All clinical isolates were obtained as previously described in Russell and Kyle (2022). Trophozoites were cultured axenically as previously described with some changes according to desired uses. Shortly, trophozoites were grown from stocks axenically at 34°C and 5% CO₂ in non-vented 75-cm² tissue culture flasks (Olympus, El Cajon, CA, United States; cat#:25-208) with Nelson’s complete medium (NCM) supplemented with 10% fetal bovine serum (FBS; Corning, Oneonta, NY, United States; cat#:35-016-CV) and 1,000 U/mL penicillin and 1,000 mg/mL streptomycin (penstrep; Gibco, Gaithersburg, MD, United States; cat#:15140-122) until 80–90% confluent.

All mammalian cells were grown in vented 75-cm² tissue culture flasks (ThermoFisher, Waltham, MA, United States; cat#:156367) at 37°C and 5% CO₂. A549 cells (Human lung carcinoma cells; ATCC CCL-185) were purchased from American Type Culture Collection (ATCC) and grown in F12K media (Corning, Oneonta, NY, United States; cat#:10-025-CV) supplemented with 10% FBS and 1% penstrep. B103 cells (rat neuroblastoma cells) were purchased from AddexBio (cat#:C0005003) and were grown in DMEM (Corning, Oneonta, NY, United States; cat#:10-013-CM) supplemented with 10% FBS and 1% penstrep. HFF cells (Human foreskin fibroblasts; HFF-1 ATCC SCRC-1041) were purchased from ATCC and were cultured in DMEM supplemented with 15% FBS and 1% penstrep. Vero cells (green monkey kidney cells; E6; ATCC CRL-1586) were purchased from ATCC and were cultured in DMEM supplemented with 10% FBS and 1% penstrep. For passaging, all cell types were first washed with pre-warmed PBS to remove residual serum and then incubated with pre-warmed 0.25% Trypsin–EDTA (Gibco, Gaithersburg, MD, United States; cat#:25200-056) for 5 min at 37°C. Following incubation, flasks were lightly tapped until all cells were detached and respective pre-warmed media supplemented with FBS was added to inactivate the trypsin prior to centrifugation at 37°C for 5 min at 3,900 rpm and resuspension in respective media.

For imaging axenic cultures, 5 × 10⁵ amoebae were allowed to attach to 13 mm Thermanox coverslips (Electron Microscopy Sciences, Hatfield, PA, United States; cat#:50-949-480) in 12-well plates prior to fixation. For feeding assays, 2.5–6 × 10⁵ mammalian cells were seeded into plates containing coverslips and allowed to attach/grow for 18–24 h before adding 3–6 × 10⁵ amoebae that were allowed to feed for various timepoints prior to fixation. Media was initially gently removed from samples and replaced with 2.5% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, United States; cat#:16220) diluted in respective base media (no FBS or antibiotics added) and incubated for 1 h at RT. This was replaced with 2.5% glutaraldehyde diluted in 0.1 μm filtered 1X PBS (Gibco, Gaithersburg, MD, United States; cat#:10010-023) and incubated for 10 min at RT prior to being washed with PBS two times, each with a 10 min incubation period. Samples were then incubated for 5 min each with serially increasing concentrations (30, 50, 70, 90%) of molecular biology grade absolute ethanol (EtOH; Fisher BioReagents, Pittsburgh, PA; cat#:BP2818-500) diluted in 0.1 μm filtered MilliQ H2O. This was followed by two incubation periods of 5 min each in 100% EtOH. A final incubation step of 5 min in hexamethyldisilazane (Electron Microscopy Sciences, Hatfield, PA, United States; cat#:16700) was performed, liquid was removed, and samples were allowed to air dry in fume hood for 30 min – 1 h before being mounted on carbon-conductive tape (Ted Pella, Redding, CA, United States; cat#:16084-8) on microscope slides. Samples were transported to SEM facility, carbon-coated and imaged with a Hitachi SU70 scanning electron microscope at 2 kV. For preparation of EVs for imaging, EVs were diluted in 1 mL of 0.1 μm-filtered 1X PBS and this dilution was passed over a 13 mm 0.2 μm Whatman Nucleopore Track Etch Membrane (Cytiva, Marlborough, MA, United States; cat#:10417001) mounted in a 13 mm Swinnex Filter Holder (MilliporeSigma, Burlington, MA, United States; cat#:SX0001300) using a 3 mL syringe. Following this step, a new syringe containing 2.5% glutaraldehyde in 0.1 μm filtered PBS was attached and some of the contents were passed over the filter before capping and allowing to rest for 2 h at 4°C. Subsequent steps through imaging were performed in parallel with previously described samples in plates starting at 2.5% glutaraldehyde in PBS 10 min incubation step. Measurements were either performed upon pre-calibrated micrographs using the Quartz PCI SEM software (v8) or by manually calibrating to scale bars with ImageJ (v1.53k). Figures were generated using Adobe Photoshop (v23.5.2).

