The actin-based machinery of Trichomonas vaginalis mediates flagellate-amoeboid transition and migration across host tissue

Authors


For correspondence. E-mail gould@hhu.de; Tel. (+49) 211 811 3983; Fax (+49) 211 811 3554.

Summary

Trichomonas vaginalis is the most widespread non-viral pathogen of the human urogenital tract, infecting ∼ 3% of the world's population annually. At the onset of infection the protist changes morphology within minutes: the flagellated free-swimming cell converts into the amoeboid-adherent stage. The molecular machinery of this process is not well studied, but is thought to involve actin reorganization. We have characterized amoeboid transition, focusing in particular on TvFim1, the only expressed protein of the fimbrin family in Trichomonas. Addition of TvFim1 to actin polymerization assays increases the speed of actin filament assembly and results in bundling of F-actin in a parallel and anti-parallel manner. Upon contact with vaginal epithelial cells, the otherwise diffuse localization of actin and TvFim1 changes dramatically. In the amoeboid TvFim1 associates with fibrous actin bundles and concentrates at protrusive structures opposing the trailing ends of the gliding amoeboid form and rapidly redistributes together with actin to form distinct clusters. Live cell imaging demonstrates that Trichomonas amoeboid stages do not just adhere to host tissue, rather they actively migrate across human epithelial cells. They do so in a concerted manner, with an average speed of 20 μm min−1 and often using their flagella and apical tip as the leading edge.

Introduction

Trichomonas vaginalis is the most widespread non-viral and sexually transmitted human parasite worldwide (Benchimol, 2004). Its genome encodes approximately 60 000 proteins, at least twice as many as its mammalian host (Carlton et al., 2007). Of the many hundred million T. vaginalis infections that occur annually, the majority is asymptomatic, while a minority results in trichomoniasis with urogenital tract swelling and inflammatory discharge. However, asymptomatic infections decrease fertility, increase the risk of acquiring HIV and elevate the risk of prostate and cervical cancer (Petrin et al., 1998; Stark et al., 2009; Ryan et al., 2011). Trichomoniasis is commonly treated with nitroimidazole derivates, but resistant strains are on the rise (Kulda, 1999; Upcroft and Upcroft, 2001; Benchimol, 2008).

As an essential component of T. vaginalis urogenital tract colonization, minutes after contact with human urogenital tract cells, T. vaginalis undergoes a radical change in cell morphology (Lal et al., 2006). The free-swimming flagellate cell flattens and spreads out over the host cell tissue to become an amoeba. This ability to shift from a flagellate-motile to an adherent-amoeboid cell stage is rare among protists and, outside the trichomonads, only known among a few species such as Naegleria gruberi and Physarum polycephalum (Fritz-Laylin et al., 2010a; Ryan et al., 2011). With only a few minutes to complete, this amoeboid transition is the fastest described and far more rapid than the N. gruberi transformation, which can require more than an hour (Fulton, 1993; Fritz-Laylin et al., 2010b).

Little is known about the molecular machinery behind amoeboid transition, how it is orchestrated or the components involved. Clues come however from the circumstance that the amoeboid transition connects the two types of cellular locomotion known among eukaryotes: tubulin-based flagellar swimming and actin-based amoeboid gliding. Actin and tubulin are the core components of these locomotion types and of the eukaryotic cytoskeleton in general. Although the proteins associated with these two types of locomotion are ubiquitous among eukaryotic genomes, few species exhibit both types during their life cycle (Fig. 1). In amoeba and apicomplexan parasites, the actin-based cytoskeleton is the key component of gliding locomotion (Fukui, 2002; Baum et al., 2008). The apicomplexan machinery is known as the glideosome and becomes active when the parasites need to overcome biological barriers or invade new host cells (Santos et al., 2009). Previous reports localized the actin-binding proteins alpha-actinin and coronin to the periphery of T. vaginalis amoeboid cells, suggesting that parabasalian amoeboid transformation might involve rearrangement of the actin cytoskeleton (Brugerolle et al., 1996; Addis et al., 1998; Bricheux et al., 1998; 2000) and actin polymerization blockers influence the ability of the parasite to adhere to plastic surfaces (Gold and Ofek, 1992).

Figure 1.

Heat map of actin, tubulin and selected associated proteins. Clusters of orthologous genes (query sequences Table S1) are sorted by the sum of average similarity to all cluster members and represented by the heat map in percent. The cladogram on the left shows a schematic relationship within the five eukaryotic kingdoms and earlier branching is not shown due to its uncertainty. The SAR group unites stramenopiles, alveolates and rhizaria. Symbols indicate whether organisms of the individual groups display flagellate and/or amoeboid phenotypes (pyriform-shape and star-shape respectively).

