Fresh C. parvum oocysts (Iowa strain) were supplied by Bunch Grass Farm (Deary, Idaho USA), stored at 4°C in PBS with penicillin (1000 U.I./ml) and streptomycin (1 mg/ml). The excystation procedure was performed as follows: aliquots of 1x10⁷ oocysts per ml were pelleted at 376 × g, for 5 min, resuspended in 1 ml of 10 mM HCl, and then incubated at 37°C for 10 min. Oocysts were pelleted again as above and resuspended in 1 ml of excystation medium (D-MEM containing 2 mM sodium taurocholate); they were maintained at 15°C, for 10 min, and then moved to 37°C to induce the excystation. Excystation mixtures were sampled at various times after the induction.

The CpRom1 and CpGRASP coding sequence (monoexonic genes without introns) were directly amplified from C. parvum genomic DNA (Iowa strain).

Sporozoite microvesicles were prepared from the excystation medium of fresh oocyst aliquots after 2 h incubation at 37°C; excystation was blocked by placing samples on ice for 5 min. All subsequent manipulations were performed on ice and centrifugation steps were performed at 4°C. To remove most of oocysts and sporozoites, the excystation medium was centrifuged at 376 × g, for 20 min; the resulting supernatant was centrifuged again at 2,348 × g, for 10 min, to remove residual oocysts, sporozoites, and excystation debris. To isolate EVs, supernatant was transferred to 1.5 ml tubes (Eppendorf), centrifuged for 20 min at 10,000 × g and the pellet containing LEVs resuspended in 100 μl of ice-cold PBS. To isolate SEVs, the supernatant was transferred to new tubes (Beckman Coulter) and ultracentrifuged on a TLA 120.2 rotor (Beckman Coulter) at 100,000 × g for 3 h. The supernatant was discarded, and the pellet was resuspended in 1.2 ml of ice-cold PBS and recentrifuged as above. The final pellet containing SEVs was resuspended in 100 μl of ice-cold PBS. All the procedure is schematized in Figure 1 .

Proteins in the excystation medium and supernatants after ultracentrifugation were precipitated with 10% ice-cold trichloroacetic acid (TCA)/0.015% (w/v) sodium deoxycholate (DOC). After 10 min on ice, samples were microfuged and the pellets were washed twice with 0.5 ml of ice-cold acetone, air dried, resuspended in 50 μl of lysis buffer (1% w/v SDS, 1% v/v Triton X-100, 0.5% w/v DOC, 10 mM 1,4-dithiothreitol (DTT), 1 mM EDTA) containing 1 μl/ml of Protease Inhibitor Cocktail (Sigma-Aldrich), and incubated for 5 min at 70°C in ThermoMixer® (Eppendorf).

Pellets after centrifugation (LEVs) and ultracentrifugation (SEVs) were resuspended in 50 μl of lysis buffer containing 1 μl/ml of Protease Inhibitor Cocktail (Sigma-Aldrich) as above. Samples were boiled in Laemmli sample buffer before separation by SDS-PAGE on 4-20% TGX™ precast gels (Bio-Rad, Hercules, CA, USA). Gels were transferred to nitrocellulose membranes (Bio-Rad), which were then blocked in PBS with 5% non-fat milk in 0.1% Tween-20. Blots were incubated with primary antibodies, at 4°C, overnight. For recombinant 6His-CpRom1 and 6His-CpGRASP, blots were incubated with mouse monoclonal RGS-His antibody (Qiagen) diluted 1:1,000. Blots for CpRom1 were probed with mouse pre-immune and CpRom1immune serum 1 diluted 1:250, while blots for CpGRASP were probed with mouse pre-immune and anti-CpGRASP serum diluted 1:100. For all blots, incubation with secondary antibodies was conducted for 1 h, at 25°C, with goat anti-mouse IgG-HRP conjugate (Bio-Rad, Hercules, CA USA) diluted 1:3,000. Detection of proteins was performed using Pierce ECL substrate (ThermoFisher) and bands were visualized with a ChemiDoc MP Imaging System (Bio-Rad, Hercules, CA, USA).

