Extracellular Vesicle‐Bound Bacterial Toxin Pneumolysin Triggers Membrane Engagement and Damage Beyond Canonical Pore Formation

ABSTRACT

Pneumolysin (PLY) is a cholesterol‐dependent pore‐forming toxin and a key virulence factor of Streptococcus pneumoniae, the leading cause of pneumonia worldwide in children aged below 5 years. At sublytic toxin doses, host cells shed PLY‐laden extracellular vesicles (EVs) during the membrane repair response. However, it remains unclear how these toxin‐bearing EVs engage and damage target cell membranes. Here, we combine molecular dynamics simulations, liposome fusion assays and cell‐based experiments to elucidate the membrane interaction potential of vesicle‐bound PLY. Simulations indicate that EV‐embedded PLY uses an exposed helix to bind target cell membranes, inducing pronounced curvature, bilayer thinning and water influx. Liposome fusion assays by 3D confocal imaging and Cryo‐EM analysis demonstrate that both PLY and membrane cholesterol promote vesicle‐membrane interactions. Characterisation of vesicle subpopulations released from PLY‐challenged monocytes by western blotting and immunogold electron microscopy demonstrated that plasma membrane‐derived microvesicles are preferentially enriched in membrane‐bound PLY compared to small extracellular vesicles. Consistent with these findings, MVs purified from wild‐type PLY‐challenged monocytes, fuse with human peripheral blood mononuclear cells, delivering toxin and causing membrane damage, which was significantly lower with the toxoid mutant, PLYW433F. Among PBMCs, CD4⁺ T cells showed higher PLY positivity upon MV co‐incubation and higher cell death compared to monocytes at same dose. Together, our results reveal a noncanonical mode of toxin dissemination, in which vesicle‐bound pneumolysin fuses and destabilises target cell membranes, representing an alternative pathway for EV‐mediated toxin activity and a potential target for therapeutic intervention.

Bacterial toxin hijacks host microvesicles to breach cell membranes Microvesicles derived from challenged host cells carrying the pneumococcal pore‐forming toxin pneumolysin promote fusion with target cells and membrane permeabilisation. Our integrated experimental and computational study reveals a mechanism by which PLY‐lipid interactions enable microvesicle‐mediated propagation of toxin activity, potentially amplifying membrane damage and immune cell dysfunction during pneumococcal infection.

1. Introduction

Streptococcus pneumoniae is a leading cause of community‐acquired pneumonia and a major contributor to morbidity and mortality worldwide, particularly among children under 5 years of age (Infections and Antimicrobial Resistance 2024; Kong et al. 2025). In 2024, the WHO classified pneumococci as one of 15 priority pathogens for which new antibiotics are urgently needed. Due to rapidly increasing antibiotic resistance, new antimicrobial strategies targeting both bacteria and the host need to be developed. Pneumolysin (PLY), a pore‐forming toxin of the cholesterol‐dependent cytolysin family, is a major pneumococcal virulence determinant. It binds to cholesterol‐containing host cell membranes, oligomerises into large prepore assemblies and undergoes a dramatic conformational transition to form β‐barrel transmembrane pores that drive cytolysis (Tilley et al. 2005; van Pee et al. 2017; Vogele et al. 2019). PLY is organised into four structurally and functionally distinct domains D1–D4, whose coordinated rearrangements drive the structural transition from soluble monomers to membrane‐inserted β‐barrel pores (Marshall et al. 2015). The membrane‐binding domain D4 has conserved undecapeptide residues that mediate the initial, cholesterol‐dependent binding to target bilayers. Domain 3 comprises the α‐helical bundles that refold into amphipathic β‐hairpins to form the transmembrane β‐barrel during pore insertion (Tilley et al. 2005). Domains 1 and 2 form the scaffold that mediates monomer‐monomer contacts during prepore assembly and act as a mechanical linkage that transmits membrane engagement to the D3 refolding event. This mechanism explains the toxin activity when PLY oligomerises on the target cell membrane from soluble monomers. However, increasing evidence indicates that during infection, PLY may also operate outside this classical paradigm. At sublytic concentrations, host cells activate intrinsic plasma‐membrane repair pathways that sequester and remove assembled pore complexes via calcium‐dependent membrane remodelling and microvesicle shedding (Wolfmeier et al. 2016).

The susceptibility of immune cells to PLY has been previously shown to be dependent on the microvesicle shedding capacity (Larpin et al. 2020; Koffel et al. 2018). Extracellular vesicles (EVs) represent a heterogeneous population of membrane‐bound particles ranging from the nano‐ to micron‐scale particles released by cells. They include small endosome‐derived exosomes (∼30–150 nm) and larger plasma membrane‐derived microvesicles (∼0.1–1 µm), both of which encapsulate biological cargo such as proteins, lipids and nucleic acids that modulate recipient cell function and mediate intercellular signalling (Jeppesen et al. 2023). A recent study from our group demonstrated that PLY challenge markedly upregulates the release of host EVs enriched in inflammatory cargo that mediate PLY‐dependent lung injury in infected mice (Parveen et al. 2024). These studies establish that EV shedding is a prominent host response during pneumococcal infection and indicate that host EVs can carry bacterial toxin in vivo. However, the relative distribution of toxin in EV subpopulations and the membrane interactions of vesicle‐bound PLY were not analysed in these studies.

Together, these observations raise the possibility that PLY presented on host‐derived vesicles may represent more than a passive byproduct of cell membrane repair. Instead, vesicle‐associated PLY may constitute a distinct, membrane‐active form of the toxin, capable of engaging recipient membranes in ways that differ fundamentally from pore formation by soluble monomers. Because PLY oligomers embedded in a curved vesicular membrane expose a different set of protein surfaces to the target cell membrane, vesicle presentation has the potential to alter toxin–lipid interactions, membrane deformation and downstream cellular outcomes. Such a mechanism would provide a means for lateral toxin dissemination and bystander membrane damage, extending the impact of PLY beyond the primary target cell.

In this study, we test this hypothesis by combining in silico coarse‐grained molecular dynamics simulations, in vitro reconstituted liposome fusion assays and functional analyses in human peripheral blood mononuclear cells exposed to purified monocyte‐derived PLY‐containing microvesicles. We demonstrate that PLY is more enriched in plasma membrane‐derived MVs relative to small extracellular vesicles (sEVs) and localised on the vesicle membrane. Computational simulations indicated that vesicle‐bound PLY acts as a mediator of cholesterol‐dependent membrane engagement through interactions of the exposed helix with membrane lipids, inducing membrane curvature, thinning and permeability, and facilitating toxin transfer to recipient cells. Liposome fusion, cellular fusion and membrane damage assays in human PBMCs using wild‐type and mutant toxoid concur with the computational model. These findings support a revised view of pneumolysin biology in which EVs serve as platforms for the dissemination of membrane‐active toxin, with important implications for pneumococcal pathogenesis and the development of host‐directed therapeutic strategies.

