Peptides as surface coatings of nanoparticles that penetrate human cystic fibrosis sputum and uniformly distribute in vivo following pulmonary delivery

Abstract

Therapeutic delivery of drug and gene delivery systems have to traverse multiple biological barriers to achieve efficacy. Mucosal administration, such as pulmonary delivery in cystic fibrosis (CF) disease, remains a significant challenge due to concentrated viscoelastic mucus, which prevents drugs and particles from penetrating the mucus barrier. To address this problem, we used combinatorial peptide-presenting phage libraries and next-generation sequencing (NGS) to identify hydrophilic, net-neutral charged peptide coatings that enable penetration of human CF mucus ex vivo with ~600-fold better penetration than control, improve uptake into lung epithelial cells compared to uncoated or PEGylated-nanoparticles, and exhibit enhanced uniform distribution and retention in the mouse lung airways. These peptide coatings address multiple delivery barriers and effectively serve as excellent alternatives to standard PEG surface chemistries to achieve mucus penetration and address some of the challenges encountered using these chemistries. This biomolecule-based strategy can address multiple delivery barriers and hold promise to advance efficacy of therapeutics for diseases like CF.

Introduction

For successful treatment of mucosal-associated diseases, such as cystic fibrosis (CF), asthma, and chronic obstructive pulmonary disease, it is critical for drug and gene therapies to cross several biological barriers to enter the target cells at therapeutic concentrations. One of the initial and primary barriers is the protective mucus layer lining the epithelia in the eyes, airways, gastrointestinal and cervicovaginal tracts [1]. Mucus is a complex biopolymer composed mainly of water, mucin glycoproteins, lipids, DNA, non-mucin proteins, salts, and cell debris [1]. The mucus layer acts as a transport barrier and can greatly hinder the diffusion of drugs, particles, and other molecules via size filtering, mucociliary clearance, and intermolecular interactions, including electrostatic, hydrophobic, and/or other specific binding interactions [2]. In CF, the hyperconcentrated mucus layer traps foreign pathogens, resulting in chronic bacterial infections and concomitantly prevents up to 91% of model drugs and 99% of polymeric nanoparticles from penetration needed for successful therapy [3, 4]. As a result, it is a long-standing challenge for drugs to permeate the mucus barrier and reach the underlying epithelia.

It has been demonstrated that the use of net-neutral charge hydrophilic polymers can improve the transport of drugs and drug carriers and minimize interactions with respiratory mucus [5, 6], gastrointestinal mucus [7, 8], and cervical mucus [9]. However, it has been estimated that only 35% of 200 nm nanoparticles functionalized with poly(ethylene glycol) (PEG) 3.4 kDa are capable of penetrating through a 10 μm CF sputum layer within 20 min [6]. There have also been numerous clinical studies that have demonstrated that people possess pre-existing anti-PEG antibodies [10, 11] or develop an immune response after repeated administration of PEGylated proteins [12] and nanoparticles [13]. As a result, there are potential challenges for the safety and feasibility of PEG carriers prior to their use for mucus penetrating delivery, and there is a need to find alternative mucus-penetrating chemistries.

After mucus penetration, drug delivery systems must enter target cells to efficiently deliver their therapeutic payload. Although PEGylated systems demonstrate improved diffusion in mucus via decreased hydrophobic and electrostatic interactions, PEGylation can hinder cellular uptake of the drug carrier system [14]. In genetic diseases such as CF, it is essential that delivery systems can deliver small molecules or nucleic acid therapeutics intracellularly. Current strategies mostly focus on either mucus penetration or cell penetration, but there has not been a single surface chemistry that addresses both mucus and cellular barriers. However, in nature, viruses have evolved to possess properties for favorable interactions for mucus transport [15]. Complex virus-like particles that are hydrophilic and display highly densely charged, alternating cationic and anionic amino acids on their surface have demonstrated unhindered, diffusive transport through mucus comparable to saline [15]. It has also been demonstrated that peptides with asymmetric charge properties had improved transport in reconstituted mucin [16]. Collectively, these results suggest that biomolecules have diverse physicochemical properties (i.e. charge, hydrophilicity, their spatial distribution), which can be leveraged to achieve mucus-penetrating transport.

In particular, bacteriophage are promising biological molecules that can be engineered to provide a large diversity of physicochemical properties that can be screened to identify new surface coatings that overcome the mucosal and cellular barriers. Bacteriophage (phage), or viruses that infect bacteria, have been engineered as display systems to express combinatorial libraries of up to 10⁹ different peptides or proteins on their viral coat proteins (i.e. different peptide per phage). These phage libraries effectively function as a collection of surface chemistries with diverse physicochemical properties and functionalities [17]. These peptide-presenting phage libraries can be used for high-throughput, iterative screening against a variety targets to identify peptide coatings that have favorable interactions with mucus for penetration and other desired functionalities. By performing repetitive rounds of selection against the target, or selection pressure, it is possible to collapse the highly diverse libraries to a few leads (i.e. biopanning) [18].

Here, we screened peptide-presenting phage libraries against a mucus barrier, used NGS to identify the genetically-encoded peptides from a vast sequence space, and then validated their mucus-penetration ex vivo using clinical samples from the sputum of cystic fibrosis patients. In CF, an increase in mucus concentration and reduced mucociliary clearance creates a physical barrier that traps pathogens and drastically decreases drug diffusion, absorption, bioavailability, and consequently limiting therapeutic outcomes of mucosal drug delivery systems [19]. In this study, we present a general strategy to identify mucus-penetrating peptides, validate identified peptides in CF sputum samples from patients, and confirm the cellular uptake of nanoparticles functionalized with mucus-penetrating peptides. In addition, we evaluated the distribution and retention of peptide coated nanoparticles in vivo in the mouse lungs compared to controls following inhalation. This biomolecule-based approach can be translated to therapeutic delivery strategies of drug and gene carriers across mucosal barriers and the dense mucus barrier present in disease states like cystic fibrosis, chronic pulmonary obstructive disease, and asthma.

Materials and Methods

Library construction.

A combinatorial library of cyclic heptapeptides flanked by a pair of cysteine residues displayed on T7Select415–1 phage (Novagen, WI; catalog number 70015) was constructed according to the manufacturer’s protocol. Briefly, degenerate NNK-oligonucleotides (Supplementary Table 1) encoding a library of heptapeptides with the general structure CX7C, were synthesized and obtained from Integrated DNA Technologies (IDT, IL). To create the library insert, oligonucleotides (primers 1 and 2, Supplementary Table 1) were annealed, extended with DNA polymerase I Klenow fragment (NEB, MA; catalog number M0210S), double digested with HindIII-HF/EcoRI-HF (NEB, MA; catalog numbers R3104S and R3101S), and cloned into the T7Select 415–1 HindIII/EcoRI vector arms (Novagen, WI product number RC0135) to enable peptide display on the 415 copies of the T7 gp10A phage capsid protein as a C-terminal fusion. Next, the peptide-displayed phage library was amplified once in liquid culture with BL21 E. coli (Novagen, WI; product number RC0131). The resulting lysate was clarified by centrifugation at 8,000 × g and stored long-term in 0.1 volume sterile 80% glycerol at −80°C. The constructed T7 phage CX7C library had a diversity (i.e. number of phage displaying unique peptide sequences) of 3.7×10⁵ distinct clones, as determined by NGS (Supplementary Table 2) and a phage concentration of 1.24 × 10¹¹ plaque forming units per mL (pfu/mL), as determined by standard double-layer plaque assay.