For negative staining in transmission electron microscopy (TEM), samples of LEVs and SEVs were suspended in 50 μl of 2.5% w/v glutaraldehyde in 0.1 M sodium cacodylate and incubated at 4°C, overnight. Samples (20 μl) were deposited by successive applications on carbon-coated grids for electron microscopy; they were left to adsorb for 20 min, and the excess fluid was blotted with filter paper. A contrasting solution (5 μl) made of 2% w/v phosphotungstic acid was added on grids and air dried. Samples were observed by a PHILIPS Morgagni 268 TEM (FEI - Thermo Fisher) (Théry et al., 2006). For immunolocalization, LEVs and SEVs pellets were suspended in 100 μl of PBS, and aliquots of 20 μl were adsorbed on carbon-coated grids for electron microscopy, for 20 min. Samples were air dried and transferred to a drop of 2% w/v paraformaldehyde, for 5 min. After two washes with PBS, the grids were floated on PBS drops containing 0.1 M glycine for 30 min, washed with PBS, blocked with PBS containing 5% goat serum and 1% w/v BSA, for 30 min, and washed with PBS containing 0.05% w/v TWEEN 20 and 1% w/v BSA (PBS/BSA/TW). Then grids were floated on rabbit polyclonal anti-ID2 serum (1:10) in PBS/BSA/TW, for 1 h, at 25°C, rinsed in the same buffer, incubated on 10 nm gold-conjugated goat anti-rabbit IgG (SIGMA) (1:50), for 1 h, rinsed again in buffer PBS/BSA/TW and treated with 1% glutaraldehyde, for 5 min. Finally, the samples were contrasted and embedded in a 1:9 mixture of 4% uranyl acetate and 2% methyl cellulose, respectively, for 10 min, on ice. The samples were air dried and analyzed by a PHILIPS Morgagni 268 TEM (FEI - Thermo Fisher) (Théry et al., 2006).

For ultrathin sections, 1 ml of excystation mixture (taken 30 min after induction) was loaded onto a 1 ml syringe and filtered with a 5 μm disposable filter Millex-SV (Millipore) to remove residual non-excysted and empty oocysts. Purified sporozoites were fixed in 2.5% w/v glutaraldehyde, 2% paraformaldehyde, 2 mM CaCl₂ in 0.1 M sodium cacodylate, pH 7.4, and processed according to Perry and Gilbert (1979). Parasites were washed in cacodylate buffer and post-fixed with 1% OsO₄ in 0.1 M sodium cacodylate, for 1 h, at 25°C; then, they were treated with 1% tannic acid in 0.05 M cacodylate buffer, for 30 min, and rinsed in 1% sodium sulphate, 0.05 M cacodylate, for 10 min. Post-fixed specimens were washed, dehydrated through a graded series of ethanol solutions (from 30% to 100% ethanol) and embedded in Agar 100 (Agar Scientific Ltd, UK). Ultrathin sections, obtained by an UC6 ultramicrotome (Leica), were stained with uranyl acetate and Reynolds lead citrate and examined at 100 kV with a PHILIPS EM208S TEM (FEI - ThermoFisher) equipped with the Megaview III SIS camera (Olympus). Statistical analysis of vesicle sizes was performed with GraphPad Prism 10 (Dotmatics).

