The T. vaginalis clones containing TVV1, TVV2, and TVV3 viruses (TV79-49c1⁺) and isogenic TVV-free clone TV79-49c1^– were used in this study (Narayanasamy et al., 2021). The TVV-free clone was derived from TV79-49c1⁺ by 2′-C-methylcytidine treatment as described (Narayanasamy et al., 2021). Trichomonads were cultivated in the tryptone-yeast extract-maltose (TYM) medium at pH 6.2, supplemented with 10% inactivated horse serum (Gibco Life Technologies, Penrose, New Zealand) at 37°C (Diamond, 1957). Human spontaneously transformed aneuploid immortal keratinocyte cells (HaCaT, a kind gift of V. Vonka, Institute of Hematology and Blood Transfusion, Prague; RRID:CVCL_0038; Vonka et al., 1995) were grown in RPMI 1640 complete medium supplemented with 10% fetal bovine serum, 2‰ NaHCO₃, penicillin (100 U/ml), and streptomycin (100 μl/ml). The absence of mycoplasma contamination was tested by PCR as described (Van Kuppeveld et al., 1994).

A partial gene coding for 78 amino acid residues of the soluble domain of transmembrane protein Tsp1 (TVAG_019180) was amplified from T. vaginalis genomic DNA by PCR (primers are given in Supplementary Table 1) and cloned into bacterial vector pET42b (Novagen, Sigma-Aldrich, St. Louis, Unites States). Recombinant Tsp1 peptide was expressed with His-tag in Escherichia coli strain DE3 using autoinduction medium (Studier, 2005) and affinity purified with His-tagged fusion proteins purification system (ThermoFisher Scientific, Waltham, Unites States). Purified protein was used for immunization of Wistar Han IGS rats (BIOCEV Core facility, Vestec, Czechia), and polyclonal anti Tsp1 antibody was registered in Antibody Registry (RRID:AB_2910119).

The T. vaginalis culture of 1 × 10⁹ cells in the logarithmic phase of growth was harvested by centrifugation for 15 min at 2000 × g and 25°C, and the cell pellet was washed twice with serum-free TYM medium. Washed trichomonads were resuspended in serum-free TYM medium to a concentration of 1 × 10⁶ cells/ml and incubated for 2.5 h at 37°C. Cell integrity was checked microscopically using the trypan blue exclusion test (Strober, 2015). Then, the cells were pelleted by centrifugation for 15 min at 2000 × g and 25°C. The supernatant with exosomes and other vesicles was filtered through a 0.22 μm pore size filter to remove any non-sedimented cells. The filtered supernatant was concentrated by tangential filtration using a Vivaflow 200 system with 100,000 MWCO PES membrane (Sartorius, Göttingen, Germany) to a final volume of approximately 30 ml. The concentrated filtrate was centrifuged for 1 h at 100,000 × g and 4°C, and proteins in the supernatant were precipitated with methanol and chloroform (4:1). The precipitate was analyzed by mass spectrometry to subtract contaminating proteins from EV proteome. The pellet (EV) containing sEVs was resuspended in Sucrose-Tris (ST) buffer (250 mM sucrose, 10 mM Tris, 0.5 mM KCl, pH 7.2) and fractionated on a linear gradient of 5–50% OptiPrep (5% OptiPrep was made using 250 mM sucrose, 1 mM EDTA, 10 mM Tris, pH 7.4, and 50% OptiPrep was prepared using 250 mM sucrose, 6 mM EDTA, 60 mM Tris, pH 7.4) according to OptiPrep™protocol (Axis-Shield PoC AS, Norway) on a Beckman SW 41 swing-out rotor at 100,000 × g and 4°C for 15 h. The gradient was fractionated using a Beckman Fraction recovery system (Beckman Coulter Inc., Brea, United States) connected to an FPLC pump, a UV-Vis flow spectrophotometer, and a fraction collector. The individual gradient fractions (0.5 ml) were diluted at 1:10 in ST buffer and centrifuged for 1 h at 100,000 × g and 4°C. The washed pellets were analyzed by immunoblotting using antibodies against TVV1 (RRID:AB_2910120) and TVV3 (RRID:AB_2910121) capsid proteins (a kind gift from Dr. J-H. Tai, Academia Sinica, Taiwan) and Tsp1 (dilution of antibodies 1:1000). The goat anti-rabbit and anti-rat IgG-peroxidase antibodies (1:2000; Sigma-Aldrich, St. Louis, United States) were used as secondary antibodies. Proteins were visualized using chemiluminescence (Immobilon Forte, Merck, Kenilworth, United States), images were obtained with the Amersham Imager 600 (GE Healthcare, Chicago, Unites States), and signals were quantified using ImageJ/Fiji software (Schindelin et al., 2012).

