A platform for discovery of functional cell-penetrating peptides for efficient multi-cargo intracellular delivery

Abstract

Cell penetrating peptides (CPPs) offer great potential to deliver therapeutic molecules to previously inaccessible intracellular targets. However, many CPPs are inefficient and often leave their attached cargo stranded in the cell's endosome. We report a versatile platform for the isolation of peptides delivering a wide range of cargos into the cytoplasm of cells. We used this screening platform to identify multiple "Phylomer" CPPs, derived from bacterial and viral genomes. These peptides are amenable to conventional sequence optimization and engineering approaches for cell targeting and half-life extension. We demonstrate potent, functional delivery of protein, peptide, and nucleic acid analog cargos into cells using Phylomer CPPs. We validate in vivo activity in the cytoplasm, through successful transport of an oligonucleotide therapeutic fused to a Phylomer CPP in a disease model for Duchenne's muscular dystrophy. This report thus establishes a discovery platform for identifying novel, functional CPPs to expand the delivery landscape of druggable intracellular targets for biological therapeutics.

Figure 1

CPP screening and selection process. (a) Selection begins with a T7 phage library displaying a fusion of a cargo (in this example, an EGFR receptor binding domain, EBD), an Avitag and Phylomer peptides; (b) Avitagged-Phylomer sequences with potential for CPP activity are internalized into cells; this uptake can also be facilitated by binding to specific cell types via a cell surface receptor (CSR) and uptake into endosomes by receptor-mediated endocytosis; (c) peptides with capacity for cytosolic delivery allow the phage to enter the cytoplasm; (d) selection is performed in mammalian cells expressing the bacterial biotin ligase BirA in the cytoplasm, which ligates free biotin to the lysine residue of the phage-displayed Avitag sequence; this step produces selectively labeled T7 phage that have internalized into the cytoplasm by virtue of the CPP; (e) sodium pyrophosphate (PPi), a specific inhibitor of BirA, is added to the cells to terminate the biotinylation reaction; (f) cells are lysed and streptavidin-coated magnetic beads (SAV) are added to the lysate to selectively capture and concentrate biotinylated T7 phage. Enriched phage are then amplified in E. coli and subjected to further rounds of selection. Identification of specific CPPs is achieved by deep sequencing of early selection rounds or by Sanger-sequencing of individual phage clones after 3–4 rounds of selection.

Figure 2

Uptake of Phylomer CPP_EBD_S11 fusion proteins validated by GFP complementation. (a) Thirteen Phylomer peptides showed a positive GFP complementation signal, evidence of intracellular delivery. These functionally validated CPPs were of various sizes, net charges and origin. (b,c) Dose-dependent uptake of recombinant Phylomer CPP_EBD_S11 fusion proteins was determined by GFP complementation in (b) HEK-293 or (c) CHO-K1 cells transiently transfected with a hGFP1–10 expressing plasmid. (d,e) Dose-dependent uptake of recombinant CPP fusion proteins was confirmed in stable cell lines by GFP complementation of (d) EBD_S11 proteins in HEK-293/GFP1–10 and (e) TRX_S11 proteins in CHO-K1/GFP1–10 where hGFP1–10 is stably expressed. The greater sensitivity of the stable cell lines enables comparison of uptake at low-dose concentrations. “No CPP” control is shown at the highest concentration (40 µM). In flow cytometry experiments GFP complementation is measured as % fluorescent cells. “No CPP” control is EBD_S11 (b-d) or TRX_S11 (e) protein with no CPP moiety. Results are representative of two independent experiments. Error bars represent standard deviation from the mean between duplicates. (f) Fluorescence microscopy visually confirms the CPP-dependent GFP complementation (FITC channel) in CHO-K1 cells transiently transfected with GFP1–10, comparing 0084_EBD_S11 fusion protein (10 µM) to the “No CPP” _EBD_S11 control protein (10 µM). Cells are counter-stained for endogenous cytoplasmic β-Actin (TRITC) and nuclei (DAPI); bar scale is 25 µm.

