EEP13) sequence that enhanced intracellular delivery of both proteins and nucleic acids. Of note, EEPs as a component of ELP nanoparticles displayed enhanced potency with minimal cytotoxicity as compared to free EEP shuttles. Moreover, this system provided an effective means for cytosolic delivery of siRNA, mRNA, plasmid DNA, proteins, as well as Cas9 and ABE8e RNP in cell lines as well as primary fibroblasts, macrophages, T cells, and hematopoietic stem cells. Delivery of interferon regulatory factor 5 (IRF5), an M1 macrophage lineage-defining transcription factor, to macrophages in vitro significantly upregulated the expression of key pro-inflammatory cytokines. Finally, we demonstrate that intranasal administration of ELP-EEP13 combined with Cre recombinase protein achieved effective editing of lung epithelial cells in reporter mice.

Figure 1 |. Elastin-based Nanoparticles for Therapeutic delivERy (ENTER).

a) Schematic of micelle-forming ELPs containing (from N- to C-terminus) a hydrophilic block, cysteine-based crosslinking block, hydrophobic block, and endosomolytic block designed for siRNA, mRNA, pDNA, and protein cargo delivery. Created using BioRender. b) ELP block designs of each ELP version. The listed hydrophilic and hydrophobic block designs are repeated n = 10 and n = 15 times respectively in the full ELP sequences. Underlined residues denote the guest residue X selected for use in the elastin-derived VPGXG motif. c) Particle size distribution by volume of V2-ELP at varying pH ([ELP] = 10 μM, n = 3 technical replicates). d) % RFP+ HEK293-RFP cells following treatment with Cre protein either conjugated to V1-ELP or His-containing V2-ELP, with and without chloroquine (CRQ) supplementation. [Cre-ELPs] = 8 μM, n = 3 biological replicates. e) % RFP+ and viability of HEK293-RFP cells following treatment with Cre protein conjugated to V3 (n = 2 biological replicates). f) Representative negative stain TEM image of dehydrated V4-ELP nanoparticles. Scale bar = 50 nm, inset scale bar = 10 nm. Assessment was repeated once with similar results. g) % RFP+ and viability of HEK293-RFP cells following treatment with V4-ELP-EEPs and free Cre protein. [ELP] = 8 μM, [Cre protein] = 2 μM, n = 3 biological replicates. h) % RFP+ and viability of HEK293-RFP cells following treatment with V4-ELP-S10 or S10 peptide with free Cre protein at 2 μM (n = 3 biological replicates). Data represented by mean ± SEM.

Results

Formation and characterization of ELP-protein nanoparticles

We have previously reported a first-generation micelle-forming ELP (V1-ELP) bearing an N-terminal triglycine (GGG) motif for sortase-mediated conjugation.²⁹ Micelle self-assembly was observed above the transition temperature of the hydrophobic block (10°C)³³ with dynamic light scattering (DLS) demonstrating an average particle diameter of ~60 nm above a critical micelle concentration (CMC) of ~500 nM (Supp. Fig. 1a–b).³³ V1-ELPs were effectively internalized by HeLa cells and HEK293 cells without cytotoxicity (Supp. Fig. 1c–d).

We assessed the ability to conjugate Cre recombinase and Streptococcus pyogenes Cas9 (SpCas9) protein modified with a C-terminal LPETG motif and a protease cleavable (GGS)₉ linker³⁴ to ELP nanoparticles by evolved sortase A (eSrtA), an optimized variant of the bacterial transpeptidase SrtA.³⁵ SDS-PAGE confirmed the ability to conjugate Cre protein to about 50% of accessible GGG motifs (Extended Data Fig. 1a–b) with an increase in nanoparticle diameter to 70-80 nm (Extended Data Fig. 1c). Cathepsin B, an endosomal protease, effectively liberated Cre and induced degradation of the ELP protein polymer (Supp. Fig. 1e–f). Similarly, Cas9 can be conjugated to V1-ELPs (Extended Data Fig. 1d, Supp. Fig. 1g) with an increase in particle diameter, further augmented by the introduction of single guide RNA (sgRNA) with a commensurate reduction in zeta potential to generate ELP-bound Cas9 ribonucleoproteins (RNP) (Extended Data Fig. 1e–f). We also demonstrated the ability of an anti-CD117 antibody functionalized with an LPETG-containing linker to be conjugated to V1-ELPs. CD117 (c-kit) is upregulated on human and murine hematopoietic stem cells (HSC) and has previously been used for selective HSC targeting.^36,37 SDS-PAGE and band densitometry analysis confirmed antibody conjugation (Extended Data Fig. 1g–h). Antibody mediated ELP targeting to P815 cells, a CD117+ murine mast cell line, was evaluated by conjugating Alexa Fluor 633 to a CD117 mAb-functionalized V1-ELP variant via a C-terminal cysteine domain. Significantly increased binding was observed compared to a non-antibody functionalized control, confirming selective receptor engagement (Extended Data Fig. 1i–j, Supp. Fig. 1h–i).

ELP-mediated intracellular protein delivery

We employed an iterative design strategy to optimize ELP-mediated intracellular delivery of protein cargo (Fig. 1b). We first replaced the valine residues present in the hydrophobic block of V1-ELP with histidine (V2-ELP). Histidine protonation due to endosomal acidification is expected to dramatically increase the hydrophilicity of the hydrophobic block, thereby, facilitating micelle disassembly and promoting endosomal swelling and osmolysis through a proton sponge effect.³⁸ Measurement of V2-ELP micelle size at varying pH demonstrated effective micelle formation at neutral pH, and disassembly at pH below the pKa of the histidine side chain (~6.8) (Supp. Fig. 2a, Fig. 1c). HEK293 Cre reporter cells, containing a floxed stop cassette preceding an RFP gene (HEK293-RFP), were treated with V2-ELP nanoparticles functionalized with Cre recombinase protein (Cre-V2-ELP) via sortase conjugation. Cre-functionalized V2-ELP formed monodisperse nanoparticles with increased (~5 nm) diameter compared to unfunctionalized V2-ELP nanoparticles (Supp. Fig. 2b–d). RFP expression was only observed in the presence of chloroquine, a cationic amphipathic drug with known endosomolytic capacity³⁹, but enhanced by the incorporation of histidines in V2-ELP (Fig. 1d).

To endow ELPs with membrane disruption capabilities, V2-ELPs were modified with endosomal escape peptides (EEPs), known to enhance intracellular delivery of macromolecular cargo (V3-ELPs).^{15,18,40–42} V3-ELPs were generated containing one of five EEPs positioned at the N-terminus of the protein polymer; each constructs demonstrated stable size at room temperature over the course of 72 hours and was amenable to sortase-mediated functionalization with Cre recombinase. (Supp. Fig. 2e–i). HEK293-RFP cells treated with Cre-V3-ELPs for 48 hours displayed dose-dependent RFP expression (Fig. 1e). The most effective ELPs contained Tat¹⁵ or S10¹⁸, enabling RFP expression in up to 30% of cells at a concentration of 16 μM, though some cytotoxicity was observed at this dose (Fig. 1e). To further optimize Cre release, six linkers with known susceptibility to a variety of endolysosomal proteases were screened (Supp. Fig. 3a).^43–46 HEK293-RFP cells were treated with V3-ELPs bearing S10 and conjugated to linker-modified Cre. A furin cleavable linker flanked by GGS repeats (GGSGGSRVRRGGSGGS) demonstrated the highest release efficiency, as evident by the ratio of RFP+ cells after treatment with Cre protein in ELP-conjugated versus unconjugated formats (Supp. Fig 3b). This linker was utilized in subsequent studies.

Cytotoxicity remains a shortcoming of all existing cell penetrating and membrane disruptive peptides, which has been attributed to excessive disruption of the plasma cell membrane.⁴⁷ To address this limitation, we designed V4-ELPs with the EEP sequence positioned at the C-terminus of the protein polymer, so as to be hidden within the nanoparticle core. We postulated that once concentrated within an endosome, protonation of histidines would lead to micelle disassembly, exposing EEPs and enabling endosomal disruption. We further hypothesized that these cationic EEPs would also facilitate encapsulation of nucleic acids. To circumvent electrostatic interactions between the C-terminal cationic EEP and the N-terminal anionic hydrophilic block, glutamic acid residues in hydrophilic block were replaced by alanine and glycine at a 3:2 ratio (Fig. 1b). V4-ELPs form monodisperse spherical micelles with a hydrodynamic diameter of ~70 nm as determined by DLS, with a CMC of ~500 nM (Supp. Fig. 2j). Negative-stain TEM imaging, collected under dehydrated conditions, shows clusters of monodisperse spherical particles that are smaller in diameter (~30 nm, Fig. 1f). This phenomenon likely occurs because the hydrophilic corona of the particles dehydrates and self-associates.