2. Results

2.1. Computational Simulation of Interactions Between Vesicle‐Embedded Pneumolysin and Lipid Bilayer

First, we performed in silico studies by simulating how vesicle‐embedded PLY engages a target cell membrane. To capture this encounter in a controlled yet biomimetic setting, we built a vesicle‐bilayer model system (Figure 1). The model comprises a coarse‐grained representation (see Section 4 for details) of a 20 nm vesicle with an exosome‐mimicking lipid composition (Cholesterol (CHOL) 40%, Dipalmitoylphosphatidylcholine (DPPC) 30%, Dioleoylphosphatidylserine  (DOPS) 20%, Dipalmitoyl‐sphingomyelin (DPSM) 10%) (Sakai‐Kato et al. 2020). The vesicle was embedded with two PLY pentamers in membrane inserted confirmation that was derived from the reported Cryo‐EM structures of PLY pore complex (van Pee et al. 2017; Vogele et al. 2019) that were oriented towards a lipid bilayer comprising 1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phosphocholine (POPC) 80% and CHOL 20% mimicking a general plasma membrane (Rog and Pasenkiewicz‐Gierula 2006b). The PLY pentamer structure was used since it was reported to be the minimal stable functional oligomeric form that still reflects the geometry and interactions necessary for pore initiation (Vogele et al. 2019). The orientation of PLY on the vesicle was fixed with the cholesterol‐binding D4‐domain docked onto the vesicle, and D1 opposed to the bilayer membrane, mimicking a vesicle containing cell‐extruded PLY oligomers (Figure 1). The PLY‐D4 domain contains the undecapeptide loop essential for pore formation on the budding membrane. The D1 domain of PLY contains several α‐helices. α‐helical motifs are widely recognised for their capacity to promote membrane interactions by stabilising protein‐lipid contacts and facilitating insertion into lipid bilayers (Kabelka and Vacha 2021). We hypothesised that the alpha helices in PLY‐D1 extending outward from the vesicle may play a role in mediating interactions with lipid bilayer. The control system consisted of a vesicle‐bilayer system without any embedded PLY oligomers (Figure S1 and Table S1). Both systems were subjected to a multi‐microsecond‐long MD simulation. In the control system, no prominent changes in the bilayer or the vesicle were observed during the simulation period of 8 µs (Figure 2A). Biophysically, two opposing membranes are unlikely to fuse in the absence of fusion proteins due to the repulsion of the associated water molecules (Baoukina and Tieleman 2010). However, in the case of the PLY‐vesicle‐bilayer system, we found that pneumolysin can mediate contact between the vesicle and the bilayer. Changes in membrane curvature of the bilayer were observed from 1.5 µs in the system (Figure 2B). This membrane curvature and its site were maintained through the simulation. By 8 µs, a pronounced positive curvature was observed (Figure 2B). To check if this was due to undulations of the bilayer or due to apposition of the vesicle, mean curvature for all three systems was calculated, suggesting that the curvature was likely induced by PLY (Figure S2). The PLY‐induced membrane curvature was accompanied by local thinning (∼36–37 Å vs. bulk ∼40–41 Å) and water permeation (Figure 2C,D). This was captured in a simulation snapshot shown in Figure 2C and Video S1. Analysis of the planar membrane fragment within 10 nm of PLY showed that thinning and water leakage occur over similar spatial ranges, but thinning is most pronounced at the periphery of the protein‐containing region, whereas water permeation is distributed more uniformly across both leaflets (Figure S3). Analysis of the residence time (τ) of contacts between PLY and POPC head groups calculated for the last 1 µs from the 8 µs coarse‐grained simulation identified amino acids in PLY domain D1 alpha helix‐ Leu130, Trp134 and His135 as the top interacting residues (Figure S4). Residues L130, W134 and H135 were mutated to L130D, W134D and H135G using the CHARMM‐GUI mutation module to reduce local hydrophobicity and perturb the native secondary structure (Figure 3A and Figure S5). Maximum protein–lipid contacts over the final 1 µs of simulation were analysed for both wild‐type and mutant PLY. Mutation of key PLY residues led to a marked reduction in both lipid occupancy and residence time at the POPC bilayer compared with the wild‐type model (Figure 3B,C). Figure 3B,C report changes in lipid occupancy (ΔO) and residence time (Δτ) of wild‐type PLY relative to the mutant, where larger values of ΔO and Δτ indicate a greater loss of interaction upon mutation. Notably, mutations at the most strongly interacting residues Leu130, Trp134 and His135 resulted in pronounced reduction in POPC interactions, reflected by substantially increased Δτ and ΔO values. These observations suggest that hydrophobic interactions at the lipid‐water interface may play a critical role in stabilising vesicle‐embedded PLY‐membrane association. Consistently, substitution of D1 residues Leu130 and Trp134, to Asp and His135 to Gly disrupted key membrane‐anchoring interactions, leading to a marked attenuation of membrane protrusion in the mutant PLY‐vesicle simulation (Figure S6). Additionally, no interactions and generation of curvature were observed for the mutant‐PLY system throughout the simulation (Video S2).

FIGURE 1.

Simulation model of pneumolysin (PLY) embedded vesicle. The membrane inserted D4 domain is positioned onto the vesicle bilayer membrane, and the D1 domain is opposed to the bilayer membrane. The pentameric PLY Cryo‐EM structure reported by Vögele et al. (2019) used for the simulation is depicted in the cartoon and surface representation.

FIGURE 2.

Progressive snapshots of the simulation for (A) control with a vesicle and bilayer and (B) a vesicle with pneumolysin embedded and a bilayer. (C) Pronounced bilayer membrane curvature was accompanied by displacement of water. (D) Water displacement and thickness of the bilayer were calculated for the simulation. The total time of the simulation was 8 µs for both the vesicle‐bilayer and pneumolysin‐vesicle‐bilayer systems.

FIGURE 3.

Difference in occupancy and residence time per residue of vesicle‐bound wild type and mutant PLY with bilayer POPC lipid. (A) Schematic of top interacting residues from D1 domain for which the mutations W134D, L130D and H135G were introduced. (B) Difference in occupancy (∆O) was obtained by subtracting the mutant PLY from the wild‐type PLY residues interacting with POPC; occupancy is the sum of contacts/total number of frames. The more positive the ∆O, the lesser is the effect of mutant‐PLY interaction with POPC. (C) Difference in duration (∆τ) of contacts obtained by subtracting the τ of mutant PLY from wild‐type PLY residue interactions with POPC. The higher the τ, greater is the duration of contact for wild‐type PLY residues. The standard deviations were calculated for ∆O and ∆τ plots.

Further, to study the changes in the bilayer curvature induced by PLY‐vesicle, we defined a z‐normal along the bilayer membrane (Figure S7A) and assessed the curvature of the bilayer membrane for the plain bilayer, vesicle‐wild‐type PLY‐bilayer and vesicle‐mutant PLY‐bilayer models (Figure S7B). The wild‐type PLY‐embedded vesicle bilayer model indicated pronounced fluctuations of the bilayer, characterised by both positive and negative deviations along the membrane normal (z‐axis), while the mutant‐PLY system showed substantially reduced bilayer perturbations, with fluctuations largely confined to positive curvature along the z‐normal (Figure S7B). Additionally, we derived a multi‐dimensional free energy surface along the reaction coordinate RC‐1 as bilayer thickness and RC‐2 as water leakage counts. Figure S7C suggests a possible disruption of the bilayer induced by PLY and the progression is thermally favourable as more water leakage occurs over the simulation period. Upon quantification, the membrane thickness did not change overall with the mutant PLY system. However, the water leakage was similar for the wild‐type PLY and mutant (Figure S8). Analysis of the localisation tendency of this defect revealed different patterns of water displacement, prompting us to investigate how the barrier was impacted by lipid tail protrusion. Based on the unbiased simulation, it was observed that lipid tail exposure is higher in the wild‐type PLY system in comparison to the mutant system, as inferred from the mean squared displacement and Solvent Accessible Surface Area (SASA) (Figure S9A–B). Taken together, the four unbiased MD simulations suggest a possible role for the PLY Domain 1 (D1) arm in modulating membrane curvature during interaction with the lipid bilayer, compared with the control and mutant systems.

