A Trichomonas vaginalis Rhomboid Protease and Its Substrate Modulate Parasite Attachment and Cytolysis of Host Cells

Trichomonas vaginalis is an extracellular eukaryotic parasite that causes the most common, non-viral sexually transmitted infection worldwide. Although disease burden is high, molecular mechanisms underlying T. vaginalis pathogenesis are poorly understood. Here, we identify a family of putative T. vaginalis rhomboid proteases and demonstrate catalytic activity for two, TvROM1 and TvROM3, using a heterologous cell cleavage assay. The two T. vaginalis intramembrane serine proteases display different subcellular localization and substrate specificities. TvROM1 is a cell surface membrane protein and cleaves atypical model rhomboid protease substrates, whereas TvROM3 appears to localize to the Golgi apparatus and recognizes a typical model substrate. To identify TvROM substrates, we interrogated the T. vaginalis surface proteome using both quantitative proteomic and bioinformatic approaches. Of the nine candidates identified, TVAG_166850 and TVAG_280090 were shown to be cleaved by TvROM1. Comparison of amino acid residues surrounding the predicted cleavage sites of TvROM1 substrates revealed a preference for small amino acids in the predicted transmembrane domain. Over-expression of TvROM1 increased attachment to and cytolysis of host ectocervical cells. Similarly, mutations that block the cleavage of a TvROM1 substrate lead to its accumulation on the cell surface and increased parasite adherence to host cells. Together, these data indicate a role for TvROM1 and its substrate(s) in modulating attachment to and lysis of host cells, which are key processes in T. vaginalis pathogenesis.


Introduction
Trichomonas vaginalis is an extracellular, eukaryotic parasite that is the causative agent of trichomoniasis, the most common non-viral sexually transmitted infection in the world [1]. Approximately 276 million people worldwide become newly infected each year [1]. In the United States, an estimated 3.7 million people are currently infected [2]. Symptoms and outcomes of infection include vaginitis, urethritis, prostatitis, infertility, and adverse pregnancy outcomes (reviewed in Petrin et al. 1998) [3,4]. T.vaginalis infection is associated with an increased risk of HIV acquisition [5] and potential transmission [6] due to HIV target cell recruitment to the site of infection [7] and increased viral shedding upon co-infection [8,9]. T. vaginalis is also associated with cervical cancer [10,11] and aggressive prostate cancer [12,13]. Due to the high burden, threat of illness, and understudied nature of T. vaginalis infection, trichomoniasis has been recently recognized as one of the United States' neglected parasitic infections [4,14,15].
Although the magnitude of parasite infection is high, little is known about how T. vaginalis colonizes the human host and causes disease [3,16]. As an extracellular organism that thrives in the changing and physiologically diverse environment of the urogenital tract of men and women, T. vaginalis likely utilizes multiple mechanisms to establish an infection and persist. The parasite attaches to multiple host cell types such as vaginal and prostate epithelial cells [17], red blood cells [18], and is capable of acquiring nutrients from them through host cell lysis. T. vaginalis can also bind together to form clusters [19]. Parasite cell aggregates are observed upon axenic growth and when placed on monolayers of host cells. However, only a few of the molecular players that mediate and regulate these parasite-parasite or host-parasite interactions have been identified [16,19,20].
Recent genomic, transcriptomic, and proteomic studies of T. vaginalis have aided the identification of protein families that may play important roles in T. vaginalis cell biology and pathogenesis [21][22][23][24]. In studies analyzing the surface proteome of T. vaginalis, we identified a rhomboid-like protein [23]. Rhomboid proteases are polytopic serine proteases that are localized to membranes where they encounter and cleave their substrates. Thus they belong to the intramembrane-cleaving protease (I-CLiPs) class of enzymes [25]; however, rhomboid proteases differ from most I-CLiPs in that their cleavage products are often released to the outside of the cell. Rhomboid proteases were first described in Drosophila melanogaster, where they shed growth factors from the membrane to initiate cell signaling in neighboring cells [26][27][28].
One of the greatest challenges in uncovering the biological functions of rhomboid proteases has been the identification of their substrates. For example, the substrate for the most studied bacterial rhomboid, Escherichia coli GlpG, is still unknown [45]. Rhomboid proteases from evolutionarily divergent organisms are capable of recognizing and cleaving a common set of substrates [26,31,34,39,44,46]. This property has allowed the identification of active rhomboid proteases, from diverse organisms, in the absence of a known substrate.
Rhomboid proteases and other I-CLiPs evolved separately from soluble proteases [25], hence the mechanism(s) they use for substrate recognition and cleavage is also predicted to be different, as recent studies have begun to reveal [47][48][49]. Selection of a specific rhomboid substrate is dictated by transmembrane (TM) segment dynamics of the substrate [49]. In particular, helix-destabilizing residues in the TM domain appear to be required for the substrate to exit the membrane and reside in the enzyme's inner active site [49]. There is also a preference for certain amino acids surrounding the cleavage site. Cleavage of the substrate occurs between small amino acids at positions designated P1 and P1' [46,50,51]. Additional preferences for particular amino acids at the P4 and P2' sites for the eubacterial AarA rhomboid led to the prediction of a universal recognition motif for substrate binding by rhomboid proteases [46]. However, further studies revealed that most eukaryotic rhomboid proteases do not conform to this prediction [49]. The discovery of additional substrates will aid in defining specific substrate features that promote rhomboid protease cleavage and how these may differ between different classes of rhomboid proteases.