1x10⁷ C. parvum oocysts (see above) were induced to excystation. At various times (from 0 to 30 min), the samples were fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, and 30 μl were left to adhere to polylysine-treated round glass coverslips (Ø 10mm), for 1 h, at 25°C. Samples were processed for scanning electron microscopy (SEM) as previously described with slight modifications (Shively and Miller, 2009). Briefly, samples were post-fixed with 1% OsO₄ in 0.1 M sodium cacodylate, for 1 h, at 25°C, and were dehydrated through a graded series of ethanol solutions (from 30% to 100%). Then, absolute ethanol was gradually replaced with a 1:1 solution of hexamethyldisilazane (HMDS) in absolute ethanol, for 30 min, and successively by pure HMDS, for 1 h. Finally, HMDS was completely removed, and samples were left to dry in a desiccator, at room temperature, for 2 h, overnight. Then dried samples were mounted on stubs, carbon coated and analyzed with a FE-SEM Quanta Inspect F (FEI, Thermo Fisher Scientific).

Fluorescent labelling of proteins of LEVs and SEVs was conducted as follows: pellets were resuspended in 50 μl PBS containing 10 µM Alexa Fluor 647 NHS Ester (Life Technologies) and incubated for 30 min, at 25°C. Then pellets were resuspended again in 50 μl PBS containing 10 μM CFDA-SE (CFSE) (Life Technologies), for 30 min, at room temperature. Reactions were stopped by adding 2 μl of 100 mM L-glutamine. Fluorescent LEVs and SEVs from the corresponding pellets were analysed by flow cytometry (FC) with a Gallios Flow Cytometer (Beckman Coulter) using an optimised procedure, as previously described (Coscia et al., 2016). Briefly, the instrument was set using the Flow Cytometry Sub-micron Particle Size Reference Kit providing green-fluorescent beads of different sizes (Invitrogen™) to establish the correct threshold value to apply on the 525/40 nm fluorescent channel (FL1). This procedure allowed to create gates of the following dimensions: ≤200 nm, 500 nm and 1 µm. Sample analysis was performed by mixing 5 µl of fluorescent vesicles with 20 µl of Flow-Count Fluorospheres (Beckman Coulter) in a final volume of 200 µl of PBS to determine absolute counts. Fluorescent populations were analysed by plotting fluorescence at 525/40 nm (FL1) versus log scale side scatter (SSarea). Sample acquisition was stopped at 2,000 Flow-Count Fluorospheres events. The total number of fluorescent vesicles was established according to the formula: x = {[(y × a)/b]/c} × d, where y = events counted at 2,000 counting beads; a = number of counting beads in the sample; b = number of counting beads registered (2,000); c = volume of sample analysed; and d = total volume of exosome preparation. Kaluza Software v. 2.0 (Beckman Coulter) was used for FC analysis.

Freshly labelled fluorescent LEVs and SEVs from 1 x 10⁷ oocysts were diluted in 0.3 ml of PBS. Then, fluorescent samples were mixed with 1 ml of 60% iodixanol solution (OptiprepTM, Sigma-Aldrich), overlaid with 0.5 mL of 40%, 0.5 mL of 30% and 1.8 mL of 10% of iodixanol solutions and floated into the gradient by ultracentrifugation in a SW60Ti rotor (Beckman) at 192,000 × g, for 18 h, stopping without brake. After centrifugation, 12 fractions of 330 µl were collected from the top of the tube, diluted 40- to 80-fold with PBS and analysed by FC. Fraction densities were determined by refractometry. Gradient solutions were produced from the working solution (WS) by dilution with the HM solution. WS was prepared by mixing 5 vol of OptiPrep™ with 1 vol of 0.25 M sucrose, 6 mM EDTA, 60 mM Tris-HCl, pH 7.4; HM solution also contained 0.25 M sucrose, 1 mM EDTA, 10 mM Tris-HCl, pH 7.4. Gradient fractions of fluorescent LEVs and SEVs were analysed by FC with a Gallios Flow Cytometer (Beckman Coulter) using an optimised procedure, as previously described (Coscia et al., 2016).