Amoebae were grown on glass coverslips under normal culturing procedures in 6-well plates until ~70–80% confluent. For fluorescence microscopy, samples were first fixed with 4% PFA and 0.5% glutaraldehyde for 15 min followed by a wash with PBS and a subsequent incubation with 2 μg/mL Hoescht 33342 (Invitrogen, Waltham, MA, United States; cat#:H21492) for 30 min. This was followed by another wash and a subsequent incubation step with 4 μL of the 1,000X DMSO suspension of SPY620-actin stain (Cytoskeleton Inc., Denver, CO, United States; cat#:CY-SC505) in a 1 mL stain solution for 30–45 min. A final wash with PBS was performed prior to mounting and imaging with either the DeltaVision II (pd20621) microscope or the Carl Zeiss Elyra 7 microscope. For the DeltaVision II, we show the maximum intensity projection of z-stacks imaged with the 100X objective that were deconvolved using the SoftWorx software (settings: enhanced ratio(aggressive), 10 cycles, medium(200 nm) noise filtering). For the Zeiss Elyra 7, a z-stack of 106 slices (9.546 μm) was obtained with the 63X objective and the Lattice SIM² reconstruction algorithm was used to reconstruct the images and generate a 3D rendering (Grating Period: 617.32 nm, Processing: 3D, Input SNR: Medium, Iterations: 16, Regularization Weight: 0.065, Processing and Output Sampling: 4, Filter: Median, Detrend: No, Sectioning: 100, Baseline: Yes).

The RealTime-Glo MT Cell Viability Assay (RTG; Promega, Madison, WI; cat#:G9712) previously described by Colon et al. and Rice et al., was used to assess whether any differences in viability of Vero, A549, or B103 cells occurred in response to exposure to Nf-secreted EVs (Colon et al., 2019; Rice et al., 2020). As shown in Supplementary Figure S5, we first determined the optimal seeding density of cells within 96-well plates for each mammalian cell line by performing dilution series testing (2,500 to 40,000 cells/well) at various timepoints (6 h, 12 h, 24 h and 36 h) using the CellTiter-Glo 2.0 assay as described in our previous publication (Russell and Kyle, 2022). Being that 5,000 cells fell within the linear range of the assay at all timepoints tested (except for HFFs which were excluded from this assay due to a lack of linearity), we selected this concentration and seeded cells with four technical replicates per concentration tested in a volume of 35 μL per well. Cells were allowed to attach and then serum-starved by adding non-supplemented base media 1–2 h prior to adding EVs. The protein concentration of freshly extracted brain passaged Nf69 EVs was determined, and dilutions were made as necessary to allow a standard volume of 15 μL to be added to each well. The same process was followed for frozen stocks that were pulled from −80°C storage and thawed on ice for testing. Prior to adding EVs, the RTG substrate and enzyme were equilibrated to 37°C and mixed in respective base growth medias of each cell line. EVs were added to cells and 50 μL of the 2X RTG enzyme-substrate mixture was immediately added to obtain a 1X concentration. Plates were quickly sealed and incubated in a SpectraMax i3x plate reader (Molecular Devices, Sunnyvale, CA, United States) at 37°C and relative luminescence units (RLUs) were recorded every 2.5 min for 20 h. Negative/untreated controls were seeded with four replicates per cell line and included 15 μL PBS + cells, 15 μL media + cells, 50 μL media only, and 50 μL PBS only. EV only controls were included for each replicate and consisted of 1 μg of EVs (15 μL) in 35 μL of PBS. Initial dilution series testing of varying EV protein concentrations exposed to B103 cells was performed (Supplementary Figure S6) prior to proceeding with 1 and 0.5 μg. Data was analyzed and graphed using GraphPad Prism 9 (GraphPad Software, La Jolla, CA, United States; v9.5.0).

