Targeted delivery of peptidoglycan hydrolases (PGH) via engineered extracellular vesicles (EVs) to clear intracellular bacterial infections

Abstract

Bacterial infectious diseases remain a major global health threat due to increasing antibiotic resistance. Further complicating effective treatment, some pathogens such as Staphylococcus aureus persist within host cells. Peptidoglycan hydrolases (PGHs), including bacteriophage-derived endolysins, represent a promising class of novel antimicrobials due to their rapid bacteriolytic activity and low risk of resistance emergence. However, their clinical application is limited by an unfavourable PK profile and inefficient delivery into infected host cells. Here, we engineered extracellular vesicles (EVs) for the targeted, intracellular delivery of GH15, a novel endolysin against S. aureus. GH15-loaded EVs, functionalized with antibodies targeting α_vβ₃ integrin on bacterially infected cells, efficiently delivered GH15 into S. aureus-positive endothelial cells and macrophages in vitro and eliminated intracellular bacteria. In addition, GH15-loaded EVs promoted the clearance of intracellular bacteria in macrophages of infected zebrafish larvae in vivo. Our findings establish EVs as effective vehicles for intracellular delivery of antimicrobial enzymes and highlight their potential for the treatment of intracellular bacterial infections.

Graphical abstract

Overview of strategy for intracellular delivery of engineered peptidoglycan hydrolases (PGHs) to bacteria-infected target cells. Extracellular vesicles (EVs) harvested from GPI-avidin expressing donor cells are loaded with GH15. Through functionalization with a biotinylated antibody targeting α_vβ₃ integrin, a receptor upregulated on bacterially infected cells, the drug-loaded EVs are directed to these target cells to release their antimicrobial cargo. Here, the antimicrobial effect of engineered EVs (eEVs) was evaluated using an in vitro (cell based) infection model in HUVECs and THP-1 cells. Intracellular delivery and bacterial clearance were assessed by fluorescence microscopy and plaque assay-based quantification of intracellular bacteria. In vivo GH15 delivery and antimicrobial efficacy were further assessed in a zebrafish model of S. aureus infection. Created in BioRender.

Highlights

• •Engineered, bacteriophage-derived endolysin GH15 effectively lyses multidrug-resistant S. aureus.
• •GH15-loaded EVs deliver their payload into S. aureus infected cells and promote killing of intracellular bacteria.
• •Surface-functionalization of EVs with antibodies targeting α_vβ₃ integrins potentiates their uptake by infected cells.
• •EVs loaded with GH15 clear intracellular bacterial infections in vivo in zebrafish larvae.
• •EVs constitute an efficient platform for intracellular delivery of antimicrobial proteins.

1. Introduction

Bacterial infections still constitute a major threat to global health in the 21st century [1,2]. Despite their key role in controlling bacterial infections, antibiotics are becoming less effective due to the rise of antimicrobial resistance (AMR), expected to cause 10 million deaths by 2050 [3]. Staphylococcus aureus, a Gram-positive opportunistic pathogen, frequently causing hospital infections, is the second leading microorganism after Escherichia coli, contributing to the global AMR burden. In 2019, methicillin-resistant S. aureus (MRSA), was linked to over 100,000 deaths [4]. Besides showing antibiotic resistance, S. aureus can grow within abscesses or biofilms, i.e. three-dimensional bacterial aggregates encapsulated within an extracellular polymer matrix, which prevent efficient penetration of antimicrobials [5]. Typically, bacteria within biofilms reside within a dormant state, rendering them less susceptible to antibiotic treatment [6]. Further complicating antibiotic therapy, these bacteria gain access to intracellular compartments of host cells [7], thereby evading treatment of antibiotics, which typically do not cross cell membranes [8]. Thus, there is an urgent need for the development of novel antimicrobial agents capable of targeting intracellular, antibiotic-resistant S. aureus.

In recent years, peptidoglycan hydrolases (PGHs), a class of enzymes specifically inducing bacterial lysis by binding to the cell wall and degrading the peptidoglycan (PG) of the bacterial cells, have emerged as promising alternatives to traditional antibiotics [9]. PGHs, in particular bacteriophage-derived endolysins mediating bacterial cell wall degradation at the end of a phage's multiplication cycle [10], exhibit effective bactericidal activity also against drug-resistant and persistent strains. In addition, due to their high specificity for their target strains, they are less prone to resistance development [11]. Protein engineering has generated PGH variants with improved efficacy and optimized for activity in specific environments [[12], [13], [14]]. The recently described engineered endolysin CHAPGH15_SH3bALE1 (GH15) comprising the enzymatic CHAP endopeptidase domain from the bacteriophage LysGH15 and an SH3b cell wall binding domain (CBD) from the bacteriocin ALE1, is effectively lysing MRSA strains [15]. Yet, native PGHs typically show limited in vivo activity against intracellular S. aureus, due to i) poor in vivo stability and ii) their poor cell permeability [13]. Genetic engineering approaches have been explored to mitigate these challenges. For example, fusion of PGHs to cell-penetrating peptides (CPPs) significantly enhances cellular uptake and increases intracellular bactericidal activity [16]. Similarly, attachment of an albumin-binding domain (ABD) prolongs systemic circulation by extending the blood half-life, while fusion to specific targeting sequences can enable tissue-directed delivery [17,18]. Despite these advantages, genetic and chemical modification strategies carry inherent risks. Structural alterations may compromise enzymatic activity, reduce therapeutic efficacy or affect protein folding and stability, necessitating tedious, case-by-case evaluation of each engineered construct. Consequently, the effective therapeutic application of PGHs against intracellular infections requires delivery strategies, which enhance intracellular access, while improving in vivo stability and biodistribution without directly modifying the enzyme itself.

A promising alternative is the protective encapsulation of PGHs within lipid-based nanoparticles, including engineered extracellular vesicles (EVs). Such systems offer the potential to shield enzymes from degradation, facilitate cellular uptake, and enable targeted delivery, while preserving the native structure and activity of the therapeutic enzyme.