The hydrodynamic diameter of sEVs in selected fractions from the OptiPrep gradient was measured using a light scattering system with Malvern Zetasizer Ultra (Malverm Pananalytical Ltd., Malvern, United Kingdom). The measurement parameters were: size measurement mode – backscattering (173°), viscosity – water, temp –20°C. 50 μl of the sample was measured in a ZEN2112 quartz cuvette. Data analysis was performed by particles size distributed by number.

For protein protection assay, isolated sEVs were treated with trypsin (5 μg/ml) for 10 min at 37°C. In parallel, Triton X-100 was added to the final concentration of 0.1% to dissolve the sEV membrane before trypsin treatment. Treatment with trypsin was stopped by methanol/chloroform protein precipitation, and proteins were analyzed by immunoblotting. For RNA protection assay, samples of sEVs, untreated or treated with 0.1% Triton X-100, were incubated with RNAse A (10 mg/ml; Thermo Fisher, Waltham, MA, United States) for 10 min at 37°C. After the incubation, sEVs RNA was isolated using Trizol reagent (Merck, Darmstadt, Germany).

Label-free quantitative mass spectrometry (LFQ-MS) was performed as described previously (Zimmann et al., 2021). The samples were digested with trypsin and the peptides obtained were subjected to nano-liquid chromatography-mass spectrometry. The MS/MS spectra were searched against the T. vaginalis database (TrichDB, www.trichdb.org, release 2020-05-27, 60,330 entries), and 92 TVV protein sequences available in NCBI non-redundant protein database. The quantifications were performed with the label-free algorithms using MaxQuant LFQ version 1.6.2.0 (Max-Planck-Institute of Biochemistry, Planegg, Germany; Cox et al., 2014). The MS data were obtained from three independent sEV isolations and deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2019) with the dataset identifier PXD031790.

MaxQuant output file was processed using (Rstudio, Boston, MA, United States). To filter the initial dataset we used three criteria: we removed all single peptide identifications, all proteins identified in less than two independent experiments, and soluble proteins that were enriched in the supernatant after separation of sEVs with at least 1.5 fold change (p-value 0.02) in comparison to proteins in sEV fraction. To compare TVV plus and TVV minus sEV proteomes we conducted an imputation of all missing values (NA) with the normal distribution of the spectrometer limit and the final dataset was used to construct a volcano plot with a significance threshold p-value of 0.01 in RStudio.

Each protein identified in the proteome of exosomes was annotated based on the TrichDB database (www.trichdb.org, release 52), and searched against the HMMs profiles of UniRef30_2020_06_hhsuite database (from http://gwdu111.gwdg.de/c̃ompbiol/uniclust/2020_06/) by hhblits version 3.3.0 (Steinegger et al., 2019). Conserved domains were predicted using the Pfam 34.0 database.¹ The proteins were clustered based on assignment to Clusters of Orthologous Groups of proteins (COGs; Galperin et al., 2021) with the eggNOG-mapper v. 2.1.7 using the inbuilt eukaryotic database (Cantalapiedra et al., 2021). The presence of signal peptides was predicted using the TargetP version 2.0 (Emanuelsson et al., 2007). Transmembrane domains were predicted using the TMHMM server version 2.0 (Krogh et al., 2001).

Total RNA was isolated from TV79-49c1⁺ and TV79-49c1^– using the Highpure RNA isolation kit (Roche Diagnostics, Mannheim, Germany) following the manufacturer’s protocol. The RNA was diluted to 50 ng/μl and 100 ng was used in each reaction for cDNA synthesis. RT-qPCR was performed using the Kapa SYBR FAST one-step Universal kit (Sigma-Aldrich, MI, United States). DNA topoisomerase II (TVAG_038880) was used for normalization as described (Narayanasamy et al., 2021). The primers are listed in Supplementary Table 1. All the data were analyzed in GraphPad Prism 7 (La Jolla, CA, United States) and the experiments were performed in triplicates.