We utilized SEM to visualize and compare five axenically cultured clinical isolates of N. fowleri. When analyzing each isolate, we made morphometric observations and took measurements of structures that are labeled throughout Figure 1 and summarized in Figure 1F. When comparing between culturing conditions (axenic or fed over mammalian cells), we found no significant differences in the average width of each individual isolate, however we did discover statistically significant differences when comparing the overall widths among isolates as determined by one-way analysis of variance (ANOVA) [F(4,893) = 6.983; p < 0.0001]. The average combined width of the five clinical isolates (composed of at least 3 measurements of length and/or width per amoeba ± s.e.m.) was 10.3 ± 0.6 μm (n = 898 amoebae measured), with Nf69 being significantly larger than both 6088 (p = 0.0012) and V631 (p = 0.0001), and HB4 also being larger than V631 (p = 0.0012; post hoc Tukey comparisons; Table 1; Figure 1F).

We took note of one structural motif known as the uroid—an organelle previously observed on the posterior end of N. fowleri trophozoites with trailing filaments or processes (Carter, 1970; Visvesvara et al., 2005). It is referred to as an excretory organelle in limax amoebae (Vickerman, 1962), the potential location of the contractile vacuole in N. fowleri (Martinez et al., 1971), and a structure potentially involved in releasing vacuoles, waste, and excretory granules in Entamoeba histolytica (Hopkins and Warner, 1946). Our current work shows that the uroid of N. fowleri consists of clusters of hemispherical bulges of the membrane (rounded blebs or clusters of membranous invaginations) and was identified in all clinical isolates analyzed except for 6088, although it was only visible in a subset of the amoebae. The size of the uroid regions (reported in Table 1 and Figure 1F) seemingly correlated with the width of the isolates as evidenced by the largest isolate, Nf69, boasting a significantly larger uroid region compared to V067 (p = 0.0081), to V631 (p = 0.0002), and to HB4 (p = 0.0109) [one-way ANOVA and post hoc Tukey comparisons; F(3,132) = 6.540; p = 0.0004].

Often, we observed bundles of thin filaments with occasional bulbous tips—herein referred to as uroid filaments—that extruded from the uroid region, adhered to the substrate, and were repeatedly found broken off on either the mammalian cell or the substrate surface surrounding imaged amoebae. Other filaments with occasional bulbous tips that we suggest are adhesive filaments, often extended from the membrane around the entire cell periphery. We measured the diameters of these filaments in axenic cultures and report them in Table 1. Lastly, we have confirmed the observation made by Antonios et al. that no or very few organized “suckers” or food-cups/amoebastomes were identified on the trophozoite stages examined (Antonios, 2010). Contrary to some reports of one or more food cups per amoeba (John et al., 1984; Sohn et al., 2010), easily identifiable food cups were rare, and only identified in a small fraction of images taken of axenically cultured amoebae as well as those cultured over mammalian cells (Figures 1A,F; Table 1; Supplementary Figures S1A,D,H,J).