For nanoparticle tracking analysis (NTA), 20 × 10⁶ oocysts were excysted and EVs were collected in two fractions (LEVs and SEVs respectively) as above. LEVs and SEVs pellets were diluted in 500 µl PBS and quantified by NTA performed with a NanoSight NS300 system (Malvern Instruments, UK). Camera level was set at 14-15 for all recordings, until all particles were distinctly visible, and camera focus was adjusted to make particles appear as sharp individual dots. Three 60-second videos were recorded for each sample under the following conditions: cell temperature: 23°C; syringe flow: 40 µl/s. Detection threshold was set at 5 and other settings were kept at default. After capture, the videos have been analysed by the in-build NanoSight Software NTA 3.4.4.

Proteins from LEVs and SEVs were separated on a precast 4-15% T SDS-PAGE (Bio-Rad, Hercules, CA USA) and stained with Coomassie G-250. Gel lanes were cut into 9 slices that were separately in-gel reduced, alkylated with iodoacetamide and digested with trypsin, as previously reported (Salzano et al., 2013). Peptide mixtures were then desalted by μZipTipC18 tips (Millipore) before nano-liquid chromatography-electrospray ionization-Orbitrap tandem mass spectrometry (nLC-ESI-Q-Orbitrap-MS/MS) analysis, which was performed on a Q-Exactive Plus mass spectrometer equipped with UltiMate 3000 HPLC RSLC nano system (Thermo Fischer Scientific, USA). Peptides were separated on an Easy C18 column (100 × 0.075 mm, 3 μm) at a flow rate of 300 nl/min, using a linear gradient from 5 to 40% of acetonitrile containing 0.1% formic acid (solvent B) in 0.1% formic acid (solvent A), over 60 min. Mass spectra were acquired at nominal resolution 70,000 in the range m/z 350-1500, and data-dependent automatic MS/MS acquisition was applied to the ten most abundant ions (Top10), enabling dynamic exclusion with repeat count 1 and exclusion duration 30 s. The mass isolation window and the collision energy for peptide fragmentations were set to m/z 1.2 and 32%, respectively. Raw data from nLC-ESI-Q-Orbitrap-MS/MS analysis were searched by MASCOT v2.6.1, (Matrix Science, UK) node within Proteome Discoverer suite, against a C. parvum (strain Iowa II) database of protein sequences (4076 sequences) retrieved from CryptoDB site (https://cryptodb.org), and results were merged into a single mgf file to obtain proteins identified in LEVs and SEVs fractions. The following parameters were used for protein identification: mass tolerance values of 20 ppm and 0.05 Da for precursor and fragment ions, respectively; trypsin as proteolytic enzyme with maximum missed-cleavage sites of 2; Cys carbamidomethylation as fixed modification; Met oxidation, Asn/Gln deamidation and pyroglutamate formation at Gln/Glu as variable modifications. A significance threshold of p<0.05 was set for protein identification, and molecular candidates with at least 2 significantly matched peptide sequences and a protein Mascot score >50 were further considered for definitive assignment, after manual spectra visualization and verification.

The sequence of CpRom1 (CryptoDB: cgd6_760) and CpGRASP (CryptoDB: cgd7_340) were downloaded at CryptoDB site (https://cryptodb.org). Similarity searches with DNA and protein sequences were conducted on non-redundant GenBank databases using the BLAST program (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The prediction of transmembrane domains was performed at TMHMM-2.0 (http://www.cbs.dtu.dk/services/TMHMM/) and at Phobius (https://phobius.sbc.su.se/index.html). Prediction of structural and functional domains was performed at Pfam (http://pfam.xfam.org) and Prosite (https://prosite.expasy.org). The bidimensional representation of CpRom1 ( Figure 2 ) was realized with TMRPres2D software (http://bioinformatics.biol.uoa.gr/TMRPres2D/download.jsp). Searches for potential cleavage sites by proteases were performed at PeptideCutter (https://www.expasy.org/resources/peptidecutter) and at ProP 1.0 server (http://www.cbs.dtu.dk/services/ProP/) for Arg and Lys propeptide cleavage sites (furin-like). Searches for potential GPI-anchor were performed at GPI Modification Site Prediction (https://mendel.imp.ac.at/gpi/gpi_server.html). Subcellular localization was predicted by DeepLoc-1.0 (http://www.cbs.dtu.dk/services/DeepLoc/) and was reported only when the probability was higher than 50%. Protein-protein interaction networks were obtained with STRING v.12.0 (https://string-db.org). Searches for proteins associated with extracellular vesicles were performed in the following databases: ExoCarta (http://www.exocarta.org) and Vesiclepedia (http://www.microvesicles.org).