EVs are naturally secreted, membrane-enclosed nanovesicles, effectively shuttling proteins, lipids, and nucleic acids between cells. Their physiological role in intercellular communication, based on their ability to carry diverse biological cargo, has triggered interest in EVs as highly promising drug-delivery systems [19]. EVs have been shown to efficiently deliver their contents into recipient cells either through endocytosis or direct membrane fusion, making them particularly attractive for intracellular drug delivery applications [20]. Owing to their superior biocompatibility and minimal immunogenicity, EVs surpass synthetic nanocarriers, including liposomes, in suitability for their therapeutic use [21]. While EVs exhibit a certain degree of intrinsic biodistribution and tissue-targeting capacity determined by their donor cell origin, they generally display short blood half-lives and are rapidly cleared by the mononuclear phagocyte system (MPS) [22]. This facilitates EV targeting to macrophages. Yet, in order to enrich EVs in specific tissues and promote their uptake by specific cell types, EV functionalization with targeting ligands is indispensable [23].

Here we exploited EVs for the target cell-specific, intracellular delivery of the recently described PGH GH15 into S. aureus-infected host cells [15]. To this end, GH15 was loaded into EVs purified from previously described mouse C51 donor cells expressing glycosylphosphatidylinositol (GPI)-anchored avidin, presenting avidin on their outer membrane amenable for surface functionalization with any type of biotinylated ligand [24]. Loading of GH15 into EVs was achieved through incubation in the presence of saponin, which facilitates transient permeabilization of the vesicle membrane [25]. Intracellular S. aureus infection typically occurs within host macrophages or endothelial cells [26,27]. To promote EV uptake by bacteria-infected host cells, PGH-loaded EVs were functionalized with biotinylated antibodies targeting α_vβ₃ integrin, a cell-surface-marker overexpressed by infected cells [28]. PGH-loaded EVs were evaluated regarding their capacity to clear intracellular S. aureus infection in cultured endothelial cells (human umbilical vein endothelial cells (HUVECs)) and macrophages (THP-1 cells) in vitro, and in a zebrafish larval in vivo model of S. aureus infection. Our findings demonstrate the potential of engineered EVs as versatile nanotherapeutics for the intracellular delivery of biologics, particularly in the treatment of intracellular bacterial infections.

2. Material and methods

2.1. Cell lines

C51 mouse colon carcinoma cells [29], stably expressing a construct driving the expression of a glycosylphosphatidyl-anchored avidin (GPI-Av) under the control of three hypoxia response elements (HRE) [30], were cultured in Dulbecco's modified Eagle's medium high glucose (DMEM, Sigma-Aldrich), containing 4.5 mg/mL glucose and supplemented with 10% heat-inactivated fetal bovine serum (FBS, GIBCO; Invitrogen) and 400 μg/mL geneticin (G-418, GIBCO; Invitrogen). To induce expression of GPI-Av, C51 cells were incubated in hypoxia (1% oxygen, 48 h) to induce HIF stabilization as described [31]. Human umbilical vein endothelial cells (HUVEC-c pooled, C-12203, PromoCell) were cultured in endothelial cells basal medium (PromoCell) supplemented with endothelial cell growth medium (GM) (PromoCell). THP-1 cells (ATCC# TIB-202) were cultured in RPMI 1640 medium (BioConcept) supplemented with 10% of heat-inactivated fetal bovine serum (FBS, GIBCO; Invitrogen) and with 1% of penicillin/streptomycin (Thermo Fisher Scientific).

2.2. Bacterial strains and culture conditions

S. aureus EGFP (pBSU100 carrying egfp) [32] were grown in tryptic soy broth (TSB, Merck), supplemented with spectinomycin dihydrochloride pentahydrate (120 μg/mL, 1002978570, Sigma-Aldrich) at 37°C. All E. coli strains used for protein expression were grown under standard cultivation conditions in Luria-Bertani (LB) medium (10 g/L tryptone, 5 g/L yeast extract, 8 g/L NaCl [pH 7.4]) and on LB agar (LB medium plus 14 g/L agar) supplemented with suitable antibiotics where necessary (100 μg/mL ampicillin, 30 μg/mL tetracycline, 50 μg/mL kanamycin).

2.3. GH15 production

Recombinant proteins were expressed in E. coli ClearColi BL21(DE3) (Lucigen, Middleton, WI, United States) and purified as previously described under endotoxin free conditions [18]. In brief, cultures were grown in modified LB medium (LB-PE; 15 g/L tryptone, 8 g/L yeast extract, 5 g/L NaCl [pH 7.8]) and cooled on ice. Protein expression was induced with 0.5 mM isopropyl-D-thiogalactopyranoside (IPTG, for 18 h at 19°C under agitation). Cells were harvested by centrifugation and frozen at −80°C. The bacterial cells pellets were resuspended in Buffer A (20 mM NaH₂PO₄, 20% glycerol [pH 7.4]) and lysed by sonication using a Bandelin Sonopuls HD 2076 (5 × 1 min intervals, 1 min breaks, 1:1 pulses, 50% power). The lysate was digested with 1 U DNase I per ml of lysate, sterile filtered and loaded onto a HiTrap Sepharose fast-flow cation exchange column (Cytiva) in an ÄKTA fast-performance liquid chromatography device (Cytiva). A gradient of 1%/min of Buffer B (1 M NaCl, 20 mM NaH₂PO₄, 20% glycerol [pH 7.4]) was used to elute the protein. The protein-containing fractions were collected, pooled and dialyzed against two changes of 1000x excess phosphate buffered saline (PBS). Purity of the constructs was assessed by SDS gel electrophoresis (Criterion TGX stain-free gel in 25 mM Tris, 192 mM glycine, 0.1% SDS, pH 8.3 running buffer) and subsequent staining with Coomassie (READYBLUE® protein gel stain, Sigma-Aldrich). Impure preparations underwent SEC in a Superdex 200 10/300 GL column (Cytiva) using gel filtration buffer (50 mM NaH₂PO₄, 500 mM NaCl, 10% glycerol [pH 7.4]) followed by dialysis against two changes of 1000 x excess PBS, sterile filtration and lyophilization.