In this study, we report the identification and characterization of two active T. vaginalis rhomboid proteases (TvROM1 & TvROM3). Exogenous expression of TvROM1 increases attachment to and cytolysis of host ectocervical cells, which are two phenotypes implicated in pathogenesis. Using quantitative proteomics and bioinformatics, we identified and verified two TvROM1 substrates, TVAG_166850 and TVAG_280090. Neither TvROM1 substrate is cleaved by TvROM3, hence the two enzymes have different natural substrates, consistent with their specificity for different model substrates and different subcellular localizations. Importantly, the T. vaginalis substrates identified for TvROM1 belong to a family of putative adhesins previously implicated in pathogenesis [20,23]. Together, our observations demonstrate a role for TvROM1 and its substrate(s) in modulating parasite attachment and cytolysis of host cells.

Identification of active rhomboid proteases in T. vaginalis
To assess the function of rhomboid proteases in the extracellular parasite T. vaginalis, we interrogated its genome (http://trichdb.org/trichdb) and found nine genes annotated as "rhomboidlike". Bioinformatic analysis of these genes revealed that only four encode both a serine (Ser) and a histidine (His) at the top of TMs 4 and 6 which are predicted to form the catalytic dyad for substrate cleavage (S1 Fig). The proteins encoded by these four genes also have the Gly-X-Ser-X motif, (X = amino acid) that surrounds and identifies the catalytic serine residue [52]. We named these putative active T. vaginalis rhomboid proteases TvROM1-4 (TrichDB accession numbers TVAG_112900, TVAG_359500, TVAG_476950, TVAG_161010). The other five gene products (TrichDB accession numbers TVAG_282180, TVAG_378960, TVAG_  233140, TVAG_183030, TVAG_037580) lack one or both catalytic residues and are likely proteolytically inactive. TVAG_233140, which lacks a catalytic His, was detected in the surface proteome of T. vaginalis [23]. We focused our attention on the rhomboid proteins predicted to be active proteases. S1 Fig summarizes additional features found in the four predicted active T. vaginalis rhomboid proteases. The alignment in S1 Fig also reveals differences in the N-and Ctermini amongst TvROMs 1-4, a feature that has been previously noted when comparing rhomboid protease sequences from multiple organisms [52].

Localization of TvROMs in T. vaginalis
The genes encoding the predicted active rhomboids were cloned under the control of the strong α-SCS promoter with N-terminal hemagglutinin (HA) tags and transfected into T. vaginalis. TvROMs were tagged at their N-terminus based on reports that HA-N-terminally-tagged rhomboid proteases localize properly and are active, while addition of a C-terminal tag can abolish proteolytic activity [34,39,44]. Furthermore, all TvROMs lack predicted N-terminal signal peptides as analyzed by several signal peptide and membrane topology prediction programs. As has been observed for other multi-spanning membrane proteins, TvROMs 1-4 are predicted to be properly targeted and inserted in the membrane via their first TM domain [53]. The subcellular localization of the proteins was determined in transfectants using indirect immunofluorescence assays (IFA) with an anti-HA antibody. Initial analyses indicated a cell surface localization of HA-TvROM1. To confirm this localization, transfectants were labeled using membrane-impermeable biotin as described previously [23] and a co-localization of the HA-TvROM1 signal with the biotinylated cell surface was observed (Fig 1A-1C). No background staining was observed in non-biotinylated parasites used as a negative control.
In contrast to the cell surface localization of HA-TvROM1, HA-TvROM2 and HA-Tv-ROM3 were located in a line structure next to the nucleus (Fig 1D, 1E, 1F, and 1G, respectively), which is likely the Golgi apparatus. Many rhomboid proteases have been localized in the Golgi apparatus, including the founding member of the rhomboid family, Drosophila Rhomboid-1 (DmRho1) [26] and the Toxoplasma gondii TgROM2 [54]. Prior to mitosis, the T. vaginalis Golgi apparatus elongates and then divides through medial fission of each cisternae in a process called Golgikinesis, forming two Golgi ribbons [55]. We observed a similar pattern of staining adjacent to the nucleus in dividing cells where HA-TvROM2 (Fig 1D and 1E) and HA-TvROM3 localize, providing additional support for a Golgi localization of these proteins. We were unable to detect expression of HA-TvROM4 in transfectants and therefore did not continue to analyze this protein.

TvROM1 and TvROM3 are active proteases
We tested the proteolytic activities of TvROM1-3 using an established heterologous cell cleavage assay [39]. Synthetic open reading frames of each TvROM were designed to match codon bias of human cells to facilitate expression in human cells. These proteases were HA-tagged and co-expressed in HEK293 cells with model rhomboid protease substrates tagged at the Nterminus with GFP or doubly tagged with an N-terminal GFP tag and a C-terminal FLAG tag. Cleavage was then assessed by detection of released GFP-tagged substrates of the predicted size. Rhomboid proteases known to cleave the model substrate being tested were also co-transfected with the substrate as a positive control. Cells transfected only with the model substrate or co-transfected with catalytically inactive rhomboid mutants served as negative controls.
We first tested whether TvROMs cleave the D. melanogaster rhomboid substrate Spitz [26], a canonical rhomboid substrate cleaved by rhomboid proteases from most organisms tested [34,39,44,51,56]. We discovered that TvROM3 was the only TvROM that could consistently cleave Spitz, albeit weakly. TvROM3 cleavage was more efficient with APP+Spi7, a chimeric substrate in which the first 7 residues of the amyloid precursor protein (APP) TM segment were replaced with those from Spitz (Fig 2A-lane 5). The positive control DmRho1 cleaved APP+Spi7 (Fig 2A-lane 2) releasing the cleaved fragment into the media. Cleavage by TvROM3 was specific as no cleavage product was detected with the TvROM3 His181Ala mutant (Fig 2A-lane 6).