CpRom1 rhomboid was previously identified using the 46 amino acids long ID2 peptide (Trasarti et al., 2007). A prediction based on the protein sequence showed that CpRom1 consists of 990 residues and has a theoretical mass of 109.3 kDa; this makes CpRom1 the largest rhomboid ever described. Computational predictions showed that CpRom1 has a large N-terminal region of approximately 500 amino acids, which includes the ID2 peptide ( Figure 2 ). This large portion of the protein is exposed towards the extracellular environment and was not associated with a recognizable domain. A well-distinguishable rhomboid domain, consisting of 450 amino acids arranged in six transmembrane helices that compose the proteolytic site, was also predicted occurring at the protein C-terminal region.

To identify the native CpRom1 rhomboid, we prepared a mouse polyclonal serum reacting with the full-length protein. To this end, a fusion protein with six histidine-tag (6His-CpRom1) was expressed in bacteria, purified by Ni-NTA chromatography and used to immunize Balb-C mice. The polyclonal mouse antiserum was used in Western blot experiments on sporozoite lysates, in which a band of approximately 110 kDa was recognized, in agreement with the above-reported theoretical mass ( Figure 3 ). However, the serum also detected a smaller band migrating at about 75 kDa, which might represent a cleaved form of the 110 kDa protein. These CpRom1 forms were present in resting oocyst at time 0 of excystation ( Figure 3A ), with a similar amount of the 110 kDa form in unexcysted oocyst and sporozoites, whereas a lower amount of the 75 kDa species was detected after excystation. The search for putative cleavage sites was ineffective by the presence of hundreds of possible proteolytic cuts along the amino acid sequence. Probably, the latter protein form is quickly degraded after excystation. A significant amount of CpRom1 was also detectable in the supernatant of excystation medium after eliminating sporozoites and residual oocysts. In fact, TCA precipitates of excystation medium after 30 min of incubation, when probed with anti-CpRom1 serum, revealed a distinctive band of 110 kDa ( Figure 3B ). Given the presence of six transmembrane domain in CpRom1, which eventually hampers its release in the excystation medium, we hypothesized that CpRom1 is released in association with membranous vesicles.

To test the hypothesis reported above, we developed an ultracentrifugation protocol to separate extracellular vesicle populations from the excystation medium, after preventive removal of oocyst-sporozoite residues. A schematic representation of the procedure is shown in Figure 1 . Depending on centrifugation conditions, two types of C. parvum EVs were selectively recovered that were tentatively classified as large extracellular vesicles (LEVs) and small extracellular vesicles (SEVs). Protein extracts from LEVs and SEVs were then assayed for the presence of CpRom1 by dedicated immunoblotting. Results showed that a protein band of 110 kDa was observed in the LEVs extracts, but not in the SEVs counterparts ( Figure 4A ).

To demonstrate that CpRom1 is associated with EVs, we assayed C. parvum LEVs and SEVs by negative staining immunoelectron microscopy using a polyclonal rabbit serum directed against the ID2 peptide (Trasarti et al., 2007). In agreement with the immunoblotting results, numerous strong immunogold signals were detected on the surface of large vesicles (> 150 nm) occurring in LEVs ( Figure 4 and Supplementary Figure S1 ). A negative control with pre-immune rabbit serum showed sporadic signals with one or two gold particles for microscopic field ( Figure 4C ). We also tested SEVs by negative staining immunoelectron microscopy with the same anti-ID2 serum, and we observed small vesicles that were not labelled by immunogold particles ( Figure 4D ).