V4-ELP constructs were created that were further recombinantly modified with the EEPs noted above, as well as nonaarginine (R₉)⁴⁸, transportan-10⁴⁹, and L17E M-lycotoxin⁵⁰. Treatment of HEK293-RFP cells with V4-ELP-EEPs (8 μM) and free, unbound Cre protein (2 μM) revealed that V4-ELPs containing S10 enabled RFP expression in nearly all cells without cytotoxicity (Fig. 1g). Notably, Cre protein delivery by V4-ELP-S10 was superior to that achieved using S10 as an untethered peptide, with over 20-fold less ELP being required to achieving 50% RFP+ cells under serum-containing transfection conditions (~1.5 μM vs 32 μM, Fig 1h). Cell viability was unaffected by treatment with V4-ELPs at all doses tested, in contrast to free S10 peptide, which was associated with substantial toxicity at its EC₅₀ (Fig. 1h). V4-ELP-S10 also outperformed S10 peptide in Cre protein delivery to HEK293-RFP cells under serum-free transfection conditions (Supp. Fig. 4a). Direct comparison of the delivery efficiency of V3- and V4-ELPs containing S10 at either the surface-facing N-terminus (S10-V3-ELP, S10-V4-ELP) or core-facing C-terminus (V3-ELP-S10, V4-ELP-S10) revealed that the V4-ELP-S10 configuration facilitated the highest Cre protein delivery efficiency to HEK293-RFP cells at both Cre doses assessed (Supp. Fig. 4b). V4-ELPs outperformed Lipofectamine 3000 in the delivery of Cre protein, with RFP expression observed in only 40% of HEK293-RFP cells under optimal Lipofectamine conditions (Supp. Fig. 4c). These findings confirm that V4-ELPs are effective vectors for intracellular delivery of protein cargo.

Discovery of enhanced EEPs via in silico screening

To further improve the efficacy of V4-ELPs, we sought to design EEPs with enhanced activity. The most effective peptide assessed was S10; in contrast to other polycationic EEPs that are either unstructured or contain a single α-helical domain, S10 is comprised of amphipathic α-helical peptide variants of CM18⁵¹ and PTD4⁵², separated by a Gly-Ser linker. The amphipathic α-helical structure is thought to facilitate intercalation within lipid bilayers, particularly negatively charged bilayers, characteristic of late endosomes.^53,54

Motivated by this insight, we compiled and screened an α-helix peptide database to identify next generation EEPs with enhanced functionality. To enable prediction of EEP efficacy, a multiple linear regression model was trained on a published dataset of 73 membrane disruptive peptides encoded by a set of 15 physicochemical, structural, and compositional features, labeled with efficacy values derived from an in vitro GFP protein delivery assay (Supp. Table 1).⁵⁵ F scores of each feature with respect to peptide efficacy revealed specific physicochemical properties, including hydrophobic moment, net charge, and amphipathic α-helical structure or lack thereof, which were highly correlated with delivery efficacy (Supp. Fig. 5a). The model with the top performance on held-out data points (RMSE = 12.4) is described by the equation (Fig. 2a):

$$\text{Predicted efficacy}=5.9\times (\mathrm{Hydrophobic\; Moment})+3.2\times (\mathrm{Net\; Charge})-3.2\times (\mathrm{pl})-4.3\times (\mathrm{\%\; Negative\; Residues})+20.8\times (\mathrm{is\; Amphipathic\; Helix}).$$

Figure 2 |. Discovery of enhanced endosomolytic peptides via in silico screening.

a) Comparison of observed and predicted % GFP+ HeLa cells treated with GFP proteins and a library of membrane disruptive peptides. Data from Ref 57. b) In silico procedure for screening of alpha-helix peptide libraries. c) % RFP+ HEK293-RFP cells following treatment with V4-ELP-EEP (8 μM) and free Cre recombinase protein (100 nM) for 48 hours (n = 3 biological replicates). d) Predicted AlphaFold2 structure, helical wheel diagrams, and peptide sequence of EEP13. e) Diameter of V4-ELP-[EEP13]x nanoparticles (x = 1-3, where x is the copy number of EEP13). [ELP] = 10 μM, n = 3 technical replicates. f) % RFP+ and viability of HEK293-RFP cells following treatment with V4-ELP-[EEP13]_0-3 and free Cre recombinase protein for 48 hours ([Cre recombinase] = 2 μM, n = 3 biological replicates). g) % RFP+ HEK293-RFP cells following treatment with V4-ELP-EEP13 formulated with Cre recombinase protein. Cells in the +BafA1 condition were pre-treated with or without 50 nM Bafilomycin A1 for 1 hour to inhibit endosomal acidification prior to transfection. n = 3 biological replicates. h) Editing efficiency in HEK293 cells 72 hours after treatment with SpCas9 RNP containing HPRT1 sgRNA and V4-ELP-EEP13 assessed by Sanger sequencing. [ELP] = 8 μM, n = 3 biological replicates. i) A to G conversion efficiency in HEK293 cells 72 hours after treatment with ABE8e RNP containing HEK3 sgRNA and V4-ELP-EEP13 assessed by Sanger sequencing. CM: CRISPRMAX, [ELP] = 8 μM, n = 3 biological replicates. *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 unless otherwise specified, one-way ANOVA with Dunnett’s (c, comparison to S10) or Tukey’s (e) correction for multiple comparisons; two-way ANOVA with Dunnett’s correction for multiple comparisons (f, comparison to V4-ELP-EEP13); (g) Unpaired two-tailed t-tests with Holm-Šídák correction for multiple comparisons. All data represented by mean ± SEM.

This model was used in a multistage screen of a library of peptides compiled from the Antimicrobial Peptide Database (APD)⁵⁶, the Database of Antimicrobial Activity and Structure of Peptides (DBAASP)⁵⁷, and the Therapeutic Peptide Design database (TP-DB)⁵⁸, a database containing all α-helical segments extracted from protein structures compiled in the Protein Data Bank (Fig. 2b). After filtering to remove peptides 1) containing non-canonical amino acids, 2) without cationic residues, and 3) duplicates, a total of 11,084 helical peptides were evaluated using the model. Peptides with predicted efficacies greater than 10% were combined into dimers separated by a GGSGGGS linker and then reassessed. Seven lead peptides were selected (Supp. Table 2) based on predicted efficacy.

V4-ELPs containing each of these derived EEPs were generated and assessed for their ability to deliver free Cre protein at a dose of 100 nM to avoid the editing saturation observed in prior screening studies at 2 μM. The highest performing first generation EEP (EEP5) led to RFP expression in 25.7% of HEK293-RFP cells, but underperformed S10 (44.1%, Fig. 2c). The two highest performing EEPs (EEP1 and EEP5) were enriched in lysine over arginine among positively charged residues. Motivated by this insight, a second generation of EEPs was identified by rescreening the peptide library to select peptides enriched in lysine (Supp. Fig. 5b–c). The top eight EEP peptides derived from this in silico screen were evaluated for their capacity to deliver Cre protein to HEK293-RFP cells. Two of these constructs, EEP13 and EEP14, significantly outperformed S10 (Fig. 2c). The lead EEP, EEP13, is a dimeric amphipathic α-helical peptide composed of two synthetic antimicrobial peptides (Fig. 2d).^59,60 As a free peptide alone, S10 significantly outperformed EEP13 in Cre protein delivery and demonstrated less cytotoxicity in a peptide dose escalation study (Supp. Fig. 6a–b). The separation in performance between the two peptides emerges at doses where EEP13 displayed increased cytotoxicity, suggesting that elevated cytotoxicity may contribute to reduced potency. These findings indicate that by mitigating cytotoxicity, ELP nanoparticles designed to shield highly potent EEPs until localized within the endosome can enable the discovery of EEPs with greater levels of endosomal escape efficiency that would otherwise not be achievable with screening of free peptides.

To further characterize the V4-ELP-EEP13 delivery system, nanoparticle size and stability were monitored at room temperature (25°C) and under physiologic (37°C) conditions for up to 72 hours, in the presence or absence 5% w/v bovine serum albumin (BSA) as a mimic of human serum. V4-ELP-EEP13 nanoparticles maintained a stable size of ~60 nm at both temperatures in the absence of BSA. In the presence of BSA, nanoparticle diameter increased modestly (~50-70 nm), likely due to surface adsorption of BSA, but remained stable over 72 hours without further increase in size or evidence of aggregation (Supp. Fig. 2k–l).