2.4. EV isolation

EV isolation was performed as previously described [24]. C51 GPI-Av cells were allowed to adhere to 15 cm tissue culture dishes (1 × 10⁷ cells/dish; 2 h, 37°C, TPP). Cells were subsequently gently rinsed with pre-warmed PBS, before being incubated in exosome-free cell medium (DMEM, containing 4.5 mg/mL glucose (GIBCO; Thermo Fisher Scientific) supplemented with 1% exosome-depleted FBS (GIBCO; Thermo Fisher Scientific)). To induce GPI-Av expression cells were transferred to hypoxia (1% O₂, 72 h). Cellular supernatants were then collected, centrifuged (10 min, 300 g) and filtered (0.45 μm Stericup filter, Merck Millipore). Filtered supernatants were concentrated using 100 kDa centrifugal filters (Centricon Plus-70, Merck Millipore; 30 min, 4000 g, 4°C). EVs were purified by size exclusion chromatography (SEC, qEVoriginal, Izon) according to the manufacturer's procedures and, if necessary (particle concentration < 1 × 10¹⁰ particles/mL), concentrated using 100 kDa centrifugal filters (Amicon Ultra-2 mL, Merck Millipore).

2.5. Intracellular bacterial infection of THP-1 and HUVEC cells

THP-1 cells (2x10⁵ cells/well) were differentiated to M0-like macrophages with PMA (100 nM, Thermo Fisher Scientific) for 72 h. To induce proinflammatory activation, cells were incubated with lipopolysaccharide (LPS) from E. coli (100 μg/mL, Sigma-Aldrich) for 24 h. HUVEC cells (1x10⁵ cells/well) were grown on a coated surface with PureCol type I bovine collagen solution (50 μg/mL, Advanced BioMatrix) and human fibronectin solution (5 μg/cm², C-43060, PromoCell) for 24 h and subsequently treated with either tumor necrosis factor-α (TNF-α) (10 ng/mL, Sigma-Aldrich) or LPS (2.5 μg/mL, Sigma-Aldrich) for 24 h. Alternatively, cells were infected with S. aureus expressing EGFP. Overnight (ON) bacteria cultures were grown in TSB with spectinomycin dihydrochloride penta-hydrate (120 μg/mL, MedChemExpress) for 16 h and diluted to reach an optical density at 600 nm (OD₆₀₀) of 0.05. Then, bacteria were grown at 37°C to an OD₆₀₀ of 0.6, centrifuged (2′000 g, 10 min) and the pellet was resuspended in the corresponding cell medium. Bacteria were added to the cells at a multiplicity of infection (MOI) of 5 and incubated for 1 h. Next, vancomycin (5 μg/mL, 0000161542, Sigma-Aldrich) was added to the cells for 1 h and cells were washed to clear extracellular bacteria.

2.6. Immunofluorescence staining of α_vβ₃ integrin on THP-1 and HUVEC cells

For immunofluorescence staining, cells were incubated with mouse antibody anti-α_vβ₃ (LM609, Abcam, 5 μg/ml in PBS containing 10% normal-goat-serum (NGS, Thermo Fisher Scientific)) for 30 min on ice. Next, cells were paraformaldehyde (PFA)-fixed (4% in PBS, 12 min at RT, Sigma-Aldrich) and immunocomplexes were detected with goat antibody anti-mouse Alexa Fluor 488 (A11001, Invitrogen, 0.4 μg/mL in PBS with 10% NGS for 30 min at room temperature (RT)). Cell nuclei were stained with Hoechst 33342 (10 μM, Thermo Fisher Scientific). Coverslips were mounted on microscopic slides using ProLong Diamond Antifade Mountant (Invitrogen) and images acquired with a confocal laser scanning microscope (Olympus FV1000). Image analysis was performed with Fiji/ImageJ.

2.7. Flow-cytometric analysis of α_vβ₃ integrin expression on HUVEC and THP-1 cells

For flow cytometric analysis, HUVECs were grown to confluence and incubated for 24 h with TNF-α (10 ng/ml). Cells were collected by incubation with ethylenediaminetetraacetic acid (EDTA, 1 mM, Invitrogen) solution at 37°C and stained with either Alexa Fluor 488-conjugated mouse anti-α_vβ₃ antibody (FAB3050G, R&D Systems, 0.8 μg/ml in PBS with 10% NGS) or with a mouse IgG Alexa Fluor 488-conjugated isotype control antibody (IC002G, R&D Systems, 0.8 μg/ml in PBS with 10% NGS) for 30 min on ice. Next, cells were washed with ice-cold PBS (supplemented with 1 mM EDTA and 0.5% bovine serum albumin (BSA, Millipore)). Cell suspensions were added to 5 mL polystyrene tubes with CellStrainer caps (Falcon), incubated on ice and analysed on a BD FACS Aria™ IIu Cytometer. Data analysis was performed by FlowJo.

2.8. EV loading with GH15 and surface functionalization

For loading with the antimicrobial protein GH15, 5x10¹¹ isolated EVs were incubated with 500 μg GH15 in 0.1% saponin/PBS for 18 h at RT under constant rotation (PTR35, Grant Instrument). Following protein loading, EVs were functionalized for targeting by incubation with a biotinylated anti-human CD51/CD61 antibody (13-0519-82, Invitrogen; 1:100) for 30 min at RT under constant rotation. Free GH15 and unbound antibodies were subsequently removed by SEC (qEVoriginal, Izon) as described previously. GH15 loading efficiency was estimated based on enzymatic activity measurements in turbidity reduction assays (TRAs).

2.9. Turbidity reduction assays to assess antimicrobial efficacy of GH15-loaded EVs

Turbidity reduction assay were performed as previously described [16]. Briefly, a dilution series of GH15 ranging from 0.4 μg/mL to 0.00625 μg/mL was prepared. 100 μL of bacterial suspensions of frozen S. aureus USA300 substrate cells were mixed with 100 μL of each dilution of the GH15 in a 96-well plate so that the OD₆₀₀ of the mixed suspension was 1. GH15 loaded EVs were incubated with 0.1% Triton X-100 in PBS to free the encapsulated GH15 and directly added to bacterial suspensions. The decrease in optical density over time was monitored (for 1 h, 30 s intervals, with a FLUOstar Omega plate reader (BMG Labtech)).

