Introduction

Three trypanosomatids—Trypanosoma brucei, Trypanosoma cruzi, and Leishmania major—have been sequenced so far (Berriman et al. 2005; El-Sayed et al. 2005b; Ivens et al. 2005). They are kinetoplastids according to the presence of a DNA-containing organelle known as kinetoplast in their single large mitochondrion. They are primitive eukaryotes (Sogin 1991) and often categorized into the super group Excavates (Adl et al. 2005; Simpson and Roger 2004). T. brucei and T. cruzi are the aetiological agents of African sleeping sickness (HAT) and Chagas’ disease in South and Central America, respectively. Leishmania spp. causes a variety of diseases throughout the tropics and sub-tropics (Burri and Brun 2003). There are half a billion people, primarily in tropical and subtropical areas of the world, who are at the risk of contacting these diseases, and there are more than 20 million infections and 100,000 deaths per year caused by trypanosome infection (Stuart et al. 2008).

Although having diverged 200 to 500 million years ago (Douzery et al. 2004; Stevens et al. 2001; Overath et al. 2001), predating the emergence of mammals (O'Brien et al. 1999), the genomes of the trypanosomatid species, exhibits high synteny. A comparison of gene content and genome architecture of T. brucei, T. cruzi, and L. major revealed a conserved core proteome of ~6,200 genes that are largely syntenic polycistronic gene clusters (El-Sayed et al. 2005a) and gene families (Brenchley et al. 2007). As far as the Leishmania species are concerned, the orthologous chromosomes of L. major, Leishmania braziliensis, and Leishmania infantum are remarkably conserved in both gene content and order (Peacock et al. 2007), even though they diverged 40 to 80 million years ago (Fernandes et al. 1993).

For Leishmania spp., GP63 (also known as leishmanolysin) is considered to be a major virulence factor (Chaudhuri et al. 1989; Joshi et al. 2002; Yao et al. 2003). Unlike Leishmania spp., T. brucei employs a variant surface glycoprotein (VSG) as its main virulence factor and T. cruzi uses mucins and transsialidases/sialidases. Among the known virulence factors involved in host–parasite interaction, GP63 proteases are shared by the three trypanosomatids (Santos et al. 2006; Yao 2010). The T. brucei GP63 proteases are predominantly expressed in the bloodstream form, and the experimental evidence implicated that they may protect bloodstream trypanosomes against complement-mediated lysis by impairing a protein-processing function (LaCount et al. 2003; El-Sayed and Donelson 1997; Grandgenett et al. 2007). Neutralization assays have indicated that anti-GP63 serum has a significant inhibitive effect on T. cruzi infection (Kulkarni et al. 2009; Cuevas et al. 2003). Genome content and structure comparisons among T. cruzi, T. brucei, and L. major demonstrated that among the large-scale syntenic regions, gene divergence, acquisition and loss, and rearrangement have shaped each genome significantly (El-Sayed et al. 2005a). In order to further understand the virulence factor of these three parasites, we focused our analysis on their GP63 proteases, taking the advantage of the extremely homologous Leishmania spp.(Smith et al. 2007) and the knowledge about the parasitic trypanosomatid species (Lipoldova and Demant 2006). We also scrutinized sequence variations that may be involved in parasite–host interactions.

Materials and methods

Data sources and analysis tools

We downloaded 562 sequences annotated as “leishmanolysin” from Genbank, Refseq of NCBI, EMBL, DDBJ, and PDB, and Swiss-Prot on June 23 2008 and used HMMER (Krogh et al. 1994) version 2.3.2 (Linux), a freely distributable implementation of profile Hidden Markov Models software, to discover GP63 sequences from the Pfam database. After removal of short length sequences, we obtained 290 GP63 sequences from 63 species (more details are summarized in Online Resources 3). Information for all the species’ name and sequences’ accession can be referred to Online Resources 1 and 2, respectively.

We predicted glycosylphosphatidylinositol (GPI) modification site, using big-PI Predictor (http://mendel.imp.ac.at/sat/gpi/gpi_server.html). When sequences are predicted to have GPI modification, P and S indicate higher and lower confidence, respectively. We display three-dimensional protein structures using the molecular graphics system PyMOL version 0.96 (http://www.pymol.org/).