Multimerization of cell-penetrating peptides enhances membrane disruption by increasing binding affinity of multimeric cationic peptides for anionic proteoglycans in the plasma membrane.^61,62 Given that endosomal membranes are enriched in anionic phospholipid bis(monoacylglycero)phosphate (BMP)⁶³, we hypothesized that EEP13 multimerization would further increase potency. V4-ELPs containing EEP13 as a monomer (V4-ELP-EEP13), dimer (V4-ELP-[EEP13]₂), or trimer (V4-ELP-[EEP13]₃) were designed and expressed for evaluation. All three constructs formed uniform nanoparticles between 60 and 80 nm in diameter, with micelle size increasing as additional copies of EEP13 were incorporated into the ELP (Fig. 2e). The monomeric, dimeric, and trimeric V4-ELP-EEP13 constructs, as well as a V4-ELP construct without EEP13 as a control, were next evaluated for their ability to deliver free Cre protein to HEK293-RFP. As expected, V4-ELPs containing an EEP13 dimer or trimer outperformed those containing an EEP13 monomer at ELP doses between 500 nM and 2 μM, though increased cytotoxicity relative to the monomer was observed at and above 1 μM (Fig. 2f). In contrast, V4-ELP construct without EEP13 was ineffective in Cre protein delivery, highlighting the critical role of EEP13 in mediated endosomal escape. Notably, V4-ELP-[EEP13]₂ displayed an EC₅₀ of ~500 nM in the presence of 2 μM Cre, while 30 nM Cre protein was sufficient to induce a 50% RFP+ rate when delivered with 8 μM of all three ELPs (Supp. Fig. 7a–b). To examine the ability of V4-ELP-EEPs to deliver cargo larger than Cre recombinase (38 kDa), FITC-labeled dextrans with molecular weight up to 2 megadaltons were tested as cargo in a co-formulation study. Both V4-ELP-EEP13 and V4-ELP-[EEP13]₂ effectively facilitated cytoplasmic entry of dextrans of all tested sizes (Supp. Fig. 8). Collectively, these findings support the ability of V4-ELP-EEP13 nanoparticles to deliver large macromolecules.

We hypothesize that V4-ELP-EEP13-mediated intracellular protein cargo delivery is dependent on endosomal acidification, as this reduction in pH during endosomal trafficking (1) induces V4-ELP nanoparticle disassembly due to histidine protonation, enabling EEP13 exposure and endosomolytic action, and (2) triggers endosomal water influx and osmolysis due to the proton buffering capacity of the ELP. To test this hypothesis, we assessed the effect of Bafilomycin A1 (BafA1), an inhibitor of the vacuolar-type H+-ATPase proton pump that drives endosomal acidification, on the efficacy of V4-ELP-EEP13-mediated delivery of Cre recombinase protein to HEK293-RFP cells. HEK293-RFP cells were pretreated with 50 nM BafA1 immediately prior to treatment with V4-ELP-EEP13 formulated with Cre protein. As expected, treatment with BafA1 significantly reduced delivery efficiency across all treatment conditions (Fig. 2g). This data supports the mechanism that ENTER-mediated intracellular cargo delivery is dependent on endosomal acidification and histidine protonation, likely related to both protonation-induced micelle disassembly, as well as endosomal swelling and disruption.

Transient delivery of gene editors in protein form has the potential to reduce off-target editing by limiting the lifetime of the editing agent within treated cells.⁶⁴ However, the large size and complex charge of gene editor RNPs presents multiple delivery challenges. To test the ability of our ELP delivery system to overcome these challenges, we first used V4-ELP-EEP13 to deliver free SpCas9 RNP formulated with a chemically modified sgRNA targeting HPRT1 to HEK293 cells. At an ELP concentration of 8 μM, V4-ELP-EEP13 facilitated efficient gene editing at a level (~65%) comparable to the positive control CRISPRMAX, a commercial lipid-based system optimized for RNP delivery (Fig. 2h). We next assessed the ability of ELP nanoparticles to deliver an adenine base editor (ABE) RNP by co-formulating V4-ELP-EEP13 with ABE8e complexing an sgRNA targeting the HEK3 locus in HEK293. Consistent with the Cas9 delivery results, V4-ELP-EEP13 facilitated potent ABE8e delivery and effective editing at both target adenines, enabling 83% A>G editing at both bases with an RNP dose of 300 nM (Fig. 2i). Notably, the ABE delivery performance of V4-ELP-EEP13 exceeded that of CRISPRMAX at both RNP doses assessed (30 and 300 nM). Collectively, these studies establish the potential of ENTER to serve as a delivery vehicle for gene editors.

ELP-mediated protein delivery to primary cells ex vivo

The ability of ELP nanoparticles to deliver Cre protein was evaluated in a variety of primary cells, including lung fibroblasts, peritoneal macrophages, splenic CD4+ T cells and hematopoietic stem cells from Ai9 reporter mice. Ai9 mice carry a loxP-flanked stop codon cassette located upstream of a gene encoding tdTomato. Upon successful Cre-mediated recombination, the stop cassette is excised, allowing tdTomato expression. In fibroblasts, co-treatment with V4-ELP-EEP13 and Cre protein resulted in a dose-dependent increase in tdTomato+ cells of up to 70% (Fig. 3a–b). In contrast, Lipofectamine 3000 exhibited lower effectiveness in Cre protein delivery, achieving tdTomato+ expression in fewer than 15% of cells at the highest tested dose (Supp. Fig. 9a–b). In macrophages, an optimal dose of 8 μM V4-ELP-EEP13 with 8 μM Cre resulted in editing in over 80% of treated cells (Fig. 3c–d). Similarly, in CD4+ T cells, Cre protein and V4-ELP-EEP13 dose-dependently increased tdTomato-positive cells, reaching up to 90% (Fig. 3e–f). Finally, V4-ELP-EEP13 facilitated effective delivery of Cre protein into Lin⁻Sca-1⁺ckit⁺ hematopoietic stem cells, enabling ~20% tdTomato+ cells at optimal dosing (Fig. 3g–h). Minimal ELP associated cytotoxicity was observed for all cell types and conditions (Supp. Fig. 10).

Figure 3 |. ELP-mediated protein delivery to primary cells.

a-b) % tdTomato+ in Ai9-derived mouse lung fibroblasts, c-d) peritoneal macrophages, e-f) splenic CD4+ T cells, and g-h) Lin⁻Sca-1⁺c-kit⁺ (LSK) hematopoietic stem cells following Cre protein and V4-ELP-EEP13 co-treatment (n = 3 biological replicates). In a, c, e, g - [Cre] = 8 μM. b, d, f, h – [ELP] = 8 μM. i) Fluorescence images with nuclear staining (blue) taken 2 hours after treatment of murine peritoneal macrophages with 100 nM AF488-labeled IRF5 alone (top) and in combination with 4 μM V4-ELP-EEP13 (bottom). Scalebar = 25 μm. j) IFNg and k) IL12b, and l) IL1b expression in murine macrophages assessed via qPCR 72 hours after treatment with IRF5 alone (500 nM), V4-ELP-EEP13 alone (8 μM), and both agents (n = 3 biological replicates. Significance was assessed using a two-way ANOVA with Tukey’s correction for multiple comparisons (j-l). Data represented by mean ± SEM.

We next assessed the ability of ENTER to facilitate delivery of transcription factors and enable reprogramming of cell state.⁶⁵ Tumor-associated macrophages have been reprogrammed towards an antitumor phenotype by delivery of mRNA-encoded interferon regulatory factor 5 (IRF5), a proposed master transcription factor for pro-inflammatory M1 macrophages.^66,67 However, the delivery of protein-form IRF5 has not been previously achieved and could be advantageous compared to mRNA due to improved cargo stability, dosing accuracy, and its functionality independent of translational activity of targeted cells. Motivated by this, V4-ELP-EEP13 was evaluated for its ability to deliver IRF5 protein to primary mouse macrophages in vitro. V4-ELP-EEP13 effectively enabled the internalization of AF488-labeled IRF5 into murine peritoneal macrophages after 2 hours (Fig. 3i). Macrophages were then treated with V4-ELP-EEP13 alone, IRF5 alone, or both for 48 hours. Expression of Ifng, Il12b, and Il1b, all genes under IRF5 regulation, significantly increased after co-treatment with IRF5 and V4-ELP-EEP13 (Fig. 3j–l). Limited increases in cytokine expression were observed following treatment with V4-ELP-EEP13 or recombinant IRF5 alone, suggesting enhanced production is related to IRF5 transcriptional activity enabled by ELP-mediated cytosolic entry.

Evaluation of ELP nanoparticles for in vitro siRNA delivery

Towards enabling the use of ELP nanoparticles for nucleic acid cargo delivery, we hypothesized that V4-ELPs containing a cationic C-terminal EEP would effectively encapsulate nucleic acids through combining electrostatic and physical complexation (Fig. 4a) and the EEP domain would remain effective as an endosomolytic agent following pH-induced ELP disassembly within the endosome, enabling potent RNA delivery.