Phage selection in CF mucus.

The generated T7 phage CX7C library was iteratively screened against a cystic fibrosis mucus model (CF mucus model) in a 3.0 μm polyester membrane 24-well transwell system (Corning, MA; catalog number 3472) to identify mucus-penetrating peptides. The CF mucus model was prepared according to literature reports of composition to model sputum of CF patients [20]. The model is an aqueous mixture of 40 mg/mL mucin from porcine stomach type III (Sigma Aldrich, MO; catalog number M1778), 1.4 mg/mL herring sperm DNA (Promega, WI; catalog number D1811), 5 mg/mL egg yolk from chicken (Sigma Aldrich, MO; catalog number E0625), 0.9 mg/mL lactoferrin human (Sigma Aldrich, MO; catalog number L4040), 5 mg/mL sodium chloride (Thermo Fisher Scientific, MA; catalog number S271–500), and 2.2 mg/mL potassium chloride (Thermo Fisher Scientific, MA; catalog number P217–500) to give the concentration of ions found in CF sputum. The solution was adjusted with 0.5 M hydrogen chloride (Thermo Fisher Scientific, MA; catalog number AC423795000) to pH 6.5, which is the estimated pH of CF airway mucus. The reagents were transferred to a sterile tube, mixed for 2 h in a tube rotator at 5 rpm (Thermo Fisher Scientific, MA), and used within 4 h.

To set up the transwell system, the basolateral compartment (receiver) in the transwell was filled with 600 μL of phosphate buffered saline (PBS; Corning, MA; catalog number 21–040-CV), and the apical compartment (donor) was filled with 100 μL of CF mucus. For the first round of selection, an initial phage amount of 4.2 × 10⁹ phage plaque forming units (pfu) was added on top of the mucus layer in the donor compartment and incubated for 1 h at room temperature (25°C). At 5, 10, 15, 30, and 45 minutes, 10 μL aliquots of the eluates were taken for further quantification. After 1 h, the entire eluate was collected from the basolateral side and titered using standard double-layer plaque assay to quantify phage concentration. The eluted phage library was amplified in BL21 E. coli as described before, which was then quantified by plaque assay prior to the next round of selection. Two subsequent rounds of iterative selection were performed with initial amounts of 5.2 × 10⁶ pfu and 4.2 × 10⁹ pfu, respectively.

Next-generation sequencing of phage libraries selected against CF mucus

Library preparation for NGS.

After the third round of selection against the CF mucus model, the eluates from 15, 30 and 60 minutes from each round and replicate were amplified in BL21 E. coli. In addition, the naïve library (i.e. starting library) was amplified over three rounds without selection for further use as controls to account for bias due to amplification growth. Next, the amplified phage eluates were prepared for NGS following a library preparation workflow to the manufacturer’s protocol (Illumina, CA); the naïve library was also similarly prepared in order to determine the diversity of the CX7C library (i.e. how many unique clones). Samples were heat denatured at 100°C for 15 minutes in a heated dry bath (Thermo Fisher Scientific, MA), and their DNA was amplified by PCR using 2x KAPA HiFi HotStart ReadyMix (KAPA Biosystems, MA; catalog number KK2601) with 2.5 μL of phage DNA template and 10 μL of 1μM primers (3 and 4, see Supplementary Table 1). The PCR conditions were carried out as recommended by the manufacturer. The obtained PCR amplicons were purified with AMPure XP beads (Beckman Coulter, IN; catalog number A63881), and size was confirmed by gel electrophoresis in a 2% agarose gel. Next, sequencing adaptors and unique dual barcodes combinations were attached to each library PCR amplicons using the Nextera XT Index Kit primers (index 1 N7XX and index 2 S5XX, Illumina, CA; catalog number FC-131–1001), and the amplicon size and concentration were confirmed using a Bioanalyzer (Agilent) and Nanodrop (Thermo Fisher Scientific, MA), respectively. Samples were diluted to the final concentration of 40 nM in 10 mM Tris buffer, pH 8.5. One microliter aliquots of each sample diluted DNA with unique indices (i.e. barcodes) were pooled together and submitted for sequencing to The University of Texas at Austin GSAF sequencing facility on a MiSeq system to run single reads of 300 nucleotides.

Translating sequencing reads to peptides and frequency counts.

Sequences were automatically demultiplexed by MiSeq Reporter software and the obtained FASTQ files were analyzed by FastQC version 0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) for quality control checks on raw sequence data. Next, sequences were converted from fastq to fasta format using FASTX-Toolkit fastq_to_fasta tool version 0.0.13.2 (http://hannonlab.cshl.edu/fastx_toolkit/index.html), and the resulting fasta files were processed using a custom Bash script to find the variable library region by aligning to sequences flanking both sides of the displayed CX7C peptide (available upon request). For each sample, sequences corresponding to the library insert region (which includes DNA encoding for the constrained cysteines) were transcribed and translated into amino acids using a custom Python script, version 2.7.12 (available upon request). Then, sequence reads encoding unique peptide sequences had their frequency count calculated and sorted according to their abundance in each sample. An asterisk (*) indicates the presence of a stop codon. Sequences were disregarded in subsequent analysis if they contained a premature stop codon in the first three amino acid positions in the variable region. An overview of the number of reads that passed each processing step can be found in Supplementary Table 2. For each sample, the top 20 most abundant peptide sequences were compiled, and their physicochemical properties were determined in silico using Protein Calculator v3.4 (The Scripps Research Institute, La Jolla, CA), and Kyte-Doolittle to measure the grand average of hydropathy (GRAVY), a score indicative of the hydrophobicity of a sequence (i.e. negative GRAVY score values are indicative of hydrophilic peptide sequences whereas positive GRAVY score values indicate hydrophobic peptide sequences) [21]. Sequence database analysis was performed using Basic Local Alignment Search Tool (BLASTp, U.S. National Library of Medicine, Bethesda, MD). In addition, the selected peptide sequences were compared to databases encompassing previously isolated peptides known as “target-unrelated peptides” (TUP) motifs [22, 23]. These databases have identified a large set of sequences with undesired proliferation advantages and false positive target-unrelated peptides. Peptide sequences that appeared in the database search were excluded from subsequent analysis.

Cloning of selected sequences into T7 phage vector.

Oligonucleotides encoding the peptides of interest were designed and obtained from IDT (Supplementary Table 1). Primers (1 and 5–13, Supplementary Table 1) were diluted to 50 μM in TE buffer, annealed, extended with DNA polymerase I Klenow fragment (NEB, MA), double digested with HindIII-HF/EcoRI-HF (NEB, MA), and cloned into the T7Select 415–1 HindIII/EcoRI vector arms according to the manufacturer’s protocol (Novagen, WI). Individual phage plaques (i.e. areas of bacterial cell lysis on overlaid agar plates from infection by individual phage clones) were selected from titer plates and grown in liquid culture of BL21 E. coli (Novagen, WI) to amplify the amount of phage. To confirm cloning, DNA of the individual clones were PCR amplified for 20 cycles following the manufacturer’s temperature settings with T7 primers (14 and 15, Supplementary Table 1), purified, and sequences encoding for selected peptides were confirmed by Sanger DNA sequencing. To then obtain sufficient amounts of phage needed for subsequent studies, the peptide-presenting phage clones were amplified in liquid culture with BL21 E. coli and purified twice by precipitation with sterile 50% PEG-8000 (Thermo Fisher Scientific, MA; catalog number BP233–100) on ice followed by centrifugation at 10,000 rpm.