All together, these results suggested that sporozoites release two different types of extracellular vesicles, among which those present in LEVs are generally larger than those in SEVs. Immunolabelling for CpRom1 showed the presence of this protein only in LEVs.

To further characterize sporozoite EVs, we labelled vesicles in LEVs and SEVs sequentially with two different fluorescent probes: at first, we used the hydrophilic dye NHS-AF647, which labels proteins on the external surface of the vesicles; then, we utilized the lipophilic dye CFSE, which passively diffuses into vesicles and labels also internal vesicular proteins. Then, vesicles contained in LEVs and SEVs were separated by density gradient centrifugation on an iodixanol gradient, and the fluorescent content of the resulting fractions was analysed by flow cytometry. FC analysis showed that LEVs comprised a heterogeneous population of intact vesicles ranging in size from 50 nm to over 250 nm ( Figures 5A, B ). Notably, LEVs were often clumped together ( Figure 6A ), contributing to the highest density of extracellular vesicles according to FACS analysis. In contrast, SEVs displayed a more homogeneous composition ( Figure 5C ), consisting of smaller intact vesicles with sizes below 150 nm ( Figure 6B ).

EVs were also quantified by FC. We observed that LEVs are more numerous than SEVs; moreover, both vesicle populations presented a peak in fractions with density ranging between 1.08-1.14 g/ml ( Figure 6D ). To establish more specifically the size of vesicles occurring in LEVs and SEVs, the diameter of a hundred vesicles was measured through electron microscopy, and the corresponding micrographs were subjected to statistical analysis. This experiment revealed that LEVs ( Figure 6A ) and SEVs ( Figure 6B ) had a different mean size, with the former and latter vesicles showing an average diameter of 150 nm and 60 nm, respectively ( Figure 6C ).

Finally, LEVs and SEVs were also measured and quantified by NTA; results of this experiment revealed a different size distribution between the two types of vesicles. LEVs ( Figure 6D ) were distributed in three different peaks, with the main one at 176 nm; the corresponding modal class was of 181.3 nm ( Figure 6F ). Differently, most of SEVs were accumulated in a peak at 102 nm ( Figure 6E ) and the modal class was of 105 nm ( Figure 6F ). Full details of NTA are reported in Supplementary Report 1 for LEVs and Supplementary Report 2 for SEVs. Importantly, NTA allowed to estimate the total number of EVs released by sporozoites ( Figure 6F ), which were 5.8 10⁸ LEVs and 5.8 10⁸ of SEVs from the excystation of 2 10⁷ oocysts. Overall, this result showed that each sporozoite releases at least a dozen LEVs and SEVs after the excystation.

To capture images of vesicles at the time of their release from sporozoites, we performed SEM and TEM analysis of the the excystation mixture at various time points (from 0 to 30 min) after the induction. Hence, we observed some sporozoites immediately after the egress from two oocysts ( Figure 7A ). Oocysts appeared as thin, empty envelopes; oocysts and sporozoites were surrounded by numerous vesicles of varying sizes scattered throughout the microscopic field. Similarly, we captured images at higher magnification showing two vesicles of different sizes at the very moment of their release from the sporozoite membrane ( Figure 7B ). Finally, we observed emerging vesicles from the apical part of a sporozoite ( Figure 7C ) as well as a vesicle of approximately 200 nm just released from the posterior part of another sporozoite ( Figure 7D ). All together, these images supported the origin of EVs, at least the larger ones, through budding of the external membrane.

Since SEVs have physical parameters like exosomes, which originate from intracellular multi-vesicular bodies (MVBs) (Van Niel et al., 2018; Barreca et al., 2023), we looked for a similar structure in the sporozoite cytoplasm. Electron microscopy images showed a vacuolar structure with internal membranes that resemble MVBs ( Figures 8A, B ). A similar MVB-like structure has not yet been described in Cryptosporidium spp. so far.