Phylogeny

Since the most conserved portion of GP63 proteases across multiple species is about 100 aa in length (between amino acid position 295 and 393 of Leima90), we constructed the phylogenic tree based on this segment of 290 GP63 sequences, using a maximum likelihood method implemented in the phylogeny inference package (version 3.67; Retief 2000). The robustness of the phylogenic inferences was tested by bootstrapping, using 100 replications of the data based on the criterion of 50% majority-rule consensus.

Positive selection test

We used Codeml in the PAML package (Yang 1997) for the evolutionary analysis of the GP63 protease in Leishmania spp. There are three models in Codeml: branch, site, and branch–site. In particular, the branch model allows the ω ratio to vary among branches in the phylogenic tree and is useful for detecting positive selection acting on particular lineages. The site model allows the ω ratio to vary among different sites, whereas the branch–site model allows the ω ratio to vary among both sites and lineages and attempts to detect positively selected sites along a limited number of sequences. We analyzed 17 Leishmania spp. GP63 sequences that have potential GPI structure to investigate if ω ratio varies among branches. After constructing free-ratio and one-ratio models, we used likelihood ratio test (LRT) to evaluate the opposing hypotheses. Since the null hypothesis that ω ratio varies among branches made better sense, we applied the branch–site model to find lineages (branches) and sites where positive selections are at work. We set the parameters as recommended (the Bayes empirical Bayes results of NSsites = 2 and NSsites = 8; Yang et al. 2005) for the site model analysis.

Results

A conserved motif implies functional role for GP63 proteases

We started our analysis by dividing GP63 proteases into four groups according to their taxonomy and pathogenicity: (1) non-pathogenic protists (such as Tetrahymena thermophila and Paramecium tetraurelia), (2) pathogenic protists (such as Leishmania spp., T. cruzi, and T. brucei), (3) multicellular animals, and (4) plants (Fig. 1a). Since the pfam GP63 model is built with majority ruling, sequences (~500 aa) from group 2, group 3, and group 4 showed higher similarity, whereas sequences of group 1 have much shorter conserved length. Nevertheless, all sequences do share a 100-aa segment that follows a zinc-binding motif (HEXXH): between amino acid positions 340–430 of the pfam GP63 model and 295–393 of Leima90. The relative positions of this segment to the zinc-binding motif appeared group specific: 25 aa in pathogenic protists, 63 aa in animals, and 34 aa in plants; it is relatively variable in non-pathogenic protists: 30 to 50 aa. There are only a handful sequences (18 sequences: five from T. thermophila, one from P. tetraurelia, nine from T. cruzi, one from Tetraodon nigroviridis, one from Rattus norvegicus, and one from Pan troglodytes; see Online Resources 4) whose conserved segments are interrupted. We investigated the similarity matrix of the shared segment through pair-wise blast among the 290 sequences. The similarity score is significantly higher in Leishmania spp. and among multicellular eukaryotes (Fig. 1b).

Fig. 1
figure 1

The domain structure of GP63 proteases. a GP63 proteases were divided into four groups according to the taxonomy and pathogenicity: red group 1, black group 2, blue group 3, and green group 4. The y-axis indicates the length of the GP63 protease domain of the 294 sequences. There are an apparent motif absence of around 500 aa in T. cruzi and a less evident absence of around 350 aa. b Similarity matrix of the shared segment. Pair-wise blast was carried out among the 290 amino acid sequences, and the similarity score was colored. Le Leishmania spp., Tb T. brucei, Tc T. cruzi

Full size image

We aligned the conserved segment of the 290 sequences and found 12 highly conserved sites (we numbered them sequentially; Online Resources 5). We suggest that His3 and Met5 are the two amino acid residues within the active site for proteolysis function. In particular, His3 is the third histidine for zinc binding; Met5 is Met-turn-like, similar to other Metzincin class zinc proteases such as astacin and collagenase (Schlagenhauf et al. 1998). Among the remaining residues, cysteine may function in maintaining protease structure and the potential roles for the rest of the residues remain to be elucidated. Given the high degree of conservation and the presence of active sites, this shared segment is most likely to be a functional motif of GP63 protease in supporting catalytic reaction but not species-associated functions.