CF sputum sample collection.

Spontaneously expectorated sputum samples were collected from patients at the CF clinic at Dell Children’s Medical Center of Central Texas Cystic Fibrosis Center, Dell Children’s Pulmonology Clinic, Seton Healthcare Family (n = 21) following protocols approved by the Institutional Review Board of the University of Texas at Austin under study number 2016-03-0104. Informed consent was obtained from all study participants. Samples were stored at −20 to −80°C immediately after collection and thawed on ice for subsequent experiments. Age, gender, and any inhaled medications taken by the patients were recorded. Sputum samples with visible amounts of saliva were excluded from experiments.

Bulk diffusion of T7 phage clones.

Diffusion studies of peptide-presenting phage clones were performed in a 3.0 μm polyester membrane 24-well transwell across CF mucus or CF sputum from patients at room temperature (25°C) to maintain consistency and avoid fluctuation of temperature and humidity during the experiments. Briefly, 100 μL of CF mucus or CF sputum was transferred to the donor compartment in the transwell system containing 0.6 mL PBS in the receiver compartment. An initial phage concentration of 2*10⁹ genomic copies (gc) was added to the donor compartment on top of the CF mucus or CF sputum layer. At 15, 30, 45, 60, and 120 minutes, the entire eluate was collected from the receiver compartment, which was then replenished with fresh PBS. Collected eluates were quantified by quantitative PCR (qPCR) as previously described [24] to determine phage concentration (in genomic copies (gc)) using the forward and reverse qPCR primers (16 and 17, Supplementary Table 1). Standard curves were generated with serial dilutions of T7Select 415–1b packaging control DNA (Novagen, WI; product number RC0132) for each run. The data presented is the average of three independent experiments.

Functionalization of nanoparticles with mucus-penetrating peptides.

Red (excitation/emission 580/605 nm) or yellow-green (excitation/emission 505/515 nm) fluorescent 100-nm diameter carboxylate-modified polystyrene (PS) nanoparticles (Molecular Probes FluoSpheres®; Thermo Fisher Scientific, MA; catalog numbers F8801 and F8803) were conjugated with either methoxy-PEG-amine 1KDa (Alfa Aesar, MA; catalog number 46643) or custom synthesized peptides (LifeTein, NJ) via covalent attachment to COOH groups on the particles via standard EDC chemistry using 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (Thermo Fisher Scientific, MA; catalog number 22980) and N-hydroxysulfosuccinimide sodium salt (Thermo Fisher Scientific, MA; catalog number 24510) in 50mM MES buffer (pH 6.0; Alfa Aesar, MA; catalog number J60763), according to the manufacturer protocol. A chemical reaction scheme is depicted in Supplementary Figure S3. Peptides were synthesized with the exact same peptide sequence and cyclic configuration (C-C constrained with a disulfide bridge at positions 1–9) as the phage-displayed peptides. However, when having a stop codon, synthetic peptide sequences were truncated to better recapitulate phage-displayed peptides that presented a stop codon in their sequence. Stop codons signal the termination of protein translation [25, 26]. Therefore, the presence of a premature stop codon will terminate the translation of the peptide, resulting in a shortened peptide. The molar ratio for conjugation between COOH groups and mPEG-amine 1KDa or peptides was 1:5. After conjugation, nanoparticles were dialyzed for 24 hours against 1X PBS using 100 kDa MWCO Float-A-Lyzer G2 dialysis membranes (Spectrum Labs, MA; catalog number G235035). Freshly prepared nanoparticle dispersions were characterized by dynamic light scattering (DLS) (Zetasizer Nano, Malvern Instruments, MA). Size and zeta potential measurements were performed in water or PBS at 25°C and BEGM cell culture media at 37°C, at a scattering angle of 173°.

The coating density of peptides conjugated per nanoparticle was quantified by High Performance Liquid Chromatography (HPLC) and 1-pyrenylyldiazomethane (PDAM) assay as previously described [27]. The unconjugated peptides were removed from the conjugated nanoparticles by ultrafiltration with an Amicon Ultra-0.5 mL filter device MWCO 100 kDa (Millipore, MA; catalog number UFC510024). The nanoparticles were washed three times with PBS, the washes containing unconjugated peptides were collected and analyzed by HPLC using an Agilent Technologies 1260 Infinity II LC Bio Inert system with a XSelect Peptide HSS T3, 100 Å, 5 μm, 4.6 × 150 mm column (Waters Corporation, MA; catalog number 186004791) set for detection at 220 nm wavelength. A standard curve was generated using known concentrations of peptides. The density of peptides conjugated to the nanoparticles was calculated by subtracting the unconjugated amount from the amount of peptide added initially to the nanoparticles. Samples were tested in triplicate of at least three independent experiments. The residual carboxylic groups present on the conjugated polystyrene nanoparticles were quantified using 1-pyrenylyldiazomethane (PDAM; Thermo Fisher Scientific, MA; catalog number P1405), as previously described [27]. Briefly, 1 μL of the conjugated nanoparticles was diluted in 20 μL of a 15 mg/mL Pluronic F127 solution (Sigma Aldrich, MO; catalog number P2443) in a black 96 well plate. Then, 10 μL of a saturated PDAM solution in methanol (~0.3 mg/mL) was added to each well. A standard curve was generated using known concentrations of unmodified PS nanoparticles. Fluorescence intensities of PDAM and nanoparticles were measured at excitation and emission wavelengths of 340/395 and 480/520 nm (Infinite M200, Tecan, Switzerland), respectively. The residual carboxylic groups density in the conjugated nanoparticles was calculated from the unmodified PS particles standard curves. The density of peptides conjugated to the nanoparticles was calculated by subtracting the total carboxylic groups present on the unmodified PS nanoparticles from the residual carboxylic groups in the nanoparticles. Samples were tested in triplicate of at least three independent experiments.

Cell culture.

CuFi-1 cells (ATCC CRL-4013) were maintained as monolayer cultures in flasks pre-coated with 60 μg/mL solution of human placental collagen type IV (Sigma Aldrich, MO; catalog number C7521) and grown in bronchial epithelial growth medium (BEGM) supplemented with SingleQuot additives from Lonza (BEGM Bullet Kit, reference CC-3170) and 50 μg/mL G-418 (Sigma Aldrich, MO; catalog number A1720) at 37°C and 5% CO₂.

Transwell co-culture cell uptake assay.

A co-culture diffusion experiment with CF sputum from patients and CuFi-1 cells was performed in 3.0 μm polyester membrane 24-well transwells at 37°C and 5% CO₂. Prior to co-culture, cells were seeded in the receiver compartment of a transwell at a cell density of 15,000 cells/cm² and grown for 24 hours. After, 100 μL of CF sputum was transferred to the donor compartment in the transwell system, with 0.6 mL BEGM cell culture media and the CuFI-1 cell monolayer in the receiver compartment. Nanoparticle suspensions were added to the donor compartment on top of the CF sputum layer at an initial concentration of 1 mg/mL. After 2 hours, the transwells and cell culture media were removed, and cells were washed 3 times with ice-cold 1X PBS. To detach the cells, 150 μL of 0.25% trypsin-EDTA (Gibco, MA; catalog number 25200056) was added to each well and incubated at 37°C for 10 minutes. Next, 150 μL of ice-cold 1% FBS in Dulbecco’s phosphate buffered saline was added and cells were transferred to centrifuge tubes. Cells were spun at 125 x g for 5 to 10 minutes and the supernatant was discarded. Cells were resuspended in 250 μL 1X PBS with 1.25 μL 1 mg/mL propidium iodide (PI; Sigma Aldrich, MO; catalog number P4864). Cell uptake of nanoparticles was evaluated by flow cytometry (Accuri, Becton-Dickinson, CA). Data presented are the average of three independent experiments.