Figure 4 |. Complexation and delivery of siRNA by ELP nanoparticles.

a) Schematic of ELP-siRNA nanoparticle formation via combined electrostatic and heat-induced hydrophobic interactions. b) Gel shift and c) Ribogreen assay evaluating V4-ELP-EEP13 and V4-ELP-[EEP13]₂ siRNA encapsulation efficiency (n = 3 technical replicates for Ribogreen, [siRNA] = 50 nM). d) Agarose gel depicting scrambled siRNA exposed to RNAse A for 6 hours at 37°C with or without complexation with V4-ELP-EEP13 at varying ELP:siRNA molar ratios. e) Diameter and f) zeta potential of V4-ELP-EEP13 and V4-ELP-[EEP13]₂ nanoparticles loaded with siRNA vs ELP nanoparticles or siRNA alone (n = 3 technical replicates). [ELP] = 5 μM, [siRNA] = 50 nM. g) MFI of HeLa cells treated with Alexa Fluor 647-labeled siRNA complexed with V4-ELP-EEP13 series compared to RNAiMAX after 1, 2, 4, and 24 hours of treatment. [ELP] = 5 μM, [siRNA] = 50 nM, n = 3 biological replicates. Flow cytometry assessment was conducted in 0.2% v/v trypan blue to quench signal from cell membrane-associated siRNA. h) % GFP+ of HeLa-d2eGFP cells following treatment with siGFP complexed with V4-ELP-EEP13 series at varying doses (n = 3 biological replicates, [siRNA] = 100 nM). Expression levels for all conditions are normalized to scrambled siRNA (siScr)-treated controls. i) GAPDH mRNA expression in Ai9-derived murine lung fibroblasts following 48 hours of treatment with siGAPDH, assessed via qPCR. Expression levels are normalized to cells treated with siRNA alone. [ELP] = 10 μM, [siRNA] = 500 nM, n = 3 biological replicates. j) MFI (bars) and viability (points) of HEK293 cells treated with Alexa Fluor 647-labeled siRNA (50 nM) complexed with V4-ELP-EEP13 (5 μM) following treatment with chemical endocytosis inhibitors. Cells were pretreated with inhibitors for 1 hour then treated with ELP-siRNA nanoparticles for an additional 2 hours (n = 3 biological replicates). *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 unless otherwise specified, two-way (e-h, j) or one-way (i) ANOVA with Dunnett’s or Šídák correction for multiple comparisons as appropriate. All data represented by mean ± SEM.

Both V4-ELP-EEP13 and V4-ELP-[EEP13]₂ were first assessed for their ability to complex siRNA. Both ELPs displayed effective dose-dependent complexation of siRNA via gel shift and RiboGreen RNA binding dye exclusion assays (Fig. 4b–c). Near 100% encapsulation efficiency was observed at a V4-ELP protein polymer:siRNA molar ratio above 10 for both constructs. To assess the extent to which ELP complexation protects RNA cargo from enzymatic degradation, ELP-siRNA nanoparticles formulated at varying ELP:siRNA molar ratios were incubated with RNAse A for 6 hours at 37°C. siRNA integrity was subsequently assessed by agarose gel electrophoresis following heparin-mediated decomplexation. siRNA formulation with V4-ELP-EEP13 at molar ratios at and above 10 provided increased protection from RNAse-mediated degradation as compared to unformulated controls (Fig. 4d). After siRNA complexation, V4-ELP-EEP13 and V4-ELP-[EEP13]₂ nanoparticles showed a modest (~10-20 nm) increase in size (Fig. 4e). This increase in size was accompanied by a decrease in zeta potential (Fig. 4f). We subsequently assessed the ability of V4-ELPs to enable intracellular delivery of Alexa Fluor 647-labeled siRNA to HeLa cells. V4-ELPs containing EEP13 and [EEP13]₂ enabled rapid internalization, with nearly double that of a lipid-based transfection reagent (RNAiMAX) (Fig. 4g).

To evaluate functional siRNA delivery, V4-ELP-EEP13 and V4-ELP-[EEP13]₂ were used to deliver an siRNA mediating GFP knockdown in HeLa cells expressing a destabilized GFP (HeLa-d2eGFP). As observed in Cre protein delivery studies, the dimeric construct displayed increased potency compared to the monomeric construct, mediating up to 60% GFP knockdown at an ELP concentration of 5 μM with 100 nM siGFP following treatment for 24 hours. Minimal cytotoxicity was observed for both constructs (Fig. 4h, Supp. Fig. 11a–b). Further increasing ELP treatment concentration to 10 μM substantially reduced the efficacy of V4-ELP-[EEP13]₂ nanoparticles, while V4-ELP-EEP13 particles displayed further improvement (~75% GFP knockdown, Fig. 4h, Supp. Fig. 11c). This effect may be related to a decrease in siRNA endosomal de-complexation efficiency from V4-ELP-[EEP13]₂ at high ELP concentrations, as the increased charge of V4-ELP-[EEP13]₂ results in higher avidity interactions with nucleic acid cargo compared to V4-ELP-EEP13. Notably, V4-ELP-EEP13 also effectively facilitated delivery of a GAPDH-targeted siRNA into murine lung fibroblasts, inducing ~95% knockdown of mRNA expression, comparable to RNAiMAX (Fig. 4i). Collectively, these data demonstrate that ELPs serve as an effective vehicle for in vitro siRNA delivery.

We next interrogated the endocytic pathways facilitating ELP nanoparticle internalization. HEK293 cells were pretreated for 1 hour with chemical inhibitors known to primarily affect specific pathways: chlorpromazine (clathrin-mediated endocytosis, CME), nystatin (caveolin and clathrin independent carrier pathways), Dynasore (dynamin-dependent pathways, including CME), and amiloride (macropinocytosis)⁶⁸. HEK293 cells were then treated with Alexa Fluor 647-labeled siRNA complexed with V4-ELP-EEP13 for 2 hours. Of the four inhibitors assessed, both chlorpromazine and Dynasore resulted in a reduction of nanoparticle uptake (Fig. 4j). This finding suggests that clathrin- and dynamin-dependent endocytic pathways are central to ELP nanoparticle internalization in HEK293 cells.

Next, we assessed whether endosomal or lysosomal rupture occurs following nanoparticle uptake. HEK293 cells were labeled with LysoTracker Red, which stains acidic organelles, including late endosomes and lysosomes and is expected to diminish in intensity if compartment pH is neutralized due to membrane disruption. Indeed, treatment with V4-ELP-EEP13/siRNA-AF647 caused a gradual reduction in LysoTracker fluorescence, reaching near-complete quenching at 70 minutes post-treatment. In contrast, LysoTracker fluorescence remained stable following treatment with free siRNA-AF647, indicating V4-ELP-EEP13-dependent disruption of endosomal and lysosomal membranes (Supp. Fig. 12). Moreover, rapid binding of V4-ELP-EEP13/siRNA-AF647 to the cell membrane was observed within 5 minutes, followed by an accumulation of punctate siRNA-AF647 signals over time, suggesting robust endocytic uptake. No binding or uptake was observed with free siRNA, confirming ELP-dependent delivery. As punctate AF647 signal increased in the cytoplasm, LysoTracker fluorescence diminished with minimal colocalization, indicating membrane permeabilization and the loss of LysoTracker staining in the affected compartments. At later time points, the appearance of diffused siRNA-AF647 signals indicated successful escaped cargo (Supp. Fig. 12). Overall, these observations align with our hypothesized mechanism of endosomal escape.

ELP nanoparticles for in vitro mRNA and pDNA delivery

To further establish ELPs as a platform for therapeutic macromolecule delivery, we investigated their ability to deliver protein-encoding mRNA. The mRNA binding capacity of V4-ELP-EEP13 and V4-ELP-[EEP13]₂ was assessed through gel shift and Ribogreen assays. Cre mRNA (1335 nt) was added to V4-ELPs at varying molar ratios. As expected, higher ELP:mRNA molar ratios were required for effective complexation (~150) as compared to ELP-siRNA complexation (~25), given the larger size of the mRNA cargo (Fig. 5a–b). Similar to siRNA, mRNA complexation results in a modest increase in size (~10-20 nm) with minimal to no change in zeta potential (Fig. 5c–d). The maintained positive surface zeta potential supports that the majority mRNA is shielded within the nanoparticle core.

Figure 5 |. Complexation and delivery of mRNA by ELP nanoparticles.

a) Comparison of V4-ELP-EEP13 and V4-ELP-[EEP13]₂ Cre mRNA complexation efficiency via gel shift and b) RiboGreen assays (n = 3 technical replicates, [Cre mRNA] = 0.5 ng/μL). c) Diameter and d) zeta potential of V4-ELP-EEP13 and V4-ELP-[EEP13]₂ nanoparticles encapsulating Cre mRNA ([ELP] = 5 μM, [Cre mRNA] = 0.5 ng/μL, n = 3 technical replicates). e) % RFP+ of HEK293-RFP cells following co-treatment with V4-ELP-EEP constructs (8 μM) and mRNA-encoded Cre recombinase (1 ng/μL) for 48 hours (n = 3 biological replicates). f) Correlation between protein and mRNA-form Cre delivery efficacy by V4-ELP EEPs (30 constructs assessed). g) % RFP+ of HEK293-RFP cells following treatment with Cre mRNA complexed with V4-ELP-EEP13 and V4-ELP-[EEP13]₂ at varying ELP (constant [Cre mRNA] = 1 ng/μL) and h) Cre mRNA concentrations (constant [ELP] = 10 μM) (n = 3 biological replicates). i) % RFP+ of HEK293-RFP cells following treatment with Cre pDNA at escalating [pDNA] (n = 3 biological replicates, [ELP] = 10 μM). j) Delivery of Cas9 mRNA and k) pDNA with a chemically modified sgRNA targeting HPRT1 to HEK293 cells at varying mRNA/pDNA:sgRNA molar ratios. [mRNA] = 6 ng/μL in mRNA study, [ELP] = 20 μM in pDNA study, n = 3-4 biological replicates. l) % tdTomato+ of Ai9 lung fibroblasts and m) peritoneal macrophages following treatment with Cre mRNA encapsulated by V4-ELP-EEP13 (n = 3). *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 unless otherwise specified, two-way ANOVA with Dunnett’s or Šídák correction for multiple comparisons as appropriate (c-e, g-i). All data represented by mean ± SEM.