Figure 3

Initial optimization of a Phylomer CPP sequence using cluster alignments and a PAP-based viability bioassay. (a) Diagrams of SpyTag (SpyT)/SpyCatcher (SpyC) fusion protein conjugates and other cargos used in this study. In protein ligation a short peptide sequence, SpyTag (SpyT), forms an isopeptide bond with the SpyCatcher (SpyC) partner protein in an irreversible peptide-protein coupling. (b) SpyC_PAP/CPP_SpyT conjugate uptake has a dose dependent decrease in CHO-K1 cell viability for Phylomer and Penetratin-delivered protein, assessed by resazurin reduction potential. The 1746_SpyT/SpyC_PAP conjugate (1746_PAP) showed greatest potency, followed by 0084_SpyT/SpyC_PAP conjugate (0084_PAP). Penetratin_SpyT/SpyC_PAP conjugate (Pen_PAP) is included as a positive control for the assay. SpyC_PAP treatment showed no activity at all concentrations tested up to 30 µM. Calculated IC₅₀ values are presented below the graph. Data are from N independent experiments. Error bars represent standard deviation from the mean. Significance was assessed by one-way ANOVA. (c) Uptake of 1746_PAP and potent variants with reduced charge conjugated to SpyC_PAP (1746del_PAP and 1746c27_PAP) show dose-dependent decrease in CHO-K1 cell viability, assessed by resazurin reduction potential. Delivery using 1746del and 1746c27 variants showed improved, lower IC₅₀ compared to 1746. Calculated IC₅₀ values are shown below the graph. Data are from N independent experiments. Error bars represent standard deviation from the mean. Significance was assessed by one-way ANOVA. (d) Summary of 1746 variant testing and selection. Parameters include peptide length, sequence charge, IC₅₀ in viability assays where 1746_SpyT variants deliver conjugated SpyC_PAP into CHO-K1 cells. CPP derivatives are aligned to the 1746 parental sequence with an explanation of variant type. N is the number of independent PAP assays in which each sequence was assessed.

Figure 4

Phylomer CPPs show minimal toxicity. Phylomer CPP toxicity was assessed in cells and in mice. (a,b) Cell viability effect on CHO-K1 cells of 1746 and 1746c27, alongside TAT, following 24 h (a) and 48 h (b) incubation with peptides. Cell viability was assessed by resazurin reduction potential. (c,d) Membrane integrity of CHO-K1 cells were assessed by LDH release following 2 h (c) and 24 h (d) incubation with peptides. All CPPs remained non-toxic to cells. Results are representative of three independent experiments. Error bars represent standard deviation from the mean of triplicate samples. (e) 1746c27 toxicity in vivo was assessed by measuring the urea, creatinine, aspartate transaminase (AST) and alanine transaminase (ALT) concentrations in serum from mice (n = 6) treated with daily IP injections of 1746c27 for 7 days based on EMA- and FDA-approved standard preliminary toxicity testing guidelines. Minimal evidence of toxicity was seen following treatment. AST and urea levels showed no significant difference compared to untreated control mice (p = 0.94 and p = 0.21, respectively). ALT levels showed no significant difference and were lower compared to untreated control mice (p = 0.054), and creatinine levels were significantly lower compared to untreated mice (p = 0.01), which supports the finding of minimal toxicity. Significance was assessed by unpaired, two-tailed T-test.