No homologues of the most common mammalian proteins associated with exosomes, namely tetraspanins (e.g., CD63, CD81 or CD9) or endosomal sorting complex required for transport (ESCRT) proteins (eg, Alix and Tsg101) (Théry et al., 2018), were identified neither in the translated products of the Cryptosporidium genome through a dedicated bioinformatic analysis (data not shown) nor in the proteome of this parasites (Snelling et al., 2007) and its EVs (this study, see below). On the other hand, well-conserved Cryptosporidium homologs of proteins involved in vesicle trafficking were traced among those associated with the Golgi apparatus. In all eukaryotes, the Golgi complex is a central hub for the vesicle trafficking (Doyle and Wang, 2019), and Golgi proteins might represent markers for tracking various types of vesicles. The Golgi reassembly and stacking proteins (GRASPs) belong to a protein family with a conserved similarity in all eukaryotes, and the Cryptosporidium homolog of these components (CpGRASP) was easily identified in CryptoDB. Moreover, following a search in different protein databases of extracellular vesicles and in the literature, GRASPs have been found in different types of EVs (Chua et al., 2012; Peres da Silva et al., 2018).

Accordingly, we proceeded to clone the gene for the C. parvum homolog of GRASPs to express a histidine-fusion protein in bacteria. The purified recombinant protein (6his-CpGRASP) was then used to generate a specific antiserum in mice. When probed with the anti-GRASP serum in Western blot experiments, a protein with an apparent mass of about 80 kDa band was identified in the whole oocyst-sporozoite lysate ( Figure 9 ); its measured molecular mass matched the expected value of the C. parvum GRASPs homolog (82.49 kDa). The same component with mass of 80 kDa band was also observed in LEVs extracts, while no band was detected in the SEVs extracts or in the supernatant following ultracentrifugation. The latter result suggested that the biogenesis of LEVs involves Golgi-derived vesicles unlike SEVs, which conversely did not show any evidence for the occurrence of CpGRASP.

A first analysis regarding the protein content of sporozoite vesicles was performed by comparing the electrophoretic profiles of the two types of vesicles. To this end, we double-labelled LEVs and SEVs with NHS-AF647 (red) and CFSE (green) and analysed the corresponding protein profiles by SDS-PAGE (Dehghani and Gaborski, 2020). Figure 10 shows the image obtained by overlaying the green and red channels onto the electrophoretic protein profiles of LEVs and SEVs. Upon comparing the electrophoretic profiles, we identified distinct common bands (black arrows) as well as some prominent unique bands (white arrows) in SEVs. This method allowed a rapid generation of distinctive patterns for LEVs and SEVs and confirmed the occurrence of differences in the protein content of them. Of note, the SEVs profile exhibited a greater number of red bands indicating the likely presence of a greater number of surface proteins.

To identify C. parvum proteins present in above-reported vesicles, components of LEVs and SEVs were resolved by SDS-PAGE and stained with Coomassie G250. The resulting whole gel lanes (see Supplementary Figure S2 ) were divided into 9 slices that were further subjected to trypsinolysis and then to nLC-ESI-Q-Orbitrap-MS/MS analysis. Overall, we identified 60 C. parvum proteins of which 39 were in common between the two vesicle types; conversely, 5 were identified only in LEVs, and 16 were found exclusively in SEVs. All identified proteins are listed in Table 1 , and details of the corresponding proteomic analysis are reported in Supplementary Table S1 . The full annotation from CriptoDB for all assigned proteins is reported in Supplementary Table S2 , which also includes presumptive information on the corresponding function, gene ontology (GO terms ID), assignment as membrane/soluble component, and cellular localisation.