2.10. Assessment of EV-mediated delivery of GH15 in vitro cell assays

THP-1 and HUVEC cells were bacterially infected as described above. Upon their treatment with vancomycin (5 μg/mL, Sigma-Aldrich) to eliminate extracellular bacteria, cells were incubated with functionalized GH15-loaded EVs (5x10⁸ particles) for 2 or 24 h. After their trypsin-EDTA-mediated detachment (3 min at 37°C, Sigma-Aldrich), cells were lysed with 0.1% Triton X-100 (in PBS, 3 min at 37°C, Sigma-Aldrich) and a serial dilution was performed. Cell lysates were plated on agar containing spectinomycin dihydrochloride pentahydrate (120 μL/mL, MedChemExpress) and incubated at 30°C overnight. Colony counting was performed the next day. As an alternative, intracellular bacteria were quantified by fluorescence microscopy. To this end, treated and untreated control cells were PFA-fixed (12 min, RT) and stained with Hoechst 33342 (10 μM). Images were captured by confocal microscopy (Olympus FV1000) or widefield fluorescence imaging Cytation 5 (Agilent BioTek). Image analysis and intensity quantifications were performed with Fiji/ImageJ.

2.11. Zebrafish husbandry and microinjection procedure

Adult tg(mpeg1:mCherry) zebrafish expressing mCherry in macrophages, adult tg(mpeg1:EGFP) zebrafish expressing EGFP in macrophages, adult tg(flk-1:GFP) expressing GFP in blood vessel endothelial cells were maintained under standard condition at 28°C with a 10/14 h dark/light cycle. Adult animals were set up pairwise and embryos and larvae were kept at 28°C in E2 medium (5 mM NaCl, 0.25 mM KCl, 0.5 mM MgSO₄, 0.15 mM KH₂PO₄, 0.05 mM Na₂HPO₄, 0.5 mM CaCl₂, and 0.71 mM NaHCO₃) containing 0.003% (wt/vol) 1-phenyl 2-thiourea (PTU) to prevent pigmentation for the whole experiments. Forty-eight hours post fertilization (hpf), zebrafish larvae were dechorionated and anesthetized using 0.01% (wt/vol) buffered tricaine. Prior to injection and imaging, larvae were then embedded in 0.2% (wt/vol) buffered agarose supplemented with 0.01% (wt/vol) tricaine for immobilization. Injection needles were produced using 1.2 × 0.69 mm x 10 cm capillaries (GBF120-69-10, WPI) and a pipette puller (PUL-1000, WPI). The zebrafish larvae were individually injected using a micromanipulator (M3301R, WPI), a pneumatic PicoPump (PV830, WPI), and a Leica microscope (504941, WPI). Larvae were injected systemically into the Duct of Cuvier. Zebrafish husbandry and procedures were performed in accordance with Swiss animal welfare regulations.

2.12. Bacterial infection model in zebrafish larvae

S. aureus expressing EGFP overnight cultures were grown in tryptic soy broth (TSB, Merck), supplemented with spectinomycin dihydrochloride pentahydrate (120 μL/mL, MedChemExpress) and diluted to an OD₆₀₀ of 0.05. Subsequently, bacteria were re-cultured to an OD₆₀₀ of 0.6, centrifuged and resuspended in PBS to reach an estimated concentration of 12.5x10¹⁰ CFU/mL for the bacterial infection model and an estimated concentration of 5x10¹⁰ CFU/mL for the intracellular bacterial infection model. Tg(mpeg1:mCherry) larvae were injected with 1 nl bacterial suspension into the Duct of Cuvier and incubated for 2 h at 28°C. Next, larvae were treated with vancomycin (7.4 ng/fish, Sigma-Aldrich), GH15 (7.4 ng/fish) or GH15-EVs (2.0x10¹¹ particle/mL) by injecting 2 nl into the Duct of Cuvier. Bacterial reduction was monitored via live imaging at widefield fluorescence imaging Cytation 5 (Agilent, BioTek).

2.13. Biodistribution of EVs in zebrafish larvae model

Isolated EVs were labelled with CellMask deep red (5 μg/mL, 30 min at 37°C, Invitrogen). Free dye was removed by SEC (qEVoriginal, Izon) as described previously. Labelled tg(flk-1:GFP) or tg(mpeg1:EGFP) larvae were injected with 5 nl of labelled EVs (1.3x10¹¹ particle/mL) into the Duct of Cuvier. EV biodistribution was monitored on a multiphoton microscope (Miltenyi Biotec).

2.14. Statistical analysis

Statistical analyses of the data were performed with Prism 8.0 (GraphPad Software Inc.) by using an unpaired t-test with Welch's correction for all p-values. Results are shown as mean ± SEM.

3. Results

3.1. Saponin treatment enhances the loading of the engineered endolysin GH15 into extracellular vesicles (EVs) derived from C51 cells

We have previously established a versatile and modular platform for EV surface functionalization aimed at drug-delivery applications, based on the GPI-Av system. EVs isolated from C51 donor cells genetically engineered to express a glycosylphosphatidylinositol-anchored avidin (GPI-Av) present avidin on their surface, allowing flexible and efficient post-isolation modification through the avidin-biotin interaction [24]. In the present study, we set out to explore whether this platform can be applied to the intracellular delivery of novel antimicrobial proteins into bacteria-infected host cells. Specifically, we aimed to encapsulate GH15, which demonstrated efficacy against methicillin-resistant Staphylococcus aureus (MRSA) [15], but poor eukaryotic membrane penetration. A schematic illustration of the GH15 domain architecture is provided in Fig. 1A. Three approaches for loading of the therapeutic proteins into EVs were compared: passive incubation, saponin permeabilization, and sonication. The fluorescent protein EGFP, fused to the cell penetrating peptide (CPP) Trans-Activator of Transcription (TAT) [33], termed TAT-EGFP, served as a proxy for GH15, and was loaded into EVs isolated from mouse C51 cells, stably expressing GPI-Av on their surface [24]. Among the different loading approaches explored, EV and TAT-EGFP co-incubation in the presence of 0.1% saponin for 14 h yielded the highest efficient protein incorporation, as confirmed by fluorescence measurement (Fig. S1B). This method was hence used to load GH15 into EVs, isolated from C51 donor cells, expressing GPI-Av (Fig. 1B). EV size distribution and morphology was assessed by Nanoparticle Tracking Analysis (NTA) and cryo-transmission electron microscopy (cryo-TEM) and revealed phospholipid bilayer enclosed nanovesicles with a median diameter of ca. 130 nm (Fig. S2A and D). Post-loading and functionalization EVs maintained their size distribution and stability, as confirmed by dynamic light scattering (DLS) and nanoparticle tracking analysis (NTA). An increase in ζ-potential was observed for modified EVs, yet the surface charge remained in the negative range of −10 mV (Fig. S2B and C). To estimate loading efficiency and assess the antimicrobial activity of GH15 following EVs encapsulation, turbidity reduction assays were performed. GH15 was either released from EVs by triton-mediated lysis or directly added to bacterial suspensions at defined concentrations in PBS containing 0.1% of Triton X-100. GH15 recovered from EVs reduced bacterial growth with similar efficacy as GH15 added directly to bacterial suspensions, indicating that GH15 retains its activity after encapsulation into EVs (Fig. 1C). Moreover, intact GH15-EVs did not show any significant reduction of optical density in turbidity reduction assays (TRAs), indicating that there is no extravesicular GH15 present and no major leakage of the enzyme out of the EVs during (short-term, 3 days) storage at 4°C. The biosafety profiles of unloaded EVs, GH15-EVs and free GH15 were evaluated using an MTT cell viability assay. EVs and GH15-loaded EVs did not alter the viability of THP-1 cells (Fig. S4A). Consistently, free GH15 exhibited no cytotoxic effects on either THP-1 cells or normal human dermal fibroblast (Fig. S4B and C). In a next step, we set out to determine which target cell receptor could be explored for directing GH15-loaded EVs to S. aureus infected endothelial cells and/or macrophages.