2.2. Cholesterol‐Dependent Interactions of PLY‐Bound Donor With Recipient Liposomes

Encouraged by the computational results, we performed experiments to validate the membrane activity of vesicle‐bound PLY upon interaction with recipient cells. To study the role of pneumococcal toxin, PLY, on vesicle fusion, we used purified recombinant GFP‐tagged PLY and synthetic 1,2‐Diacyl‐sn‐glycero‐3‐phosphocholine (DOPC)/cholesterol (70:30) liposomes prepared by the thin‐film hydration method. Recipient liposomes were labelled with the lipophilic dye Nile red, while the donor unlabelled liposomes were pre‐incubated with GFP‐PLY at 37°C for 20 min. The excess non‐liposome‐bound PLY was removed by adding an equimolar ratio of cholesterol and subsequent washes. Confocal microscopy confirmed the binding of GFP‐PLY to liposomes in a cholesterol‐dependent manner, while cholesterol‐free liposomes showed negligible GFP signal (Figure 4A). Subsequently, the preformed GFP‐PLY liposomes were incubated with cholesterol to remove free PLY and then co‐incubated with recipient Nile Red‐labelled liposomes at 37°C for 30 min. Under these conditions, multiple vesicle fusion events were observed with GFP‐PLY located at the fusion interface (Figure 4A). Quantification of GFP signal in liposomes from three independent experiments showed that ∼60% of cholesterol‐containing donor liposomes and Nile‐red labelled recipient liposomes co‐incubated with donor liposomes were positive for GFP‐PLY (Figure S10). To confirm that the effect was due to fusion with liposome‐bound PLY rather than carryover of soluble PLY that dissociated from donor liposomes and bound to Nile red liposomes, a subset of cholesterol‐quenched GFP‐PLY liposomes was incubated in PBS under identical conditions for 30 min. The liposome‐free supernatant was incubated with recipient Nile‐red liposomes. No GFP signal was observed on the recipient Nile‐red liposomes under these conditions, suggesting that the PLY transfer was primarily mediated by fusion with donor PLY‐bound liposomes (Figure 4A and Figure S10). Co‐incubation of Nile‐red liposomes with cholesterol‐free donor GFP‐PLY liposomes did not result in transfer of GFP signal, indicating the role of cholesterol in mediating vesicle membrane fusion. In addition, the localisation of GFP‐PLY at the contact sites between the donor and recipient liposomes was confirmed by 3D Z‐stack confocal imaging and orthogonal projections (Figure 4B,C). Line intensity profile measured in the region of interest at the liposome contact sites showed the GFP signal flanked by the Nile‐red signals, indicating its presence at the fusion site between liposomes (Figure 4D). To further validate PLY‐mediated interactions between donor and recipient liposomes, we performed high‐resolution negative‐stain TEM and cryo‐EM analyses in the presence or absence of recombinant PLY. Negative‐stain TEM revealed extensive liposome‐liposome contacts in the presence of rPLY, with electron‐dense oligomeric structures consistent with PLY localised at membrane interface junctions (Figure 5A and Figure S11). In contrast, control liposomes incubated without rPLY exhibited minimal membrane contacts and no detectable oligomeric structures (Figure S12). Consistent with these observations, cryo‐EM analysis identified additional electron density at the liposome interface in rPLY‐treated samples, supporting the localisation of PLY at sites of membrane interaction (Figure 5B,C). Taken together, our data suggest that the liposome‐bound PLY toxin can mediate interaction with neighbouring vesicles upon contact.

FIGURE 4.

Cholesterol‐dependent interactions of PLY‐bound donor with recipient liposomes. (A) Confocal microscopy images of Nile Red‐labelled recipient liposomes co‐incubated with donor GFP‐PLY‐bound liposomes, prepared with or without cholesterol (‐Chol). Scale bars, 5 µm. (B, C) Orthogonal projections from 3D Z‐stacks in the XY and XZ axis, showing the localisation of GFP‐PLY at the contact sites with the recipient liposome, indicated by white arrows. (D) Line intensity profiles of GFP and Nile red signals were measured using Nikon A1R HD software along with the Mander's overlap coefficient in the boxed contact area between the fusing liposomes. Confocal data are representative of three independent experiments.

FIGURE 5.

High‐resolution TEM and cryo‐EM analysis of donor and recipient liposomes in the presence of GFP‐rPLY. (A) Negative‐stain TEM image of donor and recipient liposomes in the presence of GFP‐rPLY. The green‐dotted line indicates the donor liposome, whereas the white dotted line indicates the recipient liposome. GFP‐rPLY molecules are marked with blue arrows. (B) Cryo‐EM image of donor and recipient liposomes in the presence of GFP‐rPLY. GFP‐rPLY molecules are marked with yellow arrows. (C) Reference‐free 2D class averages showing the junction between donor and recipient liposomes. An extra density is observed between the two liposomes.

2.3. Membrane Localisation of PLY on Microvesicles Shed From Challenged Human Monocytes

Next, we investigated the distribution of PLY across two different populations of EVs released by human THP‐1 monocyte‐like cells challenged with a sublytic dose of 0.5 µg/mL recombinant PLY. The microvesicle (MVs) fraction was isolated by centrifugation of the conditioned culture supernatant at 10,000 × g. The sEVs were isolated by iodixanol cushion‐density gradient ultracentrifugation and pooling the fractions F5‐F9. The purity of pooled sEV fractions was confirmed by probing for the tetraspanin proteins CD63 and CD81, and the absence of the abundant serum protein albumin, indicating the purity of EVs (Figure 6A). The sEVs also contained PLY toxin. Nanoparticle tracking analysis (NTA) of Naïve and rPLY sEVs revealed a mean particle diameter of 142.6 and 151.7 nm, respectively (Figure 6B). The MVs were positive for the classical marker, Annexin A1 (Jeppesen et al. 2019) and the PLY toxin (Figure 6C). The MVs showed a slightly higher particle diameter compared to sEVs, with mean values of 156.3 and 176.8 nm for Naïve and rPLY MVs, respectively (Figure 6D).

FIGURE 6.

Membrane localisation of PLY on microvesicles shed from challenged human THP‐1 cells. (A) Western blot showing the presence of characteristic sEV markers, CD63 and CD81 and the pneumococcal toxin, PLY. Albumin was used as a negative control to show EV purity. (B) Nanoparticle tracking analysis (NTA) showing the size distribution of sEVs isolated from THP‐1 cells challenged with (0.5 µg/mL) rPLY (sEVs rPLY) and untreated (sEVs Naïve). (C) Western blotting showing the typical microvesicle (MV) marker, Annexin A1 and PLY toxin in rPLY and Naïve MVs. (D) Size distribution of Naïve and rPLY‐MVs using NTA. (E) Western blotting showing the relative enrichment of PLY toxin in purified sEVs and MVs. β‐actin was used as the loading control. (F) Quantification of MV particle concentration purified from THP‐1 cells challenged with GFP‐PLY and the mutant PLYW433F. (G) The percentage of MVs containing GFP‐PLY was quantified using fluorescence NTA on the FL500 channel. (H) Representative immunogold TEM images showing the membrane localisation of PLY on the MVs. Arrows indicate the (∼12 nm) gold beads conjugated to the secondary antibody, which are present on rPLY‐MVs but absent on naïve MVs. Scale bars, 200 nm (Naïve), 100 nm (rPLY MVs). (I) Distribution of rPLY‐positive MVs (N = 67 fields). The graph displays the percentage of rPLY‐positive MVs and is presented as the mean ± SEM. **** represents p < 0.0001 by unpaired t‐test. Data are representative of three independent experiments.