Cell uptake of nanoparticles.

Cellular uptake studies were performed at optimized conditions based on previously published studies [28]. Prior to cell uptake experiments, 24-well plates were seeded with CuFi-1 cells at a cell density of 15,000 cells/cm² and grown for 24 hours at 37°C and 5% CO₂. On the day of experiment, 25 μg/mL suspension of fluorescent nanoparticles was added to cells in 0.5 mL BEGM cell culture media. After 2 hours, the cell culture media was removed, and cells were washed 3 times with ice-cold 1X PBS to ensure the removal of non-internalized nanoparticles adhering on the cell membrane and that the remaining cell fluorescence is from nanoparticles that have been internalized by the cells [29]. To detach the cells, 150 μL of 0.25% trypsin-EDTA solution was added to each well and incubated at 37°C for 10 minutes. Next, 150 μL of ice-cold 1% FBS in Dulbecco’s phosphate buffered saline was added, and cells were transferred to centrifuge tubes. Cells were spun at 125 x g for 5 to 10 minutes and the supernatant was discarded. Cells were resuspended in 250 μL 1X PBS with 1.25 μL PI. Cell uptake of nanoparticles was analyzed by flow cytometry.

Distribution of nanoparticles in mouse lung airways.

All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Texas at Austin. Lung distribution studies were performed as previously described [30]. Briefly, Balb/c mice (female, 6–8 weeks, Charles River) were anesthetized under a continuous flow of 2% isoflurane, and 50 μL of red fluorescent 100 nm nanoparticles (2 mg/mL) conjugated with either mucus-penetrating peptides or mPEG in sterile hypotonic PBS were administered intratracheally. After 30 min, mice were sacrificed, and the lungs were carefully harvested, fixed in 4% PFA (Sigma Aldrich, MO; catalog number 158127), embedded with optimum cutting temperature (OCT) compound, mounted, and cryosectioned. The sections were stained with 4′,6-diamidino-2-phenylindole (DAPI), and imaged using an Olympus IX-83 inverted fluorescence microscope.

Retention of nanoparticles in mouse lung airways.

Lung retention studies were performed as previously described [30]. Briefly, Balb/c mice (female, 6–8 weeks) were anesthetized under a continuous flow of 2% isoflurane, and 50 μL of red fluorescent 100 nm nanoparticles (2 mg/mL) conjugated with either mucus-penetrating peptides or mPEG in sterile hypotonic PBS were administered intratracheally. At pre-determined time points, mice were sacrificed, and the lungs carefully harvested, rinsed briefly in PBS, laid on a labeled petri dish, and stored at −20°C until imaging. Imaging was performed with an IVIS Spectrum In Vivo Imaging System, with fluorescence excitation/emission wavelengths of 570/620 nm with an exposure time of 0.5 sec. Measurement of fluorescence (total radiant efficiency [p/s] / [μW/cm²]) was done using Living Image 4.3 software (PerkinElmer). Lung retention over time was calculated by normalizing fluorescence at varying time points with initial fluorescence.

Statistical analysis

Unless otherwise indicated, data are presented as mean ± standard deviation. All statistical analyses were performed using GraphPad Prism 8 (GraphPad Software, La Jolla, CA), at a significance level of p ≤ 0.05, and are indicated in each figure legend.

Results

High-throughput sequencing of phage libraries screened against CF mucus indicates a convergence for unique peptides

To identify peptide-presenting phage that diffuse across the barrier, the T7 phage CX7C library was iteratively screened against a CF mucus model (Fig. 1A). There was a significant 232-fold increase in the output to input of phage (C/C₀) that penetrated through CF mucus from the third round compared to the first round (Fig. 1B). This increase under selection pressure of mucus barrier suggests there is possible enrichment for clones with desired properties (i.e. mucus penetration) in CF mucus [18].

Figure 1.

Overview of the iterative phage library selection in CF mucus model. (A) A T7 phage peptide library with an initial diversity of 10⁵ distinct clones is added on top of a CF mucus layer in a Transwell system (Corning, MA) and incubated in the donating reservoir with 1X PBS. Phage that penetrate through CF mucus are collected, quantified by standard double-layer plaque assay, and amplified by E. coli for use either in the next selection round or subjected to PCR and NGS. (B) Enrichment of phage plaque forming units recovered from the basolateral side over three rounds of screening (R1 to R3, n = 3), quantified by standard double-layer plaque assay (unpaired, two-tailed Student’s t-test, p<0.01). (C) Net charge at pH 7.0 and (D) GRAVY score of top 20 peptide hits in 60 minutes eluates, found by high-throughput sequencing of phage display selection against CF mucus. Negative GRAVY score values are indicative of hydrophilic peptide sequences whereas positive GRAVY score values indicate hydrophobic peptide sequences. Net charge was calculated by Protein Calculator v3.4 (The Scripps Research Institute, La Jolla, CA. Available at http://protcalc.sourceforge.net/), and GRAVY score was calculated according to the literature [21]. (B, D) mean ± SD. (C) median, IQR. One-way analysis of variance (ANOVA) with post hoc Tukey’s multiple comparisons test (p<0.05).

Next, the phage DNA from all three selection rounds in replicates was isolated, PCR amplified, and sequenced by high-throughput sequencing. An average of 635,004 ± 124,391 DNA sequences was obtained from the samples. From these sequences, sequences that did not encode for the CX7C peptide sequence or included stop codons before the first cysteine residue were excluded from subsequent analysis and experiments; otherwise, reads would contain a significant fraction of sequences with TAG stop codons or not display any peptide sequence. The number of unique peptide sequences in each sample decreased considerably with each subsequent round of selection (Supplementary Table 2), which has been observed by others [31, 32].

To determine the distribution and abundance of peptide sequences and reproducibility (i.e. similarity between replicates) with each round of selection, we analyzed density plots of peptide frequencies (i.e. counts) (Fig. S1). With each successive round of selection (denoted as R1–R3), the frequencies of specific peptide sequences in both replicates (denoted as replicate 1 and replicate 2) were reproducible; peptides observed with 100 counts in replicate 1 were observed at similar frequencies in replicate 2. Moreover, from the heat map, there is a convergence of the most frequent peptides. With each successive round of selection, the number of unique sequences decreases, and there is a convergence of abundant peptides (Fig. S1), which suggests there is selection and enrichment for unique peptide sequences [18].

To ensure that the frequency of identified sequences was due to screening and not growth bias, these sequences were compared to the naïve library grown without selection [33]. The naïve phage library was amplified without any selection pressure for three rounds, and these samples were sequenced by NGS to identify overrepresented sequences indicative of fast growers. Selected peptide sequences from Table 1 were not amongst the overrepresented sequences present in the amplified rounds of the naïve library. In addition, we compared the selected peptide sequences to databases that have identified sequences with growth bias and/or non-specific affinity, also known as “target-unrelated peptides” [23]; the selected peptide sequences (Table 1) were not found in these comprehensive databases. These results indicate that selected peptides are not artefactual and can be further validated for penetrating transport through the mucus barrier.

Table 1.