In contrast to the intestinal parasite Giardia intestinalis, in which common eukaryotic acting regulators are absent (Paredez et al., 2011), the T. vaginalis genome encodes a variety of standard actin-associated and regulating proteins (Fig. 1). The only putative actin-bundling proteins of Trichomonas appear to be alpha-actinins (up to five copies) and two proteins of the fimbrin family, also known as plastins (Aurrecoechea et al., 2009; Elmendorf et al., 2010). These two actin-bundling families contain calponin homology (CH) domains in tandem – each pair making up one actin binding domain (ABD) – and additional Ca++ binding EF-hands (Korenbaum and Rivero, 2002). But in contrast to alpha-actinin, in which the EF-hands are usually located at the C-terminus of the protein and separated from the ABDs by a coiled-coil region, the EF-hand in fimbrin occurs N-terminal, directly adjacent to the ABDs (Korenbaum and Rivero, 2002). For human and Dictyostelium fimbrin it has been shown that binding of Ca++ to the EF-hand blocks the protein's ability to bind to filamentous actin (Namba et al., 1992; Prassler et al., 1997). Only one of the two putative fimbrin proteins of T. vaginalis harbours the N-terminal EF-hand and only for that one, TvFim1, expression evidence exists at TrichDB (Aurrecoechea et al., 2009).

To address the question put forward in the original genome paper of T. vaginalis, whether the structural remodelling of amoeboid parasites really is actin-based (Carlton et al., 2007), we characterized the localization shift of the actin cytoskeleton during infection. We focused on TvFim1, that we demonstrate to bundle actin filaments, and provide detailed evidence for the actin machinery not only to promote flagellate-amoeboid transition upon infection, but also to subsequently mediate active and concerted gliding across host tissue.

Results

Trichomonas encodes the majority of canonical actin-associated proteins and only one functional fimbrin

Based on human proteins associated with the actin and tubulin-based cytoskeleton, which are characterized and annotated the best, we screened for homologues among a range of eukaryotes with representatives from all major phylogenetic groups. T. vaginalis encodes the most canonical actin and tubulin cytoskeleton-associated proteins (Fig. 1). Among the actin-based cytoskeleton this includes genes for capZ, cofilin, formin and the Arp2/3 complex, in addition to at least 29 genes (five orthologous groups) encoding proteins of the actin family itself. Overall, all actins share a 40% global sequence identity and expression evidence exists for almost all of them at TrichDB (Aurrecoechea et al., 2009). The major actin-bundling protein of filopodia, fascin (Jansen et al., 2011), appears absent from the T. vaginalis genome (Fig. 1). The parasite encodes two putative proteins of the fimbrin family; TvFim1 (TVAG_351310) and TvFim2 (TVAG_116370). Both contain two tandem actin-binding domains (ABDs), but only TvFim1 contains the N-terminal Ca++-binding EF-hand, typical for proteins of the fimbrin/plastin family (Korenbaum and Rivero, 2002). Compared with TvFim1 and the human homologue HsFim1 (L-Plastin), TvFim2 appears to have experienced a 5′ truncation, which extends into the N-terminal region of the first CH domain (Fig. S1B). Overall, the two Trichomonas fimbrins share a sequence identity of 74%, which increases to 82% when considering only the region containing the two ABDs.

At TrichDB we identified 145 expressed sequence tags (ESTs) for TvFim1, but none for TvFim2, despite the ESTs originating from mRNA obtained under different conditions such as normal and low-glucose culture conditions or fibronectin-mediated cytoadherence. To confirm we performed reverse-transcriptase PCR on isolated RNA from exponentially growing cells of T. vaginalis, both flagellate and amoeboid stage, and were only able to demonstrate expression of TvFim1 (Fig. 2A). To determine whether the only expressed fimbrin gene is upregulated upon contact with host tissue we performed quantitative real-time PCR on RNA isolated from the free-swimming stage and at 5, 20 and 60 min after fibronectin induced morphogenesis. In two individual experiments, each using biological and technical triplicates, we observed no significant upregulation of TvFim1 over the first hour compared with the slightly upregulated 40S ribosomal protein S5 (Fig. 2B). Hence, the parasite expresses only its full-length fimbrin copy and this gene does not appear to experience a significant upregulation upon contact with host tissue.

Figure 2.

A. Reverse transcriptase (RT) PCR on complementary DNA (cDNA) generated from motile and adherent cells (indicated by symbols as in Fig. 1) in the presence (+ RT) or, as control, absence of the reverse transcriptase enzyme (− RT), demonstrates the expression of TvFim1 only. Control PCR was performed on genomic DNA (gDNA).

B. Quantitative real-time PCR shows TvFim1 not to be significantly upregulated during infection compared with a copy of the ribosomal protein S5 (TVAG_158720).

Overall, TvFim1 shares a global sequence identity with the human homologue HsFim1 of 43%. An alignment of TvFim1 and its homologues from a range of eukaryotes shows that a high amount of conservation is found among the actin binding CH domains themselves (Fig. S1A). The amount of conservation of the individual amino acids thought to interact with actin (Klein et al., 2004; Galkin et al., 2008) is comparable to that of other fimbrins. Reconstructing the interaction of the actin-binding domains of TvFim1 with actin using the known crystal structure of the Schizosaccharomyces pombe fimbrin protein (PDB accession 1RT8) and the F-actin-fimbrin ABD2 complex (PDB accession 3BYH), confirms the alignment sequence identity. The superposition of the two fimbrins of S. pombe and T vaginalis reveals no obvious difference and interacting residues within the CH3 and CH4 domain, such as E448, E523 or F602 (positions based on the alignment of Fig. S1A) are well conserved (Fig. 3).