The ability of V4-ELP-EEP13 to facilitate mRNA delivery was investigated through the delivery of mRNA-encoded Cre recombinase to HEK293-RFP cells. To assess whether the efficacy of ELP-EEPs is cargo dependent, ELPs containing EEPs generated through three generations of iterative design were assessed for their ability to deliver Cre mRNA to HEK293-RFP cells and compared to their Cre protein delivery efficiency (Fig. 5e–f). Consistently, both V4-ELP-EEP13 and V4-ELP-EEP14 significantly outperformed V4-ELP-S10. A strong correlation was noted for protein and mRNA delivery (R = 0.9062, Fig. 5f) with EEP13 identified as most effective, despite the disparate nature of the two cargos. This suggests the existence of a common mechanistic pathway by which macromolecule endosomal escape is triggered, independent of the cargo’s physiochemical properties.

In a follow-up dose-escalation assessment of V4-ELP-EEP13, a maximum % RFP+ cells of 86% was achieved following treatment with 10 μM ELP with 1 ng/μL Cre mRNA (Fig. 5g). Consistent with the results observed with siRNA cargo, V4-ELP-[EEP13]₂ outperformed V4-ELP-EEP13 at lower ELP concentrations (1 and 2 μM) but underperformed at higher concentrations. As with siRNA cargo, this effect may be due to less efficient release of mRNA cargo within the endosome as a result of the higher nucleic acid binding affinity exhibited by dimeric constructs. Both constructs demonstrated minimal cytotoxicity (Supp. Fig. 13a). Since V4-ELP-EEP13 achieved the highest overall efficacy, it was selected for follow-up assessment. Notably, at optimal ELP doses (10 μM), V4-ELP-EEP13 enabled RFP+ expression at a level comparable to that achieved by MessengerMAX, a commercial lipid-based transfection agent optimized for mRNA delivery, with significantly less toxicity (Fig. 5h, Supp. Fig. 13b). Similarly, V4-ELP-EEP13 performed comparably to Lipofectamine in delivering plasmid DNA-encoded Cre recombinase to HEK293-RFP cells with minimal cytotoxicity (Fig. 5i, Supp. Fig. 13c). BafA1 pre-treatment to inhibit endosomal acidification significantly reduced V4-ELP-EEP13-mediated Cre mRNA delivery efficiency to HEK293-RFP cells, consistent with the mechanism that ELP-mediated delivery of both protein and mRNA cargos is dependent on endosomal acidification (Supp. Fig. 13d).

To further evaluate ENTER against state-of-the-art nucleic acid delivery vehicles, we compared its performance to SM-102 ionizable lipid-containing LNPs and PEI as representative lipid and synthetic polymeric nanoparticles. HEK293-RFP cells were treated with escalating doses of Cre mRNA doses (0.01–2.0 ng/μL) formulated with linear PEI (MW 25000 Da), SM-102 LNP (50:10:38.5:1.5 ionizable lipid:helper lipid:cholesterol:DMG-PEG2k molar ratio), or V4-ELP-EEP13. While SM-102 LNP demonstrated superior efficiency at lower mRNA doses, V4-ELP-EEP13 induced comparable levels of RFP expression at doses at or above 1 ng/μL (88-90% RFP+ in ELP-treated conditions vs 92-94% in SM-102 LNP-treated conditions, Supp. Fig. 13e). V4-ELP-EEP13 achieved significantly higher transfection efficiency compared to PEI at all mRNA doses assessed. These data support the potential of ENTER as an efficient and scalable delivery system for genetically-encoded cargo.

We next assessed the ability of V4-ELP-EEP13 to mediate mRNA- and pDNA-encoded Cas9 delivery as a representative therapeutic gene editor cargo. SpCas9 mRNA or pDNA were co-complexed with a chemically modified sgRNA targeting HPRT1 at varying mRNA/pDNA:sgRNA weight ratios. In both formats, a weight ratio of 2:1 facilitated optimal delivery efficiency to HEK293 cells, resulting in 26% and 29% editing efficiency for mRNA (20 μM V4-ELP-EEP13 with 6 ng/μL total nucleic acid) and pDNA cargo (20 μM V4-ELP-EEP13 with 20 ng/μL total nucleic acid), respectively (Fig. 5j–k). V4-ELP-EEP13 also demonstrated effective Cre mRNA delivery to primary Ai9 lung fibroblasts and macrophages in a dose-dependent manner, facilitating 54% and 56% tdTomato expression respectively at optimal dosing with minimal cytotoxicity (Fig. 5l–m, Supp. Fig. 13f–g). Altogether, these experiments demonstrate the ability of V4-ELP-EEP13 to deliver therapeutically relevant genetically encoded cargo to diverse cell types, including primary cells.

Delivery of Cre recombinase to lung epithelium in vivo

Gene editing holds great promise for the treatment of monogenic disorders such as cystic fibrosis, in which mutations in the cystic fibrosis transmembrane conduct regulator (CFTR) protein results in lifelong pulmonary dysfunction. As a proof-of-concept study for in vivo lung delivery, we evaluated ENTER’s delivery efficiency to the airway epithelium in mice using Cre recombinase as a model cargo. Ai9 mice received doses of V4-ELP-EEP13 nanoparticles (100 μM, ~24 mg/kg) co-formulated with Cre recombinase protein (50 μM, ~8 mg/kg) or Cre protein alone via intranasal instillation for three consecutive days. Lungs were harvested for image-based evaluation of tdTomato expression 5 days following the final treatment (Fig. 6a). Compared to the group treated with Cre protein alone, mice co-treated with Cre protein and V4-ELP-EEP13 displayed significantly higher percentage of tdTomato+ cells in the bronchial epithelium in both large airways (Fig. 6b, d) and small airways (Fig. 6c, e). We observed that 25.4 ± 3.7% of large airway epithelial cells and 9.2 ± 0.9% of small airway epithelial cells were tdTomato+ in the co-treatment group. Notably, editing was also observed in several important epithelial cell subtypes, including ciliated epithelial cells (α-tubulin⁺), mucus-producing goblet cells (Mucin 5AC⁺), and repopulating progenitor cell types such as club cells (CCSP⁺) and basal cells (Keratin-5⁺) (Fig. 6f). Histopathologic scoring from buffer, Cre, and combined Cre and V4-ELP-EEP13-treated mice was performed to assess the extent of any peribronchial or parenchymal inflammatory reaction (Supp. Fig. 14a–c). The observed mean response of the co-treatment cohort was 1.08 (mild) on a 0 to 4 severity scale, with focally mild lymphohistiocytic aggregates around bronchioles and adjacent alveolar parenchyma without acute inflammation.

Figure 6 |. ELP-mediated delivery of Cre protein to the bronchial epithelium in vivo.

a) Schematic of V4-ELP-EEP13-mediated Cre protein delivery study to the bronchial epithelium via intranasal instillation in Ai9 mice. b) Editing efficiency observed in large and c) small airway epithelial cells following administration of Cre protein ± V4-ELP-EEP13 (n = 3 mice, ≥ 5 large airway images/mouse, ≥ 20 small airway images/mouse). d) Representative images of large airway and e) small airways following co-treatment with Cre protein and V4-ELP-EEP13 nanoparticles. Large airway scale bars: 200 μm (left), 50 μm (right). Small airway scale bars: 50 μm. f) Representative images of edited epithelial cell types. Arrows point to tdTomato+ cells that demonstrate colocalization with the marker of interest (stained in green). Orange arrows point to cell magnified in inset image. Scale bars: 25 μm (inset scale bar 10 μm). Significance was assessed using an unpaired two-tailed t-test (b-c). Data represented by mean ± SEM.

To compare the efficacy of ENTER against LNPs in protein cargo delivery, we evaluated the ability of SM-102-containing LNPs to deliver Cre recombinase protein both in vitro and in vivo. SM-102-LNPs demonstrated lower Cre protein delivery potency with increased toxicity compared to V4-ELP-EEP13 in HEK293 cells in vitro (Supp. Fig. 15a–b) Using the same Cre protein dose and intranasal dosing scheme described above, SM-102 LNPs demonstrated minimal delivery to Ai9 mouse bronchial epithelial cells in vivo (3.9 ± 1.13% of large airway epithelial cells, 0.347 ± 0.073% of small airway epithelial cells) (Supp. Fig. 15c). This reduced efficacy may be attributed to both lower LNP uptake by bronchial epithelial cells compared to ELPs and instability of the Cre protein during microfluidic formulation with lipids. Altogether, these experiments demonstrate that ENTER can be leveraged for the intracellular delivery of therapeutic protein cargo in vivo.