Physicochemical properties of selected mucus-penetrating peptides. Amino acid sequences, net charge at 7.0, and GRAVY score of peptide hits found by high-throughput sequencing of phage display selection against CF mucus. An asterisk (*) indicates the presence of a stop codon. Net charge was calculated by Protein Calculator v3.4 (The Scripps Research Institute, La Jolla, CA. Available at http://protcalc.sourceforge.net/), and GRAVY score was calculated according to the literature [21]. Negative GRAVY score values are indicative of hydrophilic peptide sequences whereas positive GRAVY score values indicate hydrophobic peptide sequences.

Phage clone	Sequence	Feature	Net charge at pH 7.0	GRAVY score
Clone 1	CGGQDLKSC	Top 20, MUC16 hit	0	−0.411
Clone 2	CSNLTSP*C	Top 20, MUC2 precursor	−0.1	−0.157
Clone 3	CPSSSREKC	Top 20, MUC19 precursor	1	−1.211
Positively charged (+)	CRRRRKSAC	Control, positive charge	5	−1.767

Next, the amino acid composition of peptide sequences from each round of selection was compared to the original library (Fig. S2). Here, mostly hydrophobic amino acids had an overall decrease in their frequency during selection compared to the original library. There was an enrichment in glycine (G), serine (S), and in the acidic residues glutamic acid (E) and aspartic acid (D) over rounds compared to the original library. Basic residues such as histidine (H), lysine (K), and arginine (R) had an overall decrease in frequency after three rounds of selection against CF mucus.

Identification of CF mucus-penetrating peptides by high throughput sequencing leads to negative to neutral net-charge and hydrophilic sequences

After high-throughput sequencing of DNA isolated from phage eluates, sequences were translated, and the top 20 most abundant peptide sequences and their physicochemical properties (i.e. charge and hydropathy GRAVY score) were calculated (Fig. 1, C and D). The median net-charge was close to −1.0 in the three rounds; however, the interquartile range indicates there is a trend for the distribution to approach neutral charge with each round of selection. The GRAVY score gradually increased with subsequent rounds of selection, but overall, sequences were mostly hydrophilic to neutral.

Next, BLASTp local alignment search was done on the top 20 most abundant peptide sequences to identify their sequence to an existing database of known proteins. Of interest, peptide sequences CGGQDLKSC, CSNLTSP*C, and CPSSSREKC had homology to mucin 16 (MUC16), mucin 2 precursor (MUC2), and mucin 19 precursor (MUC19), respectively (Table 1; an asterisk (*) indicates the presence of a stop codon). MUC16 is a cell surface associated mucin present in the airways, salivary glands, cervix, and eyes; MUC2 and MUC19 are secreted, gel-forming mucins present either in the airways, salivary glands, intestine, cervix, and/or eyes [34]. The aforementioned sequences possess either slightly negative or close to neutral net charges at pH 7.0 and were close to neutral or hydrophilic as indicated by the calculated GRAVY index score (Table 1) [21].

Immunogenicity is a concern in the development of drug delivery systems and could limit the medical applicability and commercial success of drug products. Therefore, we have analyzed the discovered peptide sequences using algorithms to predict antigenic determinant sites in proteins and peptides, such as the antigen prediction tool that follows the method of Kolaskar and Tongaonkar [35] (available at http://imed.med.ucm.es/Tools/antigenic.pl), as well as other databases that search for peptides that bind to major histocompatibility complex (MHC) molecules, such as EPIMHC (available at http://imed.med.ucm.es/epimhc/) and IEDB Analysis Resource Epitope Prediction and Analysis Tools (available at http://tools.iedb.org/main/). The discovered peptide sequences from the phage display screenings (Table 1) did not present any predicted antigenic determinants, therefore it is expected minimal immunogenicity related to the peptide sequences.

Permeation assays of phage clones displaying the identified peptides indicates selection for phage particles with enhanced diffusion across CF mucus

Permeation assay across CF mucus model.

We performed a permeation assay in a transwell system similar to the phage screening to confirm that identified phage-displayed peptides (Table 1) improve diffusion of phage across the CF mucus model. Each phage clone (2.0 × 10⁹ genomic copies (gc)) displaying CGGQDLKSC, CSNLTSP*C, or CPSSSREKC (denoted as Clones 1–3, respectively; an asterisk (*) indicates the presence of a stop codon), along with a positive charge phage control (Clone (+)) was added on top of a CF mucus layer on the donor compartment of a transwell system. Phage that diffused across the mucus layer were recovered from the receiver compartment, and the amount of phage (Q) in the eluates was quantified by qPCR. qPCR method was used to facilitate the processing of large sample numbers, to reduce variability, and for greater accuracy. It has been previously reported that qPCR technique provides accurate titer estimation that reflected the biologic titers of viruses and rapid turnaround time for fast titer quantification [36]. Moreover, it has been shown that the qPCR method for the quantification of bacteriophage produces more accurate quantitative data with a wider linear range than those obtained by the plaque assay [36, 37]. As shown in Fig. 2A, phage clones 2 and 3 permeated across CF mucus 9-and 13-fold higher, respectively, compared to the positively charged control clone (+). Clone 1 showed 3-fold improved permeation compared to the positively charged control clone (+). Due to the presence of carboxyl and sulfate groups on the mucin proteoglycans, mucus has an overall negative net-charge. Therefore, it is expected that positively charged molecules diffuse slower compared to neutral to negatively charged molecules due to strong electrostatic-driven binding to mucin fibers [3, 38, 39]. These findings suggest that the identified clones 1–3 have the ability to facilitate phage diffusion across CF mucus.

Figure 2.

Amount of phage clones that permeate across (A) CF mucus model, (B) CF sputum from patients. Each phage clone (2.0 × 10⁹ genomic copies (gc)) was added on top of a CF mucus layer on the donor compartment of a transwell system. Phage that diffused across the mucus layer were recovered from the basolateral side at 2 hours, and phage amounts (Q) in the eluates were determined by qPCR. Data represents mean ± SD (n = 3). Kruskal-Wallis test with post hoc Dunn’s multiple comparisons, *p<0.05; (B) p=0.067 Clone 3 vs Clone (+).

Permeation assay of phage-displayed peptides against CF sputum from patients.

We validated the ability of the identified phage-displayed peptides to improve diffusion of phage ex vivo across patient samples. Here we performed a permeation assay in a transwell system similar to the CF-like mucus permeation assay but used CF sputum from patients (n = 15) pooled together to minimize patient-to-patient variation. Phage were incubated and quantified similar to the CF-like permeation assay. As shown in Fig. 2B, phage clones 1, 2, and 3 permeated across CF sputum 590-, 375-, and 244-fold more than the positively charged control clone (+), respectively, suggesting that the identified phage-displayed peptides have the ability to facilitate phage diffusion across CF sputum from patients.

Surface physicochemical properties affect phage permeation across CF sputum from patients.

