Creating Designer Engineered Extracellular Vesicles for Diverse Ligand Display, Target Recognition, and Controlled Protein Loading and Delivery

Abstract

Efficient and targeted delivery of therapeutic agents remains a bottleneck in modern medicine. Here, biochemical engineering approaches to advance the repurposing of extracellular vesicles (EVs) as drug delivery vehicles are explored. Targeting ligands such as the sugar GalNAc are displayed on the surface of EVs using a HaloTag-fused to a protein anchor that is enriched on engineered EVs. These EVs are successfully targeted to human primary hepatocytes. In addition, the authors are able to decorate EVs with an antibody that recognizes a GLP1 cell surface receptor by using an Fc and Fab region binding moiety fused to an anchor protein, and they show that this improves EV targeting to cells that overexpress the receptor. The authors also use two different protein-engineering approaches to improve the loading of Cre recombinase into the EV lumen and demonstrate that functional Cre protein is delivered into cells in the presence of chloroquine, an endosomal escape enhancer. Lastly, engineered EVs are well tolerated upon intravenous injection into mice without detectable signs of liver toxicity. Collectively, the data show that EVs can be engineered to improve cargo loading and specific cell targeting, which will aid their transformation into tailored drug delivery vehicles.

Keywords: exosomes; extracellular vesicles; genetic engineering; protein loading; targeting.

Figure 1

Generation and characterization of engineered EVs for cell targeting and protein loading. a) Schematic representation of Expi239F cell‐derived EV floatation on an iodixanol density gradient. Expi293F cell supernatant was collected and subjected to differential centrifugation. Small EVs (100k x g pellet) were subsequently bottom‐loaded in high‐resolution iodixanol density gradients (Optiprep) with decreasing densities (50%−10%, bottom to top). Nine fractions of 1 or 2 mL each were collected from top to bottom and their densities were analyzed by measuring the absorbance at 340 nm. b) Individual fractions (F1–F9) were analyzed using a nanoparticle tracking analyzer (NTA) to determine the particle number and c) the number of particles per µg of protein. Data are plotted as the mean of three independent experiments ± s.e.m. d) Representative western blot analysis of F1−F9 density fractions (12 µL/each lane). Membranes were blotted with the following antibodies: Alix, CD63, CD81, and CD9. Protein markers commonly associated with small EVs were enriched in the low‐density fractions. Based on the particle count, particle purity, and protein markers, fractions F1–F3 were pooled and considered as small EVs for further analysis. e) Representative negative staining TEM image of pooled fractions F1–F3 showing clean preparations of highly pure vesicles with a cup‐shaped morphology characteristic of the technique. Scale bar = 100 nm. On the right side, representative zoomed‐in immuno‐gold EM images of vesicles from low‐density fractions positive for CD63, and CD81 protein markers. Scale bar = 20 nm. Five microliters were loaded onto the grids. f) Cryo‐EM images of pooled EV samples (F1–F3) depicting EV size and morphology. Scale bar = 100 nm.

Figure 2

HaloTag enables a versatile and modular display of multiple targeting ligands on the surface of EVs. a) EV engineering construct design and the protein display topology within EVs. b) Schematic representation of the HaloTag designed to covalently bind various synthetic chloroalkane‐based targeting ligands. c) Immuno‐gold labeling of EVs collected from Expi293F cells transiently transfected with a plasmid coding for CLIC1‐HaloTag and naïve EVs as control. EVs were incubated with primary CD63 and CD81 antibodies followed by secondary antibodies conjugated with 6 nm gold particles (black arrow), and primary HaloTag antibody followed by a secondary antibody with 15 nm particles (white arrow). Scale bars = 100 nm. d) Nano‐flow cytometry analysis using the CytoFLEX system of EVs isolated from naïve Expi293F or HaloTag‐Myc‐CLIC1 transfected cells. HaloTag Ligand conjugated with Oregon green dye was used for HaloTag visualization. CellTrace Far Red Cell dye was used to label the whole EV fraction. Shown is an average of 3 independent experiments ± s.e.m. e) Representative western blot analysis of EVs from CLIC1‐HaloTag or mock‐transfected cells and corresponding cell lysates. f) EVs carrying EGFP and HaloTag were incubated with the HaloTag ligand – TMR and imaged on the surface of glass bottom plates. Shown are representative co‐localization images of EGFP and TMR fluorescent signals. Scale bar = 15 µm. g) Correlation of EGFP and TMR peak intensities counts for individual EVs from (f). The dotted line illustrates the 1:1 intensity ratio. Both EFGP and TMR co‐localize in the vague of observed particles. Over 850 EVs were analyzed in each of the 3 independent experiments. h) Quantification of the number of TMR molecules per EV. The peak intensity of the point‐spread function of each detected EV was extracted and divided by the single‐molecule signal to quantify the copy number of functional HaloTag molecules in each EV. Shown is an average of 3 independent experiments ± s.e.m. i) Chemical structure of the synthetic GalNAc derivate synthesized in‐house containing a HaloTag reactive chloroalkane linker (shown in red) for HaloTag biding and display on the EV surface. j) Representative image of the primary human hepatocyte (PHH) spheroids used as a human 3D liver model for the uptake experiments. Scale bar = 200 µm. k) Evaluation of the binding efficiency of engineered EVs to PHH. EVs display GalNAc at the surface and carry NanoLuc (Nluc) protein in the lumen for luminescent tracking. Shown is the average of luciferase signals from each spheroid (n = 6) in relative luminescence units ± s.e.m. The P‐value was calculated using a two‐sided Student's T‐test. p<0.05.