Figure 3.

Tertiary structure prediction of TvFim1.

A. Ribbon representation of the predicted tertiary conformation of TvFim1 (darker colours) superimposed onto the known crystal structure of the S. pombe fimbrin (lighter colours). Four exemplary amino acids that have been implicated to interact with F-actin are highlighted (grey, sphere representation) in the yeast sequence.

B. Semitransparent surface representation of TvFim1 (same colour code as in A) showing the interaction of the two CH domains 3 and 4 with an actin of T. vaginalis (grey; TVAG_337240). For details please refer to the text.

TvFim1 increases the actin polymerization rate and bundles F-actin

Previous work on human and yeast fimbrins showed this protein family to bundle actin filaments (Namba et al., 1992; Prassler et al., 1997; Skau et al., 2011). To verify TvFim1's function as a fimbrin protein, and analyse the potential influence of TvFim1 on actin polymerization, we heterologously overexpressed a full-length C-terminally HIS-tagged version of TvFim1 in Escherichia coli and purified the protein using fast protein liquid chromatography (Fig. S2). All attempts to purify T. vaginalis actin failed to deliver sufficient quantities for in vitro studies, and we hence used actin purified from rabbit muscle and the amoeboid protist Acanthamoeba instead. We measured assembly kinetics of pyrene-labelled actin by fluorescence spectrophotometry and performed total internal reflection fluorescence microscopy (TIRF).

The pyrene assay showed that only Acanthamoeba actin polymerization significantly increased in the presence of TvFim1 and saturation of polymerization was reached several minutes earlier (Fig. 4A). TvFim1 seems to stimulate the rate of Acanthamoeba actin assembly, but with no apparent polymerization effect on rabbit actin. However, TIRF imaging performed with rabbit actin demonstrated an intense parallel and anti-parallel bundling of actin filaments through TvFim1 (Fig. 4B). Dense bundles of many individual filaments were observed that continued to elongate at both ends of the bundles (Fig. 4C). We did not notice a favoured orientation of the aligned filaments, which is consistent with the bundling behaviour observed for yeast fimbrin (Skau et al., 2011).

Figure 4.

Effect of TvFim1 on in vitro actin polymerization.

A. Polymerization assays of pyrene-labelled Acantamoeba (AcACT) and rabbit actin (RbACT) in the presence and absence of 4 μM TvFim1. Whereas no noticeable difference was observed for the polymerization rate of rabbit actin, the polymerization rate of the protozoan Acantamoeba actin was increased.

B. Rabbit muscle actin polymerization was monitored in the absence (Actin) and presence (3 and 5 μM) of purified TvFim1 using total internal reflection fluorescence microscopy and revealed a higher polymerization rate, as well as a strong parallel and anti-parallel bundling of actin filaments (see also Movies S2 and S3).

C. Details of F-actin bundled by TvFim1. Parallel and anti-parallel bundling is revealed by bundled filaments continuing to polymerize on both ends, marked by yellow arrow heads and at many branching points, marked by red arrow heads.

Fimbrin and actin dynamics upon contact with host tissue

We generated a polyclonal peptide antibody directed against an epitope derived from the first ABD. The antibody detects a single band migrating below the 70 kDa marker lane, fitting well with the predicted mass of TvFim1 of approximately 68 kDa (Fig. S3A and B). Additionally we generated Trichomonas clones expressing haemagglutinin (HA) and green fluorescent protein (GFP)-tagged TvFim1 to further validate antibody specificity, avoid potential cross-talk of our fimbrin antibodies with the host cells and perform live imaging. In protein extract from cells expressing C-terminally HA-tagged TvFim1, a band migrating slightly slower compared with the endogenous copy – due to the HA-tag – was identified plus two faster migrating bands indicating putative proteolytic processing of the fusion protein (Fig. S3B).

In the free-swimming stage, TvFim1 localized to the cell periphery in a gradient-like manner and did not distinctly colocalize with actin (Fig. 5A–C). Identical localization was observed for the endogenously expressed protein, as well as for the HA-tagged copy, indicating that the C-terminal HA-tag did not influence the fusion-proteins localization (Fig. S3D). In adherent-amoeboid cells, which were induced through the exposure to a monolayer of vaginal epithelial cells (VECs), the observed localization pattern of TvFim1 changed dramatically. TvFim1 now begins to colocalize more specifically with actin, also in the still pyriform stage (Fig. 5E–G), but is predominantly observed to form clusters subtending the plasma membrane (Fig. 5M–O) and to associate with filamentous arrays (Figs 5I–K and S4E–G). We further observed filamentous structures (Figs 5I–K and S4E–G) similar to ‘fibrous arrays’ previously observed during the analyses of a coronin homologue in Trichomonas (Bricheux et al., 2000). Live imaging shows that the peripheral clusters at the migration front actually correspond to waves of fimbrin (Fig. 6). Specifically, at the contact sites clusters were observed and further around a contractile ring-like structure of a daughter cell budding from a multinuclear cell.