The significant variation in the composition and arrangement of the peptide sequences due to the diversity of amino acids and their physical and chemical properties can impact mucus transport. Rationally designed peptides have been previously investigated to study the effects of charge and spatial charge distribution in purified pig gastric mucin [16]. In order to study the effect of charge, hydrophilicity, and spatial distribution on the diffusion of peptide-presenting phage in CF sputum from patients, we performed the permeation assay in a transwell system with CF sputum from patients using phage clones displaying peptides with different surface physicochemical characteristics (Table 2). We tested six different phage clones displaying peptides rationally designed as either positive charges, negative charges, hydrophobic, hydrophilic, or as two neutral-charged peptides in different configurations: block (i.e. a combination of three positive amino acids followed by three negative amino acids) and alternate (i.e. a combination of positive and negative amino acids arranged in an alternating sequence) (Table 2). As shown in Fig. 3A, the negative-charged and hydrophilic phage-displayed peptides exhibited better diffusion in CF sputum. The negative-charged phage clone (−) showed the highest diffusion up to 600-fold higher compared to the positive-charged phage clone (+). The increase in hydrophilicity led to 2.5-fold enhanced transport compared to the hydrophobic phage clone and 270-fold enhanced transport compared to the positive-charged phage clone (+). The combination of negative and positive surface charges in block and alternate configurations with overall net-neutral charges improved permeation 50- and 110-fold of phage clones, respectively, compared to the positive-charged phage clone (+). Collectively, these results indicate that CF sputum strongly hinder free diffusion of the phage-displayed peptide with positive net charge, and corroborate with previous studies that demonstrate that net charge and hydrophilicity play a key role in transport across the mucus barrier [3, 16, 40].

Table 2.

Phage-displayed peptides with different net charge, hydropathy, and spatial charge distributions to investigate the effect of surface physicochemical properties and spatial distribution on phage diffusion across CF sputum from patients.

Phage clone	Position							Net charge at pH 7.0	GRAVY score
	1	2	3	4	5	6	7
Clone (+)	R	R	R	R	K	S	A	5	−1.767
Clone (−)	E	E	E	E	E	E	E	−7	−2.167
Clone hydrophobic	A	A	A	A	A	A	A	0	1.956
Clone hydrophilic	N	N	N	N	N	N	N	0	−2.167
Clone block	K	K	K	A	E	E	E	0	−1.711
Clone alternate	K	E	K	E	K	E	A	0	−1.711

Figure 3.

Effects of physicochemical properties and spatial distribution on phage-displayed peptides diffusion in CF sputum. (A) Phage-displayed peptides with different net charge, hydropathy, and spatial charge distributions diffusion across CF sputum. Data represents mean ± SD (n = 3). Kruskal-Wallis test with post hoc Dunn’s multiple comparisons, *p<0.05 vs clone (+), ***p<0.001 vs clone (+). (B–C) Regression and correlation coefficients between (B) net charge or (C) GRAVY score of phage-displayed peptides and phage amount that permeate CF sputum from patients. Negative GRAVY score values are indicative of hydrophilic peptide sequences whereas positive GRAVY score values indicate hydrophobic peptide sequences.

To better understand how charge and hydrophobicity of phage-displayed peptides affect phage permeation across CF sputum, we analyzed correlations between net charge and GRAVY score (i.e., which indicates the degree of hydrophilicity or hydrophobicity of a peptide sequence) from each phage-displayed peptide (Table 2) vs the amount of phage clones that permeated across CF sputum from patients. Figs. 3B and 3C depict the Pearson correlation coefficients (r) for permeation of each phage clone against charge and GRAVY score, respectively. The increase in charge of phage-displayed peptides was significantly correlated with a decrease in phage permeation (Fig. 3B). There was not a significant correlation between an increase in GRAVY score (i.e., increase in hydrophobicity) and phage permeation (Fig. 3C).

Mucus barrier and cell uptake co-culture assay indicate higher uptake of mucus- penetrating peptide-conjugated nanoparticles

Building on our work to identify and characterize peptides that transport through the CF-like and CF patient sputum, we wanted to confirm their ability as surface chemistries to facilitate intracellular uptake of nanoparticles. Nanoparticles are promising carriers as for transmucosal delivery due to their ability to deliver and protect large amounts of therapeutic cargo while minimizing off-target toxicity. However, these nanoparticles must overcome the transport barriers presented by both the mucus and cell surface. The enhanced transport of nanoparticles across mucus barriers using different surface chemistries, such as hydrophilic, net-neutral polymer coatings, has been extensively demonstrated [41, 42]. However, even after penetration through the mucus barrier, nanoparticles must bind and enter cells to deliver their payload. Previous studies have demonstrated that the cellular uptake of nanoparticles depends on their physicochemical properties such as size, shape and surface chemistry [43, 44]. While existing surface coatings on nanoparticles improve mucus penetration, there are challenges associated with achieving intracellular uptake in CF-affected cells. Consequently, we tested to confirm that our mucus penetrating peptides as surface coatings could permeate through CF sputum and achieve cell uptake. Towards this end, we incubated 1 mg/mL of fluorescent polystyrene (PS) nanoparticles functionalized with our identified mucus-penetrating peptides or mPEG in a transwell co-culture with CF sputum on the donor compartment and CuFi-1 bronchial epithelial cells in the receptor compartment for 2 hours at 37°C, and uptake was measured by flow cytometry (Fig. 4A). Prior to incubation in the co-culture setup, the nanoparticles were monodisperse and stable in 1X PBS and in bronchial epithelial growth medium (BEGM) (Supplementary Table 3). The coating density of the conjugated PS nanoparticles was on average 60.95 ± 11.03 %, which represents an average of 6.62 00d7 10⁴ peptide molecules per nanoparticle. It has been reported in previous literature using PEGylated enhanced cell penetrating peptide nanoparticles that coating densities of 40% and 60% demonstrated higher diffusivities in human CF sputum samples [45]. Another study using PEGylated nanoparticles demonstrated that a coating density of at least 5% or higher is required to effectively shield the nanoparticle core from interacting with mucus components in vitro and ex vivo in cervicovaginal mucus [46]. From the transwell co-culture assay of nanoparticle uptake, there was approximately 3-fold higher uptake of peptide CGGQDLKSC conjugated nanoparticles compared to non-modified, carboxylated nanoparticles and mPEG 1KDa conjugated nanoparticles (Fig. 4A). Peptide CPSSSREKC conjugated nanoparticles exhibited 2.6- and 2.2-fold significantly improved uptake compared to carboxylated nanoparticles and mPEG 1kDa conjugated nanoparticles, respectively (Fig. 4A).

Figure 4.

In vitro bronchial epithelial cell uptake and in vivo distribution and retention in the mouse lungs of mucus-penetrating peptides and mPEG conjugated fluorescent nanoparticles. (A-B) CuFi-1 cells co-cultured with (A) CF sputum and (B) without CF sputum were incubated with 100 nm carboxyl-modified PS nanoparticles conjugated with either clones 1–3 peptides or mPEG 1kDa for 2 hours at 37°C prior to flow cytometry. Data represents mean ± SD (n = 3). (C-E) The distribution and retention of conjugated 100 nm PS red fluorescent nanoparticles in the mouse lungs. (C) Tracheal distribution at 30 min after administration (n=3). (D) Retention in the entire lung over time. The percentage of retention is reported as the percentage of the initial fluorescence for each nanoparticle after administration. (E) Representative images of the mouse lungs harvested at varying times after intratracheal administration of uncoated and mucus-penetrating CPSSSREKC-conjugated PS nanoparticles. Data represent mean ± SEM (n = 4). *p < 0.05, **p < 0.01. One-way analysis of variance (ANOVA) with post hoc Tukey’s multiple comparisons test. An asterisk (*) indicates the presence of a stop codon.