Plasmodium falciparum rhomboid 4 (PfROM4) displays 'atypical' specificity, because it cannot process Spitz yet cleaves many Plasmodium adhesins [39]. Since TvROM1 was unable to cleave Spitz, we tested its ability to cleave Plasmodium adhesins erythrocyte-binding antigen 175 (EBA-175) protein ( Fig 2B) Fig -lane 4). The ability of TvROM1 to cleave Plasmodium adhesins but not Spitz protein categorizes it as having atypical substrate specificity [39]. Importantly, in addition to PfROM4 and Entamoeba histolytica rhomboid 1 (EhROM1), TvROM1 is the third parasite rhomboid protease that displays atypical substrate specificity [39,44] suggesting this may be a common feature of eukaryotic parasite rhomboid proteases. Since TvROM1 displayed similar   TvROM catalytic activity analyses. The activity of T. vaginalis rhomboid proteases were tested by co-transfecting the proteases with known model rhomboid substrates in a heterologous cell cleavage assay using HEK293 cells. Proteases were HA tagged and substrates contained an N-terminal GFP tag substrate specificity as PfROM4, we determined whether TvROM1 also cleaves EBA-175 at the same site as PfROM4. The FLAG-tagged C-terminal cleavage product produced by TvROM1 and PfROM4 co-transfectants was immunoprecipitated and the cleavage site was determined by MALDI-TOF analysis. We found that TvROM1 cleaves EBA-175 between Ala and Gly ( Fig  2C) as previously found for PfROM4 [42]. Therefore, TvROM1 also cleaves its substrate between two small amino acids.
TvROM1 could also cleave human EphrinB3 ( . Taken together, these results identify different substrate specificities for TvROM1 and TvROM3 (summarized in Fig 2F).
TvROM2 produced cleavage fragments of Spitz (S2D Fig-lane 4), however, the fragments were also produced by the TvROM2 His270Ala mutant (S2D Fig -lane 5). As endogenous metalloproteases have been shown to cleave and release rhomboid substrates in the heterologous cell cleavage assay [26,34,39], we added a wide-spectrum metalloprotease inhibitor Serine protease activity and TvROM1 contributes to T. vaginalis attachment to and cytolysis of host ectocervical cells To investigate the biological role of TvROMs and their possible contribution to parasite-host interactions, we tested the effect of the serine protease inhibitor 3,4-dichloroisocoumarin (3,4-DCI) on parasite attachment to and lysis of host cells. 3,4-DCI has been shown to inhibit the activity of rhomboid proteases from multiple organisms [33,34,40,58]. The effect of 3,4-DCI on parasite attachment to host cells was tested by exposing T. vaginalis to ectocervical cell monolayers for 30 min in the presence of increasing concentrations of 3,4-DCI. Unattached parasites were then washed away and the number of attached parasites was quantified to allow detection. (A, B, D and E) Whole cell lysates (WCL) and conditioned media (CM) collected from co-transfectants was analyzed by Western blot analyses [39]. Top panels: rhomboid protease detected in WCL using an anti-HA antibody; middle panels: full-length (FL, filled arrowheads) and cleaved substrates (open arrowheads) detected in WCL using an anti-GFP antibody; bottom panels: cleaved substrate fragments detected in CM using an anti-GFP antibody (open, red arrowheads). The substrates tested were (A) APP+7 residues of the Drosophila melanogaster Spitz protein encompassing the rhomboid protease cleavage site (APP+Spi7) (as previously described [17]). A dose-response effect was observed even with low concentrations of 10 and 15 μM 3,4-DCI, which decreased attachment to host cells 40% and 64%, respectively, compared to vehicle-treated parasites ( Fig 3A). To test the effect of 3,4-DCI on cytolysis of host cells, we exposed T. vaginalis to ectocervical cell monolayers for three hours in the presence of different 3,4-DCI concentrations and then assessed host cell lysis by measuring LDH released from damaged cells as described [17]. Similar to the observed reduction in attachment, cytolysis of host cells was reduced by 25%, 61% and 73% in the presence of 5, 10 and 15 μM 3,4-DCI, respectively ( Fig 3B). No effect on the viability of the parasites or host cells was observed at the concentrations used for both the attachment and cytolysis assays. These results suggest that serine proteases likely play a role in two critical processes that contribute to T. vaginalis destruction of host cells.
To test the role of individual TvROMs specifically, we compared the ability of parasites exogenously expressing HA-tagged TvROMs to attach to and lyse host cells. Parasites expressing additional exogenous HA-TvROM1 had a 1.6-fold increase in attachment to host cells compared to empty vector control transfectants (p-value<0.01) ( Fig 3C). Additionally, exogenous expression of HA-TvROM1 resulted in a 4.2-fold increase in host cell lysis (p-value<0.01) (Fig 3D). The involvement of TvROMs in these processes was limited to TvROM1, as exogenous expression of TvROM2 and TvROM3 did not lead to increased attachment to or lysis of host cells.

TvROM1 substrate identification using quantitative proteomics
Having demonstrated that TvROM1 is an active protease that modulates attachment to and lysis of host cells, we sought to identify its substrates using a quantitative proteomics approach. Given the cell surface localization of TvROM1, we identified proteins that were differentially released into the media by TvROM1 transfectants treated with the serine protease inhibitor 3,4-DCI compared to DMSO vehicle control (see flow chart in Fig 4A). Proteins released into the media were collected and differentially labeled using stable isotope dimethyl labeling [59] to allow quantitative comparisons of protein abundance, with the prediction that inhibition of TvROM1 activity would reduce the release of its plasma membrane substrates. S1 Table lists and Fig 4B shows the distribution of all the proteins identified in a combined analysis of two independent mass spectrometry experiments. Of the proteins identified, seventeen protein groups decreased by more than 50% with 3,4-DCI vs. vehicle control treatment, with a statistically significant decrease observed for seven protein groups (see S1 Table). Five proteins of the seven protein groups (TVAG_166850, TVAG_573910, TVAG_245580, TVAG_425470, TVAG_393390) have a predicted TM domain and type 1 topology (N-terminus outside cell-TM domain-C-terminus inside cell), which constitute the minimal features of rhomboid protease substrates. The abundance of the five candidate substrates in the T. vaginalis media was reduced 60-67% in the presence of 50 μM 3,4-DCI ( Fig 4C). Mass spectrometry data from these experiments have been deposited to Chorus and are available for download (https:// chorusproject.org, project ID 969).