Evidence shown by John and John (1994) indicates that allowing N. fowleri clinical or environmental isolates to feed over mammalian cells for multiple passages confers higher levels of pathogenicity, thus we reasoned that this mechanism could be leveraged in a controlled visual assay for understanding the structural differences arising from actively feeding versus growing axenically in culture. We exposed the amoebae to three different food sources: human foreskin fibroblasts (HFF), B103 rat neuroblastoma cells (B103), and Vero green monkey kidney cells. All isolates, except for 6088, readily consumed HFF and Vero monolayers, while feeding on B103s was rare and commonly resulted in amoeboid encystation. Initial criteria for the selection of mammalian cells to monitor sub-micron level interactions among amoebae and their food allowed for the exclusion of B103s and HFFs as these both produce long/fibrous cellular connections and raise from the substrate at times when growing in a semi-confluent to confluent monolayer (Supplementary Figures S2A,C). Vero cells were selected as these produced uniformly flat monolayers with little to no visual fibrous intracellular connections or extracellular nanotube structures, thereby allowing for the least confounding identification and differentiation between amoebae and mammalian cells (Supplementary Figures S2E,G). Cytopathic effects were noted for all strains and cell lines when cultured with amoebae (Figure 2C; Supplementary Figure S2) and consisted of initial apoptotic bleb formation (inset in Figure 2C) followed by rounding of the mammalian cell accompanied by partial detachment from the substrate and the formation of long thin dendritic fibrils extending from main cell body. When these thin fibrils are formed, the amoebae attach to them and cluster along them (regardless of whether they are still associated with the substrate), and a common phenomenon seen when culturing over mammalian cells are networks of floating mammalian cell connections with dense clusters of amoebae dispersed along their lengths (example in right side of Figure 2C).

Prior work on N. fowleri by Antonios et al., described a thin surface extension when comparing the morphology of axenically cultured amoebae versus those freshly extracted from a brain and speculated that they played a role in adherence to brain tissues (Antonios, 2010). In our study, we also noted the formation of elongated thin filopodia extending from amoebae to mammalian cells (Figure 2Ai,ii). Due to their thick nature and extension over mammalian cells, we speculate that these could provide a basis for material exchange (inset in Figure 2Ai). Additionally, in an axenically cultured V631 prep, we noted a thick filopodial extension connecting three different trophozoites (Figure 2E) which could contribute to biofilm formation. We also observed pseudopodia with raised topographies extended onto mammalian cells, sometimes with thinner tubular structures connecting them to the main cell body (Figure 2B), but mostly with the canonical fan-shaped pseudopodia (Figure 2Bi,ii,iii; Supplementary Figures S1C,F,L). We speculate that, due to the raised nature of the pseudopodia, there are likely amoebastomes or other structures facilitating material exchange being formed at the interface between the pseudopodia and the mammalian cell, but further characterization was not possible due to the limitations of SEM.

Another phenomenon that we noted was the formation of thin branching filaments—sometimes emanating from the uroid region (Figures 2C,D), sometimes surrounding the entire cell periphery (Figure 2F; Supplementary Figure S1C)—that seem to be in contact with other amoebae and/or mammalian cells. We hypothesize that this could be a method of probing of the environment employed by the amoebae. We show that these thin filopodial extensions are formed via actin polymerization and can be visualized with high resolution fluorescence microscopy (Figures 3A–G; Supplementary Video S1). Moreover, we also observed the formation of thin filamentous connections from cysts to neighboring trophozoites (Supplementary Figures S1F,I). We speculate that this is either a tethering technique or a technique for intracellular signaling between cysts and/or trophozoites and the environment. The cysts exhibited the canonical rounded/oval form—sometimes with an exit pore (Supplementary Figure S1F) – and a finely reticulated membrane texture previously described by Lastovica (1974).

To ascertain the structure and mechanism of secretion of Nf-EVs, we designed an ultracentrifugation and purification technique following the Minimal Information for the Study of Extracellular Vesicles (MISEV) guidelines (Théry et al., 2018), and adapted from Bayer-Santos et al. (2013) and Szempruch et al. (2016) to extract vesicles from amoeba-conditioned media of four of the five isolates analyzed in this study. We extracted EVs from all isolates except 6088 as it was particularly sensitive to the transition from normal culture media to the EV-depleted fetal bovine serum (FBS) supplemented media. Nanoparticle tracking analysis revealed an average diameter of 152.6 nm for Nf-EVs across all isolates (Table 1; Figure 1F; Supplementary Figures S3A–D). Upon examination of freshly extracted Nf69 and V631 EV suspensions passed over a filter, we observed clusters of Nf-EVs with spherical morphologies (Figures 4A,B). We took measurements of individual Nf-EVs found on the filter, including those with easily interpretable edges and excluding those associated with clusters, and present them in Table 1. Following this, we examined the SEM images of amoebae and took measurements of potential EVs being secreted from the cell membrane or from filaments around the cell that are summarized in Table 1 and Figure 1F. Our results coincide with the range of 131–172 nm reported by Lertjuthaporn et al. (2022) and 43.88–216 nm reported by Retana Moreira et al. (2022). When comparing the sizes of measured Nf-EVs across the 3 measurement techniques and also among the 5 isolates, we found them to be significantly different via two-way ANOVA (across measurement techniques: [F(2,314,849) = 15.91; p < 0.0001]; among isolates: [F(4,314,849) = 27.42; p < 0.0001]; Figure 1F; Supplementary Table S1). In short, among the NTA preparations, V631, HB4 and V067 EVs were significantly larger than Nf69 EVs; V631 EVs were significantly larger than V067 EVs; and HB4 EVs were significantly larger than V067 EVs. When comparing NTA to filter measurements, Nf69 and V631 NTA EVs were significantly larger than their counterparts measured on the filter. This could be due to clusters of EVs being detected as a single EV in NTA and our exclusion of these clusters from our manual measurements of SEM images.