To confirm that improved uptake was due to cellular uptake and not decreased delivery due to hindered diffusion through CF sputum, we tested the ability of conjugated nanoparticles to enter cells directly without permeation across CF sputum. Here, we incubated the CuFi-1 cells with 25 μg/mL of each fluorescent nanoparticle suspension for 2 hours at 37°C and quantified uptake by flow cytometry (Fig. 4B). As shown in Fig. 4B, CuFi-1 cell uptake of carboxylated, peptide CGGQDLKSC, and peptide CSNLTSP*C functionalized nanoparticles was approximately 2.1-fold higher compared to 1kDa mPEG conjugated nanoparticles. Peptide CPSSSREKC conjugated nanoparticles exhibited approximately 1.6-fold higher cell uptake compared to 1 kDa mPEG conjugated nanoparticles. Our results corroborate with previous studies which demonstrated that PEGylated systems impaired cellular uptake [14, 45, 47–49].

To investigate the distribution of the mucus-penetrating peptides in vivo in the airways over short time periods, we administered peptide-conjugated nanoparticles of 100 nm diameter intratracheally to mice. As shown in Fig. 4C, thirty minutes following administration into mouse lungs, particles coated with peptides CGGQDLKSC and CPSSSREKC, and mPEG 1 kDa were uniformly distributed along the mucus layer in the airway epithelium, whereas uncoated carboxylated nanoparticles were found mostly aggregated and sparsely distributed in the lung airways. We observed a higher accumulated amount of CPSSSREKC peptide coated nanoparticles uniformly distributed in the airway epithelium. Peptide CSNLTSP was found partially distributed along the airway epithelium but not uniformly. Moreover, we have performed high-resolution confocal laser scanning microscopy of the mouse lung cryosections to determine particle colocalization within the tissue (Supplementary Information, Figure S4). From the confocal images it is possible to observe colocalization of the red fluorescent nanoparticles within the airway tissue as well lining the epithelia. In vivo as well as in vitro epithelial cellular uptake of polymeric nanoparticles has also been previously investigated and demonstrated by confocal laser scanning microscopy in previous published studies (Xia, Kovochich et al. 2008, Kuhn, Vanhecke et al. 2014, Du, Wang et al. 2018).

Next, we measured retention of mucus-penetrating peptide-conjugated nanoparticles in the lung airways compared to PEGylated and uncoated nanoparticles following pulmonary administration. At time “zero”, mice were sacrificed immediately after intratracheal administration and lungs were harvested for subsequent IVIS imaging and measurement of fluorescence. At time “zero”, the fluorescence is considered the peak fluorescence intensity (=100%). The subsequent fluorescence intensities were calculated by normalizing fluorescence at each time point over the initial fluorescence at time “zero”, and plotted as the percentage of the initial fluorescence for each nanoparticle after administration. We found that over 90% of the CPSSSREKC peptide coated nanoparticles initially deposited in the mouse lungs remained in the airways at 24 h post-administration, whereas the concentration of uncoated carboxylated nanoparticles decreased to ~ 48% of the initial value. These findings suggest that uncoated nanoparticles, unlike our peptide-coated nanoparticle, were rapidly eliminated from the mouse lungs most likely due to homeostatic mucociliary clearance (Fig. 4, D and E). Nanoparticles coated with peptides CGGQDLKSC and CSNLTSP, and mPEG had ~ 68%, ~75%, and ~67% retention in the airways after 24 h post-administration, respectively (Fig. 4D).

Discussion

For successful delivery, drug and gene therapies must overcome poor diffusion through the mucus barrier and after crossing the mucus layer, must also achieve intracellular uptake. To address these challenges, we have developed a phage transport assay to screen T7 phage-displayed cysteine constrained heptapeptide libraries in CF mucus to identify mucus-penetrating peptides as surface coatings that facilitate transport through mucus and ultimately, penetrate the affected cells. Cysteine-constrained peptides are more stable due to their reduced conformational freedom, which results in higher receptor selective affinities [50, 51]. As a result, the cysteines ensure that the peptide will be in the same orientation when interacting with mucus; linear peptides have more degrees of freedom and are more likely to have variable interactions with mucus due to its conformational freedom. The short length of a heptapeptide does not possess secondary structures, as opposed to a longer length peptide, that could potentially elicit variable interactions with the mucus barrier depending on its current conformation. Moreover, their constrained structural conformation confers greater proteolytic stability [51]. The T7 phage vector can be genetically engineered to multivalently display 415 copies of a peptide on its gp10 protein coat [52] and has size features (~60 nm diameter) on the scale of conventional synthetic nanoparticles. We used short length heptapeptides to avoid a significant increase in the size of phage or nanoparticles, which could affect the transport and uptake studies; as a result, any interactions with mucus would be due to the surface physicochemical properties. After repeated screening of the T7 phage library (~10⁵ diversity of random peptides) against CF mucus (Fig. 1A), we identified phage able to transport through mucus with more than 200-fold increased output in the third round eluates compared to the first round of screening (Fig. 1B). This increase in output is indicative of enrichment and selection of phage clones with favorable properties [18], in this case for mucus penetration. Previously, using a different M13 phage library against CF mucus-like reconstituted mucin, we identified phage with up to 17.5-fold improved output ratios in mucus in the fourth round of screening, and phage-displayed peptides demonstrated up to 2.6-fold improved transport in mucus compared to wild-type control (i.e. non-recombinant, insertless phage) [53]. However, this low-throughput sample size (20–100 clones) from traditional Sanger sequencing, while providing valuable information, presents a limited repertoire of the complete sequence space (less than 0.01% of available sequences), and as a result, it is likely that better performing peptides are not identified [32]. Second, the low copy display of peptides on the p3 coat of M13 (i.e. 5 copies versus 415 for T7) has a small surface area and does not allow for multiple interactions with the mucus substrate, which is needed to identify peptides that have desired weak but reversible interactions with mucus needed for penetrating transport [39, 54, 55]. Finally, while previous work focused on phage transport through reconstituted mucin, it is necessary to validate transport in sputum samples of CF patients; building insights from patient samples advances translational aspects in therapeutic applications of the mucus- and cell-penetrating peptide carriers.

We utilized a bioinformatics workflow to count and sort the frequencies of individual phage-displayed peptides in each eluate from the rounds of screening against CF mucus. A large diversity of peptides was observed in the output eluates in the first-round replicates (Supplementary Table 2). However, there were fewer unique peptides after each round, with a ~12- and 15-fold decrease after the second and third rounds compared to the first round, respectively (Supplementary Table 2). This observation is in agreement with the expected process of selection, where there is an increase in specificity of phage-displayed peptides over additional rounds of selection and a decrease in their sequence diversity [18, 31]. These findings suggest that mucus acts as a selection pressure for potential mucus-penetrating peptides after each round of screening. From selection, a small set of mostly hydrophilic and slightly negative or close to neutral net-charge peptide sequences were selected for subsequent transport studies in patient samples (Table 1). Of interest, the three discovered peptides presented homology to human mucin proteins in a basic local alignment search tool (BLASTp) database search, indicating the potential importance of the peptide sequence. Sequence homology of peptide sequences to their targets have been previously reported in phage display biopanning experiments against the human vasculature and collagen, among others [56, 57]. It is feasible that mucus biomimetic peptides might shield or have less hindered intermolecular interactions with mucins [58], enabling their diffusion across mucus.