Figure 5.

Immunolocalization comparison of TvFim1 in free-swimming and adherent-amoeboid cells. In flagellated motile cells fimbrin localizes to the periphery of the cells in a gradient-like manner with no obvious actin colocalization (A–D). In cells exposed to vaginal epithelial cells (E–P) fimbrin colocalizes with actin, associates with structures reminiscent of actin cables (arrows in I and J) and shows peripheral clustering together with actin (M–O). In (E–G) a still pyriform cell can be seen, in which actin and fimbrin already localize in a sharp ring at the cells periphery (arrow) and next to a fully adherent cell with a very different labelling pattern (arrow head). The dashed line in (P) outlines the parasite based on the fluorescent channels of fimbrin and actin. Scale: 10 μm.

Figure 6.

Live imaging of TvFim1::GFP on host tissue. In the top panel a daughter cell can be seen to bud from the mother cell. TvFim1::GFP concentrates around a contractile ring-like structure (indicated by an asterisk) towards the basal end of the daughter cell during separation. During amoeboid migration across host tissue individual waves of TvFim1::GFP were clearly visible (wave indicated by arrow heads in the central panel). Clustering of TvFim1::GFP was predominantly found to occur around the initial host contact sites (indicated by arrow heads in the bottom panel). Images were taken 15 min (for clustering) and 70 min (for division and migration) after inoculation of host tissue with T. vaginalis respectively. Scale: 10 μm.

The time-lapse videos allowed to document different motion patterns of T. vaginalis, most importantly amoeboid migration across host tissue, only minutes upon exposure to VECs (Movies S7–S10). T. vaginalis actively roams across the VEC-monolayer with an average speed of 20.2 μm min−1 (Table S3) and appears to use its flagella and apical tip as the guiding end (Fig. S5), which also appears to downright penetrate areas of adjacent and adherent VECs. One hour post infection we observed some amoeboid cells to massively increase their cell mass and adherent surface (Fig. S5); a process we refer to as juggernauting. The cells display multiple flagella pockets and nuclei (Figs S5 and S4K), which is consistent with previous observations (Yusof and Kumar, 2011). We observed individual T. vaginalis cells rapidly budding off from multinuclear cells, similar to what was described for Tritrichomonas foetus (Pereira-Neves and Benchimol, 2009). Overall, the cells appear highly agile while they migrate and scavenge from host tissue.

To determine whether the association of fimbrin and actin is specific for phenotypic plasticity induced through VECs, we investigated what happens during phagocytosis of Saccharomyces. Exponentially growing parasites were exposed to yeast cells in a 1:50 parasite/yeast ratio and incubated for 60 min in the CO2 incubator – to be comparable to the other experiment – before fixation. We observed T. vaginalis to incorporate entire and multiple yeast cells at the same time, as previously documented (Pereira-Neves and Benchimol, 2007). As for the amoeboid form of the parasite, fimbrin now clearly colocalizes with actin, forming distinct rings around the large phagocytic vesicle, but at the same time highlights a range of other subcellular structures, including some punctuate staining and again fibrous patterns (Fig. 7).

Figure 7.

Fimbrin and actin localization during phagocytosis. During phagocytosis of yeast cells actin and its bundling protein fimbrin colocalize and predominantly cluster around bordering phagocytic vesicles (asterisks). Other structures, in parts similar to those observed upon contact with host tissue, are also apparent, including smaller vesicular particles (arrow heads). Scale: 10 μm.

Discussion

Through the characterization of the fimbrin protein family and live cell imaging of T. vaginalis we provide detailed evidence that the parasite's amoeboid morphogenesis during infection is accompanied by a rapid reorganization of the actin cytoskeleton. The flagellate to amoeboid transformation occurs several times faster in T. vaginalis than in N. gruberi (Arroyo et al., 1993; Fritz-Laylin et al., 2010a). We suggest that this might be attributable to the prey-induced activation of ‘sleeping’ components, such as the T. vaginalis fimbrin protein TvFim1. Accordingly, the abundance and gradient-like localization of the actin-bundling protein below the plasma membrane in the flagellate form (Fig. 5A–C) appears to be a prearrangement that keeps the free-swimming parasite primed for actin cytoskeleton mediated phenotypic plasticity upon contact with host tissue. With the molecular machinery required already in place, only the final signal at the end of the activation cascade (such as calcium depletion or an altered phosphorylation state of the protein; see below) is required to trigger morphogenesis. This would explain why no significant upregulation of TvFim1 during infection is detected, when compared with a copy of the 40S ribosomal protein S5. By contrast, upregulation of alpha-actinin using semi-quantitative PCR has been reported to occur during infection (Noël et al., 2010), whereby expression analysis in Trichomonas is in general impaired by the many gene duplicates present.