Fig. 1.

GH15 retains activity upon encapsulation within EVs. (A) Schematic representation of the architecture of GH15, including its molecular weight (MW) and length in amino acids and SDS PAGE showing the purified GH15. (B) Schematic representation of the procedure established to load GH15 into EVs isolated from mouse C51 cells, expressing GPI-Av, in presence of 0.1% saponin/PBS. Created with BioRender. (C) Turbidity reduction assay (TRA) of GH15 added to bacterial (S.aureus USA300) suspensions at different concentrations or upon triton-mediated release from EVs. A significant reduction of optical density was observed in the presence of GH15, with higher doses resulting in stronger inhibition of bacterial growth. Addition of GH15 released from EVs by triton treatment to bacterial suspensions resulted in a significant reduction of optical density, indicating encapsulated GH15 to retain its antimicrobial activity.

3.2. α_vβ₃ integrin is upregulated on endothelial cells and macrophages upon bacterial infection or inflammatory stimulation

Previous reports suggest α_vβ₃ integrin, a major cell adhesion molecule expressed on endothelial cells and leukocytes [28], to be upregulated in response to pro-inflammatory stimuli or upon bacterial infection [34,35]. To verify these results in the cell models utilized throughout this report, we studied the expression of α_vβ₃ integrin on endothelial cells (HUVECs) and THP-1 monocytes, exposed to pro-inflammatory stimuli (TNF-α or LPS, 24 h) or in response to S. aureus infection. HUVECs and THP-1 were treated with TNF-α (10 ng/mL, HUVECs only) or LPS (2.5 μg/mL and 100 μg/mL respectively), as well as infection with S. aureus for 1 h (MOI 5), led to a twofold upregulation of α_vβ₃ integrin expression in immunofluorescence staining (Fig. 2A, Fig. S5) and flow cytometry (Fig. 2B) experiments. Increased α_vβ₃ surface expression levels were detected in cells treated with pro-inflammatory stimuli or exposed to bacterial infection compared to untreated controls. These results support the use of α_vβ₃ as suitable receptor for directing drug-loaded EVs into S. aureus infected cells.

Fig. 2.

S. aureus infection of macrophage (THP-1) or endothelial cells (HUVECs) results in an increased expression of α_vβ₃ integrin. (A) Immunofluorescence staining of endothelial cells (HUVECs) and macrophage-like cells (THP-1) treated with LPS (2.5 μg/mL and 100 μg/mL respectively) or infected by EGFP expressing S. aureus. LPS treatment or bacterial infection led to an upregulation of α_vβ₃ integrin on the cell surface. (B) Flow cytometry of HUVEC cells after treatment with TNFα (10 ng/mL) or PBS as vehicle control confirmed an inflammation-mediated upregulation of α_vβ₃ integrin. Histograms and geometric means are shown (n = 3). Bar, 30 μm.

3.3. GH15-loaded GPI-Av EVs, functionalized with an α_vβ₃ integrin antibody, are selectively internalized by S. aureus-infected cells and effectively eradicate intracellular bacteria

We investigated to what extent GPI-Av EVs, loaded with GH15 and functionalized with an antibody targeting α_vβ₃ integrins on bacteria-infected cells, could kill intracellular bacteria. HUVEC and THP-1 cells were infected with S. aureus expressing the fluorescent reporter EGFP. 1 h after infection, extracellular bacteria were eliminated using vancomycin (5 μg/mL) for 1 h, after which the medium was replaced. Subsequently, cells were treated with GH15 loaded EVs and control EVs for 2 h. Fixed cells were analysed by confocal microscopy, and a significant reduction of the bacterial fluorescence intensity was observed for the cell treated with GH15-loaded EVs compared to empty control EVs (Fig. 3A and B). Cells treated with antibody-functionalized GPI-Av EVs loaded with GH15 further reduced intracellular fluorescence intensities compared to GH15-loaded GPI-Av EVs, functionalized with biotin only or PBS control treatment (Fig. 3C). These results suggest that EVs effectively deliver GH15 into bacteria-infected cells, and that this intracellular delivery is further enhanced by functionalizing EV surfaces with antibodies specific for the infected cells. To further confirm the antimicrobial effect of GH15-loaded EVs within infected cells, quantification of bacterial loads was performed by plating extracellular supernatants and intracellular cell contents on agar 24 h post EV addition (Fig. 3D and E). In line with the image-quantification results, a significant reduction in colony-forming units (CFU) was observed for intracellular bacterial loads in infected cells treated with antibody-functionalized GH15-EVs compared with the respective controls. Interestingly, extracellular bacteria were detected despite vancomycin treatment. Their abundance was similarly reduced following treatment of cells with GH15-loaded, antibody-functionalized extracellular vesicles (EVs). These extracellular bacteria likely originated from intracellular reservoirs and were released upon host cell apoptosis or necrosis.

Fig. 3.