The ability of TvROM1 to cleave the TM domains of the five candidate substrates was tested using the HEK293 heterologous cell cleavage assay. Initial attempts to express the full-length T. vaginalis substrates resulted in retention of these large proteins in the ER. Therefore, to assess cleavage at the cell surface, a chimeric protein of GFP-EBA-175 with its TM domain replaced with each of the five putative substrate's TM domain was constructed and co-expressed with either TvROM1 or a TvROM1 His316Ala mutant. We found that only the TM domain of TVAG_166850 could be specifically cleaved by TvROM1 (Fig 4D,  Although we observed release of TVAG_573910 into the media, cleavage was not dependent on co-transfection with TvROM1 (S3A Fig -lanes 1 and 2). We also tested whether any of the putative substrates could be cleaved by TvROM3, the other active rhomboid protease identified (Fig 2), but TvROM3 failed to cleave any of the five putative substrates (see S3A-S3D Fig). Inspection of the amino acids in the TM domain of the five putative substrates (Fig 4C) revealed that the four proteins that were not cleaved (TVAG_573910, TVAG_245580, TVAG_425470, and TVAG_393390) contain residues at the top of the TM domain that fit the previously published Strisovsky et al. motif [46]. Therefore, the substrate specificity of both TvROM1 and TvROM3 appear to differ from the bacterial-like rhomboid specificity defined by the Strisovsky et al. motif. Instead, the TM domain of the cleaved substrate TVAG_166850, has a predominant presence of the small amino acid alanine, and the helix relaxing residue glycine [60,61] (Fig 4C). This finding led us to investigate possible substrate determinants that might aid in identification of additional T. vaginalis rhomboid substrates.
Screening the T. vaginalis surface proteome for TM domains similar to TMs in parasite rhomboid protease substrates identifies an additional TvROM1 substrate and reveals features of T. vaginalis substrate specificity Rhomboid proteases cleave near or at the external face of the TM domain of the substrate. Since the substrate's overall TM dynamics has been found to be the governing feature defining a rhomboid substrate [49], a subset of amino acids surrounding the cleavage site are likely to play a major role in promoting a substrate conformation that leads to cleavage [51]. Thus, we compiled and compared established and predicted cleavage sites for 20 rhomboid substrates from other eukaryotic parasites: Toxoplasma gondii, Plasmodium falciparum, and Entamoeba histolytica [34-37, 39, 41, 42, 44, 51, 54]. The resulting comparison revealed a preference for  specific residues (Fig 5A) from which we created a consensus called the parasite search motif (Fig 5A). It is notable that this motif has little overlap with the consensus motif of the P4-P2' sites preferred by bacterial rhomboid proteases [46].
Using the parasite search motif, we screened the T. vaginalis surface proteome [23] to identify putative TvROM1 substrates (see flowchart in Fig 5B). Of 314 proteins searched, seven proteins had the motif positioned in the top part of the TM domain by at least two of three TM prediction programs (S2 Table) and five of these were type 1, single TM proteins, typical of rhomboid protease substrates. The predicted TM domains of the five putative substrates are shown in Fig 5C. TVAG_166850, the only TvROM1 substrate identified in the quantitative proteomics approach (Fig 4D-lane 2) is also one of the five candidate substrates identified by this bioinformatics approach.
After confirming that the five candidate substrates are localized to the plasma membrane, we performed heterologous cell cleavage assays to test whether their TM domains are cleaved by TvROM1. In addition to the TM domain of TVAG_166850 (Fig 4D), we found that the TM domain of TVAG_280090 is cleaved by TvROM1 (Fig 5D-lane 2, bottom panel). It is notable that the TVAG_280090 protein has overall 55% identity and 69% similarity to TVAG_166850. The TvROM1 His316Ala mutant could not cleave TVAG_280090's TM domain (Fig 5D-lane  3, bottom panel), demonstrating specific cleavage. DmRho1 can also process GFP-TVAG_280090 (S4F Fig -lane 4), providing further support that this protein is a rhomboid protease substrate. Similar to that observed for TVAG_166850 (S4A Fig -lane 6) TvROM3 could not cleave the TM of TVAG_280090 (S4B Fig -lane 6). None of the remaining candidate substrates were cleaved by TvROM1 or TvROM3 (S4A-S4E Fig).
Mutation of predicted rhomboid cleavage site in the putative substrate TVAG_166850 increases its presence at the cell surface and leads to greater parasite attachment to host cells We have previously reported that the TvROM1 substrate identified in this study, TVAG_ 166850, increases parasite attachment to host cells when exogenously expressed in a poorly adherent T. vaginalis strain [23]. To address whether cleavage by TvROM1 influences TVAG_166850's ability to mediate host cell attachment we introduced mutations at the predicted rhomboid cleavage site. The predicted P1-P1' site Ala672-Gly673 (AG) residues were mutated to Phe-Phe (FF) residues, since we previously found that mutation of small amino acids at the predicted cleavage site to bulky Phe significantly decrease rhomboid cleavage of the T. gondii AMA1 substrate [38]. Fig 6A shows a graphical representation of GFP-TVAG_ 166850 and the predicted rhomboid cleavage site. To verify that the mutations introduced into the GFP-TVAG_166850 AG/FF mutant affect TvROM1 cleavage, the HEK293 heterologous cell cleavage assay was used to compare TvROM1 cleavage of the chimeric EBA-175 protein containing either the wild type or mutant TM domain of TVAG_166850. The GFP-EBA-175-TVAG_166850TM AG/FF mutant chimeric protein was cleaved 90% less than the chimeric protein containing the wild type TM domain (Fig 6B-lane 4 vs. lane 9). Cleavage of the wild type and mutant TM domains, above background levels, was not observed using the His316Ala mutant of TvROM1 (Fig 6B-lane 5 and lane 10). Similarly, the Toxoplasma gondii TgROM5 that can also cleave the wild type TM domain of TVAG_166850 (Fig 6B-lane 3) and exhibited limited cleavage of the mutant TM domain (Fig 6B-lane 8).