The majority of known core components of actin regulation are encoded by Trichomonas (Fig. 1), and some gene families have been significantly expanded in the course of genome duplication (Carlton et al., 2007). Paralogous copies of many genes are present and simultaneously expressed. Proteins of the Trichomonas actin family are encoded in 29 copies and expression evidence exists for 24 of them at TrichDB. Only two genes encode proteins of the actin-bundling fimbrin family (TVAG_351310 and TVAG_116370) and the second, TvFim2, appears to be a pseudogene, a 5′ truncated paralogous copy of TvFim1. During or after gene duplication, TvFim2 lost the N-terminal EF-hand – the characteristic and regulatory module of fimbrin proteins – and with it its function. We could neither detect expression of TvFim2 on RNA level through RT-PCR, supporting the lack of expression evidence on TrichDB, nor on protein level through Western blot analysis using our fimbrin antibody. The epitope of TvFim1 differs from that of TvFim2 by only one amino acid at position 185, which suggests the antibody should recognize both proteins. However, only one band of approximately 68 kDa and none of 50 kDa was detected (Fig. S3), corresponding to the predicted sizes of TvFim1 and TvFim2 respectively.

TvFim1 increases the polymerization rate of free Acanthamoeba actin, albeit weakly, and served as a potent filament cross-linker, bundling rabbit F-actin in a parallel and antiparallel manner (Fig. 4). Generally, fimbrin is not considered to be a canonical actin nucleator, such as formin or the ARP complex (Butler and Cooper, 2009). Our results are not yet able to determine the exact actin polymerization potency of TvFim1; however, we can exclude ‘false positive polymerization’ observed through scattering due to the settings in the pyrene assays. Experiments with purified Trichomonas actin are currently hindered, because it does not express well in any system tested. How, specifically, fimbrin increases actin polymerization is unknown for any model system, but independent experiments on homologues from yeast and Arabidopsis suggest that fimbrin might lower the necessary critical concentration for actin to polymerize (Cheng et al., 1999; Kovar et al., 2000). Intriguingly, only the polymerization of Acanthamoeba G-actin was significantly increased, but not of rabbit G-actin, although rabbit F-actin was clearly bundled by TvFim1. Amino acid residues that have been implicated in the interaction with actin are as well conserved in Trichomonas as in other eukaryotes (Fig. S1). Furthermore, the predicted tertiary structure of TvFim1 is almost identical to that of yeast (Fig. 3), and thereby fails to offer obvious clues as to the structural basis of the differential behaviour of TvFim1 to the different actins investigated here. However, our results provide further indirect evidence that fimbrin might be a weak nucleator, too, as the polymerization increase observed was not just due to a side-effect of the bundling of filamentous actin.

Hence, T. vaginalis expresses only one protein of the fimbrin family that, as shown here, fulfils all functions attributed to the fimbrin family and relocates during morphogenesis and phagocytosis. A knockout or knockdown of TvFim1 to underpin its importance during infection was at this point not successful. Knockouts could not be established – which only twice has been successful for T. vaginalis and only once reported to reveal a noticeable phenotype (Land et al., 2004; Pereira-Brás et al., 2013) – and expressing TvFim1 antisense RNA did not lead to a decrease of TvFim1 transcript (Fig. S7). Yet, modifications of fimbrin seemed to influence the transfected cells: (i) the HA-tagged line duplicated with only about half the speed, (ii) in a HaloTag line we could not detect any protein, albeit the construct was identified on RNA level (Fig. S6), suggesting either transcriptional inhibition or immediate post-translational degradation, (iii) the HA-tagged copy was also observed to show a degradation product not observed for the endogenous copy (Fig. S3B). In summary, fimbrin seems to fulfil pleiotropic roles, as indicated by its colocalization with actin during adherence to and migration across host cells and during phagocytosis of yeast, and a knockout or knockdown appears to influence the parasites viability.

The behaviour of TvFim1 during infection suggests the protein to be predominantly active during the amoeboid stage and most likely during phase transition from flagellate to the amoeboid. Clustering, specific aggregation and more defined localization of the protein were only observed in the adherent-amoeboid form of Trichomonas (Figs 5, 6 and S4) or during phagocytosis (Fig. 7). The signals that trigger morphogenesis and surface attachment are unknown, but TvFim1/actin re-localization and TvFim1-dependent actin polymerization provide new tools towards their investigation. Trichomonas can spontaneously adhere to glass and plastic surfaces and the amount of adherent cells increases when the surfaces are coated with fibronectin (Brugerolle et al., 1996). Only on VECs however have we observed that the vast majority of parasite cells undergo transition from a motile to an adherent stage over time. This suggests multiple signals and a complex cascade involving the correct recognition of host tissue. This is further supported by the fact that the parasite is also able to completely engulf and phagocytose other eukaryotes such as yeast – a process also involving a co-ordinated, but different re-localization of actin and fimbrin (Fig. 7). Phagocytosis must hence require a different downstream behaviour after prey recognition in comparison to adhering to VECs, and further unknown regulatory units orchestrating the different processes. Whether the activation of TvFim1 requires the EF-hand to not be bound to Ca++ (Namba et al., 1992; Prassler et al., 1997), or whether it is regulated through the phosphorylation of certain serine residues as shown for the human L-plastin (Shinomiya, 2012), remains to be investigated. TvFim1 has about 20 potential phosphorylation sites as predicted by NetPhos alone.