Western blot analysis revealed a markedly greater enrichment of PLY toxin in the MVs compared to the sEVs, suggesting preferential incorporation of PLY protein on plasma‐membrane‐derived MVs (Figure 6E). Fluorescence‐mode NTA analysis of MVs isolated from GFP‐PLY challenged THP‐1 cells showed ∼2.5‐fold increase in particle concentration compared to Naïve MVs. In contrast, MVs purified from cells challenged with the toxoid mutant, PLYW433F, with reduced membrane binding and hemolytic activity (Shewell et al. 2014) did not show a significant difference in particle concentration compared to Naïve EVs (Figure 6F). Direct quantification of GFP‐PLY on the MVs by fluorescence‐NTA revealed ∼2‐fold enrichment of PLY in MVs relative to sEVs (Figure 6G), in agreement with Figure 6E. To visualise the localisation of PLY on the MVs, we performed immunogold labelling followed by transmission electron microscopy analysis, showing membrane‐bound PLY (Figure 6H). Quantification of TEM images revealed that ∼50% of the MVs exhibited PLY positivity (Figure 6I). Taken together, the above findings suggest that pneumococcal toxin PLY upregulates shedding of plasma membrane‐derived MVs that traffic PLY toxin shed from challenged cells.

2.4. Transfer of MV‐Bound PLY to Recipient PBMCs Upon Co‐Incubation

We next tested whether the PLY‐bound MVs would interact and fuse with recipient cells, resulting in the transfer of PLY toxin to the target cell membrane. Purified MVs from wild‐type PLY‐challenged THP‐1 cells were co‐incubated with healthy donor peripheral blood mononuclear cells (PBMCs) at 37°C for 1 h. The presence of PLY toxin in the MVs purified from challenged monocytes was validated by western blotting (Figure 7A). The MV‐free supernatant was also probed as a control, which showed negligible carryover of free soluble PLY into EV preparations. THP‐1 cells challenged with the PLY toxoid mutant, PLYW433F, which exhibits compromised haemolytic activity relative to wild‐type PLY (Paton et al. 1991) also shed MVs containing PLY, albeit significantly less compared to WT‐PLY. This was quantified by densitometry analysis upon normalisation to the loading control, β‐actin (Figure S13A), showing ∼2.5‐fold higher binding of PLY‐WT compared to the mutant.

FIGURE 7.

Transfer of MV‐bound PLY to recipient PBMCs upon co‐incubation. (A) Immunoblotting showing the presence of PLY in MVs purified from THP‐1 cells challenged with wild‐type PLY and the toxoid mutant PLYW433F. MV‐free supernatants from PLY‐MVs and naïve MVs were used as controls, showing negligible non‐MV‐bound PLY after purification. β‐actin served as the loading control. (B) Western blot analysis of human PBMCs treated with rPLY MVs at 1 h post‐MV incubation. PBMCs treated with wild‐type PLY‐MVs, showed transfer of PLY. PLYW433F mutant and MV‐free supernatant were used as controls. β‐actin was used as the loading control. (C) Confocal microscopy of PBMCs incubated with 30 µg of Nile Red‐labelled GFP‐rPLY‐MVs, GFP‐rPLY‐MVs supernatant, PLYW433F MVs or naïve MVs for 1 h. Images show that GFP‐rPLY‐MVs fuse with PBMCs, leading to cell disruption (arrows), while PLYW433F MVs and naïve MVs did not disrupt cell morphology. DAPI was used to counterstain the cells. Scale bars, 5 µm. (D, E) Orthogonal projections from 3D Z‐stacks in the XY and XZ plane, showing colocalisation of PLY‐GFP signal with MV‐Nile red bound to the PBMCs. (F) Line intensity profiles of GFP and Nile red signals were measured using Nikon A1R HD software along with Mander's overlap coefficient in the selected region of interest. All data are representative of experiments from three independent donors.

To study the transfer of MV‐associated PLY to recipient PBMCs upon coincubation, we performed western blotting on the treated PBMC lysates, which confirmed the transfer of PLY on the treated cells (Figure 7B). Incubation with MV‐free supernatant and MVs containing PLYW433F toxoid showed negligible PLY transfer to PBMCs. Densitometry quantification normalised to β‐actin showed significantly higher PLY‐band intensity in PBMCs co‐incubated with WT‐PLY bound MVs compared to PLYW433F‐bound MVs (Figure S13B). To visualise the MV‐mediated PLY transfer on the cell membrane, we performed confocal microscopy on PBMCs incubated with MVs isolated from monocytes challenged with GFP‐tagged PLY. The MVs were also counterstained with the lipophilic dye Nile red. Results showed that PBMCs incubated with Nile‐red labelled MVs containing GFP‐PLY showed co‐localised Nile red and GFP signals, confirming the MV‐mediated PLY transfer to PBMCs (Figure 7C). Moreover, the cells showed a ruptured cell membrane consistent with the seeding of PLY‐pore. 3D Z‐stack confocal imaging and orthogonal projections were also performed to confirm the membrane localisation of the signals (Figure 7D,E). Further, the line intensity profile of fluorescent signals measured at the region of interest on the PBMC membrane indicated colocalisation of GFP and Nile‐red signals with a Mander's coefficient of 0.78 (Figure 7F). Naïve MVs isolated from untreated cells also showed binding, but the cells retained rounded and intact morphology. The MV‐free supernatant as well as the MVs containing the toxoid PLYW433F, showed very little cellular binding and intact cells.

2.5. Transfer of MV‐Bound PLY to Recipient Cells Compromises Plasma Membrane Integrity and Induces Cell Death

Next, we wanted to study the impact of MV‐mediated transfer of PLY toxin onto recipient cells. We performed the propidium iodide (PI) staining assay to assess the membrane integrity of PBMCs upon MV coincubation at 37°C for 1 h. We found that upon PLY‐MV coincubation, there was a 3‐fold increase in the percentage of PI positivity of treated PBMCs, relative to Naïve‐MVs, indicating compromised membrane integrity (Figure 8A). The fold change in the median intensity was calculated from three independent experiments showing a clear trend, despite not achieving statistical significance (Figure 8B). This was consistent with the higher binding of PLY‐GFP to the PBMCs upon co‐incubation with PLY‐MV, when compared to MVs containing the toxoid PLYW433F, as measured by flow cytometry (Figure 8C and Figure S14A). Cell viability analysis using fixable viability dye e‐fluor 780 showed a significant increase in cell death induced by WT‐PLY MVs compared to Naïve MVs and the mutant PLY MVs (Figure 8D and Figure S14B). Transfer of supernatant from MV‐free supernatant showed negligible cell death, suggesting MV‐specific effects. Next, we compared the differential binding of PLY to CD14⁺ monocytes and CD4⁺ T cells upon co‐incubation with MVs and resulting cytotoxicity. We found that CD4⁺ T cells showed higher GFP‐PLY positivity (Figure 8E and Figure S15A) and higher cell death (Figure 8F and Figure S15B) compared to monocytes. Naïve MVs and MV‐free supernatants showed negligible cell death.