To determine the potential source of the secreted EVs, we scanned our SEM images for secretions from amoebae that resembled the EVs visualized on filters and identified three probable sources (Figures 4C–G). Firstly, we observed that the uroid/adhesive filaments commonly separate into vesicles via beading or pearling (Rilla, 2021) and break off onto the substrate leaving behind clusters of material (Figures 4C,D,F,G). This source of EVs from filopodial extensions has been described for numerous types of mammalian cells (Lai et al., 2015; Mathieu et al., 2019; Rilla, 2021). To support this phenomenon in N. fowleri, we mined the images of Nf-EVs passed over filters and found an example of a filament associated with clusters of EVs in a similar manner to those that we visualized in amoebae cultures (Figure 4H). Secondly, we noted the release of clusters of EVs from the periphery of the cell membrane (Figure 4E) which could indicate emission of vesicles via multivesicular bodies (MVBs). Lastly, we observed clusters of materials being released from regions of the plasma membrane (Supplementary Figure S1D). Upon taking measurements of the adhesive/uroid filaments, the putative vesicles formed via beading of filaments, and the clusters of putative vesicles secreted from the plasma membrane, we found that the size range overlaps with measurements that we obtained via NTA and manually on filters.

To determine whether Nf-EVs are taken up by mammalian cells, we labeled Nf69-EVs with the lipophilic fluorophore octa-decyl rhodamine B (R18) and performed fluorescence dequenching assays to monitor time-dependent increases in fluorescence that would be indicative of EV uptake by recipient cells. We first exposed B103 cells to Nf69-secreted R18-stained EVs for various timepoints and imaged the resulting cells (Figure 5). At the 5-min timepoint we noted initial dim lipid equilibration with B103 cell body membranes indicated by dim/diffuse staining through the 30-min timepoint (Figures 5E–H) compared to the R18-stained PBS control cells (Figures 5A–D). As more EVs were taken up by host cells over time, more of the R18 dye was diluted resulting in brighter fluorescent signals on the host cell membranes (Figures 5I–P). Initial punctate staining patterns occurred over the cell body, and this escalated to bright coverage of the full membrane (including the axons and dendrites) until fluorescence reached a plateau at ~6 h (Figures 5Q–T). We next quantified levels of Nf69-EV uptake for different cell lines at the population level (as a function of integrated fluorescence intensity) by using different concentrations of Nf69-EVs with B103, HFF, Vero, and A549 cells. We developed a 96-well plate assay using high-content imaging to detect host cell nuclei which were defined as Hoescht+ and then host cytosol defined as R18−/+. When quantifying the mean fluorescent intensity of double positive cells (host nuclei and cytosol), we observed a time and concentration-dependent increase in mean integrated fluorescence intensity per treatment group that plateaued at approximately 6–12 h (Supplementary Figure S4).

To determine whether EVs are taken up by other amoebae, we grew the amoebae in suspension and treated V631 with 1 μg of Nf69-secreted R18-stained EVs for various timepoints and performed imaging flow cytometry. We used Hoescht+ gating to select for viable amoebae followed by R18+ gating to quantify fluorescence levels of each treatment population of amoebae. Similar to mammalian cells, we saw initial rapid lipid equilibration with amoeba cell membranes (Figure 6A, 5 min) and this was followed by increasing numbers of punctate staining patterns (Figure 6A, 15 min – 2 h) indicative of vacuolized EVs taken up by the amoebae via phagocytosis with increasing levels of uptake over time (Figure 6B).