Figure 3

Display of antibody‐binding proteins on the surface of EVs. a) Schematic representation of the EV construct design and protein topology within EVs for the display of antibody‐binding proteins. b) Representative western blot analysis of EVs from cells transfected with Nanobody‐CD81, Protein A‐CD81, and Protein G‐CD81 or naïve cells and corresponding cell lysates. Asterisk indicates the binding of antibodies by proteins A and G. Cross marks bands from antibodies bound to Nanobody. c) Protein A‐CD81 and Protein G‐CD81 vesicles were incubated with IgG–gold antibodies and representative immuno‐gold EM images are shown. Both types of EVs have the binding capacity to IgG‐gold antibodies. Scale bar = 100 nm. d) Schematics representation of the EV immunoassay. EVs were first immobilized on the surface of Poly‐L‐Lysine coated plates, blocked to avoid unspecific binding of the antibodies to their surface, and incubated with HRP‐labeled antibodies for detection. The level of the HRP luminescent signal corresponds to the number of antibodies bound to the EV surface. e) EV surface binding of IgG1 antibody. EVs derived from cells transfected with Nanobody‐CD81, Protein A‐CD81, and Protein G‐CD81 or control (Ctrl) naïve cells (0.3, 3, or 30 µg of the protein) were immobilized followed by their incubation with IgG1‐HRP antibodies as described in (d). The average HRP signal in relative luminescence units ± s.e.m corresponding to the amount of bound IgG1 antibodies to the vesicles is shown (n = 4). No significant binding of IgG1 antibodies was shown in Ctrl vesicles. f) Evaluation of the binding efficiency of engineering EVs displaying anti‐GLP1 receptor (GLP1R) antibody to HEK293 cells overexpressing GLP1R. Protein A EVs or naïve EVs were first incubated with blocking solution followed by incubation with anti‐GLP1R antibodies, labeled with NHS ester Alexa Fluor 594 r dye. After washes, EVs were added to the cells for 1 h at 4 °C. Unbound EVs were washed out, and fluorescent signal was measured. Normalized to Ctrl average fluorescent signal from three biological replicates ± s.e.m. is shown. P‐value was calculated using a two‐sided Student's T‐test. Asterisks indicate that p<0.05.

Figure 4

Protein loading of genetically engineered EVs. a) Schematic representation of the design of the DnaB mini‐intein fusion protein construct and the EV loading system. The tetraspanin CD63 (violet) was fused with a DnaB mini‐intein cassette consisting of a mini‐intein protein domain (yellow) and flanking extein (orange). The C‐term of the mini‐intein cassette was connected with Cre recombinase (green) with an N‐terminal nuclear localization sequence (light green) (NLS‐Cre). NLS‐Cre recombinase is then recruited in the EV lumen during biogenesis as a fusion with CD63. DnaB mini‐intein consists of the splicing domain and is modified to catalyze cleavage at its termini. Mini‐intein‐mediated cleavage reaction releases Cre recombinase from CD63‐Cre fusion protein. b) Representative western blot analyses of the Cre recombinase in EVs loaded with the DnaB mini‐intein system as a function of time and temperature. EVs isolated from Expi293F cells transfected with the mini‐intein loading system were incubated for 0, 24, or 72 h at 22 or 37 °C. GAPDH was used as a loading control and CD9 as an EV marker. c) Schematic representation of the design of a protein loading system with a TimeSTAMP (Time‐Specific Tagging for the Age Measurement of Proteins) drug‐controllable cargo release and loading in EVs. As above, between CD63 (violet) and NLS‐Cre recombinase (green) TimeSTAMP protein (brown) was inserted. The epitope tag (brown) is rapidly removed from the protein of interest by a sequence‐specific protease unless a protease inhibitor Asunaprevir (ASV) is present (shown in yellow). This approach allows the release of Cre recombinase from the fusion with CD63 protein in the EV lumen, once ASV is absent. d) Representative western blot analyses of the Cre recombinase EVs as a function of time and temperature. EVs were isolated from Expi293F cells transfected with the TimeSTAMP loading system. Cells were cultured with and without ASV (3 µm). EVs were incubated for 0, 24, or 72 h at 22 or 37 °C. GAPDH was used as a loading control and CD9 as an EV marker. e) Representative western blot analyses of cells and EVs secreted by Expi293F cells expressing the DnaB mini‐intein or TimeSTAMP loading systems. Cells were transfected with mini‐intein and TimeSTAMP constructs, plasmid overexpressing Cre recombinase or mock (Ctrl sample), and cultured with 3 M of ASV in indicated conditions. Cell supernatant was collected 48 h after transfection and subjected to differential centrifugation. Both loading approaches successfully brought cargo protein inside EVs. Levels of free Cre recombinase inside EVs can be estimated by the intensity of Cre protein bands ≈38 kDa. Bands ≈70 kDa correspond to CD63‐Cre fusion protein. GAPDH was used as a loading control and CD9 as an EV marker.