RNA was extracted from sEVs using Trizol reagent and purified on the Zymo IIC spin column (RNA clean & Concentrator-25) that yielded 25–40 ng/μl of sEV RNA. Six libraries were prepared from three independent experiments (A, B, C), each with sEV RNA isolated in parallel from TV79-49c1⁺ and TV79-49c1^– cells. RNA integrity was checked using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, United States), and concentration was determined with Qubit^® RNA Assay Kit in Qubit^® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, United States). Small RNA-Seq libraries were prepared from 200 ng of RNA using the NEXTFLEX^® Small RNA-Seq Kit v3 for Illumina^® Platforms (Perkin Elmer, Waltham, Unites States). Libraries were pooled in equimolar amounts, a 9 pM solution was loaded on the Illumina sequencer MiSeq (Illumina, San Diego, CA, United States) and sequenced uni-directionally, generating 50 bases long reads. Library preparation and sequencing were performed at the GeneCore facility of the European Molecular Biology Laboratory, Heidelberg, Germany.

The sEVs RNA (200 ng) was used for cDNA synthesis with the Verso cDNA synthesis kit (Thermo Fisher Scientific, Waltham, MA, United States) following the manufacturer’s instruction. The cDNA was used as a template for PCR, which was performed with Phusion Green Hotstart II High-Fidelity PCR Mastermix (Thermo Fisher Scientific, Waltham, MA, United States). The following cycling conditions were applied: Initial denaturation at 98°C for 2 min, 30 cycles of denaturation 98°C/10 s, 60°C/30 s, 72°C/45 s, and final extension at 72°C for 5 min. The PCR amplicons were separated on 1% agarose gel and the images were captured in InGenius3 (Syngene, Cambridge, UK).

Cutadapt v1.8.3 (https://cutadapt.readthedocs.io/en/v1.9.1/) was used to remove the adapter (TGGAATTCTCGGGTGCCAAGG), first 4 nt and last 4 nt, and to filter out sequences < 15 nt. Sequences were normalized per million reads for comparison between experiments. RNA fragments were mapped to reference TVV genomes sequences TVV1 NC_027701.1, TVV2 NC_003873.1, TVV3 NC_004034.1, and TVV4 NC_038700.1, and a reference genome of T. vaginalis at TrichoDB (Amos et al., 2022; TrichDB-52_TvaginalisG3.gff) using Bowtie2 version 2.4.4 with end-to-end alignment option (Langmead and Salzberg, 2012). The mapped reads were assigned to their respective genes using FeatureCounts. Differential expression analysis was conducted using DESeg2 package in Rstudio. To create the bar charts and pie charts of RNA and tRNA distribution, genes were assigned to different groups according to their Trich-DB annotation. All reads and filtered tRNA reads were grouped by their lengths to create the length distribution graphs. To calculate the coverage and distribution of different tRNAs, coverageBed version 2.30.0 (https://bedtools.readthedocs.io/en/latest/content/tools/coverage.html) was used.

Sequences were deposited in the European Nucleotide Archive, under accession number PRJEB50674.

The negative staining of sEV was performed using 5 μl of a sample that was incubated on glow discharged carbon-coated copper grid for 5 min. The grid was then washed in Milli-Q-H₂O and stained with 2% uranyl acetate. The samples were observed in a transmission electron microscope (TEM) JEOL JEM 2100Plus (Akishima, Japan) equipped with an XF416 CMOS camera (TVIPS GmbH, Gauting, Germany) at 200 kV. For the cryoEM analysis, the samples were prepared using Automatic Plunge Freezer EM GP2 (Leica Microsystems GmbH, Wetzlar, Germany). The sample (4 μl) was placed on the glow discharged Quantifoil^® carbon grid (R 1.2/1.3), blotted with filter paper for 4 s, and directly plunged in liquid ethane tempered to –143°C with liquid nitrogen. The JEOL JEM-2100Plus, at 200 kV was used for the analysis under cryo conditions with Gatan 626 cryo holder (Gatan, Inc., Pleasanton, CA, United States) and SerialEM software version 3.9 beta1 (https://bio3d.colorado.edu/SerialEM/).

Gene for BspA (TVAG_268070) was amplified by PCR from T. vaginalis genomic DNA and cloned into plasmid TagVag for episomal expression with C-terminal hemagglutinin tag in T. vaginalis. Trichomonads were transformed and selected as described (Hrdy et al., 2004). Immunofluorescent microscopy was performed as described (Stáfková et al., 2018). BspA was detected in trichomonads by the combination of mouse monoclonal antibody against 2x-HA tag (Exbio, Czechia), and secondary donkey anti-mouse antibody Alexa Fluor 488 (Thermo Fisher Scientific, Waltham, MA, United States). Structured illumination microscopy (SIM) was performed as described (Stáfková et al., 2018).