Discussion

We report Elastin-based Nanoparticles for Therapeutic delivERy (ENTER), a non-viral, protein nanoparticle system as a versatile platform for effective cytosolic delivery of multiple types of biomacromolecules of therapeutic interest, including siRNA, mRNA, pDNA, protein, and RNPs. Notably, the ability to deliver diverse biomacromolecules is achieved without the need for cargo-specific vehicle modifications, whereas optimization of lipid selection and formulation ratios have been necessary to adapt LNP formulations for protein (as opposed to nucleic acid cargo) delivery.^69,70

Our fourth-generation multiblock micelle copolymer design differs in multiple ways from previous ELP systems utilized for cytosolic delivery of nucleic acids, such as pDNA^31,71,72, siRNA⁷³, and CpG⁷⁴. First, unlike the amorphous polyion complexes (PICs) formed by these systems^74–76, ENTER displays an organized micellar structure. The hydrophilic shell ensures well-controlled surface chemistry, allowing for flexibility in modification with functional modalities, including targeting moieties through sortase-mediated conjugation or genetic fusion. Second, unlike previous approaches that lack an efficient endosomal escape mechanism^73,75,76, we have discovered and implemented potent cationic endosomal escape peptides (EEP) through in silico screening of peptide databases. The core-facing cationic EEP remains shielded, facilitating the encapsulation of nucleic acid cargo in the core. Third, in contrast to previous pH-unresponsive designs^73,74,76, we rationally incorporated histidine into the ELP hydrophobic block, facilitating micelle disassembly under acidic conditions. This enables the exposure of EEP domains and cargo release only within endosomes and lysosomes and is a critical component leading to low cytotoxicity and high delivery efficiency. Finally, in addition to delivering siRNA and pDNA as demonstrated in prior studies^31,71–73, ENTER demonstrated a capacity to also deliver protein, mRNA, and gene editor ribonucleoproteins. ENTER shares similarities with other state-of-the-art delivery systems, such as lipid nanoparticles, viruses, and virus-like particles, each exhibiting organized structures with surface modules separated from core-encapsulated cargo, accompanied by a triggered release mechanism..

The effectiveness of ENTER is significantly enhanced by the identification of potent EEPs. Recombinant fusion of membrane-disruptive EEPs to ELPs offers several advantages over using untethered peptides for macromolecular delivery. One benefit is the impact of EEP multivalency on delivery potency. ELP-S10 demonstrated a 21.3-fold increase in potency in Cre protein delivery compared to S10 alone (Fig. 1h). Increased local EEP concentration afforded by ELP micelle multivalency is presumed to directly enhance membrane disruptive activity. Secondly, the fusion of ELP and EEP significantly reduced EEP-associated cytotoxicity. In conventional transfection using standalone endosomal escape peptides, high cytotoxicity related to plasma membrane disruption necessitates the removal of these peptides from the cell culture media after a brief (< 5 minute) incubation, limiting the in vivo applicability of this approach.⁷⁷ However, in the reported ELP-EEP system, prolonged peptide incubation was conducted in various cell types without significant cytotoxicity, in part because ELP nanoparticles require endocytosis and acidification-induced disassembly to expose the encapsulated EEP. The kinetics of this exposure and the specific disruption of the endosomal membrane likely contribute to substantially less cytotoxicity.

In a broader context, the seclusion of EEPs from the surface of ELP nanoparticles establishes a system that allows the evaluation of endosomolytic potential of membrane-disruptive peptides independent of cell uptake or cytotoxicity. This feature enables the discovery of more potent EEPs such as EEP13 that may not emerge in a screening of free peptides due to their cytotoxic potential—an ability crucial to surpassing the current efficiency limitations in endosomolytic peptide discovery. The identified EEPs have the potential to facilitate endosomal escape in other nanoparticle systems, both protein and non-protein based.

ENTER may overcome the scalability and immunogenicity challenges faced by existing delivery platforms. As compared to LNPs, which require complex chemical conjugation and downstream purification processes to incorporate additional functional domains, ENTER’s recombinant basis enables programmable synthesis with genetically defined precision. In comparison to viral vectors that require bespoke mammalian producer cell lines with limited scalability, here we produce ELPs via recombinant synthesis in E. coli, a simpler manufacturing approach with demonstrated scalability^78,79. The ELPs were purified using combination of phase separation- and chromatography-based methods followed by treatment to remove endotoxin; removal of endotoxin from E. coli-derived protein products is essential for avoiding innate immune responses.⁸⁰ This work demonstrates that ELP treatment is benign in terms of inflammatory response, as evidenced following both in vitro treatment of macrophages and in vivo treatment of lungs. These data aligns with prior studies, including human clinical trials evaluating therapeutic peptide-ELP fusions, where ELPs demonstrated low immunogenicity and redosability.³² Similarly, therapeutic enzymes conjugated to an ELP induced fewer antibodies as compared to their PEG-conjugated counterparts in a murine model.⁸¹ The non-inflammatory nature and low immunogenicity of ELPs represent another potential advantage of the ENTER system compared to LNPs and virus-based delivery systems.

We acknowledge several limitations and unresolved questions that warrant further investigation. This study focused on a local route for in vivo delivery and does not harness the potential of protein ligand-mediated targeting. Additional optimization of multiple design parameters is needed for achieving efficient targeted delivery through systemic administration. This includes addressing challenges, such as controlling non-specific adsorption of plasma proteins, ensuring particle stability upon systemic administration, and demonstrating in vivo organ and cell type-specific targeting. Additional optimization of the potency of ELP-mediated nucleic acid cargo delivery is also necessary, which will require the identification and use of EEPs optimized for nucleic acid delivery, among other efforts.

Methods

Plasmid cloning

DNA sequences encoding ELPs were codon optimized for E. coli expression, synthesized, and cloned into a pQE-80L backbone using commercial services from GenScript.³³ Cre and Cas9 with an LPETG tag and GGS linker were initially subcloned using previously reported genome editor coding plasmids.³⁴ Cre with variable cleavable linkers were cloned by GenScript. A constitutively active mutant of human IRF5 splice variant 5 carrying five serine to aspartic acid mutation (IRF5/4D) with a furin cleavable linker and LPETG tag was codon optimized for E. coli expression, synthesized, and cloned into pET backbone by GenScript.⁸²

Protein expression and purification

All plasmids were transformed into E. coli strain BL21(DE3) (Invitrogen, Thermo Fisher Scientific; C601003) for expression. E. coli were grown in LB medium supplemented with 100 μg/mL of ampicillin at 37 °C with continuous shaking at 225 rpm until OD₆₀₀ reached 0.6 - 0.8. For ELP expression, 1 mM of isopropyl ß-D-1-thiogalactopyranoside (IPTG) (UBPBio; P1010100) was added to induce expression over 4 hours at 37 °C. Harvested E. coli pellets were resuspended in denaturing buffer containing 8 M Urea, 1 M NaCl, and 10 mM TCEP in PBS at pH 7.0, (GoldBio; 51805459) with Protease Inhibitor Cocktail (Sigma-Aldrich; P8849). Cells were lysed by sonication on ice and cell lysates were incubated at 4 °C overnight to ensure complete solubilization of ELPs from inclusion bodies. Lysates were centrifuged at 20,000 g for 30 min and 4 °C and soluble supernatant collected. His-tag carrying ELPs were purified by Immobilized Talon Metal Affinity chromatography (IMAC, Takara Bio; 635503) under denaturing conditions following a commercial protocol. This fraction was used in high throughput screening of ELP-EEPs. For follow up assessment of V1-V3 ELPs, an additional round of inverse thermal cycling was carried out to improve purity.²¹ Briefly, ELPs were precipitated in a hot cycle by incubating with 3M NaCl at 30 °C for 50 min. Following centrifugation, ELP pellets were resolubilized in PBS with 10 mM DTT at 4 °C during the cold cycle. Non-resoluble impurities were removed by centrifugation at 20,000 g at 4 °C for 30 min. For lead V4 constructs (V4-EEP13 series), cationic exchange chromatography (HiTrap SP HP, Cytiva) was utilized as the second step of purification following the manufacturer’s protocol under denaturing conditions. All final ELP eluants were thoroughly dialyzed in deionized water, lyophilized, and stored at −20 °C until used.

For the expression of all variants of Cre and Cas9, 0.5 mM IPTG was added at OD 0.6 to 0.8. For IRF5 expression, 0.1 mM IPTG and 3% ethanol were added at OD 0.6 to 0.8 to improve the solubility of the recombinant protein. Expression was carried out overnight at 20 °C. Cell pellets were resuspended in lysis buffer containing 1 M NaCl, 20% glycerol and 10 mM TCEP in 50 mM Tris-HCl at pH 8.0, followed by sonication and IMAC purification, as described above under non-denaturing conditions. Cre and Cas9 were further purified by SP column and IRF5 by Q column following commercial protocols. Protein products were concentrated using centrifugal filtration to >100 μM. Small aliquots were snapped frozen in liquid nitrogen and stored at −80 °C until used.

Prior to primary cell treatment and in vivo experimentation, the Tris based buffer in protein cargo samples was exchanged to HBSS. Both ELP and protein cargo samples were further purified by Pierce High-Capacity Endotoxin Removal Resin (Fisher) and sterilized by passage through 0.22 μM filter. Final endotoxin levels in a typical 8 μM sample were quantified by Pyrogent-5000 kinetic turbidimetric LAL assay (Lonza) to < 5 EU/mL.