FIGURE 8.

Transfer of MV‐bound PLY to recipient cells compromises plasma membrane integrity. (A) Flow cytometry histograms showing propidium iodide (PI) staining of human PBMCs at 1 h post‐treatment with 100 µg of rPLY MVs, as measured on the PE‐A channel. (B) Quantification of fold change (relative to untreated) in median fluorescence intensity of PI uptake by human PBMCs in panel A. p values were computed by an unpaired t‐test. (C) Quantification of the percentage of PLY and PLYW433F transfer to PBMCs upon 100 µg MV challenge. p values determined by one‐way ANOVA, ****p < 0.0001, ***p < 0.0005, *p < 0.05. (D). Percentage of dead cells quantified by FVD staining at 1 h post 100 µg MV challenge. PLYW433F‐MVs and MV‐free supernatant were used as controls. Data were analysed by one‐way ANOVA. ***p < 0.005, ns; non‐significant. (E) Graph showing the percentage of GFP‐PLY binding to CD4+ T cells and monocytes upon 100 µg PLY‐MV treatment. p values were analysed using the two‐way ANOVA Sidak's multiple comparison test. *p < 0.05, ns; non‐significant. (F). Quantification of the percentage of dead cells of CD4+ T cells and monocytes upon PLY‐MV treatment. p values determined with two‐way ANOVA. **** <0.0001, ns; non‐significant (G). PI staining of human PBMCs treated with 100 µg of rGFP‐PLY, PLYW433F, or naïve MVs in the presence of 25 mM EDTA. (H). Quantification of log10 fold change (relative to untreated) in median fluorescence intensity of PI uptake by human PBMCs treated with Naive MVs, PLYW433F and rPLY MVs in panel D. **** represents p < 0.0001 by one‐way ANOVA. Data are representative of experiments from three independent donors and are presented as mean ± SEM.

When the plasma membrane of eukaryotic cells is mechanically injured, Ca²⁺ influx triggers a rapid repair process that promotes membrane resealing (Bhattacharya et al. 2022; Andrews and Corrotte 2018). To test whether Ca⁺²‐dependent intrinsic membrane repair could reduce the impact of PLY‐MVs, we next performed the experiment in the presence of 25 mM EDTA to chelate divalent cations, including Ca²⁺ and Mg²⁺. We found that this caused a drastic increase in the PI positivity of treated PBMCs up to 90% (Figure 8G) with a 4‐log fold increase in the median fluorescence intensity (Figure 8H), suggesting the role of the membrane repair process in mediating resistance to PLY‐containing MVs. The mutant PLYW433F did not show any significant change compared to Naïve MVs.

Taken together, our results show that MVs shed from PLY‐challenged cells implant the toxin to recipient cells upon fusion, causing membrane damage.

3. Discussion

This study identifies extracellular vesicle‐bound pneumolysin (PLY) as an active mediator of membrane engagement and instability, extending the classical view of PLY action beyond pore formation by soluble toxin. Our findings support a previously unrecognised model in which PLY displayed on microvesicles shed from challenged host cells has unique membrane‐interactive properties that may facilitate vesicle‐cell membrane fusion and toxin transfer, thereby contributing to bystander cell damage during pneumococcal infection. Canonical models of cholesterol‐dependent cytolysins (CDC) emphasise the central role of domain 4 (D4) in cholesterol recognition and domain 3 (D3) in β‐barrel pore insertion (Tilley et al. 2005). Our data suggest that when PLY is presented on a curved vesicular membrane, additional structural elements may become functionally relevant. In particular, the exposure of domain 1 (D1) α‐helical regions may enable direct interactions with an opposing bilayer, promoting membrane curvature and destabilisation. This expands current understanding of CDC biology by indicating that membrane‐bound toxin oligomers can engage target membranes through non‐canonical interfaces, distinct from those used during classical pore formation.

The cholesterol dependence of vesicle‐vesicle interaction and toxin transfer observed here is consistent with the well‐established role of cholesterol in CDC binding (Giddings et al. 2003; Farrand et al. 2015; Johnstone et al. 2022) and its ability to modulate bilayer mechanics and fusion energetics (Rog and Pasenkiewicz‐Gierula 2006a). These findings suggest that PLY may exploit cholesterol‐rich membranes not only as docking platforms but potentially also as permissive environments for membrane fusion, thereby enabling lateral dissemination of toxin. At the cellular level, the preferential enrichment of PLY in plasma membrane‐derived microvesicles aligns with previous work identifying microvesicle shedding as a major component of calcium‐dependent membrane repair in response to CDC attack (Larpin et al. 2020; Wolfmeier et al. 2016). While the cellular toxin shedding response is generally viewed as protective, our results indicate that the released PLY‐bearing microvesicles retain pathogenic activity by transferring toxin to recipient immune cells. This work also complements the emerging idea in the field that EVs act as carriers of bacterial virulence factors and inflammatory mediators during infection (Jeppesen et al. 2023; Parveen et al. 2024). Koffel et al. (2018) demonstrated that host‐derived microvesicles carrying pore‐forming toxins, including PLY, can deliver toxin‐associated signals to macrophages and modulate downstream immune responses. Importantly, our findings complement these studies by providing mechanistic insight into how vesicle‐associated PLY can directly engage and destabilise recipient membranes, besides acting through the reported endocytic uptake pathway (Letsiou et al. 2021). The reduced activity observed with the toxoid mutant, PLYW433F further underscores the requirement for intact toxin‐membrane interactions even in the vesicle‐bound context. Interestingly, PLY‐bearing MVs preferentially targeted CD4⁺ T cells over CD14⁺ monocytes, resulting in greater PLY binding and cytotoxicity. This is consistent with the findings of Larpin et al. (2020), who showed that T cells are intrinsically more susceptible to soluble pneumolysin than monocytes due to enhanced toxin binding and less efficient Ca^{2 +}‐dependent membrane repair. Our findings extend this observation by demonstrating that MV‐associated PLY remains biologically active and retains this cell type‐specific tropism, highlighting a potential mechanism by which PLY‐bearing host‐derived MVs contribute to immune dysregulation during pneumococcal infection.

While the approaches used here provide mechanistic insight into vesicle‐bound PLY‐membrane interactions, they also have limitations and highlight potential scope for future work. The coarse‐grained molecular dynamics simulations employed in this study are well suited to capturing mesoscopic membrane deformations and long‐time scale lipid rearrangements, but higher‐resolution atomistic simulations will be valuable for resolving the detailed protein‐lipid interactions and transient fusion intermediates that underlie these events. A further consideration is that the modelled vesicle (∼20 nm) is more highly curved than physiological PLY‐bearing microvesicles, whose mean diameters measured here by NTA were ∼150–170 nm. Because membrane curvature is itself an on‐pathway contributor to fusion, coarse‐grained free‐energy calculations show that curvature lowers the barrier to stalk formation, with curved vesicular geometries exhibiting the lowest fusion barrier relative to apposed planar membranes (Kawamoto et al. 2015). The curvature generated at the contact site is consistent with an early priming event rather than an artefact of system size. Importantly, the membrane engagement observed here is not a generic consequence of vesicle curvature. On an identically curved vesicle, the PLY D1 triple mutant (L130D/W134D/H135G) failed to generate curvature or membrane protrusion and showed markedly reduced PLY‐POPC occupancy and residence time, indicating that the effect is driven by specific D1 α‐helix residues rather than by vesicle size or curvature per se. The absolute magnitude of any curvature‐dependent reduction in the fusion barrier is nonetheless expected to differ for larger physiological vesicles, and larger systems together with higher‐resolution atomistic models will be required to quantify this dependence. Similarly, the liposome fusion assays were intentionally designed as reductionist systems to isolate the role of pneumolysin and cholesterol. Our complementary cryo‐EM analyses provide direct evidence of PLY‐mediated liposome fusion and localisation of PLY at the membrane interface, supporting the mechanism predicted by atomistic simulations. Future studies employing vesicles with more complex and asymmetric lipid compositions, incorporating membrane‐associated proteins, and optimised high‐resolution cryo‐EM analyses will further define the structural basis and stoichiometry of PLY‐mediated membrane fusion in physiologically relevant settings. Together, these approaches will advance our understanding of pneumolysin‐driven vesicle‐membrane interactions and may inform therapeutic strategies targeting vesicle‐membrane fusion or membrane‐active surfaces of vesicle‐bound PLY to mitigate pneumococcal virulence in the face of rising antibiotic resistance.