EVs were isolated from TV79-49c1⁺ as described above. The equivalent of EVs produced by 1.5 × 10⁸ Tv79-49c1⁺ cells (5 mg of protein) was added to the suspension of 1.5 × 10⁵ TV79-49c1^– cells per ml (total volume of 10 ml of TYM media). Trichomonads with microvesicles were incubated in TYM medium for 24 h to reach cell density 1.5 × 10⁶ per ml, then 1 × 10⁶ cells were transferred to 10 ml of the fresh medium, and 9 × 10⁶ cells were harvested and used for RNA isolation. The same procedure was performed for consecutive four subcultures.

To test the TVV transmission between trichomonads, we prepared two cell lines. The donor line was TV79-49c1⁺ that was transformed with pTagVagV5-Pur (Rada et al., 2019), enabling the expression of biotin ligase (BirA) as a marker under puromycin selection (TV79-49c1⁺ PAC-BirA). The recipient strain was TV79-49c1^– that was transfected with pTagVag2 (Rada et al., 2015) for expression of cytosolic HA-tagged phosphofructokinase under geneticin selection (TV79-49c1^– G418-PFK). Both trichomonad strains were mixed based on their doubling times (1.1 × 10⁵ of TV79-49c1⁺cells and 5.3 × 10⁴ of TV79-49c1^–cells per ml) to reach the density of 1.5 × 10⁶ cells per ml after 24 h of co-cultivation. After five transfers, geneticin was added to the culture (200 μg/ml) to eliminate the TVV donor strain, and the selection with geneticin was maintained for 40 transfers. Total RNA was isolated at selected time points from washed cells for RT PCR analysis as above.

HaCaT cells were grown to 80-90% confluency in RPMI 1640 medium in a total volume of 5 ml. Subsequently, the cells were incubated in fresh RPMI 1640 medium with the sEVs and at 37°C for 24 h in the 5% CO₂ atmosphere. The treatment dose of sEVs was calculated to be equivalent to a load of parasites during vaginal infection (4 × 10⁵ trichomonads/ml, 27 μg sEVs per 1 ml HaCaT cell culture). After incubation, the conditioned medium was centrifuged for 10 min at 1000 × g and 4°C, divided into aliquots, and stored at –80°C. The isolations of sEVs were performed in triplicate for each cell line (TV79-49c1^±) and each batch of harvested sEVs was co-incubated with technical triplicate of HaCaT cells. Unstimulated HaCaT cells conditioned medium was used as a negative control.

Levels of selected immune mediators (IL-6, IL-8, RANTES, CCL-2, and IL-1β) were analyzed by ELISA kits, performed in triplicates according to the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, Unites States).

To test the hypothesis that TVV is exported from T. vaginalis cells via EVs, we isolated EVs from the clone TV79-49c1⁺ that harbors TVV1, TVV2, and TVV3 (Narayanasamy et al., 2021). EVs were isolated based on the combination of filtration and centrifugation of conditioned medium using a linear gradient of 5–50% OptiPrep in the final step. For monitoring of sEVs in the gradient fractions, we developed a polyclonal antibody against tetraspanin Tsp1 (TVAG_019180) that has been identified in the proteome of T. vaginalis exosomes (Twu et al., 2013). The major signal on the western blot for Tsp1 was observed in fractions 17–19 of 24 collected fractions (Figures 1A,B). Dynamic light scattering analysis of isolated EVs in fractions 17–19 revealed peak sizes 108–146 nm (Figure 1D; hereafter sEVs), which is in the range of T. vaginalis exosomes (Twu et al., 2013). The presence of TVVs was monitored by antibodies against the capsid protein of TVV1 (Liu et al., 1998) and TVV3 (Bessarab et al., 2011; Narayanasamy et al., 2021). The maximal signals for both TVVs were in the fractions enriched in sEVs, however, there was another maximum in more dense fractions (10–12; Figures 1A,B). We assumed that in the sEV fractions, the viral particles are inside of the vesicles whereas the second maximum may represent naked particles outside of sEVs. To test the TVV topology, we treated pooled sEV fractions (17–19), and more dense fractions (10–12) with trypsin (Figure 1C). The trypsin treatment had a minor effect on the capsid protein signal in the exosome fractions; the capsid protein disappeared only upon the addition of the membrane-dissolving detergent Triton X-100. The capsid protein in the more dense sample (fractions 10–12) was not membrane-protected and was cleaved without adding the detergent. Next, we used sEV fractions for analysis by cryo-electron microscopy which is optimal to observe vesicles in shape close to their physiological stage. The sEVs appeared as spherical vesicles surrounded by a single membrane of 122 ± 20 nm (n = 20) in diameter (Figure 1Ea). In some vesicles we observed TVV particles of about 40 nm (Figure 1Eb), infrequently the viral particles were detected also outside of exosomes (Figure 1Ec). Similarly, we found individual TVV particles enclosed in vesicles of expected size using negative staining (Figure 1Fa–e), rarely do we find larger vesicles surrounding 2–3 viral particles (Figure 1Ff). Altogether, these results indicated that TVV particles resided inside of sEVs released by T. vaginalis TV79-49c1⁺.