Diameter and zeta potential measurements

The diameter and zeta potential of ELP nanoparticles were evaluated using a ZetaSizer NanoSeries (Nano ZS, Malvern Instruments, UK). ELP nanoparticles were formed in 1x PBS. Dynamic light scattering measurements were recorded at a scattering angle of 173°. Samples were incubated at 37 °C for five min prior to measurement 37 °C to recapitulate cell treatment conditions, unless otherwise noted. Three measurements with at least 10 runs were recorded for each sample.

TEM imaging

TEM grids were prepared by adding 5 μL V4-ELP in HBSS (50 μM) was adsorbed for 1 minute onto a glow-discharged, 200 mesh carbon Type-B copper grid with ~97 μm grid hole size (Ted Pella). Excess liquid was removed with filter paper, then the grid was floated briefly on a drop of water, blotted again on filtered paper, then stained with 0.75% uranyl formate for 30 seconds. After removing excess stain, grids were examined using a Tecnai T12 TEM microscope (Thermo Scientific).

ELP-RNA nanoparticle formation

To form ELP-RNA nanoparticles, siRNA or mRNA was added to ELPs in HBSS at varying molar ratios. An ELP stock concentration of 200 μM was used unless otherwise indicated. For siRNA particle formation, nanoparticles were incubated at RT with continuous shaking at 500 rpm for 30 min. siRNAs used in the described studies include a GFP-targeted siRNA (green fluorescent protein) (Thermo Fisher Scientific, AM4626), AF647-labeled siRNA (Qiagen, 1027295), and a scrambled siRNA (Applied Biosystems, Thermo Fisher Scientific; 4390843). For mRNA particle formation, the nanoparticles were first incubated on ice with continuous shaking at 500 rpm for 30 min, followed by a 30 min RT incubation at 500 rpm. 5-Methoxyuridine-modified Cre and Cas9 mRNA (TriLink) were used.

Measurement of ELP-RNA encapsulation efficiency

mRNA encapsulation efficiency of each ELP formulation was calculated using the Quant-iT RiboGreen (Thermo Fisher Scientific) or electrophoretic mobility gel shift assay. For electrophoretic mobility gel shift assays, ELP-RNA nanoparticles were formed as described and run in a 1% agarose gel at 100 V for 30 min. For the RiboGreen assay, ELP-RNA nanoparticles were formed in 1x PBS at varying molar ratios and diluted 1:1 in 1x TE buffer containing the fluorescent RiboGreen reagent. ELP-RNA nanoparticles, RNA standards, and ELP standards were plated in triplicate then incubated with the RiboGreen reagent at 37 °C for 5 min. Fluorescence intensity was read on a SpectraMAX iD5 microplate reader at an excitation wavelength of 485 nm and an emission wavelength of 535 nm. Encapsulation efficiency was calculated by normalizing ELP-RNA nanoparticle fluorescence by subtracting baseline ELP fluorescence, and using the RNA standard curve to convert fluorescence values to estimated free RNA concentration.

$$\text{Encapsulation efficiency}(\%)=(1-[\mathrm{Free\; mRNA}]/[\mathrm{Total\; mRNA}])\times 100.$$

In vitro siRNA delivery

HeLa cells for assessment of AF647-labeled siRNA uptake or HeLa-d2eGFP cells for evaluation of GFP knockdown were seeded in 96 well-plates at 10,000 cells/well and incubated overnight. siRNA-containing ELPs were used to treat cells for 24 hours. GFP expression and siRNA uptake were evaluated by flow cytometry. In nanoparticle uptake studies, flow cytometry assessment was conducted in 0.2% v/v trypan blue to quench signal from cell membrane-associated siRNA. Cell viability was monitored by Zombie Aqua live/dead cell staining (Thermo Fisher Scientific) prior to flow analysis. RNAiMAX (Thermo Fisher Scientific) was used as a positive control. Gating scheme is shown in Supplementary Figure 16.

In vitro Cre and Cas9 mRNA, pDNA and protein delivery

HEK293-RFP cells were seeded in 96 well plates at 20,000 cells/well and incubated overnight. Cre pDNA (pCAG-Cre, Addgene Plasmid #13775) or Cre mRNA (Trilink) containing ELPs, or Cre protein-ELP nanoparticles were used to treat cells for 48 hours (mRNA) or 72 hours (pDNA) and RFP expression evaluated by flow cytometry. All ELP transfections were conducted in full serum media. Transfections using untethered peptides were conducted in either full serum media or using a 15-minute serum-free incubation step followed by addition of media. Cell viability was monitored by Zombie Aqua live/dead cell staining (Thermo Fisher Scientific) prior to flow analysis. Gating scheme is shown in Supplementary Figure 16.

For mRNA transfection studies, MessengerMAX (Thermo Fisher Scientific) was used as a positive control. In studies interrogating the effect of endosomal acidification on ELP-mediated Cre mRNA and protein delivery efficacy, cells were pretreated for 1 hour in media containing 50 nM Bafilomycin A1 (Thermo Fisher Scientific); Bafilomycin A1-containing media was replaced with fresh media immediately prior to nanoparticle addition. For LNP transfection studies, lipids (SM-102, 1,2-DSPC, cholesterol, and DMG-PEG2k at a 50:10:38.5:1.5 molar ratio) dissolved in ethanol were formulated with Cre mRNA or protein dissolved in 10 mM acetate buffer through microfluidic mixing via a NanoAssemblr (Cytiva). Lipids and mRNA or protein were formulated at a weight ratio of 10:1 lipid:mRNA, at a 1:3 ethanol:aqueous volume ratio. For PEI transfection studies, 0.25 mg/mL of linear MW 25000 Da polyethylenimine (PEI) (Thermo Fisher Scientific) was complexed with Cre mRNA at 2:1 PEI:RNA weight ratio via pipette mixing.

For Cas9 delivery studies, HEK293-RFP cells were treated with ELP nanoparticles containing either (1) pDNA (CAG-Cas9, Addgene Plasmid #89995) or mRNA-encoded SpCas9 (Trilink) and a chemically modified sgRNA targeting HPRT1 (Synthego), or (2) co-formulated with a True SpCas9 (Thermo Fisher Scientific) RNP complexed with the identical HPRT1 sgRNA. For ABE8e RNP delivery studies, HEK293-RFP cells were treated with ELP nanoparticles co-formulated with ABE8e protein (Aldevron) complexed with an sgRNA targeting HEK3. Genomic DNA was isolated from treated cells after 72 hours and the target locus amplified by PCR and sequenced via Sanger sequencing. Editing efficiency was quantified using TIDE analysis (Cas9)⁸³ or BEAT analysis (ABE8e).⁸⁴

EEP design

A multiple linear regression model was trained on a dataset derived from a library screen of membrane disruptive peptides tested for their ability to facilitate GFP protein delivery to HeLa cells in vitro⁵⁵. The Pearson correlation coefficient r between the values for each feature and the peptide efficacy was calculated and used to calculate the F score for each feature. F scores were used to rank features according to their predictive value, where a higher score indicates a stronger correlation to peptide efficacy. Regression models using the top k features (from k = 1 to k = 15) were trained and evaluated by leave-one-out cross validation. For each value of k, this process was repeated with one member of the dataset held out for testing in each iteration, and the average root mean square error (RMSE) of all iterations calculated. The model with the lowest RMSE was used for peptide library screening.

The Alpha Helix Database (AHDB) screened to identify enhanced EEPs was constructed by combining peptides from the: (1) Antimicrobial Peptide Database (APD)⁵⁶, (2) Database of Antimicrobial Activity and Structure of Peptides (DBAASP)⁵⁷, and (3) Therapeutic Peptide Design database (TP-DB)⁵⁸. Filters were applied to select peptides labeled as helical from the APD (491 peptides) and without chemical modifications from the DBAASP (5,836 peptides), as well as those peptides likely to remain stable as helices when isolated from a parent protein in the TP-DB. Such peptides defined as having 1) a length greater than 10 residues, 2) a helical propensity greater than 1 standard deviation above the mean, and 3) a contact number less than 1 standard deviation below the mean (8,956 peptides).⁵⁸ After removing duplicates, peptides with non-canonical amino acids, and peptides without cationic residues (Lys or Arg), the AHDB contained a total of 11,084 peptides. The predicted efficacy of each peptide was calculated. A library of dimers was formed by combining peptide monomers with predicted efficacies >10% (dimers were composed of two sequences linked by a GGSGGGS linker). From these dimers, the first generation of 7 EEP sequences to be recombinantly incorporated into ELP constructs for in vitro screening were manually identified. Selected dimeric peptides were those with a predicted efficacy greater than the most effective peptide in the training dataset, composed of sequences of non-microbial origin, and < 50 residues in length.

A second generation of EEPs with cationic residues enriched in lysine over arginine were generated through a second in silico screen. Using a dataset derived from a screen of V4-ELPs for siRNA delivery, a second linear regression model was trained, which was used to rescreen the compiled α-helical database. To retain key insights derived from the first model and improve diversification, additional structural and physicochemical feature filters were applied during the screening process (Supp. Fig. 5c). The top 8 peptides from this screen were selected for in vitro testing.