According to functional predictions, the largest group of annotated proteins (14 in number) included enzymes related to cell metabolism, while the second set (12 in number) consisted of components involved in protein folding, such as heat shock proteins (HSPs) and molecular chaperones ( Figure 11A ). This latter group included 4 heat shock proteins 70 (HSP70s) and 2 heat shock proteins 90 (HSP90s) that may have a different subcellular localization, as inferred by the variable presence of signals and/or transmembrane domains. Additional protein categories were related to protein synthesis (9 in number), regulatory function (6 in number), cytoskeleton (3 in number), redox homeostasis (2 in number) and proteolysis (2 in number) ( Figure 11A ). The presumptive function remained unknown for 12 proteins.

All the proteins were also analysed in silico to identify signal peptide, transmembrane domain and endoplasmic reticulum retention signal portions, which are indicative traits of their subcellular localization. Thus, the largest group was composed of soluble components (34 in number) with a putative cytoplasmic localization ( Figure 11B ). A second group of 7 proteins showed a signal for sorting and/or retention in the endoplasmic reticulum. Six secretory proteins were identified due to the occurrence of a signal peptide at their N-terminus, whereas three showed a transmembrane domain. Finally, 2 proteins were predicted as nuclear components. Location of eight proteins remained unclear, as the potential location was assigned to subcellular compartments (i.e. mitochondrion, plastid, and lysosome), the existence of which is still unknown in C. parvum.

A putative interactome of EVs proteins was modelled by STRING interaction analysis. This search revealed a big network connecting 51 components ( Supplementary Figure S3A and Supplementary Table S3 ), which was identified with a medium confidence (0.4). The involvement of most (85%) of the identified proteins in this network emphasized the occurrence of a functional assembly bridging different molecular processes, which seemed highly represented in EVs. When higher confidence (0.9) was used for the analysis, this interaction network was limited to include only 34 proteins (57%) and three main sub-networks ( Supplementary Figure S3 ) related to carbohydrate metabolism. Therefore, the above-reported interaction data highlighted the possible existence of functional macromolecular complexes in the C. parvum EVs involved in specific biological processes, as evidenced by the coherence with the most represented predicted functional categories ( Figure 11A ).

The occurrence of protein homologs assigned to EVs in other organisms was verified for 48 out of the 60 species identified in this study ( Table 1 , results from comparison with data reported in Vesiclepedia and ExoCarta database). Accordingly, this investigation confirmed that proteins generally active within the C. parvum cell’s interior environment can also be detected in microvesicles or exosome-like vesicles. On the other hand, this proteomic study originally identified 12 proteins never assigned to EVs so far, among which all the ones (4 in number) here predicted being membrane components ( Supplementary Table S2 ). An in-silico prediction of the schematic structure of these 12 proteins is shown in Figure 12 . The occurrence of a signal peptide was evidenced in 10 proteins, among which 4 also contained a transmembrane domain, thus suggesting an extracellular/membrane localization for these components; the remaining 2 proteins did not contain localization motifs, as typical of cytoplasmic/luminal proteins. These 12 proteins included: i) three large molecules, such as the cysteine-rich secretory protein (cgd7_4310-RA-p1) including repeated CAP domains related to secretory proteins of metazoans (Gibbs et al., 2008) and allergen V5/Tpx-1-related (cgd5_2020_RA-p1); ii) a protein (cgd6_1630_RA) containing a CS domain, which is a binding module for HSP90, possibly involved in recruiting HSPs to multiprotein assembly (Lee et al., 2004); iii) membrane-linked aspartyl-protease (cgd1_3690-RA-p1) that was detected only in the SEVs proteome. Additional proteins were uncharacterized proteins (cgd1_660-RA, cgd2_3780-RA, cgd3_1540-RA, cgd3_3370-RA, cgd6_1080-RA, cgd6_4460-RA, cgd6_710-RA, cgd7_400-RA) that do not show similarity with already known protein families. This group of Cryptosporidium-specific proteins also included the glycoprotein GP60 (cgd6_1080-RA-p1), which is a well-known immunodominant antigen of the parasite (Wanyiri et al., 2007), here detected in both types of EVs.