To determine if there are any measurable effects on cellular metabolism in real-time after treating cells with Nf69-EVs, we first optimized the seeding densities of tested cells using the CellTiter-Glo 2.0 kit in 96-well plates to confirm that the luminescence readings fell within the linear range of the RealTime-Glo MT cell viability assay at various timepoints throughout the planned incubation period (Supplementary Figure S5). The RealTime-Glo MT kit allows real-time monitoring of cell viability depending on the ability of exposed cells to take up and metabolize/reduce the prosubstrate to a substrate that is exported from the cell to bind to the NanoLuc Enzyme and produce the bioluminescence (in RLUs) that is measured as an output for the assay. We anticipated a decrease in cellular viability in response to Nf-EV uptake according to prior work showing that the EVs of another free-living amoeba, Acanthamoeba castellanii, induce cytotoxic effects in mammalian cells (Goncalves et al., 2018). On the contrary, we observed an increase in RLUs in the treatment groups compared to control groups regardless of low versus high doses, or the cell line used (Supplementary Figure S6; Figure 7). We observed the same phenomenon when we compared the efficacy of freshly extracted and processed EVs to those that were frozen immediately after processing and stored at −80°C for 14 weeks prior to thawing for the assay (Supplementary Figure S7).

To date, little is known about the protein composition of Nf-EVs. A preliminary visualization of protein contents of Nf69-EVs via SDS-PAGE gel is shown in Supplementary Figure S3E and the presence of at least 24 bands indicates that many different protein populations are present in the secreted EV proteome. To elucidate these proteins, we performed LC–MS/MS on two independent preparations of Nf69-EVs extracted from amoeba-conditioned media and identified 2,270 proteins present within both samples that represent the EV proteome (Supplementary Table S2). This represents a subset of 16.7% of the full 13,596 protein N. fowleri proteome (Zysset-Burri et al., 2014). These EV proteins were run through the Blast2GO suite, and the resulting annotations are also provided in Supplementary Table S2. We then compared our proteomic profile to the previously reported N. fowleri EV proteome and found that out of the 184 proteins reported by Retana Moreira et al. (2022), 18 proteins were specific to the non-pathogenic Naegleria gruberi. Of the remaining 166 N. fowleri specific proteins, 16 were not recapitulated in our study, resulting in an overlap of 150 proteins (highlighted in yellow in Supplementary Table S2) that are reiterated in our proteomic profile. We utilized the Panther classification system (Thomas et al., 2022) to identify enriched protein classes or functionalities. Out of the 2,270 proteins submitted to Panther, 1,876 were recognized for each of the 3 classification schemas. Graphical summaries of the Panther results are presented in Figure 8 and Supplementary Figures S9A,B with 1,430 protein class hits, 1,312 molecular function hits, and 2,109 biological process hits. The 5 categories in the Protein Class schema with the highest prevalence were: metabolite interconversion enzyme (26.6%; 380 proteins), protein-binding activity modulator (14.8%; 211 proteins), transporter (8.5%; 121 proteins), translational protein (7.8%; 112 proteins), and membrane traffic protein (6.9%; 98 proteins). The 5 categories in the Biological Process schema with the highest prevalence were: cellular process (38.3%; 808 proteins), metabolic process (22.0%; 465 proteins), localization (12.7%; 267 proteins), biological regulation (11.7%; 246 proteins), and response to stimulus (7.0%; 147 proteins). Lastly, the 5 categories in the Molecular Function schema with the highest prevalence were: catalytic activity (45%; 591 proteins), binding (31.5%; 413 proteins), transporter activity (5.9%; 77 proteins), molecular function regulator (5.4%; 71 proteins), and ATP-dependent activity (3.3%; 43 proteins). A direct comparison of PANTHER protein class hit categories for the entire N. fowleri proteome (Supplementary Figure S8) compared to hits for the Nf-EV proteome is provided in Supplementary Table S3.