GP63 proteases of pathogenic protists bear more similarity to those of multicellular eukaryotes than non-pathogenic protists

We aligned the entire length of GP63 proteases from the pathogenic parasites and animals. In addition to the highly conserved segment, we found fifteen conserved Cys and four conserved Pro residues between multicellular eukaryotes and pathogenic protists (Fig. 2). The group 3 (animals) members are clustered into two clades: insects and others; the insect clade bears more similarity to that of the trypanosomatids than other animals (Online Resources 3). In addition, the presence or absence of GPI anchorage is also diverged among GP63 proteases; it is absent in group 1 and group 4, but in group 2, 11% and 5% of the proteases were predicted to have GPI at P and S levels, respectively. In group 3, the predictions are much higher, 21% and 38% at P and S levels, respectively. All the P level sequences belong to the insect clade.

Fig. 2
figure 2

Alignment of GP63 sequences between group 2 and group 3. We divided group 3 into two clades: insect and other animals. We selected ten representative sequences for each group and trimmed the portions with low similarity. We color-coded the sites that are conserved in more than 80% sequences, underlined the common segment, and used asterisks to indicate the 15 conserved Cys and four conserved Pro residues

Full size image

The T. cruzi GP63 proteases share a unique conserved domain

In T. cruzi, GP63 fell into four clades: TcGP63-a, TcGP63-b, TcGP63-c, and TcGP63-d (Fig. 3). The between-clade and within-clade amino acid similarities are 30% and 80%, respectively, as domain absence was frequently observed, involving both terminal and internal domains. Each clade has its signature deletions: 25 members of the TcGP63-a clade have N-terminal deletions, the 22 members of the TcGP63-b clade have deletions at both ends, eight members of the TcGP63-c clade have internal deletions of 10 aa (360–370 aa according to the pfam GP63 model), and 29 members of the TcGP63-d clade miss 80 aa (480–560 aa) when compared to the pfam model. We found that this missing 80 aa segment is highly conserved (data not shown).

Fig. 3
figure 3

Maximum likelihood (ML) tree of 290 GP63 sequences from 63 species. T. cruzi has four main groups: TcGP63-a, TcGP63-b, TcGP63-c, and TcGP63-d. T. brucei has three groups: TbMSP-A, TbMSP-B, and TbMSP-C. Sequence is named with an abbreviation of the species name plus a serial number. For example, Leima90 is a GP63 sequence from the species L. major and 90 stands for the sequence series in our data. Clades with more than eight sequences of the same species were condensed. Names for all condensed branches were in the parentheses. The ML tree was constructed by using Phylip (also see Online Resource 6 for all data)

Full size image

Previous studies have suggested that TcGP63-b is involved in parasite infection. Three GP63 proteases in two groups were identified based on Northern blotting: Tcgp63-I, Tcgp63-II, and Tcgp63-III (Cuevas et al. 2003). Tcgp63-I was thought to be differentially expressed depending on the life cycle of T. cruzi, and the antibodies against Tcgp63-I were demonstrated to partially block the infection of Vero cells by trypomastigotes. Tcgp63-Ia and Tcgp63-Ib are GP63-1 and GP63-4, respectively, according to Grandgenett’s classification (Grandgenett et al. 2000); both are grouped into TcGP63-b (Trycr355 and Trycr362) in our analysis.

In addition, among the 2,784 T. cruzi proteins (Atwood et al. 2005), we found that all GP63 proteases fell into one group: TcGP63-b. Therefore, TcGP63-b is also abundantly expressed in T. cruzi. However, the remaining three GP63 groups of T. cruzi (TcGP63-a, TcGP63-c, and TcGP63-d) have not yet been studied despite the fact that they share the highly conserved eighteen Cys and eight Pro residues (except a few members of TcGP63-d, which have lost the twelfth Cys residue). In addition, the predicted N-glycosylation sites seem to be specific for each group; the TcGP63-b members have two potential N-glycosylation sites, whereas TcGP63-a members have three (Cuevas et al. 2003).

We did not define new GP63 family members in T. brucei, where nine non-redundant sequences are grouped into three as reported previously (LaCount et al. 2003). One striking feature of the T. brucei GP63 groups is that they are scattered into the Leishmania spp. groups and the T. cruzi groups. It suggests that the GP63 genes (paralogs) were duplicated and started to diverge from each other before their speciation events. TbMSP-A (Trybr231, Trybr237, and Trybr238 in this study) has a unique extended C-terminal region rich in serines and glutamates. TbMSP-B (Trybr228, Trybr227, and Trybr230) is constitutively expressed in both the procyclic and the bloodstream stages, whereas TbMSP-A and TbMSP-C (Trybr239) only express at the bloodstream stage. Experimental evidence showed that depletion of TbMSP-B inhibits the release of VSG and thus increases the steady level of cell-associated VSG (LaCount et al. 2003).