4. Methods

4.1. Preparation of Vesicle‐Membrane Systems for Simulations

We constructed two membrane systems using the Martini builder in CHARMM‐GUI (Jo et al. 2008): (1) A planar bilayer (60 × 60 nm²) of ∼4 nm thickness mimicking the lipid composition of planar bilayer membranes (POPC:80%, CHOL:20%), and (2) A liposome (∼20 nm diameter) resembling lipid vesicles (CHOL:40%, DPPC:30%, DOPS:20%, DPSM:10%) (Rog and Pasenkiewicz‐Gierula 2006b; Sakai‐Kato et al. 2020) were prepared. A vesicle diameter of ∼20 nm was used to retain the complete vesicle‐protein‐bilayer assembly at a system size tractable over the multi‐microsecond trajectories required to observe membrane engagement. This is consistent with established coarse‐grained membrane‐fusion studies, in which small unilamellar vesicles of ∼10–50 nm radius are routinely employed, and vesicles of comparable or smaller size have been used to resolve fusion intermediates (Marrink and Mark 2003; Kasson et al. 2006). Simulating a full physiological‐diameter vesicle with an embedded protein over these timescales remains computationally prohibitive. A total of four molecular systems (Table S1) were prepared viz. (1) plain bilayer membrane mimicking host cell membrane, (2) the control model with vesicle at 3 nm distance from lipid bilayer (Figure S1), (3) vesicle embedded pneumolysin (PLY) positioned towards the lipid bilayer at 3 nm distance (Figure 1) and (4) a mutant system similar to that of model 3, with PLY embedded on vesicle and oriented towards the bilayer (Figure S4). PLY pentamers (5‐mer) in the membrane‐inserted conformation were extracted from the Cryo‐EM structure (van Pee et al. 2017) and docked onto the vesicle with D4 inserted into the vesicle (Figure 1) using CGSB tool (Andreasen et al. 2025). The two PLY pentamers were placed with the per‐oligomer orientation taken from the cryo‐EM pore complex (D4 docked into the vesicle, D1 facing the target bilayer) and separated by ∼15.9 nm (centre‐to‐centre), the maximum permitted by the ∼20 nm vesicle geometry, to avoid artificial inter‐oligomer contacts. For preparing the molecular systems, initially the membranes were desolvated, combined using PyMOL (PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC and resolved via the CGSB tool (Andreasen et al. 2025).

4.2. Coarse‐grained Simulations of Vesicle‐Bilayer Interactions

All coarse‐grained simulations (Table S1) were performed in GROMACS 2021.4 (Abraham et al. 2015) with the Martini v2.2 force field (Michaud‐Agrawal et al. 2011) for proteins, lipids, water and ions. The ion concentration used after solvation was 0.15 M NaCl. Energy minimisation used the steepest descent algorithm (Berendsen et al. 1995) (100 kJ/mol/nm tolerance). Equilibration involved gradual restraint release, increasing time steps from 10 to 30 fs. Periodic boundary conditions were applied; dielectric constant was 15; Coulombic and Lennard‐Jones (Marrink et al. 2007) cutoffs were 1.1 nm. Pressure was maintained at 1 bar using Berendsen coupling during equilibration, with isotropic coupling for production runs for all systems other than the plain bilayer where semi‐isotropic coupling was used. Temperature control employed the V‐rescale thermostat (Bussi et al. 2007). Neighbour lists were updated every 20 steps. Analysis were conducted using MDAnalysis (Michaud‐Agrawal et al. 2011) and Prolint python modules (Sejdiu and Tieleman 2021).

4.3. Liposome Preparation

Liposomes were prepared by dissolving 70% 1,2‐Diacyl‐sn‐glycero‐3‐phosphocholine (DOPC; Sigma, P3556) with or without 30% cholesterol (Sigma, C8667) into 1 mL of dichloromethane as the solvent. The thin lipid film was formed using a rotary flask stirred at a 45° angle in a water bath at 40°C. The thin film formed was further hydrated using 1 mL PBS and sonicated for 15 min at 37 kHz at 37°C. The resulting liposome suspension was centrifuged at 5000 × g for 5 min at room temperature, and the pellet was resuspended in 100 µl PBS. The liposomes were labelled with 1 mM Nile red (Invitrogen) that was added separately to the dichloromethane solvent during the preparation stage.

4.4. Liposome Fusion Assay

Liposomes with or without cholesterol were incubated with 0.5 µg (1:2000) of recombinant GFP‐PLY (gift from Prof. Anirban Banerjee, Indian Institute of Technology Bombay) and 1 mM DTT for 20 min. It was centrifuged at 5000 × g for 5 min and washed to remove unbound GFP‐PLY. The pellet was reconstituted with cholesterol at an equimolar concentration to PLY and incubated for 20 min at 37°C to bind free soluble PLY. Following a wash, the liposomes were resuspended in PBS. Subsequently, the Nile red‐labelled liposomes were co‐incubated with GFP‐PLY bound liposomes with or without cholesterol for 30 min at 37°C. It was washed and resuspended in PBS. The GFP‐PLY‐bound liposomes with cholesterol were co‐incubated with PBS for 30 min at 37°C as a control to test for PLY detachment. The liposomes were then incubated with Nile red liposomes for 30 min at 37°C and washed with PBS. The liposomes were analysed using the Nikon A1R confocal microscopy system on their respective FITC and TRITC channels. The acquired data were analysed by optimising on NIS AR1 HD software and 3D projection, and Z‐stack was performed to confirm the co‐localisation of GFP‐PLY on the liposomes.

4.5. Negative Staining Sample Preparation and Visualisation by TEM

A total of 3 µL of rGFP‐PLY incubated with DOPC‐Cholesterol lipid mixture (as described above) was added to a thin film of carbon‐coated copper grids (Carbon film 300 mesh, Copper, Electron Microscopy Sciences) at room temperature, which was freshly glow discharged (negative polarity) for 30 s, 20 mA. The sample was allowed to settle down on grid for 1 min, before removing excess buffer. The excess solution was then carefully blotted off using Whatman filter paper. 1% uranyl acetate (Polysciences, Inc.) was used for staining and excess stain was blotted out. Talos L120C transmission electron microscope operating at 120 kV was used for imaging. Images were collected with a bottom‐mounted Ceta camera (4000×4000 pixels) at magnifications of x57,000–x73,000.