To test whether surface substrates of TvROM1 play a role in host: pathogen interactions, Nterminally tagged GFP-TVAG_166850 WT or the GFP-TVAG_166850 AG/FF mutant were exogenously expressed in the poorly adherent T. vaginalis strain G3. The levels of wild type and mutant proteins expressed and presented on the surface of transfectants were then measured   Predicted topology of the putative substrate TVAG_166850 using the Spoctopus TM prediction program, illustrated using the TOPO2 graphical representation program. TVAG_166850 WT was tagged at the N-terminus with a GFP tag (green box). The predicted rhomboid cleavage site (scissors) and the predicted P1-P1' cleavage site residues are highlighted in blue. The surrounding parasite search motif residues are highlighted in red. The predicted TM residues are shown below. A rhomboid cleavage site mutant was generated by mutating the predicted Ala-Gly P1-P1' residues (underlined in sequence) to Phe-Phe residues to using flow cytometry. The percent of cells expressing the GFP-tagged proteins were similar in the GFP-TVAG_166850 WT and GFP-TVAG_166850 AG/FF mutant populations (Fig 6C, top  panel). However, the mean fluorescence intensity (MFI) of the GFP-TVAG_166850 AG/FF mutant population was consistently at least three-fold higher than the wild type population in three independent experiments (a representative experiment is shown in Fig 6C, bottom  panel). The increased MFI signal observed in the GFP-TVAG_166850 AG/FF population continued to be observed after growing the parasites in culture for several days and after freeze-thawing of the population, providing evidence for a robust and stable phenotype. Higher surface staining of the rhomboid substrate MIC2 in T. gondii rhomboid knockout cells has also been reported by Shen et al. [62]. Our data indicate that increased surface detection of the GFP-TVAG_166850 AG/FF mutant protein results from lack of cleavage by TvROM1, allowing for its greater accumulation at the cell surface.
To determine the phenotypic effect of mutating the predicted rhomboid cleavage site in GFP-TVAG_166850, we compared the attachment properties of the GFP-TVAG_166850 WT and GFP-TVAG_166850 AG/FF mutant populations. Exogenous expression of GFP-TVAG_ 166850 WT led to a 1.3-fold increased attachment to host cells compared to empty vector transfectants (p-value<0.05), while exogenous expression of the GFP-TVAG_166850 AG/FF mutant led to an even greater increase in attachment (1.7-fold) compared to empty vector (p-value<0.01) (Fig 6D). The change in attachment between the mutant and wild type was statistically significant (p-value<0.01). These data indicate that reduced TvROM1 cleavage leads to the increased surface levels of the GFP-TVAG_166850 AG/FF mutant protein (Fig 6C) which in turn leads to increased attachment to host cells (Fig 6D). Incubation of GFP-TVAG_166850 WT and GFP-TVAG_166850 AG/FF mutant populations with ectocervical cells in a four hour cytolysis assay did not reveal a greater ability of the transfectants to lyse host cells (S5 Fig), indicating additional factors likely modulate T. vaginalis cytolysis of host cells.

Discussion
We have characterized three putative rhomboid proteases in the human parasite T. vaginalis and demonstrated that two, TvROM1 and TvROM3, are catalytically active and exhibit distinct substrate specificity. TvROM3 cleaves the canonical rhomboid substrate Spitz in a heterologous cell cleavage assay while TvROM1 does not cleave Spitz but does cleave several known Plasmodium rhomboid substrates [39,44]. Immunofluorescence studies of exogenously expressed HA-tagged TvROM1 and TvROM3 indicate that TvROM1 is a plasma membrane protease whereas TvROM3 localizes to Golgi-like structures. Cellular localization studies of the endogenous proteins are yet to be done to corroborate the localization determined for the HAtagged proteins.
As a step towards defining the function of TvROMs, we identified nine putative T. vaginalis surface substrates using both quantitative proteomic and bioinformatic approaches. Two of these candidate substrates were shown to be TvROM1 substrates: TVAG_166850 and TVAG_280090. These proteins have a high degree of similarity and are members of a gene family consisting of over 150 members [20]. Interestingly, two other predicted substrates (TVAG_244130 and TVAG_335250) are paralogs of TVAG_166850 and TVAG_280090; however their TM domains were not cleaved by TvROM1 in the heterologous cell cleavage assay. Whether these putative substrates are cleaved by other proteases not examined in this study remains to be determined. TvROM3 was not able to cleave any of the nine putative T. vaginalis substrates identified, confirming that TvROM1 and TvROM3, which display different subcellular localizations, also appear to have different substrate specificities.
We were not able to identify T. vaginalis substrates for either TvROM2 or TvROM3. However, given that the candidate substrates identified are members of large protein families, untested members of these families or other unidentified proteins may serve as substrates for these TvROMs. This is consistent with our observation that T. vaginalis surface proteomes contain a subset of proteins encoded by large multi-gene families, with different subsets expressed by different strains [23].
Amino acids surrounding the predicted cleavage sites in model rhomboid substrates cleaved by TvROM1 and the T. vaginalis substrates are similar. There is an overall presence of small amino acid residues surrounding the predicted cleavage site and indeed restriction of a small amino acid at the P4 and P2' sites in our parasite search motif was critical for bioinformatic identification of the two T. vaginalis substrates cleaved by TvROM1. Nevertheless, finding that TvROM1 cleaves only two of the five candidate substrates identified by bioinformatics underscores the difficulty in identifying substrates using search motifs. Overall, these analyses indicate that TM dynamics may play a more important role in substrate selection than the presence of specific amino acids surrounding the cleavage site.