Identification of TVV particles in sEVs prompted us to test whether the sEVs may serve for transmission of TVV to TVV-free T. vaginalis. Initially, we isolated EVs from the TV79-49c1⁺ clone and added access (equivalent that produces 1.5 × 10⁸ of Tv79-49c1⁺) to the culture of TV79-49c1^– (1.5 × 10⁶ cells in 10 ml). Like in Tv79-49c1⁺ cells, all three TVVs were detected in the isolated EVs by RT PCR (Supplementary Figure 1). However, a negligible amount of TVV1 and no TVV2 and TVV3 were detected in T. vaginalis cells after 24 h incubation with TVV positive EVs. The cells in the following four subcultures were negative for all three TVVs (Figure 2A and Supplementary Figure 1).

Next, we rationalized that isolation of EVs may affect their biological properties, and interfere with TVV transmission, thus we decided to directly co-cultivate TVV plus and TVV minus trichomonad clones. For this purpose, we derived the donor TV79-49c1⁺ clone, which was transformed with pTagVagV5-Pur (Rada et al., 2019), expressing a puromycin N-acetyl-transferase (PAC) for puromycin resistance and biotin ligase (BirA) as a marker (TV79-49c1⁺ PAC-BirA). The recipient strain TV79-49c1^– was transformed with pTagVag2 (Rada et al., 2015), providing resistance to geneticin (TV79-49c1^– G418-PFK). The clones were mixed, co-cultivated for five subcultures, and then geneticin was added to ablate the donor trichomonads. RT PCR signal for pac disappeared after two subcultures with geneticin, and neither pac nor birA was detectable after five subcultures indicating an absence of the donor trichomonads (Figures 2B,C and Supplementary Figure 2). TVV1 signal gradually declined, but it was still detectable after 30 subcultures and disappeared after 35 subcultures. Thus, the co-cultivation did not lead to the stable TVV infection of TVV-free clones under our experimental conditions.

To investigate whether TVV infection modifies exosomal cargo, we first compared proteomes of exosomes produced by isogenic T. vaginalis clones TV79-49c1⁺, and its virus-free derivative TV79-49c1^–. The size of sEVs produced by TV79-49c1^– was similar to those from TV79-49c1⁺ (113–150 nm, Supplementary Figure 3), and both isogenic clones produced a comparable amount of sEVs, estimated as the production of sEV proteins. The TV79-49c1⁺, and TV79-49c1^– clones released sEVs corresponding to 13.02 ± 1.29 μg, and 14 ± 0,69 μg (n = 3) of sEV proteins per 1 h by 1 × 10⁶ cells. In total, we identified 1633 proteins including conserved exosomal proteins such as Tsp1 and five other tetraspanins, Rab small GTPases (146 proteins), Hsp70 (20 proteins), Tsg101, and cytoskeletal proteins actin (7 paralogs), and tubulin (Supplementary Table 2). The proteins were classified based on clusters of orthologous groups (COGs) into 18 functional categories (Figure 3A). The largest functional category formed proteins that are involved in intracellular trafficking (COG category U, 17%), namely SNARE proteins (putative synaptobrevins, syntaxins), vacuolar protein sorting-associated proteins (Vps), and multiple Rab paralogs of twelve Rab families, as well as unclassified Rabs. The most prominent group of proteins in the COG category “posttranslational modification, protein turnover, chaperones” (COG O, 14%) were peptidases (65 in total). Of these, membrane-associated peptidases included serine peptidases (10 paralogs), two calcium-dependent cysteine proteases (calpains), and leishmanolysin-like metallopeptidases (GP63, 11 paralogs), while the most frequent soluble peptidases were cysteine peptidases (CPs, 18 proteins) of various families, including legumains, calpain-like CPs, ubiquitin-hydrolase-like CPs, and metacaspases. The third-largest COG category included proteins for signal transduction mechanism (COG T, 9%) with 35 protein kinases, 13 proteins with predicted phosphatase activity, and 15 calcium-binding proteins including 7 calmodulin paralogs (Supplementary Table 2). Our dataset was considerably larger in comparison to the previously reported exosomal proteome of 215 proteins, of which 47% were shared with our dataset (Figure 3B). Comparison with the proteome of microvesicles shed from T. vaginalis plasma membrane (ectosomes; Nievas et al., 2018) and the surface proteome (De Miguel et al., 2010) revealed 45 and 43% common proteins, respectively (Figure 3B).