4.6. Cryo‐EM Sample Preparation

R1.2/1.3 300 mesh copper grids (Quantifoil; Electron Microscopy Sciences) were glow discharged for 90 s at 20 mA following cryo‐sample freezing process. Three microliters of solution from that rGFP‐PLY‐lipid mixture were applied onto the freshly glow discharged grids and blotted for 6.5 s with blot force of positive 5 using FEI Vitrobot Mark IV plunger. The grid was plunged into the liquid ethane instantly after blotting.

4.7. Cryo‐EM Data Collection

200 kV Talos Arctica transmission electron microscope (Thermo Scientific) equipped with K2 Summit Direct Electron Detector (Gatan Inc) was used to Cryo‐EM data acquisition. Movies were recorded automatically using Latitude‐S (Digital Micrograph—GMS 3.5) at 42,900× magnification and the effective pixel size of 1.17 Å (Kumar et al. 2021). A total of 1154 movies were collected in counting mode with a total dose of 45 e⁻/Ų, with an exposure time of 8 sec distributed for 20 frames.

4.8. Cryo‐EM Data Processing

Cryo‐EM data was processed using the cryoSPARC 4.5 (Punjani et al. 2017). Beam induced motion was corrected for 1154 movies using Patch Motion Corr followed by estimation of Contrast transfer function (CTF) parameters using Patch Ctf estimation. Micrographs were manually checked, and after curating exposure with a resolution cutoff of 7 Å, we selected the best 337 micrographs for further processing. A small subset of particles was manually picked using manual picker to generate the reference‐free 2D class averages. The best 2D classes were selected to use as a template for template‐picker based picking. During inspect pick we have kept lower histogram values and higher NCC score for more accurate picking. The particles were extracted with the two different box sizes, for example, 360‐ and 480‐pixel to see the individual particles and particles at the inter lipid membrane area. Several rounds of 2D classification were run to remove junk particles and reduce background noise.

4.9. Cell Culture

Human THP‐1 monocyte‐like cells were cultured in suspension at a density of 1 × 10⁶ cells/mL in T175 flasks (Corning) using RPMI 1640 medium (HiMedia) supplemented with 10% fetal bovine serum (FBS) (Invitrogen). The cells were maintained at 5% CO₂ and 37°C medium upon subculturing every 3 days.

4.10. Isolation of EVs From THP‐1 Cells Post PLY Challenge

Briefly, 6 × 10⁷ THP‐1 cells were seeded in a T175 flask (Corning) with 12 mL media containing RPMI 1640 supplemented with 1% penicillin‐streptomycin. The cells were challenged with a sublytic dose of 0.5 µg/mL of wild‐type rPLY, rGFP‐PLY and mutant PLYW433F (kind gifts from Prof. Anirban Banerjee, IIT Bombay) or left untreated at 37°C for 24 h. The conditioned media collected at 24 h post‐PLY challenge were subjected to differential centrifugation to remove cells (300 × g, 10 min), cell debris and apoptotic bodies (2000 × g, 20 min) and larger microvesicles (10,000 × g, 30 min) at 4°C. The resulting supernatant was filtered through a 0.45 µm filter (HiMedia) to remove soluble non‐EV proteins. The filtrate was layered on 2 mL of 60% iodixanol solution and subjected to cushioned density ultracentrifugation at 160,000 × g for 3 h at 4°C. The supernatant was discarded, and up to 3 mL of the bottom layer containing 40% iodixanol and small EVs was overlaid with 3 mL of 20%, 10% and 5% sucrose‐iodixanol gradient and centrifuged at 100,000 × g for 18 h at 4°C using SW40Ti rotor on the Optima XPN‐100 ultracentrifuge (Beckman Coulter). Each 1 mL fractions were collected from the top, and F5‐F9 fractions that were rich in sEVs were pooled and concentrated using 100 kDa cut‐off columns (Amicon, Merck). The purified EVs were stored in −80°C until use.

4.11. Nanoparticle Tracking Analysis of EVs

NTA were performed using Zeta view x30 Quatt instrument equipped with four lasers, 405, 488, 520, and 640 nm. The EV samples were diluted (1:1000) with 10% 1x PBS in milli‐Q water for analysis. The GFP‐PLY EVs and rPLY EVs was analysed using FL500 and SC488 Channels and the particle number and size distribution were analysed at 11 individual positions inside the measuring cell under a set sensitivity and a shutter value of 100. The data were analysed by Zetaview NTA software version 1.0.2.7.

4.12. Western Blotting

The isolated EVs were lysed using equal volume of RIPA lysis buffer (HiMedia) supplemented with Protease inhibitor cocktail (Roche) on ice for 30 min, followed by centrifugation at 16,000 × g for 30 min at 4°C. Total EV lysate protein was quantified using Bicinchoninic acid assay (Pierce). Thirty micrograms of total EV protein was boiled at 70°C for 10 min in 1x Laemmli buffer (Biorad) containing 50 mM DTT and resolved on 12%–15% SDS polyacrylamide gel electrophoresis and transferred onto PVDF membrane. The membrane was then blocked with 5% BSA at room temperature for 1 h and incubated overnight at 4°C with respective primary antibody (mouse anti‐CD81 (1:1000), rabbit anti‐CD63 (1:1000), rabbit anti‐Annexin A1 (1:1000) and mouse anti‐PLY (1:500), followed by washes with TBS containing 0.1% Tween 20 and HRP‐conjugated secondary antibody (anti‐mouse IgG and anti‐rabbit IgG (1:3000), Biorad). The chemiluminescence signal was visualised with an iBright FL1500 Imaging System (Invitrogen). Subsequently, the membrane was stripped and reprobed using mouse anti‐β‐actin antibody at a dilution of 1:1000, followed by goat anti‐mouse antibody at a dilution of 1:3000. β‐actin served as an internal control to confirm uniform protein loading across the samples.

4.13. Isolation of PBMCs From Healthy Donors

A total of 20 mL of venous blood was drawn from healthy volunteer donors at BRIC‐RGCB in EDTA Vacutainer tubes (BD Bioscience) to prevent clotting. Informed consent was obtained in accordance with the approved ethical protocol (Ref. IHEC/12/2020_E/14). The blood was then diluted with an equal volume of recommended media (endotoxin‐free PBS containing 2% FBS and 1 mM EDTA. The diluted blood was carefully layered over HiSep density gradient media (HiMedia) and centrifuged at 1200 × g for 20 min with low brakes. The buffy coat layer was carefully collected into a new tube and topped with recommended media, followed by centrifugation at 300 × g for 10 min with low brakes. The washes were repeated thrice. A low‐spin centrifugation at 120 × g for 10 min with low brakes was performed in between the washes to remove platelets. Residual RBCs were lysed with 1x RBC lysis buffer (HiMedia) for 10 min at room temperature. The PBMC pellet was resuspended in RPMI containing 2.5% EV‐depleted FBS (Gibco), counted and seeded in 12‐well plates for EV challenge.