We found that overexpression of TvROM1 significantly increases parasite attachment to and lysis of host cells, which are two properties critical for the pathogenesis of T. vaginalis. These data may be explained several ways. Signaling induced by cleavage of a TvROM substrate during the initial phase of parasite-host interaction may lead to an increase in adherent surface proteins on the parasite or host cell. This might be mediated by an initial low-affinity binding, with cleavage driving a higher affinity binding, which in turn triggers host cell lysis. A role for TvROM1 at the junction of adherence and cytolysis is consistent with the observation that once a threshold of parasite attachment is reached, host cell cytolysis is triggered [17]. Exogenous expression of the TVAG_166850 substrate with the predicted P1-P1' rhomboid cleavage site mutated to reduce TvROM1 cleavage leads to an even greater increase in parasite attachment relative to exogenous expression of the wild type protein, consistent with a direct role for this substrate in adherence and regulation of its function by TvROM1 cleavage. In this regard, it is notable that using the Phyre2 program, both TVAG_166850 and TVAG_280090 have cadherin-like predicted secondary structures which are known to mediate cell: cell interactions in eukaryotes [63].
The proteins identified in supernatants of T. vaginalis (S1 Table) together with proteins we previously identified to be released in exosomes [64], compose, to our knowledge, the most comprehensive analyses lists of proteins released/secreted by T. vaginalis to date. There is overlap between these two proteomes, with 70 of the 188 proteins identified here also being present in the T. vaginalis exosome proteome (S1 Table). We identified cysteine proteases, which are considered to be established and secreted virulence factors of T. vaginalis (reviewed in [65]).
This list also provides experimental evidence for the release of proteins previously predicted to be T. vaginalis secreted proteins, such as cysteine protease inhibitors called cystatins [66] and a pore-forming/saposin-like protein hypothesized to help lyse host cells [20,21,67]. Other soluble factors uncovered in this proteomic analysis may be candidates contributing to T. vaginalis virulence.
In summary, we have identified two active T. vaginalis rhomboid proteases (TvROM1 and TvROM3) and two TvROM1 substrates. We have shown that modulation of expression of TvROM1 and its substrate plays a role in regulating host cell attachment and cytolysis dynamics. These findings are significant as few proteins have been shown to contribute to both T. vaginalis attachment and cytolysis of host cells (reviewed in [16]). The results of this study also expand the finding that rhomboid proteases contribute to pathogenesis-associated processes in parasites that pose important public health problems (reviewed in [68] and [69]). As both human and TvROM1 substrate specificities are better defined, it may be possible to selectively target T. vaginalis rhomboid proteases as therapeutic agents [70].

TvROMs plasmid construction and exogenous expression in T. vaginalis
TvROMs were PCR amplified from G3 genomic DNA, using the primer pairs listed in S6 Fig. PCR fragments were cloned into the Nt-HA-MasterNeo plasmid [23]. To increase the detection of tagged TvROMs, two additional HA tags were inserted to generate 3xHA-TvROMs by ligation of hybridized oligos encoding two HA-tags into the NdeI restrictions sites. The 3xHA-TvROM-MasterNeo plasmid was then digested with ClaI to remove the fragment encoding the neomycin phosphotransferase (Neo) selectable marker and its flanking 5 0 and 3 0 beta-tubulin untranslated regions (UTRs) and ligated to close the vector. A fragment encoding the puromycin N-acetyltransferase gene flanked with alpha-SCS 5 0 and 3 0 UTRs previously described in [73] was cloned into the Master (-Neo) plasmid using the ApaI restriction site.

TvROMs indirect immunofluorescence assays
The subcellular localization of 3xHA-TvROM1 in RU393 transfectants was determined by cell surface biotinylation of transfectants followed by IFA and imaging as described [23] except for use of fewer parasites, 3.18×10 7 parasites in 30 ml. For localization of 3xHA-TvROM2 and 3xHA-TvROM3 in G3 transfectants, IFA was performed as described [75].

HEK293 heterologous cell cleavage assays and PfEBA-175 cleavage site determination
The in vitro activities of TvROMs were analyzed with a heterologous cleavage assay derived from [39]. Briefly, HEK293T (ATCC CRL-11268) cells were seeded in 12-well plates and cotransfected at~80% confluency with X-tremeGENE HP (Roche) and pcDNA3.1 (Invitrogen) expressing N-terminal GFP-tagged and/or C-terminal Flag-tagged substrates and N-terminal 3xHA-tagged rhomboid proteases. Cells were washed once and then conditioned with serumfree DMEM 18-h post-transfection. Media and cell lysates were harvested 24-h later in Laemmli buffer. Protein from lysates and media were separated on 4%-12% gradient Bis-Tris and 4%-20% Tris-glycine gels, respectively, (Life Technology) and transferred onto nitrocellulose membranes. Membranes were probed with anti-GFP (Abcam), anti-HA (Roche), or anti-FLAG (Sigma) primary antibodies, anti-rabbit/rat/mouse secondary antibodies conjugated to infrared fluorophores (Li-COR Biosciences) and imaged on an Odyssey infrared scanner (Li-COR Biosciences). For cleavage site determination, HEK293T cells were transfected as described, lysed in RIPA buffer, subjected to anti-FLAG immunopurification (Sigma) and analyzed by MALDI-TOF mass spectrometry as previously described [76]. TvROMs were amplified from G3 genomic DNA via PCR and cloned into the 3xHA pcDNA expression vector. Like other parasitic rhomboid enzymes [39], TvROMs did not express in HEK293T cells and were thus recoded for human expression (GeneArt).