However, the different patterns of actin and fimbrin localization within the parasite during different morphological shifts reflect the complexity of actin and fimbrin dynamics in Trichomonas. In the amoeboid form actin and fimbrin colocalize in patches opposing the trailing end of the amoeba (Fig. 5M–O, Fig. 6, Fig. S4A–C), and with fibrous structures similar to those previously observed (Bricheux et al., 2000). The latter could represent actin cable-like structures. These were also observed during phagocytosis of yeast to a minor degree, albeit here the clustering around the phagocytic vacuole was the predominant structure (Fig. 7). The punctate localization of fimbrin observed during phagocytosis (Fig. 7I–K) might represent endocytic vesicles that are known to associate with fimbrin-bundled actin filaments in mammals and yeast (Hagiwara et al., 2011; Skau et al., 2011). This would then furthermore suggest an interaction of TvFim1 with proteins of the Rab family, as observed for the mammalian Rab5 and fimbrin during endocytosis (Hagiwara et al., 2011).

The bundling of actin through fimbrin furthermore reflects only a fraction of the proteins likely associated with infection- and phagocytosis-related actin remodelling. Coronin has been localized to ‘various dynamic subcortical zones’ of the parasite, too, and suggested to play an important role during morphogenesis (Bricheux et al., 2000). In other eukaryotic systems actin-regulating proteins are known to together orchestrate a variety of processes, in particular those associated with motion (Golsteyn et al., 1997; Eichinger et al., 1999; dos Remedios et al., 2003; Xue et al., 2010). As mentioned earlier, the majority of these proteins are encoded by T. vaginalis (Fig. 1). One must predict that only their concerted effort allows the different responses observed upon contact with host tissue and yeast cells (Figs 5/6 and 7 respectively), and the elaborate patterns of motion revealed, which include swift migration across host tissue and a sweeping of the substrate through the constricted apical tip of the parasite (Movies S4–S10). The latter also supports previous findings, which suggested a flagella-localized tetraspanin (TvTSP6) to serve sensory reception (de Miguel et al., 2012). T. vaginalis might therefore offer an alternative system to study the actin dynamics of a pathogenic protist.

In summary our results demonstrate that fimbrin likely assists actin bundling across a multitude of different processes and it appears that many of these functions are conserved among a diverse range of evolutionary distant eukaryotes. In the parasite T. vaginalis one of the primary function of the early actin-based – and fimbrin accompanied – morphogenesis includes phenotypic plasticity during host cell attachment: the increase of surface interactions with VECs to scavenge substrate while gliding across host tissue. Another function could include the feeding on, and defence against, other microorganisms of the vaginal flora the parasite likely encounters during infection. Our results confirm the actin machinery to accompany phagocytosis and the parasite has been shown to phagocytose a broad range of prokaryotic and eukaryotic cells (Street et al., 1984; Benchimol and de Souza, 1995; Rendon-Maldonado et al., 1998; Pereira-Neves and Benchimol, 2007). These observations shift the way we view parasite–host tissue interaction and offer a system and molecular proxy to study cytoskeletal actin dynamics in a protozoan parasite during infection-associated phenotypic morphogenesis.

Experimental procedures

Cultures

Trichomonas vaginalis strains T016, FMV1 and T1 were cultivated in tryptone–yeast extract maltose medium {2.22% (w/v) tryptose, 1.11% (w/v) yeast extract, 15 mM maltose, 9.16 mM l-cysteine, 1.25 mM L(+)ascorbic acid, 0.77 mM KH2PO4, 3.86 mM K2HPO4, 10% (v/v) horse serum, 0.71% (v/v) iron solution [= 1% (w/v) Fe(NH4)2(SO4) × 6H2O, 0.1% (w/v) 5-sulfosalicylacid]} at 37°C and 5% CO2 in a Galaxy 48R (Eppendorf, Germany). Immortalized VECs (VECs MS-74) were cultivated in 45% DMEM (Invitrogen, #31885), 45% Keratinocyte-SFM (Invitrogen, #37010022) and 10% fetal calf serum (FCS) in standard cell culture flasks (75 cm2) and at the same conditions as T. vaginalis. At high confluency, cells were washed twice with Dulbecco's PBS (PAA, #H15-001), digested with trypsin (Invitrogen, #25300-054) for 5 min, before inactivation with FCS. Cells were then pelletized and resuspended in fresh media and split 1:10 into new flasks and medium. To prevent bacterial contamination a penicillin/streptomycin mix was added to a final concentration of 100 μg ml−1 to both media.