4.14. Propidium Iodide Staining

Propidium Iodide (PI) was used to measure the membrane permeability induced by PLY‐bound MVs on PBMCs. One hundred micrograms of MVs, including GFP‐rPLY, GFP‐rPLY supernatant, PLYW433F and naive EVs were co‐incubated with 2 × 10⁵ PBMCs for 1 h at 37°C. After incubation, unbound EVs were removed by wash and centrifugation for 5 min at 2000 × g. The cell pellet was resuspended in 200 µL of FACS buffer and subsequently transferred to a 96‐well plate. The PBMCs were incubated with 10 µM PI stain (Invitrogen) for 30 min at 37°C. Samples were washed thrice by flicking the supernatant at 2000 × g for 1 min. The pellet was resuspended in 200 µL of FACS buffer and analysed on the PE‐A channel on the BD FACS Aria II flow cytometer (BD Biosciences) to quantify the percentage of PI‐positive population. A total of 5% H₂O₂ was used as a positive control to induce membrane damage to PBMCs. The data obtained were further analysed using Kaluza software (Beckman Coulter).

4.15. Flow Cytometry to Measure PLY‐Transfer to PBMCs Upon MV Treatment

Flow cytometry was performed to quantify the percentage of PLY‐positive PBMCs following MV challenge. Following 1 h incubation with 100 µg PLY‐MVs, the cells were washed and analysed for GFP tagged WT‐PLY using the CytoFLEX S flow cytometer on the FITC channel. For PLYW433F‐MV, the cells were harvested and blocked with FACS buffer (PBS containing 2% FBS) for 30 min on ice to minimise non‐specific antibody binding. After blocking, the cells were incubated with the primary antibody, anti‐pneumolysin (anti‐PLY) antibody, diluted 1:50 in FACS buffer (2 µL antibody in 98 µL FACS buffer), for 1 h on ice. Following two washes with FACS buffer, the cells were incubated with Alexa Fluor 594‐conjugated secondary antibody (1:200 dilution) for 30 min on ice in the dark. The stained cells were then washed twice with FACS buffer to remove excess antibody and immediately analysed using the CytoFLEX S flow cytometer on the ECD fluorescence channel. Untreated cells served as the negative control.

4.16. Immunogold Labelling of MVs with PLY Antibody

A total of 5 µL of the EV‐rich fractions were used for immunogold labelling using transmission electron microscopy. The samples were fixed with 2% paraformaldehyde (Himedia) to preserve the EV membrane integrity and 10 µL of fixed EV samples was adsorbed onto formvar‐coated 400‐mesh copper grids (Electron Microscopy Sciences) overnight at 4°C in a humidified chamber. The excess sample was removed using a wick and blocked with filtered 0.1% BSA in PBS for 30 min at room temperature. The grid was incubated with 10 µL of primary antibody against PLY (1:10) for 1 h at room temperature. The grids were washed thrice with PBS and incubated with 10 µL of 12 nm gold conjugated anti‐mouse antibody (1:10) for 30 min at room temperature. The samples were fixed using 2.5% glutaraldehyde for 10 s, washed twice with de‐ionised water and negatively stained using UranyLess (Electron Microscopy Sciences) for 10 s. The excess stain was removed using a wick and allowed to dry. The grids were then examined at 200 kV and 74k magnification using Talos F200i transmission electron microscope (Thermo Scientific).

4.17. PLY‐MV Fusion With PBMCs

A total of 30 µg of Naïve, PLYW4333F, GFP‐PLY‐MVs supernatant, and GFP‐PLY MVs were fluorescently labelled with Nile‐red (10 µg/mL) and incubated for 30 min at room temperature. The samples were centrifuged at 10,000 × g for 30 min to remove excess dye, followed by two consecutive washes with PBS. The labelled MVs were co‐incubated with 7.5 × 10⁴ PBMCs for 1 h at 37°C. Following the incubation, the cells were washed at 300 × g for 10 min to remove unbound MVs. The final cell pellet was resuspended into 30 µL of PBS, and glass slides were mounted using ProLong Antifade Glass Mounting media (Invitrogen) with DAPI. Confocal imaging was performed using a laser‐scanning confocal microscope, and the images were processed and analysed using Nikon AR1HD software to assess fluorescence signals and observe the localisation of the GFP‐PLY protein on the PBMC membrane.

4.18. Primary CD4+ T Cell and Monocyte Isolation

CD4⁺ T cells and monocytes were independently isolated from the peripheral blood of the same healthy donor using CD4⁺ T cell and monocyte enrichment cocktails (Stem Cell Technologies), respectively, following the manufacturer's instructions. Briefly, whole blood (10 mL) was incubated with the respective enrichment cocktail at a concentration of 50 µL/mL for 20 min at room temperature. Following incubation, the blood was diluted with an equal volume of recommended buffer (PBS supplemented with 2% FBS and 1 mM EDTA) and carefully layered over Lymphoprep (Stem Cell Technologies). Density gradient centrifugation was performed at 1200 × g for 20 min with low acceleration and deceleration settings. The resulting white buffy coat layer, enriched for CD4⁺ T cells or monocytes, was carefully collected and washed 3–4 times with the recommended buffer. Red blood cell (RBC) lysis was performed during the washing steps if required. The monocytes and CD4⁺ T cells were validated by flow cytometry.

4.19. Fixable Viable Dye (FVD) Staining

FVD staining was performed to determine the percentage of dead cells following EV challenge. Freshly isolated PBMCs, monocytes, or CD4⁺ T cells were resuspended in RPMI‐1640 medium supplemented with 10% EV‐depleted FBS and 1% Penstrep. Cells were challenged with 100 µg of MVs for 1 h, while untreated cells served as a negative control. Following incubation, cells were harvested and stained with FVD. Briefly, 2 × 10⁵ cells were resuspended in 98 µL of FACS buffer (PBS containing 2% FBS), and 2 µL of FVD working solution (prepared by diluting the stock 1:1000 in PBS) was added, after which excess dye was removed by washing the cells twice with FACS buffer. The stained cells were immediately analysed on a CytoFLEX S flow cytometer, and dead cell populations were assessed using the APC‐A750 detection channel.

4.20. Statistical Analysis

Data were statistically analysed using GraphPad Prism v.9.1.2. Data represent mean ± SEM unless otherwise specified. Experiments were performed with three biological replicates with triplicates unless specified otherwise. Pairwise comparison of normalised data was analysed using unpaired t‐tests. Differences were considered significant at ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001.

Author Contributions

Aswathy C. Sagilkumar: conceptualization, methodology, formal analysis, investigation, data curation, Writing – original draft, visualization. Avinashi Lal Kushwaha: methodology, formal analysis, investigation, data curation, writing – original draft, visualization. Sruthika Sukumar: investigation. Jagannath Mondal: methodology, formal analysis, writing – review and editing, supervision, funding acquisition. Karthik Subramanian: conceptualization, methodology, writing – original draft, writing – review and editing, supervision, funding acquisition. Somnath Dutta: methodology, formal analysis, investigation, data curation, visualization. Anushka Shitut: methodology, data curation, investigation, formal analysis, visualization, writing – original draft. Subhajit Mahanty: methodology, formal analysis, data curation, investigation. Dheeraj Kumar Sarkar: methodology, investigation, visualization, formal analysis, writing – original draft.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information: jev270387‐sup‐0001‐SuppMat.docx

Supporting Information: jev270387‐sup‐0002‐VideoS1.mpg

Supporting Information: jev270387‐sup‐0003‐VideoS2.mpg

Sagilkumar, A. C. , Kushwaha A. L., Shitut A., et al. 2026. “Extracellular Vesicle‐Bound Bacterial Toxin Pneumolysin Triggers Membrane Engagement and Damage Beyond Canonical Pore Formation.” Journal of Extracellular Vesicles 15, no. 10: e70387. 10.1002/jev2.70387

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.