Quantitative analysis of differentially expressed proteins in TV79-49c1⁺and TV79-49c1^– sEVs with a statistical cut-off p-value of 0.01 revealed 12 significantly enriched and 8 unique proteins in TVV plus sEVs (Figure 4A and Supplementary Table 2). The unique proteins that were absent in TV79-49c1^– sEVs included four TVV capsid-RdRp fusion proteins. Mapping of individual peptides against 92 available protein sequences of TVVs revealed that the majority mapped to capsid protein of TVV1 (21 peptides), 13 peptides to TVV2, 13 peptides to TVV3, and a single peptide corresponded to RdRp of TVV1 (Supplementary Table 3). The protein sequence alignment also revealed the presence of unique peptides that mapped to different variants of TVV species: four variants of TVV1, and two variants of each TVV2 and TVV3 species (Supplementary Figure 4). The other unique proteins in TV79-49c1⁺ sEVs included the repair of iron centers protein 2 (RIC2, TVAG_167830; Nobre et al., 2016), a membrane protein TVAG_109130 with N-terminal subtilisin-like peptidase (IPR036852), and galactose-like binding (IPR008979) domains, and two proteins of unknown function (Figure 4A). The enriched proteins included putative adhesin BspA (TVAG_268070) with the C-terminal transmembrane domain (TMD). This protein is of particular interest as exosomes have been shown to modulate T. vaginalis adherence (Twu et al., 2013). In addition to TVAG_268070, the proteome contains 18 other BspA-like paralogs, however, these were not significantly enriched in TVV plus sEVs (Supplementary Table 2). To verify cellular localization of BspA TVAG_268070, the recombinant protein with the C-terminal HA tag was expressed in TV79-49c1⁺. Structural illumination microscopy showed the presence of BspA in the membrane of large vesicles reminiscent of multivesicular bodies (Figure 4B; Twu et al., 2013). Western blot analysis showed a band corresponding to BspA of the expected size (87 kDa) in cell lysate as well as in shed sEVs (Figure 4C). The other enriched proteins included discoidin-like adhesin (TVAG_004030), putative calcium-dependent phospholipid-binding protein (TVAG_069450), Rabs (TVAG_320200, TVAG_044270), protein kinase domain-containing protein (TVAG_037610), vacuolar V-ATPase subunit B and C (TVAG_037610, TVAG_324980), a metallopeptidase M24 (TVAG_410220), and three unknown proteins. The V-ATPase is a multisubunit complex, thus we searched for other subunits insEV proteome. We identified ten additional proteins corresponding to ATPase subunits B, C, D, E, and H, of which nine were enriched in TV79-49c1⁺sEVs at p-value ≤ 0.05. The proteome of TV79-49c1^– sEVs revealed 12 unique and 42 enriched proteins. Of these, the largest functional groups were formed by proteins involved in intracellular trafficking and exocytosis (11 proteins, 20%), and hydrolases (8 proteins, 15%) including three GP63 peptidases and alpha-1,3-glucosidase. The group of proteins with unknown functions included 17 members (31%; Figure 4A and Supplementary Table 2).