Figure 5

EV‐mediated functional delivery of Cre recombinase in reporter cells. a) HEK293 cell line expressing loxP‐GFP‐stop‐loxP‐RFP cassette. The CMV promoter initiates the expression of GFP, while the downstream RFP open reading frame (ORF) is inhibited by a stop codon located after the GFP ORF. Upon the presence of Cre protein in the nucleus, the DNA fragment between two LoxP sites is excised, and this enables the expression of the RFP ORF resulting in the fluorescent switch from GFP to the RFP in the cells. b) Image‐based quantification of the percentage of RFP positive cells following a 48 h co‐treatment with control EVs (without Cre recombinase), Cre recombinase loaded EVs via the DnaB mini‐intein system (Cre EVs), both at a concentration of 3.5 × 10⁹ EVs per well, and 0.5 µL commercial Cre recombinase Gesicles (Cre Gesicles, Takara) with indicated concentrations of chloroquine. Data is shown as an average of 3 independent experiments ± s.e.m. c) Representative image of the cells 48 h after the treatment described in (b) with 50 µM or 0 µM Chloroquine. Scale bar = 50 µm. d) Schematic overview of the Nanoprofiler Galectin‐9 (GAL9) reporter assay for endosomal escape evaluation. Cytosolic mCherry‐GAL9 recruits to the sites of endosomal damage via the binding of β‐galactoside sugar molecules. This results in the redistribution of the mCherry signal from the cytoplasm to endolysosome structures, giving the appearance of the mCherry puncta in the cells. e) Quantification of GAL9 puncta in HEK293‐GAL9 reporter cells treated with vehicle, naïve EVs, Cre EVs, Cre Gesicles, 50 µM of chloroquine, and a combination of naïve EVs, Cre EVs and Cre Gesicles with 50 µM of chloroquine. Cells were imaged at indicated times up to 14 h and GAL9 puncta were quantified. Data represents the average log2 fold change to the vehicle condition from three independent experiments ± s.e.m. f) Representative images of HEK293‐GAL9 cells 14 h after treatment described in (e). Scale bar = 50 µm.

Figure 6

Engineered EVs have a safe liver toxicity profile in vivo. a) Schematics for the experimental design. EVs (1 × 10¹¹ particles) or AV5 (1.4 × 10⁹ PFU) were injected into the tail vein of mice. Liver tissues were collected 1 h after injection for histology analysis. Blood from the saphenous vein was collected 6 h after particle tail vein injection for toxicity analyses. b) The levels of selected liver toxicity markers were compared among PBS control animals and control EVs (without Cre recombinase), adenovirus 5 with Cre recombinase protein (Cre AV5), EVs carrying Cre recombinase (Cre EVs), EVs carrying Cre recombinase and decorated with GalNAc molecules (Cre EVs+GalNAc) (n = 6). The average values indicated for each panel parameter ± s.e.m. are shown. The P‐value was calculated using a two‐sided Student's T‐test. Significant differences are indicated by asterisks (*** = p<0.001). c) Representative liver histology images of the animals treated with HaloTag engineered EVs with or without GalNAc. An anti‐human CD63 antibody assay was developed to detect injected particles in the tissue (purple dots) 1 h after the treatment. Kupffer cells were stained with F4/80 antibodies (bright blue). Arrows indicate the accumulation of CD63 signals inside cells. Scale bar = 50 µm in the wide‐field images and 25 µm in the magnifications. d) Quantification of the results described in (c). For both treatment conditions, most of the CD63‐positive cells were liver macrophages (Kupffer cells) (right panel). The average percentage ± s.e.m. of CD63‐positive cells of other types (CD63 signal is outside macrophages) is indicated on the left panel. Each data point represents different liver sections (n = 6).