Positive selection of Leishmania spp. GP63 sites

Leishmania spp. infection is known to lead to different sicknesses: cutaneous leishmaniasis (CL), mucocutaneous leishmaniasis (ML), and visceral leishmaniasis (VL). CL heals without any treatment, ML causes disfiguring, and VL is fatal if left untreated. Using the tools in the PAML, we examined the sequence variations of Leishmania GP63 proteases to evaluate sites of positive selection as well as their functional implications. To focus on a set of limited number and relevant sequences, we selected those predicted to be GPI-anchored based on big-PI algorithm because previous evidence suggested that most GP63 are anchored to the plasma membranes through a GPI anchor (Bordier et al. 1986). We obtained 17 sequences from eight species (L. major, Leishmania amazonensis, L. infantum, Leishmania chagasi, Leishmania donovani, Leishmania guyanensis, Leishmania panamensis, and L. braziliensis), and a segment from Val100 to Asn577 according to Leima90 was used for our analysis (Schlagenhauf et al. 1998).

We used Consurf (Landau et al. 2005) to investigate how variable sites have distributed on the tertiary structure of the GP63 sequences. Our results showed that the zinc-binding surface is quite variable and its protuberant positions, also highly variable, seem to be advantageous for interacting with host proteins (Fig. 4b), which agree with the non-synonymous variations that occur externally and around the active site (Alvarez-Valin et al. 2000).

Fig. 4
figure 4

Positive selection analysis of the Leishmania spp. GP63 proteases. We used a segment (Val100 to Asn577 according to Leima90) of the 17 mature sequences. a Neighbor-joining (NJ) tree was built for a Leibr395 and b the branch of Leigu73, Leigu74, Leipa66, Leipa67, and Leibr112. b Positively selected sites were mapped to the tertiary structure of the GP63 protease: a and d, the back side of the zinc-binding surface; b, c, and e, the zinc-binding surface. The zinc atom is shown as a blue pellet. In a and b, different colors were used to represent the conservation condition for each site. In c, d, and e, the yellow-coded sites are positively selected (P > 95%). b and c shows the positive selected sites of branch–site model. e shows the positive selected sites based on the site model. All the pictures were processed by using PyMOL

Full size image

In the test for positive selections, we put forward two opposite hypotheses based on two different models. One assumes that all 17 sequences have the same ω ratio, based on the one-ratio model. The other assumes that some sequences may have different ω ratios, based on the free-ratio model. The LRT results are shown in Table 1. The lnL values for the one-ratio and free-ratio models are −7,806.594318 and −7,759.450251, respectively. Therefore, free-ratio model was considered better in this case. Given that df was 30, providing a value of 94.288134 for 2ΔL and the P value less than 0.01, our test suggested that some sequences may evolve at a different ratio. Since Leibr395 represents a more divergent lineage as compared to the remaining sequences (Fig. 4a), we supposed that Leibr395 has a different evolutionary pace in the branch–site model and our tests accepted the hypothesis. We thus defined 45 variation sites (105, 119, 123, 127, 153, 154, 166, 176, 181, 182, 206, 220, 235, 246, 248, 255, 256, 259, 267, 301, 302, 305, 307, 309, 314, 318, 332, 351, 392, 401, 468, 483, 484, 486, 496, 506, 532, 536, 537, 542, 547, 555, 570, 575, and 576, which are positioned according to the mature sequence of Leima90. Positions in bold have a P > 99%, while the remaining are 99% >P > 95%. These P values are different from the P value in LRT and represent the possibility for a site to be positively selected) that are supposed to contribute to the different ω ratios. After having localized all positions on the tertiary structure, we noticed that most of the positions are at the back of the zinc-binding surface (these sites are at the front if you look from the opposite angle; Fig. 4b). We also examined the three species (L. braziliensis, L. guyanensis, and L. panamensis) in lineage b, which have longer branches when compared to the rest and were reported to have a close phylogenic status and all induce mucocutaneous leishmaniasis (Osorio et al. 1998; Di Lella et al. 2006), but free-ratio test (16 sequences without Leibr395) showed insignificant positive selection.