TvROMs cytolysis and attachment assays
T. vaginalis adherence and cytolysis assays were performed as described [17]. To test the effect of 3,4-dichloroisocoumarin (3,4-DCI, SIGMA) on parasite adherence, parasites were incubated with 3,4-DCI or DMSO vehicle control in completed K-SFM media and added to ectocervical cell monolayers. Adherence and cytolysis assays were performed as described [17], using three or four hour incubation times for wildtype RU393 and RU393 3x-HA-TvROM1 transfectants, respectively.
Stable isotope dimethyl labeling of cell supernatants and quantitative proteomics 1×10 7 log-phase 3xHA-TvROM1 RU393 T. vaginalis transfectants in 100 ml of PBS+5% sucrose were treated with 50 μM 3,4-DCI or DMSO vehicle control for one hour at 37°C. Parasites were pelleted and the supernatant was filtered through a 0.22 μm Steriflip™ 50 ml filter (Millipore) and stored at -80°C. Upon thawing, supernatants were concentrated to~300 μl using an Amicon1 Ultra Centrifugal filter unit (3 kDa cutoff) and lyophilized. 25 μg of protein from each sample was reduced, alkylated, and digested with Lys-C endopeptidase (Wako, Richmond, VA) for four hours and trypsin for 14 hours (Thermo Scientific) as previously described [77]. Digested peptides were desalted by HPLC using a Michrom Bioresources Microtrap column (2-20 μg binding capacity). Stable isotope dimethyl labeling was performed as described [59], with the addition of an HPLC desalting step after peptide labeling followed by lyophilization. Peptides from the vehicle control were labeled with the "light label" and 3,4-DCI samples were labeled with the "intermediate label." Labeled peptides were reconstituted in 0.2% formic acid, and equal protein amounts were mixed. For mass spec analysis, 8 μl of a mixture containing 1 μg of each sample was separated on an analytical (75 μm ID) HPLC column packed inhouse with ReproSil-Pur C 18 AQ 3 μm resin (120 Å pore size, Ammerbuch, Germany) coupled to an LTQ-Orbitrap Classic mass spectrometer (Thermo Scientific, Bremen, Germany) as described [78] with the following modifications. The nanoflow LC system EASY nLC II gradient was as follows: 2-30% B (80 min), 30-100% B (1 min) and 100% B (8 min).

MS data analysis
MaxQuant (v. 1.4.1.2) was used to search Thermo RAW files [79]. Spectra were searched against T. vaginalis (50623 entries) and a contaminant database (245 entries). MaxQuant generated decoy sequences (reversed peptide sequences) to estimate the false discovery rate. Search parameters included variable oxidation of methionine, variable protein N-terminal acetylation, fixed carboxyamidomethylation of cysteine and dimethyl labeling of peptide N-terminus and lysine. Trypsin was specified as the digestion enzyme with up to two missed cleavages. A 1% false discovery rate threshold was applied for protein and peptide identifications. Precursor mass tolerance was 7 ppm (or less for individual peptides). Fragment mass tolerance was 0.5 Da. Confidence intervals and p-values were calculated using a hierarchical model with bootstrap resampling and pooled variance estimates as described [80]. P-values were adjusted using the Benjamini and Hochberg method to correct for multiple hypothesis testing [81].

Heterologous cell cleavage assays
The transmembrane domains from candidate substrates were inserted into EBA-175 by inverse PCR (Phusion, NEB) with primers encoding the new transmembrane domains at their 5 0 ends. PCR products were purified (QiaQuick, Qiagen) and blunt ends were phosphorylated by T4 polynucleotide kinase (NEB), and ligated with T4 ligase (Roche). The chimeric molecules were confirmed by DNA sequencing and analyzed for cleavage in the heterologous cleavage assay as described above.
Site-directed mutagenesis, expression, and phenotypic analysis of GFP-TVAG_166850 WT and rhomboid cleavage-site mutant GFP-TVAG_166850 AG/FF The plasmid encoding for TVAG_166850 with an N-terminal enhanced green fluorescent protein (eGFP) tag was constructed by cloning a eGFP PCR-amplified fragment into the NdeI and SacII restriction sites of Nt-HA-MasterNeo, followed by ligation of the TVAG_166850 gene inframe using the SacII and BamHI restriction sites. The TVAG_166850 AG/FF putative rhomboid cleavage site mutant was generated by site-directed mutagenesis of the GFP-TVAG_166850 construct using the QuikChange kit (Stratagene) and primers that encoded the Ala672Phe and Gly673Phe mutations (primers listed in S6 Fig). Plasmids were transfected into the poorly adherent G3 strain for gain of function analysis, and selected with G418. Attachment to and cytolysis of host cells was assayed as described above except the cytolysis assay was incubated for 18 hrs.