Functionalized GH15 loaded EVs reduce intracellular S. aureus in vitro. (A, B, C) HUVECs were infected with EGFP-S. aureus for 1 h and later treated with vancomycin (5 μg/mL) for 1 h to eliminate extracellular bacteria in cellular supernatants. Subsequently, cells were incubated with empty EVs, GH15-EVs functionalized with biotin only, a biotinylated anti-α_vβ₃ antibody (5x10⁸ particle) or PBS for 2 h. Upon fixation of cells, confocal images were acquired to analyse intracellular bacterial loads. EGFP fluorescence intensity was significantly decreased when cells were treated with GH15-EVs functionalized with biotin or the anti-α_vβ₃ antibody. (D and E)In vitro antimicrobial efficacy of GH15-EVs (5x10⁸ particle) was assessed by plating extracellular supernatants as well as cell lysates (for intracellular cell content) after 24 h of EV incubation. N = 3, one representative dataset is shown (number of replicates = 4; statistics: unpaired t-test with Welch's correction with p value marked: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, mean ± SEM). Bar, 100 μm (overviews) and 30 μm (inserts).

In conclusion, encapsulation of GH15 within EVs, combined with surface functionalization using a target cell-specific antibody, enables efficient intracellular delivery to bacteria infected cells and promotes the elimination of intracellular bacteria.

3.4. An in vivo zebrafish larval model of S. aureus infection shows intracellular bacteria escaping GH15 treatment

For validation in vivo, we employed a previously established zebrafish larval model of bacterial infection [12]. 2 h after systemic injection of EGFP-labelled bacteria into tg(mpeg1:mCherry) zebrafish larvae that are stably expressing mCherry in macrophages, the larvae were treated with either vancomycin (7.4 ng/fish larva) or GH15 (7.4 ng/fish larva). Bacterial distribution and burden within fish larvae was monitored for 19 h using widefield fluorescence microscopic imaging. Both antimicrobial therapies resulted in a significantly reduced bacterial burden indicated by reduced EGFP fluorescence intensities (Fig. 4B and C, Movie S7, 8, 9) and increased larval survival (Fig. 4D). However, high resolution fluorescence imaging at experimental endpoints at 13.5 h post injection revealed persistent bacteria surviving therapy within macrophages (Fig. 4B, arrowheads). The zebrafish larval infection model thus recapitulates key features of human disease, including the presence of intracellular bacteria residing within macrophages, where they are able to evade antimicrobial therapy. This highlights the need for a biodistribution study after systemic administration to determine whether EVs reach their intended targets within the fish larvae.

Fig. 4.

GH15 increased survival of bacteria-infected zebrafish larvae. (A) Zebrafish larvae tg(mpeg1:mCherry) 48 hpf were systemically injected with EGFP-positive S. aureus (12.5x10¹⁰ CFU/mL). 2 h later, the fish larvae were treated with PBS as vehicle (n = 8), vancomycin (7.4 ng/fish, n = 12) or GH15 (7.4 ng/fish, n = 12). (B) Live images acquired at 2 h intervals allowed to monitor the bacterial distribution and burden of EGFP positive bacteria within the larvae. Magnification images of the tail indicate the presence of intracellular bacteria in macrophages. (C) EGFP fluorescence intensity was significantly decreased when larvae were treated with either vancomycin or GH15. Statistics, unpaired t-test with Welch's correction, (**p < 0.01, ***p < 0.001, mean ± SEM). (D) Survival of infected larvae, injected with PBS as vehicle control (n = 14), vancomycin (n = 65) and GH15 (n = 54). Curve comparison was performed using the log-rank test (****p < 0.0001). Bar 500 μm (overviews) and 100 μm (inserts). Graphical figure created in BioRender.

3.5. EVs are actively taken up by macrophages in zebrafish larvae

To assess in vivo biodistribution patterns of not functionalized EVs within zebrafish larvae, EVs were labelled with CellMask deep red, a cell membrane fluorescent dye, and injected into i) tg(flk-1:GFP) fish larvae, expressing GFP in blood endothelial cells and ii) tg(mpeg1:EGFP) larvae, stably expressing EGFP in macrophages. Labelled EVs were injected systemically (via the Duct of Cuvier) in 48 hpf larvae and their distribution imaged using multiphoton microscopy (Fig. 5). Images acquired 15 min post injection showed the arrest of EVs in the venous plexus and their fast uptake by endothelial cells (Fig. 5B, Movie S10). However, after 20 h, EVs appeared to accumulate in cellular compartments not colocalized with endothelial cells indicating the uptake of EVs by another cell type. Indeed, analysing EV biodistribution within tg(mpeg1:EGFP) fish larvae showed EV fluorescence signals to colocalize with EGFP positive macrophages, suggesting EV endocytosis into these cells. Uptake of EVs by macrophages was detected as early as 1 h post injection (Movie S11), but it was only at late timepoints (20 h post injection) that EVs showed significant accumulation in macrophages (Fig. 5C). Systemic injection of EVs into fish larvae thus results in EV accumulation within monocytes and macrophages, passively directing therapeutic content such as GH15 into these cells.

Fig. 5.

EVs are actively taken up by endothelial cells and macrophages in zebrafish larvae. (A) CellMask deep red labelled EVs (1.3x10¹¹ particles/mL) were injected intravenously into 48 h old transgenic zebrafish larvae, expressing GFP in endothelial cells or EGFP in macrophages. (B) Multiphoton images of distribution of EVs revealed rapid EV uptake by endothelial cells, as early as 15 min post injection. (C) 20 h post EV administration, EVs showed prominent uptake within macrophages. Second harmonics generation (SHG) imaging was used for anatomical orientation and indicated tail muscle fibres of the zebrafish. Bar 100 μm (overviews) & 10 μm (inserts). Graphical figure created in BioRender.

3.6. GH15-loaded EVs can significantly reduce bacterial burden within fish macrophages

After observing that systemically injected EVs accumulate in the monocyte/macrophage compartment of 48 h-old zebrafish larvae, we investigated whether EV-delivered GH15 promotes bacterial clearance within macrophages in infected larvae. To this end, tg(mpeg1:mCherry) zebrafish larvae, were infected with S. aureus expressing EGFP by systemic injection of bacteria into the fish larvae 48 hpf. As previous experiments in fish larvae demonstrated that low-dose bacterial infection leads to predominant bacterial accumulation within macrophages and endothelial cells at early time points (Fig. S6), only 5 × 10³ CFUs were injected in this experiment. The proliferation of bacteria and the number of bacteria within the macrophage compartment was determined by multiphoton microscopic imaging of the fish tail region. The injection of EVs, loaded with GH15, resulted in a significant reduction of EGFP fluorescence intensity per macrophage, compared to treatment with empty (unloaded) EVs or PBS (vehicle control), 24 h post injection (Fig. 6B and C). GH15-loaded EVs thus effectively reduce intracellular bacterial burden in an in vivo model of S. aureus infection.