Database screening and structure prediction

We compiled a list of 77 actin and tubulin associated genes, mainly from Homo sapiens (Table S1). These query sequences were matched against the complete genomes of 25 eukaryotes representing the different eukaryotic phyla (Table S2). The proteomes were obtained from RefSeq (Pruitt et al., 2007) except Bigelowiella natans (http://genome.jgi-psf.org/Bigna1/), Cyanidioschyzon merolae (Matsuzaki et al., 2004) and Cyanophora paradoxa (Price et al., 2012), which were downloaded from the corresponding genome project homepages. For five species without complete genomic sequences, ESTs were obtained downloaded from dbEST (Boguski et al., 1993; see also Table S2). Clusters of homologous proteins were reconstructed from a total of 453 696 proteins encoded within the eukaryotic chromosomes. A blast (Altschul et al., 1997) search analysis yielded 1497 best blast hits. All protein pairs were globally aligned using Needleman–Wunsch algorithm with needle program (Rice et al., 2000). A total of 12 758 protein pairs having global amino acids identities ≥ 20% were clustered into protein families using MCL algorithm (Enright et al., 2002) with the default parameters. The query proteins within each cluster were BLASTed against ESTs from Alexandrium tamarense, Oxyrrhis marina, Porphyra yezoensis, Reclinomonas americana and Physarum polycephalum, and matching ESTs were translated to amino acid sequences and added to the clusters based on the highest similarity. Clusters were further refined by splitting of paralogous clusters, merging of orthologous clusters, removing of single gene clusters and obvious paralogous sequences from a cluster, based on protein annotations and examination of phylogenetic trees reconstructed using PHYML (Guindon et al., 2010). This procedure yielded 62 protein families with a total of 891 genes.

The tertiary structure of TvFim1 (without the EF-hand) was generated on the basis of the known crystal structure of the Schizosaccharomyces pombe fimbrin PDB accession 1RT8 (Klein et al., 2004), using the MODELLER software (Eswar et al., 2006). A local alignment of TvFim1 to 1RT8 in MODELLER revealed an identity of 46% over an alignment length of 456 amino acids, representing the highest level of sequence similarity in the PDB database (http://www.rcsb.org/pdb/). In order to model the actin/ABD2 complex structure we used the structure of the F-actin-fimbrin/plastin ABD2 complex of Homo sapiens PDB accession 3BYH (Galkin et al., 2008) together with TvFim1 and an abundantly expressed actin of T. vaginalis (TVAG_337240).

Gene cloning and heterologous overexpression of TvFim1

TvFim1 gene sequence was amplified by TvFim1_NdeI_FOR (5′-CTGACGCATATGGCTGTAAACGCTGCG-3′) and TvFim1_BamHI_REV (5′-GACGTGGATCCTTGATCCATGGCCATAAGAGA-3′) using a proof-reading polymerase and ligated into expression vector pTagvag2 for IFAs, pTvGFP [based on pTagvag2 with a Trichomonas codon-optimized green fluorescent protein (GFP)-tag replacing the HA] for live cell imaging of FMV1 and pETEV21a for overexpression respectively, and verified by sequencing. Thirty micrograms of the plasmid DNA was used for transfection of 2.5 × 108 T. vaginalis cells using standard electroporation (Delgadillo et al., 1997). After 4 h of incubation neomycine (G418) was added to a final concentration of 100 μg ml−1 for selection. For heterologous overexpression E. coli C41(DE3) was used and transformed with pETEV21a including the gene of interest. Briefly, 1 l of culture was incubated on an orbital shaker at 37°C until an optical density of OD600 0.4–0.6 was reached. Overexpression was induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) followed by a 4 h incubation at 37°C. Cells were pelletized, washed once with phosphate-buffered saline (PBS) and again pelletized. Cells were lysed by plotting and subsequent disruption using the OneShot disruptor (Constant Systems Limited). HIS-tagged protein was isolated using a HISTrap column (HisTrap HP 5 ml, GE Healthcare) and standard fast protein liquid chromatography on an Äkta P-920 (GE Healthcare).

Quantitative real-time PCR

All experiments were carried out using a StepOnePlus and Power SYBR Green master mix (Applied Biosystems). T. vaginalis RNA was isolated from biological triplets using TRIzol (Invitrogen) from the motile-flagellated and three adherent-amoeboid stages at 5, 20 and 60 min after fibronectin-induced morphogenesis (fibronectin from human plasma; Sigma F0895). RNA was transcribed into cDNA with iScript cDNA Synthesis Kit (Bio-Rad) and used as a template for real-time quantitative PCR. Primers used were: Fimbrin: TvFim1_qFOR: 5′-ACAACCTTTACGACGGCATC-3′ and TvFim1_qREV: 5′-GCTTTGTCTTGTTGGCCTTC-3′, 40S ribosomal protein S5: 40SRibo_qFOR: 5′-GCATTGATCAGGCTCTCTCC-3′, 40SRibo_qREV: 5′-ATGCGCTCAAGTTCGTCTTT-3′. For absolute quantification, a linear relationship of Ct and log (DNA weight) was plotted for the target transcripts and used to infer its corresponding amount. As the target gene sequence was known, the copy numbers implied in the quantity may be calculated by the molecular weight of the sequence, which led to the estimate of the transcript number in the unknown sample (Lu et al., 2012).

Western blotting and immunofluorescence

Protein samples were separated through standard SDS-PAGE and blotted onto nitrocellulose membrane. Membranes were blocked in 5% milk powder in Tris-buffered saline pH 7 (blocking buffer) for 30 min. Blots were incubated with the primary antibody at a dilution of 1:1000 or 1:5000 in blocking buffer either overnight at 4°C or for 1 h at room temperature (RT) and then washed 3× with TBS-T (TBS + 0.1% Tween 20), followed by the incubation with the secondary antibody (1:2000 or 1:10 000) and identical subsequent washes. Detection of the chemiluminescence signal was performed through the SuperSignal West Pico Chemiluminescent Substrate Kit (Thermo Scientific) according to the manufacturer's protocol.