Figure 5

CPP-mediated intracellular delivery of a range of cargos into cells. (a) Uptake of ^DPMIα peptide into cells shows a dose dependent decrease in T47D cell viability when ^DPMIα was delivered by 1746, and to a lesser extent, by TAT. Cell viability was assessed by resazurin reduction potential after 48 h incubation with peptides. (b) Comparison of 10 µM peptide treatments shows CPP_^DPMIα peptides are significantly more potent compared to ^DPMIα peptide alone. 1746_^DPMIα treatment also shows significantly reduced cell viability compared to TAT. Results are representative of 3 independent experiments. Error bars represent standard deviation from the mean of duplicate samples. Significance was assessed by one-way ANOVA with Dunnett’s multiple comparison test (*p < 0.05; ***p < 0.001; ****p < 0.0001). (c) β-lactamase flow cytometry assay shows dose-dependent cell entry of SpyC_BLA/1746c27_SpyT conjugates at concentrations with the limit of detection as low as 500 nM in CHO-K1 cells; no marked signal is observed for the control protein SpyC_BLA. Data shown are from two independent experiments. Error bars represent standard deviation from the mean. (d–f) Treatment of (d) AMO-1 (plasmacytoma), (e) HL-60 (promyelocytic leukemia), and (f) T47D (breast cancer) cell lines with 1746c27_Omomyc protein. After 48 h incubation, cell viability was assessed by measuring ATP activity. Results show strong, similar efficacy of 1746c27_Omomyc across all three cell lines. The average IC₅₀ values for 1746c27_Omomyc treatment in the three cell lines were calculated as 1.28 µM (AMO-1), 1.88 µM (HL-60) and 1.67 µM (T47D). Peptide 1746c27 alone shows no significant cytotoxicity. Control protein Omomyc exhibits a minor effect on cell viability only at the highest concentrations (mid to high micromolar potencies). Results are representative of two independent experiments. Error bars represent standard deviation from the mean of duplicate samples.

Figure 6

Phylomer CPP delivery is compatible with cell specific targeting and half-life extension approaches. (a) CHO-K1 cells stably expressing EGFR receptor or (b) CHO-K1 cells were treated with 1746del_SpyT conjugated to EGFRAffibody_Bouganin_SpyC (EGFRAffbd_Boug_SpyC) toxin. After 48 h incubation, the cell viability was assessed according to the resazurin reduction potential. Comparison of 100 nM immunotoxin treatment in CHO-K1_EGFR (a, right) vs CHO-K1 (b, right) cells shows that conjugation to 1746del improved delivery compared to the EGFRAffibody alone, and that the 1746del-conjugate retains EGFRAffibody-encoded specificity. Results are representative of 3 independent experiments. Error bars represent standard deviation from the mean of duplicate samples. Significance was assessed by one-way ANOVA with Dunnett’s multiple comparison test (**p < 0.01; ***p < 0.001). (c) T47D cells were treated with 1746c27_PAP_linker_SpyT with and without conjugation to the PAS_SpyC fusion protein. 1746c27-dependent PAP-induced cytotoxicity was detected for all PAS conjugates in comparison to the buffer control (Tris). The Furin-cleavable conjugate exhibited the greatest potency, but all PAS conjugates showed dose-dependent cell toxicity in a comparable concentration range. Linkers are Cathepsin B FKFL cleavage motif (BF), Cathepsin B Valine-Citrulline cleavage motif (Ba) and Furin RKKR cleavage motif (Fur). Results are representative of two independent experiments. Error bars represent standard deviation from the mean of duplicate samples.

Figure 7

1746c27 delivery of a PMO therapeutic in vivo in a disease model of DMD. (a) Intracellular delivery of 1746c27_M23D(+7–18) induces dose-dependent skipping of exon 23 of the dystrophin gene in differentiated murine H-2K^b-tsA58 myogenic cultures. Exon skipping (marked by arrow) can be detected by RT-PCR from doses of 50 nM 1746c27_M23D(+7–18), but is not detected at any dose of M23D(+7–18) morpholino alone or the untreated cells (UT). (b) Tissue staining for dystrophin expression shows in vivo treatment of C57BL/10ScSn^mdx mice (5 treatments over two weeks, with 4 nmol per dose) of 1746c27_M23D(+7–18) causes improved dystrophin protein levels and muscle architecture in the diaphragm, and to a lesser extent the tibialis anterior (samples taken from two independently-treated mice, n = 2). This improvement is compared to untreated C57BL/10ScSn^mdx mice (Mdx untreated control) or those treated with the M23D(+7–18) morpholino alone (M23D(+7–18)-PMO). Treatment with Pip6-morpholino conjugate (Pip6_M23D(+7–18)-PMO) was a positive control for antisense-induced dystrophin expression. Tissue staining for dystrophin expression in C57BL/10ScSn mice (C57 untreated control) shows normal muscle architecture for comparison. Bar scale is 100 µm.