Primary cell isolation and culture

All primary cells were isolated from adult male SauSpyAi9 reporter mouse between 8 to 12 weeks of age. Murine pulmonary fibroblasts were isolated as previously described.⁸⁵ The lung was surgically removed, minced to 1 mm pieces, and digested with Liberase (Sigma) at 37 °C for 30 min. After neutralization and wash, digested tissue fragments were plated in 10 cm tissue culture plates in DMEM/F12 media with 15% FBS, 1X antibiotic/antimycotic for 1 week to allow fibroblast egression.

Elicited murine peritoneal macrophages were lavaged from the peritoneal cavity 4 days following intraperitoneal injection of thioglycolate as previously described.⁸⁶ Macrophages were enriched by plate adherence and unattached cells were removed. Macrophages were cultured in complete Dulbecco’s Modified Eagle Media (DMEM) supplemented with 10% FBS (Hyclone, GE Healthcare, IL). All treatments were conducted within 1 week of culture.

Murine CD4+ T cells were isolated as previously described.⁸⁷ Spleens were harvested and red blood cells were lysed by osmotic shock, and single-cell suspensions were collected using a 70 μm cell strainer. CD4+ T cells were isolated by negative selection using a T cell isolation kit (Miltenyi Biotec) and stimulated with plate-bound CD3 antibody (10 μg/mL, #553057, BD Biosciences) and soluble CD28 antibody (5 μg/mL, #553294, BD Biosciences) in RPMI medium supplemented with 10% FBS, 1% penicillin/streptomycin, 50 μM 2-mercaptoethanol, and 50 U/mL IL-2 for 4 days before treatment.

Murine bone marrow Sca1+ progenitor cells were isolated and cultured as previously described.⁸⁸ Briefly, the tibia, femur and iliac crests were dissected from reporter mice and crushed by mortar and pestle to release bone marrow cells. Sca1+ cells were isolated using a Sca1 enrichment kit (Miltenyi Biotech). Sca1 enriched cells were cultured in serum free StemSpanTM SFEM II medium (Stemcell technologies) supplied with mouse SCF (50 ng/mL), mouse TPO (50 ng/mL), mouse Flt3L (50 ng/mL) and human IL-11 (50 ng/mL) (Peprotech) for 3 days before treatment.

In vitro primary cell transfection

Murine protocols were approved by the Institutional Animal Care and Use Committee of Beth Israel Deaconess Medical Center. A total of 10,000 fibroblast cells/well, 100,000 macrophage cells/well, 25,000 CD4+ T cells/well, and 50,000 Sca-1 enriched bone marrow cells/well were seeded in 96 well plates in 50 μL of their corresponding full medium with serum. A total of 50 μL of HBSS containing ELP and Cre protein or mRNA cargo were added to achieve the final concentration as indicated, and incubated for 48 hours before harvesting, unless otherwise specified. For protein delivery studies with macrophages, T cells, and Sca-1 enriched bone marrow cells, a serum free pre-incubation phase was used to improve efficiency. For these experiments, cells were pre-treated with protein cargo and ELP in HBSS for 15 min before adding 50 μL of the corresponding full medium. Editing efficiency was typically determined in the bulk cell population. For HSPC studies, Sca-1+ enrich bone marrow cells were stained with an eFluor 450 conjugated hematopoietic lineage antibody cocktail (#50-112-9718, Fisher Scientific), anti-CD117-APC (#105811, Biolegend) or anti-Sca-1-FITC (#15238139, Fisher Scientific), and the LSK (Lin-Sca-1+ c-Kit+) HSPC population examined by flow cytometry. Gating scheme is shown in Supplementary Figure 16.

Intranasal administration to Ai9 mice

Murine protocols were approved by the Institutional Animal Care and Use Committee of Beth Israel Deaconess Medical Center (Protocol # 052-2022). V4-ELP-EEP13 (100 μM, ~24 mg/kg) was formulated with Cre protein (50 μM, ~8 mg/kg) in HBSS and administered to 8-week-old male Ai9 mice via intranasal instillation of a 50 μL volume. Mice were anesthetized with ketamine by intraperitoneal injection. Mice were treated once daily for 3 days. Lungs were harvested for cryopreservation and sectioning 5 days after the last treatment.

Assessment of tdTomato+ epithelial cells in mouse airways

Lung tissues were fixed with 4% formaldehyde by tracheal instillation and cryopreserved. The tissue was placed ventral side down for sectioning. After reaching a depth of sectioning at which all lung lobes were visible, a total of 16 sections (8 μm thick) were collected at 4 levels, spaced ~112 μm apart (4 sections per level, 2 sections per slide). One slide at each level was stained by H&E for histopathologic assessment. The other slide was systemically imaged for quantification of tdTomato expression in the bronchial epithelium. Airways were defined as large or small airways based on size and anatomical position. The left and right main bronchi were categorized as large airways. Small bronchiole airways were defined as airways with a long axis diameter < 400 μm. All small airways imaged had lumen areas < 150,000 μm² (mean area 48,000 μm²) as determined using an ellipsoid approximation (Area = Pi × Length × Width). Small and large airway views to be imaged were identified through evaluation of the DAPI-stained nuclei pattern alone to prevent biased selections. A total of 5 randomly selected small airways and 2 large airway views (one of each main bronchi) were imaged per slide. The total number of cells in the visualized epithelial layer was assessed through counting the DAPI-stained nuclei in the bronchial epithelial layer. The total number of tdTomato+ bronchial epithelial cells were also counted. Editing efficiency was defined as the number of tdTomato+ cells/number of nuclei × 100.

For immunostaining, additional sections were blocked in blocking buffer (1x PBS, 5% goat serum, 0.3% Triton X-100) for 1 hour. They were then separately incubated overnight with primary antibodies in dilution buffer (1x PBS, 1% BSA, 0.3% Triton X-100) at the following ratios: 1:200 for anti-acetyl-α-tubulin antibody (#5335, Cell Signaling), 1:50 for anti-Cytokeratin 5 antibody (#ab193895, Abcam), 1:100 for anti-Mucin-5AC antibody (#MA5-12178, Fisher Scientific), and 1:100 for anti-Clara Cell Secretory Protein antibody (#07-623, Millipore Sigma). After washing, sections were incubated with goat anti-rabbit (#4414, Cell Signaling) or goat anti-mouse (#4410, Cell Signaling) secondary antibodies conjugated with Alexa Fluor 647 at a ratio of 1:1000 in dilution buffer for 1 hour. Subsequently, sections were mounted in SlowFade Diamond antifade Mountant with DAPI (S36964, Fisher Scientific) and imaged using a Zeiss LSM 980 confocal microscope with a 40x water immersion objective lens.

Histological evaluation of mouse lung tissues

Four spaced 8 μm tissue sections were collected per mouse and stained by H&E as described above. Slides containing the sections were assessed for presence of (1) peribronchial and parenchymal immune cell infiltrates; and (2) alveolar thickening by a blinded board-certified pathologist. Sections were assigned a score using a previously described scoring system, in which a score of 0 indicates no evidence of immune cell infiltration or alveolar thickening, while a score of 4 indicates diffuse and severe infiltration and alveolar thickening (see Supp. Fig. 13 for full description of scoring system and representative images for each level observed).⁸⁹ Images were taken using an Olympus DP23 microscope camera.

Statistical analysis

Statistical analyses were performed in GraphPad Prism 10 using unpaired t-tests or ANOVA with corrections for multiple comparisons as appropriate. All values were expressed as mean ± standard error.

Extended Data

Extended Data Figure 1 |. Formation of ELP-protein nanoparticles.

a) PAGE gel image depicting result of sortase-mediated Cre-LPETG conjugation to V1-ELP at varying ratios. The reaction solution underwent centrifugal filtration against a 100 kDa filter to remove unreacted Cre. b) Quantification of Cre-V1-ELP conjugation efficiency via densitometry analysis, measured as percentage of total ELP conjugated to Cre. c) Diameter of Cre-V1-ELP nanoparticles (n = 3 technical replicates). d) PAGE gel image depicting result of sortase-mediated SpCas9-LPETG conjugation to V1-ELP at varying ratios. e) Diameter and f) Zeta potential of SpCas9-V1-ELP nanoparticles (n = 3 technical replicates). g) PAGE gel image and h) Densitometry analysis of CD117 mAb-V1-ELP conjugation efficiency. i) MFI and j) Fluorescent images of CD117+ P815 mast cells treated with CD117 mAb-functionalized V1-ELP nanoparticles compared to unfunctionalized controls (n = 3). Scalebars = 10 μm. For all experiments, [V1-ELP] = 8 μM unless otherwise noted. Significance assessed via one-way ANOVA with Dunnett’s correction for multiple comparisons. Data represented by mean ± SEM.

Data availability

Data used for training EEP predictive models and the compiled alpha helical peptide database are available on Github (https://github.com/sayoeweje/elp-eep-discovery). Source data for all main body and extended data figures are provided with this paper. All additional data are available upon request.

Code availability

All scripts and data used for EEP design can be found on GitHub at https://github.com/sayoeweje/elp-eep-discovery.⁹⁰