Flow cytometry analysis of surface GFP-TVAG_166850 WT and GFP-TVAG_166850 AG/FF mutant proteins 2×10 5 log-phase G3 transfectants were resuspended in FACS buffer (5% fetal bovine serum/ 0.1% sodium azide/1× PBS) and assayed in triplicate. Parasites were stained with 1 μg/ml mouse-anti-GFP (Clontech) and Empty vector transfectants were stained with only the secondary (2°) Ab as a control for background. Staining was detected with 1 μg/ml goat antimouse-Alexa Fluor 1 488 2°Ab antibody. Antibody incubations were performed on ice for 30 min. Samples were analyzed on a Becton Dickinson Fortessa flow cytometer in the FITC channel. Empty vector 2°Ab alone samples were used to set the GFP + population gates. The mean fluorescence values were calculated for each sample, using the FlowJo 8.4 software. The experiment was performed three times and representative data from one experiment is shown. ("~" marks sites of manual insertion/deletion to aid alignment of critical residues, numbers within the sequences indicate the number of residues omitted to allow optimal alignment). Shading indicates 75% or greater sequence similarity. Lines above the TvROM1 sequence marks predicted TM domains for TvROM1 using the Spoctopus signal peptide and membrane protein topology prediction program. Similar TM domain predictions were obtained for TvROM2 and TvROM4. TvROM3 lacks a predicted 7 th TM domain and thus the C-terminal tail is predicted to be intracellular. No signal peptides were predicted for TvROMs 1-4, and all the proteins are predicted to have their N-terminus located inside the cell. Bold red letters indicate residues participating in nucleophilic catalysis, while the green residues function in electrophilic catalysis (oxyanion stabilization). Blue lettering highlights residues that form four conserved Keystones that play key roles in maintaining rhomboid architecture (as defined experimentally for bacterial rhomboid proteases in R.P. Baker & S. Urban, Nature Chemical Biology, 2012). Briefly, the arginine (R) of Keystone I (E/QxWRxxS/TxxxxH) helps to stabilize the L1 loop 'hairpin' by donating several hydrogen bonds to neighboring residues, while GxxxExxxG of Keystone II stabilizes cytoplasmic interaction of TMs 1, 2 and 3 into an apex. Keystone III (surrounding the catalytic serine) and Keystone IV (with a GxxxG dimerization motif following the catalytic histidine base) mediate close apposition of TMs 4 and 6 at the core of the enzyme. The ability of T. vaginalis rhomboid proteases to cleave known model rhomboid substrates was tested using the HEK293 heterologous cell cleavage assay. Proteases were HA tagged and substrates contained an N-terminal GFP tag to allow detection. Whole cell lysates (WCL) and conditioned media (CM) were collected from co-transfectants and analyzed by Western blot analyses (Baker et al. 2006). Top panels: rhomboid protease detected in WCL using an anti-HA antibody; middle panels: full-length (FL-filled arrowheads) and cleaved substrates (open arrowheads) detected in WCL using an anti-GFP antibody; bottom panels: cleaved substrate fragments detected in CM using an anti-GFP antibody (red, open arrowheads). The substrates tested were (A) Plasmodium falciparum BAEBL, (B) human Ephrin-B3, (C) Plasmodium falciparum EBA-175, (D) Drosophila melanogaster Spitz, and (E) Providencia stuartii TatA. The positive control protease for testing cleavage of Spitz and TatA (D and E) was DmRho1; the positive control protease for Plasmodium BAEBL (A) and EBA-175 (C) was PfROM4; the positive control protease for EphrinB3 was TvROM1 (B). Negative controls lacked co-transfection with a TvROM (Negative). TvROM1/TvROM2/TvROM3 = wild type protease; TvROM1 mut/ TvROM2 mut/TvROM3 mut = protease with the catalytic histidine mutated to alanine. In (B) the heterologous cell cleavage assay was performed in the absence and presence of 10 μM Batimastat (-or + BB-94), a metalloprotease inhibitor. TvROM1 can cleave Plasmodium BAEBL (A-lane 3) and the TvROM1mut cannot (A-lane 4) indicative of TvROM-1 specific cleavage. EphrinB3 and TatA are released into the media by both TvROM2 and TvROM2mut co-transfectants (B-lanes 3 and 4 and E-lanes 4 and 5), therefore it does not appear to be TvROM2-specific cleavage. To further investigate this, TvROM2 cleavage of EphrinB3 was tested in the presence of a metalloprotease inhibitor and release of EphrinB3 above background was no longer observed (B-lanes 7 and 8). References: Baker, R. P., R. Wijetilaka and S. Urban (2006). PLoS Pathog 2(10): e113. (TIF) S3 Fig. TvROM1 and TvROM3 are not able to cleave remaining putative substrates from dimethyl labeling mass spectrometry experiments. The HEK293 heterologous cell cleavage assay was used to test cleavage of the TM domain of the putative substrates identified in the quantitative proteomics experiment (Fig 4). A plasmid encoding a chimeric protein composed of GFP-P. falciparum EBA-175 with the TM domain replaced with that of the putative substrates, was co-transfected with a plasmid encoding for wild type HA-TvROM1 or HA-Tv-ROM3, or catalytic His to Ala mutants (mut). The chimeric protein tested is indicated above each blot. Negative controls lacked co-transfection with a TvROM (Negative). Western blot analysis of whole cell lysates (WCL) and conditioned media (CM) from co-transfectants was performed with an anti-GFP antibody to test for the presence of a smaller GFP-EBA-175 fragment  5). A plasmid encoding a chimeric protein composed of GFP-P. falciparum EBA-175 with the TM domain replaced with that of the putative substrates (A-E), was co-transfected with a plasmid encoding for wildtype HA-TvROMs 1-3 or catalytic His to Ala mutants (mut). The chimeric protein tested is indicated above each blot. Negative controls lacked co-transfection with a TvROM (Negative). Western blot analysis of whole cell lysates and conditioned media from co-transfectants was performed with an anti-GFP antibody to test for the presence of a smaller GFP-EBA-175 fragment released into the media by TvROM1 cleavage or detected in cell lysates if cleaved by TvROM2 or TvROM3. An anti-HA antibody was used to confirm expression of the HA-TvROMs wt and mut proteins (top panel). Full-length chimeric substrate is annotated with a filled arrowhead (FL-middle panel). A red open arrowhead marks the expected molecular weight of the cleavage product. As positive controls, the cleavage of TVAG_166850 and TVAG_280090 by TvROM1 is shown (A and B-lanes 2 and 3). The chimeric protein EBA-175 with the TM domain of TVAG_280090 (B-lanes 4 and 5), TVAG_244130 (C-lanes 4 and 5), TVAG_335250 (D-lanes 4 and 5), and TVAG_189150 (E-lanes 4 and 5) was released into the media of HA-TVROM2 wt and catalytic mutant co-transfectants, thus these are not considered to be specific TvROM2-generated cleavage products.  Table. Results from surface proteome screen using the parasite search motif.