The combination of phage display libraries with high-throughput sequencing provided a more in-depth study of the phage selection process and identification of the most abundant peptide sequences in each round of screening. We determined the overall amino acid composition of the peptides from each round of selection (Fig. S2). There was an increase in glycine, serine, and acidic residues, such as glutamic acid and aspartic acid, and a decrease in hydrophobic and basic residues, such as histidine, lysine, and arginine. The changes in amino acid composition during selection can be attributed due to intermolecular interactions (or lack thereof) between peptide substrates with mucins present in mucus. Mucins, the main non-aqueous component of mucus, are heavily glycosylated proteins with an overall net-negative charge due to high sialic acid and sulfate content [1]. Moreover, we calculated the physicochemical properties (i.e. charge and hydrophobicity) of the most twenty abundant sequences in each round of screening, and the in silico analysis indicated that the peptide pool presented mostly negative net-charge and hydrophilic sequences over rounds of selection. The prevalence of hydrophilic residues is consistent with prior work where hydrophilic polymers provided an inert surface for mucus, or “muco-inert”, which minimizes mucin interactions for improved particle transport [6].

We confirmed that identified phage-displayed peptides improved diffusion of phage in reconstituted CF mucus and sputum from CF patients (Fig. 2). We pooled samples from patients to minimize patient-to-patient variability. Patient heterogeneity for CF samples has been previously reported in the literature, with sample variations even between the same patient intra-day, or within the same sample [59], which suggests that CF sputum properties at the microscale are patient-specific. We were able to demonstrate up to ~600-fold greater permeation of the phage-displayed peptides in CF sputum, compared to the positively charged control phage clone. Since mucus has an overall negative net-charge due to the presence of carboxyl or sulfate groups on the mucin proteoglycans, it is expected that neutral to negatively charged molecules diffuse faster compared to positively charged molecules due to lack of strong electrostatic-driven binding to mucin fibers [41]. Indeed, we observed significant higher diffusion of phage-displayed peptides with net-neutral and negative charges in CF sputum compared to a positive charged phage-displayed peptide (Fig. 3A). These findings are in agreement with previous studies where neutral and negatively charged nanoparticles have demonstrated higher diffusivities in mucus compared to positively charged particles [9, 60]. It has been demonstrated that positively-charged molecules typically exhibit low permeability across the mucus barrier due to interactions with the glycoproteins and lipids in the mucus [3, 38, 39]. Furthermore, the diffusion of hydrophobic molecules in mucus is hindered compared to hydrophilic molecules due to adhesive interactions with hydrophobic domains stabilized by disulfide bonds present in mucins [61]. These findings suggest that charge and hydrophobicity can influence diffusion in mucus due to intermolecular interactions with mucins. While some basic physicochemical principles, such as net charge and hydrophobicity are understood, there is still much to be elucidated on mucus binding and permeation mechanisms, such as the effects of spatial variation of charge and hydrophobicity. It is known that higher positive charge correlates with higher mucus binding, while neutral to negatively charged molecules have higher mucus penetration [39]. However, net charge is not the only predictor for mucus penetration because different molecules with the same net charge may interact in very different ways with mucus, making rational design of peptides for mucus penetration challenging. For instance, Li et al. reported that a slightly negative charged peptide with separated blocks of positive and negative charge showed decreased transport in mucin, while a peptide with the same amino acid composition but alternated in positive and negative charged configuration did not interact with mucins [16]. Our observations support findings reported by Li et al [16], whereby changing the spatial order of a phage-displayed peptide containing both positively and negatively charged amino acids affected the transport of the peptide through mucus because of the change in spatial distribution of the peptides and interactions with mucus.

After mucus penetration, it is highly desirable that transmucosal drug delivery systems can achieve cell penetration in order to deliver their therapeutic payload intracellularly. Overcoming these two barriers with single chemistries requires balance between two different or even contradictory surface properties [41, 47]. Recent studies have shown hydrophilic, net-neutral charge PEG polymer and polymer conjugates can improve transport of particles in CF sputum up to 90-fold compared to uncoated particles and improve therapeutic activity of antibiotics such as tobramycin against bacterial biofilms commonly associated with CF infections [6, 20]. While the primary strategy for achieving mucus penetration consists in a hydrophilic and net-neutral surface that prevents both hydrophobic and electrostatic interactions, overcoming the cellular barrier traditionally requires positively charged surfaces, due to binding interactions with anionic functional groups on the cell surface [62, 63]. Cell-penetrating peptides (CPPs) have recently demonstrated the potential to enhance the mucosal delivery of drugs and particles [64]; however these positively charged peptides form strong interactions with the negatively charged mucin glycoproteins, thereby hindering transmucosal transport and ultimately, preventing their ability to achieve cell penetration [65]. Recently, a study demonstrated that PEGylated CPP-DNA nanoparticles had improved particle mobility in CF sputum samples and showed uniform distribution in healthy mice airways. However, decreased transgene expression in lung epithelial cells was observed with increased PEGylation rates [45]. In contrast, despite the use of rationally designed peptides with overall negative charges might have a potential application in mucus penetration, the cellular uptake of negatively charged formulations has been previously investigated, and a strong relationship between negative zeta potential and decreased cellular uptake has been demonstrated [66, 67]. While net-neutral and negatively charged polymers and polymer conjugates demonstrate improved mucus transport and may be promising, they may face hindered cellular internalization and escape [14], thereby dampening the enthusiasm for their therapeutic use.

Here, we circumvented these challenges by using peptides to achieve both mucus penetration and cellular internalization. We measured transport of the identified mucus-penetrating peptides conjugated to 100 nm nanoparticles in a co-culture assay of CF sputum and lung epithelial cells derived from CF patients. From our transport studies, we found that select mucus-penetrating peptides not only improved the bulk diffusion of phage in CF sputum from patients ~600-fold more than a positively charged control phage clone but also achieved up to 3-fold significantly higher cell uptake of conjugated nanoparticles compared to non-modified carboxylated- and mPEG-conjugated nanoparticles. We also found that the PEG coating significantly decreased cell uptake of nanoparticles by up to 2-fold compared to mucus-penetrating peptides- and non-coated carboxylated nanoparticles. Collectively, these results indicate that the identified hydrophilic, net-neutral charge peptides that have sequence similarity to mucins successfully traverse the complex CF sputum barrier and improve intracellular uptake of conjugated nanoparticles to lung epithelial cells.

For effective pulmonary drug delivery, uniform distribution into the airway epithelia is beneficial [30]. For example, poor lung distribution has been attributed as a cause of reduced gene transfer from polymeric gene carriers in vivo [68]. Thus, we investigated how the mucus-penetrating peptide coatings influence nanoparticle distribution and retention in vivo in mouse lung airways. Particles with mucus-penetrating peptide coatings exhibited more uniform distribution and retention in the mouse lung airways compared to uncoated nanoparticles, most likely due to improved mucus penetration and subsequent improved lung residence time by avoiding mucociliary clearance. These findings have clinical implications because improved mucus penetration, lung distribution, and lung retention will also increase the probability that nanoparticles and drugs reach the underlying epithelia for therapeutic pulmonary delivery [69].

Conclusions

Our findings demonstrate that using phage display libraries with high throughput sequencing may be an effective strategy to identify mucus-penetrating peptides that improve transport across CF sputum from patients and enhance cell internalization of drug delivery systems. These peptides may effectively serve as conjugates or surface modifications to potentially increase the amount of drug and drug carriers delivered to the CFTR-mutated epithelia to improve clinical outcomes. While this work focuses on mucus environment in CF, the use of mucus-penetrating peptides can be extended to develop the next generation of therapeutics to advance gene and drug delivery through various mucosal barriers, including the lung airways, eyes, gastrointestinal tract, and cervicovaginal tract.