For immunofluorescent labelling all wells of CultureSlides (BD Falcon, #354114) were loaded with about 1.5 × 105 VECs 48 h prior to fixation and incubated at 37°C and 5% CO2. Medium was discarded, followed by the inoculation with 1.5 × 106 T. vaginalis cells for a minimum of 15 min at 37°C and 5% CO2. Supernatant was discarded, adhesive cells washed gently with PBS and fixed and permeabilized with a solution containing 4% PFA and 0.1% Triton X-100 in PBS for 15 min at RT. After discarding the solution, cells were blocked (1% BSA, 0.25% Gelatine, 0.05% Tween 20 in PBS) for 30 min at RT. Slides were incubated with the primary antibody (1:50 to 1:500 in blocking buffer) for 1 h at RT, supernatant discarded and slides washed at least twice with PBS before incubation with the secondary antibody at 1:2500 to 1:5000 for 1 h at RT in the dark. Cells were then mounted in Fluoroshield with DAPI (Sigma #F6057). Samples were stored at 4°C, dark, until imaging using a Zeiss LSM 710 confocal microscope. Primary antibodies: polyclonal peptide antibody TvFim produced in rat (Eurogentec) against amino acid sequence CRKFVGPREIVKGNQR, monoclonal HA-antibody (Sigma #H9658), monoclonal actin-antibody (Sigma #A4700) and mouse anti-GFP (Invitrogen #332600). Secondary antibodies: AlexaFluor488-anti-mouse IgG (Invitrogen #A11001), AlexaFluor594-anti-rat IgG (Invitrogen #A11007), ImmunoPure Goat Anti-Mouse IgG and Anti-Rat IgG (Pierce #31430 and #31470 respectively). Additional actin staining through TexasRed-X phalloidin (Invitrogen #T7471).

Yeast powder (RUF Lebensmittelwerk KG, Quakenbrück, Germany) was resuspended in sterile water (37°C) and washed three times with 0.1 M PBS. The cells were then resuspended in TYM-medium without serum at a concentration of 5 × 107 cells ml−1 and immediately used for the phagocytosis assay. T. vaginalis and yeast were mixed with a ratio of 1:50 in TYM without serum and incubated on CultureSlides (BD Falcon, #354114) for 1 h at 37°C and at 5% CO2. Immunofluorescent labelling and microscopy was identical to the steps described above.

Live imaging

For live imaging μ-Slide VI 0.4 (Ibidi, #80606) was used. All six slide chambers were pre-loaded with 9 × 103 VECs, 48 h prior to infection and incubated at 37°C and 5% CO2. Medium was discarded, followed by the inoculation with 9 × 104 T. vaginalis cells and immediate microscopy, using a Zeiss AxioObserver.Z1 microscope with AxioCam MRm (real-time) and AxioCam ICC1 (time-lapse) at 37°C and 5% CO2.

Actin polymerization assays

TIRF microscopy was carried out using the Zeiss Laser TIRF3 microscope at the Summer School on Actin Dynamics in Regensburg (DFG program SPP 1464). Actin polymerization was performed using 1.5 μM rabbit muscle actin alone as a control and in the presence of 3 and 5 μM TvFim1 in FPLC elution buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl, 500 mM imidazole, 1 mM NaN3). Protein mixtures were diluted in freshly prepared fluorescence buffer containing 10 mM imidazole-HCl (pH 7.8), 50 mM KCl, 1 mM MgCl2, 100 mM dithiothreitol, 3 mg ml−1 glucose, 20 μg ml−1 catalase, 100 mg ml−1 glucose oxidase and 0.5% methylcellulose to induce actin polymerization. Actin polymerization was induced in a solution containing 1.5 μM actin monomers (7% labelled with Alexa 568) and 3 or 5 μM TvFim1 respectively. Actin polymerization assays were monitored using the microplate reader Infinite200 PRO (Tecan Group, Ltd) and a Nunclon 96 flat black well plate. G-actin mix [4 μM unlabelled Acanthamoeba actin, 5% pyrene-actin in G-buffer (2 mM Tris pH 8.0, 0.5 mM DTT, 0.2 mM ATP, 0.1 mM CaCl2, 0.01% NaN3)] was measured alone and including 4 μM of TvFim1 with the following parameters: excitation = 365 nm, emission = 410 nm, Z position = 17 000, gain = 135, read every 3 s for 30 min.

Acknowledgements

We thank G. Landan and S. Nelson-Sathi for help with MatLab, all supervisors of the Actin Summer School of 2011 at the Universitätsklinikum Regensburg (Germany), in particular Margot Quinlan (University of California, USA) for helpful discussions. We thank J. Alderete and M. Benchimol for sharing T. vaginalis strains. This work was funded by a DFG grant (GO 1825/3-1) and the support of the ‘Stiftung zur Erforschung infektiös-immunologischer Erkrankungen’ to S.B.G.

Ancillary