Fig. 6.

GH15-loaded EVs reduce intracellular S. aureus in vivo. (A) Experimental setup: zebrafish larvae tg(mpeg1:mCherry) were systemically injected with EGFP expressing S. aureus (5x10³ CFU/mL) and treated with PBS as vehicle, EVs (4.0x10⁵ particles) or GH15- loaded EVs (4.0x10⁵ particles) 2 h post infection. Images were analysed by segmenting macrophages. EGFP fluorescence was analysed within individual macrophages. (B) Live images of the tail region showed the presence of EGFP positive bacteria within endogenous macrophages, persisting over time. (C) EGFP fluorescence intensity within endogenous macrophages was significantly decreased when fish larvae were treated with GH15-loaded EVs compared to unloaded EV or vehicle controls. Number of larvae per group, 14; number of macrophages analysed per larvae, 50-200. Macrophage fluorescence intensities were normalized to intensities measured at 0.3 h post injection. Data shown as means of 3 independent experiments, statistics, unpaired t-test with Welch's correction (**p < 0.01, mean ± SEM). Bar, 200 μm (overviews), 30 μm (inserts). Graphical figure created in BioRender.

4. Discussion

The increasing prevalence of antibiotic-resistant pathogenic bacteria poses a critical global health threat and is predicted to become one of the leading causes of death by 2050 [2]. Novel, bacteriophage-derived antimicrobials, including genetically engineered endolysins, have the potential to overcome resistant strains. Yet their clinical translation is currently hampered by challenges relating to immunogenicity, short plasma half-life and, in particular, poor membrane permeability, preventing effective treatment of intracellular bacterial infections [13], [36]. These are caused by a number of pathogenic bacteria including certain S. aureus, L. monocytogenes and E. coli strains, which can persist within host cells, thereby escaping antibiotic therapy [13], [37]. Genetic engineering of endolysins offers novel strategies to overcome several limitations associated with their therapeutic application [13], [38]. In an attempt to improve the cell permeability of novel engineered endolysins to eliminate intracellular bacteria, Roehrig et al. fused peptidoglycan hydrolases (PGHs), pre-selected to show high activity within mammalian cells, to different cell penetrating peptides (CPP) and thereby achieved effective lysis (up to 4.5 log units) of intracellular S. aureus [16]. Despite their considerable promise in treating intracellular bacterial infections, CPP protein fusions still present notable limitations including their intrinsic proteolytic instability. As such, these fusion constructs are highly susceptible to proteolytic degradation by both intracellular and extracellular proteases, which shortens their plasma half-life and hinders the attainment of effective therapeutic concentrations at the target tissue [39]. Further engineering of endolysins through fusion with albumin-binding domains or tissue-specific targeting peptides can overcome these limitations by substantially prolonging blood circulation time and enhancing accumulation in specific target tissues [17,18]. However, such modifications require laborious, case-by-case evaluation of their effects on enzymatic activity and therefore cannot readily be implemented as a universal platform technology.

To avoid tedious and complex genetic engineering while preventing premature protein degradation and maximizing therapeutic concentrations at target tissues, we sought to encapsulate GH15, a potent genetically engineered endolysin with demonstrated activity against intracellular S. aureus [15,16] in extracellular vesicles (EVs) as drug nanocarriers. GH15 was loaded into EVs, derived from GPI-Av expressing C51 donor cells, in the presence of saponin for 18 h. Our data indicate that empty extracellular vesicles (EVs) remain stable for up to 11 weeks and that GH15 encapsulated within EVs retains enzymatic activity for several weeks after loading, as demonstrated by turbidity reduction assays (TRAs). Moreover, we have analysed GH15-EVs suspensions regarding extravesicular GH15 activity after 3 days of storage (4°C) via TRAs and not detected any significant enzyme activity at this time (Fig. S3). This suggests no major enzyme leakage within short time periods. Yet for future pharmaceutical applications, potential drug leakage from the EVs and the long-term stability of EV-encapsulated GH15 during storage at −80°C has to be systematically evaluated.

Functionalization of GH15-loaded, GPI-Av EVs with an antibody targeting α_vβ₃ integrin, promoted effective intracellular delivery of GH15 and successful killing of intracellular bacteria in endothelial (HUVECs) or macrophage (THP-1) cells in vitro. In vivo, in zebrafish larvae, passive targeting by non-functionalized GH15-loaded GPI-Av EVs was sufficient to reduce intracellular bacterial loads within macrophages. This observations suggests that, for macrophages, which are part of the mononuclear phagocyte system (MPS) and known to efficiently take up nanoparticulate drug formulations [40], the intrinsic cell targeting properties of EVs may be sufficient to achieve therapeutic intracellular delivery and antibacterial efficacy. However, it remains to be determined to what extent this effect can be further enhanced through EV functionalization with antibodies specific for infected fish macrophages.

The encapsulation of bacteriophage-derived endolysins within phospholipid-based nanoparticles, specifically the synthetic counterparts of EVs, i.e. liposomes, has been analysed before. Liposomal endolysin formulations promoted an increased stability of encapsulated endolysins [41,42], in line with literature reports suggesting liposomes to improve the in vivo stability and overall pharmacokinetic properties of any therapeutic protein cargo they carry [43]. At the same time, liposome uptake by the mononuclear phagocyte system (MPS) can also be exploited for their passive, in vivo targeting to macrophages. Indeed, Bartlett et al. successfully delivered a phage-derived enzymatic cocktail, including two endolysins, into Mycobacteroides abscessus- and Mycobacterium avium-infected macrophages ex vivo using macrophage-targeted liposomes, thereby rescuing these cells from mycobacteria-induced necrosis [41].

In contrast to phospholipid-based liposomes, EVs only show minor immunogenicity and are fully biocompatible [44]. In addition, they have been suggested to be more effectively taken up by macrophages or target cells than synthetic liposomes [45].