To investigate whether the increased level of proteins in sEVs correlates with the level of the corresponding mRNA in T. vaginalis cells, we isolated total RNA from TV79-49c1⁺and TV79-49c1^–, and quantified RNA using qRT PCR for five proteins enriched or unique in TVV plus and TVV minus sEVs. This analysis revealed over 2-fold higher RNA level in TV79-49c1⁺ for RIC2 and BspA in comparison to TV79-49c1^–, which correlated with a higher level of corresponding proteins, however, no significant difference or decrease was found for TVAG_320200, TVAG_267930, and TVAG_004030 (Figure 4D). Similarly, a limited correlation between protein and mRNA levels was observed for five selected proteins that were unique in TV79-49c1^– (Figure 4E). Particularly, TVAG_064330 (GP63) and TVAG_098820 displayed significantly lower mRNA levels in TV79-49c1^– in comparison to TV79-49c1⁺ although these proteins were present only in TV79-49c1^– sEVs (Figure 4A). The positive correlation between mRNA and protein levels for RIC2 and BspA suggested that TVV infection may have an impact on gene transcription or mRNA stabilization. The higher mRNA level with the absence of corresponding proteins (GP63, and TVAG_098820) in TV79-49c1⁺ suggested that TVV infection may interfere with the protein translation or packaging of the exosomal cargo in T. vaginalis cells, however, the proposed explanation of observed phenomena needs experimental verification.

To investigate whether TV79-49c1⁺sEVs contain TVV-derived RNA and how the presence of TVV modulates sEV RNA cargo, aliquotes of exosomal samples from TV79-49c1⁺ and TV79-49c1^– isolated for proteomic analysis were in parallel used for RNA isolation (Figure 5A RNA analysis). The majority of isolated RNAs appeared to be of a small size ranging between 25 and 200 nt (Figure 5B). We did not observe a signal corresponding to the size of TVV genomes of 4.6–4.8 kbp, however, we presumed that the amount of complete genomic RNA was under the limit of electrophoretic detection or only fragments of genomes were present. Indeed, we amplified from isolated RNA complete ORFs for RdRp from TVV1, 2, and 3 (2037, 2130, and 2127 bp), and for capsid proteins (2205, 2106, and 2046 bp; Figure 5C) by reverse transcription PCR. To test whether sEVs-associated RNA is membrane-protected, we treated EVs with RNase A before RNA isolation, which had a negligible effect on RNA sizes. The addition of Triton X-100 together with RNase resulted in partial RNA digestion, particularly the disappearance of the dominant peak of approximately 100 nt (Figure 5B). As an unprotected control, we isolated cellular RNA from T. vaginalis with characteristic peaks of ribosomal RNAs on the electrophoretogram that disappeared upon treatment with RNase A without the detergent addition (Figure 5C). These results confirmed that exosomal RNAs were protected by a membrane envelope.

In total, six RNA-seq libraries were prepared from three independent experiments. Sequencing of each library generated on average 2.8 million reads per library, each read 50 nt long. The total number of filtered reads ranged from 1.8 to 3.9 million (Table 1). Mapping of reads to reference TVV1-4 genomes revealed 248-6029 reads in TV79-49c1⁺sEVs cargo matching TVV1, TVV2, and TVV3 genomes. The reads were distributed along the whole genome sequence covering 45.7–82.1% of the genome sequences (Supplementary Figure 5). A negligible number of reads (1–11) derived from TV79-49c1^– EVs library randomly mapped to TVVs.

Next, the reads were mapped to reference T. vaginalis G3 genome (Table 1 and Supplementary Table 4). The majority of mapped reads were assigned to rRNA with a higher contribution in TV79-49c1^–sEVs (on average 89%) than in TV79-49c1⁺ sEVs (70%; Table 1). The second group included tRNA with about a 2.5-fold higher number of total tRNA reads in TV79-49c1⁺ (on average 10%) than in TV79-49c1^– sEVs (27%; Table 1). Other gene coding sequences (CDS) represented on average 1.4% in TVV minus and 2.3% in TVV plus sEVs. Unmapped reads accounted for 41–48% in total (Table 1 and Supplementary Table 4). Mapping of tRNA reads revealed the presence of twenty tRNA types in both TVV plus and TVV minus sEVs with the majority of RNA reads mapping to tRNA-Gly, tRNA-Val, and tRNA-Glu (Figure 6A). The other tRNA types were represented in less than 2% each. Statistical analysis of tRNAs, rRNAs, and CDS RNAs with a p-value of 0.01 showed significant enrichment of tRNAs in TV79-49c1⁺ sEVs (Figure 6B). The most frequent type of enriched tRNA fragments corresponded to tRNA-Gly (51.8% of enriched tRNA fragments), tRNA-Val (20.1%), and tRNA-Glu (19.5%; Figure 6C). The other RNA fragments enriched in TV79-49c1⁺ sEVs corresponded to sixteen mRNAs with the most enriched fragments mapping to two genes for putative hydroxylamine reductases (HAR1-2). The unique fragments belonged to TVV RNAs in TV79-49c1⁺(Figure 6B and Supplementary Table 4).