Table 1 LRT for positive selection of Leishmania spp.
Full size table

We used the site model to analyze the remaining 16 sequences and found 12 positively selected sites, 136, 275, 278, 396, 424, 426, 429, 431, 432, 469, 474, and 502 (the positions in bold have a P > 99% and the rest are 99% >P > 95). The ω ratios of these sites are all greater than one. They were again mapped to the surrounding area of the zinc-binding site in the tertiary structure and mostly in protuberant positions (Fig. 4b). Although the two models, the branch–site and the site models, give rise to non-overlapping positively selected positions and have different locations on the tertiary structure, the nonrandom distributions suggested that these variations or sites may have functional implications in parasite–host interactions.

To determine whether the positively selected sites correlate with any specific leishmaniasis, we extracted the positively selected positions defined by the site model from each sequence to form a positive selection pattern for the sequences. We also divided the species into three groups according to different types of leishmaniasis: CL (L. major and Leishmania aethiopica), CL and ML (L. braziliensis, L. panamensis, and L. guyanensis), and CL and VL (L. chagasi, L. infantum, and L. donovani). The results are quite significant, where each group not only has their own distinct patterns but also more than one sub-patterns (Table 2).

Table 2 Positively selected sites of Leishmania spp.
Full size table

Discussion

In this study, we found that GP63 proteases evolved differently among trypanosomatids, especially in T. cruzi, whose GP63 proteases were duplicated extensively and brought in several novel GP63 domains as compared to those of T. brucei and most Leishmania spp. In addition, we found that the trypanosomatid GP63 protease domain bears more similarity to that of its hosts than to that of their non-pathogenic counterparts. Compared to the vectors (insects) and hosts (human and rodents), the parasitic GP63 proteases are more similar to those of their vectors. These results suggest that GP63 proteases may indeed play significant roles in host–parasite interactions.

We also identified different positive selection patterns for the three types of leishmaniasis. It was reported that patients cured from visceral leishmaniasis or cutaneous leishmaniasis respond to GP63 proteases differently: the former induces Th2-like response and the latter shows Th1-like response (Kurtzhals et al. 1994; Kemp et al. 1994). Furthermore, four Leishmania promastigote species, L. donovani, L. amazonensis, L. infantum, and L. major, exhibit different sensitivities to human complement (Dominguez et al. 2002). In each type of symptoms, we obtained unique patterns that can be subdivided into a limited number of sub-patterns. In fact, Leishmania species that induce the same disease symptoms also differ in immunoreactions (McMahon-Pratt and Alexander 2004; Wilson et al. 2005). The three species L. chagasi, L. infantum, and L. donovani all induce VL and CL, but only L. chagasi leads to common CL lesions. Dermal lesions at the inoculation site of L. infantum resemble lesions caused by dermatotropic species. L. donovani is unique in producing disseminated post-kala-azar dermal leishmaniasis, which occurs in a percentage of cured VL cases (Launois et al. 2008). In addition, even different strains of the same species can manifest different disease sequelae (Sharma et al. 2005). For instance, two L. major strains, MRHP/SU/59 Neals and WHOM/IR/-/173, not only differ in virulence but also respond differently to the same vaccine. Moreover, there are possibilities that disease diversity may be contributed by different host immune systems. For instance, in a mouse model, resistance and susceptibility to infection were found to be correlated with the development of CD4+ Th1 and CD4+ Th2 responses, respectively (Reiner and Locksley 1995). As a final note, GP63 protease expression levels may also result in different disease symptoms as Leishmania spp. display variable GP63 expression pattern even at the same stages (Yao et al. 2007).

Current therapies for trypanosomatid-associated diseases are costly, often poorly tolerated, and not always effective. Therefore, the development of alternative therapies, including vaccines, is of essence in the prevention of trypanosomatid infection. For Leishmania, the pathogenic mechanisms have been thoroughly elucidated and the biggest obstacle now for vaccine development is the variability of antigens among different Leishmania species (Spitzer et al. 1999). Our analysis provides a fresh look at the conserved and variable sequences, which may lead to new designs for the vaccines against the pathogenic trypanosomatids.