Sharing important physiological processes and molecular pathways with humans and recapitulating important features of human infectious disease, including intracellular bacteria, zebrafish larvae have become an attractive model for the characterization of novel antimicrobial drug products [46]. The model is particularly suited for analysing the biodistribution and efficacy of novel antimicrobials due to its transparency at early larval stages allowing real-time microscopic visualization of drug distribution and bacterial spread [47]. However, several limitations should be considered when interpreting these findings. At the larval stage, the adaptive immune system is not fully developed, and host-pathogen interactions therefore primarily reflect innate immune responses. Furthermore, zebrafish are maintained at 28°C, which may influence bacterial virulence and disease progression. Differences in anatomy, metabolism, and drug pharmacokinetics (PK) may affect the absorption, distribution, metabolism, and excretion of therapeutics compared with mammals [46,48]. Thus, although the present results provide first in vivo proof-of-concept for the antibacterial activity of GH15-loaded EVs, further validation in mammalian infection models will be required.

Strikingly, surface functionalization of GH15-loaded EVs with an antibody targeting α_vβ₃ on bacteria-infected cells, further increased the efficacy of intracellular bacterial killing in vitro. This suggests that EVs are highly suitable vehicles for intracellular delivery of novel antimicrobials to infected endothelial cells and macrophages, and that tailoring their surface composition could further enhance the efficiency of cargo delivery. Although we have demonstrated increased α_vβ₃ integrin expression following S. aureus infection of endothelial cells and macrophages, α_vβ₃ is not exclusively expressed on bacterially infected cells. This integrin family member has also been reported to be present on inflammation-activated endothelial cells and macrophages as well as certain leukocyte populations and osteoclasts [49,50]. Consequently, functionalization of EV surfaces with α_vβ₃-binding antibodies may promote interactions with non-infected α_vβ₃-expressing cells and tissues. Our MTT assays performed on non-infected THP-1 cells incubated with empty EVs, GH15-loaded EVs, and GH15 alone suggest that the treatments do not induce detectable cytotoxicity in this particular cell type (Suppl. Fig. 4). However, these experiments do not directly address potential safety concerns associated with EV biodistribution and off-target accumulation following their systemic in vivo administration. Previous experiments in animal models indicate EVs to predominantly accumulate in the liver, spleen and lungs upon systemic administration [51]. Overall, EV-based formulations have, in experimental animals, been associated with favourable safety profiles and good tolerability with only minimal signs of toxicity [[52], [53], [54]]. Nevertheless, biodistribution and safety characteristics of each formulation must be assessed individually. Therefore, further PK and biodistribution studies in mammalian infection models will be required to evaluate the biodistribution, off-target accumulation, and long-term safety of α_vβ₃-targeted EVs following systemic administration.

In earlier experiments, we have observed fluorescently labelled C51 GPI-Av EVs to be internalized by endocytosis and endocytosis-independent mechanisms [24]. Following cellular uptake by endocytosis, EVs will traffic through the endolysosomal system, where vesicle degradation and membrane remodeling may expose luminal cargo to the endosomal or phagosomal compartment in macrophages. Direct fusion of EVs with the plasma membrane or the endosome membrane, may promote release of EV cargo proteins in the cytosol [42]. With S. aureus persisting inside the endosomal compartment within professional phagocytes [40] and replicating within phagosomes or the cytosol in non-professional phagocytes [41], EV-mediated delivery is expected to promote GH15 accumulation at intracellular sites relevant to bacterial persistence.

Our platform is based on the avidin-biotin system to achieve efficient and versatile surface functionalization for targeted delivery. Donor cells stably express glycosylphosphatidylinositol-anchored avidin [30] which allows the isolation of EVs displaying avidin on their outer membrane [24]. For our initial proof of concept experiments, GPI-avidin C51 colon cancer cells were utilized to generate EVs. These cells were primarily selected as EV donor cells based on practical considerations. Specifically, we had already established our functionalization platform in this model, which is easy to culture and to expand, making it well suited for the initial validation of our delivery platform. However, due to their murine origin and tumorigenic nature, C51 cells are strictly constrained to early stage proof-of-concept and are not suited for clinical translation [55]. For future translational and clinical development, a non-tumorigenic alternative should be used. Mesenchymal stem cell (MSCs) are particularly attractive for this purpose, given their high EV production yield, high biocompatibility and intrinsic immunosuppressive properties [56].

In summary, we demonstrate extracellular vesicles (EVs) to constitute a promising platform for the intracellular delivery of antimicrobial proteins, enabling efficient elimination of bacteria within infected macrophages and endothelial cells both in vitro and in a zebrafish larval infection model in vivo. EVs may also serve as effective carriers for the intracellular delivery of other antimicrobial agents, including glycopeptide antibiotics. Collectively, our findings underscore the potential of EVs as versatile and adaptable vehicles for targeted intracellular delivery of antimicrobial therapeutics. The delivery vehicle characterized in this study is likely to open new therapeutic avenues for the target cell-specific, intracellular delivery of drugs in other disease contexts.

CRediT authorship contribution statement

Besmira Sabani: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. Martina Leone: Investigation, Methodology, Visualization. Lynn Gasser: Data curation, Investigation, Methodology, Visualization. Eric Sumrall: Data curation, Formal analysis, Investigation, Methodology. Susanne Meile: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Joelle Inderbitzin: Data curation, Formal analysis, Investigation, Methodology, Visualization. Daniel Razansky: Conceptualization, Funding acquisition, Investigation, Writing – review & editing. Mathias Schmelcher: Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Supervision, Writing – review & editing. Martin J. Loessner: Conceptualization, Funding acquisition, Investigation, Supervision, Writing – review & editing. Rainer Riedl: Conceptualization, Funding acquisition, Investigation, Supervision, Writing – review & editing. Steffi Lehmann: Conceptualization, Data curation, Funding acquisition, Resources, Supervision, Writing – original draft, Writing – review & editing.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the authors used Microsoft Copilot to assist with grammar and spelling correction. The authors reviewed and edited all AI-generated suggestions as appropriate and take full responsibility for the content of the published article.

Funding

This work was supported by Innosuisse – Swiss Innovation Agency (Project no.: 59414.1 IP-LS) and by the Swiss 3R Competence Center (3RCC).

Declaration of competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

figs1.

figs2.

figs3.

figs4.

figs5.

figs6.

Data availability

Data will be made available on request.