Transfer RNA fragments (tsRNA) are of particular interest as they belong to the emerging category of non-coding regulatory RNAs that are conserved in all domains of life including T. vaginalis (Li et al., 2018). They are sorted according to the tRNA region, from which they are derived by cleavage within T-loop, D-loop, or anticodon-loop (Dou et al., 2019). Thus, we sorted tsRNAs based on their gene mapping, which revealed two major peaks of tsRNAs with sizes of ∼ 21 nt and ∼ 27–29 nt with the higher number of all fragments in TV79-49c1⁺ sEVs (Figure 7A). In TV79-49c1⁺ sEVs, majority of tsRNA (87.6%) were classified as 5′tsRNA, 8.6% mapped to the middle of tRNA including anticodons region (internal tsRNA), and 3.8% were 3′tsRNAs (Figure 7B). A similar distribution but a lower quantity of tRNA fragments were found in TV79-49c1^– sEVs. For detailed classification, we grouped tsRNA based on tRNA type and anticodons (Figures 7C–E, Supplementary Figures 6, 7 and Supplementary Table 5). The most dominant tsRNAs in TV79-49c1^– sEVs were 5′tsRNA-Gly with GCC anticodon (58%), 5′tsRNA-Glu with TCC anticodon(13%), and internal tsRNA-Val with GAC anticodon(7%; Supplementary Table 5). In TV79-49c1⁺, the pattern of tsRNA was similar with a slightly higher contribution of 5′tsRNA-Gly (GCC; 60%), and a lower number of 5′tsRNA-Glu (TTC; 9%), and internal tsRNA-Val (GAC; 5%; Supplementary Table 5). Collectively, a common pattern of tsRNA in TVV plus and minus exosomes indicated that tsRNA biogenesis operates similarly in both cell lines, however, the tsRNA content is significantly enhanced in exosomes produced by TV79-49c1⁺, which suggests more efficient packaging of tsRNA cargo into exosomes upon TVV infection.

Unlike for tsRNA, mapping of fragments derived from 16S rRNA and 28S rRNA revealed even distribution along with the complete genes without enrichment of specific fragments. This distribution is more consistent with rRNA degradation than with the packaging of functional fragments (Supplementary Figure 8A). Interestingly, fragments derived from 5.8 rRNA appear to be 28 and 42 nt long (Supplementary Figure 8B–D), which suggests their specific processing.

Trichomonas vaginalis sEVs were shown to modulate host-parasite interactions (Twu et al., 2013; Govender et al., 2020). Thus, we were interested in whether and how the sEVs from TV79-49c1⁺ modulate the immune response of human cells in comparison with sEVs derived from TVV minus TV79-49c1^–. To address this question, we co-incubated human keratinocytes (HaCaT cell line) with TVV plus and TVV minus sEVs for 24 h and quantified the production of five proinflammatory markers, including C-C motif chemokine ligand 2 (CCL2, syn. MCP-1), and RANTES (syn. CCL5) that are known to respond to dsRNA (Goodman et al., 2011; Sokol and Luster, 2015), proinflammatory cytokines IL-6 and IL-8 that were shown to be suppressed by sEVs from TVV positive T. vaginalis (Govender et al., 2020), and IL-1β, a proinflammatory factor during viral infection (Poeck et al., 2010). An advantage of HaCaT cells is that they are spontaneously immortal cells (Boukamp et al., 1988), without transformation with human papillomavirus oncogenes E6 and E7, which are often used for immortalization of various cell lines (Fichorova et al., 1997), and may interfere with signaling pathways involved in immunity and other cellular processes (Bienkowska-Haba et al., 2017). TV79-49c1⁺ EVs stimulated production of RANTES, while its production upon incubation with TV79-49c1^–sEVs was close to the baseline as for control cells. TV79-49c1⁺ EVs also induced a significant increase in the production of IL-8, IL-6, and IL-1β in comparison to TV79-49c1^– EVs (Figure 8). These results indicate that sEVs derived from TVV-infected T. vaginalis stimulate a proinflammatory response in human HaCaT cells